Hospitalists play key roles in many types of clinical services, including teaching, nonteaching, consultative, and comanagement services.14 While the impact of hospitalist programs on LOS for inpatient medicine services has been studied,58 less work has focused on the impact of hospitalists in other types of service delivery, such as in short‐stay or observation units.

While many hospitals now have short‐stay units to care for observation patients, most are adjuncts of the emergency department. A Canadian hospitalist‐run short‐stay unit that targeted patients with an expected LOS of less than 3 days has been described.9 The experience of a single, chest‐painspecific service has also been reported.10

In August 2005, we introduced a hospitalist‐run observation unit, the Clinical Decision Unit (CDU), at University Hospital, the primary teaching affiliate of the University of Texas Health Science Center at San Antonio (San Antonio, TX). The rationale was that observation‐level care in a dedicated short‐stay unit would be more efficient than in an inpatient general medicine service. Through the creation of this unit, we consolidated the care of all medical observation patients, including patients previously evaluated in a cardiology‐run chest pain unit.

In this brief report, we present a description of the unit as well as a preliminary analysis of the impact of the unit on LOS for the most common CDU diagnoses.

Methods

Results

Discussion and Conclusions

Implementation of a hospitalist‐run observation unit was associated with an overall decreased LOS for patients with the 5 most common CDU discharge diagnoses of chest pain, cellulitis, asthma, pyelonephritis, and syncope. The lack of statistically significantly differences in patient acuity in the preimplementation and postimplementation periods suggests this result is not due to acuity differences, but rather to unit implementation. We believe this reduction resulted from the greater efficiencies of care that occur from clustering observation patients in a geographically separate unit with dedicated nursing staff and efficient workflow. The reduction of 0.2 days over 2148 patients (total number of postimplementation discharges) led to an additional 429.6 days of capacity without adding additional beds. Thus, what might appear to be a modest LOS reduction has a larger impact when patient volume is considered.

For individual diagnoses, significant differences in LOS were seen for patients with cellulitis and asthma The lack of a difference for chest pain may be related to the fact that these patients were cared for in a chest pain unit prior to CDU creation, which likely fostered similar efficiencies. This finding may suggest that hospitalists are as efficient as cardiologists in assessing patients with chest pain. The lack of a difference in LOS for syncope may have reflected a bottleneck in obtaining echocardiogram tests. Finally, the lack of a difference for pyelonephritis may indicate that it is not a diagnosis for which observation is beneficial.

While our use of administrative data over the year‐long preimplementation and postimplementation periods allows for the inclusion of a large number of discharges, the retrospective study design limits the strength of our results. A prospective study would more definitively reduce the possibility of bias and ensure the validity of our finding of reduced LOS.

The creation of a hospitalist‐run observation unit may represent an alternative to emergency departmentrun units. It allows physicians with greater expertise in inpatient medicine to make admission and discharge decisions, allowing emergency department physicians to concentrate on the care of other patients. This can be particularly critical for high‐volume emergency departments. The CDU also offers an alternative to specialist‐run chest pain units. Because patients either stay for only the observation period or are admitted and typically moved off the unit, there is little need for provider continuity, and the discontinuous shift staffing model works well.

In addition to the geographic localization, several aspects of the CDU model may be critical to the successful implementation of similar hospitalist‐run observation units. Dedicated nursing staff with expertise in caring for high‐turnover patients with a more limited spectrum of diagnoses may be a factor. Another factor may be that the lack of less‐experienced trainees in a nonteaching service leads to more efficient care.

A potential area of further exploration includes understanding the differences between CDU patients who are discharged within 23 hours and those who are later admitted. This understanding may help us better differentiate patients appropriate for CDU admission, allowing the creation of more formal admission criteria.

Hospitalists play key roles in many types of clinical services, including teaching, nonteaching, consultative, and comanagement services.14 While the impact of hospitalist programs on LOS for inpatient medicine services has been studied,58 less work has focused on the impact of hospitalists in other types of service delivery, such as in short‐stay or observation units.

While many hospitals now have short‐stay units to care for observation patients, most are adjuncts of the emergency department. A Canadian hospitalist‐run short‐stay unit that targeted patients with an expected LOS of less than 3 days has been described.9 The experience of a single, chest‐painspecific service has also been reported.10

In August 2005, we introduced a hospitalist‐run observation unit, the Clinical Decision Unit (CDU), at University Hospital, the primary teaching affiliate of the University of Texas Health Science Center at San Antonio (San Antonio, TX). The rationale was that observation‐level care in a dedicated short‐stay unit would be more efficient than in an inpatient general medicine service. Through the creation of this unit, we consolidated the care of all medical observation patients, including patients previously evaluated in a cardiology‐run chest pain unit.

In this brief report, we present a description of the unit as well as a preliminary analysis of the impact of the unit on LOS for the most common CDU diagnoses.

Methods

Results

Discussion and Conclusions

Implementation of a hospitalist‐run observation unit was associated with an overall decreased LOS for patients with the 5 most common CDU discharge diagnoses of chest pain, cellulitis, asthma, pyelonephritis, and syncope. The lack of statistically significantly differences in patient acuity in the preimplementation and postimplementation periods suggests this result is not due to acuity differences, but rather to unit implementation. We believe this reduction resulted from the greater efficiencies of care that occur from clustering observation patients in a geographically separate unit with dedicated nursing staff and efficient workflow. The reduction of 0.2 days over 2148 patients (total number of postimplementation discharges) led to an additional 429.6 days of capacity without adding additional beds. Thus, what might appear to be a modest LOS reduction has a larger impact when patient volume is considered.

For individual diagnoses, significant differences in LOS were seen for patients with cellulitis and asthma The lack of a difference for chest pain may be related to the fact that these patients were cared for in a chest pain unit prior to CDU creation, which likely fostered similar efficiencies. This finding may suggest that hospitalists are as efficient as cardiologists in assessing patients with chest pain. The lack of a difference in LOS for syncope may have reflected a bottleneck in obtaining echocardiogram tests. Finally, the lack of a difference for pyelonephritis may indicate that it is not a diagnosis for which observation is beneficial.

While our use of administrative data over the year‐long preimplementation and postimplementation periods allows for the inclusion of a large number of discharges, the retrospective study design limits the strength of our results. A prospective study would more definitively reduce the possibility of bias and ensure the validity of our finding of reduced LOS.

The creation of a hospitalist‐run observation unit may represent an alternative to emergency departmentrun units. It allows physicians with greater expertise in inpatient medicine to make admission and discharge decisions, allowing emergency department physicians to concentrate on the care of other patients. This can be particularly critical for high‐volume emergency departments. The CDU also offers an alternative to specialist‐run chest pain units. Because patients either stay for only the observation period or are admitted and typically moved off the unit, there is little need for provider continuity, and the discontinuous shift staffing model works well.

In addition to the geographic localization, several aspects of the CDU model may be critical to the successful implementation of similar hospitalist‐run observation units. Dedicated nursing staff with expertise in caring for high‐turnover patients with a more limited spectrum of diagnoses may be a factor. Another factor may be that the lack of less‐experienced trainees in a nonteaching service leads to more efficient care.

A potential area of further exploration includes understanding the differences between CDU patients who are discharged within 23 hours and those who are later admitted. This understanding may help us better differentiate patients appropriate for CDU admission, allowing the creation of more formal admission criteria.

Hospitalists play key roles in many types of clinical services, including teaching, nonteaching, consultative, and comanagement services.14 While the impact of hospitalist programs on LOS for inpatient medicine services has been studied,58 less work has focused on the impact of hospitalists in other types of service delivery, such as in short‐stay or observation units.

While many hospitals now have short‐stay units to care for observation patients, most are adjuncts of the emergency department. A Canadian hospitalist‐run short‐stay unit that targeted patients with an expected LOS of less than 3 days has been described.9 The experience of a single, chest‐painspecific service has also been reported.10

In August 2005, we introduced a hospitalist‐run observation unit, the Clinical Decision Unit (CDU), at University Hospital, the primary teaching affiliate of the University of Texas Health Science Center at San Antonio (San Antonio, TX). The rationale was that observation‐level care in a dedicated short‐stay unit would be more efficient than in an inpatient general medicine service. Through the creation of this unit, we consolidated the care of all medical observation patients, including patients previously evaluated in a cardiology‐run chest pain unit.

In this brief report, we present a description of the unit as well as a preliminary analysis of the impact of the unit on LOS for the most common CDU diagnoses.

Methods

Results

Discussion and Conclusions

Implementation of a hospitalist‐run observation unit was associated with an overall decreased LOS for patients with the 5 most common CDU discharge diagnoses of chest pain, cellulitis, asthma, pyelonephritis, and syncope. The lack of statistically significantly differences in patient acuity in the preimplementation and postimplementation periods suggests this result is not due to acuity differences, but rather to unit implementation. We believe this reduction resulted from the greater efficiencies of care that occur from clustering observation patients in a geographically separate unit with dedicated nursing staff and efficient workflow. The reduction of 0.2 days over 2148 patients (total number of postimplementation discharges) led to an additional 429.6 days of capacity without adding additional beds. Thus, what might appear to be a modest LOS reduction has a larger impact when patient volume is considered.

For individual diagnoses, significant differences in LOS were seen for patients with cellulitis and asthma The lack of a difference for chest pain may be related to the fact that these patients were cared for in a chest pain unit prior to CDU creation, which likely fostered similar efficiencies. This finding may suggest that hospitalists are as efficient as cardiologists in assessing patients with chest pain. The lack of a difference in LOS for syncope may have reflected a bottleneck in obtaining echocardiogram tests. Finally, the lack of a difference for pyelonephritis may indicate that it is not a diagnosis for which observation is beneficial.

While our use of administrative data over the year‐long preimplementation and postimplementation periods allows for the inclusion of a large number of discharges, the retrospective study design limits the strength of our results. A prospective study would more definitively reduce the possibility of bias and ensure the validity of our finding of reduced LOS.

The creation of a hospitalist‐run observation unit may represent an alternative to emergency departmentrun units. It allows physicians with greater expertise in inpatient medicine to make admission and discharge decisions, allowing emergency department physicians to concentrate on the care of other patients. This can be particularly critical for high‐volume emergency departments. The CDU also offers an alternative to specialist‐run chest pain units. Because patients either stay for only the observation period or are admitted and typically moved off the unit, there is little need for provider continuity, and the discontinuous shift staffing model works well.

In addition to the geographic localization, several aspects of the CDU model may be critical to the successful implementation of similar hospitalist‐run observation units. Dedicated nursing staff with expertise in caring for high‐turnover patients with a more limited spectrum of diagnoses may be a factor. Another factor may be that the lack of less‐experienced trainees in a nonteaching service leads to more efficient care.

A potential area of further exploration includes understanding the differences between CDU patients who are discharged within 23 hours and those who are later admitted. This understanding may help us better differentiate patients appropriate for CDU admission, allowing the creation of more formal admission criteria.

Nearly half of the hospital beds in the United States are occupied by the elderly,1 whose numbers are increasing.2 The odds of a hospitalized Medicare patient being cared for by a hospitalist are increasing by nearly 30% per year.3 Hospitalists require competence in geriatrics to serve their patients and to teach trainees. Train‐the‐Trainer (TTT) programs both educate health care providers and provide educational materials, information, and skills for teaching others.4 This model has been successfully used in geriatrics to impact knowledge, attitudes, and self‐efficacy among health care workers.46

A prominent example of a geriatrics TTT program is the University of Chicago Curriculum for the Hospitalized Aging Medical Patient (CHAMP),7 which requires 48 hours of instruction over 12 sessions. To create a less time‐intensive learning format for busy hospitalists, the University of Chicago developed Mini‐CHAMP, a streamlined 2‐day workshop with web‐based components for hospitalist clinicians, but not necessarily hospitalist educators.7

We created The Donald W. Reynolds Program for Advancing Geriatrics Education (PAGE) at the University of California, San Francisco (UCSF), in light of the time intensity of CHAMP, to integrate geriatric TTT sessions within preexisting hospitalist faculty meetings. This model is consistent with current practices in faculty development.8 This paper describes the evaluation of the PAGE Model, which sought answers to 3 research questions: (1) Does PAGE increase faculty confidence in teaching geriatrics?, (2) Does PAGE increase the frequency of hospitalist teaching geriatrics topics?, and (3) Does PAGE increase residents' practice of geriatrics skills?

Methods

Results

The hospitalist group grew from 31 to 36 members in June of 2008. On average, 14 hospitalists (M = 14.40, standard deviation [SD] = 2.41, range 1119) attended each session, with all hospitalists (n = 36) attending 1 session (M = 3.83, SD = 2.35, range 19). At each session, an average of 72% completed a post‐session evaluation form. Overall, faculty were likely to use the PAGE teaching tools (M = 4.61, SD = 0.53) and would recommend PAGE to other hospitalists (M = 4.63, SD = 0.51).

Thirteen hospitalist trainees of 36 (36%) completed a post‐PAGE online questionnaire. Respondents taught on faculty for an average of 5 years (mean (M) = 5.08, SD = 3.52). Faculty perceived self‐efficacy at teaching residents about geriatrics improved significantly with a large effect size (pretest M = 3.05, SD = .60; posttest M = 3.96, SD = .36, d = 1.52; P < 0.001). Session attendance was positively correlated with the increase in geriatrics teaching self‐efficacy (r = .62, P < 0.05), while teaching experience was not (r = 0.05, P = 0.88). Hospitalist trainees found the PAGE model more useful after participating (M = 4.62, SD = 0.65), than they had expected (M = 3.92, SD = 0.76; P < 0.05).

