iScore for Predicting Institutional Care after Ischemic Stroke: A Population-Based Study Yannick Bejot, MD, PhD, Benoit Daubail, MD, Benedicte Sensenbrenner, MD, ^me Durier, MSc, and Maurice Giroud, MD, PhD Nicolas Legris, MD, Jero
Background: We assessed whether the iScore could predict the need for poststroke institutional care. Methods: Patients with acute ischemic stroke living in Dijon, France, were recorded between 2006 and 2011, using a population-based stroke registry. The iScore was calculated for each patient. A logistic regression model was used to assess the performance of the iScore for predicting the need for placement in a care institution. The discrimination and calibration of the model were assessed using the c statistic and the Hosmer–Lemeshow goodness-of-fit test, respectively. Results: Of the 1199 patients recorded, 124 were excluded because of early death and 95 because of missing for variables included in the iScore. Of the remaining 980 patients, 522 (53.3%) returned home and 458 (46.7%) required placement in a care institution. The median iScore was 123 (interquartile range, 97-148), and the proportion of patients who required placement in a care institution increased with each quintile of risk score. The discrimination of the model was good with a c statistic of .75 (95% confidence interval, .72-.78), as was calibration (P 5 .35). Conclusions: The iScore could be useful for predicting the need for placement in a care institution in ischemic stroke patients. Further studies are required to confirm this finding. Key Words: Stroke—stroke outcome—epidemiology—stroke registry— predictors—discharge planning. Ó 2015 by National Stroke Association
Introduction Recent improvements in the management of acute ischemic stroke have led to increasing the number of patients who survive after the event, especially, elderly people because of the progressive aging population.
From the Dijon Stroke Registry, EA4184, Department of Neurology, University Hospital and Medical School of Dijon, University of Burgundy, Dijon, France. Received August 26, 2014; revision received October 31, 2014; accepted November 14, 2014. The Dijon Stroke Registry is supported by the French Institute for Public Health Surveillance (InVS) and Inserm. The authors have nothing to disclose. Address correspondence to Yannick Bejot, MD, PhD, Service de Neurologie, Bocage Central, 14 rue Gaffarel, BP77908, 21079 Dijon cedex, France. E-mail: [email protected]. 1052-3057/$ - see front matter Ó 2015 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.11.010
694
Among stroke survivors, 45%-75% are unable to return home at once and need to be discharged to either a rehabilitation facility or a nursing home, with consequences in terms of both quality of life and economic costs for the society.1-7 From a health care organization point of view, better identifying the need for poststroke institutional care is a difficult but great challenge so as to adjust necessary resources in terms of number of beds and health care professionals, to estimate costs, and to make projection on future needs. Although several predictors of not being discharged to home have been identified including age, stroke severity, and cognitive impairment,1,4,7–9 no reliable score to predict poststroke care requirements is available currently. The iScore system has been demonstrated to be useful for predicting 30day mortality, the effectiveness of thrombolytic therapy, and functional outcome, when applied to ischemic stroke patients.10-17 We recently externally validated this score in patients included in the population-based registry of Dijon, France.17 The aim of this study was to assess
Journal of Stroke and Cerebrovascular Diseases, Vol. 24, No. 3 (March), 2015: pp 694-698
ISCORE FOR PREDICTING POSTSTROKE CARE
whether the iScore could also be of interest for predicting the need for poststroke institutional care in ischemic stroke survivors.
Methods Study Population All patients with a stroke diagnosed between January 1, 2006, and December 31, 2011, among the residents of the city of Dijon, France (151,543 inhabitants), were identified from the population-based Dijon Stroke Registry.1,17 This registry complies with epidemiologic criteria for stroke incidence studies.18 To ensure exhaustiveness, case collection relies on both hot and cold pursuit procedures so as to identify hospitalized and not-hospitalized patients in the catchment area, as described elsewhere.1,17 Stroke was defined according to the World Health Organization diagnostic criteria,19 and only ischemic stroke patients were considered for the present study.
