- Email: [email protected]
Evaluation of Charlson's comorbidity index in elderly living in nursing homes
Journal of Clinical Epidemiology 55 (2002) 1144–1147
Evaluation of Charlson’s comorbidity index in elderly living in nursing homes F. Buntinxa,c,*, L. Niclaesa, C. Suetensb, B. Jansb, R. Mertensb, M. Van den Akkerc a Department of General Practice–KUL, Kapucijnenvoer 33, Blok J, B-3000 Leuven, Belgium Scientific Institute for Public Health, Louis Pasteur, Juliet Wytsmanstreet 14-B1050 Brussels, Belgium c University of Maastricht, The Netherlands Department of General Practice, PB 616-6200 MD Maastricht, The Netherlands Received 28 February 2001; received in revised form 9 November 2001; accepted 10 July 2002 b
Abstract The object of this article was to validate the predictive value of Charlson’s comorbidity index for the prediction of short-term mortality or morbidity in elderly people. The design was a cohort study comparing survival and hospitalization in institutionalized elderly people with different levels of comorbidity at baseline. The setting was 16 Flemish nursing homes for the elderly. The subjects were 2,727 inhabitants of which full data were available for 2,624. The outcome measures were hazard ratios resulting from Cox regression analysis, comparing 6 months survival in patients with moderate and a high level to low level of comorbidity. Odds ratios resulting from multiple logistic regression analysis comparing the occurrence of at least one hospitalization during the follow-up period in surviving patients of the same groups. Mortality adjusted for age group was significantly increased in patients with a moderate (HR 2.00) and even more in those with a high level (HR 3.62) of comorbidity. Hospitalization was more frequent in both groups (OR 1.54 and 2.19, respectively), with statistical significance only being reached for the highest group. Adjustment for age, gender, mobility status, and disorientation did not change the general picture. Charlson’s comorbidity index is a predictor of short-term mortality in institutionalized elderly patients and, to a lesser extend, also of hospitalization. These results support its use as a measure for introducing comorbidity as a covariable in longitudinal studies with a geriatric population. © 2002 Elsevier Science Inc. All rights reserved. Keywords: Comorbidity; Elderly people; Charleson’s comorbidity index
1. Introduction As comorbidity is a determinant of mortality as well as disability [1–6], some measure of comorbidity is an important covariable when studying prognosis or outcome of treatment [7,8]. As the incidence and prevalence of cooccurrence of diseases are sharply rising with increasing age [2–5,9] and people are progressively becoming older, this also has consequences for health care organization and planning. A number of researchers have tried to develop and validate instruments to register comorbidity [9–14]. From the literature, Charlson’s comorbidity index appears to have a growing popularity. A Medline search identified 58 studies between 1995–1999 in which the index or some adaptation was used as a determinant or covariable. From a series of 30 comorbid diseases, some of them indicating different severity categories of the same disorder, Charlson et al. selected 19 with a 1-year mortality relative risk of 1.2 or more, adjusted for the contribution of all coexistent diseases, illness
* Corresponding author. Tel: 32 16 33 74 93; fax: 32 16 33 74 80. E-mail address: [email protected] (F. Buntinx).