All session leaders (n = 15) completed the questionnaire after PAGE (9 hospitalists, 5 geriatricians, 1 urologist). Two‐thirds had 5 years on faculty; eight had no prior experience as a faculty development trainer. Over 80% indicated that they found their coteaching experience, enjoyable, useful and collaborative. Only 1 participant did not commit to interdisciplinary teaching again. Most hospitalist session leaders reported that coteaching with a geriatrician enhanced their knowledge; they were more likely to consult a geriatrician regarding patients. All but 2 session leaders felt that the model fostered a collaborative environment between their 2 divisions.

Of the 56 residents, 41% (16 PGY1, 7 PGY2) completed a pretest; 43% (15 PGY1, 9 PGY2) completed a posttest. Residents reported receiving inpatient teaching on geriatrics skills significantly more frequently from hospitalists vs. nonhospitalist attendings both before PAGE (hospitalists M = 2.18, SD = 0.37; nonhospitalists M = 2.00, SD = 0.53, P < 0.05), and after (hospitalists M = 2.39, SD = 0.46; nonhospitalists M = 2.05, SD = 0.57, P < 0.05; see Fig. 1). Although hospitalists taught more frequently about geriatrics than nonhospitalists before PAGE, our findings suggest that they increased their teaching by a greater magnitude than nonhospitalists (P < 0.01, P > 0.05, respectively). Residents reported increased geriatric skill practice after PAGE with a medium effect size (pretest M = 2.92, SD = 0.55, posttest M = 3.28, SD = 0.66, P = 0.052, d = 0.66). There was greater mean reported practice for all skills with the exception of hospice care, which already was being performed between often and very often before PAGE. The largest increases in skill practice were (descending order, most increased first): assessing polypharmacy, performing skin exams, prognostication, performing functional assessments and examining Foley catheter use.

Figure 1

Discussion

Our aging population and a shortage of geriatricians necessitates new, feasible models for geriatric training. Similar to the CHAMP model,7 PAGE had a favorable impact on faculty perceived behavioral change; after the PAGE sessions, faculty reported significantly greater self‐efficacy of teaching geriatrics. However, this study also examined the impact of the PAGE Model on 2 groups not previously reported in the literature: faculty session leaders and medicine residents.

To our knowledge, this is the first study about a hospitalist TTT program codeveloped with nonhospitalists aimed at teaching geriatrics skills to residents, though smaller scale programs for medical students exist.24 We believe codevelopment was important in our model for many reasons. First, using hospitalist peers and local geriatricians likely increased trust in the educational curricula and allowed for strong communication channels between instructors.25, 26 Second, coteaching allowed for hospitalist mentorship. Hospitalists acknowledged their coleaders as mentors and several hospitalists subsequently engaged in new geriatric projects. Third, coteaching was felt to enhance patient care and increase geriatrician consultations. Coteaching may have applicability to other hospitalist faculty development such as intensive care and palliative care, and hospitalist programs may benefit from creating faculty development programs internally with their colleagues, rather than using online resources.

Another important finding of this study is that training hospitalists to teach about geriatrics seems to result in an increase in both the geriatric teaching that residents receive and residents' practice of geriatric skills. This outcome has not been previously demonstrated with geriatric TTT activities.27 This trickle‐down effect to residents likely results from both the increased teaching efficacy of hospitalists after the PAGE Model and the exportable nature of the teaching tools.

Several continuing medical education best practices were used which we believe contributed to the success of PAGE. First, we conducted a needs assessment, which improves knowledge outcomes.28, 29 Second, sessions included cases, lectures, and discussions. Use of multiple educational techniques yields greater knowledge and behavioral change as compared to a single method, such as lecture alone.24, 25, 30, 31 Finally, sessions were sequenced over a year, rather than clustered in short, intensive activity. Sequenced, or learn‐work‐learn opportunities allow education to be translated to practice and reinforced.8, 27, 30, 32

We believe that the PAGE Model is transportable to other hospitalist programs due to its cost and flexible nature. In economically‐lean times, hospitalist divisions can create a program similar to the PAGE Model essentially at no cost, except for donated faculty preparation time. In contrast, CHAMP was expensive, costing nearly $72,000 for 12 faculty to participate in the 48‐hour curriculum,7, 33 and volunteering physicians were compensated for their time. Though Mini‐CHAMP is a streamlined 2‐day workshop that offers free online lectures and slide sets, there may be some benefit to producing a faculty development program internally, as we stated above, and PAGE included additional topics (urinary catheters and decubitus ulcers/wound care) not covered in mini‐CHAMP.

There were several limitations to this study. First, some outcomes of the PAGE Model were assessed by retrospective self‐report, which may allow for recall bias. Although self‐report may or may not correlate with actual behavior,34 faculty and resident perspectives of their teaching and learning experiences are themselves important. Furthermore, a retrospective presurvey allows for content of an educational program or intervention to be explained prior to a survey, so that participants first assess their new level of understanding or skill on the post test, then reflectively assess the level of understanding or skill they had prior to the workshop. This avoids response shift bias and can improve internal validity.21, 35

Second, the small numbers of session leaders, hospitalist trainees, and residents restricted statistical power to detect small effects. The fact that we found significant improvements enhances the likelihood that the differences observed were not due to chance.

Third, the low response rates from the hospitalist trainee post‐intervention questionnaire and the residents' questionnaires may affect the validity of our results. For the resident survey, the subjects were not matched, and we cannot state that an individual's geriatric skill practice changed due to PAGE, though the results suggest the residency program as a whole improved the frequency of geriatric skill practice.

Finally, the residents were required to report the frequency of teaching on and practice of geriatric skills practice over the prior year and accuracy of recall may be an issue. However, frequencies were queried both pre and post intervention and favorable change was noted. Furthermore, because the high end of the 3‐point teaching scale was limited to more than once, the true amount of teaching may have been underestimated if more than once actually represented high frequencies.

Future studies are needed to replicate these findings at other institutions to confirm generalizability. It would be beneficial to measure patient outcomes to determine whether increased teaching and skill practice benefits patients using measures such as reduction in catheter related urinary tract infections, falls, and inadequate pain management. Further investigations of cotaught faculty development programs between hospitalists and other specialists help emphasize why internally created TTT programs are of greater value than online resources.

Conclusions

This time‐sensitive adaptation of a hospitalist geriatric TTT program was successfully implemented at an academic medical center and suggests improved hospitalist faculty self‐efficacy at teaching geriatric skills, increased frequency of inpatient geriatric teaching by hospitalists and increased resident geriatric skill practice. Confidence to care for geriatric patients and a strong skill set to assess risks and manage them appropriately will equip hospitalists and trainees to provide care that reduces geriatric patients' in‐hospital morbidity and costs of care. As hospitalists increasingly care for older adults, the need for time‐efficient methods of teaching geriatrics will continue to grow. The PAGE Model, and other new models of geriatric training for hospitalists, demonstrates that we are beginning to address this urgent need.

Nearly half of the hospital beds in the United States are occupied by the elderly,1 whose numbers are increasing.2 The odds of a hospitalized Medicare patient being cared for by a hospitalist are increasing by nearly 30% per year.3 Hospitalists require competence in geriatrics to serve their patients and to teach trainees. Train‐the‐Trainer (TTT) programs both educate health care providers and provide educational materials, information, and skills for teaching others.4 This model has been successfully used in geriatrics to impact knowledge, attitudes, and self‐efficacy among health care workers.46

A prominent example of a geriatrics TTT program is the University of Chicago Curriculum for the Hospitalized Aging Medical Patient (CHAMP),7 which requires 48 hours of instruction over 12 sessions. To create a less time‐intensive learning format for busy hospitalists, the University of Chicago developed Mini‐CHAMP, a streamlined 2‐day workshop with web‐based components for hospitalist clinicians, but not necessarily hospitalist educators.7

We created The Donald W. Reynolds Program for Advancing Geriatrics Education (PAGE) at the University of California, San Francisco (UCSF), in light of the time intensity of CHAMP, to integrate geriatric TTT sessions within preexisting hospitalist faculty meetings. This model is consistent with current practices in faculty development.8 This paper describes the evaluation of the PAGE Model, which sought answers to 3 research questions: (1) Does PAGE increase faculty confidence in teaching geriatrics?, (2) Does PAGE increase the frequency of hospitalist teaching geriatrics topics?, and (3) Does PAGE increase residents' practice of geriatrics skills?

Methods

Results

The hospitalist group grew from 31 to 36 members in June of 2008. On average, 14 hospitalists (M = 14.40, standard deviation [SD] = 2.41, range 1119) attended each session, with all hospitalists (n = 36) attending 1 session (M = 3.83, SD = 2.35, range 19). At each session, an average of 72% completed a post‐session evaluation form. Overall, faculty were likely to use the PAGE teaching tools (M = 4.61, SD = 0.53) and would recommend PAGE to other hospitalists (M = 4.63, SD = 0.51).

Thirteen hospitalist trainees of 36 (36%) completed a post‐PAGE online questionnaire. Respondents taught on faculty for an average of 5 years (mean (M) = 5.08, SD = 3.52). Faculty perceived self‐efficacy at teaching residents about geriatrics improved significantly with a large effect size (pretest M = 3.05, SD = .60; posttest M = 3.96, SD = .36, d = 1.52; P < 0.001). Session attendance was positively correlated with the increase in geriatrics teaching self‐efficacy (r = .62, P < 0.05), while teaching experience was not (r = 0.05, P = 0.88). Hospitalist trainees found the PAGE model more useful after participating (M = 4.62, SD = 0.65), than they had expected (M = 3.92, SD = 0.76; P < 0.05).

All session leaders (n = 15) completed the questionnaire after PAGE (9 hospitalists, 5 geriatricians, 1 urologist). Two‐thirds had 5 years on faculty; eight had no prior experience as a faculty development trainer. Over 80% indicated that they found their coteaching experience, enjoyable, useful and collaborative. Only 1 participant did not commit to interdisciplinary teaching again. Most hospitalist session leaders reported that coteaching with a geriatrician enhanced their knowledge; they were more likely to consult a geriatrician regarding patients. All but 2 session leaders felt that the model fostered a collaborative environment between their 2 divisions.

Of the 56 residents, 41% (16 PGY1, 7 PGY2) completed a pretest; 43% (15 PGY1, 9 PGY2) completed a posttest. Residents reported receiving inpatient teaching on geriatrics skills significantly more frequently from hospitalists vs. nonhospitalist attendings both before PAGE (hospitalists M = 2.18, SD = 0.37; nonhospitalists M = 2.00, SD = 0.53, P < 0.05), and after (hospitalists M = 2.39, SD = 0.46; nonhospitalists M = 2.05, SD = 0.57, P < 0.05; see Fig. 1). Although hospitalists taught more frequently about geriatrics than nonhospitalists before PAGE, our findings suggest that they increased their teaching by a greater magnitude than nonhospitalists (P < 0.01, P > 0.05, respectively). Residents reported increased geriatric skill practice after PAGE with a medium effect size (pretest M = 2.92, SD = 0.55, posttest M = 3.28, SD = 0.66, P = 0.052, d = 0.66). There was greater mean reported practice for all skills with the exception of hospice care, which already was being performed between often and very often before PAGE. The largest increases in skill practice were (descending order, most increased first): assessing polypharmacy, performing skin exams, prognostication, performing functional assessments and examining Foley catheter use.

Figure 1

Discussion

Our aging population and a shortage of geriatricians necessitates new, feasible models for geriatric training. Similar to the CHAMP model,7 PAGE had a favorable impact on faculty perceived behavioral change; after the PAGE sessions, faculty reported significantly greater self‐efficacy of teaching geriatrics. However, this study also examined the impact of the PAGE Model on 2 groups not previously reported in the literature: faculty session leaders and medicine residents.

To our knowledge, this is the first study about a hospitalist TTT program codeveloped with nonhospitalists aimed at teaching geriatrics skills to residents, though smaller scale programs for medical students exist.24 We believe codevelopment was important in our model for many reasons. First, using hospitalist peers and local geriatricians likely increased trust in the educational curricula and allowed for strong communication channels between instructors.25, 26 Second, coteaching allowed for hospitalist mentorship. Hospitalists acknowledged their coleaders as mentors and several hospitalists subsequently engaged in new geriatric projects. Third, coteaching was felt to enhance patient care and increase geriatrician consultations. Coteaching may have applicability to other hospitalist faculty development such as intensive care and palliative care, and hospitalist programs may benefit from creating faculty development programs internally with their colleagues, rather than using online resources.

Another important finding of this study is that training hospitalists to teach about geriatrics seems to result in an increase in both the geriatric teaching that residents receive and residents' practice of geriatric skills. This outcome has not been previously demonstrated with geriatric TTT activities.27 This trickle‐down effect to residents likely results from both the increased teaching efficacy of hospitalists after the PAGE Model and the exportable nature of the teaching tools.

Several continuing medical education best practices were used which we believe contributed to the success of PAGE. First, we conducted a needs assessment, which improves knowledge outcomes.28, 29 Second, sessions included cases, lectures, and discussions. Use of multiple educational techniques yields greater knowledge and behavioral change as compared to a single method, such as lecture alone.24, 25, 30, 31 Finally, sessions were sequenced over a year, rather than clustered in short, intensive activity. Sequenced, or learn‐work‐learn opportunities allow education to be translated to practice and reinforced.8, 27, 30, 32

We believe that the PAGE Model is transportable to other hospitalist programs due to its cost and flexible nature. In economically‐lean times, hospitalist divisions can create a program similar to the PAGE Model essentially at no cost, except for donated faculty preparation time. In contrast, CHAMP was expensive, costing nearly $72,000 for 12 faculty to participate in the 48‐hour curriculum,7, 33 and volunteering physicians were compensated for their time. Though Mini‐CHAMP is a streamlined 2‐day workshop that offers free online lectures and slide sets, there may be some benefit to producing a faculty development program internally, as we stated above, and PAGE included additional topics (urinary catheters and decubitus ulcers/wound care) not covered in mini‐CHAMP.