Data Collected and iScore Calculation The following variables were recorded to assign a score to each patient, as previously described11: age, sex, preadmission dependence, cancer, atrial fibrillation, congestive heart failure, renal dialysis, stroke subtype (lacunar, nonlacunar, and undetermined), blood glucose on admission, and stroke severity. In our database, stroke severity was quantified using the National Institutes of Health Stroke Scale. It was therefore converted to the Canadian Neurological Scale in accordance with previously reported methods,20 so as to calculate the iScore. Other vascular risk factors were collected as previously described.1,17
695
Table 1. Risk scoring system to calculate the iScore Variables
30-day score
Age, y Sex Female Male Risk factors Atrial fibrillation Congestive heart failure Comorbid conditions Cancer Renal dialysis Preadmission dependence Stroke type Lacunar Nonlacunar Undetermined Stroke severity CNS $8 (NIHSS #8) CNS 5-7 (NIHSS, 9-13) CNS 1-4 (NIHSS, 14-22) CNS 0 (NIHSS $23) Glucose on admission, mmol/L ,7.5 $7.5
1Age (y) 0 110 110 110 110 135 115 0 130 135 0 140 165 1105 0 115
Abbreviations: CNS, Canadian Neurological Scale; NIHSS, National Institutes of Health Stroke Scale.
outcome in even severely disabled stroke patients by reducing poststroke mortality.
Statistical Analysis Outcomes Measured Outcome was poststroke care requirement. We distinguished between patients who returned home and those who required placement in a care institution including either an inpatient rehabilitation institution, a convalescent home (defined as an establishment where patients receive temporary care with no specific rehabilitation program before either going back home or being admitted to a long-term nursing home), or a long-term nursing facility. In our community, the medical attitude toward making decision on patients’ destination of discharge relies on a multidisciplinary approach involving both the neurologists and the neurorehabilitation physicians, based on clinical evaluation and testing by physiotherapists, occupational therapists, and speech therapists. Admission to the most appropriate health care facility is guided by clinical criteria found in the national conference of experts.21 The selection of moderately impaired patients for rehabilitation based on the assumption that these patients may be more likely to benefit from this type of management, which is a controversial matter with regard to data suggesting that rehabilitation could improve
Individual scores were calculated according to the published risk scoring system (Table 1).11 A logistic regression model was used to assess the performance of the iScore for predicting poststroke need for placement in a care institution. Receiver operator characteristic analysis was performed, and the c statistic representing the area under
Figure 1.
Study flow chart.
Y. BEJOT ET AL.
696
Table 2. Baseline characteristics of the 980 ischemic stroke patients Baseline characteristics Demographics Age, y, mean 6 SD* Age, y, median (IQR) Male gender* Medical history Hypertension Diabetes Hypercholesterolemia Current smoker Atrial fibrillation* Previous stroke or TIA Previous myocardial infarction Congestive heart failure* Peripheral vascular disease Cancer* Renal dialysis* Prestroke dependence* Stroke type* Lacunar Nonlacunar Undetermined Stroke severity* NIHSS score, mean 6 SD NIHSS score, median (IQR) Categories NIHSS #8 (CNS $8) NIHSS 9-13 (CNS 5-7) NIHSS 14-22 (CNS 1-4) NIHSS $23 (CNS 0) Glucose on admission, mmol/L* Mean 6 SD Median (IQR) Categories ,7.5 $7.5
n
% (95% CI)
74.4 6 15.7 78.7 (66.8-85.4) 458
46.7 (43.6-49.9)
684 165 379 360 166 152 100 114 51 131 0 70
69.8 (66.9-72.7) 16.8 (14.5-19.2) 38.7 (35.6-41.7) 36.7 (33.7-39.7) 16.9 (14.6-19.3) 15.5 (13.2-17.8) 10.2 (8.3-12.1) 11.6 (9.6-13.6) 5.2 (3.8-6.6) 13.4 (11.2-15.5) 0.0 7.1 (5.5-8.8)
194 526 260
19.8 (17.3-22.3) 53.7 (50.5-56.8) 26.5 (23.8-29.3)
6.4 6 6.2 4 (2-8) 745 101 103 31
76.0 (73.3-78.7) 10.3 (8.4-12.2) 10.5 (8.6-12.4) 3.2 (2.1-4.3)
6.9 6 2.7 6.2 (5.4-7.5) 725 255
74.0 (71.2-76.7) 26.0 (23.3-28.8)
Abbreviations: CI, confidence interval; CNS, Canadian Neurological Scale; IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale; SD, standard deviation; TIA, transient ischemic attack. *Variables included for the calculation of the iScore.
the receiver operator characteristic curve for the models was evaluated to assess discrimination. The optimal cutoff point representing the highest product of sensitivity and specificity was determined. The positive predictive value and the negative predictive value were calculated. Calibration was evaluated with the Hosmer– Lemeshow goodness-of-fit test. Statistical analysis was performed with STATA 10.0 software (StataCorp LP, College Station, TX).
