Venous thromboembolism (VTE), including deep venous thrombosis (DVT) and pulmonary embolism (PE), is estimated to affect 900,000 Americans each year and is a cause of significant morbidity and mortality with associated high healthcare costs.[1] Accordingly, the comparative effectiveness and safety of interventions for the prevention and treatment of VTE are among the national priorities for comparative effectiveness research.[2] Whereas we have evidence‐based guidelines for the prophylaxis of VTE in the general population, there are no guidelines informing the care of select patient populations. Select populations are those patients in whom there is decisional uncertainty about the optimal choice, timing, and dose of VTE prophylaxis. Not only do these patients have an increased risk of DVT and PE, but most are also at high risk of bleeding, the most important complication of VTE prophylaxis.[3, 4, 5, 6]

The objectives of this systematic review were to define the comparative effectiveness and safety of pharmacologic and mechanical strategies for VTE prevention in some of these select medical populations including obese patients, patients on concomitant antiplatelet therapy, patients with renal insufficiency, patients who are underweight, and patients with coagulopathy due to liver disease.

METHODS

The methods for this comparative effectiveness review (CER) follow the guidelines suggested in the Agency for Healthcare Research and Quality (AHRQ) Methods Guide for Effectiveness and Comparative Effectiveness Reviews.[7] The protocol was publically posted.[8]

RESULTS

We identified 30,902 unique citations and included 9 studies (Figure 1). There were 5 RCTs with relevant subgroups and 4 observational studies (Table 1). Two studies reported on the risk of bleeding in patients given pharmacologic prophylaxis while they are concomitantly taking nonsteroidal anti‐inflammatory drugs (NSAIDS) or antiplatelet agents/aspirin, 1 RCT and 1 prospective observational study reported on obese patients, and 5 studies described outcomes of patients with renal insufficiency (see Supporting Information, Table 1, in the online version of this article). No study tested prophylaxis in underweight patients or those with liver disease.

Figure 1

DISCUSSION

We found that the evidence was very limited about prevention of VTE in these select and yet prevalent patient populations. Despite the fact that there is an increasing number of obese patients and patients who are on antiplatelet therapies, most clinical practice guidelines do not address the care of these populations, which may be entirely appropriate given the state of the evidence.

The ACCP practice guidelines[21] suggest using a higher dose of enoxaparin for the prevention of VTE in obese patients. The subgroup analysis by Kucher et al.[11] showed effect attenuation of dalteparin when given at a fixed dose of 5000 IU/mL to patients with a BMI of >40 kg/m². The Freeman study[12] showed that extremely obese patients (average BMI >62.1 kg/m²) who are given a fixed dose of enoxaparin achieved target anti‐factor Xa levels significantly less often than those who received a higher dose of enoxaparin. The 2 separate findings, although not conclusive, lend some credence to the current ACCP guidelines.[21]

The studies we reviewed on VTE prophylaxis in patients who are concomitantly on antiplatelets including aspirin reported no major increased risk of bleeding; however, in the Friedman et al. study,[15] 3.0% of patients who were put on enoxaparin while still on aspirin had a bleeding event compared to 1.2% of those on enoxaparin alone. This difference is not statistically significant but is a trend possibly worth noting, especially when one looks at the lower RR of bleeding at 0.55 compared to 0.82 when dabigatran is compared with enoxaparin with and without concomitant aspirin therapy, respectively (Table 1). The highest dose of aspirin used in either of the studies was 160 mg/day, and neither study addressed other potent antiplatelets such as clopidogrel or ticlopidine separately, which limits the generalizability of the finding to all antiplatelets. Current ACCP guidelines do not recommend aspirin as a sole option for the prevention of VTE in orthopedic surgery patients.[22] Concerns remain among clinicians that antiplatelets, including aspirin, on their own are unlikely to be fully effective to thwart venous thrombotic processes for most patients, and yet the risk of bleeding is not fully known when these agents are combined with other anticoagulants for VTE prophylaxis.

Our review has several limitations, including the possibility that we may have missed some observational studies, as the identification of relevant observational studies in electronic searches is more challenging than that of RCTs. The few studies made it impossible to quantitatively pool results. These results, however, have important implications, namely that additional research on the comparative effectiveness and safety of pharmacologic and mechanical strategies to prevent VTE is needed for the optimal care of these patient subgroups. This might be achieved with trials dedicated to enrolling these patients or prespecified subgroup analyses within larger trials. Observational data may be appropriate as long as attention is paid to confounding.

Venous thromboembolism (VTE), including deep venous thrombosis (DVT) and pulmonary embolism (PE), is estimated to affect 900,000 Americans each year and is a cause of significant morbidity and mortality with associated high healthcare costs.[1] Accordingly, the comparative effectiveness and safety of interventions for the prevention and treatment of VTE are among the national priorities for comparative effectiveness research.[2] Whereas we have evidence‐based guidelines for the prophylaxis of VTE in the general population, there are no guidelines informing the care of select patient populations. Select populations are those patients in whom there is decisional uncertainty about the optimal choice, timing, and dose of VTE prophylaxis. Not only do these patients have an increased risk of DVT and PE, but most are also at high risk of bleeding, the most important complication of VTE prophylaxis.[3, 4, 5, 6]

The objectives of this systematic review were to define the comparative effectiveness and safety of pharmacologic and mechanical strategies for VTE prevention in some of these select medical populations including obese patients, patients on concomitant antiplatelet therapy, patients with renal insufficiency, patients who are underweight, and patients with coagulopathy due to liver disease.

METHODS

The methods for this comparative effectiveness review (CER) follow the guidelines suggested in the Agency for Healthcare Research and Quality (AHRQ) Methods Guide for Effectiveness and Comparative Effectiveness Reviews.[7] The protocol was publically posted.[8]

RESULTS

We identified 30,902 unique citations and included 9 studies (Figure 1). There were 5 RCTs with relevant subgroups and 4 observational studies (Table 1). Two studies reported on the risk of bleeding in patients given pharmacologic prophylaxis while they are concomitantly taking nonsteroidal anti‐inflammatory drugs (NSAIDS) or antiplatelet agents/aspirin, 1 RCT and 1 prospective observational study reported on obese patients, and 5 studies described outcomes of patients with renal insufficiency (see Supporting Information, Table 1, in the online version of this article). No study tested prophylaxis in underweight patients or those with liver disease.

Figure 1

DISCUSSION

We found that the evidence was very limited about prevention of VTE in these select and yet prevalent patient populations. Despite the fact that there is an increasing number of obese patients and patients who are on antiplatelet therapies, most clinical practice guidelines do not address the care of these populations, which may be entirely appropriate given the state of the evidence.

The ACCP practice guidelines[21] suggest using a higher dose of enoxaparin for the prevention of VTE in obese patients. The subgroup analysis by Kucher et al.[11] showed effect attenuation of dalteparin when given at a fixed dose of 5000 IU/mL to patients with a BMI of >40 kg/m². The Freeman study[12] showed that extremely obese patients (average BMI >62.1 kg/m²) who are given a fixed dose of enoxaparin achieved target anti‐factor Xa levels significantly less often than those who received a higher dose of enoxaparin. The 2 separate findings, although not conclusive, lend some credence to the current ACCP guidelines.[21]

The studies we reviewed on VTE prophylaxis in patients who are concomitantly on antiplatelets including aspirin reported no major increased risk of bleeding; however, in the Friedman et al. study,[15] 3.0% of patients who were put on enoxaparin while still on aspirin had a bleeding event compared to 1.2% of those on enoxaparin alone. This difference is not statistically significant but is a trend possibly worth noting, especially when one looks at the lower RR of bleeding at 0.55 compared to 0.82 when dabigatran is compared with enoxaparin with and without concomitant aspirin therapy, respectively (Table 1). The highest dose of aspirin used in either of the studies was 160 mg/day, and neither study addressed other potent antiplatelets such as clopidogrel or ticlopidine separately, which limits the generalizability of the finding to all antiplatelets. Current ACCP guidelines do not recommend aspirin as a sole option for the prevention of VTE in orthopedic surgery patients.[22] Concerns remain among clinicians that antiplatelets, including aspirin, on their own are unlikely to be fully effective to thwart venous thrombotic processes for most patients, and yet the risk of bleeding is not fully known when these agents are combined with other anticoagulants for VTE prophylaxis.