There were several limitations to this study. First, some outcomes of the PAGE Model were assessed by retrospective self‐report, which may allow for recall bias. Although self‐report may or may not correlate with actual behavior,34 faculty and resident perspectives of their teaching and learning experiences are themselves important. Furthermore, a retrospective presurvey allows for content of an educational program or intervention to be explained prior to a survey, so that participants first assess their new level of understanding or skill on the post test, then reflectively assess the level of understanding or skill they had prior to the workshop. This avoids response shift bias and can improve internal validity.21, 35

Second, the small numbers of session leaders, hospitalist trainees, and residents restricted statistical power to detect small effects. The fact that we found significant improvements enhances the likelihood that the differences observed were not due to chance.

Third, the low response rates from the hospitalist trainee post‐intervention questionnaire and the residents' questionnaires may affect the validity of our results. For the resident survey, the subjects were not matched, and we cannot state that an individual's geriatric skill practice changed due to PAGE, though the results suggest the residency program as a whole improved the frequency of geriatric skill practice.

Finally, the residents were required to report the frequency of teaching on and practice of geriatric skills practice over the prior year and accuracy of recall may be an issue. However, frequencies were queried both pre and post intervention and favorable change was noted. Furthermore, because the high end of the 3‐point teaching scale was limited to more than once, the true amount of teaching may have been underestimated if more than once actually represented high frequencies.

Future studies are needed to replicate these findings at other institutions to confirm generalizability. It would be beneficial to measure patient outcomes to determine whether increased teaching and skill practice benefits patients using measures such as reduction in catheter related urinary tract infections, falls, and inadequate pain management. Further investigations of cotaught faculty development programs between hospitalists and other specialists help emphasize why internally created TTT programs are of greater value than online resources.

Conclusions

This time‐sensitive adaptation of a hospitalist geriatric TTT program was successfully implemented at an academic medical center and suggests improved hospitalist faculty self‐efficacy at teaching geriatric skills, increased frequency of inpatient geriatric teaching by hospitalists and increased resident geriatric skill practice. Confidence to care for geriatric patients and a strong skill set to assess risks and manage them appropriately will equip hospitalists and trainees to provide care that reduces geriatric patients' in‐hospital morbidity and costs of care. As hospitalists increasingly care for older adults, the need for time‐efficient methods of teaching geriatrics will continue to grow. The PAGE Model, and other new models of geriatric training for hospitalists, demonstrates that we are beginning to address this urgent need.

Nearly half of the hospital beds in the United States are occupied by the elderly,1 whose numbers are increasing.2 The odds of a hospitalized Medicare patient being cared for by a hospitalist are increasing by nearly 30% per year.3 Hospitalists require competence in geriatrics to serve their patients and to teach trainees. Train‐the‐Trainer (TTT) programs both educate health care providers and provide educational materials, information, and skills for teaching others.4 This model has been successfully used in geriatrics to impact knowledge, attitudes, and self‐efficacy among health care workers.46

A prominent example of a geriatrics TTT program is the University of Chicago Curriculum for the Hospitalized Aging Medical Patient (CHAMP),7 which requires 48 hours of instruction over 12 sessions. To create a less time‐intensive learning format for busy hospitalists, the University of Chicago developed Mini‐CHAMP, a streamlined 2‐day workshop with web‐based components for hospitalist clinicians, but not necessarily hospitalist educators.7

We created The Donald W. Reynolds Program for Advancing Geriatrics Education (PAGE) at the University of California, San Francisco (UCSF), in light of the time intensity of CHAMP, to integrate geriatric TTT sessions within preexisting hospitalist faculty meetings. This model is consistent with current practices in faculty development.8 This paper describes the evaluation of the PAGE Model, which sought answers to 3 research questions: (1) Does PAGE increase faculty confidence in teaching geriatrics?, (2) Does PAGE increase the frequency of hospitalist teaching geriatrics topics?, and (3) Does PAGE increase residents' practice of geriatrics skills?

Methods

Results

The hospitalist group grew from 31 to 36 members in June of 2008. On average, 14 hospitalists (M = 14.40, standard deviation [SD] = 2.41, range 1119) attended each session, with all hospitalists (n = 36) attending 1 session (M = 3.83, SD = 2.35, range 19). At each session, an average of 72% completed a post‐session evaluation form. Overall, faculty were likely to use the PAGE teaching tools (M = 4.61, SD = 0.53) and would recommend PAGE to other hospitalists (M = 4.63, SD = 0.51).

Thirteen hospitalist trainees of 36 (36%) completed a post‐PAGE online questionnaire. Respondents taught on faculty for an average of 5 years (mean (M) = 5.08, SD = 3.52). Faculty perceived self‐efficacy at teaching residents about geriatrics improved significantly with a large effect size (pretest M = 3.05, SD = .60; posttest M = 3.96, SD = .36, d = 1.52; P < 0.001). Session attendance was positively correlated with the increase in geriatrics teaching self‐efficacy (r = .62, P < 0.05), while teaching experience was not (r = 0.05, P = 0.88). Hospitalist trainees found the PAGE model more useful after participating (M = 4.62, SD = 0.65), than they had expected (M = 3.92, SD = 0.76; P < 0.05).

All session leaders (n = 15) completed the questionnaire after PAGE (9 hospitalists, 5 geriatricians, 1 urologist). Two‐thirds had 5 years on faculty; eight had no prior experience as a faculty development trainer. Over 80% indicated that they found their coteaching experience, enjoyable, useful and collaborative. Only 1 participant did not commit to interdisciplinary teaching again. Most hospitalist session leaders reported that coteaching with a geriatrician enhanced their knowledge; they were more likely to consult a geriatrician regarding patients. All but 2 session leaders felt that the model fostered a collaborative environment between their 2 divisions.

Of the 56 residents, 41% (16 PGY1, 7 PGY2) completed a pretest; 43% (15 PGY1, 9 PGY2) completed a posttest. Residents reported receiving inpatient teaching on geriatrics skills significantly more frequently from hospitalists vs. nonhospitalist attendings both before PAGE (hospitalists M = 2.18, SD = 0.37; nonhospitalists M = 2.00, SD = 0.53, P < 0.05), and after (hospitalists M = 2.39, SD = 0.46; nonhospitalists M = 2.05, SD = 0.57, P < 0.05; see Fig. 1). Although hospitalists taught more frequently about geriatrics than nonhospitalists before PAGE, our findings suggest that they increased their teaching by a greater magnitude than nonhospitalists (P < 0.01, P > 0.05, respectively). Residents reported increased geriatric skill practice after PAGE with a medium effect size (pretest M = 2.92, SD = 0.55, posttest M = 3.28, SD = 0.66, P = 0.052, d = 0.66). There was greater mean reported practice for all skills with the exception of hospice care, which already was being performed between often and very often before PAGE. The largest increases in skill practice were (descending order, most increased first): assessing polypharmacy, performing skin exams, prognostication, performing functional assessments and examining Foley catheter use.

Figure 1

Discussion

Our aging population and a shortage of geriatricians necessitates new, feasible models for geriatric training. Similar to the CHAMP model,7 PAGE had a favorable impact on faculty perceived behavioral change; after the PAGE sessions, faculty reported significantly greater self‐efficacy of teaching geriatrics. However, this study also examined the impact of the PAGE Model on 2 groups not previously reported in the literature: faculty session leaders and medicine residents.

To our knowledge, this is the first study about a hospitalist TTT program codeveloped with nonhospitalists aimed at teaching geriatrics skills to residents, though smaller scale programs for medical students exist.24 We believe codevelopment was important in our model for many reasons. First, using hospitalist peers and local geriatricians likely increased trust in the educational curricula and allowed for strong communication channels between instructors.25, 26 Second, coteaching allowed for hospitalist mentorship. Hospitalists acknowledged their coleaders as mentors and several hospitalists subsequently engaged in new geriatric projects. Third, coteaching was felt to enhance patient care and increase geriatrician consultations. Coteaching may have applicability to other hospitalist faculty development such as intensive care and palliative care, and hospitalist programs may benefit from creating faculty development programs internally with their colleagues, rather than using online resources.

Another important finding of this study is that training hospitalists to teach about geriatrics seems to result in an increase in both the geriatric teaching that residents receive and residents' practice of geriatric skills. This outcome has not been previously demonstrated with geriatric TTT activities.27 This trickle‐down effect to residents likely results from both the increased teaching efficacy of hospitalists after the PAGE Model and the exportable nature of the teaching tools.

Several continuing medical education best practices were used which we believe contributed to the success of PAGE. First, we conducted a needs assessment, which improves knowledge outcomes.28, 29 Second, sessions included cases, lectures, and discussions. Use of multiple educational techniques yields greater knowledge and behavioral change as compared to a single method, such as lecture alone.24, 25, 30, 31 Finally, sessions were sequenced over a year, rather than clustered in short, intensive activity. Sequenced, or learn‐work‐learn opportunities allow education to be translated to practice and reinforced.8, 27, 30, 32

We believe that the PAGE Model is transportable to other hospitalist programs due to its cost and flexible nature. In economically‐lean times, hospitalist divisions can create a program similar to the PAGE Model essentially at no cost, except for donated faculty preparation time. In contrast, CHAMP was expensive, costing nearly $72,000 for 12 faculty to participate in the 48‐hour curriculum,7, 33 and volunteering physicians were compensated for their time. Though Mini‐CHAMP is a streamlined 2‐day workshop that offers free online lectures and slide sets, there may be some benefit to producing a faculty development program internally, as we stated above, and PAGE included additional topics (urinary catheters and decubitus ulcers/wound care) not covered in mini‐CHAMP.

There were several limitations to this study. First, some outcomes of the PAGE Model were assessed by retrospective self‐report, which may allow for recall bias. Although self‐report may or may not correlate with actual behavior,34 faculty and resident perspectives of their teaching and learning experiences are themselves important. Furthermore, a retrospective presurvey allows for content of an educational program or intervention to be explained prior to a survey, so that participants first assess their new level of understanding or skill on the post test, then reflectively assess the level of understanding or skill they had prior to the workshop. This avoids response shift bias and can improve internal validity.21, 35

Second, the small numbers of session leaders, hospitalist trainees, and residents restricted statistical power to detect small effects. The fact that we found significant improvements enhances the likelihood that the differences observed were not due to chance.

Third, the low response rates from the hospitalist trainee post‐intervention questionnaire and the residents' questionnaires may affect the validity of our results. For the resident survey, the subjects were not matched, and we cannot state that an individual's geriatric skill practice changed due to PAGE, though the results suggest the residency program as a whole improved the frequency of geriatric skill practice.

Finally, the residents were required to report the frequency of teaching on and practice of geriatric skills practice over the prior year and accuracy of recall may be an issue. However, frequencies were queried both pre and post intervention and favorable change was noted. Furthermore, because the high end of the 3‐point teaching scale was limited to more than once, the true amount of teaching may have been underestimated if more than once actually represented high frequencies.

Future studies are needed to replicate these findings at other institutions to confirm generalizability. It would be beneficial to measure patient outcomes to determine whether increased teaching and skill practice benefits patients using measures such as reduction in catheter related urinary tract infections, falls, and inadequate pain management. Further investigations of cotaught faculty development programs between hospitalists and other specialists help emphasize why internally created TTT programs are of greater value than online resources.

Conclusions

This time‐sensitive adaptation of a hospitalist geriatric TTT program was successfully implemented at an academic medical center and suggests improved hospitalist faculty self‐efficacy at teaching geriatric skills, increased frequency of inpatient geriatric teaching by hospitalists and increased resident geriatric skill practice. Confidence to care for geriatric patients and a strong skill set to assess risks and manage them appropriately will equip hospitalists and trainees to provide care that reduces geriatric patients' in‐hospital morbidity and costs of care. As hospitalists increasingly care for older adults, the need for time‐efficient methods of teaching geriatrics will continue to grow. The PAGE Model, and other new models of geriatric training for hospitalists, demonstrates that we are beginning to address this urgent need.

Promotion of safer healthcare by patient organizations has led to an expansion of studies aimed at understanding medical errors to minimize injury through systemic improvement. These efforts have focused on identifying patient‐related factors, reducing technology failures, and improving communication.1 In contrast, factors related to cognitive errors by healthcare providers have received relatively little attention, although such errors may be an important source of preventable harm.1, 2

Limited information is available on the types and prevalence of cognitive factors in cases of medical injury, although cognitive factors may be a major risk for medical injury. If these factors were confirmed to be important factors for medical injury, better educational strategies may be needed to reduce cognitive errors among physicians and to enhance quality improvement and patient safety. Better understanding of these cognitive factors may also help to implement educational programs aimed at the improvement of cognitive performance in medical schools or teaching hospital.35

Closed‐claim files for cases of medical injury contain valuable information for investigation of the factors involved in medical errors.3 In Japan, court claims were tried and closed orders were issued by judges without a jury system until 2009. Under this system, representatives for defense and plaintiffs can present medical experts. Courts can also appoint experts independent of either party. Court opinions in Japan are considered as neutral judgments for conflicts between plaintiffs and defendants. Usually there are 3 judges who are required to be involved with each judgment in Japanese courts.