Nationale de l’Informatique et des Libertes’’ (National Commission for the Protection of the Privacy of Electronic Data) was obtained.
Ethics The Dijon Stroke Registry was approved by the Comite National des Registres (French National Committee of Registers) and the InVS (French Institute for Public Health Surveillance). Authorization of the ‘‘Commission
Figure 2. Proportion of patients who required placement in a care institution after acute ischemic stroke by iScore quintiles.
ISCORE FOR PREDICTING POSTSTROKE CARE
697
Figure 4. The optimal cutoff of the iScore was 134 or greater, with a sensitivity and specificity of 57.0% and 81.2%, respectively. The positive predictive value was 72.7% and the negative predictive value was 68.3%. The positive likelihood ratio was 3.03 and the negative likelihood ratio was .17. Additional analyses were performed to assess the performance of the iScore for predicting either poststroke need for placement in a care institution or early death. Data were available for 1092 (91%) patients. As a result, the discrimination of the model was still good with a c statistic of .79 (95% confidence interval, .76-.82) as was calibration (P 5 .19 for the Hosmer–Lemeshow test). Figure 3. Receiver operator characteristic analysis for placement in a care institution.
Results Over the study period, 1199 patients with acute ischemic stroke were recorded. Of these, 124 (10.4%) were excluded because of early death (either at stroke presentation or during acute hospitalization). Among the 1075 stroke survivors, 95 (8.8%) were then excluded because there was at least 1 item of data missing for variables included in the iScore (Fig 1). Of the 980 remaining patients, 522 (53.3%) returned home and 458 (46.7%) required placement in a care institution including 189 (19.3%) in an inpatient rehabilitation center, 139 (14.2%) in a convalescence home, and 130 (13.2%) in a long-term nursing facility. Characteristics of patient are listed in Table 2. The median iScore was 123 (interquartile range, 97148). The proportion of patients who required placement in a care institution increased with each quintile of risk score (Fig 2). The discrimination of the model was good with a c statistic of .75 (95% confidence interval, .72-.78; Fig 3) as was calibration (P 5 .35 for the Hosmer– Lemeshow test). The observed versus predicted need for placement in a care institution are shown in
Figure 4. Observed versus predicted need for placement in a care institution according to the iScore (Pearson coefficient correlation, .99).
Discussion This study demonstrates that the iScore system is useful for predicting the need for placement in a care institution in patients who survive an ischemic stroke. This finding adds to those from previous studies, which showed the accuracy of the iScore for predicting mortality, the effectiveness of thrombolytic therapy, and functional outcome.11-17 To the best of our knowledge, no other easy-to-administer prognostic scores have been shown to reliably predict the need for poststroke care. The results of previous studies on this topic conducted in different settings with various health care systems demonstrated that several factors were associated with institutionalization or discharge to a rehabilitation facility. Older age, greater stroke severity or greater discharge disability, cognitive impairment, or dependence before stroke increased the risk of not being discharged.1,4,7–9 Of note, all these factors except cognitive impairment are included in the calculation of the iScore, which is consistent with the good accuracy of this score for predicting poststroke disposition. Taken together, these results are of interest for decision making in health care policy. They indicate that the iScore system could be an interesting tool for predicting poststroke outcomes in terms of both vital and functional
698
prognosis and resource needs, so as to better plan the need for dedicated services. In addition, because variables included in the iScore are easily collected, it makes this score very easy to use in routine practice. The strength of our study is the population-based design with a small proportion of patients excluded for missing data. Our cohort was, thus, representative of all ischemic strokes occurring in the entire population. However, given the variations in practices with regard to resource use after stroke, mainly because of differences in health care policy, social factors, and the availability of poststroke beds, the generalization of our findings may be limited, and further studies are required to validate the usefulness of the iScore to identify patients who need placement in a care institution after ischemic stroke.
