Donna M. Fick, PhD

Delirium (acute confusion) is common in older adults and leads to poor outcomes, such as death, clinician and caregiver burden, and prolonged cognitive and functional decline.[1, 2, 3, 4] Delirium is extremely costly, with estimates ranging from $143 to $152 billion annually (2005 US$).[5, 6] Early detection and management may improve the poor outcomes and reduce costs attributable to delirium,[3, 7] yet delirium identification in clinical practice has been challenging, particularly when translating research tools to the bedside.[8, 9, 10]As a result, only 12% to 35% of delirium cases are detected in routine care, with hypoactive delirium and delirium superimposed on dementia most likely to be missed.[11, 12, 13, 14, 15]

To address these issues, we recently developed and published the three‐dimensional Confusion Assessment Method (3D‐CAM), the 3‐minute diagnostic assessment for CAM‐defined delirium.[16] The 3D‐CAM is a structured assessment tool that includes mental status testing, patient symptom probes, and guided interviewer observations for signs of delirium. 3D‐CAM items were selected through a rigorous process to determine the most informative items for the 4 CAM diagnostic features.[17] The 3D‐CAM can be completed in 3 minutes, and has 95% sensitivity and 94% specificity relative to a reference standard.[16]

Despite the capabilities of the 3D‐CAM, there are situations when even 3 minutes is too long to devote to delirium identification. Moreover, a 2‐step approach in which a sensitive ultrabrief screen is administered, followed by the 3D‐CAM in positives, may be the most efficient approach for large‐scale delirium case identification. The aim of the current study was to use the 3D‐CAM database to identify the most sensitive single item and pair of items in the diagnosis of delirium, using the reference standard in the diagnostic accuracy analysis. We hypothesized that we could identify a single item with greater than 80% sensitivity and a pair of items with greater than 90% sensitivity for detection of delirium.

METHODS

RESULTS

Characteristics of the patients are shown in Table 1. Subjects had a mean age of 84 years, 62% were female, and 28% had a baseline dementia. Forty‐two (21%) had delirium based on the clinical reference standard. Twenty (10%) had less than a high school education and 100 (49%) had at least a college education.

DISCUSSION

Identifying simple, efficient, bedside case‐identification methods for delirium is an essential step toward improving recognition of this highly morbid syndrome in hospitalized older adults. In this study, we identified a single cognitive item, Months of the year backwards, that identified 83% of delirium cases when compared with a reference standard diagnosis. Furthermore, we identified 2 items, Months of the year backwards and What is the day of the week? which when used in combination identified 93% of delirium cases. The same 1 and 2 items also worked well in patients with dementia, in whom delirium is often missed. Although these items require further clinical validation, the development of an ultrabrief 2‐item test that identifies over 90% of delirium cases and can be completed in less than 1 minute (recently, we administered the best 2‐item screener to 20 consecutive general medicine patients over age 70 years, and it was completed in a median of 36.5 seconds), holds great potential for simplifying bedside delirium screening and improving the care of hospitalized older adults.

Our current findings both confirm and extend the emerging literature on best screening items for delirium. Sands and colleagues (2010)[26] tested a single test for delirium, Do you think (name of patient) has been more confused lately? in 21 subjects and achieved a sensitivity of 80%. Han and colleagues developed a screening tool in emergency‐department patients using the LOC question from the Richmond Agitation‐Sedation Scale and spelling the word lunch backwards, and achieved 98% sensitivity, but in a younger emergency department population with a low prevalence of dementia.[27] O'Regan et al. recently also found Months of the year backwards to be the best single‐screening item for delirium in a large sample, but only tested a 1‐item screen.[28] Our study extends these studies in several important ways by: (1) employing a rigorous clinical reference standard diagnosis of delirium, (2) having a large sample with a high prevalence of patients with dementia, (3) use of a general medical population, and (4) examining the best 2‐item screens in addition to the best single item.

Systematic intervention programs[29, 30, 31] that focus on improved delirium evaluation and management have the potential to improve patient outcomes and reduce costs. However, targeting these programs to patients with delirium has proven difficult, as only 12% to 35% of delirium cases are recognized in routine clinical practice.[11, 12, 13, 14, 15] The 1‐ and 2‐item screeners we identified could play an important role in future delirium identification. The 3D‐CAM combines high sensitivity (95%) with high specificity (94%)[16] and therefore would be an excellent choice as the second step after a positive screen. The feasibility, effectiveness, and cost of administering these screeners, followed by a brief diagnostic tool such as the 3D‐CAM, should be evaluated in future work.