Closed‐claim files in cases of medical injury contain information about the types and prevalence of cognitive factors suggested to be causally related to the injuries by verdicts in district courts. Thus, by analyzing these files, an unbiased description of the characteristics and epidemiology of cognitive factors can be obtained for cases of medical injury, with minimization of potentially biased claims indicated by both parties; ie, plaintiffs vs. hospitals. Therefore, in this study, by using information from closed claims files at district courts in Tokyo and Osaka, Japan, we aimed to determine the important cognitive factors associated with cases of medical injury from such factors as judgment, vigilance, memory, technical competence, or knowledge. Since we anticipated that cognitive factors would dominate among the causative factors, we also explored the association of these factors with cases in which a judgment of paid compensation was made.

Methods

Results

In a total of 274 closed cases of medical injury, the mean age of the patients was 49 years old and 45% were women (Table 1). The reviews performed by the coinvestigators were all confirmed by the chief investigator without discordance of the reviews between the coinvestigators and the chief investigator. The claims involved death of patients in 45% of cases; injuries that caused significant or major disability in 10% and 24%, respectively (a total of 34%); and minor adverse outcomes of medical care in 21% (57 cases). Closing verdicts supporting the plaintiffs (patients or family) by the district courts were given in 103 claims (38%), with compensation at a median of 8,000,000 JY (100 JY = $1 US in 2005). The compensation ranged from 20,000 JY to 222,710,251 JY. The highest compensation was ordered to be paid to a 36‐year‐old woman with an obstetrics‐related major injury and the court indicated the injury was causally related to the following 3 cognitive factors: error in judgment, failure of vigilance, and lack of technical competence.

Operative injury was the most frequent reason for claims, followed by delayed diagnosis, medication error, and missed diagnosis. General surgery, orthopedics, internal medicine, and obstetrics/gynecology were the most frequently involved specialties, comprising 30% of all cases (Table 2). The verdicts suggested cognitive factors were the most prevalent factors associated with cases of medical injury: 73% of the injuries were judged to be the result of an error in judgment (Table 3), followed by failure of vigilance (65%), lack of technical competence (34%), and lack of knowledge (31%). Verdicts indicated systemic factors in only a few cases, including poor teamwork in 4% and technology failure in 2%. Patient‐related factors were suggested in 32% of the claims.

In a multivariable‐adjusted logistic regression analysis of cognitive factors with a potential association with the claims with paid compensation (Table 4), only error in judgment showed a significant association (odds ratio, 1.9; 95% confidence interval [CI], 1.01‐3.40). The other four cognitive factors in the model were not associated with these claims. The odds ratio for failure of memory was high (2.8), but this factor was identified by the courts in only 5 cases and was not significantly associated with claims with paid compensation.

Discussion

In this study of closed claims files, we identified 2 important cognitive factors involved in cases of medical injury. Error in judgment was the most common factor, comprising about 70% of all claims, and was significantly associated with cases with paid compensation for medical injury. The second cognitive factor was failure of vigilance, which was found in 65% of the claims. Other cognitive factors, such as lack of technical competence and knowledge or failure of memory, as well as systemic factors (poor teamwork and technology failure) were less frequently found to be causally related to cases with medical injury in the verdicts examined in the study.

Reasons for the low frequency of systemic factors involved in cases of medical injury in our study are unclear. This may be the cultural characteristics such as greater emphasis to working in teams and following rules of an organization in Japan. Another possibility is that plaintiffs might have tended to generate lawsuits in cases with suspected higher frequency of individual physicians' factors in Japan. Moreover, among cognitive factors, lack of technical competence and knowledge or failure of memory was also less frequently related to cases with medical injury in our study compared to those of the previous studies.3, 11

The study design of analyzing closed claims files of cases of medical injury is noteworthy for its methodology of error assessment and provides valuable information on errors related to medical injury.3, 7 Moreover, the system of court verdicts in Japan based on decisions by a professional judge allows elimination of potential bias from stakeholders (plaintiffs vs. hospitals) involved in cases of medical injury. Thus, probable causes related to adverse events can be determined from a neutral position. Previous studies of medical error have focused on medical record reviews, surveys, and interviews;12, 13 our study corroborates and extends the findings in these studies that cognitive errors are the most frequent source of medical injury.

Error in judgment is commonly made in the course of decision making in multiple clinical areas. This type of error is referred to recently as cognitive dispositions to respond,14 which is different from bias or heuristics, since not all heuristics are biased and not all errors in judgments come from bias. There is a well‐established value of heuristics in medical diagnosis. Moreover, the properties of this type of error are likely to be distinct from those associated with performance of procedures (lack of technical competence), such as operative injury, which are directly visible and can be prevented through rapid dissemination of information on safety procedures among a medical team. However, the consequences of error in judgment are important for patients, family, and healthcare providers, and these errors are also largely preventable by implementation of educational programs.15

Possible solutions for improving clinical judgment skills may be derived from recent education theory. The theory provides a means for minimizing errors in judgment through the process of meta‐cognition, in which cognitive forcing strategies can be developed through thinking that involves active control over the process of one's own thinking.14, 15 For example, reflective practice has been suggested to be an important instrument for improving clinical judgment and may particularly improve diagnoses in situations of uncertainty and uniqueness, thereby reducing diagnostic errors.16 The capability of critical reflection in real‐time practice (reflection‐in‐action) and on our own practice (reflection‐on‐action) appears to be a key requirement for developing and maintaining medical expertise.17, 18 For instance, case‐based discussion with clinician educators can be an opportunity for enhancing critical thinking skills of medical trainees.

Based on a context‐based approach that focuses on the nature of the clinical problem, potential systemic solutions have recently been proposed for reducing errors in judgment.1 These solutions utilize advanced technology, including symptom‐oriented diagnostic decision support, internet search engines for information on possible diagnoses, and automated reminders in electronic health records.1, 19 Previous studies have shown that long work hours and sleep deprivation can decrease cognitive function, leading to failure of vigilance and increased medical errors,20 and several systemic solutions provide models for avoidance of failure of vigilance. For instance, eliminating extended work shifts and reducing the number of work hours per week was shown to reduce serious medical errors through increased sleep and decreased failure of vigilance during night work in an intensive care unit.21, 22 Taking a brief nap during work hours has also been associated with decreased medical errors in a recent study conducted in Japan.23 Despite the well‐known importance of factors of physicians' workloads, our study did not analyze these factors and thus further studies are needed to confirm their importance in Japanese medical practice.

There were also 32% of patient‐related factors suggested as contributory factors to medical injury in verdicts of the closed claims. This finding may be also important in planning educational intervention strategies to reduce medical errors. Although our data did not include the relative frequency of components related to these factors, major components of patient‐related factors may include age, severity of illnesses, comorbidity, functional status, or mental status. Educational intervention programs may help healthcare providers to evaluate patients with these risk factors and to implement preventive strategies to avoid incidents among these patients.

General surgery, orthopedic surgery, internal medicine, and obstetrics‐gynecology were the most frequently involved specialties in our study. The reasons why these specialties were highly involved in the claims are unclear and our study could not analyze these issues. However, these specialties may be related to patients with greater clinical severity and thus they may have subsequently higher risk for receiving claims. Further, physicians in these specialties may be at higher risk for having various errors because of the complexity of care for patients.

Our study has several limitations. First, the closed claims are more likely to represent cases with severe injury.3 Therefore, it is unclear if we can generalize our findings beyond cases with severe injury.3 Second, certain contributory factors may not have been suggested by the verdicts, even though they played a role. Among these potential factors, poor teamwork and communication issues are unlikely to be identified as causative in verdicts, unless the allegation of the plaintiffs documented these issues. Moreover, the Japanese courts did not open the medical records to the public and so we could not analyze the medical records of the cases. Third, we only evaluated closed verdicts given by professional judges of district courts, who are unlikely to be medical experts. However, the closed verdicts underwent an extensive process involving testimony from medical professionals and academic societies. Fourth, we, as investigators, had few members with surgical backgrounds in this study so we might have underestimated issues related to technical competence among the claims. Finally, although a small percentage of closed‐ claim cases involving team performance were identified in our study, the plaintiffs might have indicated this point to the court claims, since it might have been difficult to describe this issue as a reason for requesting compensations from defendants. Thus, despite a low proportion of team performance involvement in the verdicts, we still believe that poor team performance is a factor related to most medical injuries.

In summary, causal factors obtained from closed claims files suggest the importance of cognitive factors in cases of medical injury. Among the cognitive factors, error in judgment and failure of vigilance were the most frequent. These findings may help leaders of medical schools and hospitals to allocate more resources for research into strategies to improve cognitive performance and thereby ensure patient safety. Further research is needed to better understand the cognitive mechanisms involved in medical errors and to translate this into educational strategies.

Promotion of safer healthcare by patient organizations has led to an expansion of studies aimed at understanding medical errors to minimize injury through systemic improvement. These efforts have focused on identifying patient‐related factors, reducing technology failures, and improving communication.1 In contrast, factors related to cognitive errors by healthcare providers have received relatively little attention, although such errors may be an important source of preventable harm.1, 2

Limited information is available on the types and prevalence of cognitive factors in cases of medical injury, although cognitive factors may be a major risk for medical injury. If these factors were confirmed to be important factors for medical injury, better educational strategies may be needed to reduce cognitive errors among physicians and to enhance quality improvement and patient safety. Better understanding of these cognitive factors may also help to implement educational programs aimed at the improvement of cognitive performance in medical schools or teaching hospital.35

Closed‐claim files for cases of medical injury contain valuable information for investigation of the factors involved in medical errors.3 In Japan, court claims were tried and closed orders were issued by judges without a jury system until 2009. Under this system, representatives for defense and plaintiffs can present medical experts. Courts can also appoint experts independent of either party. Court opinions in Japan are considered as neutral judgments for conflicts between plaintiffs and defendants. Usually there are 3 judges who are required to be involved with each judgment in Japanese courts.

Closed‐claim files in cases of medical injury contain information about the types and prevalence of cognitive factors suggested to be causally related to the injuries by verdicts in district courts. Thus, by analyzing these files, an unbiased description of the characteristics and epidemiology of cognitive factors can be obtained for cases of medical injury, with minimization of potentially biased claims indicated by both parties; ie, plaintiffs vs. hospitals. Therefore, in this study, by using information from closed claims files at district courts in Tokyo and Osaka, Japan, we aimed to determine the important cognitive factors associated with cases of medical injury from such factors as judgment, vigilance, memory, technical competence, or knowledge. Since we anticipated that cognitive factors would dominate among the causative factors, we also explored the association of these factors with cases in which a judgment of paid compensation was made.

Methods

Results

In a total of 274 closed cases of medical injury, the mean age of the patients was 49 years old and 45% were women (Table 1). The reviews performed by the coinvestigators were all confirmed by the chief investigator without discordance of the reviews between the coinvestigators and the chief investigator. The claims involved death of patients in 45% of cases; injuries that caused significant or major disability in 10% and 24%, respectively (a total of 34%); and minor adverse outcomes of medical care in 21% (57 cases). Closing verdicts supporting the plaintiffs (patients or family) by the district courts were given in 103 claims (38%), with compensation at a median of 8,000,000 JY (100 JY = $1 US in 2005). The compensation ranged from 20,000 JY to 222,710,251 JY. The highest compensation was ordered to be paid to a 36‐year‐old woman with an obstetrics‐related major injury and the court indicated the injury was causally related to the following 3 cognitive factors: error in judgment, failure of vigilance, and lack of technical competence.

Operative injury was the most frequent reason for claims, followed by delayed diagnosis, medication error, and missed diagnosis. General surgery, orthopedics, internal medicine, and obstetrics/gynecology were the most frequently involved specialties, comprising 30% of all cases (Table 2). The verdicts suggested cognitive factors were the most prevalent factors associated with cases of medical injury: 73% of the injuries were judged to be the result of an error in judgment (Table 3), followed by failure of vigilance (65%), lack of technical competence (34%), and lack of knowledge (31%). Verdicts indicated systemic factors in only a few cases, including poor teamwork in 4% and technology failure in 2%. Patient‐related factors were suggested in 32% of the claims.

In a multivariable‐adjusted logistic regression analysis of cognitive factors with a potential association with the claims with paid compensation (Table 4), only error in judgment showed a significant association (odds ratio, 1.9; 95% confidence interval [CI], 1.01‐3.40). The other four cognitive factors in the model were not associated with these claims. The odds ratio for failure of memory was high (2.8), but this factor was identified by the courts in only 5 cases and was not significantly associated with claims with paid compensation.

Discussion

In this study of closed claims files, we identified 2 important cognitive factors involved in cases of medical injury. Error in judgment was the most common factor, comprising about 70% of all claims, and was significantly associated with cases with paid compensation for medical injury. The second cognitive factor was failure of vigilance, which was found in 65% of the claims. Other cognitive factors, such as lack of technical competence and knowledge or failure of memory, as well as systemic factors (poor teamwork and technology failure) were less frequently found to be causally related to cases with medical injury in the verdicts examined in the study.

Reasons for the low frequency of systemic factors involved in cases of medical injury in our study are unclear. This may be the cultural characteristics such as greater emphasis to working in teams and following rules of an organization in Japan. Another possibility is that plaintiffs might have tended to generate lawsuits in cases with suspected higher frequency of individual physicians' factors in Japan. Moreover, among cognitive factors, lack of technical competence and knowledge or failure of memory was also less frequently related to cases with medical injury in our study compared to those of the previous studies.3, 11

The study design of analyzing closed claims files of cases of medical injury is noteworthy for its methodology of error assessment and provides valuable information on errors related to medical injury.3, 7 Moreover, the system of court verdicts in Japan based on decisions by a professional judge allows elimination of potential bias from stakeholders (plaintiffs vs. hospitals) involved in cases of medical injury. Thus, probable causes related to adverse events can be determined from a neutral position. Previous studies of medical error have focused on medical record reviews, surveys, and interviews;12, 13 our study corroborates and extends the findings in these studies that cognitive errors are the most frequent source of medical injury.