References 1. Bejot Y, Troisgros O, Gremeaux V, et al. Poststroke disposition and associated factors in a population-based study: the Dijon Stroke Registry. Stroke 2012;43:2071-2077. 2. Fang J, Alderman MH. Trend of stroke hospitalization, United States, 1988-1997. Stroke 2001;32:2221-2226. 3. Treger I, Ring H, Schwartz R, et al. Hospital disposition after stroke in a national survey of acute cerebrovascular diseases in Israel. Arch Phys Med Rehabil 2008; 89:435-440. 4. Rundek T, Mast H, Hartmann A, et al. Predictors of resource use after acute hospitalization: the Northern Manhattan Stroke Study. Neurology 2000;55:1180-1187. 5. Gregory PC, Han E. Disparities in postacute stroke rehabilitation disposition to acute inpatient rehabilitation vs. home: findings from the North Carolina Hospital Discharge Database. Am J Phys Med Rehabil 2009; 88:100-107. 6. Gregory PC, Han E, Morozova O, et al. Do racial disparities exist in access to inpatient stroke rehabilitation in the state of Maryland? Am J Phys Med Rehabil 2006; 85:814-819. 7. Brown RD Jr, Ransom J, Hass S, et al. Use of nursing home after stroke and dependence on stroke severity: a population-based analysis. Stroke 1999;30:924-929. 8. Portelli R, Lowe D, Irwin P, et al, Intercollegiate Stroke Working Party. Institutionalization after stroke. Clin Rehabil 2005;19:97-108.
Y. BEJOT ET AL. 9. Schlegel D, Kolb SJ, Luciano JM, et al. Utility of the NIH Stroke Scale as a predictor of hospital disposition. Stroke 2003;34:134-137. 10. Saposnik G, Black SE, Hakim A, et al. Age disparities in stroke quality of care and delivery of health services. Stroke 2009;40:3328-3335. 11. Saposnik G, Kapral MK, Liu Y, et al, Investigators of the Registry of the Canadian Stroke Network, Stroke Outcomes Research Canada (SORCan) Working Group. IScore: a risk score to predict death early after hospitalization for an acute ischemic stroke. Circulation 2011;123:739-749. 12. Saposnik G, Raptis S, Kapral MK, et al, Investigators of the Registry of the Canadian Stroke Network and the Stroke Outcome Research Canada Working Group. The iScore predicts poor functional outcomes early after hospitalization for an acute ischemic stroke. Stroke 2011; 42:3421-3428. 13. Zhang N, Liu G, Zhang G, et al, On behalf of the China National Stroke Registry (CNSR) Investigators. External validation of the iScore for predicting ischemic stroke mortality in patients in China. Stroke 2013;44:1924-1929. 14. Park TH, Saposnik G, Bae HJ, et al. The iScore predicts functional outcome in Korean patients with ischemic stroke. Stroke 2013;44:1440-1442. 15. Saposnik G, Fang J, Kapral MK, et al, Investigators of the Registry of the Canadian Stroke Network (RCSN), Stroke Outcomes Research Canada (SORCan) Working Group. The iScore predicts effectiveness of thrombolytic therapy for acute ischemic stroke. Stroke 2012;43:1315-1322. 16. Park TH, Park SS, Ko Y, et al. The iScore predicts clinical response to tissue plasminogen activator in Korean stroke patients. J Stroke Cerebrovasc Dis 2014;23:367-373. 17. Bejot Y, Jacquin A, Daubail B, et al. Population-based validation of the iScore for predicting mortality and early functional outcome in ischemic stroke patients. Neuroepidemiology 2013;41:169-173. 18. Sudlow CL, Warlow CP. Comparing stroke incidence worldwide: what makes studies comparable? Stroke 1996;27:550-558. 19. WHO. The world health report 2000: Health Systems improving performance. Geneva: WHO 2000. 20. Nilanont Y, Komoltri C, Saposnik G, et al. The Canadian Neurological Scale and the NIHSS: development and validation of a simple conversion model. Cerebrovasc Dis 2010;30:120-126. 21. Pelissier J. The management of stroke patients. Conference of experts with a public hearing. Mulhouse (France), 22 October 2008. Ann Phys Rehabil Med 2010;53:124-147.
