Medical centers are faced with multiple competing priorities when deciding how to focus their improvement efforts and meet the ever expanding menu of publicly reported and regulatory issues. In this article we expand on the rationale for supporting inpatient glycemic control programs as a priority that should be moved near the top of the list. We review the evidence for establishing glycemic range targets, and also review the limitations of this evidence, acknowledging, as does the American Diabetes Association (ADA), that in both the critical care and non‐critical care venue, glycemic goals must take into account the individual patient's situation as well as hospital system support for achieving these goals.1, 2 We emphasize that inpatient glycemic control programs are needed to address a wide variety of quality and safety issues surrounding the care of the inpatient with diabetes and hyperglycemia, and we wish to elevate the dialogue beyond arguments surrounding adoption of one glycemic target versus another. The Society of Hospital Medicine Glycemic Control Task Force members are not in unanimous agreement with the American Association of Clinical Endocrinologists (AACE)/ADA inpatient glycemic targets. However, we do agree on several other important points, which we will expand on in this article:

• 1

  Uncontrolled hyperglycemia and iatrogenic hypoglycemia are common and potentially dangerous situations that are largely preventable with safe and proven methods.

• 2

  The current state of care for our inpatients with hyperglycemia is unacceptably poor on a broad scale, with substandard education, communication, coordination, and treatment issues.

• 3

  Concerted efforts with changes in the design of the process of care are needed to improve this state of affairs.

DIABETES AND HYPERGLYCEMIA ARE VERY COMMON INPATIENT CONDITIONS

Diabetes mellitus (DM) has reached epidemic proportions in the United States. A reported 9.3% of adults over 20 years of age have diabetes, representing over 20 million persons. Despite increasing awareness, diabetes remains undiagnosed in approximately 30% of these persons.3 Concurrent with the increasing prevalence of diabetes in the U.S. population from 1980 through 2003, the number of hospital discharges with diabetes as any listed diagnosis more than doubled, going from 2.2 to 5.1 million discharges.4 Hospital care for patients with diabetes and hyperglycemia poses a significant health economic burden in the United States, representing over 40 billion dollars in annual direct medical expenditures.5

Hyperglycemia in the hospital may be due to known diabetes, to previously unrecognized diabetes, to prediabetes, and/or to the stress of surgery or illness. Deterioration in glycemic control in the hospital setting is most commonly associated with one or more factors, including stress‐induced release of insulin counterregulatory hormones (catecholamines, cortisol, glucagon, and growth hormone), exogenous administration of high dose glucocorticoids, and suboptimal glycemic management strategies.68 In a Belgian medical intensive care unit (MICU) randomized controlled trial (RCT) of strict versus conventional glycemic control, mean blood glucose (BG) on admission to the unit in the intention to treat group was 162 70 mg/dL (n = 1200),9 and in this group's RCT of 1548 surgical intensive care unit (SICU) patients, BG > 110 mg/dL was observed in over 70% of subjects.10 Mean BG of >145 mg/dL has been reported in 39%11 and BG >200 mg/dL in anywhere from 11% to 31% of intensive care unit (ICU) patients.10, 12 For general medicine and surgery, 1 study of 2030 patients admitted to a teaching hospital revealed that 26% of admissions had a known history of DM and 12% had new hyperglycemia, as evidenced by an admission or in‐hospital fasting BG of 126 mg/dL or more or a random BG of 200 mg/dL or more on 2 or more determinations.13 National and regional estimates on hospital use maintained by the Agency for Healthcare Research and Quality include data concerning diabetes diagnoses alone, without hyperglycemia, and may be displayed by querying its Web site.14 In cardiovascular populations almost 70% of patients having a first myocardial infarction have been reported to have either known DM, previously unrecognized diabetes, or impaired glucose tolerance.15

THE EVIDENCE SUPPORTS INPATIENT GLYCEMIC CONTROL

INPATIENT GLYCEMIC CONTROL IS INCREASINGLY INCORPORATED INTO PUBLIC REPORTING, GUIDELINES, REGULATORY AGENCY, AND NATIONAL QUALITY INITIATIVE PRIORITIES

National quality initiatives, public reporting, pay‐for‐performance, and guideline‐based care continue to play an increasingly important role in the U.S. healthcare system. Over the years these initiatives have focused on various disease states (venous thromboembolism, congestive heart failure, community‐acquired pneumonia, etc.) in an attempt to standardize care and improve patient safety and quality. Inpatient hyperglycemic control is also increasingly being incorporated into public reporting, regulatory compliance, and national quality initiatives.

Professional organizations such as the ADA2 and AACE50 have published guidelines supporting improved glycemic control, the safe use of insulin, and other measures to improve care for hyperglycemic inpatients. The AACE has a Web site dedicated to hospital hyperglycemia.52 The Society of Hospital Medicine48 has created a resource room on its Web site and a workbook for improvement49 on optimizing the care of inpatients with hyperglycemia and diabetes. The guidelines and Web sites help raise awareness and educate physicians and healthcare workers in inpatient glucose management. The American Heart Association has incorporated specific recommendation regarding inpatient diabetic management in its Get With the Guidelines.53

The Joint Commission54 has developed an advanced disease‐specific certification on inpatient diabetes. Disease management programs are important components of complex healthcare systems that serve to coordinate chronic care, promote early detection and prevention, and reduce overall healthcare costs. Certification is increasingly important to providers, payers, and healthcare institutions because it demonstrates a commitment to quality and patient safety. The Joint Commission disease‐specific care certification is a patient‐centered model focusing on the delivery of clinical care and relationship between the practitioner and the patient. The evaluation and resulting certification by the Joint Commission is based on 3 core components: (1) an assessment of compliance with consensus‐based national standards; (2) the effective use of established clinical practice guidelines to manage and optimize care; and (3) an organized approach to performance measurement and improved activities.55 For inpatient diabetes, the Joint Commission program has 7 major elements following the ADA recommendations, including general recommendations regarding diabetic documentation, BG targets, preventing hypoglycemia, diabetes care providers, diabetes self‐management education, medical nutrition therapy, and BG monitoring.54 This mirrors the Call to Action Consensus Conference essential elements for successful glycemic control programs.1

Other organizations such as the Surgical Care Improvement Partnership (SCIP) and National Surgical Quality Improvement Program (NSQIP) have included perioperative glycemic control measures, as it impacts surgical wound infections. The University HealthSystem Consortium (UHC) has benchmarking data and endorses perioperative glycemic control measures, whereas the Institute for Healthcare Improvement (IHI) has focused on safe use of insulin practices in its 5 Million Lives campaign.

HOSPITALIZATION IS A MOMENT OF OPPORTUNITY TO ASSESS AND INTERVENE

The benefits of outpatient glycemic control and quality preventive care are well established, and the reduction of adverse consequences of uncontrolled diabetes are a high priority in ambulatory medicine.5658 Hospitalization provides an opportunity to identify previously undiagnosed diabetes or prediabetes and, for patients with known diabetes, to assess and impact upon the long term course of diabetes.

As a first step, unless a recent hemoglobin A1C (HbA1c) is known, among hospitalized hyperglycemic patients an HbA1C should be obtained upon admission. Greci et al.59 showed that an HbA1c level >6.0% was 100% specific (14/14) and 57% sensitive (12/21) for the diagnosis of diabetes. Among patients having known diabetes, an HbA1C elevation on admission may justify intensification of preadmission management at the time of discharge. If discharge and postdischarge adjustments of preadmission regimens are planned in response to admission A1C elevations, then the modified long‐term treatment strategy can improve the A1C in the ambulatory setting.60 Moreover, the event of hospitalization is the ideal teachable moment for patients and their caregivers to improve self‐care activities. Yet floor nurses may be overwhelmed by the tasks of patient education. For ideal patient education, both a nutritionist and a diabetes nurse educator are needed to assess compliance with medication, diet, and other aspects of care.6163 There also is need for outpatient follow‐up education. Finally, at the time of discharge, there is a duty and an opportunity for the diabetes provider to communicate with outpatient care providers about the patient's regimen and glycemic control, and also, based on information gathered during the admission, to convey any evidence that might support the need for a change of long‐term strategy.64 Unfortunately, the opportunity that hospitalization presents to assess, educate, and intervene frequently is underused.1, 8, 51, 65

LARGE GAPS EXIST BETWEEN CURRENT AND OPTIMAL CARE

Despite the evidence that inpatient glycemic control is important for patient outcomes, and despite guidelines recommending tighter inpatient glycemic control, clinical practice has been slow to change. In many institutions, inpatient glycemic management has not improved over the past decade, and large gaps remain between current practice and optimal practice.

Studies of individual institutions provide several insights into gaps in care. For example, Schnipper et al.66 examined practices on the general medicine service of an academic medical center in Boston in 2004. Among 107 prospectively identified patients with a known diagnosis of diabetes or at least 1 glucose reading >200 mg/dL (excluding patients with diabetic ketoacidosis, hyperglycemic hyperosmolar state, or pregnancy), they found scheduled long‐acting insulin prescribed in 43% of patients, scheduled short‐acting/rapid‐acting insulin in only 4% of patients, and 80 of 89 patients (90%) on the same sliding scale insulin regimen despite widely varying insulin requirements. Thirty‐one percent of glucose readings were >180 mg/dL compared with 1.2% of readings 60 mg/dL (but 11% of patients had at least 1 episode of hypoglycemia). Of the 75 patients with at least 1 episode of hyperglycemia or hypoglycemia, only 35% had any change to their insulin regimen during the first 5 days of the hospitalization.

Other studies have confirmed this concept of clinical inertia (ie, recognition of the problem but failure to act).67 A study by Cook et al.68 of all hospitalized non‐ICU patients with diabetes or hyperglycemia and length of stay of 3 days between 2001 and 2004 showed that 20% of patients had persistent hyperglycemia during the hospitalization (defined as a mean glucose >200 mg/dL). Forty‐six percent of patients whose average glucose was in the top tertile did not have their insulin regimen intensified to a combination of short‐acting/rapid‐acting and long‐acting insulin, and 35% of these patients either had no change in their total daily insulin dose or actually had a decrease in their dose when comparing the last 24 hours with the first 24 hours of hospitalization, a concept they term negative therapeutic momentum.

Perhaps the most well‐balanced view of the current state of medical practice comes from the UHC benchmarking project.69 UHC is an alliance of 90 academic health centers. For the diabetes project, each institution reviewed the records of 50 randomly selected patients over 18 years of age with at least a 72‐hour length of stay, 1 of 7 prespecified Diagnosis Related Group (DRG) codes, and at least 2 consecutive glucose readings >180 mg/dL or the receipt of insulin any time during the hospitalization. Patients with a history of pancreatic transplant, pregnant at the time of admission, receiving hospice or comfort care, or receiving insulin for a reason other than glucose management were excluded. The study showed widespread gaps in processes and outcomes (Table 1). Moreover, performance varied widely across hospitals. For example, the morning glucose in the ICU on the second measurement day was 110 mg/dL in 18% of patients for the median‐performing hospital, with a range of 0% to 67% across all 37 measured hospitals. In the non‐ICU setting on the second measurement day, 26% of patients had all BG measurements = 180 mg/dL in the median‐performing hospital, with a range of 7% to 48%. Of note, hypoglycemia was relatively uncommon: in the median hospital, 2.4% of patient‐days had 1 or more BG readings 50 mg/dL (range: 0%8.6%). Finally, in the median‐performing hospital, effective insulin therapy (defined as short‐acting/rapid‐acting and long‐acting subcutaneous insulin, continuous insulin infusion, or subcutaneous insulin pump therapy) was prescribed in 45% of patients, with a range of 12% to 77% across measured hospitals.

FREQUENT PROBLEMS WITH COMMUNICATION AND COORDINATION

Those who work closely with frontline practitioners striving to improve inpatient glycemic management have noticed other deficiencies in care.1, 70 These include: a lack of coordination between feeding, BG measurement, and insulin administration, leading to mistimed and incorrectly dosed insulin; frequent use of sliding‐scale only regimens despite evidence that they are useless at best and harmful at worst;6, 47, 60, 71 discharge summaries that often do not mention follow‐up plans for hyperglycemic management; incomplete patient educational programs; breakdowns in care at transition points; nursing and medical staffs that are unevenly educated about the proper use of insulin; and patients who are often angry or confused about their diabetes care in the hospital. Collectively, these gaps in care serve as prime targets for any glycemic control program.

HYPOGLYCEMIA IS A PROMINENT INPATIENT SAFETY CONCERN

Hypoglycemia is common in the inpatient setting and is a legitimate safety concern. In a recently reported series of 2174 hospitalized patients receiving antihyperglycemic agents, it was found that 9.5% of patients experienced a total 484 hypoglycemic episodes (defined as 60 mg/dL).72 Hypoglycemia often occurred in the setting of insulin therapy and frequently resulted from a failure to recognize trends in BG readings or other clues that a patient was at risk for developing hypoglycemia.73 A common thread is the risk created by interruption of carbohydrate intake, noted by Fischer et al.73 and once again in the recent ICU study by Vriesendorp et al.74 Sources of error include: lack of coordination between feeding and medication administration, leading to mistiming of insulin action; lack of sufficient frequency in BG monitoring; lack of clarity or uniformity in the writing of orders; failure to recognize changes in insulin requirements due to advanced age, renal failure, liver disease, or change in clinical status; steroid use with subsequent tapering or interruption; changes in feeding; failure to reconcile medications; inappropriate use of oral antihyperglycemic agents, and communication or handoff failures.

It has been difficult to sort out whether hypoglycemia is a marker of severity of illness or whether it is an independent factor leading to poor outcomes. Observational studies lend credibility to the concept that patients having congestive heart failure or myocardial infarction may be at risk for excessive mortality if their average BG resides in the low end of the normal range.7578 Sympathetic activation occurs as the threshold for hypoglycemia is approached, such as occurs at BG = 70 or 72 mg/dL.79 Patients living with BG levels observed to be in the low end of the normal range might experience more severe but unobserved and undocumented episodes of neuroglycopenia. Arrhythmia and fatality have been directly attributed to strict glycemic control.80, 81 We are confronted with the need to interpret well conducted observational studies, evaluating subgroups at risk, and using multivariate analysis to assess the impact of hypoglycemia upon outcomes.82 In such studies, we will need to examine high‐risk subgroups, including cardiac patients, in particular, for the possibility that there is a J‐shaped curve for mortality as a function of average BG.

Unfortunately, clinical inertia exists in response to hypoglycemia just as it does with hyperglycemia. One recent study examined 52 patients who received intravenous 50% dextrose solution for an episode of hypoglycemia.83 Changes to insulin regimens were subsequently made in only 21 patients (40%), and diabetes specialists agreed with the changes for 11 of these patients. The other 31 patients (60%) received no changes in treatment, and diabetes specialists agreed with that decision for only 10 of these patients.