severity, and reason for admission [10]. Also, a combined age–comorbidity index is available, in which each decade of age over 40 adds one point to the score. The influence of the disorders on the index was weighted according to the magnitude of the 1-year mortality relative risk. The index was developed from an inception cohort study of 604 patients admitted to a hospital’s medical service and initially tested for its ability to predict death from comorbid disease in a cohort of 685 patients with primary breast cancer. The index was later also tested in patients with hypertension or diabetes who underwent elective surgery. In this study the relative risk of death within 5 years was 1.46 for each unit of the comorbidity score, 1.42 for each decade of age over 40, and 1.45 for each unit of the combined age–comorbidity score [15]. In the mean time, the index has been used for studying treatment allocation in cancer patients [6,7], and it has been adapted to enable it to be used with ICD 9-based billing data that are routinely collected [16,17], although the validity of such adaptations has been questioned [18,19]. Although the Charlson index is well validated against mortality over a period of 1 year or longer, no research is available with respect to the prediction of short-term mortality and only one study on morbidity [14], and no studies have been per-
0895-4356/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved. PII: S0895-4356(02)00 4 8 5 - 7
F. Buntinx et al. / Journal of Clinical Epidemiology 55 (2002) 1144–1147
formed in old-age people. We therefore examined the predictive value of the index in a large population of people living in a nursing home for the elderly studying mortality and hospitalization (as a proxy measure for new or worsened morbidity) during a follow-up period of 6 months. 2. Methods 2.1. Patients From April to June 1997, a cross-sectional study was carried out in all 2,570 residents of 16 nursing homes evenly spread over the Flemish territory. Four institutions were public and 15 were private; all were run by nonprofit organizations. According to statistical information the proportion of inhabitants of Belgian nursing homes requiring intensive care and help for activities of daily life is estimated to be 58%. The status of all patients with respect to mortality and first hospitalization was followed up during a period of 6 months. 2.2. Procedures Data on age and sex were taken from the administrative files. Identification of mobility status (mobile, chairbound, bedridden) and degree of disorientation in time and space were based on clinical information. In Belgium, the latter is registered routinely by nursing home staff for social security matters, using a five-item scale, ranging from “well” to “totally disoriented” [11]. Scores were dichotomized as “well to seldom disoriented” or “daily to totally disoriented.” Death and first hospitalization were registered intermittently during the follow-up period. Comorbidity was studied using Charlson’ s comorbidity index. Data were extracted from the medical files that are kept for all patients. The index-relation with the outcome variables was categorized in three categories: low (0–1), medium [2–4], and high [5–12]. As the follow-up period was restricted to 6 months only and the age window in this elderly population was small compared to an unselected population, we did not use the combined comorbidity–age score. However, age was used as a covariable. 2.3. Analysis Data were analyzed using Stata statistical software. A two-tailed P-value of .05 was chosen as the cut point for statistical significance. The sample size enables the detection of a statistically significant risk ratio of 1.4 or more with a power of 0.90, considering a 6-month mortality rate of 10% in the reference group. Follow-up data were collected at 6 months after baseline measurement. Follow-up time was set at 183 days for all survivors. For deceased patients the follow up time was calculated as the date of death minus the date at which the baseline data were collected. Two patients who were reported to have died shortly after day 183 were classified as survivors for this analysis. For five deceased patients for whom the death date was unknown, the follow-up time was
1145
substituted by the median survival time for those who died during the follow-up period, i.e., 95 days. Differences in survival according to comorbidity status at baseline were assessed using the Kaplan-Meier method. Cox’s proportional hazard regression was used to calculate the hazard ratio (HR) with 95% confidence limits expressing the risk of dying for each comorbidity group at baseline compared to the group with the lowest level of comorbidity. The hazard ratio was calculated before and after adjustment for age. Additionally odds ratios based on logistic regression were calculated with and without adjustment for age, gender, disorientation in time and morbidity status. The impact of comorbidity on the presence of at least one hospitalization during the follow-up period was analyzed using multiple logistic regression analysis with and without adjustment for the same covariables as used in the mortality analysis. It included surviving patients only. 2.4. Ethical considerations: The study was approved by the ethical committee of the medical school of the Catholic University of Leuven.
3. Results 3.1. Description of the study population Two thousand seven hundred twenty-seven inhabitants of 16 nursing homes were included in the study. Data at 6 months of follow-up were available for 2,624 inhabitants. Two hundred fifty-four (9.3%) patients died during the follow-up period and 366 (13.9%) were at least once hospitalized. The number of residents by nursing home ranged from 14 to 291, with a median of 131. The mean age was 84 years (range 49–103). For the analysis age was categorized in three groups 80, 80–89, 90 years, including 670 (26%), 1274 (49%), and 678 (26%) people, respectively. Men represented only 22% of the study population and had a mean age of 81, compared to 85 in women. The median duration of stay in the home was 33 months. Fifty-three percent of the patients were mobile, 43% were chairbound, and 4.4% were bedridden. On the Comorbidity Index 1,210 (46%) scored low, 1,226 (47%) moderate, and 188 (7%) high. 3.2. Mortality Compared to the patients with a low level of comorbidity, mortality adjusted for age group was significantly increased in patients with a moderate (HR 2.00) and even more in patients with a high level (HR 3.62) (Table 1). Mortality was also significantly and independently increased in patients of both oldest age groups compared with patients that are younger than age 80 (HR 1.54 and 2.19, respectively) (Table 1). Odds ratios adjusted for age, gender, disorientation, and mobility status were 1.62 (95% CI 1.20–2.18) and 2.92 (95% CI 1.87–4.58), respectively.