Error in judgment is commonly made in the course of decision making in multiple clinical areas. This type of error is referred to recently as cognitive dispositions to respond,14 which is different from bias or heuristics, since not all heuristics are biased and not all errors in judgments come from bias. There is a well‐established value of heuristics in medical diagnosis. Moreover, the properties of this type of error are likely to be distinct from those associated with performance of procedures (lack of technical competence), such as operative injury, which are directly visible and can be prevented through rapid dissemination of information on safety procedures among a medical team. However, the consequences of error in judgment are important for patients, family, and healthcare providers, and these errors are also largely preventable by implementation of educational programs.15

Possible solutions for improving clinical judgment skills may be derived from recent education theory. The theory provides a means for minimizing errors in judgment through the process of meta‐cognition, in which cognitive forcing strategies can be developed through thinking that involves active control over the process of one's own thinking.14, 15 For example, reflective practice has been suggested to be an important instrument for improving clinical judgment and may particularly improve diagnoses in situations of uncertainty and uniqueness, thereby reducing diagnostic errors.16 The capability of critical reflection in real‐time practice (reflection‐in‐action) and on our own practice (reflection‐on‐action) appears to be a key requirement for developing and maintaining medical expertise.17, 18 For instance, case‐based discussion with clinician educators can be an opportunity for enhancing critical thinking skills of medical trainees.

Based on a context‐based approach that focuses on the nature of the clinical problem, potential systemic solutions have recently been proposed for reducing errors in judgment.1 These solutions utilize advanced technology, including symptom‐oriented diagnostic decision support, internet search engines for information on possible diagnoses, and automated reminders in electronic health records.1, 19 Previous studies have shown that long work hours and sleep deprivation can decrease cognitive function, leading to failure of vigilance and increased medical errors,20 and several systemic solutions provide models for avoidance of failure of vigilance. For instance, eliminating extended work shifts and reducing the number of work hours per week was shown to reduce serious medical errors through increased sleep and decreased failure of vigilance during night work in an intensive care unit.21, 22 Taking a brief nap during work hours has also been associated with decreased medical errors in a recent study conducted in Japan.23 Despite the well‐known importance of factors of physicians' workloads, our study did not analyze these factors and thus further studies are needed to confirm their importance in Japanese medical practice.

There were also 32% of patient‐related factors suggested as contributory factors to medical injury in verdicts of the closed claims. This finding may be also important in planning educational intervention strategies to reduce medical errors. Although our data did not include the relative frequency of components related to these factors, major components of patient‐related factors may include age, severity of illnesses, comorbidity, functional status, or mental status. Educational intervention programs may help healthcare providers to evaluate patients with these risk factors and to implement preventive strategies to avoid incidents among these patients.

General surgery, orthopedic surgery, internal medicine, and obstetrics‐gynecology were the most frequently involved specialties in our study. The reasons why these specialties were highly involved in the claims are unclear and our study could not analyze these issues. However, these specialties may be related to patients with greater clinical severity and thus they may have subsequently higher risk for receiving claims. Further, physicians in these specialties may be at higher risk for having various errors because of the complexity of care for patients.

Our study has several limitations. First, the closed claims are more likely to represent cases with severe injury.3 Therefore, it is unclear if we can generalize our findings beyond cases with severe injury.3 Second, certain contributory factors may not have been suggested by the verdicts, even though they played a role. Among these potential factors, poor teamwork and communication issues are unlikely to be identified as causative in verdicts, unless the allegation of the plaintiffs documented these issues. Moreover, the Japanese courts did not open the medical records to the public and so we could not analyze the medical records of the cases. Third, we only evaluated closed verdicts given by professional judges of district courts, who are unlikely to be medical experts. However, the closed verdicts underwent an extensive process involving testimony from medical professionals and academic societies. Fourth, we, as investigators, had few members with surgical backgrounds in this study so we might have underestimated issues related to technical competence among the claims. Finally, although a small percentage of closed‐ claim cases involving team performance were identified in our study, the plaintiffs might have indicated this point to the court claims, since it might have been difficult to describe this issue as a reason for requesting compensations from defendants. Thus, despite a low proportion of team performance involvement in the verdicts, we still believe that poor team performance is a factor related to most medical injuries.

In summary, causal factors obtained from closed claims files suggest the importance of cognitive factors in cases of medical injury. Among the cognitive factors, error in judgment and failure of vigilance were the most frequent. These findings may help leaders of medical schools and hospitals to allocate more resources for research into strategies to improve cognitive performance and thereby ensure patient safety. Further research is needed to better understand the cognitive mechanisms involved in medical errors and to translate this into educational strategies.

Promotion of safer healthcare by patient organizations has led to an expansion of studies aimed at understanding medical errors to minimize injury through systemic improvement. These efforts have focused on identifying patient‐related factors, reducing technology failures, and improving communication.1 In contrast, factors related to cognitive errors by healthcare providers have received relatively little attention, although such errors may be an important source of preventable harm.1, 2

Limited information is available on the types and prevalence of cognitive factors in cases of medical injury, although cognitive factors may be a major risk for medical injury. If these factors were confirmed to be important factors for medical injury, better educational strategies may be needed to reduce cognitive errors among physicians and to enhance quality improvement and patient safety. Better understanding of these cognitive factors may also help to implement educational programs aimed at the improvement of cognitive performance in medical schools or teaching hospital.35

Closed‐claim files for cases of medical injury contain valuable information for investigation of the factors involved in medical errors.3 In Japan, court claims were tried and closed orders were issued by judges without a jury system until 2009. Under this system, representatives for defense and plaintiffs can present medical experts. Courts can also appoint experts independent of either party. Court opinions in Japan are considered as neutral judgments for conflicts between plaintiffs and defendants. Usually there are 3 judges who are required to be involved with each judgment in Japanese courts.

Closed‐claim files in cases of medical injury contain information about the types and prevalence of cognitive factors suggested to be causally related to the injuries by verdicts in district courts. Thus, by analyzing these files, an unbiased description of the characteristics and epidemiology of cognitive factors can be obtained for cases of medical injury, with minimization of potentially biased claims indicated by both parties; ie, plaintiffs vs. hospitals. Therefore, in this study, by using information from closed claims files at district courts in Tokyo and Osaka, Japan, we aimed to determine the important cognitive factors associated with cases of medical injury from such factors as judgment, vigilance, memory, technical competence, or knowledge. Since we anticipated that cognitive factors would dominate among the causative factors, we also explored the association of these factors with cases in which a judgment of paid compensation was made.

Methods

Results

In a total of 274 closed cases of medical injury, the mean age of the patients was 49 years old and 45% were women (Table 1). The reviews performed by the coinvestigators were all confirmed by the chief investigator without discordance of the reviews between the coinvestigators and the chief investigator. The claims involved death of patients in 45% of cases; injuries that caused significant or major disability in 10% and 24%, respectively (a total of 34%); and minor adverse outcomes of medical care in 21% (57 cases). Closing verdicts supporting the plaintiffs (patients or family) by the district courts were given in 103 claims (38%), with compensation at a median of 8,000,000 JY (100 JY = $1 US in 2005). The compensation ranged from 20,000 JY to 222,710,251 JY. The highest compensation was ordered to be paid to a 36‐year‐old woman with an obstetrics‐related major injury and the court indicated the injury was causally related to the following 3 cognitive factors: error in judgment, failure of vigilance, and lack of technical competence.

Operative injury was the most frequent reason for claims, followed by delayed diagnosis, medication error, and missed diagnosis. General surgery, orthopedics, internal medicine, and obstetrics/gynecology were the most frequently involved specialties, comprising 30% of all cases (Table 2). The verdicts suggested cognitive factors were the most prevalent factors associated with cases of medical injury: 73% of the injuries were judged to be the result of an error in judgment (Table 3), followed by failure of vigilance (65%), lack of technical competence (34%), and lack of knowledge (31%). Verdicts indicated systemic factors in only a few cases, including poor teamwork in 4% and technology failure in 2%. Patient‐related factors were suggested in 32% of the claims.

In a multivariable‐adjusted logistic regression analysis of cognitive factors with a potential association with the claims with paid compensation (Table 4), only error in judgment showed a significant association (odds ratio, 1.9; 95% confidence interval [CI], 1.01‐3.40). The other four cognitive factors in the model were not associated with these claims. The odds ratio for failure of memory was high (2.8), but this factor was identified by the courts in only 5 cases and was not significantly associated with claims with paid compensation.

Discussion

In this study of closed claims files, we identified 2 important cognitive factors involved in cases of medical injury. Error in judgment was the most common factor, comprising about 70% of all claims, and was significantly associated with cases with paid compensation for medical injury. The second cognitive factor was failure of vigilance, which was found in 65% of the claims. Other cognitive factors, such as lack of technical competence and knowledge or failure of memory, as well as systemic factors (poor teamwork and technology failure) were less frequently found to be causally related to cases with medical injury in the verdicts examined in the study.

Reasons for the low frequency of systemic factors involved in cases of medical injury in our study are unclear. This may be the cultural characteristics such as greater emphasis to working in teams and following rules of an organization in Japan. Another possibility is that plaintiffs might have tended to generate lawsuits in cases with suspected higher frequency of individual physicians' factors in Japan. Moreover, among cognitive factors, lack of technical competence and knowledge or failure of memory was also less frequently related to cases with medical injury in our study compared to those of the previous studies.3, 11

The study design of analyzing closed claims files of cases of medical injury is noteworthy for its methodology of error assessment and provides valuable information on errors related to medical injury.3, 7 Moreover, the system of court verdicts in Japan based on decisions by a professional judge allows elimination of potential bias from stakeholders (plaintiffs vs. hospitals) involved in cases of medical injury. Thus, probable causes related to adverse events can be determined from a neutral position. Previous studies of medical error have focused on medical record reviews, surveys, and interviews;12, 13 our study corroborates and extends the findings in these studies that cognitive errors are the most frequent source of medical injury.

Error in judgment is commonly made in the course of decision making in multiple clinical areas. This type of error is referred to recently as cognitive dispositions to respond,14 which is different from bias or heuristics, since not all heuristics are biased and not all errors in judgments come from bias. There is a well‐established value of heuristics in medical diagnosis. Moreover, the properties of this type of error are likely to be distinct from those associated with performance of procedures (lack of technical competence), such as operative injury, which are directly visible and can be prevented through rapid dissemination of information on safety procedures among a medical team. However, the consequences of error in judgment are important for patients, family, and healthcare providers, and these errors are also largely preventable by implementation of educational programs.15

Possible solutions for improving clinical judgment skills may be derived from recent education theory. The theory provides a means for minimizing errors in judgment through the process of meta‐cognition, in which cognitive forcing strategies can be developed through thinking that involves active control over the process of one's own thinking.14, 15 For example, reflective practice has been suggested to be an important instrument for improving clinical judgment and may particularly improve diagnoses in situations of uncertainty and uniqueness, thereby reducing diagnostic errors.16 The capability of critical reflection in real‐time practice (reflection‐in‐action) and on our own practice (reflection‐on‐action) appears to be a key requirement for developing and maintaining medical expertise.17, 18 For instance, case‐based discussion with clinician educators can be an opportunity for enhancing critical thinking skills of medical trainees.

Based on a context‐based approach that focuses on the nature of the clinical problem, potential systemic solutions have recently been proposed for reducing errors in judgment.1 These solutions utilize advanced technology, including symptom‐oriented diagnostic decision support, internet search engines for information on possible diagnoses, and automated reminders in electronic health records.1, 19 Previous studies have shown that long work hours and sleep deprivation can decrease cognitive function, leading to failure of vigilance and increased medical errors,20 and several systemic solutions provide models for avoidance of failure of vigilance. For instance, eliminating extended work shifts and reducing the number of work hours per week was shown to reduce serious medical errors through increased sleep and decreased failure of vigilance during night work in an intensive care unit.21, 22 Taking a brief nap during work hours has also been associated with decreased medical errors in a recent study conducted in Japan.23 Despite the well‐known importance of factors of physicians' workloads, our study did not analyze these factors and thus further studies are needed to confirm their importance in Japanese medical practice.

There were also 32% of patient‐related factors suggested as contributory factors to medical injury in verdicts of the closed claims. This finding may be also important in planning educational intervention strategies to reduce medical errors. Although our data did not include the relative frequency of components related to these factors, major components of patient‐related factors may include age, severity of illnesses, comorbidity, functional status, or mental status. Educational intervention programs may help healthcare providers to evaluate patients with these risk factors and to implement preventive strategies to avoid incidents among these patients.

General surgery, orthopedic surgery, internal medicine, and obstetrics‐gynecology were the most frequently involved specialties in our study. The reasons why these specialties were highly involved in the claims are unclear and our study could not analyze these issues. However, these specialties may be related to patients with greater clinical severity and thus they may have subsequently higher risk for receiving claims. Further, physicians in these specialties may be at higher risk for having various errors because of the complexity of care for patients.

Our study has several limitations. First, the closed claims are more likely to represent cases with severe injury.3 Therefore, it is unclear if we can generalize our findings beyond cases with severe injury.3 Second, certain contributory factors may not have been suggested by the verdicts, even though they played a role. Among these potential factors, poor teamwork and communication issues are unlikely to be identified as causative in verdicts, unless the allegation of the plaintiffs documented these issues. Moreover, the Japanese courts did not open the medical records to the public and so we could not analyze the medical records of the cases. Third, we only evaluated closed verdicts given by professional judges of district courts, who are unlikely to be medical experts. However, the closed verdicts underwent an extensive process involving testimony from medical professionals and academic societies. Fourth, we, as investigators, had few members with surgical backgrounds in this study so we might have underestimated issues related to technical competence among the claims. Finally, although a small percentage of closed‐ claim cases involving team performance were identified in our study, the plaintiffs might have indicated this point to the court claims, since it might have been difficult to describe this issue as a reason for requesting compensations from defendants. Thus, despite a low proportion of team performance involvement in the verdicts, we still believe that poor team performance is a factor related to most medical injuries.