Background: We assessed whether the iScore could predict the need for poststroke institutional care. Methods: Patients with acute ischemic stroke living in Dijon, France, were recorded between 2006 and 2011, using a population-based stroke registry. The iScore was calculated for each patient. A logistic regression model was used to assess the performance of the iScore for predicting the need for placement in a care institution. The discrimination and calibration of the model were assessed using the c statistic and the Hosmer–Lemeshow goodness-of-fit test, respectively. Results: Of the 1199 patients recorded, 124 were excluded because of early death and 95 because of missing for variables included in the iScore. Of the remaining 980 patients, 522 (53.3%) returned home and 458 (46.7%) required placement in a care institution. The median iScore was 123 (interquartile range, 97-148), and the proportion of patients who required placement in a care institution increased with each quintile of risk score. The discrimination of the model was good with a c statistic of .75 (95% confidence interval, .72-.78), as was calibration (P 5 .35). Conclusions: The iScore could be useful for predicting the need for placement in a care institution in ischemic stroke patients. Further studies are required to confirm this finding. Key Words: Stroke—stroke outcome—epidemiology—stroke registry— predictors—discharge planning. Ó 2015 by National Stroke Association
Introduction Recent improvements in the management of acute ischemic stroke have led to increasing the number of patients who survive after the event, especially, elderly people because of the progressive aging population.
From the Dijon Stroke Registry, EA4184, Department of Neurology, University Hospital and Medical School of Dijon, University of Burgundy, Dijon, France. Received August 26, 2014; revision received October 31, 2014; accepted November 14, 2014. The Dijon Stroke Registry is supported by the French Institute for Public Health Surveillance (InVS) and Inserm. The authors have nothing to disclose. Address correspondence to Yannick Bejot, MD, PhD, Service de Neurologie, Bocage Central, 14 rue Gaffarel, BP77908, 21079 Dijon cedex, France. E-mail: [email protected]. 1052-3057/$ - see front matter Ó 2015 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.11.010
694
Among stroke survivors, 45%-75% are unable to return home at once and need to be discharged to either a rehabilitation facility or a nursing home, with consequences in terms of both quality of life and economic costs for the society.1-7 From a health care organization point of view, better identifying the need for poststroke institutional care is a difficult but great challenge so as to adjust necessary resources in terms of number of beds and health care professionals, to estimate costs, and to make projection on future needs. Although several predictors of not being discharged to home have been identified including age, stroke severity, and cognitive impairment,1,4,7–9 no reliable score to predict poststroke care requirements is available currently. The iScore system has been demonstrated to be useful for predicting 30day mortality, the effectiveness of thrombolytic therapy, and functional outcome, when applied to ischemic stroke patients.10-17 We recently externally validated this score in patients included in the population-based registry of Dijon, France.17 The aim of this study was to assess
Journal of Stroke and Cerebrovascular Diseases, Vol. 24, No. 3 (March), 2015: pp 694-698
ISCORE FOR PREDICTING POSTSTROKE CARE
whether the iScore could also be of interest for predicting the need for poststroke institutional care in ischemic stroke survivors.
Methods Study Population All patients with a stroke diagnosed between January 1, 2006, and December 31, 2011, among the residents of the city of Dijon, France (151,543 inhabitants), were identified from the population-based Dijon Stroke Registry.1,17 This registry complies with epidemiologic criteria for stroke incidence studies.18 To ensure exhaustiveness, case collection relies on both hot and cold pursuit procedures so as to identify hospitalized and not-hospitalized patients in the catchment area, as described elsewhere.1,17 Stroke was defined according to the World Health Organization diagnostic criteria,19 and only ischemic stroke patients were considered for the present study.