1146
F. Buntinx et al. / Journal of Clinical Epidemiology 55 (2002) 1144–1147
3.3. Hospitalization The likelihood of at least one hospitalization adjusted for age group in survivors was higher in patients with a moderate (OR 1.17) and high (OR 1.61) level of comorbidity compared to those with a low level (Table 1). Statistical significance was only reached for the highest group. Adjusting for age, gender, disorientation, and mobility status did not change the results to any relevant degree: OR 1.27 (95% CI 1.00–1.61) and 1.56 (95% CI 1.02–2.38) for a moderate and high level of comorbidity. 3.4. Components of the comorbidity index Odds ratios for the relation between each component of the comorbidity index and mortality or hospitalization within six months are presented in Table 2. 4. Discussion When studying the relation between one or more exposure factors (e.g., disease or treatment) and survival or outcome, a measure of comorbidity that either can be used as a covariable or permit stratification of the population in more homogeneous subgroups is essential (7.8). Such measure may also be useful for case-mix determination, and maybe in the evaluation of health care policies [13,14]. With increasing age of the study population, this becomes even more important [1]. Today, no generally accepted measure of comorbidity is available for use in elderly people. Although Charlson’s comorbidity index seems to become a popular measure for use in the general population, we were not able to identify any validation study in the elderly. In this study we demonstrated the index to be related to shortterm mortality in institutionalized elderly people with a yearly mortality rate of around 20%. The relation was confirmed both with and without adjustment for age, gender, degree of disorientation, and mobility status. This study was restricted to a follow-up period of 6 months, so no conclusions can be drawn for longer follow-up periods. Charlson’s comorbidity index also showed to predict hospitalization within the follow-up period, although this relation was weaker and only statistically significant for the “high” group. One should realize, however, that these patients were already institutionalized with the presence of a professional
Table 1 Relation between the age-adjusted risk of dying or hospitalization and the comorbidity Index Comorbidity group
Survival (95% CI)
Hospitalization (95% CI)
Medium vs. low
2.00 (1.50–2.67) 3.62 (2.41–5.42)
1.17 (0.88–1.49) 1.61 (1.04–2.50)
High vs. low
Presented are hazard ratios (survival) or odds ratios (hospitalization) and their 95% confidence intervals.
Table 2 Relation between each of the components of the comorbidity index and mortality and hospitalization within a period of 6 months (odds ratios) Myocardial infarction Decompensatio cordis Peripheral arterial disease Cardiovascular disease Hemiplegia Dementia Chronic lung disease Diabetes mild Diabetes serious Liver disorder mild Liver disorder serious Peptic ulcer Malignancies Metastasis Systemic disease Lymphoma Leukemia Kidney disorders Chronic urinary disorders
Mortality
Hospitalization
1.21 1.23 1.06 1.09 1.81** 1.64** 1.02 1.11 1.47 0.78 2.28 1.11 2.13** 5.28** 0.60 9.18** 2.27 1.86** 1.56
2.72** 1.27* 1.17 0.93 0.91 0.85 1.11 1.07 1.39 0.53 0.94 1.20 1.23 1.45 1.11 — — 1.87* 1.36
* p .05 ** p .01 No odds ratio because not all cells of the 22 table are filled.
nursing staff and more opportunities for technical interventions. Additionally, some of them were very old. Physicians may have considered these to be reasons to restrain from hospitalization in situations where hospitalization would have been an option for patients living alone. As could be expected, an examination of the role of each component of the index identified high odds ratios for the relation between the components related to malignancies and short-term mortality and for the relation between heart diseases and hospitalization. The remaining part of the effect seems scattered allover all items. These results support the use of Charlson’s comorbidity index as a measure of comorbidity in studies with a geriatric population. Acknowledgments Data collection was made possible by the invaluable help of Eddy De Roost and the staff members in each of the nursing homes. Marina Devis was responsible for most of the data input and for secretarial assistance. The study was funded by a grant of the Flemish Research Fund (FWOVlaanderen No. G.0338.97).