In summary, causal factors obtained from closed claims files suggest the importance of cognitive factors in cases of medical injury. Among the cognitive factors, error in judgment and failure of vigilance were the most frequent. These findings may help leaders of medical schools and hospitals to allocate more resources for research into strategies to improve cognitive performance and thereby ensure patient safety. Further research is needed to better understand the cognitive mechanisms involved in medical errors and to translate this into educational strategies.

Why is SIADH Important to Hospitalists?

Disorders of body fluids, and particularly hyponatremia, are among the most commonly encountered problems in clinical medicine, affecting up to 30% of hospitalized patients. In a study of 303,577 laboratory samples collected from 120,137 patients, the prevalence of hyponatremia (serum [Na⁺] <135 mmol/L) on initial presentation to a healthcare provider was 28.2% among those treated in an acute hospital care setting, 21% among those treated in an ambulatory hospital care setting, and 7.2% in community care centers.1 Numerous other studies have corroborated a high prevalence of hyponatremia in hospitalized patients,2 which reflects the increased vulnerability of this patient population to disruptions of body fluid homeostasis. Recognizing the many possible causes of hyponatremia in hospitalized patients and implementing appropriate treatment strategies therefore are critical steps toward optimizing care and improving outcomes in hospitalized patients with hyponatremia.

In addition to its frequency, hyponatremia is also important because it has been associated with worse clinical outcomes across the entire range of inpatient care, from the general hospital population to those treated in the intensive care unit (ICU). In a study of 4123 patients age 65 years or older who were admitted to a community hospital, 3.5% had clinically significant hyponatremia (serum [Na⁺] <130 mmol/L) at admission. Compared with nonhyponatremic patients, those with hyponatremia were twice as likely to die during their hospital stay (relative risk [RR], 1.95; P < 0.05).3 In another study of 2188 patients admitted to a medical ICU over a 5‐year period, 13.7% had hyponatremia. The overall rate of in‐hospital mortality among all ICU patients was high at 37.7%. However, severe hyponatremia (serum [Na⁺] <125 mmol/L) more than doubled the risk of in‐hospital mortality (RR, 2.10; P < 0.001).4 In addition to the general hospital population, in virtually every disease ever studied, the presence of hyponatremia has been found to be an independent risk factor for increased mortality, from congestive heart failure to tuberculosis to liver failure.2

What Causes Hyponatremia in Patients with SIADH?

Hyponatremia can be caused by 1 of 2 potential disruptions in fluid balance: dilution from retained water, or depletion from electrolyte losses in excess of water. Dilutional hyponatremias are associated with either a normal (euvolemic) or an increased (hypervolemic) extracellular fluid (ECF) volume, whereas depletional hyponatremias generally are associated with a decreased ECF volume (hypovolemic). Dilutional hyponatremia can arise from a primary defect in osmoregulation, such as in SIADH, or as a result of ECF volume expansion, as seen in conditions associated with concomitant secondary hyperaldosteronism such as heart failure, hepatic cirrhosis, or nephrotic syndrome. Among some hospitalized patient groups, euvolemic hyponatremia is the most common presentation of abnormally low serum [Na⁺]. In a study of patients who developed clinically significant postoperative hyponatremia (defined as a serum [Na⁺] <130 mmol/L) in a large teaching hospital, only 8% were hypovolemic, whereas 42% were euvolemic and 21% were hypervolemic.5

Euvolemic hyponatremia results from an increase in total body water, but with normal or near‐normal total body sodium. As a result, there is an absence of clinical manifestations of ECF volume expansion, such as subcutaneous edema or ascites. It is important to recognize that although SIADH clearly represents a state of volume expansion due to water retention, it rarely causes clinically recognizable hypervolemia since the retained water is distributed across the intracellular fluid (ICF) as well as the ECF, and because volume regulatory processes act to decrease the actual degree of ECF volume expansion.6 Euvolemic hyponatremia can accompany a wide variety of pathological processes, but the most common cause by far is SIADH. Normally, increased plasma osmolality activates osmoreceptors located in the anterior hypothalamus and stimulates the secretion of arginine vasopressin (AVP), also called antidiuretic hormone (ADH), a key neurohormone that regulates fluid homeostasis. In patients with euvolemic hyponatremia due to SIADH, plasma AVP levels are not suppressed despite normal or decreased plasma osmolality.7 This can be a result of ectopic production of AVP by tumors, or stimulation of endogenous pituitary AVP secretion as a result of nonosmotic stimuli that also stimulate vasopressinergic neurons, which include hypovolemia, hypotension, angiotensin II, nausea, hypoxia, hypercarbia, hypoglycemia, stress, and physical activity. Nonsuppressed AVP levels have been documented in the majority of hyponatremic patients, including those with SIADH8 and heart failure.9

SIADH can develop as the result of many different disease processes that disrupt the normal mechanisms that regulate AVP secretion, including pneumonias and other lung infections, thoracic and extrathoracic tumors, a variety of different central nervous system disorders, the postoperative state, human immunodeficiency virus (HIV), and many different drugs (Figure 1). Given the multiplicity of disorders and drugs that can cause disrupted AVP secretion, it is not surprising that hyponatremia is the most common electrolyte abnormality seen in clinical practice.

Figure 1

What Symptoms are Associated With SIADH?

Symptoms of hyponatremia correlate both with the degree of decrease in the serum [Na⁺] and with the chronicity of the hyponatremia. Acute hyponatremia, defined as <48 hours in duration, is often associated with life‐threatening clinical features such as obtundation, seizures, coma, and respiratory arrest. These symptoms can occur abruptly, sometimes with little warning.10 In the most severe cases, death can occur as a result of cerebral edema with tentorial herniation. Hypoxia secondary to neurogenic pulmonary edema can increase the severity of brain swelling.11

In contrast, chronic hyponatremia is much less symptomatic, and the reason for the profound differences between the symptoms of acute and chronic hyponatremia is now well understood to be due to the process of brain volume regulation.12 It is essential that this process be understood in order to understand the full spectrum of hyponatremic symptoms. As the ECF [Na⁺] decreases, regardless of whether due to a loss of sodium or a gain of water, there is an obligate movement of water into the brain along osmotic gradients. That water shift causes swelling of the brain, or cerebral edema. If the increased brain water reaches approximately 8% in adults, it exceeds the capacity of the skull to accommodate brain expansion, leading to tentorial herniation and death from respiratory arrest and/or ischemic brain damage. However, if the patient survives the initial hyponatremia, a very strong volume regulatory process follows, consisting of loss of electrolytes and small organic molecules called osmolytes from brain cells into brain ECF, and eventually the peripheral ECF.12, 13 As the solute content of the brain decreases, the water content is allowed to normalize, eventually reaching a state in which brain edema is virtually absent, and as a result symptoms are markedly less than with acute hyponatremia. Although the time required for the brain to acieve a volume‐regulated state varies across patients, this process is completed within 48 hours in experinmental animal studies, and probably follows a similar time course in humans.

Despite this powerful adaptation process, chronic hyponatremia is frequently associated with neurological symptomatology, albeit milder and more subtle in nature. A recent report found a fairly high incidence of symptoms in 223 patients with chronic hyponatremia as a result of thiazide administration: 49% had malaise/lethargy, 47% had dizzy spells, 35% had vomiting, 17% had confusion/obtundation, 17% experienced falls, 6% had headaches, and 0.9% had seizures.14 Although dizziness can potentially be attributed to a diuretic‐induced hypovolemia, symptoms such as confusion, obtundation and seizures are more consistent with hyponatremic symptomatology. Because thiazide‐induced hyponatremia can be readily corrected by stopping the thiazide and/or administering sodium, this represents an ideal situation in which to assess improvement in hyponatremia symptomatology with normalization of the serum [Na⁺]; in this study, all of these symptoms improved with correction of the hyponatremia. This represents one of the best examples demonstrating reversal of the symptoms associated with chronic hyponatremia by correction of the hyponatremia, because the patients in this study did not in general have severe underlying comorbidities that might complicate interpretation of their symptoms, as is often the case in patients with SIADH.

What Is Required for Making a Diagnosis of SIADH in Hospitalized Patients?

In patients with hypotonic hypoosmolality, ascertainment of their ECF volume status (ie, hypovolemic, euvolemic, or hypervolemic) is an essential first step, as this will segregate patients into different treatment paradigms. For example, in patients who are truly clinically hypovolemic with a decreased ECF volume by clinical parameters, treatment would generally consist of solute repletion with sodium, generally isotonic saline infusion with or without potassium, until the sodium levels normalize. In patients who are hypervolemic, treatment should focus first on the underlying disease rather than addressing the serum [Na⁺] directly. In patients with clinical euvolemia, the standard diagnostic pathway should be followed to confirm a diagnosis of SIADH as described below.

Assessing ECF volume status can be difficult, even for the most experienced clinicians. Physical signs such as orthostatic decreases in blood pressure and increases in pulse rate, dry mucus membranes, and skin tenting indicate hypovolemic hyponatremia, while signs such as subcutaneous edema, ascites, or anasarca indicate hypervolemic hyponatremia. Patients without any of these findings are generally considered to be euvolemic. However, in any situation these signs are only applicable if there are no other reasons to suspect an altered ECF volume. Along with a complete history and physical examination that includes a careful neurological evaluation, several laboratory tests can help to assess the etiology of the hyponatremia, once serum sodium concentrations have been shown to be below normal ([Na⁺] <135 mmol/L):

• Urine osmolality. A urine osmolality (U_osm) less than 100 mOsm/kg H₂O can indicate low dietary solute intake, primary polydipsia, or a reset osmostat after suppression of AVP release by a decrease in plasma osmolality below the osmotic threshold for AVP secretion, usually as a result of increased water loading.

• Urine sodium concentration. Excretion of sodium, as measured by a spot urine [Na⁺] (U_Na), can indicate depletional hyponatremia if the concentration is less than 30 mmol/L.15 A low U_Na reflects a volume depleted state unless the patient has secondary hyperaldosteronism from heart failure or cirrhosis. Patients with a low U_Na are more likely to respond to isotonic saline. Euvolemic patients who have a normal dietary sodium intake will generally have spot U_Na 30 mmol/L and will not benefit from isotonic saline administration.15 In fact, in SIADH, these patients may respond to isotonic saline with a worsening of hyponatremia, since the sodium from the isotonic saline will be excreted in a concentrated urine while the free water is reabsorbed in the kidney collecting ducts. If the patient is on diuretic therapy, urine sodium values cannot always be accurately interpreted, since a U_Na 30 mmol/L may reflect the natriuretic effect of the diuretic and not a volume replete state.

• Blood tests. Additional indicators of volume status include serum blood nitrogen (BUN) and uric acid levels. A BUN <10 mg/dL and uric acid <4 mg/dL are generally consistent with a euvolemic state, particularly when there is glomerular hyperfiltration, which is often present in SIADH. Elevated serum BUN and uric acid levels (BUN >20 mg/dL and uric acid >6 mg/dL), especially if prior values are available for comparison, can also help to establish whether ineffective vascular volume status may be contributing to the pathophysiology of the hyponatremia. In certain clinical scenarios, the B‐type natriuretic protein (BNP) can be helpful to support a clinical impression of congestive heart failure.

The criteria necessary for a diagnosis of SIADH remain essentially as defined by Bartter and Schwartz16 in 1967 (Table 1), but several points deserve emphasis.17 First, true ECF hypoosmolality must be present and hyponatremia secondary to pseudohyponatremia or hyperglycemia excluded. Second, urinary osmolality must be inappropriate for plasma hypoosmolality (P_osm). This does not require a U_osm>P_osm, but simply that the urine osmolality is greater than maximally dilute (ie, U_osm>100 mOsm/kg H₂O in adults). Furthermore, urine osmolality need not be inappropriately elevated at all levels of P_osm but simply at some level under 275 mOsm/kg H₂O, since in patients with a reset osmostat, AVP secretion can be suppressed at some level of osmolality resulting in maximal urinary dilution and free water excretion at plasma osmolalities below this level.18 Although some consider a reset osmostat to be a separate disorder rather than a variant of SIADH, such cases nonetheless illustrate that some hypoosmolar patients can exhibit an appropriately dilute urine at some, though not all, plasma osmolalities. Third, clinical euvolemia must be present to diagnose SIADH, and this diagnosis cannot be made in a hypovolemic or edematous patient. Importantly, this does not mean that patients with SIADH cannot become hypovolemic for other reasons, but in such cases it is impossible to diagnose the underlying SIADH until the patient is rendered euvolemic. The fourth criterion, renal salt wasting, has probably caused the most confusion in the diagnosis of SIADH. As noted above, the importance of this criterion lies in its usefulness in differentiating hypoosmolality caused by a decreased effective intravascular volume with high aldosterone levels in which case renal Na⁺ conservation occurs, from dilutional disorders in which urine Na⁺ excretion is normal or increased due to ECF volume expansion and a suppressed renin‐angiotensin‐aldosterone system. However, U_Na can also be high in renal causes of solute depletion such as diuretic use or Addison's disease, and conversely patients with SIADH can have a low U_Na if they subsequently become hypovolemic or solute depleted, conditions sometimes produced by imposed salt and water restriction. Consequently, although high urinary Na⁺ excretion is generally the rule in most patients with SIADH, its presence does not necessarily confirm this diagnosis, nor does its absence rule out the diagnosis. The final criterion emphasizes that SIADH remains a diagnosis of exclusion, and the absence of other potential causes of hypoosmolality must always be verified. Glucocorticoid deficiency and SIADH can be especially difficult to distinguish, since both primary and secondary hypocortisolism can cause elevated plasma AVP levels in addition to direct renal effects that prevent maximal urinary dilution.19 Therefore, no patient with chronic hyponatremia should be diagnosed as having SIADH without a thorough evaluation of adrenal function, preferably via a rapid adrenocorticotropic hormone (ACTH) stimulation test. Acute hyponatremia of obvious etiology, such as postoperatively or in association with pneumonitis, may be treated without adrenal testing as long as there are no other clinical signs or symptoms suggestive of adrenal dysfunction.20

Hyponatremia is a particularly common complication in elderly hospitalized patients, increasing in prevalence from approximately 7% in the general older population to 18% to 22% among elderly patients in chronic care facilities.21 Despite the many known causes of SIADH (Figure 1), hyponatremia is often associated with idiopathic SIADH in the elderly population. In a study of 119 nursing home residents aged 60 to 103 years, 53% had at least 1 episode of hyponatremia during the previous 12 months.22 Of these patients, 26% were diagnosed with idiopathic SIADH. In another study of elderly patients with hyponatremia and SIADH, 60% were diagnosed with idiopathic SIADH. Among remaining patients, the 2 main causes identified were pneumonia (9 cases/18%) and medications (6 cases/12%).23 Therefore, more than half of elderly patients who present with hyponatremia due to SIADH may have an idiopathic form, with no detectable underlying treatable disease.