Data Collected and iScore Calculation The following variables were recorded to assign a score to each patient, as previously described11: age, sex, preadmission dependence, cancer, atrial fibrillation, congestive heart failure, renal dialysis, stroke subtype (lacunar, nonlacunar, and undetermined), blood glucose on admission, and stroke severity. In our database, stroke severity was quantified using the National Institutes of Health Stroke Scale. It was therefore converted to the Canadian Neurological Scale in accordance with previously reported methods,20 so as to calculate the iScore. Other vascular risk factors were collected as previously described.1,17
695
Table 1. Risk scoring system to calculate the iScore Variables
30-day score
Age, y Sex Female Male Risk factors Atrial fibrillation Congestive heart failure Comorbid conditions Cancer Renal dialysis Preadmission dependence Stroke type Lacunar Nonlacunar Undetermined Stroke severity CNS $8 (NIHSS #8) CNS 5-7 (NIHSS, 9-13) CNS 1-4 (NIHSS, 14-22) CNS 0 (NIHSS $23) Glucose on admission, mmol/L ,7.5 $7.5
1Age (y) 0 110 110 110 110 135 115 0 130 135 0 140 165 1105 0 115
Abbreviations: CNS, Canadian Neurological Scale; NIHSS, National Institutes of Health Stroke Scale.
outcome in even severely disabled stroke patients by reducing poststroke mortality.
Statistical Analysis Outcomes Measured Outcome was poststroke care requirement. We distinguished between patients who returned home and those who required placement in a care institution including either an inpatient rehabilitation institution, a convalescent home (defined as an establishment where patients receive temporary care with no specific rehabilitation program before either going back home or being admitted to a long-term nursing home), or a long-term nursing facility. In our community, the medical attitude toward making decision on patients’ destination of discharge relies on a multidisciplinary approach involving both the neurologists and the neurorehabilitation physicians, based on clinical evaluation and testing by physiotherapists, occupational therapists, and speech therapists. Admission to the most appropriate health care facility is guided by clinical criteria found in the national conference of experts.21 The selection of moderately impaired patients for rehabilitation based on the assumption that these patients may be more likely to benefit from this type of management, which is a controversial matter with regard to data suggesting that rehabilitation could improve
Individual scores were calculated according to the published risk scoring system (Table 1).11 A logistic regression model was used to assess the performance of the iScore for predicting poststroke need for placement in a care institution. Receiver operator characteristic analysis was performed, and the c statistic representing the area under
Figure 1.
Study flow chart.
Y. BEJOT ET AL.
696
Table 2. Baseline characteristics of the 980 ischemic stroke patients Baseline characteristics Demographics Age, y, mean 6 SD* Age, y, median (IQR) Male gender* Medical history Hypertension Diabetes Hypercholesterolemia Current smoker Atrial fibrillation* Previous stroke or TIA Previous myocardial infarction Congestive heart failure* Peripheral vascular disease Cancer* Renal dialysis* Prestroke dependence* Stroke type* Lacunar Nonlacunar Undetermined Stroke severity* NIHSS score, mean 6 SD NIHSS score, median (IQR) Categories NIHSS #8 (CNS $8) NIHSS 9-13 (CNS 5-7) NIHSS 14-22 (CNS 1-4) NIHSS $23 (CNS 0) Glucose on admission, mmol/L* Mean 6 SD Median (IQR) Categories ,7.5 $7.5
n
% (95% CI)
74.4 6 15.7 78.7 (66.8-85.4) 458
46.7 (43.6-49.9)
684 165 379 360 166 152 100 114 51 131 0 70
69.8 (66.9-72.7) 16.8 (14.5-19.2) 38.7 (35.6-41.7) 36.7 (33.7-39.7) 16.9 (14.6-19.3) 15.5 (13.2-17.8) 10.2 (8.3-12.1) 11.6 (9.6-13.6) 5.2 (3.8-6.6) 13.4 (11.2-15.5) 0.0 7.1 (5.5-8.8)
194 526 260
19.8 (17.3-22.3) 53.7 (50.5-56.8) 26.5 (23.8-29.3)
6.4 6 6.2 4 (2-8) 745 101 103 31
76.0 (73.3-78.7) 10.3 (8.4-12.2) 10.5 (8.6-12.4) 3.2 (2.1-4.3)
6.9 6 2.7 6.2 (5.4-7.5) 725 255
74.0 (71.2-76.7) 26.0 (23.3-28.8)
Abbreviations: CI, confidence interval; CNS, Canadian Neurological Scale; IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale; SD, standard deviation; TIA, transient ischemic attack. *Variables included for the calculation of the iScore.
the receiver operator characteristic curve for the models was evaluated to assess discrimination. The optimal cutoff point representing the highest product of sensitivity and specificity was determined. The positive predictive value and the negative predictive value were calculated. Calibration was evaluated with the Hosmer– Lemeshow goodness-of-fit test. Statistical analysis was performed with STATA 10.0 software (StataCorp LP, College Station, TX).