Our study has noteworthy strengths, including the use of a large purposefully challenging clinical sample with advanced age that included a substantial proportion with dementia, a detailed assessment, and the testing of very brief and practical tools for bedside delirium screening.[25] This study also has several important limitations. Most importantly, we presented secondary analysis of individual items and pairs of items drawn from the 3D CAM assessment; therefore, the 2‐item bedside screen requires prospective clinical validation. The reference standard was based on the DSM‐IV, because this study was conducted prior to the release of DSM‐V. In addition, the ordering of the reference standard and 3D‐CAM assessments was not randomized due to feasibility constraints. In addition, this study was cross‐sectional, involved only a single hospital, and enrolled only older medical patients during the day shift. Our sample was older (aged 75 years and older), and a younger sample may have had a different prevalence of delirium, which could affect the positive predictive value of our ultrabrief screen. We plan to test this in a sample of patients aged 70 years and older in future studies. Finally, it should be noted that these best 1‐item and 2‐item screeners miss 17% and 7% of delirium cases, respectively. In cases where this is unacceptably high, alternative approaches might be necessary.

It is important to remember that these 1‐ and 2‐item screeners are not diagnostic tools and therefore should not be used in isolation. Optimally, they will be followed by a more specific evaluation, such as the 3D‐CAM, as part of a systematic delirium identification process. For instance, in our sample (with a delirium rate of 21%), the best 2‐item screener had a positive predictive value of 41%, meaning that positive screens are more likely to be false positives than true positives (see Supporting Tables 2 and 3 in the online version of this article).[32] Nevertheless, by reducing the total number of patients who require diagnostic instrument administration, use of these ultrabrief screeners can improve efficiency and result in a net benefit to delirium case‐identification efforts.[32]

Time has been demonstrated to be a barrier to delirium identification in previous studies, but there are likely others. These may include, for instance, staff nihilism about screening making a difference, ambiguous responsibility for delirium screening and management, unsupportive system leadership, and absent payment for these activities.[31] Moreover, it is possible that the 2‐step process we propose may create an incentive for staff to avoid positive screens as they see it creating more work for themselves. We plan to identify and address such barriers in our future work.

In conclusion, we identified a single screening item for delirium, Months of the year backwards, with 83% sensitivity, and a pair of items, Months of the year backwards and What is the day of the week?, with 93% sensitivity relative to a rigorous reference standard diagnosis. These ultrabrief screening items work well in patients with and without dementia, and should require very little training of staff. Future studies should further validate these tools, and determine their translatability and scalability into programs for systematic, widespread delirium detection. Developing efficient and accurate case identification strategies is a necessary prerequisite to appropriately target delirium management protocols, enabling healthcare systems to effectively address this costly and deadly condition.

Delirium (acute confusion) is common in older adults and leads to poor outcomes, such as death, clinician and caregiver burden, and prolonged cognitive and functional decline.[1, 2, 3, 4] Delirium is extremely costly, with estimates ranging from $143 to $152 billion annually (2005 US$).[5, 6] Early detection and management may improve the poor outcomes and reduce costs attributable to delirium,[3, 7] yet delirium identification in clinical practice has been challenging, particularly when translating research tools to the bedside.[8, 9, 10]As a result, only 12% to 35% of delirium cases are detected in routine care, with hypoactive delirium and delirium superimposed on dementia most likely to be missed.[11, 12, 13, 14, 15]

To address these issues, we recently developed and published the three‐dimensional Confusion Assessment Method (3D‐CAM), the 3‐minute diagnostic assessment for CAM‐defined delirium.[16] The 3D‐CAM is a structured assessment tool that includes mental status testing, patient symptom probes, and guided interviewer observations for signs of delirium. 3D‐CAM items were selected through a rigorous process to determine the most informative items for the 4 CAM diagnostic features.[17] The 3D‐CAM can be completed in 3 minutes, and has 95% sensitivity and 94% specificity relative to a reference standard.[16]

Despite the capabilities of the 3D‐CAM, there are situations when even 3 minutes is too long to devote to delirium identification. Moreover, a 2‐step approach in which a sensitive ultrabrief screen is administered, followed by the 3D‐CAM in positives, may be the most efficient approach for large‐scale delirium case identification. The aim of the current study was to use the 3D‐CAM database to identify the most sensitive single item and pair of items in the diagnosis of delirium, using the reference standard in the diagnostic accuracy analysis. We hypothesized that we could identify a single item with greater than 80% sensitivity and a pair of items with greater than 90% sensitivity for detection of delirium.