Our review has several limitations, including the possibility that we may have missed some observational studies, as the identification of relevant observational studies in electronic searches is more challenging than that of RCTs. The few studies made it impossible to quantitatively pool results. These results, however, have important implications, namely that additional research on the comparative effectiveness and safety of pharmacologic and mechanical strategies to prevent VTE is needed for the optimal care of these patient subgroups. This might be achieved with trials dedicated to enrolling these patients or prespecified subgroup analyses within larger trials. Observational data may be appropriate as long as attention is paid to confounding.

Venous thromboembolism (VTE), including deep venous thrombosis (DVT) and pulmonary embolism (PE), is estimated to affect 900,000 Americans each year and is a cause of significant morbidity and mortality with associated high healthcare costs.[1] Accordingly, the comparative effectiveness and safety of interventions for the prevention and treatment of VTE are among the national priorities for comparative effectiveness research.[2] Whereas we have evidence‐based guidelines for the prophylaxis of VTE in the general population, there are no guidelines informing the care of select patient populations. Select populations are those patients in whom there is decisional uncertainty about the optimal choice, timing, and dose of VTE prophylaxis. Not only do these patients have an increased risk of DVT and PE, but most are also at high risk of bleeding, the most important complication of VTE prophylaxis.[3, 4, 5, 6]

The objectives of this systematic review were to define the comparative effectiveness and safety of pharmacologic and mechanical strategies for VTE prevention in some of these select medical populations including obese patients, patients on concomitant antiplatelet therapy, patients with renal insufficiency, patients who are underweight, and patients with coagulopathy due to liver disease.

METHODS

The methods for this comparative effectiveness review (CER) follow the guidelines suggested in the Agency for Healthcare Research and Quality (AHRQ) Methods Guide for Effectiveness and Comparative Effectiveness Reviews.[7] The protocol was publically posted.[8]

RESULTS

We identified 30,902 unique citations and included 9 studies (Figure 1). There were 5 RCTs with relevant subgroups and 4 observational studies (Table 1). Two studies reported on the risk of bleeding in patients given pharmacologic prophylaxis while they are concomitantly taking nonsteroidal anti‐inflammatory drugs (NSAIDS) or antiplatelet agents/aspirin, 1 RCT and 1 prospective observational study reported on obese patients, and 5 studies described outcomes of patients with renal insufficiency (see Supporting Information, Table 1, in the online version of this article). No study tested prophylaxis in underweight patients or those with liver disease.

Figure 1

DISCUSSION

We found that the evidence was very limited about prevention of VTE in these select and yet prevalent patient populations. Despite the fact that there is an increasing number of obese patients and patients who are on antiplatelet therapies, most clinical practice guidelines do not address the care of these populations, which may be entirely appropriate given the state of the evidence.

The ACCP practice guidelines[21] suggest using a higher dose of enoxaparin for the prevention of VTE in obese patients. The subgroup analysis by Kucher et al.[11] showed effect attenuation of dalteparin when given at a fixed dose of 5000 IU/mL to patients with a BMI of >40 kg/m². The Freeman study[12] showed that extremely obese patients (average BMI >62.1 kg/m²) who are given a fixed dose of enoxaparin achieved target anti‐factor Xa levels significantly less often than those who received a higher dose of enoxaparin. The 2 separate findings, although not conclusive, lend some credence to the current ACCP guidelines.[21]

The studies we reviewed on VTE prophylaxis in patients who are concomitantly on antiplatelets including aspirin reported no major increased risk of bleeding; however, in the Friedman et al. study,[15] 3.0% of patients who were put on enoxaparin while still on aspirin had a bleeding event compared to 1.2% of those on enoxaparin alone. This difference is not statistically significant but is a trend possibly worth noting, especially when one looks at the lower RR of bleeding at 0.55 compared to 0.82 when dabigatran is compared with enoxaparin with and without concomitant aspirin therapy, respectively (Table 1). The highest dose of aspirin used in either of the studies was 160 mg/day, and neither study addressed other potent antiplatelets such as clopidogrel or ticlopidine separately, which limits the generalizability of the finding to all antiplatelets. Current ACCP guidelines do not recommend aspirin as a sole option for the prevention of VTE in orthopedic surgery patients.[22] Concerns remain among clinicians that antiplatelets, including aspirin, on their own are unlikely to be fully effective to thwart venous thrombotic processes for most patients, and yet the risk of bleeding is not fully known when these agents are combined with other anticoagulants for VTE prophylaxis.

Our review has several limitations, including the possibility that we may have missed some observational studies, as the identification of relevant observational studies in electronic searches is more challenging than that of RCTs. The few studies made it impossible to quantitatively pool results. These results, however, have important implications, namely that additional research on the comparative effectiveness and safety of pharmacologic and mechanical strategies to prevent VTE is needed for the optimal care of these patient subgroups. This might be achieved with trials dedicated to enrolling these patients or prespecified subgroup analyses within larger trials. Observational data may be appropriate as long as attention is paid to confounding.

Healthcare workers' (HCWs) attire becomes contaminated with bacterial pathogens during the course of the workday,[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] and Munoz‐Price et al.[13] recently demonstrated that finding bacterial pathogens on HCWs' white coats correlated with finding the same pathogens on their hands. Because of concern for an association between attire colonization and nosocomial infection, governmental agencies in England and Scotland banned HCWs from wearing white coats or long‐sleeve garments,[14, 15] despite evidence that such an approach does not reduce contamination.[12]

Newly developed antimicrobial textiles have been incorporated into HCW scrubs,[16, 17, 18, 19, 20] and commercial Web sites and product inserts report that these products can reduce bacterial contamination by 80.9% at 8 hours to greater than 99% under laboratory conditions depending on the product and microbe studied.[16, 17, 19] Because there are limited clinical data pertaining to the effectiveness of antimicrobial scrubs, we performed a prospective study designed to determine whether wearing these products reduced bacterial contamination of HCWs' scrubs or skin at the end of an 8‐hour workday.

METHODS

RESULTS

We screened 118 HCWs for participation and randomized 109, 37 in the control and antimicrobial scrub group A, and 35 in antimicrobial scrub group B (during the course of the study we neglected to culture the pockets of 2 participants in the standard scrub group and 2 in antimicrobial scrub group A). Because our primary end point was total colony count from cultures taken from 3 sites, data from these 4 subjects could not be used, and all the data from these 4 subjects were excluded from the primary analysis; 4 additional subjects were subsequently recruited allowing us to meet our block enrollment target (Figure 1). The first and last participants were studied on March 12, 2012 and August 28, 2012, respectively. The trial ended once the defined number of participants was enrolled. The occupations of the 105 participants are summarized in Table 1.

Figure 1

Colony counts of all scrubs cultured prior to use never exceeded 10 colonies. The median (IQR) total colony counts from all sites on the scrubs was 99 (66182) for standard scrubs, 137 (84289) for antimicrobial scrub type A, and 138 (62274) for antimicrobial scrub type B (P=0.36). We found no significant differences between the colony counts cultured from any of the individual sites among the 3 groups, regardless of occupation (Table 2). No significant difference was observed with respect to colony counts cultured from the wrist among the 3 study groups (Table 2). Comparisons between groups were planned a priori if a difference across all groups was found. Given the nonsignificant P values across all scrub groups, no further comparisons were made.

Fifty‐five participants (52%) reported caring for patients who were known to be colonized or infected with an antibiotic‐resistant organism, 16 (46%) randomized to wear standard scrubs, and 20 (57%) and 19 (54%) randomized to wear antimicrobial scrub A or B, respectively (P=0.61). Of these, however, antibiotic‐resistant organisms were only cultured from the scrubs of 2 providers (1 with 1 colony of MRSA from the breast pocket of antimicrobial scrub A, 1 with 1 colony of MRSA cultured from the pocket of antimicrobial scrub B [P=0.55]), and from the wrist of only 1 provider (a multiresistant Gram‐negative rod who wore antimicrobial scrub B).

DISCUSSION

The important findings of this study are that we found no evidence indicating that either of the 2 antimicrobial scrubs tested reduced bacterial contamination or antibiotic‐resistant contamination on HCWs' scrubs or wrists compared with standard scrubs at the end of an 8‐hour workday, and that despite many HCWs being exposed to patients who were colonized or infected with antibiotic‐resistant bacteria, these organisms were only rarely cultured from their uniforms.