Although some increase in hypoglycemia might be expected with initiation of tight glycemic control efforts, the solution is not to undertreat hyperglycemia. Hyperglycemia creates an unsafe setting for the treatment of illness and disease. Sliding‐scaleonly regimens are ineffective in securing glycemic control and can result in increases in hypoglycemia as well as hyperglycemic excursions.6, 66 Inappropriate withholding of insulin doses can lead to severe glycemic excursions and even iatrogenic diabetic ketoacidosis (DKA). Systems approaches to avoid the errors outlined above can minimize or even reverse the increased risk of hypoglycemia expected with tighter glycemic targets.51

A SYSTEMS APPROACH IS NEEDED FOR THESE MULTIPLE COMPLEX PROBLEMS

Care is of the hyperglycemic inpatient is inherently complex. Previously established treatments are often inappropriate under conditions of altered insulin resistance, changing patterns of nutrition and carbohydrate exposure, comorbidities, concomitant medications, and rapidly changing medical and surgical status. Patients frequently undergo changes in the route and amount of nutritional exposure, including discrete meals, continuous intravenous dextrose, nil per orem (nothing by mouth status; NPO) status, grazing on nutritional supplements or liquid diets with or without meals, bolus enteral feedings, overnight enteral feedings with daytime grazing, total parenteral nutrition, continuous peritoneal dialysis, and overnight cycling of peritoneal dialysis. Relying on individual expertise and vigilance to negotiate this complex terrain without safeguards, protocols, standardization of orders, and other systems change is impractical and unwise.

Transitions across care providers and locations lead to multiple opportunities for breakdown in the quality, consistency, and safety of care.64, 65 At the time of ward transfer or change of patient status, previous medication and monitoring orders sometimes are purged. At the time of discharge, there may be risk of continuation of anti‐hyperglycemic therapy, initiated to cover medical stress, in doses that will subsequently be unsafe.

In the face of this complexity, educational programs alone will not suffice to improve care. Institutional commitment and systems changes are essential.

MARKED IMPROVEMENT IS POSSIBLE AND TOOLS EXIST: A ROADMAP IS IN PLACE

Fortunately, a roadmap is in place to help us achieve better glycemic control, improve insulin management, and address the long list of current deficiencies in care. This is imperative to develop consistent processes in order to achieve maximum patient quality outcomes that effective glycemic management offers. This roadmap entails 4 components: (1) national awareness, (2) national guidelines, (3) consensus statements, and (4) effective tools. As mentioned above, the first two components of this roadmap are now in place.

As these national guidelines become more widely accepted, the next step will be the incorporation of this into programs like Pay‐for Performance and the Physician Quality Reporting Initiative (PQRI), which will impact reimbursement to both hospitals and providers.

Regarding the third component, a recent multidisciplinary consensus conference1 outlined the essential elements needed for successful implementation of an inpatient glycemic control program which include:

• An appropriate level of administrative support.

• Formation of a multidisciplinary steering committee to drive the development of initiatives and empowered to enact change.

• Assessment of current processes, quality of care, and barriers to practice change.

• Development and implementation of interventions including standardized order sets, protocols, policies and algorithms with associated educational programs.

• Metrics for evaluation of glycemic control, hypoglycemia, insulin use patterns, and other aspects of care.

Finally, extensive resources and effective tools are now available to help institutions achieve better inpatient glucose control. The Society of Hospital Medicine (SHM), in conjunction with the ADA, AACE, the American College of Physicians (ACP), the Case Management Society of America (CMSA), the American Society of Consultant Pharmacists, nursing, and diabetic educators have all partnered to produce a comprehensive guide to effective implementation of glycemic control and preventing hypoglycemia.49 This comprehensive workbook is a proven performance improvement framework and is available on the SHM Web site.48 Details and examples of all essential elements are covered in this workbook along with opportunities for marked improvement bolstered by integration of high reliability design features and attention to effective implementation techniques. The remainder of this supplement crystallizes a substantial portion of this material. The AACE has also recently offered a valuable web‐based resource to encourage institutional glycemic control efforts.49

GLYCEMIC CONTROL INITIATIVES CAN BE COST‐EFFECTIVE

Achieving optimal glycemic control safely requires monitoring, education, and other measures, which can be expensive, labor intensive, and require coordination of the services of many hospital divisions. This incremental expense has been shown to be cost‐effective in a variety of settings.1, 84, 85 The costs of glycemic control initiatives have demonstrated a good return on investment via:

• Improved LOS, readmission rates, morbidity, and mortality.

• Improved documentation of patient acuity and related payment for acuity.

• Income generated via incremental physician and allied health professional billing.

CONCLUSION AND SUMMARY

Evidence exists that appropriate management of hyperglycemia improves outcomes, whereas the current state of affairs is that most medical centers currently manage this suboptimally. This is concerning given the magnitude of diabetes and hyperglycemia in our inpatient setting in the United States. To bring awareness to this issue, multiple initiatives (guidelines, certification programs, workbooks, etc.) are available by various organizations including the ADA, AACE, SCIP, NSQIP, IHI, UHC, the Joint Commission, and SHM. However, this is not enough. Change occurs at the local level, and institutional prioritization and support is needed to empower a multidisciplinary steering committee, with appropriate administrative support, to standardize and improve systems in the face of substantial cultural issues and complex barriers. Improved data collection and reporting, incremental monitoring, creation of metrics, and improved documentation are an absolutely necessary necessity to achieve breakthrough levels of improvement.

Now the time is right to make an assertive effort to improve inpatient glycemic control and related issues, and push for appropriate support at your institution to help achieve this in the interest of patient safety and optimal outcomes.

Medical centers are faced with multiple competing priorities when deciding how to focus their improvement efforts and meet the ever expanding menu of publicly reported and regulatory issues. In this article we expand on the rationale for supporting inpatient glycemic control programs as a priority that should be moved near the top of the list. We review the evidence for establishing glycemic range targets, and also review the limitations of this evidence, acknowledging, as does the American Diabetes Association (ADA), that in both the critical care and non‐critical care venue, glycemic goals must take into account the individual patient's situation as well as hospital system support for achieving these goals.1, 2 We emphasize that inpatient glycemic control programs are needed to address a wide variety of quality and safety issues surrounding the care of the inpatient with diabetes and hyperglycemia, and we wish to elevate the dialogue beyond arguments surrounding adoption of one glycemic target versus another. The Society of Hospital Medicine Glycemic Control Task Force members are not in unanimous agreement with the American Association of Clinical Endocrinologists (AACE)/ADA inpatient glycemic targets. However, we do agree on several other important points, which we will expand on in this article:

• 1

  Uncontrolled hyperglycemia and iatrogenic hypoglycemia are common and potentially dangerous situations that are largely preventable with safe and proven methods.

• 2

  The current state of care for our inpatients with hyperglycemia is unacceptably poor on a broad scale, with substandard education, communication, coordination, and treatment issues.

• 3

  Concerted efforts with changes in the design of the process of care are needed to improve this state of affairs.

DIABETES AND HYPERGLYCEMIA ARE VERY COMMON INPATIENT CONDITIONS

Diabetes mellitus (DM) has reached epidemic proportions in the United States. A reported 9.3% of adults over 20 years of age have diabetes, representing over 20 million persons. Despite increasing awareness, diabetes remains undiagnosed in approximately 30% of these persons.3 Concurrent with the increasing prevalence of diabetes in the U.S. population from 1980 through 2003, the number of hospital discharges with diabetes as any listed diagnosis more than doubled, going from 2.2 to 5.1 million discharges.4 Hospital care for patients with diabetes and hyperglycemia poses a significant health economic burden in the United States, representing over 40 billion dollars in annual direct medical expenditures.5

Hyperglycemia in the hospital may be due to known diabetes, to previously unrecognized diabetes, to prediabetes, and/or to the stress of surgery or illness. Deterioration in glycemic control in the hospital setting is most commonly associated with one or more factors, including stress‐induced release of insulin counterregulatory hormones (catecholamines, cortisol, glucagon, and growth hormone), exogenous administration of high dose glucocorticoids, and suboptimal glycemic management strategies.68 In a Belgian medical intensive care unit (MICU) randomized controlled trial (RCT) of strict versus conventional glycemic control, mean blood glucose (BG) on admission to the unit in the intention to treat group was 162 70 mg/dL (n = 1200),9 and in this group's RCT of 1548 surgical intensive care unit (SICU) patients, BG > 110 mg/dL was observed in over 70% of subjects.10 Mean BG of >145 mg/dL has been reported in 39%11 and BG >200 mg/dL in anywhere from 11% to 31% of intensive care unit (ICU) patients.10, 12 For general medicine and surgery, 1 study of 2030 patients admitted to a teaching hospital revealed that 26% of admissions had a known history of DM and 12% had new hyperglycemia, as evidenced by an admission or in‐hospital fasting BG of 126 mg/dL or more or a random BG of 200 mg/dL or more on 2 or more determinations.13 National and regional estimates on hospital use maintained by the Agency for Healthcare Research and Quality include data concerning diabetes diagnoses alone, without hyperglycemia, and may be displayed by querying its Web site.14 In cardiovascular populations almost 70% of patients having a first myocardial infarction have been reported to have either known DM, previously unrecognized diabetes, or impaired glucose tolerance.15

THE EVIDENCE SUPPORTS INPATIENT GLYCEMIC CONTROL

INPATIENT GLYCEMIC CONTROL IS INCREASINGLY INCORPORATED INTO PUBLIC REPORTING, GUIDELINES, REGULATORY AGENCY, AND NATIONAL QUALITY INITIATIVE PRIORITIES

National quality initiatives, public reporting, pay‐for‐performance, and guideline‐based care continue to play an increasingly important role in the U.S. healthcare system. Over the years these initiatives have focused on various disease states (venous thromboembolism, congestive heart failure, community‐acquired pneumonia, etc.) in an attempt to standardize care and improve patient safety and quality. Inpatient hyperglycemic control is also increasingly being incorporated into public reporting, regulatory compliance, and national quality initiatives.

Professional organizations such as the ADA2 and AACE50 have published guidelines supporting improved glycemic control, the safe use of insulin, and other measures to improve care for hyperglycemic inpatients. The AACE has a Web site dedicated to hospital hyperglycemia.52 The Society of Hospital Medicine48 has created a resource room on its Web site and a workbook for improvement49 on optimizing the care of inpatients with hyperglycemia and diabetes. The guidelines and Web sites help raise awareness and educate physicians and healthcare workers in inpatient glucose management. The American Heart Association has incorporated specific recommendation regarding inpatient diabetic management in its Get With the Guidelines.53

The Joint Commission54 has developed an advanced disease‐specific certification on inpatient diabetes. Disease management programs are important components of complex healthcare systems that serve to coordinate chronic care, promote early detection and prevention, and reduce overall healthcare costs. Certification is increasingly important to providers, payers, and healthcare institutions because it demonstrates a commitment to quality and patient safety. The Joint Commission disease‐specific care certification is a patient‐centered model focusing on the delivery of clinical care and relationship between the practitioner and the patient. The evaluation and resulting certification by the Joint Commission is based on 3 core components: (1) an assessment of compliance with consensus‐based national standards; (2) the effective use of established clinical practice guidelines to manage and optimize care; and (3) an organized approach to performance measurement and improved activities.55 For inpatient diabetes, the Joint Commission program has 7 major elements following the ADA recommendations, including general recommendations regarding diabetic documentation, BG targets, preventing hypoglycemia, diabetes care providers, diabetes self‐management education, medical nutrition therapy, and BG monitoring.54 This mirrors the Call to Action Consensus Conference essential elements for successful glycemic control programs.1

Other organizations such as the Surgical Care Improvement Partnership (SCIP) and National Surgical Quality Improvement Program (NSQIP) have included perioperative glycemic control measures, as it impacts surgical wound infections. The University HealthSystem Consortium (UHC) has benchmarking data and endorses perioperative glycemic control measures, whereas the Institute for Healthcare Improvement (IHI) has focused on safe use of insulin practices in its 5 Million Lives campaign.

HOSPITALIZATION IS A MOMENT OF OPPORTUNITY TO ASSESS AND INTERVENE

The benefits of outpatient glycemic control and quality preventive care are well established, and the reduction of adverse consequences of uncontrolled diabetes are a high priority in ambulatory medicine.5658 Hospitalization provides an opportunity to identify previously undiagnosed diabetes or prediabetes and, for patients with known diabetes, to assess and impact upon the long term course of diabetes.

As a first step, unless a recent hemoglobin A1C (HbA1c) is known, among hospitalized hyperglycemic patients an HbA1C should be obtained upon admission. Greci et al.59 showed that an HbA1c level >6.0% was 100% specific (14/14) and 57% sensitive (12/21) for the diagnosis of diabetes. Among patients having known diabetes, an HbA1C elevation on admission may justify intensification of preadmission management at the time of discharge. If discharge and postdischarge adjustments of preadmission regimens are planned in response to admission A1C elevations, then the modified long‐term treatment strategy can improve the A1C in the ambulatory setting.60 Moreover, the event of hospitalization is the ideal teachable moment for patients and their caregivers to improve self‐care activities. Yet floor nurses may be overwhelmed by the tasks of patient education. For ideal patient education, both a nutritionist and a diabetes nurse educator are needed to assess compliance with medication, diet, and other aspects of care.6163 There also is need for outpatient follow‐up education. Finally, at the time of discharge, there is a duty and an opportunity for the diabetes provider to communicate with outpatient care providers about the patient's regimen and glycemic control, and also, based on information gathered during the admission, to convey any evidence that might support the need for a change of long‐term strategy.64 Unfortunately, the opportunity that hospitalization presents to assess, educate, and intervene frequently is underused.1, 8, 51, 65

LARGE GAPS EXIST BETWEEN CURRENT AND OPTIMAL CARE

Despite the evidence that inpatient glycemic control is important for patient outcomes, and despite guidelines recommending tighter inpatient glycemic control, clinical practice has been slow to change. In many institutions, inpatient glycemic management has not improved over the past decade, and large gaps remain between current practice and optimal practice.

Studies of individual institutions provide several insights into gaps in care. For example, Schnipper et al.66 examined practices on the general medicine service of an academic medical center in Boston in 2004. Among 107 prospectively identified patients with a known diagnosis of diabetes or at least 1 glucose reading >200 mg/dL (excluding patients with diabetic ketoacidosis, hyperglycemic hyperosmolar state, or pregnancy), they found scheduled long‐acting insulin prescribed in 43% of patients, scheduled short‐acting/rapid‐acting insulin in only 4% of patients, and 80 of 89 patients (90%) on the same sliding scale insulin regimen despite widely varying insulin requirements. Thirty‐one percent of glucose readings were >180 mg/dL compared with 1.2% of readings 60 mg/dL (but 11% of patients had at least 1 episode of hypoglycemia). Of the 75 patients with at least 1 episode of hyperglycemia or hypoglycemia, only 35% had any change to their insulin regimen during the first 5 days of the hospitalization.

Other studies have confirmed this concept of clinical inertia (ie, recognition of the problem but failure to act).67 A study by Cook et al.68 of all hospitalized non‐ICU patients with diabetes or hyperglycemia and length of stay of 3 days between 2001 and 2004 showed that 20% of patients had persistent hyperglycemia during the hospitalization (defined as a mean glucose >200 mg/dL). Forty‐six percent of patients whose average glucose was in the top tertile did not have their insulin regimen intensified to a combination of short‐acting/rapid‐acting and long‐acting insulin, and 35% of these patients either had no change in their total daily insulin dose or actually had a decrease in their dose when comparing the last 24 hours with the first 24 hours of hospitalization, a concept they term negative therapeutic momentum.