References [1] van den Akker M, Buntinx F, Knottnerus JA. Comorbidity or multimorbidity: what’s in a name? A review of literature. European Journal of General Practice 1996;2:65–70. [2] Guralnik JM, La Croix AZ, Everett DF, Kovar MG. Aging in the eighties: the prevalence of comorbidity and its association with disability. Advance data. Hyattsville, Maryland: National Center for Health Statistics; 1989. p. 170.
F. Buntinx et al. / Journal of Clinical Epidemiology 55 (2002) 1144–1147 [3] Verbrugge LM, Lepkowski JM, Imanaka Y. Comorbidity and its impact on disability. Milbank Mem Fund Q 1989;67:450–84. [4] Van der Velden J, van der Bos GAM, Schellevis FG, van Ammers E. Comorbiditeit in AN. Volksgezondheid Toekomst Verkenning. Rijswijk: RIVM; 1993. [5] Meyboom-de Jong B. De oudere patient in beeld. Huisarts Wet 1991;34: 82–8. [6] Janssen-Heynen ML, Schipper RM, Razeberg PP, Crommelin MA, Coebergh JW. Prevalence of co-morbidity in lung cancer patients and its relationship with treatment. Lung Cancer 1998;21:105–13. [7] Coebergh JW, Janssen-Heijnen ML, Post PN, Razenberg PP. Serious co-morbidity among unselected cancer patients newly diagnosed in the southeastern part of the Netherlands in 1993–1996. J Clin Epidemiol 1999;52:1131–6. [8] van den Akker M, Buntinx F, Metsemakers JFM, Roos S, Knottnerus JA. Multimorbidity in general practice; prevalence, incidence, and determinants of co-occurring chronic and recurrent diseases. J Clin Epidemiol 1998;51:367–75. [9] Kaplan MH, Feinstein AR. The importance of clasifying initial comorbidity in evaluating the outcome of diabetes mellitus. Journal of Chronic Diseases 1974;27:387–404. [10] Charlson ME, Ales KL, Pompei P, Mac Enzie CR. A new method of classification of prognostic comorbidity for longitudinal studies: development and validation. Journal of Chronic Diseases 1987;40:373–83.
1147
[11] Flood AB, Scott WR, Ewy W. Does practice make perfect? Med Care 1984;22:98–114. [12] Greenfield S, Blanco DM, Elshoff RM, Ganz PA. Patterns of care related to age of breast cancer patients. JAMA 1987;257:2766–70. [13] Kahn KL, Park RE, Brook RH, et al. The effect of comorbidity on appropriateness ratings for two gastrointestinal procedures. J Clin Epidemiol 1988;41:115–22. [14] Brailer DJ, Kroch E, Pauly MV, Huang J. Comorbidity-afjusted complication risk. A new outcome quality measure. Med Care 1996;34:490–505. [15] Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol 1994;47:1245–51. [16] Kieszak SM, Flanders WD, Kosinski AS, Shipp CC, Karp H. A comparison of the Charlson comorbidity index derived from medical record data and administrative billing data. J Clin Epidemiol 1999;52:137–42. [17] Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613–9. [18] Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993;46:1075–9. [19] Deyo RA. Adapting a clinical comorbidity index for use with ICD9-CM administrative data: a response. J Clin Epidemiol 1993;46: 1081–2.