Nationale de l’Informatique et des Libertes’’ (National Commission for the Protection of the Privacy of Electronic Data) was obtained.
Ethics The Dijon Stroke Registry was approved by the Comite National des Registres (French National Committee of Registers) and the InVS (French Institute for Public Health Surveillance). Authorization of the ‘‘Commission
Figure 2. Proportion of patients who required placement in a care institution after acute ischemic stroke by iScore quintiles.
ISCORE FOR PREDICTING POSTSTROKE CARE
697
Figure 4. The optimal cutoff of the iScore was 134 or greater, with a sensitivity and specificity of 57.0% and 81.2%, respectively. The positive predictive value was 72.7% and the negative predictive value was 68.3%. The positive likelihood ratio was 3.03 and the negative likelihood ratio was .17. Additional analyses were performed to assess the performance of the iScore for predicting either poststroke need for placement in a care institution or early death. Data were available for 1092 (91%) patients. As a result, the discrimination of the model was still good with a c statistic of .79 (95% confidence interval, .76-.82) as was calibration (P 5 .19 for the Hosmer–Lemeshow test). Figure 3. Receiver operator characteristic analysis for placement in a care institution.
Results Over the study period, 1199 patients with acute ischemic stroke were recorded. Of these, 124 (10.4%) were excluded because of early death (either at stroke presentation or during acute hospitalization). Among the 1075 stroke survivors, 95 (8.8%) were then excluded because there was at least 1 item of data missing for variables included in the iScore (Fig 1). Of the 980 remaining patients, 522 (53.3%) returned home and 458 (46.7%) required placement in a care institution including 189 (19.3%) in an inpatient rehabilitation center, 139 (14.2%) in a convalescence home, and 130 (13.2%) in a long-term nursing facility. Characteristics of patient are listed in Table 2. The median iScore was 123 (interquartile range, 97148). The proportion of patients who required placement in a care institution increased with each quintile of risk score (Fig 2). The discrimination of the model was good with a c statistic of .75 (95% confidence interval, .72-.78; Fig 3) as was calibration (P 5 .35 for the Hosmer– Lemeshow test). The observed versus predicted need for placement in a care institution are shown in
Figure 4. Observed versus predicted need for placement in a care institution according to the iScore (Pearson coefficient correlation, .99).
Discussion This study demonstrates that the iScore system is useful for predicting the need for placement in a care institution in patients who survive an ischemic stroke. This finding adds to those from previous studies, which showed the accuracy of the iScore for predicting mortality, the effectiveness of thrombolytic therapy, and functional outcome.11-17 To the best of our knowledge, no other easy-to-administer prognostic scores have been shown to reliably predict the need for poststroke care. The results of previous studies on this topic conducted in different settings with various health care systems demonstrated that several factors were associated with institutionalization or discharge to a rehabilitation facility. Older age, greater stroke severity or greater discharge disability, cognitive impairment, or dependence before stroke increased the risk of not being discharged.1,4,7–9 Of note, all these factors except cognitive impairment are included in the calculation of the iScore, which is consistent with the good accuracy of this score for predicting poststroke disposition. Taken together, these results are of interest for decision making in health care policy. They indicate that the iScore system could be an interesting tool for predicting poststroke outcomes in terms of both vital and functional
698
prognosis and resource needs, so as to better plan the need for dedicated services. In addition, because variables included in the iScore are easily collected, it makes this score very easy to use in routine practice. The strength of our study is the population-based design with a small proportion of patients excluded for missing data. Our cohort was, thus, representative of all ischemic strokes occurring in the entire population. However, given the variations in practices with regard to resource use after stroke, mainly because of differences in health care policy, social factors, and the availability of poststroke beds, the generalization of our findings may be limited, and further studies are required to validate the usefulness of the iScore to identify patients who need placement in a care institution after ischemic stroke.