We found that HCWs in all 3 arms of the study had bacterial contamination on their scrubs and skin, consistent with previous studies showing that HCWs' uniforms are frequently contaminated with bacteria, including MRSA, VRE, and other pathogens.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] We previously found that bacterial contamination of HCWs' uniforms occurs within hours of putting on newly laundered uniforms.[12]

Literature on the effectiveness of antimicrobial HCW uniforms when tested in clinical settings is limited. Bearman and colleagues[23] recently published the results of a study of 31 subjects who wore either standard or antimicrobial scrubs, crossing over every 4 weeks for 4 months, with random culturing done weekly at the beginning and end of a work shift. Scrubs were laundered an average of 1.5 times/week, but the timing of the laundering relative to when cultures were obtained was not reported. Very few isolates of MRSA, Gram‐negative rods, or VRE were found (only 3.9%, 0.4%, and 0.05% of the 2000 samples obtained, respectively), and no differences were observed with respect to the number of HCWs who had antibiotic‐resistant organisms cultured when they were wearing standard versus antimicrobial scrubs. Those who had MRSA cultured, however, had lower mean log colony counts when they were wearing the antimicrobial scrubs. The small number of samples with positive isolates, together with differences in the extent of before‐shift contamination among groups complicates interpreting these data. The authors concluded that a prospective trial was needed. We attempted to include the scrub studied by Bearman and colleagues[23] in our study, but the company had insufficient stock available at the time we tried to purchase the product.

Gross and colleagues[24] found no difference in the mean colony counts of cultures taken from silver‐impregnated versus standard scrubs in a pilot crossover study done with 10 HCWs (although there were trends toward higher colony counts when the subjects wore antimicrobial scrubs).

Antibiotic‐resistant bacteria were only cultured from 3 participants (2.9%) in our current study, compared to 16% of those randomized to wearing white coats in our previous study and 20% of those randomized to wearing standard scrubs.[12] This difference may be explained by several recent studies reporting that rates of MRSA infections in hospitals are decreasing.[25, 26] The rate of hospital‐acquired MRSA infection or colonization at our own institution decreased 80% from 2007 to 2012. At the times of our previous and current studies, providers were expected to wear gowns and gloves when caring for patients as per standard contact precautions. Rates of infection and colonization of VRE and resistant Gram‐negative rods have remained low at our hospital, and our data are consistent with the rates reported on HCWs' uniforms in other studies.[2, 5, 10]

Only 6 of our subjects reported adverse reactions, but all were wearing antimicrobial scrubs (P=0.18). Several of the participants described that the fabrics of the 2 antimicrobial scrubs were heavier and less breathable than the standard scrubs. We believe this difference is more likely to explain the adverse reactions reported than is any type of reaction to the specific chemicals in the fabrics.

Our study has several limitations. Because it was conducted on the general internal medicine units of a single university‐affiliated public hospital, the results may not generalize to other types of institutions or other inpatient services.

As we previously described,[12] the RODAC imprint method only samples a small area of HCWs' uniforms and thus does not represent total bacterial contamination.[21] We specifically cultured areas that are known to be highly contaminated (ie, sleeve cuffs and pockets). Although imprint methods have limitations (as do other methods for culturing clothing), they have been commonly utilized in studies assessing bacterial contamination of HCW clothing.[2, 3, 5]

Although some of the bacterial load we cultured could have come from the providers themselves, previous studies have shown that 80% to 90% of the resistant bacteria cultured from HCWs' attire come from other sources.[1, 2]

Because our sample size was calculated on the basis of being able to detect a difference of 70% in total bacterial colony count, our study was not large enough to exclude a lower level of effectiveness. However, we saw no trends suggesting the antimicrobial products might have a lower level of effectiveness.

We did not observe the hand‐washing practices of the participants, and accordingly, cannot confirm that these practices were the same in each of our 3 study groups. Intermittent, surreptitious monitoring of hand‐washing practices on our internal medicine units over the last several years has found compliance with hand hygiene recommendations varying from 70% to 90%.

Although the participants in our study were not explicitly told to which scrub they were randomized, the colors, appearances, and textures of the antimicrobial fabrics were different from the standard scrubs such that blinding was impossible. Participants wearing antimicrobial scrubs could have changed their hand hygiene practices (ie, less careful hand hygiene). Lack of blinding could also have led to over‐reporting of adverse events by the subjects randomized to wear the antimicrobial scrubs.

In an effort to treat all the scrubs in the same fashion, all were tested new, prior to being washed or previously worn. Studying the scrubs prior to washing or wearing could have increased the reports of adverse effects, as the fabrics could have been stiffer and more uncomfortable than they might have been at a later stage in their use.

Our study also has some strengths. Our participants included physicians, residents, nurses, nurse practitioners, and physician assistants. Accordingly, our results should be generalizable to most HCWs. We also confirmed that the scrubs that were tested were nearly sterile prior to use.

In conclusion, we found no evidence suggesting that either of 2 antimicrobial scrubs tested decreased bacterial contamination of HCWs' scrubs or skin after an 8‐hour workday compared to standard scrubs. We also found that, although HCWs are frequently exposed to patients harboring antibiotic‐resistant bacteria, these bacteria were only rarely cultured from HCWs' scrubs or skin.

Healthcare workers' (HCWs) attire becomes contaminated with bacterial pathogens during the course of the workday,[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] and Munoz‐Price et al.[13] recently demonstrated that finding bacterial pathogens on HCWs' white coats correlated with finding the same pathogens on their hands. Because of concern for an association between attire colonization and nosocomial infection, governmental agencies in England and Scotland banned HCWs from wearing white coats or long‐sleeve garments,[14, 15] despite evidence that such an approach does not reduce contamination.[12]

Newly developed antimicrobial textiles have been incorporated into HCW scrubs,[16, 17, 18, 19, 20] and commercial Web sites and product inserts report that these products can reduce bacterial contamination by 80.9% at 8 hours to greater than 99% under laboratory conditions depending on the product and microbe studied.[16, 17, 19] Because there are limited clinical data pertaining to the effectiveness of antimicrobial scrubs, we performed a prospective study designed to determine whether wearing these products reduced bacterial contamination of HCWs' scrubs or skin at the end of an 8‐hour workday.

METHODS

RESULTS

We screened 118 HCWs for participation and randomized 109, 37 in the control and antimicrobial scrub group A, and 35 in antimicrobial scrub group B (during the course of the study we neglected to culture the pockets of 2 participants in the standard scrub group and 2 in antimicrobial scrub group A). Because our primary end point was total colony count from cultures taken from 3 sites, data from these 4 subjects could not be used, and all the data from these 4 subjects were excluded from the primary analysis; 4 additional subjects were subsequently recruited allowing us to meet our block enrollment target (Figure 1). The first and last participants were studied on March 12, 2012 and August 28, 2012, respectively. The trial ended once the defined number of participants was enrolled. The occupations of the 105 participants are summarized in Table 1.

Figure 1

Colony counts of all scrubs cultured prior to use never exceeded 10 colonies. The median (IQR) total colony counts from all sites on the scrubs was 99 (66182) for standard scrubs, 137 (84289) for antimicrobial scrub type A, and 138 (62274) for antimicrobial scrub type B (P=0.36). We found no significant differences between the colony counts cultured from any of the individual sites among the 3 groups, regardless of occupation (Table 2). No significant difference was observed with respect to colony counts cultured from the wrist among the 3 study groups (Table 2). Comparisons between groups were planned a priori if a difference across all groups was found. Given the nonsignificant P values across all scrub groups, no further comparisons were made.

Fifty‐five participants (52%) reported caring for patients who were known to be colonized or infected with an antibiotic‐resistant organism, 16 (46%) randomized to wear standard scrubs, and 20 (57%) and 19 (54%) randomized to wear antimicrobial scrub A or B, respectively (P=0.61). Of these, however, antibiotic‐resistant organisms were only cultured from the scrubs of 2 providers (1 with 1 colony of MRSA from the breast pocket of antimicrobial scrub A, 1 with 1 colony of MRSA cultured from the pocket of antimicrobial scrub B [P=0.55]), and from the wrist of only 1 provider (a multiresistant Gram‐negative rod who wore antimicrobial scrub B).