METHODS

RESULTS

Characteristics of the patients are shown in Table 1. Subjects had a mean age of 84 years, 62% were female, and 28% had a baseline dementia. Forty‐two (21%) had delirium based on the clinical reference standard. Twenty (10%) had less than a high school education and 100 (49%) had at least a college education.

DISCUSSION

Identifying simple, efficient, bedside case‐identification methods for delirium is an essential step toward improving recognition of this highly morbid syndrome in hospitalized older adults. In this study, we identified a single cognitive item, Months of the year backwards, that identified 83% of delirium cases when compared with a reference standard diagnosis. Furthermore, we identified 2 items, Months of the year backwards and What is the day of the week? which when used in combination identified 93% of delirium cases. The same 1 and 2 items also worked well in patients with dementia, in whom delirium is often missed. Although these items require further clinical validation, the development of an ultrabrief 2‐item test that identifies over 90% of delirium cases and can be completed in less than 1 minute (recently, we administered the best 2‐item screener to 20 consecutive general medicine patients over age 70 years, and it was completed in a median of 36.5 seconds), holds great potential for simplifying bedside delirium screening and improving the care of hospitalized older adults.

Our current findings both confirm and extend the emerging literature on best screening items for delirium. Sands and colleagues (2010)[26] tested a single test for delirium, Do you think (name of patient) has been more confused lately? in 21 subjects and achieved a sensitivity of 80%. Han and colleagues developed a screening tool in emergency‐department patients using the LOC question from the Richmond Agitation‐Sedation Scale and spelling the word lunch backwards, and achieved 98% sensitivity, but in a younger emergency department population with a low prevalence of dementia.[27] O'Regan et al. recently also found Months of the year backwards to be the best single‐screening item for delirium in a large sample, but only tested a 1‐item screen.[28] Our study extends these studies in several important ways by: (1) employing a rigorous clinical reference standard diagnosis of delirium, (2) having a large sample with a high prevalence of patients with dementia, (3) use of a general medical population, and (4) examining the best 2‐item screens in addition to the best single item.

Systematic intervention programs[29, 30, 31] that focus on improved delirium evaluation and management have the potential to improve patient outcomes and reduce costs. However, targeting these programs to patients with delirium has proven difficult, as only 12% to 35% of delirium cases are recognized in routine clinical practice.[11, 12, 13, 14, 15] The 1‐ and 2‐item screeners we identified could play an important role in future delirium identification. The 3D‐CAM combines high sensitivity (95%) with high specificity (94%)[16] and therefore would be an excellent choice as the second step after a positive screen. The feasibility, effectiveness, and cost of administering these screeners, followed by a brief diagnostic tool such as the 3D‐CAM, should be evaluated in future work.

Our study has noteworthy strengths, including the use of a large purposefully challenging clinical sample with advanced age that included a substantial proportion with dementia, a detailed assessment, and the testing of very brief and practical tools for bedside delirium screening.[25] This study also has several important limitations. Most importantly, we presented secondary analysis of individual items and pairs of items drawn from the 3D CAM assessment; therefore, the 2‐item bedside screen requires prospective clinical validation. The reference standard was based on the DSM‐IV, because this study was conducted prior to the release of DSM‐V. In addition, the ordering of the reference standard and 3D‐CAM assessments was not randomized due to feasibility constraints. In addition, this study was cross‐sectional, involved only a single hospital, and enrolled only older medical patients during the day shift. Our sample was older (aged 75 years and older), and a younger sample may have had a different prevalence of delirium, which could affect the positive predictive value of our ultrabrief screen. We plan to test this in a sample of patients aged 70 years and older in future studies. Finally, it should be noted that these best 1‐item and 2‐item screeners miss 17% and 7% of delirium cases, respectively. In cases where this is unacceptably high, alternative approaches might be necessary.