Which Hospital Patients With SIADH are Candidates for Treatment of Hyponatremia?

Correction of hyponatremia is associated with markedly improved neurological outcomes in patients with severely symptomatic hyponatremia. In a retrospective review of patients who presented with severe neurological symptoms and serum [Na⁺] <125 mmol/L, prompt therapy with isotonic or hypertonic saline resulted in a correction in the range of 20 mmol/L over several days and neurological recovery in almost all cases. In contrast, in patients who were treated with fluid restriction alone, there was very little correction over the study period (<5 mmol/L over 72 hours), and the neurological outcomes were much worse, with most of these patients either dying or entering a persistently vegetative state.24 Consequently, prompt therapy to rapidly increase the serum [Na⁺] represents the standard‐of‐care for treatment of patients presenting with severe life‐threatening symptoms of hyponatremia.

As discussed earlier, chronic hyponatremia is much less symptomatic as a result of the process of brain volume regulation. Because of this adaptation process, chronic hyponatremia is arguably a condition that clinicians feel they may not need to be as concerned about, and in some publications this has been called asymptomatic hyponatremia. However, such patients often do have neurological symptoms, even if milder and more subtle in nature, including headaches, nausea, mood disturbances, depression, difficulty concentrating, slowed reaction times, unstable gait, increased falls, confusion, and disorientation. Consequently, any patient with hyponatremia secondary to SIADH who manifests any neurological symptoms that could be related to the hyponatremia should be considered as appropriate candidates for treatment of the hyponatremia, regardless of the chronicity of the hyponatremia or the level of serum [Na⁺].

What Therapies are Currently Available to Manage SIADH in Hospitalized Patients?

Conventional management strategies for euvolemic hyponatremia range from saline infusion and fluid restriction to pharmacologic adjustment of fluid balance. Consideration of treatment options should include an evaluation of the benefits as well as the potential toxicities of any therapy (Table 2). Sometimes, simply stopping treatment with an agent that is associated with hyponatremia is sufficient to reverse a low serum [Na⁺].

Hyponatremia Treatment Guidelines for Hospitalized Patients With SIADH

Although various authors have published recommendations on the treatment of hyponatremia,27, 43, 45‐47 no standardized treatment algorithms have yet been widely accepted. A synthesis of existing expert recommendations for treatment of hyponatremia is illustrated in Figure 2. This algorithm is based primarily on the symptomatology of hyponatremic patients, rather than the serum [Na⁺] or the chronicity of the hyponatremia, which are often difficult to ascertain. A careful neurological history and assessment should always be done to identify potential causes for the patient's symptoms other than hyponatremia, although it will not always be possible to exclude an additive contribution from the hyponatremia to an underlying neurological condition. In this model, patients are divided into three groups based on their presenting symptoms.

Figure 2

Level 1 symptoms include seizures, coma, respiratory arrest, obtundation, and vomiting, and usually imply a more acute onset or worsening of hyponatremia requiring immediate active treatment. Therapies that will quickly raise serum sodium levels are required to reduce cerebral edema and decrease the risk of potentially fatal herniation.

Level 2 symptoms, which are more moderate, include nausea, confusion, disorientation, and altered mental status. These symptoms may be either chronic or acute, but allow time to elaborate a more deliberate approach to treatment.

Level 3 symptoms range from minimal symptoms such as a headache, irritability, inability to concentrate, altered mood, and depression, to a virtual absence of discernable symptoms, and indicate that the patient may have chronic or slowly evolving hyponatremia. These symptoms necessitate a cautious approach, especially when patients have underlying comorbidities.

Patients with severe symptoms (Level 1) should be treated with hypertonic saline as first‐line therapy, followed by fluid restriction with or without AVPR antagonist therapy. Patients with moderate symptoms will benefit from a more aggressive regimen of vaptan therapy or limited hypertonic saline administration, followed by fluid restriction or long‐term vaptan therapy. Although moderate neurological symptoms can indicate that a patient is in an early stage of acute hyponatremia, they more often indicate a chronically hyponatremic state with sufficient brain volume adaptation to prevent marked symptomatology from cerebral edema. Regardless, close monitoring of these patients in a hospital setting is warranted until the symptoms improve or stabilize. Patients with no or minimal symptoms should be managed initially with fluid restriction, although treatment with vaptans may be appropriate for a wide range of specific clinical conditions, foremost of which is a failure to improve the serum [Na⁺] despite reasonable attempts at fluid restriction (Figure 2).

Although this classification is based on presenting symptoms at the time of initial evaluation, it should be remembered that in some cases patients initially exhibit more moderate symptoms because they are in the early stages of hyponatremia. In addition, some patients with minimal symptoms are prone to develop more symptomatic hyponatremia during periods of increased fluid ingestion. In support of this, approximately 70% of 31 patients presenting to a university hospital with symptomatic hyponatremia and a mean serum [Na⁺] of 119 mmol/L had preexisting asymptomatic hyponatremia as the most common risk factor identified.48 Consequently, therapy of hyponatremia should also be considered to prevent progression from lower to higher levels of symptomatic hyponatremia, particularly in patients with a past history of repeated presentations of symptomatic hyponatremia.

How Often Should Serum [Na+] Be Monitored in Hospitalized Patients With SIADH?

The frequency of serum [Na⁺] monitoring is dependent on both the severity of the hyponatremia and the therapy chosen. In all hyponatremic patients neurological symptomatology should be carefully assessed very early in the diagnostic evaluation to assess the symptomatic severity of the hyponatremia and to determine whether the patient requires more urgent therapy. All patients undergoing active treatment with hypertonic saline for level 1 or 2 symptomatic hyponatremia should have frequent monitoring of serum [Na⁺] and ECF volume status (every 2‐4 hours) to ensure that the serum [Na⁺] does not exceeded the recommended levels during the active phase of correction,43 since overly rapid correction of serum sodium can cause damage to the myelin sheath of nerve cells, resulting in central pontine myelinolysis, also called the ODS.31 Patients treated with vaptans for level 2 or 3 symptoms should have serum [Na⁺] monitored every 6 to 8 hours during the active phase of correction, which will generally be the first 24 to 48 hours of therapy. Active treatment with hypertonic saline or vaptans should be stopped when the patient's symptoms are no longer present, a safe serum [Na⁺] (usually >120 mmol/L) has been achieved, or the rate of correction has reached 12 mmol/L within 24 hours or 18 mmol/L within 48 hours.29, 43 Importantly, ODS has not been reported either in clinical trials or with therapeutic use of any vaptan to date. In patients with a stable level of serum [Na⁺] treated with fluid restriction or therapies other than hypertonic saline, measurement of serum [Na⁺] daily is generally sufficient, since levels will not change that quickly in the absence of active therapy or large changes in fluid intake or administration.

Potential Future Indications for Treatment of Hyponatremia

Correction of hyponatremia improves other symptoms, such as gait stability, in patients who may be considered to be asymptomatic by virtue of a normal neurological exam. In 1 study, 16 patients with hyponatremia secondary to SIADH in the range of 124 mmol/L to 130 mmol/L demonstrated a significant gait instability that normalized after correction of the hyponatremia to normal ranges.49 The functional significance of the gait instability was illustrated in a study of 122 Belgian patients with a variety of levels of hyponatremia, all judged to be asymptomatic at the time of visit to an emergency department (ED). These patients were compared with 244 age‐matched, sex‐matched, and disease‐matched controls also presenting to the ED during the same time period. Researchers found that 21% of the hyponatremic patients presented to the ED because of a recent fall, compared to only 5% of the controls, resulting in an adjusted odds ratio (OR) for presenting to the ED because of a recent fall of 67 for hyponatremia (P < 0.001).49 Consequently, this study clearly documented an increased incidence of falls in so‐called asymptomatic hyponatremic patients.

The clinical significance of the gait instability and fall data were further evaluated in a study that compared 553 patients with fractures to an equal number of age‐matched and sex‐matched controls. Hyponatremia was found in 13% of the patients presenting with fractures compared to only 4% of the controls.50 Similar findings have been reported in a 364 elderly patients with large‐bone fractures in New York.51 More recently published studies have shown that hyponatremia is associated with increased bone loss in experimental animals and a significant increased OR for osteoporosis of the femoral neck (OR, 2.87; P < 0.003) in humans over the age of 50 in the NHANES III database.52 Thus, the major clinical significance of chronic hyponatremia may be the increased morbidity and mortality associated with falls and fractures in our elderly population.

Summary

Disorders of sodium and water metabolism are commonly encountered in the hospital setting due to the wide range of disease states that can disrupt the balanced control of water and solute intake and output. In particular, the prompt identification and appropriate management of abnormally low serum [Na⁺] is critical if we are to reduce the increased morbidity and mortality that accompany hyponatremia in hospitalized patients. Use of an algorithm that is based primarily on the symptomatology of hyponatremic patients, rather than the chronicity of the hyponatremia or the serum [Na⁺], will help to choose the correct initial therapy in hospitalized hyponatremic patients. However, careful monitoring of serum [Na⁺] responses is required in all cases to adjust therapy appropriately in response to changing clinical conditions. Although this approach will enable efficacious and safe treatment of hyponatremic patients with SIADH at the present time, evolving knowledge of the consequences of chronic hyponatremia will likely alter treatment indications and guidelines in the future.

Why is SIADH Important to Hospitalists?

Disorders of body fluids, and particularly hyponatremia, are among the most commonly encountered problems in clinical medicine, affecting up to 30% of hospitalized patients. In a study of 303,577 laboratory samples collected from 120,137 patients, the prevalence of hyponatremia (serum [Na⁺] <135 mmol/L) on initial presentation to a healthcare provider was 28.2% among those treated in an acute hospital care setting, 21% among those treated in an ambulatory hospital care setting, and 7.2% in community care centers.1 Numerous other studies have corroborated a high prevalence of hyponatremia in hospitalized patients,2 which reflects the increased vulnerability of this patient population to disruptions of body fluid homeostasis. Recognizing the many possible causes of hyponatremia in hospitalized patients and implementing appropriate treatment strategies therefore are critical steps toward optimizing care and improving outcomes in hospitalized patients with hyponatremia.

In addition to its frequency, hyponatremia is also important because it has been associated with worse clinical outcomes across the entire range of inpatient care, from the general hospital population to those treated in the intensive care unit (ICU). In a study of 4123 patients age 65 years or older who were admitted to a community hospital, 3.5% had clinically significant hyponatremia (serum [Na⁺] <130 mmol/L) at admission. Compared with nonhyponatremic patients, those with hyponatremia were twice as likely to die during their hospital stay (relative risk [RR], 1.95; P < 0.05).3 In another study of 2188 patients admitted to a medical ICU over a 5‐year period, 13.7% had hyponatremia. The overall rate of in‐hospital mortality among all ICU patients was high at 37.7%. However, severe hyponatremia (serum [Na⁺] <125 mmol/L) more than doubled the risk of in‐hospital mortality (RR, 2.10; P < 0.001).4 In addition to the general hospital population, in virtually every disease ever studied, the presence of hyponatremia has been found to be an independent risk factor for increased mortality, from congestive heart failure to tuberculosis to liver failure.2

What Causes Hyponatremia in Patients with SIADH?