Perhaps the most well‐balanced view of the current state of medical practice comes from the UHC benchmarking project.69 UHC is an alliance of 90 academic health centers. For the diabetes project, each institution reviewed the records of 50 randomly selected patients over 18 years of age with at least a 72‐hour length of stay, 1 of 7 prespecified Diagnosis Related Group (DRG) codes, and at least 2 consecutive glucose readings >180 mg/dL or the receipt of insulin any time during the hospitalization. Patients with a history of pancreatic transplant, pregnant at the time of admission, receiving hospice or comfort care, or receiving insulin for a reason other than glucose management were excluded. The study showed widespread gaps in processes and outcomes (Table 1). Moreover, performance varied widely across hospitals. For example, the morning glucose in the ICU on the second measurement day was 110 mg/dL in 18% of patients for the median‐performing hospital, with a range of 0% to 67% across all 37 measured hospitals. In the non‐ICU setting on the second measurement day, 26% of patients had all BG measurements = 180 mg/dL in the median‐performing hospital, with a range of 7% to 48%. Of note, hypoglycemia was relatively uncommon: in the median hospital, 2.4% of patient‐days had 1 or more BG readings 50 mg/dL (range: 0%8.6%). Finally, in the median‐performing hospital, effective insulin therapy (defined as short‐acting/rapid‐acting and long‐acting subcutaneous insulin, continuous insulin infusion, or subcutaneous insulin pump therapy) was prescribed in 45% of patients, with a range of 12% to 77% across measured hospitals.

FREQUENT PROBLEMS WITH COMMUNICATION AND COORDINATION

Those who work closely with frontline practitioners striving to improve inpatient glycemic management have noticed other deficiencies in care.1, 70 These include: a lack of coordination between feeding, BG measurement, and insulin administration, leading to mistimed and incorrectly dosed insulin; frequent use of sliding‐scale only regimens despite evidence that they are useless at best and harmful at worst;6, 47, 60, 71 discharge summaries that often do not mention follow‐up plans for hyperglycemic management; incomplete patient educational programs; breakdowns in care at transition points; nursing and medical staffs that are unevenly educated about the proper use of insulin; and patients who are often angry or confused about their diabetes care in the hospital. Collectively, these gaps in care serve as prime targets for any glycemic control program.

HYPOGLYCEMIA IS A PROMINENT INPATIENT SAFETY CONCERN

Hypoglycemia is common in the inpatient setting and is a legitimate safety concern. In a recently reported series of 2174 hospitalized patients receiving antihyperglycemic agents, it was found that 9.5% of patients experienced a total 484 hypoglycemic episodes (defined as 60 mg/dL).72 Hypoglycemia often occurred in the setting of insulin therapy and frequently resulted from a failure to recognize trends in BG readings or other clues that a patient was at risk for developing hypoglycemia.73 A common thread is the risk created by interruption of carbohydrate intake, noted by Fischer et al.73 and once again in the recent ICU study by Vriesendorp et al.74 Sources of error include: lack of coordination between feeding and medication administration, leading to mistiming of insulin action; lack of sufficient frequency in BG monitoring; lack of clarity or uniformity in the writing of orders; failure to recognize changes in insulin requirements due to advanced age, renal failure, liver disease, or change in clinical status; steroid use with subsequent tapering or interruption; changes in feeding; failure to reconcile medications; inappropriate use of oral antihyperglycemic agents, and communication or handoff failures.

It has been difficult to sort out whether hypoglycemia is a marker of severity of illness or whether it is an independent factor leading to poor outcomes. Observational studies lend credibility to the concept that patients having congestive heart failure or myocardial infarction may be at risk for excessive mortality if their average BG resides in the low end of the normal range.7578 Sympathetic activation occurs as the threshold for hypoglycemia is approached, such as occurs at BG = 70 or 72 mg/dL.79 Patients living with BG levels observed to be in the low end of the normal range might experience more severe but unobserved and undocumented episodes of neuroglycopenia. Arrhythmia and fatality have been directly attributed to strict glycemic control.80, 81 We are confronted with the need to interpret well conducted observational studies, evaluating subgroups at risk, and using multivariate analysis to assess the impact of hypoglycemia upon outcomes.82 In such studies, we will need to examine high‐risk subgroups, including cardiac patients, in particular, for the possibility that there is a J‐shaped curve for mortality as a function of average BG.

Unfortunately, clinical inertia exists in response to hypoglycemia just as it does with hyperglycemia. One recent study examined 52 patients who received intravenous 50% dextrose solution for an episode of hypoglycemia.83 Changes to insulin regimens were subsequently made in only 21 patients (40%), and diabetes specialists agreed with the changes for 11 of these patients. The other 31 patients (60%) received no changes in treatment, and diabetes specialists agreed with that decision for only 10 of these patients.

Although some increase in hypoglycemia might be expected with initiation of tight glycemic control efforts, the solution is not to undertreat hyperglycemia. Hyperglycemia creates an unsafe setting for the treatment of illness and disease. Sliding‐scaleonly regimens are ineffective in securing glycemic control and can result in increases in hypoglycemia as well as hyperglycemic excursions.6, 66 Inappropriate withholding of insulin doses can lead to severe glycemic excursions and even iatrogenic diabetic ketoacidosis (DKA). Systems approaches to avoid the errors outlined above can minimize or even reverse the increased risk of hypoglycemia expected with tighter glycemic targets.51

A SYSTEMS APPROACH IS NEEDED FOR THESE MULTIPLE COMPLEX PROBLEMS

Care is of the hyperglycemic inpatient is inherently complex. Previously established treatments are often inappropriate under conditions of altered insulin resistance, changing patterns of nutrition and carbohydrate exposure, comorbidities, concomitant medications, and rapidly changing medical and surgical status. Patients frequently undergo changes in the route and amount of nutritional exposure, including discrete meals, continuous intravenous dextrose, nil per orem (nothing by mouth status; NPO) status, grazing on nutritional supplements or liquid diets with or without meals, bolus enteral feedings, overnight enteral feedings with daytime grazing, total parenteral nutrition, continuous peritoneal dialysis, and overnight cycling of peritoneal dialysis. Relying on individual expertise and vigilance to negotiate this complex terrain without safeguards, protocols, standardization of orders, and other systems change is impractical and unwise.

Transitions across care providers and locations lead to multiple opportunities for breakdown in the quality, consistency, and safety of care.64, 65 At the time of ward transfer or change of patient status, previous medication and monitoring orders sometimes are purged. At the time of discharge, there may be risk of continuation of anti‐hyperglycemic therapy, initiated to cover medical stress, in doses that will subsequently be unsafe.

In the face of this complexity, educational programs alone will not suffice to improve care. Institutional commitment and systems changes are essential.

MARKED IMPROVEMENT IS POSSIBLE AND TOOLS EXIST: A ROADMAP IS IN PLACE

Fortunately, a roadmap is in place to help us achieve better glycemic control, improve insulin management, and address the long list of current deficiencies in care. This is imperative to develop consistent processes in order to achieve maximum patient quality outcomes that effective glycemic management offers. This roadmap entails 4 components: (1) national awareness, (2) national guidelines, (3) consensus statements, and (4) effective tools. As mentioned above, the first two components of this roadmap are now in place.

As these national guidelines become more widely accepted, the next step will be the incorporation of this into programs like Pay‐for Performance and the Physician Quality Reporting Initiative (PQRI), which will impact reimbursement to both hospitals and providers.

Regarding the third component, a recent multidisciplinary consensus conference1 outlined the essential elements needed for successful implementation of an inpatient glycemic control program which include:

• An appropriate level of administrative support.

• Formation of a multidisciplinary steering committee to drive the development of initiatives and empowered to enact change.

• Assessment of current processes, quality of care, and barriers to practice change.

• Development and implementation of interventions including standardized order sets, protocols, policies and algorithms with associated educational programs.

• Metrics for evaluation of glycemic control, hypoglycemia, insulin use patterns, and other aspects of care.

Finally, extensive resources and effective tools are now available to help institutions achieve better inpatient glucose control. The Society of Hospital Medicine (SHM), in conjunction with the ADA, AACE, the American College of Physicians (ACP), the Case Management Society of America (CMSA), the American Society of Consultant Pharmacists, nursing, and diabetic educators have all partnered to produce a comprehensive guide to effective implementation of glycemic control and preventing hypoglycemia.49 This comprehensive workbook is a proven performance improvement framework and is available on the SHM Web site.48 Details and examples of all essential elements are covered in this workbook along with opportunities for marked improvement bolstered by integration of high reliability design features and attention to effective implementation techniques. The remainder of this supplement crystallizes a substantial portion of this material. The AACE has also recently offered a valuable web‐based resource to encourage institutional glycemic control efforts.49

GLYCEMIC CONTROL INITIATIVES CAN BE COST‐EFFECTIVE

Achieving optimal glycemic control safely requires monitoring, education, and other measures, which can be expensive, labor intensive, and require coordination of the services of many hospital divisions. This incremental expense has been shown to be cost‐effective in a variety of settings.1, 84, 85 The costs of glycemic control initiatives have demonstrated a good return on investment via:

• Improved LOS, readmission rates, morbidity, and mortality.

• Improved documentation of patient acuity and related payment for acuity.

• Income generated via incremental physician and allied health professional billing.

CONCLUSION AND SUMMARY

Evidence exists that appropriate management of hyperglycemia improves outcomes, whereas the current state of affairs is that most medical centers currently manage this suboptimally. This is concerning given the magnitude of diabetes and hyperglycemia in our inpatient setting in the United States. To bring awareness to this issue, multiple initiatives (guidelines, certification programs, workbooks, etc.) are available by various organizations including the ADA, AACE, SCIP, NSQIP, IHI, UHC, the Joint Commission, and SHM. However, this is not enough. Change occurs at the local level, and institutional prioritization and support is needed to empower a multidisciplinary steering committee, with appropriate administrative support, to standardize and improve systems in the face of substantial cultural issues and complex barriers. Improved data collection and reporting, incremental monitoring, creation of metrics, and improved documentation are an absolutely necessary necessity to achieve breakthrough levels of improvement.

Now the time is right to make an assertive effort to improve inpatient glycemic control and related issues, and push for appropriate support at your institution to help achieve this in the interest of patient safety and optimal outcomes.

Medical centers are faced with multiple competing priorities when deciding how to focus their improvement efforts and meet the ever expanding menu of publicly reported and regulatory issues. In this article we expand on the rationale for supporting inpatient glycemic control programs as a priority that should be moved near the top of the list. We review the evidence for establishing glycemic range targets, and also review the limitations of this evidence, acknowledging, as does the American Diabetes Association (ADA), that in both the critical care and non‐critical care venue, glycemic goals must take into account the individual patient's situation as well as hospital system support for achieving these goals.1, 2 We emphasize that inpatient glycemic control programs are needed to address a wide variety of quality and safety issues surrounding the care of the inpatient with diabetes and hyperglycemia, and we wish to elevate the dialogue beyond arguments surrounding adoption of one glycemic target versus another. The Society of Hospital Medicine Glycemic Control Task Force members are not in unanimous agreement with the American Association of Clinical Endocrinologists (AACE)/ADA inpatient glycemic targets. However, we do agree on several other important points, which we will expand on in this article:

• 1

  Uncontrolled hyperglycemia and iatrogenic hypoglycemia are common and potentially dangerous situations that are largely preventable with safe and proven methods.

• 2

  The current state of care for our inpatients with hyperglycemia is unacceptably poor on a broad scale, with substandard education, communication, coordination, and treatment issues.

• 3

  Concerted efforts with changes in the design of the process of care are needed to improve this state of affairs.

DIABETES AND HYPERGLYCEMIA ARE VERY COMMON INPATIENT CONDITIONS

Diabetes mellitus (DM) has reached epidemic proportions in the United States. A reported 9.3% of adults over 20 years of age have diabetes, representing over 20 million persons. Despite increasing awareness, diabetes remains undiagnosed in approximately 30% of these persons.3 Concurrent with the increasing prevalence of diabetes in the U.S. population from 1980 through 2003, the number of hospital discharges with diabetes as any listed diagnosis more than doubled, going from 2.2 to 5.1 million discharges.4 Hospital care for patients with diabetes and hyperglycemia poses a significant health economic burden in the United States, representing over 40 billion dollars in annual direct medical expenditures.5

Hyperglycemia in the hospital may be due to known diabetes, to previously unrecognized diabetes, to prediabetes, and/or to the stress of surgery or illness. Deterioration in glycemic control in the hospital setting is most commonly associated with one or more factors, including stress‐induced release of insulin counterregulatory hormones (catecholamines, cortisol, glucagon, and growth hormone), exogenous administration of high dose glucocorticoids, and suboptimal glycemic management strategies.68 In a Belgian medical intensive care unit (MICU) randomized controlled trial (RCT) of strict versus conventional glycemic control, mean blood glucose (BG) on admission to the unit in the intention to treat group was 162 70 mg/dL (n = 1200),9 and in this group's RCT of 1548 surgical intensive care unit (SICU) patients, BG > 110 mg/dL was observed in over 70% of subjects.10 Mean BG of >145 mg/dL has been reported in 39%11 and BG >200 mg/dL in anywhere from 11% to 31% of intensive care unit (ICU) patients.10, 12 For general medicine and surgery, 1 study of 2030 patients admitted to a teaching hospital revealed that 26% of admissions had a known history of DM and 12% had new hyperglycemia, as evidenced by an admission or in‐hospital fasting BG of 126 mg/dL or more or a random BG of 200 mg/dL or more on 2 or more determinations.13 National and regional estimates on hospital use maintained by the Agency for Healthcare Research and Quality include data concerning diabetes diagnoses alone, without hyperglycemia, and may be displayed by querying its Web site.14 In cardiovascular populations almost 70% of patients having a first myocardial infarction have been reported to have either known DM, previously unrecognized diabetes, or impaired glucose tolerance.15

THE EVIDENCE SUPPORTS INPATIENT GLYCEMIC CONTROL

INPATIENT GLYCEMIC CONTROL IS INCREASINGLY INCORPORATED INTO PUBLIC REPORTING, GUIDELINES, REGULATORY AGENCY, AND NATIONAL QUALITY INITIATIVE PRIORITIES

National quality initiatives, public reporting, pay‐for‐performance, and guideline‐based care continue to play an increasingly important role in the U.S. healthcare system. Over the years these initiatives have focused on various disease states (venous thromboembolism, congestive heart failure, community‐acquired pneumonia, etc.) in an attempt to standardize care and improve patient safety and quality. Inpatient hyperglycemic control is also increasingly being incorporated into public reporting, regulatory compliance, and national quality initiatives.