It is important to remember that these 1‐ and 2‐item screeners are not diagnostic tools and therefore should not be used in isolation. Optimally, they will be followed by a more specific evaluation, such as the 3D‐CAM, as part of a systematic delirium identification process. For instance, in our sample (with a delirium rate of 21%), the best 2‐item screener had a positive predictive value of 41%, meaning that positive screens are more likely to be false positives than true positives (see Supporting Tables 2 and 3 in the online version of this article).[32] Nevertheless, by reducing the total number of patients who require diagnostic instrument administration, use of these ultrabrief screeners can improve efficiency and result in a net benefit to delirium case‐identification efforts.[32]

Time has been demonstrated to be a barrier to delirium identification in previous studies, but there are likely others. These may include, for instance, staff nihilism about screening making a difference, ambiguous responsibility for delirium screening and management, unsupportive system leadership, and absent payment for these activities.[31] Moreover, it is possible that the 2‐step process we propose may create an incentive for staff to avoid positive screens as they see it creating more work for themselves. We plan to identify and address such barriers in our future work.

In conclusion, we identified a single screening item for delirium, Months of the year backwards, with 83% sensitivity, and a pair of items, Months of the year backwards and What is the day of the week?, with 93% sensitivity relative to a rigorous reference standard diagnosis. These ultrabrief screening items work well in patients with and without dementia, and should require very little training of staff. Future studies should further validate these tools, and determine their translatability and scalability into programs for systematic, widespread delirium detection. Developing efficient and accurate case identification strategies is a necessary prerequisite to appropriately target delirium management protocols, enabling healthcare systems to effectively address this costly and deadly condition.

Delirium (acute confusion) is common in older adults and leads to poor outcomes, such as death, clinician and caregiver burden, and prolonged cognitive and functional decline.[1, 2, 3, 4] Delirium is extremely costly, with estimates ranging from $143 to $152 billion annually (2005 US$).[5, 6] Early detection and management may improve the poor outcomes and reduce costs attributable to delirium,[3, 7] yet delirium identification in clinical practice has been challenging, particularly when translating research tools to the bedside.[8, 9, 10]As a result, only 12% to 35% of delirium cases are detected in routine care, with hypoactive delirium and delirium superimposed on dementia most likely to be missed.[11, 12, 13, 14, 15]

To address these issues, we recently developed and published the three‐dimensional Confusion Assessment Method (3D‐CAM), the 3‐minute diagnostic assessment for CAM‐defined delirium.[16] The 3D‐CAM is a structured assessment tool that includes mental status testing, patient symptom probes, and guided interviewer observations for signs of delirium. 3D‐CAM items were selected through a rigorous process to determine the most informative items for the 4 CAM diagnostic features.[17] The 3D‐CAM can be completed in 3 minutes, and has 95% sensitivity and 94% specificity relative to a reference standard.[16]

Despite the capabilities of the 3D‐CAM, there are situations when even 3 minutes is too long to devote to delirium identification. Moreover, a 2‐step approach in which a sensitive ultrabrief screen is administered, followed by the 3D‐CAM in positives, may be the most efficient approach for large‐scale delirium case identification. The aim of the current study was to use the 3D‐CAM database to identify the most sensitive single item and pair of items in the diagnosis of delirium, using the reference standard in the diagnostic accuracy analysis. We hypothesized that we could identify a single item with greater than 80% sensitivity and a pair of items with greater than 90% sensitivity for detection of delirium.

METHODS

RESULTS

Characteristics of the patients are shown in Table 1. Subjects had a mean age of 84 years, 62% were female, and 28% had a baseline dementia. Forty‐two (21%) had delirium based on the clinical reference standard. Twenty (10%) had less than a high school education and 100 (49%) had at least a college education.

DISCUSSION

Identifying simple, efficient, bedside case‐identification methods for delirium is an essential step toward improving recognition of this highly morbid syndrome in hospitalized older adults. In this study, we identified a single cognitive item, Months of the year backwards, that identified 83% of delirium cases when compared with a reference standard diagnosis. Furthermore, we identified 2 items, Months of the year backwards and What is the day of the week? which when used in combination identified 93% of delirium cases. The same 1 and 2 items also worked well in patients with dementia, in whom delirium is often missed. Although these items require further clinical validation, the development of an ultrabrief 2‐item test that identifies over 90% of delirium cases and can be completed in less than 1 minute (recently, we administered the best 2‐item screener to 20 consecutive general medicine patients over age 70 years, and it was completed in a median of 36.5 seconds), holds great potential for simplifying bedside delirium screening and improving the care of hospitalized older adults.