DISCUSSION

The important findings of this study are that we found no evidence indicating that either of the 2 antimicrobial scrubs tested reduced bacterial contamination or antibiotic‐resistant contamination on HCWs' scrubs or wrists compared with standard scrubs at the end of an 8‐hour workday, and that despite many HCWs being exposed to patients who were colonized or infected with antibiotic‐resistant bacteria, these organisms were only rarely cultured from their uniforms.

We found that HCWs in all 3 arms of the study had bacterial contamination on their scrubs and skin, consistent with previous studies showing that HCWs' uniforms are frequently contaminated with bacteria, including MRSA, VRE, and other pathogens.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] We previously found that bacterial contamination of HCWs' uniforms occurs within hours of putting on newly laundered uniforms.[12]

Literature on the effectiveness of antimicrobial HCW uniforms when tested in clinical settings is limited. Bearman and colleagues[23] recently published the results of a study of 31 subjects who wore either standard or antimicrobial scrubs, crossing over every 4 weeks for 4 months, with random culturing done weekly at the beginning and end of a work shift. Scrubs were laundered an average of 1.5 times/week, but the timing of the laundering relative to when cultures were obtained was not reported. Very few isolates of MRSA, Gram‐negative rods, or VRE were found (only 3.9%, 0.4%, and 0.05% of the 2000 samples obtained, respectively), and no differences were observed with respect to the number of HCWs who had antibiotic‐resistant organisms cultured when they were wearing standard versus antimicrobial scrubs. Those who had MRSA cultured, however, had lower mean log colony counts when they were wearing the antimicrobial scrubs. The small number of samples with positive isolates, together with differences in the extent of before‐shift contamination among groups complicates interpreting these data. The authors concluded that a prospective trial was needed. We attempted to include the scrub studied by Bearman and colleagues[23] in our study, but the company had insufficient stock available at the time we tried to purchase the product.

Gross and colleagues[24] found no difference in the mean colony counts of cultures taken from silver‐impregnated versus standard scrubs in a pilot crossover study done with 10 HCWs (although there were trends toward higher colony counts when the subjects wore antimicrobial scrubs).

Antibiotic‐resistant bacteria were only cultured from 3 participants (2.9%) in our current study, compared to 16% of those randomized to wearing white coats in our previous study and 20% of those randomized to wearing standard scrubs.[12] This difference may be explained by several recent studies reporting that rates of MRSA infections in hospitals are decreasing.[25, 26] The rate of hospital‐acquired MRSA infection or colonization at our own institution decreased 80% from 2007 to 2012. At the times of our previous and current studies, providers were expected to wear gowns and gloves when caring for patients as per standard contact precautions. Rates of infection and colonization of VRE and resistant Gram‐negative rods have remained low at our hospital, and our data are consistent with the rates reported on HCWs' uniforms in other studies.[2, 5, 10]

Only 6 of our subjects reported adverse reactions, but all were wearing antimicrobial scrubs (P=0.18). Several of the participants described that the fabrics of the 2 antimicrobial scrubs were heavier and less breathable than the standard scrubs. We believe this difference is more likely to explain the adverse reactions reported than is any type of reaction to the specific chemicals in the fabrics.

Our study has several limitations. Because it was conducted on the general internal medicine units of a single university‐affiliated public hospital, the results may not generalize to other types of institutions or other inpatient services.

As we previously described,[12] the RODAC imprint method only samples a small area of HCWs' uniforms and thus does not represent total bacterial contamination.[21] We specifically cultured areas that are known to be highly contaminated (ie, sleeve cuffs and pockets). Although imprint methods have limitations (as do other methods for culturing clothing), they have been commonly utilized in studies assessing bacterial contamination of HCW clothing.[2, 3, 5]

Although some of the bacterial load we cultured could have come from the providers themselves, previous studies have shown that 80% to 90% of the resistant bacteria cultured from HCWs' attire come from other sources.[1, 2]

Because our sample size was calculated on the basis of being able to detect a difference of 70% in total bacterial colony count, our study was not large enough to exclude a lower level of effectiveness. However, we saw no trends suggesting the antimicrobial products might have a lower level of effectiveness.

We did not observe the hand‐washing practices of the participants, and accordingly, cannot confirm that these practices were the same in each of our 3 study groups. Intermittent, surreptitious monitoring of hand‐washing practices on our internal medicine units over the last several years has found compliance with hand hygiene recommendations varying from 70% to 90%.

Although the participants in our study were not explicitly told to which scrub they were randomized, the colors, appearances, and textures of the antimicrobial fabrics were different from the standard scrubs such that blinding was impossible. Participants wearing antimicrobial scrubs could have changed their hand hygiene practices (ie, less careful hand hygiene). Lack of blinding could also have led to over‐reporting of adverse events by the subjects randomized to wear the antimicrobial scrubs.

In an effort to treat all the scrubs in the same fashion, all were tested new, prior to being washed or previously worn. Studying the scrubs prior to washing or wearing could have increased the reports of adverse effects, as the fabrics could have been stiffer and more uncomfortable than they might have been at a later stage in their use.

Our study also has some strengths. Our participants included physicians, residents, nurses, nurse practitioners, and physician assistants. Accordingly, our results should be generalizable to most HCWs. We also confirmed that the scrubs that were tested were nearly sterile prior to use.

In conclusion, we found no evidence suggesting that either of 2 antimicrobial scrubs tested decreased bacterial contamination of HCWs' scrubs or skin after an 8‐hour workday compared to standard scrubs. We also found that, although HCWs are frequently exposed to patients harboring antibiotic‐resistant bacteria, these bacteria were only rarely cultured from HCWs' scrubs or skin.

Healthcare workers' (HCWs) attire becomes contaminated with bacterial pathogens during the course of the workday,[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] and Munoz‐Price et al.[13] recently demonstrated that finding bacterial pathogens on HCWs' white coats correlated with finding the same pathogens on their hands. Because of concern for an association between attire colonization and nosocomial infection, governmental agencies in England and Scotland banned HCWs from wearing white coats or long‐sleeve garments,[14, 15] despite evidence that such an approach does not reduce contamination.[12]

Newly developed antimicrobial textiles have been incorporated into HCW scrubs,[16, 17, 18, 19, 20] and commercial Web sites and product inserts report that these products can reduce bacterial contamination by 80.9% at 8 hours to greater than 99% under laboratory conditions depending on the product and microbe studied.[16, 17, 19] Because there are limited clinical data pertaining to the effectiveness of antimicrobial scrubs, we performed a prospective study designed to determine whether wearing these products reduced bacterial contamination of HCWs' scrubs or skin at the end of an 8‐hour workday.

METHODS

RESULTS

We screened 118 HCWs for participation and randomized 109, 37 in the control and antimicrobial scrub group A, and 35 in antimicrobial scrub group B (during the course of the study we neglected to culture the pockets of 2 participants in the standard scrub group and 2 in antimicrobial scrub group A). Because our primary end point was total colony count from cultures taken from 3 sites, data from these 4 subjects could not be used, and all the data from these 4 subjects were excluded from the primary analysis; 4 additional subjects were subsequently recruited allowing us to meet our block enrollment target (Figure 1). The first and last participants were studied on March 12, 2012 and August 28, 2012, respectively. The trial ended once the defined number of participants was enrolled. The occupations of the 105 participants are summarized in Table 1.

Figure 1

Colony counts of all scrubs cultured prior to use never exceeded 10 colonies. The median (IQR) total colony counts from all sites on the scrubs was 99 (66182) for standard scrubs, 137 (84289) for antimicrobial scrub type A, and 138 (62274) for antimicrobial scrub type B (P=0.36). We found no significant differences between the colony counts cultured from any of the individual sites among the 3 groups, regardless of occupation (Table 2). No significant difference was observed with respect to colony counts cultured from the wrist among the 3 study groups (Table 2). Comparisons between groups were planned a priori if a difference across all groups was found. Given the nonsignificant P values across all scrub groups, no further comparisons were made.