Professional organizations such as the ADA2 and AACE50 have published guidelines supporting improved glycemic control, the safe use of insulin, and other measures to improve care for hyperglycemic inpatients. The AACE has a Web site dedicated to hospital hyperglycemia.52 The Society of Hospital Medicine48 has created a resource room on its Web site and a workbook for improvement49 on optimizing the care of inpatients with hyperglycemia and diabetes. The guidelines and Web sites help raise awareness and educate physicians and healthcare workers in inpatient glucose management. The American Heart Association has incorporated specific recommendation regarding inpatient diabetic management in its Get With the Guidelines.53

The Joint Commission54 has developed an advanced disease‐specific certification on inpatient diabetes. Disease management programs are important components of complex healthcare systems that serve to coordinate chronic care, promote early detection and prevention, and reduce overall healthcare costs. Certification is increasingly important to providers, payers, and healthcare institutions because it demonstrates a commitment to quality and patient safety. The Joint Commission disease‐specific care certification is a patient‐centered model focusing on the delivery of clinical care and relationship between the practitioner and the patient. The evaluation and resulting certification by the Joint Commission is based on 3 core components: (1) an assessment of compliance with consensus‐based national standards; (2) the effective use of established clinical practice guidelines to manage and optimize care; and (3) an organized approach to performance measurement and improved activities.55 For inpatient diabetes, the Joint Commission program has 7 major elements following the ADA recommendations, including general recommendations regarding diabetic documentation, BG targets, preventing hypoglycemia, diabetes care providers, diabetes self‐management education, medical nutrition therapy, and BG monitoring.54 This mirrors the Call to Action Consensus Conference essential elements for successful glycemic control programs.1

Other organizations such as the Surgical Care Improvement Partnership (SCIP) and National Surgical Quality Improvement Program (NSQIP) have included perioperative glycemic control measures, as it impacts surgical wound infections. The University HealthSystem Consortium (UHC) has benchmarking data and endorses perioperative glycemic control measures, whereas the Institute for Healthcare Improvement (IHI) has focused on safe use of insulin practices in its 5 Million Lives campaign.

HOSPITALIZATION IS A MOMENT OF OPPORTUNITY TO ASSESS AND INTERVENE

The benefits of outpatient glycemic control and quality preventive care are well established, and the reduction of adverse consequences of uncontrolled diabetes are a high priority in ambulatory medicine.5658 Hospitalization provides an opportunity to identify previously undiagnosed diabetes or prediabetes and, for patients with known diabetes, to assess and impact upon the long term course of diabetes.

As a first step, unless a recent hemoglobin A1C (HbA1c) is known, among hospitalized hyperglycemic patients an HbA1C should be obtained upon admission. Greci et al.59 showed that an HbA1c level >6.0% was 100% specific (14/14) and 57% sensitive (12/21) for the diagnosis of diabetes. Among patients having known diabetes, an HbA1C elevation on admission may justify intensification of preadmission management at the time of discharge. If discharge and postdischarge adjustments of preadmission regimens are planned in response to admission A1C elevations, then the modified long‐term treatment strategy can improve the A1C in the ambulatory setting.60 Moreover, the event of hospitalization is the ideal teachable moment for patients and their caregivers to improve self‐care activities. Yet floor nurses may be overwhelmed by the tasks of patient education. For ideal patient education, both a nutritionist and a diabetes nurse educator are needed to assess compliance with medication, diet, and other aspects of care.6163 There also is need for outpatient follow‐up education. Finally, at the time of discharge, there is a duty and an opportunity for the diabetes provider to communicate with outpatient care providers about the patient's regimen and glycemic control, and also, based on information gathered during the admission, to convey any evidence that might support the need for a change of long‐term strategy.64 Unfortunately, the opportunity that hospitalization presents to assess, educate, and intervene frequently is underused.1, 8, 51, 65

LARGE GAPS EXIST BETWEEN CURRENT AND OPTIMAL CARE

Despite the evidence that inpatient glycemic control is important for patient outcomes, and despite guidelines recommending tighter inpatient glycemic control, clinical practice has been slow to change. In many institutions, inpatient glycemic management has not improved over the past decade, and large gaps remain between current practice and optimal practice.

Studies of individual institutions provide several insights into gaps in care. For example, Schnipper et al.66 examined practices on the general medicine service of an academic medical center in Boston in 2004. Among 107 prospectively identified patients with a known diagnosis of diabetes or at least 1 glucose reading >200 mg/dL (excluding patients with diabetic ketoacidosis, hyperglycemic hyperosmolar state, or pregnancy), they found scheduled long‐acting insulin prescribed in 43% of patients, scheduled short‐acting/rapid‐acting insulin in only 4% of patients, and 80 of 89 patients (90%) on the same sliding scale insulin regimen despite widely varying insulin requirements. Thirty‐one percent of glucose readings were >180 mg/dL compared with 1.2% of readings 60 mg/dL (but 11% of patients had at least 1 episode of hypoglycemia). Of the 75 patients with at least 1 episode of hyperglycemia or hypoglycemia, only 35% had any change to their insulin regimen during the first 5 days of the hospitalization.

Other studies have confirmed this concept of clinical inertia (ie, recognition of the problem but failure to act).67 A study by Cook et al.68 of all hospitalized non‐ICU patients with diabetes or hyperglycemia and length of stay of 3 days between 2001 and 2004 showed that 20% of patients had persistent hyperglycemia during the hospitalization (defined as a mean glucose >200 mg/dL). Forty‐six percent of patients whose average glucose was in the top tertile did not have their insulin regimen intensified to a combination of short‐acting/rapid‐acting and long‐acting insulin, and 35% of these patients either had no change in their total daily insulin dose or actually had a decrease in their dose when comparing the last 24 hours with the first 24 hours of hospitalization, a concept they term negative therapeutic momentum.

Perhaps the most well‐balanced view of the current state of medical practice comes from the UHC benchmarking project.69 UHC is an alliance of 90 academic health centers. For the diabetes project, each institution reviewed the records of 50 randomly selected patients over 18 years of age with at least a 72‐hour length of stay, 1 of 7 prespecified Diagnosis Related Group (DRG) codes, and at least 2 consecutive glucose readings >180 mg/dL or the receipt of insulin any time during the hospitalization. Patients with a history of pancreatic transplant, pregnant at the time of admission, receiving hospice or comfort care, or receiving insulin for a reason other than glucose management were excluded. The study showed widespread gaps in processes and outcomes (Table 1). Moreover, performance varied widely across hospitals. For example, the morning glucose in the ICU on the second measurement day was 110 mg/dL in 18% of patients for the median‐performing hospital, with a range of 0% to 67% across all 37 measured hospitals. In the non‐ICU setting on the second measurement day, 26% of patients had all BG measurements = 180 mg/dL in the median‐performing hospital, with a range of 7% to 48%. Of note, hypoglycemia was relatively uncommon: in the median hospital, 2.4% of patient‐days had 1 or more BG readings 50 mg/dL (range: 0%8.6%). Finally, in the median‐performing hospital, effective insulin therapy (defined as short‐acting/rapid‐acting and long‐acting subcutaneous insulin, continuous insulin infusion, or subcutaneous insulin pump therapy) was prescribed in 45% of patients, with a range of 12% to 77% across measured hospitals.

FREQUENT PROBLEMS WITH COMMUNICATION AND COORDINATION

Those who work closely with frontline practitioners striving to improve inpatient glycemic management have noticed other deficiencies in care.1, 70 These include: a lack of coordination between feeding, BG measurement, and insulin administration, leading to mistimed and incorrectly dosed insulin; frequent use of sliding‐scale only regimens despite evidence that they are useless at best and harmful at worst;6, 47, 60, 71 discharge summaries that often do not mention follow‐up plans for hyperglycemic management; incomplete patient educational programs; breakdowns in care at transition points; nursing and medical staffs that are unevenly educated about the proper use of insulin; and patients who are often angry or confused about their diabetes care in the hospital. Collectively, these gaps in care serve as prime targets for any glycemic control program.

HYPOGLYCEMIA IS A PROMINENT INPATIENT SAFETY CONCERN

Hypoglycemia is common in the inpatient setting and is a legitimate safety concern. In a recently reported series of 2174 hospitalized patients receiving antihyperglycemic agents, it was found that 9.5% of patients experienced a total 484 hypoglycemic episodes (defined as 60 mg/dL).72 Hypoglycemia often occurred in the setting of insulin therapy and frequently resulted from a failure to recognize trends in BG readings or other clues that a patient was at risk for developing hypoglycemia.73 A common thread is the risk created by interruption of carbohydrate intake, noted by Fischer et al.73 and once again in the recent ICU study by Vriesendorp et al.74 Sources of error include: lack of coordination between feeding and medication administration, leading to mistiming of insulin action; lack of sufficient frequency in BG monitoring; lack of clarity or uniformity in the writing of orders; failure to recognize changes in insulin requirements due to advanced age, renal failure, liver disease, or change in clinical status; steroid use with subsequent tapering or interruption; changes in feeding; failure to reconcile medications; inappropriate use of oral antihyperglycemic agents, and communication or handoff failures.

It has been difficult to sort out whether hypoglycemia is a marker of severity of illness or whether it is an independent factor leading to poor outcomes. Observational studies lend credibility to the concept that patients having congestive heart failure or myocardial infarction may be at risk for excessive mortality if their average BG resides in the low end of the normal range.7578 Sympathetic activation occurs as the threshold for hypoglycemia is approached, such as occurs at BG = 70 or 72 mg/dL.79 Patients living with BG levels observed to be in the low end of the normal range might experience more severe but unobserved and undocumented episodes of neuroglycopenia. Arrhythmia and fatality have been directly attributed to strict glycemic control.80, 81 We are confronted with the need to interpret well conducted observational studies, evaluating subgroups at risk, and using multivariate analysis to assess the impact of hypoglycemia upon outcomes.82 In such studies, we will need to examine high‐risk subgroups, including cardiac patients, in particular, for the possibility that there is a J‐shaped curve for mortality as a function of average BG.

Unfortunately, clinical inertia exists in response to hypoglycemia just as it does with hyperglycemia. One recent study examined 52 patients who received intravenous 50% dextrose solution for an episode of hypoglycemia.83 Changes to insulin regimens were subsequently made in only 21 patients (40%), and diabetes specialists agreed with the changes for 11 of these patients. The other 31 patients (60%) received no changes in treatment, and diabetes specialists agreed with that decision for only 10 of these patients.

Although some increase in hypoglycemia might be expected with initiation of tight glycemic control efforts, the solution is not to undertreat hyperglycemia. Hyperglycemia creates an unsafe setting for the treatment of illness and disease. Sliding‐scaleonly regimens are ineffective in securing glycemic control and can result in increases in hypoglycemia as well as hyperglycemic excursions.6, 66 Inappropriate withholding of insulin doses can lead to severe glycemic excursions and even iatrogenic diabetic ketoacidosis (DKA). Systems approaches to avoid the errors outlined above can minimize or even reverse the increased risk of hypoglycemia expected with tighter glycemic targets.51

A SYSTEMS APPROACH IS NEEDED FOR THESE MULTIPLE COMPLEX PROBLEMS

Care is of the hyperglycemic inpatient is inherently complex. Previously established treatments are often inappropriate under conditions of altered insulin resistance, changing patterns of nutrition and carbohydrate exposure, comorbidities, concomitant medications, and rapidly changing medical and surgical status. Patients frequently undergo changes in the route and amount of nutritional exposure, including discrete meals, continuous intravenous dextrose, nil per orem (nothing by mouth status; NPO) status, grazing on nutritional supplements or liquid diets with or without meals, bolus enteral feedings, overnight enteral feedings with daytime grazing, total parenteral nutrition, continuous peritoneal dialysis, and overnight cycling of peritoneal dialysis. Relying on individual expertise and vigilance to negotiate this complex terrain without safeguards, protocols, standardization of orders, and other systems change is impractical and unwise.

Transitions across care providers and locations lead to multiple opportunities for breakdown in the quality, consistency, and safety of care.64, 65 At the time of ward transfer or change of patient status, previous medication and monitoring orders sometimes are purged. At the time of discharge, there may be risk of continuation of anti‐hyperglycemic therapy, initiated to cover medical stress, in doses that will subsequently be unsafe.

In the face of this complexity, educational programs alone will not suffice to improve care. Institutional commitment and systems changes are essential.

MARKED IMPROVEMENT IS POSSIBLE AND TOOLS EXIST: A ROADMAP IS IN PLACE

Fortunately, a roadmap is in place to help us achieve better glycemic control, improve insulin management, and address the long list of current deficiencies in care. This is imperative to develop consistent processes in order to achieve maximum patient quality outcomes that effective glycemic management offers. This roadmap entails 4 components: (1) national awareness, (2) national guidelines, (3) consensus statements, and (4) effective tools. As mentioned above, the first two components of this roadmap are now in place.

As these national guidelines become more widely accepted, the next step will be the incorporation of this into programs like Pay‐for Performance and the Physician Quality Reporting Initiative (PQRI), which will impact reimbursement to both hospitals and providers.

Regarding the third component, a recent multidisciplinary consensus conference1 outlined the essential elements needed for successful implementation of an inpatient glycemic control program which include:

• An appropriate level of administrative support.

• Formation of a multidisciplinary steering committee to drive the development of initiatives and empowered to enact change.

• Assessment of current processes, quality of care, and barriers to practice change.

• Development and implementation of interventions including standardized order sets, protocols, policies and algorithms with associated educational programs.

• Metrics for evaluation of glycemic control, hypoglycemia, insulin use patterns, and other aspects of care.

Finally, extensive resources and effective tools are now available to help institutions achieve better inpatient glucose control. The Society of Hospital Medicine (SHM), in conjunction with the ADA, AACE, the American College of Physicians (ACP), the Case Management Society of America (CMSA), the American Society of Consultant Pharmacists, nursing, and diabetic educators have all partnered to produce a comprehensive guide to effective implementation of glycemic control and preventing hypoglycemia.49 This comprehensive workbook is a proven performance improvement framework and is available on the SHM Web site.48 Details and examples of all essential elements are covered in this workbook along with opportunities for marked improvement bolstered by integration of high reliability design features and attention to effective implementation techniques. The remainder of this supplement crystallizes a substantial portion of this material. The AACE has also recently offered a valuable web‐based resource to encourage institutional glycemic control efforts.49

GLYCEMIC CONTROL INITIATIVES CAN BE COST‐EFFECTIVE

Achieving optimal glycemic control safely requires monitoring, education, and other measures, which can be expensive, labor intensive, and require coordination of the services of many hospital divisions. This incremental expense has been shown to be cost‐effective in a variety of settings.1, 84, 85 The costs of glycemic control initiatives have demonstrated a good return on investment via:

• Improved LOS, readmission rates, morbidity, and mortality.

• Improved documentation of patient acuity and related payment for acuity.

• Income generated via incremental physician and allied health professional billing.

CONCLUSION AND SUMMARY

Evidence exists that appropriate management of hyperglycemia improves outcomes, whereas the current state of affairs is that most medical centers currently manage this suboptimally. This is concerning given the magnitude of diabetes and hyperglycemia in our inpatient setting in the United States. To bring awareness to this issue, multiple initiatives (guidelines, certification programs, workbooks, etc.) are available by various organizations including the ADA, AACE, SCIP, NSQIP, IHI, UHC, the Joint Commission, and SHM. However, this is not enough. Change occurs at the local level, and institutional prioritization and support is needed to empower a multidisciplinary steering committee, with appropriate administrative support, to standardize and improve systems in the face of substantial cultural issues and complex barriers. Improved data collection and reporting, incremental monitoring, creation of metrics, and improved documentation are an absolutely necessary necessity to achieve breakthrough levels of improvement.