Our current findings both confirm and extend the emerging literature on best screening items for delirium. Sands and colleagues (2010)[26] tested a single test for delirium, Do you think (name of patient) has been more confused lately? in 21 subjects and achieved a sensitivity of 80%. Han and colleagues developed a screening tool in emergency‐department patients using the LOC question from the Richmond Agitation‐Sedation Scale and spelling the word lunch backwards, and achieved 98% sensitivity, but in a younger emergency department population with a low prevalence of dementia.[27] O'Regan et al. recently also found Months of the year backwards to be the best single‐screening item for delirium in a large sample, but only tested a 1‐item screen.[28] Our study extends these studies in several important ways by: (1) employing a rigorous clinical reference standard diagnosis of delirium, (2) having a large sample with a high prevalence of patients with dementia, (3) use of a general medical population, and (4) examining the best 2‐item screens in addition to the best single item.

Systematic intervention programs[29, 30, 31] that focus on improved delirium evaluation and management have the potential to improve patient outcomes and reduce costs. However, targeting these programs to patients with delirium has proven difficult, as only 12% to 35% of delirium cases are recognized in routine clinical practice.[11, 12, 13, 14, 15] The 1‐ and 2‐item screeners we identified could play an important role in future delirium identification. The 3D‐CAM combines high sensitivity (95%) with high specificity (94%)[16] and therefore would be an excellent choice as the second step after a positive screen. The feasibility, effectiveness, and cost of administering these screeners, followed by a brief diagnostic tool such as the 3D‐CAM, should be evaluated in future work.

Our study has noteworthy strengths, including the use of a large purposefully challenging clinical sample with advanced age that included a substantial proportion with dementia, a detailed assessment, and the testing of very brief and practical tools for bedside delirium screening.[25] This study also has several important limitations. Most importantly, we presented secondary analysis of individual items and pairs of items drawn from the 3D CAM assessment; therefore, the 2‐item bedside screen requires prospective clinical validation. The reference standard was based on the DSM‐IV, because this study was conducted prior to the release of DSM‐V. In addition, the ordering of the reference standard and 3D‐CAM assessments was not randomized due to feasibility constraints. In addition, this study was cross‐sectional, involved only a single hospital, and enrolled only older medical patients during the day shift. Our sample was older (aged 75 years and older), and a younger sample may have had a different prevalence of delirium, which could affect the positive predictive value of our ultrabrief screen. We plan to test this in a sample of patients aged 70 years and older in future studies. Finally, it should be noted that these best 1‐item and 2‐item screeners miss 17% and 7% of delirium cases, respectively. In cases where this is unacceptably high, alternative approaches might be necessary.

It is important to remember that these 1‐ and 2‐item screeners are not diagnostic tools and therefore should not be used in isolation. Optimally, they will be followed by a more specific evaluation, such as the 3D‐CAM, as part of a systematic delirium identification process. For instance, in our sample (with a delirium rate of 21%), the best 2‐item screener had a positive predictive value of 41%, meaning that positive screens are more likely to be false positives than true positives (see Supporting Tables 2 and 3 in the online version of this article).[32] Nevertheless, by reducing the total number of patients who require diagnostic instrument administration, use of these ultrabrief screeners can improve efficiency and result in a net benefit to delirium case‐identification efforts.[32]

Time has been demonstrated to be a barrier to delirium identification in previous studies, but there are likely others. These may include, for instance, staff nihilism about screening making a difference, ambiguous responsibility for delirium screening and management, unsupportive system leadership, and absent payment for these activities.[31] Moreover, it is possible that the 2‐step process we propose may create an incentive for staff to avoid positive screens as they see it creating more work for themselves. We plan to identify and address such barriers in our future work.

In conclusion, we identified a single screening item for delirium, Months of the year backwards, with 83% sensitivity, and a pair of items, Months of the year backwards and What is the day of the week?, with 93% sensitivity relative to a rigorous reference standard diagnosis. These ultrabrief screening items work well in patients with and without dementia, and should require very little training of staff. Future studies should further validate these tools, and determine their translatability and scalability into programs for systematic, widespread delirium detection. Developing efficient and accurate case identification strategies is a necessary prerequisite to appropriately target delirium management protocols, enabling healthcare systems to effectively address this costly and deadly condition.