References 1. Bejot Y, Troisgros O, Gremeaux V, et al. Poststroke disposition and associated factors in a population-based study: the Dijon Stroke Registry. Stroke 2012;43:2071-2077. 2. Fang J, Alderman MH. Trend of stroke hospitalization, United States, 1988-1997. Stroke 2001;32:2221-2226. 3. Treger I, Ring H, Schwartz R, et al. Hospital disposition after stroke in a national survey of acute cerebrovascular diseases in Israel. Arch Phys Med Rehabil 2008; 89:435-440. 4. Rundek T, Mast H, Hartmann A, et al. Predictors of resource use after acute hospitalization: the Northern Manhattan Stroke Study. Neurology 2000;55:1180-1187. 5. Gregory PC, Han E. Disparities in postacute stroke rehabilitation disposition to acute inpatient rehabilitation vs. home: findings from the North Carolina Hospital Discharge Database. Am J Phys Med Rehabil 2009; 88:100-107. 6. Gregory PC, Han E, Morozova O, et al. Do racial disparities exist in access to inpatient stroke rehabilitation in the state of Maryland? Am J Phys Med Rehabil 2006; 85:814-819. 7. Brown RD Jr, Ransom J, Hass S, et al. Use of nursing home after stroke and dependence on stroke severity: a population-based analysis. Stroke 1999;30:924-929. 8. Portelli R, Lowe D, Irwin P, et al, Intercollegiate Stroke Working Party. Institutionalization after stroke. Clin Rehabil 2005;19:97-108.
Y. BEJOT ET AL. 9. Schlegel D, Kolb SJ, Luciano JM, et al. Utility of the NIH Stroke Scale as a predictor of hospital disposition. Stroke 2003;34:134-137. 10. Saposnik G, Black SE, Hakim A, et al. Age disparities in stroke quality of care and delivery of health services. Stroke 2009;40:3328-3335. 11. Saposnik G, Kapral MK, Liu Y, et al, Investigators of the Registry of the Canadian Stroke Network, Stroke Outcomes Research Canada (SORCan) Working Group. IScore: a risk score to predict death early after hospitalization for an acute ischemic stroke. Circulation 2011;123:739-749. 12. Saposnik G, Raptis S, Kapral MK, et al, Investigators of the Registry of the Canadian Stroke Network and the Stroke Outcome Research Canada Working Group. The iScore predicts poor functional outcomes early after hospitalization for an acute ischemic stroke. Stroke 2011; 42:3421-3428. 13. Zhang N, Liu G, Zhang G, et al, On behalf of the China National Stroke Registry (CNSR) Investigators. External validation of the iScore for predicting ischemic stroke mortality in patients in China. Stroke 2013;44:1924-1929. 14. Park TH, Saposnik G, Bae HJ, et al. The iScore predicts functional outcome in Korean patients with ischemic stroke. Stroke 2013;44:1440-1442. 15. Saposnik G, Fang J, Kapral MK, et al, Investigators of the Registry of the Canadian Stroke Network (RCSN), Stroke Outcomes Research Canada (SORCan) Working Group. The iScore predicts effectiveness of thrombolytic therapy for acute ischemic stroke. Stroke 2012;43:1315-1322. 16. Park TH, Park SS, Ko Y, et al. The iScore predicts clinical response to tissue plasminogen activator in Korean stroke patients. J Stroke Cerebrovasc Dis 2014;23:367-373. 17. Bejot Y, Jacquin A, Daubail B, et al. Population-based validation of the iScore for predicting mortality and early functional outcome in ischemic stroke patients. Neuroepidemiology 2013;41:169-173. 18. Sudlow CL, Warlow CP. Comparing stroke incidence worldwide: what makes studies comparable? Stroke 1996;27:550-558. 19. WHO. The world health report 2000: Health Systems improving performance. Geneva: WHO 2000. 20. Nilanont Y, Komoltri C, Saposnik G, et al. The Canadian Neurological Scale and the NIHSS: development and validation of a simple conversion model. Cerebrovasc Dis 2010;30:120-126. 21. Pelissier J. The management of stroke patients. Conference of experts with a public hearing. Mulhouse (France), 22 October 2008. Ann Phys Rehabil Med 2010;53:124-147.