Now the time is right to make an assertive effort to improve inpatient glycemic control and related issues, and push for appropriate support at your institution to help achieve this in the interest of patient safety and optimal outcomes.

At a hospital at which I work, every patient who presents to the emergency department with a suspected stroke or transient ischemic attack is evaluated by the stroke team. Per protocol, the team rapidly assesses each patient, orders diagnostic and therapeutic interventions and then refers each and every patient to the hospitalist service for admission and medical comanagement. At no point is any consideration given to whether the patients actually have medical comorbidities, or if a hospitalist will have anything meaningful to add to the care. The firmly set expectation is that hospitalists admit all stroke patients for the purposes of comanagement, while the neurologists consult.

Comanagement has become a mainstay of hospital medicine.1 It is predicated upon the assumption that surgical and specialty patients benefit when their medical comorbidities are managed by hospitalists. It differs conceptually from traditional medical consultation in that hospitalists collaboratively manage patients with surgeons or specialists, sharing responsibility and authority. In practice, however, comanagement varies widely, ranging from a model of care indistinguishable from traditional medical consultation to one where hospitalists admit and assume primary responsibility for surgical and specialty patients. This variability makes it difficult to study and make generalizations about the role and impact of hospitalist comanagement. Nonetheless, recent evidence suggests that hospitalist consultation and comanagement may not be as effective as originally anticipated.

In a 2008 observational cohort study of patients undergoing surgery at an academic medical center, Auerbach et al demonstrated that medical consultation (provided by hospitalists) did not improve glycemic control or increase the likelihood of perioperative beta‐blockade and venous thromboembolism prophylaxis.2 Patients who received consultation had longer adjusted lengths of stay (12.98% longer; 95% confidence interval, 1.61%‐25.61%) and higher adjusted costs (24.36% higher; 95% confidence interval, 13.54%‐36.34%). Notwithstanding the limited generalizability of this study to community hospitals, it has raised concerns that hospitalist consultation does not automatically improve quality of care or cost effectiveness.3

Several other recent trials have also helped to define where hospitalist comanagement may work well and where it may not. In 2004, Huddleston et al published the Hospitalist Orthopedic Team (HOT) trial, the first randomized prospective trial comparing hospitalist‐surgical comanagement to standard care.4 A total of 526 patients undergoing elective hip or knee replacement surgery at the Mayo Clinic were randomized to either standard orthopedic care with consultation as needed, or immediate hospitalist comanagement. The outcomes were disappointing. Hospitalist comanagement reduced minor complications (such as incidence of urinary tract infections, fever, and hyponatremia) but had no effect on moderate or major complications. The HOT intervention modestly reduced adjusted length of stay (LOS), defined as the point at which patients were deemed stable for discharge, by 0.5 days, but had no impact on actual LOS or cost per case. Not surprisingly, orthopedic surgeons and nurses preferred the HOT model of care over the standard model. One year later, Phy et al analyzed outcomes for patients admitted with hip fracture at the same institution.5 This retrospective cohort study compared patients who were admitted to either a standard orthopedic service or to a hospitalist team. In contrast to the HOT trial, hospitalist comanagement of hip fracture patients decreased time to surgery and lowered LOS by 2.2 days without compromising patient outcomes.

How did two trials that occurred roughly simultaneously at the same hospital, involving the same hospitalists and orthopedic surgeons generate such different outcomes? A likely answer is patient selection. Patients who undergo elective joint replacement are usually relatively healthy. They are almost always ambulatory and their comorbidities, when present, are generally reasonably compensated. As a rule, they fare well postoperatively, as evidenced by the 1.3% major complication rate demonstrated in the HOT trial.3 In contrast, hip fracture patients are older, have greater comorbidity and are at remarkably high risk for developing perioperative delirium.3, 4, 6 By definition, their urgent/emergent hip surgery stratifies them to a higher operative risk category than patients who undergo elective joint replacement.7 Half of hip fracture patients do not return to premorbid levels of function, and the 1‐year mortality rate has been estimated to be as high as 25%.6, 8 Given these differences, it is not surprising that hip fracture patients are more likely than elective joint replacement patients to respond favorably to hospitalist comanagement.

In 2007, Simon et al published a retrospective study of 739 pediatric spinal fusion patients at Childrens' Hospital in Denver.9 Beginning in 2004, hospitalists comanaged selected, high‐risk surgical patients (14 of 115 spinal fusion patients, or 12%). Over the course of the study, the mean LOS for low‐risk patients decreased by 21% but the mean LOS for the high‐risk, hospitalist‐comanaged patients decreased by 28%; a 33% relative reduction favoring hospitalist‐managed patients. By targeting selected high‐risk patients, pediatric hospitalists were able to improve upon LOS reductions that occurred systemically across the entire spinal fusion program. Also in 2007, Southern et al compared outcomes for 2,913 patients admitted by full‐time teaching hospitalists vs 6,124 patients admitted by nonhospitalists at Montefiore Medical Center, Bronx, New York.10 Mean LOS for patients admitted to the hospitalist service was 5.01 days vs 5.87 days for the nonhospitalists. Subgroup analysis demonstrated the greatest LOS differentials for patients requiring close clinical monitoring (heart failure, stroke, asthma, or pneumonia) or complex discharge planning.

Although these studies, performed at large academic medical centers, may have limited generalizability, they support the common‐sense notion that hospitalists most benefit patients who are sick, frail, and medically or socially complex. As a corollary, hospitalists probably offer relatively little benefit to surgical and specialty patients who are young or have compensated medical comorbidities and/or straightforward disposition plans. The enormous variability across healthcare institutions makes it difficult if not impossible to define a patient acuity or complexity cutoff below which hospitalist comanagement is unlikely to be beneficial. Nonetheless, some degree of common sense can be applied. As a case in point, a hospitalist probably adds little value to the care of a basically healthy patient with a hemodynamically stable upper gastrointestinal bleed. Despite this, in many institutions, hospitalists admit or comanage all gastroenterology patients, irrespective of their diagnosis, acuity, or complexity.11

One can even hypothesize that hospitalist comanagement may potentially inject risk into patient care. Admitting that patient with a stable upper gastrointestinal bleed to a hospitalist service may delay the gastroenterologist's involvement and initiation of the necessary endoscopy. Having assumed that the hospitalist is running the show, the gastroenterologist may pay insufficient attention to the patient. The hospitalist and gastroenterologist may give conflicting orders and reports that confuse patients, families, and hospital staff, ultimately increasing the likelihood of medical errors.

Ultimately, the risks inherent in adding complexity into patient care must be balanced against the potential benefits. For patients who are sick, frail, or complicated, the risk‐benefit ratio probably tilts in favor of comanagement. However, for generally healthy patients, it is conceivable that adding complexity negates (or worse yet, exceeds) the putative benefits of comanagement.

Given the potential limitations of hospitalist comanagement, why are hospitalists admitting or managing broad and unselected populations of surgical and specialty patients? Hospital leaders have suggested that hospitalist comanagement may protect overstretched surgeons and specialists and extend their capacity. A hospital with only one neurosurgeon on staff might reasonably ask its hospitalists to primarily manage carefully selected low‐acuity neurosurgical patients, allowing the neurosurgeon to serve as a consultant. However, in communities where specialists and surgeons are abundant, this justification is less credible. In such cases, it is difficult not to suspect that the primary reason that hospitalists admit surgical and specialty patients is to enhance the income and quality of life of the surgeons and specialists.

Expanding hospitalist comanagement services for no other reason than to keep specialists and surgeons happy might be justifiable if hospital medicine was not faced with its own critical manpower shortage. Hospital medicine is expected to grow from approximately 20,000 current practitioners to more than 40,000 within a decade.12 The growing shortage of qualified hospitalists has become a preoccupation for hospitalist employers across the country.13 At its 2006 strategic planning retreat, the Board of Directors of the Society of Hospital Medicine identified this issue as one of the greatest threats to the future health of hospital medicine.14 Demand for hospitalists will not abate for at least a decade, which will leave many hospitalist programs significantly understaffed for the foreseeable future. Understaffing forces hospitalist programs to lower hiring standards, jeopardizes patient care, accelerates physician burnout, and may ultimately destabilize hospital medicine.15 Understaffed hospitalist programs should be very circumspect about how and where they expand their clinical coverage.

Another principle underlying hospitalist comanagement is that it improves care by allowing surgeons and specialists to focus on their areas of expertise. Surgeons and specialists who do not have to manage their patients' medical issues can presumably spend more time focusing on their own disciplines. Although this argument is conceptually appealing, there is no evidence that this actually occurs. In fact, it is equally conceivable that hospitalist comanagement could jeopardize care by disengaging surgeons and specialists from their patients' progress (or lack thereof). Furthermore, evidence suggests that hospitalists are underprepared to manage diagnoses that have historically been the purview of surgeons and specialists. Practicing hospitalists who manage acute neurological and neurosurgical conditions, orthopedic trauma, and acute psychiatric illnesses have reported relative undertraining in all of these disease states.10, 16 Generally, hospitalists are expected to deliver this care in the absence of any regime to assess their competence, provide targeted training to fill knowledge gaps, and monitor their progress. At minimum, this should raise concerns about the quality and consistency of care that hospitalists provide to nonmedical patients.

Finally, working collaboratively with other specialties should be a major professional benefit of comanagement. In well‐designed comanagement arrangements, hospitalists and specialists work equitably under clearly defined and mutually agreed upon rules of engagement. They share responsibility for patients, collaborate to improve care, and teach and learn from each other. Unfortunately, in many instances, the power structure becomes lopsided, with surgeons and specialists dictating how, when, and why hospitalists manage their patients.17 Emergency departments have learned to default surgical and specialty patient admissions to hospitalists when surgeons and specialists balk. Hospital administrations may tacitly or overtly expect their financially subsidized hospitalists to cheerfully accept any and all referrals, irrespective of how inappropriate they may be. Practicing hospitalists frequently complain about their subordinate status and inability to control their working conditions, both of which are identified risk factors for career dissatisfaction and burnout.14, 16, 18 Once again, as a specialty facing a critical manpower shortage, hospitalist programs should be very attuned to defining work conditions that foster career satisfaction and physician retention.

REFRAMING COMANAGEMENT

The history of healthcare is laden with examples of new ideas that were widely and indiscriminately adopted only to subsequently fail to withstand rigorous scrutiny.19, 20 The unchecked expansion of hospitalist comanagement has the potential to become another case in point. In the absence of clear definitions of comanagement and good evidence to define best practices, hospitalists are left to use their best judgment to define the parameters of their comanagement services. At minimum, hospitalist leaders should ask some basic questions as they ponder potential comanagement relationships:

• Why are we being asked to provide this service?

• Do the patients have comorbidities that require our input?

• Is there a legitimate quality or efficiency case to be made to support our participation?

• Do we have the manpower to provide the service? If not, what will suffer as a result?

• Will the relationship be equitable?

• What might go wrong?

Comanagement is an appealing construct that has grown to fill many niches of healthcare delivery.10 Given compelling reasons to be skeptical about the purported benefits of comanagement, hospitalists should be circumspect about how and where they offer such services. Comanagement should be applied carefully and methodically, paying close attention to the consequences, intended and unintended. Applying comanagement in a rational, evidence‐based, and sustainable fashion will ultimately better serve patients, the healthcare community, and hospital medicine.

At a hospital at which I work, every patient who presents to the emergency department with a suspected stroke or transient ischemic attack is evaluated by the stroke team. Per protocol, the team rapidly assesses each patient, orders diagnostic and therapeutic interventions and then refers each and every patient to the hospitalist service for admission and medical comanagement. At no point is any consideration given to whether the patients actually have medical comorbidities, or if a hospitalist will have anything meaningful to add to the care. The firmly set expectation is that hospitalists admit all stroke patients for the purposes of comanagement, while the neurologists consult.

Comanagement has become a mainstay of hospital medicine.1 It is predicated upon the assumption that surgical and specialty patients benefit when their medical comorbidities are managed by hospitalists. It differs conceptually from traditional medical consultation in that hospitalists collaboratively manage patients with surgeons or specialists, sharing responsibility and authority. In practice, however, comanagement varies widely, ranging from a model of care indistinguishable from traditional medical consultation to one where hospitalists admit and assume primary responsibility for surgical and specialty patients. This variability makes it difficult to study and make generalizations about the role and impact of hospitalist comanagement. Nonetheless, recent evidence suggests that hospitalist consultation and comanagement may not be as effective as originally anticipated.

In a 2008 observational cohort study of patients undergoing surgery at an academic medical center, Auerbach et al demonstrated that medical consultation (provided by hospitalists) did not improve glycemic control or increase the likelihood of perioperative beta‐blockade and venous thromboembolism prophylaxis.2 Patients who received consultation had longer adjusted lengths of stay (12.98% longer; 95% confidence interval, 1.61%‐25.61%) and higher adjusted costs (24.36% higher; 95% confidence interval, 13.54%‐36.34%). Notwithstanding the limited generalizability of this study to community hospitals, it has raised concerns that hospitalist consultation does not automatically improve quality of care or cost effectiveness.3

Several other recent trials have also helped to define where hospitalist comanagement may work well and where it may not. In 2004, Huddleston et al published the Hospitalist Orthopedic Team (HOT) trial, the first randomized prospective trial comparing hospitalist‐surgical comanagement to standard care.4 A total of 526 patients undergoing elective hip or knee replacement surgery at the Mayo Clinic were randomized to either standard orthopedic care with consultation as needed, or immediate hospitalist comanagement. The outcomes were disappointing. Hospitalist comanagement reduced minor complications (such as incidence of urinary tract infections, fever, and hyponatremia) but had no effect on moderate or major complications. The HOT intervention modestly reduced adjusted length of stay (LOS), defined as the point at which patients were deemed stable for discharge, by 0.5 days, but had no impact on actual LOS or cost per case. Not surprisingly, orthopedic surgeons and nurses preferred the HOT model of care over the standard model. One year later, Phy et al analyzed outcomes for patients admitted with hip fracture at the same institution.5 This retrospective cohort study compared patients who were admitted to either a standard orthopedic service or to a hospitalist team. In contrast to the HOT trial, hospitalist comanagement of hip fracture patients decreased time to surgery and lowered LOS by 2.2 days without compromising patient outcomes.

How did two trials that occurred roughly simultaneously at the same hospital, involving the same hospitalists and orthopedic surgeons generate such different outcomes? A likely answer is patient selection. Patients who undergo elective joint replacement are usually relatively healthy. They are almost always ambulatory and their comorbidities, when present, are generally reasonably compensated. As a rule, they fare well postoperatively, as evidenced by the 1.3% major complication rate demonstrated in the HOT trial.3 In contrast, hip fracture patients are older, have greater comorbidity and are at remarkably high risk for developing perioperative delirium.3, 4, 6 By definition, their urgent/emergent hip surgery stratifies them to a higher operative risk category than patients who undergo elective joint replacement.7 Half of hip fracture patients do not return to premorbid levels of function, and the 1‐year mortality rate has been estimated to be as high as 25%.6, 8 Given these differences, it is not surprising that hip fracture patients are more likely than elective joint replacement patients to respond favorably to hospitalist comanagement.