Fifty‐five participants (52%) reported caring for patients who were known to be colonized or infected with an antibiotic‐resistant organism, 16 (46%) randomized to wear standard scrubs, and 20 (57%) and 19 (54%) randomized to wear antimicrobial scrub A or B, respectively (P=0.61). Of these, however, antibiotic‐resistant organisms were only cultured from the scrubs of 2 providers (1 with 1 colony of MRSA from the breast pocket of antimicrobial scrub A, 1 with 1 colony of MRSA cultured from the pocket of antimicrobial scrub B [P=0.55]), and from the wrist of only 1 provider (a multiresistant Gram‐negative rod who wore antimicrobial scrub B).

DISCUSSION

The important findings of this study are that we found no evidence indicating that either of the 2 antimicrobial scrubs tested reduced bacterial contamination or antibiotic‐resistant contamination on HCWs' scrubs or wrists compared with standard scrubs at the end of an 8‐hour workday, and that despite many HCWs being exposed to patients who were colonized or infected with antibiotic‐resistant bacteria, these organisms were only rarely cultured from their uniforms.

We found that HCWs in all 3 arms of the study had bacterial contamination on their scrubs and skin, consistent with previous studies showing that HCWs' uniforms are frequently contaminated with bacteria, including MRSA, VRE, and other pathogens.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12] We previously found that bacterial contamination of HCWs' uniforms occurs within hours of putting on newly laundered uniforms.[12]

Literature on the effectiveness of antimicrobial HCW uniforms when tested in clinical settings is limited. Bearman and colleagues[23] recently published the results of a study of 31 subjects who wore either standard or antimicrobial scrubs, crossing over every 4 weeks for 4 months, with random culturing done weekly at the beginning and end of a work shift. Scrubs were laundered an average of 1.5 times/week, but the timing of the laundering relative to when cultures were obtained was not reported. Very few isolates of MRSA, Gram‐negative rods, or VRE were found (only 3.9%, 0.4%, and 0.05% of the 2000 samples obtained, respectively), and no differences were observed with respect to the number of HCWs who had antibiotic‐resistant organisms cultured when they were wearing standard versus antimicrobial scrubs. Those who had MRSA cultured, however, had lower mean log colony counts when they were wearing the antimicrobial scrubs. The small number of samples with positive isolates, together with differences in the extent of before‐shift contamination among groups complicates interpreting these data. The authors concluded that a prospective trial was needed. We attempted to include the scrub studied by Bearman and colleagues[23] in our study, but the company had insufficient stock available at the time we tried to purchase the product.

Gross and colleagues[24] found no difference in the mean colony counts of cultures taken from silver‐impregnated versus standard scrubs in a pilot crossover study done with 10 HCWs (although there were trends toward higher colony counts when the subjects wore antimicrobial scrubs).

Antibiotic‐resistant bacteria were only cultured from 3 participants (2.9%) in our current study, compared to 16% of those randomized to wearing white coats in our previous study and 20% of those randomized to wearing standard scrubs.[12] This difference may be explained by several recent studies reporting that rates of MRSA infections in hospitals are decreasing.[25, 26] The rate of hospital‐acquired MRSA infection or colonization at our own institution decreased 80% from 2007 to 2012. At the times of our previous and current studies, providers were expected to wear gowns and gloves when caring for patients as per standard contact precautions. Rates of infection and colonization of VRE and resistant Gram‐negative rods have remained low at our hospital, and our data are consistent with the rates reported on HCWs' uniforms in other studies.[2, 5, 10]

Only 6 of our subjects reported adverse reactions, but all were wearing antimicrobial scrubs (P=0.18). Several of the participants described that the fabrics of the 2 antimicrobial scrubs were heavier and less breathable than the standard scrubs. We believe this difference is more likely to explain the adverse reactions reported than is any type of reaction to the specific chemicals in the fabrics.

Our study has several limitations. Because it was conducted on the general internal medicine units of a single university‐affiliated public hospital, the results may not generalize to other types of institutions or other inpatient services.

As we previously described,[12] the RODAC imprint method only samples a small area of HCWs' uniforms and thus does not represent total bacterial contamination.[21] We specifically cultured areas that are known to be highly contaminated (ie, sleeve cuffs and pockets). Although imprint methods have limitations (as do other methods for culturing clothing), they have been commonly utilized in studies assessing bacterial contamination of HCW clothing.[2, 3, 5]

Although some of the bacterial load we cultured could have come from the providers themselves, previous studies have shown that 80% to 90% of the resistant bacteria cultured from HCWs' attire come from other sources.[1, 2]

Because our sample size was calculated on the basis of being able to detect a difference of 70% in total bacterial colony count, our study was not large enough to exclude a lower level of effectiveness. However, we saw no trends suggesting the antimicrobial products might have a lower level of effectiveness.

We did not observe the hand‐washing practices of the participants, and accordingly, cannot confirm that these practices were the same in each of our 3 study groups. Intermittent, surreptitious monitoring of hand‐washing practices on our internal medicine units over the last several years has found compliance with hand hygiene recommendations varying from 70% to 90%.

Although the participants in our study were not explicitly told to which scrub they were randomized, the colors, appearances, and textures of the antimicrobial fabrics were different from the standard scrubs such that blinding was impossible. Participants wearing antimicrobial scrubs could have changed their hand hygiene practices (ie, less careful hand hygiene). Lack of blinding could also have led to over‐reporting of adverse events by the subjects randomized to wear the antimicrobial scrubs.

In an effort to treat all the scrubs in the same fashion, all were tested new, prior to being washed or previously worn. Studying the scrubs prior to washing or wearing could have increased the reports of adverse effects, as the fabrics could have been stiffer and more uncomfortable than they might have been at a later stage in their use.

Our study also has some strengths. Our participants included physicians, residents, nurses, nurse practitioners, and physician assistants. Accordingly, our results should be generalizable to most HCWs. We also confirmed that the scrubs that were tested were nearly sterile prior to use.

In conclusion, we found no evidence suggesting that either of 2 antimicrobial scrubs tested decreased bacterial contamination of HCWs' scrubs or skin after an 8‐hour workday compared to standard scrubs. We also found that, although HCWs are frequently exposed to patients harboring antibiotic‐resistant bacteria, these bacteria were only rarely cultured from HCWs' scrubs or skin.

Children with complex chronic conditions comprise an increasing proportion of hospital admissions, readmissions, and resource use.[1, 2, 3] Dependence on technology or medical devices is a frequent characteristic of children in this group.[4] Adverse medical device events (AMDEs) are estimated to occur in as many as 8% of all adult admissions, depending on the methods used to identify them.[5] These events may result in hospitalizations or complicate hospital stays. To date, however, the burden of AMDEs among hospitalized children is little described, even though children may be at increased risk for device events as compared to adults.[6] Although some medical devices are intended solely or primarily for use with children, most devices used with children have been initially developed for, tested with, and most frequently employed to treat adults.[6] Assessing the continued safety and effectiveness of medical devices marketed in the Unites States is the responsibility of the Center for Devices and Radiologic Health of the US Food and Drug Administration (FDA). Its existing mechanisms for postmarket device surveillance rely primarily on passive reporting systems and specific observational studies.[7]

The objective of this study was to utilize administrative data from children's hospitals to explore the prevalence and nature of AMDEs in tertiary care children's hospitals that treat significant numbers of children with complex needs requiring medical devices.

METHODS

Data were obtained from the Pediatric Health Information System (PHIS), an administrative database containing inpatient data from 44 not‐for‐profit, tertiary care, pediatric hospitals affiliated with the Children's Hospital Association. Data are deidentified at the time of submission, and are subjected to a number of reliability and validity checks.[8] Individual admission records have both a deidentified visit identification (ID) and patient ID, allowing for linkage of multiple admissions by the same patient.

AMDEs were defined by International Classification of Diseases, Ninth Revision (ICD‐9) codes, using a methodology developed by Samore et al., who identified a set of such codes that specified devices in their definitions and therefore were considered to have a high likelihood of indicating a device problem (see Supporting Information, Table S1, in the online version of this article).[5] The diagnosis codes were grouped into device categories (eg, nervous system, orthopedic, cardiac).

From the 44 hospitals, the primary study cohort consisted of any patient with an admission between January 1, 2004 and December 31, 2011 with 1 AMDE ICD‐9 code as a primary or secondary diagnosis.