Hyponatremia can be caused by 1 of 2 potential disruptions in fluid balance: dilution from retained water, or depletion from electrolyte losses in excess of water. Dilutional hyponatremias are associated with either a normal (euvolemic) or an increased (hypervolemic) extracellular fluid (ECF) volume, whereas depletional hyponatremias generally are associated with a decreased ECF volume (hypovolemic). Dilutional hyponatremia can arise from a primary defect in osmoregulation, such as in SIADH, or as a result of ECF volume expansion, as seen in conditions associated with concomitant secondary hyperaldosteronism such as heart failure, hepatic cirrhosis, or nephrotic syndrome. Among some hospitalized patient groups, euvolemic hyponatremia is the most common presentation of abnormally low serum [Na⁺]. In a study of patients who developed clinically significant postoperative hyponatremia (defined as a serum [Na⁺] <130 mmol/L) in a large teaching hospital, only 8% were hypovolemic, whereas 42% were euvolemic and 21% were hypervolemic.5

Euvolemic hyponatremia results from an increase in total body water, but with normal or near‐normal total body sodium. As a result, there is an absence of clinical manifestations of ECF volume expansion, such as subcutaneous edema or ascites. It is important to recognize that although SIADH clearly represents a state of volume expansion due to water retention, it rarely causes clinically recognizable hypervolemia since the retained water is distributed across the intracellular fluid (ICF) as well as the ECF, and because volume regulatory processes act to decrease the actual degree of ECF volume expansion.6 Euvolemic hyponatremia can accompany a wide variety of pathological processes, but the most common cause by far is SIADH. Normally, increased plasma osmolality activates osmoreceptors located in the anterior hypothalamus and stimulates the secretion of arginine vasopressin (AVP), also called antidiuretic hormone (ADH), a key neurohormone that regulates fluid homeostasis. In patients with euvolemic hyponatremia due to SIADH, plasma AVP levels are not suppressed despite normal or decreased plasma osmolality.7 This can be a result of ectopic production of AVP by tumors, or stimulation of endogenous pituitary AVP secretion as a result of nonosmotic stimuli that also stimulate vasopressinergic neurons, which include hypovolemia, hypotension, angiotensin II, nausea, hypoxia, hypercarbia, hypoglycemia, stress, and physical activity. Nonsuppressed AVP levels have been documented in the majority of hyponatremic patients, including those with SIADH8 and heart failure.9

SIADH can develop as the result of many different disease processes that disrupt the normal mechanisms that regulate AVP secretion, including pneumonias and other lung infections, thoracic and extrathoracic tumors, a variety of different central nervous system disorders, the postoperative state, human immunodeficiency virus (HIV), and many different drugs (Figure 1). Given the multiplicity of disorders and drugs that can cause disrupted AVP secretion, it is not surprising that hyponatremia is the most common electrolyte abnormality seen in clinical practice.

Figure 1

What Symptoms are Associated With SIADH?

Symptoms of hyponatremia correlate both with the degree of decrease in the serum [Na⁺] and with the chronicity of the hyponatremia. Acute hyponatremia, defined as <48 hours in duration, is often associated with life‐threatening clinical features such as obtundation, seizures, coma, and respiratory arrest. These symptoms can occur abruptly, sometimes with little warning.10 In the most severe cases, death can occur as a result of cerebral edema with tentorial herniation. Hypoxia secondary to neurogenic pulmonary edema can increase the severity of brain swelling.11

In contrast, chronic hyponatremia is much less symptomatic, and the reason for the profound differences between the symptoms of acute and chronic hyponatremia is now well understood to be due to the process of brain volume regulation.12 It is essential that this process be understood in order to understand the full spectrum of hyponatremic symptoms. As the ECF [Na⁺] decreases, regardless of whether due to a loss of sodium or a gain of water, there is an obligate movement of water into the brain along osmotic gradients. That water shift causes swelling of the brain, or cerebral edema. If the increased brain water reaches approximately 8% in adults, it exceeds the capacity of the skull to accommodate brain expansion, leading to tentorial herniation and death from respiratory arrest and/or ischemic brain damage. However, if the patient survives the initial hyponatremia, a very strong volume regulatory process follows, consisting of loss of electrolytes and small organic molecules called osmolytes from brain cells into brain ECF, and eventually the peripheral ECF.12, 13 As the solute content of the brain decreases, the water content is allowed to normalize, eventually reaching a state in which brain edema is virtually absent, and as a result symptoms are markedly less than with acute hyponatremia. Although the time required for the brain to acieve a volume‐regulated state varies across patients, this process is completed within 48 hours in experinmental animal studies, and probably follows a similar time course in humans.

Despite this powerful adaptation process, chronic hyponatremia is frequently associated with neurological symptomatology, albeit milder and more subtle in nature. A recent report found a fairly high incidence of symptoms in 223 patients with chronic hyponatremia as a result of thiazide administration: 49% had malaise/lethargy, 47% had dizzy spells, 35% had vomiting, 17% had confusion/obtundation, 17% experienced falls, 6% had headaches, and 0.9% had seizures.14 Although dizziness can potentially be attributed to a diuretic‐induced hypovolemia, symptoms such as confusion, obtundation and seizures are more consistent with hyponatremic symptomatology. Because thiazide‐induced hyponatremia can be readily corrected by stopping the thiazide and/or administering sodium, this represents an ideal situation in which to assess improvement in hyponatremia symptomatology with normalization of the serum [Na⁺]; in this study, all of these symptoms improved with correction of the hyponatremia. This represents one of the best examples demonstrating reversal of the symptoms associated with chronic hyponatremia by correction of the hyponatremia, because the patients in this study did not in general have severe underlying comorbidities that might complicate interpretation of their symptoms, as is often the case in patients with SIADH.

What Is Required for Making a Diagnosis of SIADH in Hospitalized Patients?

In patients with hypotonic hypoosmolality, ascertainment of their ECF volume status (ie, hypovolemic, euvolemic, or hypervolemic) is an essential first step, as this will segregate patients into different treatment paradigms. For example, in patients who are truly clinically hypovolemic with a decreased ECF volume by clinical parameters, treatment would generally consist of solute repletion with sodium, generally isotonic saline infusion with or without potassium, until the sodium levels normalize. In patients who are hypervolemic, treatment should focus first on the underlying disease rather than addressing the serum [Na⁺] directly. In patients with clinical euvolemia, the standard diagnostic pathway should be followed to confirm a diagnosis of SIADH as described below.

Assessing ECF volume status can be difficult, even for the most experienced clinicians. Physical signs such as orthostatic decreases in blood pressure and increases in pulse rate, dry mucus membranes, and skin tenting indicate hypovolemic hyponatremia, while signs such as subcutaneous edema, ascites, or anasarca indicate hypervolemic hyponatremia. Patients without any of these findings are generally considered to be euvolemic. However, in any situation these signs are only applicable if there are no other reasons to suspect an altered ECF volume. Along with a complete history and physical examination that includes a careful neurological evaluation, several laboratory tests can help to assess the etiology of the hyponatremia, once serum sodium concentrations have been shown to be below normal ([Na⁺] <135 mmol/L):

• Urine osmolality. A urine osmolality (U_osm) less than 100 mOsm/kg H₂O can indicate low dietary solute intake, primary polydipsia, or a reset osmostat after suppression of AVP release by a decrease in plasma osmolality below the osmotic threshold for AVP secretion, usually as a result of increased water loading.

• Urine sodium concentration. Excretion of sodium, as measured by a spot urine [Na⁺] (U_Na), can indicate depletional hyponatremia if the concentration is less than 30 mmol/L.15 A low U_Na reflects a volume depleted state unless the patient has secondary hyperaldosteronism from heart failure or cirrhosis. Patients with a low U_Na are more likely to respond to isotonic saline. Euvolemic patients who have a normal dietary sodium intake will generally have spot U_Na 30 mmol/L and will not benefit from isotonic saline administration.15 In fact, in SIADH, these patients may respond to isotonic saline with a worsening of hyponatremia, since the sodium from the isotonic saline will be excreted in a concentrated urine while the free water is reabsorbed in the kidney collecting ducts. If the patient is on diuretic therapy, urine sodium values cannot always be accurately interpreted, since a U_Na 30 mmol/L may reflect the natriuretic effect of the diuretic and not a volume replete state.

• Blood tests. Additional indicators of volume status include serum blood nitrogen (BUN) and uric acid levels. A BUN <10 mg/dL and uric acid <4 mg/dL are generally consistent with a euvolemic state, particularly when there is glomerular hyperfiltration, which is often present in SIADH. Elevated serum BUN and uric acid levels (BUN >20 mg/dL and uric acid >6 mg/dL), especially if prior values are available for comparison, can also help to establish whether ineffective vascular volume status may be contributing to the pathophysiology of the hyponatremia. In certain clinical scenarios, the B‐type natriuretic protein (BNP) can be helpful to support a clinical impression of congestive heart failure.

The criteria necessary for a diagnosis of SIADH remain essentially as defined by Bartter and Schwartz16 in 1967 (Table 1), but several points deserve emphasis.17 First, true ECF hypoosmolality must be present and hyponatremia secondary to pseudohyponatremia or hyperglycemia excluded. Second, urinary osmolality must be inappropriate for plasma hypoosmolality (P_osm). This does not require a U_osm>P_osm, but simply that the urine osmolality is greater than maximally dilute (ie, U_osm>100 mOsm/kg H₂O in adults). Furthermore, urine osmolality need not be inappropriately elevated at all levels of P_osm but simply at some level under 275 mOsm/kg H₂O, since in patients with a reset osmostat, AVP secretion can be suppressed at some level of osmolality resulting in maximal urinary dilution and free water excretion at plasma osmolalities below this level.18 Although some consider a reset osmostat to be a separate disorder rather than a variant of SIADH, such cases nonetheless illustrate that some hypoosmolar patients can exhibit an appropriately dilute urine at some, though not all, plasma osmolalities. Third, clinical euvolemia must be present to diagnose SIADH, and this diagnosis cannot be made in a hypovolemic or edematous patient. Importantly, this does not mean that patients with SIADH cannot become hypovolemic for other reasons, but in such cases it is impossible to diagnose the underlying SIADH until the patient is rendered euvolemic. The fourth criterion, renal salt wasting, has probably caused the most confusion in the diagnosis of SIADH. As noted above, the importance of this criterion lies in its usefulness in differentiating hypoosmolality caused by a decreased effective intravascular volume with high aldosterone levels in which case renal Na⁺ conservation occurs, from dilutional disorders in which urine Na⁺ excretion is normal or increased due to ECF volume expansion and a suppressed renin‐angiotensin‐aldosterone system. However, U_Na can also be high in renal causes of solute depletion such as diuretic use or Addison's disease, and conversely patients with SIADH can have a low U_Na if they subsequently become hypovolemic or solute depleted, conditions sometimes produced by imposed salt and water restriction. Consequently, although high urinary Na⁺ excretion is generally the rule in most patients with SIADH, its presence does not necessarily confirm this diagnosis, nor does its absence rule out the diagnosis. The final criterion emphasizes that SIADH remains a diagnosis of exclusion, and the absence of other potential causes of hypoosmolality must always be verified. Glucocorticoid deficiency and SIADH can be especially difficult to distinguish, since both primary and secondary hypocortisolism can cause elevated plasma AVP levels in addition to direct renal effects that prevent maximal urinary dilution.19 Therefore, no patient with chronic hyponatremia should be diagnosed as having SIADH without a thorough evaluation of adrenal function, preferably via a rapid adrenocorticotropic hormone (ACTH) stimulation test. Acute hyponatremia of obvious etiology, such as postoperatively or in association with pneumonitis, may be treated without adrenal testing as long as there are no other clinical signs or symptoms suggestive of adrenal dysfunction.20

Hyponatremia is a particularly common complication in elderly hospitalized patients, increasing in prevalence from approximately 7% in the general older population to 18% to 22% among elderly patients in chronic care facilities.21 Despite the many known causes of SIADH (Figure 1), hyponatremia is often associated with idiopathic SIADH in the elderly population. In a study of 119 nursing home residents aged 60 to 103 years, 53% had at least 1 episode of hyponatremia during the previous 12 months.22 Of these patients, 26% were diagnosed with idiopathic SIADH. In another study of elderly patients with hyponatremia and SIADH, 60% were diagnosed with idiopathic SIADH. Among remaining patients, the 2 main causes identified were pneumonia (9 cases/18%) and medications (6 cases/12%).23 Therefore, more than half of elderly patients who present with hyponatremia due to SIADH may have an idiopathic form, with no detectable underlying treatable disease.

Which Hospital Patients With SIADH are Candidates for Treatment of Hyponatremia?

Correction of hyponatremia is associated with markedly improved neurological outcomes in patients with severely symptomatic hyponatremia. In a retrospective review of patients who presented with severe neurological symptoms and serum [Na⁺] <125 mmol/L, prompt therapy with isotonic or hypertonic saline resulted in a correction in the range of 20 mmol/L over several days and neurological recovery in almost all cases. In contrast, in patients who were treated with fluid restriction alone, there was very little correction over the study period (<5 mmol/L over 72 hours), and the neurological outcomes were much worse, with most of these patients either dying or entering a persistently vegetative state.24 Consequently, prompt therapy to rapidly increase the serum [Na⁺] represents the standard‐of‐care for treatment of patients presenting with severe life‐threatening symptoms of hyponatremia.

As discussed earlier, chronic hyponatremia is much less symptomatic as a result of the process of brain volume regulation. Because of this adaptation process, chronic hyponatremia is arguably a condition that clinicians feel they may not need to be as concerned about, and in some publications this has been called asymptomatic hyponatremia. However, such patients often do have neurological symptoms, even if milder and more subtle in nature, including headaches, nausea, mood disturbances, depression, difficulty concentrating, slowed reaction times, unstable gait, increased falls, confusion, and disorientation. Consequently, any patient with hyponatremia secondary to SIADH who manifests any neurological symptoms that could be related to the hyponatremia should be considered as appropriate candidates for treatment of the hyponatremia, regardless of the chronicity of the hyponatremia or the level of serum [Na⁺].

What Therapies are Currently Available to Manage SIADH in Hospitalized Patients?

Conventional management strategies for euvolemic hyponatremia range from saline infusion and fluid restriction to pharmacologic adjustment of fluid balance. Consideration of treatment options should include an evaluation of the benefits as well as the potential toxicities of any therapy (Table 2). Sometimes, simply stopping treatment with an agent that is associated with hyponatremia is sufficient to reverse a low serum [Na⁺].