In 2007, Simon et al published a retrospective study of 739 pediatric spinal fusion patients at Childrens' Hospital in Denver.9 Beginning in 2004, hospitalists comanaged selected, high‐risk surgical patients (14 of 115 spinal fusion patients, or 12%). Over the course of the study, the mean LOS for low‐risk patients decreased by 21% but the mean LOS for the high‐risk, hospitalist‐comanaged patients decreased by 28%; a 33% relative reduction favoring hospitalist‐managed patients. By targeting selected high‐risk patients, pediatric hospitalists were able to improve upon LOS reductions that occurred systemically across the entire spinal fusion program. Also in 2007, Southern et al compared outcomes for 2,913 patients admitted by full‐time teaching hospitalists vs 6,124 patients admitted by nonhospitalists at Montefiore Medical Center, Bronx, New York.10 Mean LOS for patients admitted to the hospitalist service was 5.01 days vs 5.87 days for the nonhospitalists. Subgroup analysis demonstrated the greatest LOS differentials for patients requiring close clinical monitoring (heart failure, stroke, asthma, or pneumonia) or complex discharge planning.

Although these studies, performed at large academic medical centers, may have limited generalizability, they support the common‐sense notion that hospitalists most benefit patients who are sick, frail, and medically or socially complex. As a corollary, hospitalists probably offer relatively little benefit to surgical and specialty patients who are young or have compensated medical comorbidities and/or straightforward disposition plans. The enormous variability across healthcare institutions makes it difficult if not impossible to define a patient acuity or complexity cutoff below which hospitalist comanagement is unlikely to be beneficial. Nonetheless, some degree of common sense can be applied. As a case in point, a hospitalist probably adds little value to the care of a basically healthy patient with a hemodynamically stable upper gastrointestinal bleed. Despite this, in many institutions, hospitalists admit or comanage all gastroenterology patients, irrespective of their diagnosis, acuity, or complexity.11

One can even hypothesize that hospitalist comanagement may potentially inject risk into patient care. Admitting that patient with a stable upper gastrointestinal bleed to a hospitalist service may delay the gastroenterologist's involvement and initiation of the necessary endoscopy. Having assumed that the hospitalist is running the show, the gastroenterologist may pay insufficient attention to the patient. The hospitalist and gastroenterologist may give conflicting orders and reports that confuse patients, families, and hospital staff, ultimately increasing the likelihood of medical errors.

Ultimately, the risks inherent in adding complexity into patient care must be balanced against the potential benefits. For patients who are sick, frail, or complicated, the risk‐benefit ratio probably tilts in favor of comanagement. However, for generally healthy patients, it is conceivable that adding complexity negates (or worse yet, exceeds) the putative benefits of comanagement.

Given the potential limitations of hospitalist comanagement, why are hospitalists admitting or managing broad and unselected populations of surgical and specialty patients? Hospital leaders have suggested that hospitalist comanagement may protect overstretched surgeons and specialists and extend their capacity. A hospital with only one neurosurgeon on staff might reasonably ask its hospitalists to primarily manage carefully selected low‐acuity neurosurgical patients, allowing the neurosurgeon to serve as a consultant. However, in communities where specialists and surgeons are abundant, this justification is less credible. In such cases, it is difficult not to suspect that the primary reason that hospitalists admit surgical and specialty patients is to enhance the income and quality of life of the surgeons and specialists.

Expanding hospitalist comanagement services for no other reason than to keep specialists and surgeons happy might be justifiable if hospital medicine was not faced with its own critical manpower shortage. Hospital medicine is expected to grow from approximately 20,000 current practitioners to more than 40,000 within a decade.12 The growing shortage of qualified hospitalists has become a preoccupation for hospitalist employers across the country.13 At its 2006 strategic planning retreat, the Board of Directors of the Society of Hospital Medicine identified this issue as one of the greatest threats to the future health of hospital medicine.14 Demand for hospitalists will not abate for at least a decade, which will leave many hospitalist programs significantly understaffed for the foreseeable future. Understaffing forces hospitalist programs to lower hiring standards, jeopardizes patient care, accelerates physician burnout, and may ultimately destabilize hospital medicine.15 Understaffed hospitalist programs should be very circumspect about how and where they expand their clinical coverage.

Another principle underlying hospitalist comanagement is that it improves care by allowing surgeons and specialists to focus on their areas of expertise. Surgeons and specialists who do not have to manage their patients' medical issues can presumably spend more time focusing on their own disciplines. Although this argument is conceptually appealing, there is no evidence that this actually occurs. In fact, it is equally conceivable that hospitalist comanagement could jeopardize care by disengaging surgeons and specialists from their patients' progress (or lack thereof). Furthermore, evidence suggests that hospitalists are underprepared to manage diagnoses that have historically been the purview of surgeons and specialists. Practicing hospitalists who manage acute neurological and neurosurgical conditions, orthopedic trauma, and acute psychiatric illnesses have reported relative undertraining in all of these disease states.10, 16 Generally, hospitalists are expected to deliver this care in the absence of any regime to assess their competence, provide targeted training to fill knowledge gaps, and monitor their progress. At minimum, this should raise concerns about the quality and consistency of care that hospitalists provide to nonmedical patients.

Finally, working collaboratively with other specialties should be a major professional benefit of comanagement. In well‐designed comanagement arrangements, hospitalists and specialists work equitably under clearly defined and mutually agreed upon rules of engagement. They share responsibility for patients, collaborate to improve care, and teach and learn from each other. Unfortunately, in many instances, the power structure becomes lopsided, with surgeons and specialists dictating how, when, and why hospitalists manage their patients.17 Emergency departments have learned to default surgical and specialty patient admissions to hospitalists when surgeons and specialists balk. Hospital administrations may tacitly or overtly expect their financially subsidized hospitalists to cheerfully accept any and all referrals, irrespective of how inappropriate they may be. Practicing hospitalists frequently complain about their subordinate status and inability to control their working conditions, both of which are identified risk factors for career dissatisfaction and burnout.14, 16, 18 Once again, as a specialty facing a critical manpower shortage, hospitalist programs should be very attuned to defining work conditions that foster career satisfaction and physician retention.

REFRAMING COMANAGEMENT

The history of healthcare is laden with examples of new ideas that were widely and indiscriminately adopted only to subsequently fail to withstand rigorous scrutiny.19, 20 The unchecked expansion of hospitalist comanagement has the potential to become another case in point. In the absence of clear definitions of comanagement and good evidence to define best practices, hospitalists are left to use their best judgment to define the parameters of their comanagement services. At minimum, hospitalist leaders should ask some basic questions as they ponder potential comanagement relationships:

• Why are we being asked to provide this service?

• Do the patients have comorbidities that require our input?

• Is there a legitimate quality or efficiency case to be made to support our participation?

• Do we have the manpower to provide the service? If not, what will suffer as a result?

• Will the relationship be equitable?

• What might go wrong?

Comanagement is an appealing construct that has grown to fill many niches of healthcare delivery.10 Given compelling reasons to be skeptical about the purported benefits of comanagement, hospitalists should be circumspect about how and where they offer such services. Comanagement should be applied carefully and methodically, paying close attention to the consequences, intended and unintended. Applying comanagement in a rational, evidence‐based, and sustainable fashion will ultimately better serve patients, the healthcare community, and hospital medicine.

At a hospital at which I work, every patient who presents to the emergency department with a suspected stroke or transient ischemic attack is evaluated by the stroke team. Per protocol, the team rapidly assesses each patient, orders diagnostic and therapeutic interventions and then refers each and every patient to the hospitalist service for admission and medical comanagement. At no point is any consideration given to whether the patients actually have medical comorbidities, or if a hospitalist will have anything meaningful to add to the care. The firmly set expectation is that hospitalists admit all stroke patients for the purposes of comanagement, while the neurologists consult.

Comanagement has become a mainstay of hospital medicine.1 It is predicated upon the assumption that surgical and specialty patients benefit when their medical comorbidities are managed by hospitalists. It differs conceptually from traditional medical consultation in that hospitalists collaboratively manage patients with surgeons or specialists, sharing responsibility and authority. In practice, however, comanagement varies widely, ranging from a model of care indistinguishable from traditional medical consultation to one where hospitalists admit and assume primary responsibility for surgical and specialty patients. This variability makes it difficult to study and make generalizations about the role and impact of hospitalist comanagement. Nonetheless, recent evidence suggests that hospitalist consultation and comanagement may not be as effective as originally anticipated.

In a 2008 observational cohort study of patients undergoing surgery at an academic medical center, Auerbach et al demonstrated that medical consultation (provided by hospitalists) did not improve glycemic control or increase the likelihood of perioperative beta‐blockade and venous thromboembolism prophylaxis.2 Patients who received consultation had longer adjusted lengths of stay (12.98% longer; 95% confidence interval, 1.61%‐25.61%) and higher adjusted costs (24.36% higher; 95% confidence interval, 13.54%‐36.34%). Notwithstanding the limited generalizability of this study to community hospitals, it has raised concerns that hospitalist consultation does not automatically improve quality of care or cost effectiveness.3

Several other recent trials have also helped to define where hospitalist comanagement may work well and where it may not. In 2004, Huddleston et al published the Hospitalist Orthopedic Team (HOT) trial, the first randomized prospective trial comparing hospitalist‐surgical comanagement to standard care.4 A total of 526 patients undergoing elective hip or knee replacement surgery at the Mayo Clinic were randomized to either standard orthopedic care with consultation as needed, or immediate hospitalist comanagement. The outcomes were disappointing. Hospitalist comanagement reduced minor complications (such as incidence of urinary tract infections, fever, and hyponatremia) but had no effect on moderate or major complications. The HOT intervention modestly reduced adjusted length of stay (LOS), defined as the point at which patients were deemed stable for discharge, by 0.5 days, but had no impact on actual LOS or cost per case. Not surprisingly, orthopedic surgeons and nurses preferred the HOT model of care over the standard model. One year later, Phy et al analyzed outcomes for patients admitted with hip fracture at the same institution.5 This retrospective cohort study compared patients who were admitted to either a standard orthopedic service or to a hospitalist team. In contrast to the HOT trial, hospitalist comanagement of hip fracture patients decreased time to surgery and lowered LOS by 2.2 days without compromising patient outcomes.

How did two trials that occurred roughly simultaneously at the same hospital, involving the same hospitalists and orthopedic surgeons generate such different outcomes? A likely answer is patient selection. Patients who undergo elective joint replacement are usually relatively healthy. They are almost always ambulatory and their comorbidities, when present, are generally reasonably compensated. As a rule, they fare well postoperatively, as evidenced by the 1.3% major complication rate demonstrated in the HOT trial.3 In contrast, hip fracture patients are older, have greater comorbidity and are at remarkably high risk for developing perioperative delirium.3, 4, 6 By definition, their urgent/emergent hip surgery stratifies them to a higher operative risk category than patients who undergo elective joint replacement.7 Half of hip fracture patients do not return to premorbid levels of function, and the 1‐year mortality rate has been estimated to be as high as 25%.6, 8 Given these differences, it is not surprising that hip fracture patients are more likely than elective joint replacement patients to respond favorably to hospitalist comanagement.

In 2007, Simon et al published a retrospective study of 739 pediatric spinal fusion patients at Childrens' Hospital in Denver.9 Beginning in 2004, hospitalists comanaged selected, high‐risk surgical patients (14 of 115 spinal fusion patients, or 12%). Over the course of the study, the mean LOS for low‐risk patients decreased by 21% but the mean LOS for the high‐risk, hospitalist‐comanaged patients decreased by 28%; a 33% relative reduction favoring hospitalist‐managed patients. By targeting selected high‐risk patients, pediatric hospitalists were able to improve upon LOS reductions that occurred systemically across the entire spinal fusion program. Also in 2007, Southern et al compared outcomes for 2,913 patients admitted by full‐time teaching hospitalists vs 6,124 patients admitted by nonhospitalists at Montefiore Medical Center, Bronx, New York.10 Mean LOS for patients admitted to the hospitalist service was 5.01 days vs 5.87 days for the nonhospitalists. Subgroup analysis demonstrated the greatest LOS differentials for patients requiring close clinical monitoring (heart failure, stroke, asthma, or pneumonia) or complex discharge planning.

Although these studies, performed at large academic medical centers, may have limited generalizability, they support the common‐sense notion that hospitalists most benefit patients who are sick, frail, and medically or socially complex. As a corollary, hospitalists probably offer relatively little benefit to surgical and specialty patients who are young or have compensated medical comorbidities and/or straightforward disposition plans. The enormous variability across healthcare institutions makes it difficult if not impossible to define a patient acuity or complexity cutoff below which hospitalist comanagement is unlikely to be beneficial. Nonetheless, some degree of common sense can be applied. As a case in point, a hospitalist probably adds little value to the care of a basically healthy patient with a hemodynamically stable upper gastrointestinal bleed. Despite this, in many institutions, hospitalists admit or comanage all gastroenterology patients, irrespective of their diagnosis, acuity, or complexity.11

One can even hypothesize that hospitalist comanagement may potentially inject risk into patient care. Admitting that patient with a stable upper gastrointestinal bleed to a hospitalist service may delay the gastroenterologist's involvement and initiation of the necessary endoscopy. Having assumed that the hospitalist is running the show, the gastroenterologist may pay insufficient attention to the patient. The hospitalist and gastroenterologist may give conflicting orders and reports that confuse patients, families, and hospital staff, ultimately increasing the likelihood of medical errors.

Ultimately, the risks inherent in adding complexity into patient care must be balanced against the potential benefits. For patients who are sick, frail, or complicated, the risk‐benefit ratio probably tilts in favor of comanagement. However, for generally healthy patients, it is conceivable that adding complexity negates (or worse yet, exceeds) the putative benefits of comanagement.

Given the potential limitations of hospitalist comanagement, why are hospitalists admitting or managing broad and unselected populations of surgical and specialty patients? Hospital leaders have suggested that hospitalist comanagement may protect overstretched surgeons and specialists and extend their capacity. A hospital with only one neurosurgeon on staff might reasonably ask its hospitalists to primarily manage carefully selected low‐acuity neurosurgical patients, allowing the neurosurgeon to serve as a consultant. However, in communities where specialists and surgeons are abundant, this justification is less credible. In such cases, it is difficult not to suspect that the primary reason that hospitalists admit surgical and specialty patients is to enhance the income and quality of life of the surgeons and specialists.

Expanding hospitalist comanagement services for no other reason than to keep specialists and surgeons happy might be justifiable if hospital medicine was not faced with its own critical manpower shortage. Hospital medicine is expected to grow from approximately 20,000 current practitioners to more than 40,000 within a decade.12 The growing shortage of qualified hospitalists has become a preoccupation for hospitalist employers across the country.13 At its 2006 strategic planning retreat, the Board of Directors of the Society of Hospital Medicine identified this issue as one of the greatest threats to the future health of hospital medicine.14 Demand for hospitalists will not abate for at least a decade, which will leave many hospitalist programs significantly understaffed for the foreseeable future. Understaffing forces hospitalist programs to lower hiring standards, jeopardizes patient care, accelerates physician burnout, and may ultimately destabilize hospital medicine.15 Understaffed hospitalist programs should be very circumspect about how and where they expand their clinical coverage.