Evaluation of Charlson’s comorbidity index in elderly living in nursing homes F. Buntinxa,c,*, L. Niclaesa, C. Suetensb, B. Jansb, R. Mertensb, M. Van den Akkerc a Department of General Practice–KUL, Kapucijnenvoer 33, Blok J, B-3000 Leuven, Belgium Scientific Institute for Public Health, Louis Pasteur, Juliet Wytsmanstreet 14-B1050 Brussels, Belgium c University of Maastricht, The Netherlands Department of General Practice, PB 616-6200 MD Maastricht, The Netherlands Received 28 February 2001; received in revised form 9 November 2001; accepted 10 July 2002 b
Abstract The object of this article was to validate the predictive value of Charlson’s comorbidity index for the prediction of short-term mortality or morbidity in elderly people. The design was a cohort study comparing survival and hospitalization in institutionalized elderly people with different levels of comorbidity at baseline. The setting was 16 Flemish nursing homes for the elderly. The subjects were 2,727 inhabitants of which full data were available for 2,624. The outcome measures were hazard ratios resulting from Cox regression analysis, comparing 6 months survival in patients with moderate and a high level to low level of comorbidity. Odds ratios resulting from multiple logistic regression analysis comparing the occurrence of at least one hospitalization during the follow-up period in surviving patients of the same groups. Mortality adjusted for age group was significantly increased in patients with a moderate (HR 2.00) and even more in those with a high level (HR 3.62) of comorbidity. Hospitalization was more frequent in both groups (OR 1.54 and 2.19, respectively), with statistical significance only being reached for the highest group. Adjustment for age, gender, mobility status, and disorientation did not change the general picture. Charlson’s comorbidity index is a predictor of short-term mortality in institutionalized elderly patients and, to a lesser extend, also of hospitalization. These results support its use as a measure for introducing comorbidity as a covariable in longitudinal studies with a geriatric population. © 2002 Elsevier Science Inc. All rights reserved. Keywords: Comorbidity; Elderly people; Charleson’s comorbidity index
1. Introduction As comorbidity is a determinant of mortality as well as disability [1–6], some measure of comorbidity is an important covariable when studying prognosis or outcome of treatment [7,8]. As the incidence and prevalence of cooccurrence of diseases are sharply rising with increasing age [2–5,9] and people are progressively becoming older, this also has consequences for health care organization and planning. A number of researchers have tried to develop and validate instruments to register comorbidity [9–14]. From the literature, Charlson’s comorbidity index appears to have a growing popularity. A Medline search identified 58 studies between 1995–1999 in which the index or some adaptation was used as a determinant or covariable. From a series of 30 comorbid diseases, some of them indicating different severity categories of the same disorder, Charlson et al. selected 19 with a 1-year mortality relative risk of 1.2 or more, adjusted for the contribution of all coexistent diseases, illness
* Corresponding author. Tel: 32 16 33 74 93; fax: 32 16 33 74 80. E-mail address: [email protected] (F. Buntinx).
severity, and reason for admission [10]. Also, a combined age–comorbidity index is available, in which each decade of age over 40 adds one point to the score. The influence of the disorders on the index was weighted according to the magnitude of the 1-year mortality relative risk. The index was developed from an inception cohort study of 604 patients admitted to a hospital’s medical service and initially tested for its ability to predict death from comorbid disease in a cohort of 685 patients with primary breast cancer. The index was later also tested in patients with hypertension or diabetes who underwent elective surgery. In this study the relative risk of death within 5 years was 1.46 for each unit of the comorbidity score, 1.42 for each decade of age over 40, and 1.45 for each unit of the combined age–comorbidity score [15]. In the mean time, the index has been used for studying treatment allocation in cancer patients [6,7], and it has been adapted to enable it to be used with ICD 9-based billing data that are routinely collected [16,17], although the validity of such adaptations has been questioned [18,19]. Although the Charlson index is well validated against mortality over a period of 1 year or longer, no research is available with respect to the prediction of short-term mortality and only one study on morbidity [14], and no studies have been per-
0895-4356/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved. PII: S0895-4356(02)00 4 8 5 - 7
F. Buntinx et al. / Journal of Clinical Epidemiology 55 (2002) 1144–1147