Descriptive statistics for patient demographics and visit characteristics of AMDE admissions were generated and stratified by device category. We reported these as counts and percentages for categorical variables and as median and interquartile range for length of stay. We also reported on how frequently patients with AMDEs have a top 10 most common diagnosis and top 10 most common procedure during the AMDE admission. We also reported the presence or absence of a complex chronic condition.[9] We generated the list of most common principal diagnoses and procedures by a separate query of PHIS from 2004 to 2009. Our top 10 most common diagnoses included ICD‐9 codes 486 (pneumonia), 466.11 (acute bronchiolitis due to respiratory syncytial virus), V58.11 (chemotherapy encounter), 493.92 (asthma exacerbation), 493.91 (asthma with status asthmaticus), 466.19 (acute bronchiolitis due to other organism), 780.39 (other convulsions), 540.9 (acute appendicitis), 282.62 (sickle cell disease with crisis), and 276.51 (dehydration). Our top 10 most common procedures included ICD‐9 codes 38.93 (venous catheterization), 03.31 (spinal tap), 99.04 (packed blood cell transfusion), 99.15 (parenteral nutrition), 99.25 (cancer chemotherapy), 96.71 (invasive mechanical ventilation, 96 hours), 96.04 (endotracheal intubation), 96.72 (invasive mechanical ventilation,95 hours), 96.6 (enteral nutrition), and 99.05 (platelet transfusion). Analyses were performed using SAS Enterprise Guide version 4.2 for Windows (SAS Institute, Cary, NC).

This study was approved by Cincinnati Children's Hospital Medical Center Institutional Review Board.

RESULTS

Of the 4,115,755 admissions during the study period, 136,465 (3.3%) had at least 1 AMDE. Over our study period, AMDEs were associated with a mean 17,058 inpatient stays annually. The number of AMDE‐related admissions decreased the last 4 years of our study period despite generally increasing admissions at PHIS hospitals (Figure 1). For 55% of the admissions (75,206/136,465), this AMDE code represented the primary diagnosis. Of these visits with a primary AMDE diagnosis, 39,874 (53%) were related to nervous system devices. The visits associated with AMDEs were comprised of 88,908 unique patients, 55% of whom were male (Table 1). The median age on admission was 6 years, and the interquartile range was 1 to 14 years of age.

Figure 1

Among admissions with AMDEs, 2.9% ended in death. The mortality was 0.5% when an AMDE was the primary diagnosis and 5.7% when the AMDE was a secondary diagnosis. The median length of inpatient stays was 6 days, with an interquartile range of 2 to 17 days.

Vascular access AMDEs were the most common event associated with admissions (26.6%), followed by nervous system devices (17.8%) (Table 2). The majority (75.5%) of patients admitted with AMDEs had a complex chronic condition. Less than half (46.8%) of AMDE admissions had an associated code for 1 of the 10 most common principal procedures. A minority (14.3%) of admissions had an associated ICD‐9 code for 1 of the top 10 most common principal diagnoses.

DISCUSSION

To our knowledge, our study is the first to report the burden of AMDEs among children requiring hospitalization. AMDEs are common in this population of children cared for at tertiary care children's hospitals, accounting for or complicating 3.3% of inpatient stays in these 44 hospitals. AMDEs were associated with a mean of >17,000 total visits per year. Vascular access devices and nervous system devices were the most common device categories linked to AMDEs. Similar to published literature, we found that the youngest children accounted for the highest proportion of AMDEs.[10, 11]

The majority (>75%) of children with an AMDE admission had diagnoses indicating complex chronic conditions during the admission. Over a partially overlapping study period, Feudtner and colleagues found 25.2% of patients admitted to PHIS hospitals had complex chronic conditions.[12] This finding, combined with the uncommon association of the most prevalent diagnoses and procedures, suggests that the burden of AMDEs falls disproportionately on this population of children. Death occurred considerably less commonly when AMDE diagnosis was the primary versus a secondary diagnosis (0.5% vs 5.7%). This finding likely illustrates 2 distinct populations: children with an AMDE that causes admission who have a relatively low risk of mortality and a second group who have AMDE‐complicated hospitalizations that may have an already high risk of mortality.

Our findings complement those of Wang and colleagues who employed the National Electronic Injury Surveillance System All Injury Program database to provide national estimates of medical device‐associated adverse events.[11] Importantly, this group used a different population (patients presenting to the emergency department) and a different methodology. These authors reported on device‐associated events, as they did not collect information to discriminate the device's role in the event. A walker that malfunctioned leading to patient injury would be a device‐related event; however, a patient who has a walker suffering a fall would be device‐associated, even if the walker's role in the injury was uncertain. We believe our methodology, established by Samore et al., more accurately identifies device‐related events.[5] Wang et al. found that 6.3% of pediatric patients who presented to emergency departments with medical device‐associated events were admitted to the hospital.[11] This resulted in national estimates of 9,082 events with 95% confidence intervals of 2,990 to 25,373 hospitalizations. Our findings of >17,000 AMDE‐related inpatient stays per annum included not only AMDEs leading to admissions but also those that were complications during stays.

Our study has several limitations, most related to the possibility of misclassification present in administrative data. Our approach only captured events that led to or complicated admissions. We suspect that ICD‐9 codes likely missed some AMDEs and that our estimates may therefore under‐represent this problem in our population. Future studies should compare our methodology, which has produced the first across‐center estimates of AMDE admissions, to alternative event capture techniques. We were unable to determine which events were present on admission and which complicated hospital stays, and it is likely that differing interventions would be required to reduce these 2 types of AMDEs. Another important limitation is that the PHIS database, comprised of data on children receiving care at tertiary academic medical centers with large numbers of pediatric subspecialists, is not representative of the population of children overall. The individual ICD‐9 codes for AMDEs are sufficiently nonspecific to limit the ability to characterize device events from administrative data alone. The high prevalence of unspecified device‐related admissions is an additional limitation. Although the estimates of these types of AMDEs are important in describing the frequency of these events, the unspecified category limits the ability to fully stratify based on device type and then implement monitoring strategies and interventions based on each.

To our knowledge, this study is the first multicenter analysis of the spectrum of pediatric AMDEs in hospitalized children. The AMDE prevalence is substantial, and the burden of these events largely falls on children with complex chronic conditions. Despite its limitations, this study complements recent efforts to enhance postmarket surveillance of pediatric devices including that of the FDA's Office of Pediatric Therapeutics, the recent FDA report Strengthening Our National System for Medical Device Postmarket Surveillance (http://www.fda.gov/MedicalDevices/default.htm), and the proposed rule for a unique device identification (UDI) system.[13] Establishment of UDI systems and their eventual incorporation into electronic health‐related databases will greatly expand postmarket surveillance capabilities.[13]

Our description of AMDEs by device category and patient characteristics is a first and necessary step to understanding the public health burden associated with device use in the pediatric population. Further developments in refined coding and device designation (eg, UDI systems) are needed to refine these estimates.

Children with complex chronic conditions comprise an increasing proportion of hospital admissions, readmissions, and resource use.[1, 2, 3] Dependence on technology or medical devices is a frequent characteristic of children in this group.[4] Adverse medical device events (AMDEs) are estimated to occur in as many as 8% of all adult admissions, depending on the methods used to identify them.[5] These events may result in hospitalizations or complicate hospital stays. To date, however, the burden of AMDEs among hospitalized children is little described, even though children may be at increased risk for device events as compared to adults.[6] Although some medical devices are intended solely or primarily for use with children, most devices used with children have been initially developed for, tested with, and most frequently employed to treat adults.[6] Assessing the continued safety and effectiveness of medical devices marketed in the Unites States is the responsibility of the Center for Devices and Radiologic Health of the US Food and Drug Administration (FDA). Its existing mechanisms for postmarket device surveillance rely primarily on passive reporting systems and specific observational studies.[7]

The objective of this study was to utilize administrative data from children's hospitals to explore the prevalence and nature of AMDEs in tertiary care children's hospitals that treat significant numbers of children with complex needs requiring medical devices.

METHODS

Data were obtained from the Pediatric Health Information System (PHIS), an administrative database containing inpatient data from 44 not‐for‐profit, tertiary care, pediatric hospitals affiliated with the Children's Hospital Association. Data are deidentified at the time of submission, and are subjected to a number of reliability and validity checks.[8] Individual admission records have both a deidentified visit identification (ID) and patient ID, allowing for linkage of multiple admissions by the same patient.