Another principle underlying hospitalist comanagement is that it improves care by allowing surgeons and specialists to focus on their areas of expertise. Surgeons and specialists who do not have to manage their patients' medical issues can presumably spend more time focusing on their own disciplines. Although this argument is conceptually appealing, there is no evidence that this actually occurs. In fact, it is equally conceivable that hospitalist comanagement could jeopardize care by disengaging surgeons and specialists from their patients' progress (or lack thereof). Furthermore, evidence suggests that hospitalists are underprepared to manage diagnoses that have historically been the purview of surgeons and specialists. Practicing hospitalists who manage acute neurological and neurosurgical conditions, orthopedic trauma, and acute psychiatric illnesses have reported relative undertraining in all of these disease states.10, 16 Generally, hospitalists are expected to deliver this care in the absence of any regime to assess their competence, provide targeted training to fill knowledge gaps, and monitor their progress. At minimum, this should raise concerns about the quality and consistency of care that hospitalists provide to nonmedical patients.

Finally, working collaboratively with other specialties should be a major professional benefit of comanagement. In well‐designed comanagement arrangements, hospitalists and specialists work equitably under clearly defined and mutually agreed upon rules of engagement. They share responsibility for patients, collaborate to improve care, and teach and learn from each other. Unfortunately, in many instances, the power structure becomes lopsided, with surgeons and specialists dictating how, when, and why hospitalists manage their patients.17 Emergency departments have learned to default surgical and specialty patient admissions to hospitalists when surgeons and specialists balk. Hospital administrations may tacitly or overtly expect their financially subsidized hospitalists to cheerfully accept any and all referrals, irrespective of how inappropriate they may be. Practicing hospitalists frequently complain about their subordinate status and inability to control their working conditions, both of which are identified risk factors for career dissatisfaction and burnout.14, 16, 18 Once again, as a specialty facing a critical manpower shortage, hospitalist programs should be very attuned to defining work conditions that foster career satisfaction and physician retention.

REFRAMING COMANAGEMENT

The history of healthcare is laden with examples of new ideas that were widely and indiscriminately adopted only to subsequently fail to withstand rigorous scrutiny.19, 20 The unchecked expansion of hospitalist comanagement has the potential to become another case in point. In the absence of clear definitions of comanagement and good evidence to define best practices, hospitalists are left to use their best judgment to define the parameters of their comanagement services. At minimum, hospitalist leaders should ask some basic questions as they ponder potential comanagement relationships:

• Why are we being asked to provide this service?

• Do the patients have comorbidities that require our input?

• Is there a legitimate quality or efficiency case to be made to support our participation?

• Do we have the manpower to provide the service? If not, what will suffer as a result?

• Will the relationship be equitable?

• What might go wrong?

Comanagement is an appealing construct that has grown to fill many niches of healthcare delivery.10 Given compelling reasons to be skeptical about the purported benefits of comanagement, hospitalists should be circumspect about how and where they offer such services. Comanagement should be applied carefully and methodically, paying close attention to the consequences, intended and unintended. Applying comanagement in a rational, evidence‐based, and sustainable fashion will ultimately better serve patients, the healthcare community, and hospital medicine.

A 90‐year‐old female nursing home resident was admitted for hyponatremia and altered mental status. A Foley catheter was placed on admission. On hospital day 2, the Foley catheter was found to be draining violet urine. Urinalysis showed a pH of 9.0, numerous white cells, leukocyte esterase, and bacteria. Urine culture grew Proteus mirabilis. Purple Urine Bag Syndrome (PUBS) is a rare phenomenon associated with alkaline urine due to a urinary tract infection. The patient was treated with ciprofloxacin, and her urine returned to a pale yellow color. While alarming to patients and providers alike, PUBS is a benign herald of urinary tract infection, often in an elderly woman with constipation. In normal individuals, tryptophan is metabolized to indole by gut flora, which is in turn conjugated to indoxyl sulfate (IS) by the liver. Urine excretion of IS varies by individual. Sulfatase‐containing bacteria, notably Providencia, Klebsiella, and Proteus species, then catabolize IS to indoxyl. In an alkaline environment indoxyl isomers interact to alternately yield indigo or indirubin which jointly create the urine's characteristic violet color.0

Figure 1

A 90‐year‐old female nursing home resident was admitted for hyponatremia and altered mental status. A Foley catheter was placed on admission. On hospital day 2, the Foley catheter was found to be draining violet urine. Urinalysis showed a pH of 9.0, numerous white cells, leukocyte esterase, and bacteria. Urine culture grew Proteus mirabilis. Purple Urine Bag Syndrome (PUBS) is a rare phenomenon associated with alkaline urine due to a urinary tract infection. The patient was treated with ciprofloxacin, and her urine returned to a pale yellow color. While alarming to patients and providers alike, PUBS is a benign herald of urinary tract infection, often in an elderly woman with constipation. In normal individuals, tryptophan is metabolized to indole by gut flora, which is in turn conjugated to indoxyl sulfate (IS) by the liver. Urine excretion of IS varies by individual. Sulfatase‐containing bacteria, notably Providencia, Klebsiella, and Proteus species, then catabolize IS to indoxyl. In an alkaline environment indoxyl isomers interact to alternately yield indigo or indirubin which jointly create the urine's characteristic violet color.0

Figure 1

A 90‐year‐old female nursing home resident was admitted for hyponatremia and altered mental status. A Foley catheter was placed on admission. On hospital day 2, the Foley catheter was found to be draining violet urine. Urinalysis showed a pH of 9.0, numerous white cells, leukocyte esterase, and bacteria. Urine culture grew Proteus mirabilis. Purple Urine Bag Syndrome (PUBS) is a rare phenomenon associated with alkaline urine due to a urinary tract infection. The patient was treated with ciprofloxacin, and her urine returned to a pale yellow color. While alarming to patients and providers alike, PUBS is a benign herald of urinary tract infection, often in an elderly woman with constipation. In normal individuals, tryptophan is metabolized to indole by gut flora, which is in turn conjugated to indoxyl sulfate (IS) by the liver. Urine excretion of IS varies by individual. Sulfatase‐containing bacteria, notably Providencia, Klebsiella, and Proteus species, then catabolize IS to indoxyl. In an alkaline environment indoxyl isomers interact to alternately yield indigo or indirubin which jointly create the urine's characteristic violet color.0

Figure 1

BACKGROUND

Discharges against medical advice (AMA) account for approximately 1% of discharges for general medical patients and up to 10% and 30% for patients afflicted with HIV disease and psychiatric disorders, respectively.17 Patients discharged AMA have higher rates of readmission, longer subsequent hospital stays, and worse health outcomes.3, 5, 811 Not unexpectedly, discharges AMA are associated with overall health costs of up to 50% greater than usual discharges.2

Patients who leave AMA are more likely to have poorer social supports, to abuse alcohol, heroin, and other substances, and often have weighty psychosocial or financial concerns.1218 They are also less likely to have an established relationship with a primary care physician.19 Although studies have found that rates of discharge AMA are higher among some ethnic minorities, one recent study suggests that other patient variables, such as level of income and type of insurance, may be more closely related.7, 20 Unfortunately, many patients who leave AMA have dual sources of distress: compelling personal concerns that fuel one's wish to leave and the illness that initially caused the patient to seek care.

Physicians are often distressed by the clinical and ethical challenges of discharges AMA. How should physicians manage their conflicted obligations to respect patients' choices and to prevent harms from befalling their patients? What are physicians' obligations to their patients who leave accepting only partial or inadequate treatment plans or no treatment at all? When should physicians call into question the decision‐making capacity of patients' who make seemingly unwise or clearly dangerous judgments to leave the hospital? In addition to these sorts of concerns, physicians who discharge patients AMA enjoy no definitive legal protection from the consequences of their patients' choices.2123 In fact, good clinical judgement and careful documentation provide the best liability protection.24

Clearly, discharges AMA are problematic for patients, stressful for physicians, and resource intensive for health facilities. Therefore, efforts to understand, better manage, and ultimately decrease discharges AMA will benefit all parties. Whereas the literature on discharge AMA tends to focus on psychiatric and substance abuse patients, this review examines the professional and ethical implications of discharge AMA more generally.

Summary and Conclusions

Clinicians caring for patients who seek discharge AMA are often faced with emotionally charged and time‐pressured treatment situations. These clinicians must weigh multiple considerations for the benefit of their patients, and maintain professional standards of clinical care. Clinicians presented with these situations should (1) evaluate patients' decision‐making capacity, (2) assess the degree to which their choices are influenced by controlling external influences and mitigate these factors where possible, and (3) encourage and facilitate after‐care (Table 2).

Although discharge AMA accounts for only a small percentage of hospital discharges, its medical, emotional, and resource utilization consequences for patients as well as for physicians and hospitals is disproportionate. The clinical impacts of discharge AMA should be further investigated and specific strategies and interventions to mitigate its health effects should be validated.

BACKGROUND

Discharges against medical advice (AMA) account for approximately 1% of discharges for general medical patients and up to 10% and 30% for patients afflicted with HIV disease and psychiatric disorders, respectively.17 Patients discharged AMA have higher rates of readmission, longer subsequent hospital stays, and worse health outcomes.3, 5, 811 Not unexpectedly, discharges AMA are associated with overall health costs of up to 50% greater than usual discharges.2

Patients who leave AMA are more likely to have poorer social supports, to abuse alcohol, heroin, and other substances, and often have weighty psychosocial or financial concerns.1218 They are also less likely to have an established relationship with a primary care physician.19 Although studies have found that rates of discharge AMA are higher among some ethnic minorities, one recent study suggests that other patient variables, such as level of income and type of insurance, may be more closely related.7, 20 Unfortunately, many patients who leave AMA have dual sources of distress: compelling personal concerns that fuel one's wish to leave and the illness that initially caused the patient to seek care.

Physicians are often distressed by the clinical and ethical challenges of discharges AMA. How should physicians manage their conflicted obligations to respect patients' choices and to prevent harms from befalling their patients? What are physicians' obligations to their patients who leave accepting only partial or inadequate treatment plans or no treatment at all? When should physicians call into question the decision‐making capacity of patients' who make seemingly unwise or clearly dangerous judgments to leave the hospital? In addition to these sorts of concerns, physicians who discharge patients AMA enjoy no definitive legal protection from the consequences of their patients' choices.2123 In fact, good clinical judgement and careful documentation provide the best liability protection.24

Clearly, discharges AMA are problematic for patients, stressful for physicians, and resource intensive for health facilities. Therefore, efforts to understand, better manage, and ultimately decrease discharges AMA will benefit all parties. Whereas the literature on discharge AMA tends to focus on psychiatric and substance abuse patients, this review examines the professional and ethical implications of discharge AMA more generally.

Summary and Conclusions

Clinicians caring for patients who seek discharge AMA are often faced with emotionally charged and time‐pressured treatment situations. These clinicians must weigh multiple considerations for the benefit of their patients, and maintain professional standards of clinical care. Clinicians presented with these situations should (1) evaluate patients' decision‐making capacity, (2) assess the degree to which their choices are influenced by controlling external influences and mitigate these factors where possible, and (3) encourage and facilitate after‐care (Table 2).

Although discharge AMA accounts for only a small percentage of hospital discharges, its medical, emotional, and resource utilization consequences for patients as well as for physicians and hospitals is disproportionate. The clinical impacts of discharge AMA should be further investigated and specific strategies and interventions to mitigate its health effects should be validated.

BACKGROUND

Discharges against medical advice (AMA) account for approximately 1% of discharges for general medical patients and up to 10% and 30% for patients afflicted with HIV disease and psychiatric disorders, respectively.17 Patients discharged AMA have higher rates of readmission, longer subsequent hospital stays, and worse health outcomes.3, 5, 811 Not unexpectedly, discharges AMA are associated with overall health costs of up to 50% greater than usual discharges.2

Patients who leave AMA are more likely to have poorer social supports, to abuse alcohol, heroin, and other substances, and often have weighty psychosocial or financial concerns.1218 They are also less likely to have an established relationship with a primary care physician.19 Although studies have found that rates of discharge AMA are higher among some ethnic minorities, one recent study suggests that other patient variables, such as level of income and type of insurance, may be more closely related.7, 20 Unfortunately, many patients who leave AMA have dual sources of distress: compelling personal concerns that fuel one's wish to leave and the illness that initially caused the patient to seek care.

Physicians are often distressed by the clinical and ethical challenges of discharges AMA. How should physicians manage their conflicted obligations to respect patients' choices and to prevent harms from befalling their patients? What are physicians' obligations to their patients who leave accepting only partial or inadequate treatment plans or no treatment at all? When should physicians call into question the decision‐making capacity of patients' who make seemingly unwise or clearly dangerous judgments to leave the hospital? In addition to these sorts of concerns, physicians who discharge patients AMA enjoy no definitive legal protection from the consequences of their patients' choices.2123 In fact, good clinical judgement and careful documentation provide the best liability protection.24

Clearly, discharges AMA are problematic for patients, stressful for physicians, and resource intensive for health facilities. Therefore, efforts to understand, better manage, and ultimately decrease discharges AMA will benefit all parties. Whereas the literature on discharge AMA tends to focus on psychiatric and substance abuse patients, this review examines the professional and ethical implications of discharge AMA more generally.

Summary and Conclusions

Clinicians caring for patients who seek discharge AMA are often faced with emotionally charged and time‐pressured treatment situations. These clinicians must weigh multiple considerations for the benefit of their patients, and maintain professional standards of clinical care. Clinicians presented with these situations should (1) evaluate patients' decision‐making capacity, (2) assess the degree to which their choices are influenced by controlling external influences and mitigate these factors where possible, and (3) encourage and facilitate after‐care (Table 2).

Although discharge AMA accounts for only a small percentage of hospital discharges, its medical, emotional, and resource utilization consequences for patients as well as for physicians and hospitals is disproportionate. The clinical impacts of discharge AMA should be further investigated and specific strategies and interventions to mitigate its health effects should be validated.

The connection between idiopathic thrombocytopenic purpura (ITP) and Grave's disease is not well known in the Western hemisphere. The immunologic relationship between these 2 conditions is well reported15 but poorly defined in the literature. New‐onset hyperthyroidism in the setting of preexisting ITP can be overlooked and, if untreated, lead to worsening of the ITP, rendering it refractory to standard therapy. Early recognition and treatment of the hyperthyroid state with antithyroid medications can lead to significant improvement in the platelet count.1, 8 We report this rare but critical clinical relationship.