formed in old-age people. We therefore examined the predictive value of the index in a large population of people living in a nursing home for the elderly studying mortality and hospitalization (as a proxy measure for new or worsened morbidity) during a follow-up period of 6 months. 2. Methods 2.1. Patients From April to June 1997, a cross-sectional study was carried out in all 2,570 residents of 16 nursing homes evenly spread over the Flemish territory. Four institutions were public and 15 were private; all were run by nonprofit organizations. According to statistical information the proportion of inhabitants of Belgian nursing homes requiring intensive care and help for activities of daily life is estimated to be 58%. The status of all patients with respect to mortality and first hospitalization was followed up during a period of 6 months. 2.2. Procedures Data on age and sex were taken from the administrative files. Identification of mobility status (mobile, chairbound, bedridden) and degree of disorientation in time and space were based on clinical information. In Belgium, the latter is registered routinely by nursing home staff for social security matters, using a five-item scale, ranging from “well” to “totally disoriented” [11]. Scores were dichotomized as “well to seldom disoriented” or “daily to totally disoriented.” Death and first hospitalization were registered intermittently during the follow-up period. Comorbidity was studied using Charlson’ s comorbidity index. Data were extracted from the medical files that are kept for all patients. The index-relation with the outcome variables was categorized in three categories: low (0–1), medium [2–4], and high [5–12]. As the follow-up period was restricted to 6 months only and the age window in this elderly population was small compared to an unselected population, we did not use the combined comorbidity–age score. However, age was used as a covariable. 2.3. Analysis Data were analyzed using Stata statistical software. A two-tailed P-value of .05 was chosen as the cut point for statistical significance. The sample size enables the detection of a statistically significant risk ratio of 1.4 or more with a power of 0.90, considering a 6-month mortality rate of 10% in the reference group. Follow-up data were collected at 6 months after baseline measurement. Follow-up time was set at 183 days for all survivors. For deceased patients the follow up time was calculated as the date of death minus the date at which the baseline data were collected. Two patients who were reported to have died shortly after day 183 were classified as survivors for this analysis. For five deceased patients for whom the death date was unknown, the follow-up time was
1145
substituted by the median survival time for those who died during the follow-up period, i.e., 95 days. Differences in survival according to comorbidity status at baseline were assessed using the Kaplan-Meier method. Cox’s proportional hazard regression was used to calculate the hazard ratio (HR) with 95% confidence limits expressing the risk of dying for each comorbidity group at baseline compared to the group with the lowest level of comorbidity. The hazard ratio was calculated before and after adjustment for age. Additionally odds ratios based on logistic regression were calculated with and without adjustment for age, gender, disorientation in time and morbidity status. The impact of comorbidity on the presence of at least one hospitalization during the follow-up period was analyzed using multiple logistic regression analysis with and without adjustment for the same covariables as used in the mortality analysis. It included surviving patients only. 2.4. Ethical considerations: The study was approved by the ethical committee of the medical school of the Catholic University of Leuven.
3. Results 3.1. Description of the study population Two thousand seven hundred twenty-seven inhabitants of 16 nursing homes were included in the study. Data at 6 months of follow-up were available for 2,624 inhabitants. Two hundred fifty-four (9.3%) patients died during the follow-up period and 366 (13.9%) were at least once hospitalized. The number of residents by nursing home ranged from 14 to 291, with a median of 131. The mean age was 84 years (range 49–103). For the analysis age was categorized in three groups 80, 80–89, 90 years, including 670 (26%), 1274 (49%), and 678 (26%) people, respectively. Men represented only 22% of the study population and had a mean age of 81, compared to 85 in women. The median duration of stay in the home was 33 months. Fifty-three percent of the patients were mobile, 43% were chairbound, and 4.4% were bedridden. On the Comorbidity Index 1,210 (46%) scored low, 1,226 (47%) moderate, and 188 (7%) high. 3.2. Mortality Compared to the patients with a low level of comorbidity, mortality adjusted for age group was significantly increased in patients with a moderate (HR 2.00) and even more in patients with a high level (HR 3.62) (Table 1). Mortality was also significantly and independently increased in patients of both oldest age groups compared with patients that are younger than age 80 (HR 1.54 and 2.19, respectively) (Table 1). Odds ratios adjusted for age, gender, disorientation, and mobility status were 1.62 (95% CI 1.20–2.18) and 2.92 (95% CI 1.87–4.58), respectively.