AMDEs were defined by International Classification of Diseases, Ninth Revision (ICD‐9) codes, using a methodology developed by Samore et al., who identified a set of such codes that specified devices in their definitions and therefore were considered to have a high likelihood of indicating a device problem (see Supporting Information, Table S1, in the online version of this article).[5] The diagnosis codes were grouped into device categories (eg, nervous system, orthopedic, cardiac).

From the 44 hospitals, the primary study cohort consisted of any patient with an admission between January 1, 2004 and December 31, 2011 with 1 AMDE ICD‐9 code as a primary or secondary diagnosis.

Descriptive statistics for patient demographics and visit characteristics of AMDE admissions were generated and stratified by device category. We reported these as counts and percentages for categorical variables and as median and interquartile range for length of stay. We also reported on how frequently patients with AMDEs have a top 10 most common diagnosis and top 10 most common procedure during the AMDE admission. We also reported the presence or absence of a complex chronic condition.[9] We generated the list of most common principal diagnoses and procedures by a separate query of PHIS from 2004 to 2009. Our top 10 most common diagnoses included ICD‐9 codes 486 (pneumonia), 466.11 (acute bronchiolitis due to respiratory syncytial virus), V58.11 (chemotherapy encounter), 493.92 (asthma exacerbation), 493.91 (asthma with status asthmaticus), 466.19 (acute bronchiolitis due to other organism), 780.39 (other convulsions), 540.9 (acute appendicitis), 282.62 (sickle cell disease with crisis), and 276.51 (dehydration). Our top 10 most common procedures included ICD‐9 codes 38.93 (venous catheterization), 03.31 (spinal tap), 99.04 (packed blood cell transfusion), 99.15 (parenteral nutrition), 99.25 (cancer chemotherapy), 96.71 (invasive mechanical ventilation, 96 hours), 96.04 (endotracheal intubation), 96.72 (invasive mechanical ventilation,95 hours), 96.6 (enteral nutrition), and 99.05 (platelet transfusion). Analyses were performed using SAS Enterprise Guide version 4.2 for Windows (SAS Institute, Cary, NC).

This study was approved by Cincinnati Children's Hospital Medical Center Institutional Review Board.

RESULTS

Of the 4,115,755 admissions during the study period, 136,465 (3.3%) had at least 1 AMDE. Over our study period, AMDEs were associated with a mean 17,058 inpatient stays annually. The number of AMDE‐related admissions decreased the last 4 years of our study period despite generally increasing admissions at PHIS hospitals (Figure 1). For 55% of the admissions (75,206/136,465), this AMDE code represented the primary diagnosis. Of these visits with a primary AMDE diagnosis, 39,874 (53%) were related to nervous system devices. The visits associated with AMDEs were comprised of 88,908 unique patients, 55% of whom were male (Table 1). The median age on admission was 6 years, and the interquartile range was 1 to 14 years of age.

Figure 1

Among admissions with AMDEs, 2.9% ended in death. The mortality was 0.5% when an AMDE was the primary diagnosis and 5.7% when the AMDE was a secondary diagnosis. The median length of inpatient stays was 6 days, with an interquartile range of 2 to 17 days.

Vascular access AMDEs were the most common event associated with admissions (26.6%), followed by nervous system devices (17.8%) (Table 2). The majority (75.5%) of patients admitted with AMDEs had a complex chronic condition. Less than half (46.8%) of AMDE admissions had an associated code for 1 of the 10 most common principal procedures. A minority (14.3%) of admissions had an associated ICD‐9 code for 1 of the top 10 most common principal diagnoses.

DISCUSSION

To our knowledge, our study is the first to report the burden of AMDEs among children requiring hospitalization. AMDEs are common in this population of children cared for at tertiary care children's hospitals, accounting for or complicating 3.3% of inpatient stays in these 44 hospitals. AMDEs were associated with a mean of >17,000 total visits per year. Vascular access devices and nervous system devices were the most common device categories linked to AMDEs. Similar to published literature, we found that the youngest children accounted for the highest proportion of AMDEs.[10, 11]

The majority (>75%) of children with an AMDE admission had diagnoses indicating complex chronic conditions during the admission. Over a partially overlapping study period, Feudtner and colleagues found 25.2% of patients admitted to PHIS hospitals had complex chronic conditions.[12] This finding, combined with the uncommon association of the most prevalent diagnoses and procedures, suggests that the burden of AMDEs falls disproportionately on this population of children. Death occurred considerably less commonly when AMDE diagnosis was the primary versus a secondary diagnosis (0.5% vs 5.7%). This finding likely illustrates 2 distinct populations: children with an AMDE that causes admission who have a relatively low risk of mortality and a second group who have AMDE‐complicated hospitalizations that may have an already high risk of mortality.

Our findings complement those of Wang and colleagues who employed the National Electronic Injury Surveillance System All Injury Program database to provide national estimates of medical device‐associated adverse events.[11] Importantly, this group used a different population (patients presenting to the emergency department) and a different methodology. These authors reported on device‐associated events, as they did not collect information to discriminate the device's role in the event. A walker that malfunctioned leading to patient injury would be a device‐related event; however, a patient who has a walker suffering a fall would be device‐associated, even if the walker's role in the injury was uncertain. We believe our methodology, established by Samore et al., more accurately identifies device‐related events.[5] Wang et al. found that 6.3% of pediatric patients who presented to emergency departments with medical device‐associated events were admitted to the hospital.[11] This resulted in national estimates of 9,082 events with 95% confidence intervals of 2,990 to 25,373 hospitalizations. Our findings of >17,000 AMDE‐related inpatient stays per annum included not only AMDEs leading to admissions but also those that were complications during stays.

Our study has several limitations, most related to the possibility of misclassification present in administrative data. Our approach only captured events that led to or complicated admissions. We suspect that ICD‐9 codes likely missed some AMDEs and that our estimates may therefore under‐represent this problem in our population. Future studies should compare our methodology, which has produced the first across‐center estimates of AMDE admissions, to alternative event capture techniques. We were unable to determine which events were present on admission and which complicated hospital stays, and it is likely that differing interventions would be required to reduce these 2 types of AMDEs. Another important limitation is that the PHIS database, comprised of data on children receiving care at tertiary academic medical centers with large numbers of pediatric subspecialists, is not representative of the population of children overall. The individual ICD‐9 codes for AMDEs are sufficiently nonspecific to limit the ability to characterize device events from administrative data alone. The high prevalence of unspecified device‐related admissions is an additional limitation. Although the estimates of these types of AMDEs are important in describing the frequency of these events, the unspecified category limits the ability to fully stratify based on device type and then implement monitoring strategies and interventions based on each.

To our knowledge, this study is the first multicenter analysis of the spectrum of pediatric AMDEs in hospitalized children. The AMDE prevalence is substantial, and the burden of these events largely falls on children with complex chronic conditions. Despite its limitations, this study complements recent efforts to enhance postmarket surveillance of pediatric devices including that of the FDA's Office of Pediatric Therapeutics, the recent FDA report Strengthening Our National System for Medical Device Postmarket Surveillance (http://www.fda.gov/MedicalDevices/default.htm), and the proposed rule for a unique device identification (UDI) system.[13] Establishment of UDI systems and their eventual incorporation into electronic health‐related databases will greatly expand postmarket surveillance capabilities.[13]

Our description of AMDEs by device category and patient characteristics is a first and necessary step to understanding the public health burden associated with device use in the pediatric population. Further developments in refined coding and device designation (eg, UDI systems) are needed to refine these estimates.

Children with complex chronic conditions comprise an increasing proportion of hospital admissions, readmissions, and resource use.[1, 2, 3] Dependence on technology or medical devices is a frequent characteristic of children in this group.[4] Adverse medical device events (AMDEs) are estimated to occur in as many as 8% of all adult admissions, depending on the methods used to identify them.[5] These events may result in hospitalizations or complicate hospital stays. To date, however, the burden of AMDEs among hospitalized children is little described, even though children may be at increased risk for device events as compared to adults.[6] Although some medical devices are intended solely or primarily for use with children, most devices used with children have been initially developed for, tested with, and most frequently employed to treat adults.[6] Assessing the continued safety and effectiveness of medical devices marketed in the Unites States is the responsibility of the Center for Devices and Radiologic Health of the US Food and Drug Administration (FDA). Its existing mechanisms for postmarket device surveillance rely primarily on passive reporting systems and specific observational studies.[7]

The objective of this study was to utilize administrative data from children's hospitals to explore the prevalence and nature of AMDEs in tertiary care children's hospitals that treat significant numbers of children with complex needs requiring medical devices.