CASE REPORT

A 35‐year‐old Asian woman with a known history of stable ITP for 12 years (baseline platelet count of 40,000/mL) presented to her outpatient provider with a diffuse petechial rash, easy bruisability, and heavy menorrhagia for 2 weeks. Her new platelet count was 7000/mL. She was immediately started on prednisone at a dose of 1 mg/kg without any improvement in her platelet count. At the end of 4 weeks on prednisone, she developed fever, intractable nausea and vomiting, severe headache, hypotension, and tachycardia. She was subsequently hospitalized with the presumptive diagnosis of meningitis and sepsis syndrome. Her clinical syndrome was consistent with systemic inflammatory response syndrome. She was treated aggressively with intravenous fluids and a broad‐spectrum empirical antimicrobial regimen consisting of ceftriaxone, vancomycin, and acyclovir. Lumbar puncture was deferred because of her low platelet count. The sepsis workup, which included viral, fungal, and bacterial blood cultures, remained negative. Her peripheral smear did not show evidence of microangiopathic hemolytic anemia, therefore ruling out thrombotic thrombocytopenic purpura and disseminated intravascular coagulation. HIV and tuberculosis were also ruled out. After the initial sepsis workup turned out negative, she was started on solumedrol 125 mg IV every 6 hours. Over the next 2 weeks, she received an average of 4‐6 units of platelets per day and multiple blood transfusions to maintain her hemoglobin and platelet counts. The latter remained in the 1000‐5000 platelets/mL range throughout her hospitalization without any significant improvement. Her clinical course was further complicated by multiple small intracranial hemorrhages without major focal neurological deficits. A bone marrow biopsy was eventually done. It showed early dysplastic cells but no definite features of myelodysplasia and few large megakaryocytes. She received 1 dose of vincristine without response in the bone marrow after 2 weeks, and consideration was given to treatment with rituximab for refractory ITP. At that point, she informed her hematologist that 10 years ago, she had been treated for hyperthyroidism with antithyroid medications for 6 months, without further follow‐up. A thyroid panel was then ordered, and she was found to be hyperthyroid, with thyroid‐stimulating hormone (TSH) 0.01 mU/mL and free T4 of 3.1 ng/dL. She was subsequently started on propylthiouracil at 300 mg per day. Her platelet count dramatically improved and went up to the 50,000/mm³ range without further intervention over the next few months. After her discharge, an outpatient thyroid scan showed diffuse, homogeneous uptake of iodine, thereby confirming the diagnosis of Grave's disease. Retrospectively, her initial clinical syndrome of fever, hypotension, and tachycardia may have been the result of thyrotoxicosis or worsened by it.

DISCUSSION

The association between ITP and Grave's disease is poorly understood. Many hypotheses from observational data have been given in the literature. The leading theory to explain the coexistence of these 2 disorders is the presence of a common autoimmune pathway with production of 2 kinds of antibodies against platelets and TSH receptors. Indeed, autoimmune disorders tend to occur concurrently in individuals or families. Bizzaro et al. reported the coexistence of ITP and Grave's in 4 members of the same family.6 Hymes et al. found elevated levels of platelet‐bound IgG in 44% of 25 study patients with Grave's thyrotoxicosis.7 Most of these patients had easy bruising and/or bleeding, and 12% were thrombocytopenic. Panzer et al. reported the presence of antiplatelet IgG in patients with Grave's as well as improved platelet counts and increased mean platelet volume after successful antithyroid therapy.8

In addition to the coexistence of thyroid‐stimulatingimmunoglobulins (TSIs) and antiplatelet antibodies as a potential mechanism for Grave's‐associated thrombocytopenia, some have postulated that in Grave's patients, TSIs and other thyroid antibodies might actually bind to the platelets themselves. The postulated site for binding would be a truncated actin‐binding protein on the platelets that would link the high‐affinity Fc receptor of immunoglobulin G to the platelets' cytoskeleton, thereby accelerating their destruction.9

Another plausible mechanism is activation of the reticuloendothelial system by thyroid hormones, with increased clearance of platelets by the spleen in thyrotoxic states. This may explain the restoration of the platelet count when euthyroidism is reached.

Finally, thyrotoxicosis seems to alter platelet aggregation, partially by inhibition of myosin light‐chain kinase, and that also improves with restoration of euthyroidism.10

The coexistence of severe hyperthyroidism and thrombocytopenia can mimic severe sepsis in critically ill patients, and the hyperthyroid state in itself can worsen the thrombocytopenia of ITP. We suspect this patient's severe sepsis may actually have been an unrecognized severe thyrotoxicosis, with bone marrow dysfunction secondary to the hyperthyroidism, which might partially explain her lack of response to standard therapy.

CONCLUSIONS

This case underscores the importance of screening for and treating hyperthyroidism in patients with ITP, especially those resistant to steroid therapy, because the literature seems to indicate that treatment of the hyperthyroid state improves platelet count. This might help to prevent devastating clinical complications. Further research is necessary to define this empirical finding.

The connection between idiopathic thrombocytopenic purpura (ITP) and Grave's disease is not well known in the Western hemisphere. The immunologic relationship between these 2 conditions is well reported15 but poorly defined in the literature. New‐onset hyperthyroidism in the setting of preexisting ITP can be overlooked and, if untreated, lead to worsening of the ITP, rendering it refractory to standard therapy. Early recognition and treatment of the hyperthyroid state with antithyroid medications can lead to significant improvement in the platelet count.1, 8 We report this rare but critical clinical relationship.

CASE REPORT

A 35‐year‐old Asian woman with a known history of stable ITP for 12 years (baseline platelet count of 40,000/mL) presented to her outpatient provider with a diffuse petechial rash, easy bruisability, and heavy menorrhagia for 2 weeks. Her new platelet count was 7000/mL. She was immediately started on prednisone at a dose of 1 mg/kg without any improvement in her platelet count. At the end of 4 weeks on prednisone, she developed fever, intractable nausea and vomiting, severe headache, hypotension, and tachycardia. She was subsequently hospitalized with the presumptive diagnosis of meningitis and sepsis syndrome. Her clinical syndrome was consistent with systemic inflammatory response syndrome. She was treated aggressively with intravenous fluids and a broad‐spectrum empirical antimicrobial regimen consisting of ceftriaxone, vancomycin, and acyclovir. Lumbar puncture was deferred because of her low platelet count. The sepsis workup, which included viral, fungal, and bacterial blood cultures, remained negative. Her peripheral smear did not show evidence of microangiopathic hemolytic anemia, therefore ruling out thrombotic thrombocytopenic purpura and disseminated intravascular coagulation. HIV and tuberculosis were also ruled out. After the initial sepsis workup turned out negative, she was started on solumedrol 125 mg IV every 6 hours. Over the next 2 weeks, she received an average of 4‐6 units of platelets per day and multiple blood transfusions to maintain her hemoglobin and platelet counts. The latter remained in the 1000‐5000 platelets/mL range throughout her hospitalization without any significant improvement. Her clinical course was further complicated by multiple small intracranial hemorrhages without major focal neurological deficits. A bone marrow biopsy was eventually done. It showed early dysplastic cells but no definite features of myelodysplasia and few large megakaryocytes. She received 1 dose of vincristine without response in the bone marrow after 2 weeks, and consideration was given to treatment with rituximab for refractory ITP. At that point, she informed her hematologist that 10 years ago, she had been treated for hyperthyroidism with antithyroid medications for 6 months, without further follow‐up. A thyroid panel was then ordered, and she was found to be hyperthyroid, with thyroid‐stimulating hormone (TSH) 0.01 mU/mL and free T4 of 3.1 ng/dL. She was subsequently started on propylthiouracil at 300 mg per day. Her platelet count dramatically improved and went up to the 50,000/mm³ range without further intervention over the next few months. After her discharge, an outpatient thyroid scan showed diffuse, homogeneous uptake of iodine, thereby confirming the diagnosis of Grave's disease. Retrospectively, her initial clinical syndrome of fever, hypotension, and tachycardia may have been the result of thyrotoxicosis or worsened by it.

DISCUSSION

The association between ITP and Grave's disease is poorly understood. Many hypotheses from observational data have been given in the literature. The leading theory to explain the coexistence of these 2 disorders is the presence of a common autoimmune pathway with production of 2 kinds of antibodies against platelets and TSH receptors. Indeed, autoimmune disorders tend to occur concurrently in individuals or families. Bizzaro et al. reported the coexistence of ITP and Grave's in 4 members of the same family.6 Hymes et al. found elevated levels of platelet‐bound IgG in 44% of 25 study patients with Grave's thyrotoxicosis.7 Most of these patients had easy bruising and/or bleeding, and 12% were thrombocytopenic. Panzer et al. reported the presence of antiplatelet IgG in patients with Grave's as well as improved platelet counts and increased mean platelet volume after successful antithyroid therapy.8

In addition to the coexistence of thyroid‐stimulatingimmunoglobulins (TSIs) and antiplatelet antibodies as a potential mechanism for Grave's‐associated thrombocytopenia, some have postulated that in Grave's patients, TSIs and other thyroid antibodies might actually bind to the platelets themselves. The postulated site for binding would be a truncated actin‐binding protein on the platelets that would link the high‐affinity Fc receptor of immunoglobulin G to the platelets' cytoskeleton, thereby accelerating their destruction.9

Another plausible mechanism is activation of the reticuloendothelial system by thyroid hormones, with increased clearance of platelets by the spleen in thyrotoxic states. This may explain the restoration of the platelet count when euthyroidism is reached.

Finally, thyrotoxicosis seems to alter platelet aggregation, partially by inhibition of myosin light‐chain kinase, and that also improves with restoration of euthyroidism.10

The coexistence of severe hyperthyroidism and thrombocytopenia can mimic severe sepsis in critically ill patients, and the hyperthyroid state in itself can worsen the thrombocytopenia of ITP. We suspect this patient's severe sepsis may actually have been an unrecognized severe thyrotoxicosis, with bone marrow dysfunction secondary to the hyperthyroidism, which might partially explain her lack of response to standard therapy.

CONCLUSIONS

This case underscores the importance of screening for and treating hyperthyroidism in patients with ITP, especially those resistant to steroid therapy, because the literature seems to indicate that treatment of the hyperthyroid state improves platelet count. This might help to prevent devastating clinical complications. Further research is necessary to define this empirical finding.

The connection between idiopathic thrombocytopenic purpura (ITP) and Grave's disease is not well known in the Western hemisphere. The immunologic relationship between these 2 conditions is well reported15 but poorly defined in the literature. New‐onset hyperthyroidism in the setting of preexisting ITP can be overlooked and, if untreated, lead to worsening of the ITP, rendering it refractory to standard therapy. Early recognition and treatment of the hyperthyroid state with antithyroid medications can lead to significant improvement in the platelet count.1, 8 We report this rare but critical clinical relationship.

CASE REPORT

A 35‐year‐old Asian woman with a known history of stable ITP for 12 years (baseline platelet count of 40,000/mL) presented to her outpatient provider with a diffuse petechial rash, easy bruisability, and heavy menorrhagia for 2 weeks. Her new platelet count was 7000/mL. She was immediately started on prednisone at a dose of 1 mg/kg without any improvement in her platelet count. At the end of 4 weeks on prednisone, she developed fever, intractable nausea and vomiting, severe headache, hypotension, and tachycardia. She was subsequently hospitalized with the presumptive diagnosis of meningitis and sepsis syndrome. Her clinical syndrome was consistent with systemic inflammatory response syndrome. She was treated aggressively with intravenous fluids and a broad‐spectrum empirical antimicrobial regimen consisting of ceftriaxone, vancomycin, and acyclovir. Lumbar puncture was deferred because of her low platelet count. The sepsis workup, which included viral, fungal, and bacterial blood cultures, remained negative. Her peripheral smear did not show evidence of microangiopathic hemolytic anemia, therefore ruling out thrombotic thrombocytopenic purpura and disseminated intravascular coagulation. HIV and tuberculosis were also ruled out. After the initial sepsis workup turned out negative, she was started on solumedrol 125 mg IV every 6 hours. Over the next 2 weeks, she received an average of 4‐6 units of platelets per day and multiple blood transfusions to maintain her hemoglobin and platelet counts. The latter remained in the 1000‐5000 platelets/mL range throughout her hospitalization without any significant improvement. Her clinical course was further complicated by multiple small intracranial hemorrhages without major focal neurological deficits. A bone marrow biopsy was eventually done. It showed early dysplastic cells but no definite features of myelodysplasia and few large megakaryocytes. She received 1 dose of vincristine without response in the bone marrow after 2 weeks, and consideration was given to treatment with rituximab for refractory ITP. At that point, she informed her hematologist that 10 years ago, she had been treated for hyperthyroidism with antithyroid medications for 6 months, without further follow‐up. A thyroid panel was then ordered, and she was found to be hyperthyroid, with thyroid‐stimulating hormone (TSH) 0.01 mU/mL and free T4 of 3.1 ng/dL. She was subsequently started on propylthiouracil at 300 mg per day. Her platelet count dramatically improved and went up to the 50,000/mm³ range without further intervention over the next few months. After her discharge, an outpatient thyroid scan showed diffuse, homogeneous uptake of iodine, thereby confirming the diagnosis of Grave's disease. Retrospectively, her initial clinical syndrome of fever, hypotension, and tachycardia may have been the result of thyrotoxicosis or worsened by it.

DISCUSSION

The association between ITP and Grave's disease is poorly understood. Many hypotheses from observational data have been given in the literature. The leading theory to explain the coexistence of these 2 disorders is the presence of a common autoimmune pathway with production of 2 kinds of antibodies against platelets and TSH receptors. Indeed, autoimmune disorders tend to occur concurrently in individuals or families. Bizzaro et al. reported the coexistence of ITP and Grave's in 4 members of the same family.6 Hymes et al. found elevated levels of platelet‐bound IgG in 44% of 25 study patients with Grave's thyrotoxicosis.7 Most of these patients had easy bruising and/or bleeding, and 12% were thrombocytopenic. Panzer et al. reported the presence of antiplatelet IgG in patients with Grave's as well as improved platelet counts and increased mean platelet volume after successful antithyroid therapy.8

In addition to the coexistence of thyroid‐stimulatingimmunoglobulins (TSIs) and antiplatelet antibodies as a potential mechanism for Grave's‐associated thrombocytopenia, some have postulated that in Grave's patients, TSIs and other thyroid antibodies might actually bind to the platelets themselves. The postulated site for binding would be a truncated actin‐binding protein on the platelets that would link the high‐affinity Fc receptor of immunoglobulin G to the platelets' cytoskeleton, thereby accelerating their destruction.9

Another plausible mechanism is activation of the reticuloendothelial system by thyroid hormones, with increased clearance of platelets by the spleen in thyrotoxic states. This may explain the restoration of the platelet count when euthyroidism is reached.

Finally, thyrotoxicosis seems to alter platelet aggregation, partially by inhibition of myosin light‐chain kinase, and that also improves with restoration of euthyroidism.10

The coexistence of severe hyperthyroidism and thrombocytopenia can mimic severe sepsis in critically ill patients, and the hyperthyroid state in itself can worsen the thrombocytopenia of ITP. We suspect this patient's severe sepsis may actually have been an unrecognized severe thyrotoxicosis, with bone marrow dysfunction secondary to the hyperthyroidism, which might partially explain her lack of response to standard therapy.

CONCLUSIONS

This case underscores the importance of screening for and treating hyperthyroidism in patients with ITP, especially those resistant to steroid therapy, because the literature seems to indicate that treatment of the hyperthyroid state improves platelet count. This might help to prevent devastating clinical complications. Further research is necessary to define this empirical finding.