METHODS

Data were obtained from the Pediatric Health Information System (PHIS), an administrative database containing inpatient data from 44 not‐for‐profit, tertiary care, pediatric hospitals affiliated with the Children's Hospital Association. Data are deidentified at the time of submission, and are subjected to a number of reliability and validity checks.[8] Individual admission records have both a deidentified visit identification (ID) and patient ID, allowing for linkage of multiple admissions by the same patient.

AMDEs were defined by International Classification of Diseases, Ninth Revision (ICD‐9) codes, using a methodology developed by Samore et al., who identified a set of such codes that specified devices in their definitions and therefore were considered to have a high likelihood of indicating a device problem (see Supporting Information, Table S1, in the online version of this article).[5] The diagnosis codes were grouped into device categories (eg, nervous system, orthopedic, cardiac).

From the 44 hospitals, the primary study cohort consisted of any patient with an admission between January 1, 2004 and December 31, 2011 with 1 AMDE ICD‐9 code as a primary or secondary diagnosis.

Descriptive statistics for patient demographics and visit characteristics of AMDE admissions were generated and stratified by device category. We reported these as counts and percentages for categorical variables and as median and interquartile range for length of stay. We also reported on how frequently patients with AMDEs have a top 10 most common diagnosis and top 10 most common procedure during the AMDE admission. We also reported the presence or absence of a complex chronic condition.[9] We generated the list of most common principal diagnoses and procedures by a separate query of PHIS from 2004 to 2009. Our top 10 most common diagnoses included ICD‐9 codes 486 (pneumonia), 466.11 (acute bronchiolitis due to respiratory syncytial virus), V58.11 (chemotherapy encounter), 493.92 (asthma exacerbation), 493.91 (asthma with status asthmaticus), 466.19 (acute bronchiolitis due to other organism), 780.39 (other convulsions), 540.9 (acute appendicitis), 282.62 (sickle cell disease with crisis), and 276.51 (dehydration). Our top 10 most common procedures included ICD‐9 codes 38.93 (venous catheterization), 03.31 (spinal tap), 99.04 (packed blood cell transfusion), 99.15 (parenteral nutrition), 99.25 (cancer chemotherapy), 96.71 (invasive mechanical ventilation, 96 hours), 96.04 (endotracheal intubation), 96.72 (invasive mechanical ventilation,95 hours), 96.6 (enteral nutrition), and 99.05 (platelet transfusion). Analyses were performed using SAS Enterprise Guide version 4.2 for Windows (SAS Institute, Cary, NC).

This study was approved by Cincinnati Children's Hospital Medical Center Institutional Review Board.

RESULTS

Of the 4,115,755 admissions during the study period, 136,465 (3.3%) had at least 1 AMDE. Over our study period, AMDEs were associated with a mean 17,058 inpatient stays annually. The number of AMDE‐related admissions decreased the last 4 years of our study period despite generally increasing admissions at PHIS hospitals (Figure 1). For 55% of the admissions (75,206/136,465), this AMDE code represented the primary diagnosis. Of these visits with a primary AMDE diagnosis, 39,874 (53%) were related to nervous system devices. The visits associated with AMDEs were comprised of 88,908 unique patients, 55% of whom were male (Table 1). The median age on admission was 6 years, and the interquartile range was 1 to 14 years of age.

Figure 1

Among admissions with AMDEs, 2.9% ended in death. The mortality was 0.5% when an AMDE was the primary diagnosis and 5.7% when the AMDE was a secondary diagnosis. The median length of inpatient stays was 6 days, with an interquartile range of 2 to 17 days.

Vascular access AMDEs were the most common event associated with admissions (26.6%), followed by nervous system devices (17.8%) (Table 2). The majority (75.5%) of patients admitted with AMDEs had a complex chronic condition. Less than half (46.8%) of AMDE admissions had an associated code for 1 of the 10 most common principal procedures. A minority (14.3%) of admissions had an associated ICD‐9 code for 1 of the top 10 most common principal diagnoses.

DISCUSSION

To our knowledge, our study is the first to report the burden of AMDEs among children requiring hospitalization. AMDEs are common in this population of children cared for at tertiary care children's hospitals, accounting for or complicating 3.3% of inpatient stays in these 44 hospitals. AMDEs were associated with a mean of >17,000 total visits per year. Vascular access devices and nervous system devices were the most common device categories linked to AMDEs. Similar to published literature, we found that the youngest children accounted for the highest proportion of AMDEs.[10, 11]

The majority (>75%) of children with an AMDE admission had diagnoses indicating complex chronic conditions during the admission. Over a partially overlapping study period, Feudtner and colleagues found 25.2% of patients admitted to PHIS hospitals had complex chronic conditions.[12] This finding, combined with the uncommon association of the most prevalent diagnoses and procedures, suggests that the burden of AMDEs falls disproportionately on this population of children. Death occurred considerably less commonly when AMDE diagnosis was the primary versus a secondary diagnosis (0.5% vs 5.7%). This finding likely illustrates 2 distinct populations: children with an AMDE that causes admission who have a relatively low risk of mortality and a second group who have AMDE‐complicated hospitalizations that may have an already high risk of mortality.

Our findings complement those of Wang and colleagues who employed the National Electronic Injury Surveillance System All Injury Program database to provide national estimates of medical device‐associated adverse events.[11] Importantly, this group used a different population (patients presenting to the emergency department) and a different methodology. These authors reported on device‐associated events, as they did not collect information to discriminate the device's role in the event. A walker that malfunctioned leading to patient injury would be a device‐related event; however, a patient who has a walker suffering a fall would be device‐associated, even if the walker's role in the injury was uncertain. We believe our methodology, established by Samore et al., more accurately identifies device‐related events.[5] Wang et al. found that 6.3% of pediatric patients who presented to emergency departments with medical device‐associated events were admitted to the hospital.[11] This resulted in national estimates of 9,082 events with 95% confidence intervals of 2,990 to 25,373 hospitalizations. Our findings of >17,000 AMDE‐related inpatient stays per annum included not only AMDEs leading to admissions but also those that were complications during stays.

Our study has several limitations, most related to the possibility of misclassification present in administrative data. Our approach only captured events that led to or complicated admissions. We suspect that ICD‐9 codes likely missed some AMDEs and that our estimates may therefore under‐represent this problem in our population. Future studies should compare our methodology, which has produced the first across‐center estimates of AMDE admissions, to alternative event capture techniques. We were unable to determine which events were present on admission and which complicated hospital stays, and it is likely that differing interventions would be required to reduce these 2 types of AMDEs. Another important limitation is that the PHIS database, comprised of data on children receiving care at tertiary academic medical centers with large numbers of pediatric subspecialists, is not representative of the population of children overall. The individual ICD‐9 codes for AMDEs are sufficiently nonspecific to limit the ability to characterize device events from administrative data alone. The high prevalence of unspecified device‐related admissions is an additional limitation. Although the estimates of these types of AMDEs are important in describing the frequency of these events, the unspecified category limits the ability to fully stratify based on device type and then implement monitoring strategies and interventions based on each.

To our knowledge, this study is the first multicenter analysis of the spectrum of pediatric AMDEs in hospitalized children. The AMDE prevalence is substantial, and the burden of these events largely falls on children with complex chronic conditions. Despite its limitations, this study complements recent efforts to enhance postmarket surveillance of pediatric devices including that of the FDA's Office of Pediatric Therapeutics, the recent FDA report Strengthening Our National System for Medical Device Postmarket Surveillance (http://www.fda.gov/MedicalDevices/default.htm), and the proposed rule for a unique device identification (UDI) system.[13] Establishment of UDI systems and their eventual incorporation into electronic health‐related databases will greatly expand postmarket surveillance capabilities.[13]

Our description of AMDEs by device category and patient characteristics is a first and necessary step to understanding the public health burden associated with device use in the pediatric population. Further developments in refined coding and device designation (eg, UDI systems) are needed to refine these estimates.