1146
F. Buntinx et al. / Journal of Clinical Epidemiology 55 (2002) 1144–1147
3.3. Hospitalization The likelihood of at least one hospitalization adjusted for age group in survivors was higher in patients with a moderate (OR 1.17) and high (OR 1.61) level of comorbidity compared to those with a low level (Table 1). Statistical significance was only reached for the highest group. Adjusting for age, gender, disorientation, and mobility status did not change the results to any relevant degree: OR 1.27 (95% CI 1.00–1.61) and 1.56 (95% CI 1.02–2.38) for a moderate and high level of comorbidity. 3.4. Components of the comorbidity index Odds ratios for the relation between each component of the comorbidity index and mortality or hospitalization within six months are presented in Table 2. 4. Discussion When studying the relation between one or more exposure factors (e.g., disease or treatment) and survival or outcome, a measure of comorbidity that either can be used as a covariable or permit stratification of the population in more homogeneous subgroups is essential (7.8). Such measure may also be useful for case-mix determination, and maybe in the evaluation of health care policies [13,14]. With increasing age of the study population, this becomes even more important [1]. Today, no generally accepted measure of comorbidity is available for use in elderly people. Although Charlson’s comorbidity index seems to become a popular measure for use in the general population, we were not able to identify any validation study in the elderly. In this study we demonstrated the index to be related to shortterm mortality in institutionalized elderly people with a yearly mortality rate of around 20%. The relation was confirmed both with and without adjustment for age, gender, degree of disorientation, and mobility status. This study was restricted to a follow-up period of 6 months, so no conclusions can be drawn for longer follow-up periods. Charlson’s comorbidity index also showed to predict hospitalization within the follow-up period, although this relation was weaker and only statistically significant for the “high” group. One should realize, however, that these patients were already institutionalized with the presence of a professional
Table 1 Relation between the age-adjusted risk of dying or hospitalization and the comorbidity Index Comorbidity group
Survival (95% CI)
Hospitalization (95% CI)
Medium vs. low
2.00 (1.50–2.67) 3.62 (2.41–5.42)
1.17 (0.88–1.49) 1.61 (1.04–2.50)
High vs. low
Presented are hazard ratios (survival) or odds ratios (hospitalization) and their 95% confidence intervals.
Table 2 Relation between each of the components of the comorbidity index and mortality and hospitalization within a period of 6 months (odds ratios) Myocardial infarction Decompensatio cordis Peripheral arterial disease Cardiovascular disease Hemiplegia Dementia Chronic lung disease Diabetes mild Diabetes serious Liver disorder mild Liver disorder serious Peptic ulcer Malignancies Metastasis Systemic disease Lymphoma Leukemia Kidney disorders Chronic urinary disorders
Mortality
Hospitalization
1.21 1.23 1.06 1.09 1.81** 1.64** 1.02 1.11 1.47 0.78 2.28 1.11 2.13** 5.28** 0.60 9.18** 2.27 1.86** 1.56
2.72** 1.27* 1.17 0.93 0.91 0.85 1.11 1.07 1.39 0.53 0.94 1.20 1.23 1.45 1.11 — — 1.87* 1.36
* p .05 ** p .01 No odds ratio because not all cells of the 22 table are filled.
nursing staff and more opportunities for technical interventions. Additionally, some of them were very old. Physicians may have considered these to be reasons to restrain from hospitalization in situations where hospitalization would have been an option for patients living alone. As could be expected, an examination of the role of each component of the index identified high odds ratios for the relation between the components related to malignancies and short-term mortality and for the relation between heart diseases and hospitalization. The remaining part of the effect seems scattered allover all items. These results support the use of Charlson’s comorbidity index as a measure of comorbidity in studies with a geriatric population. Acknowledgments Data collection was made possible by the invaluable help of Eddy De Roost and the staff members in each of the nursing homes. Marina Devis was responsible for most of the data input and for secretarial assistance. The study was funded by a grant of the Flemish Research Fund (FWOVlaanderen No. G.0338.97).
References [1] van den Akker M, Buntinx F, Knottnerus JA. Comorbidity or multimorbidity: what’s in a name? A review of literature. European Journal of General Practice 1996;2:65–70. [2] Guralnik JM, La Croix AZ, Everett DF, Kovar MG. Aging in the eighties: the prevalence of comorbidity and its association with disability. Advance data. Hyattsville, Maryland: National Center for Health Statistics; 1989. p. 170.
F. Buntinx et al. / Journal of Clinical Epidemiology 55 (2002) 1144–1147 [3] Verbrugge LM, Lepkowski JM, Imanaka Y. Comorbidity and its impact on disability. Milbank Mem Fund Q 1989;67:450–84. [4] Van der Velden J, van der Bos GAM, Schellevis FG, van Ammers E. Comorbiditeit in AN. Volksgezondheid Toekomst Verkenning. Rijswijk: RIVM; 1993. [5] Meyboom-de Jong B. De oudere patient in beeld. Huisarts Wet 1991;34: 82–8. [6] Janssen-Heynen ML, Schipper RM, Razeberg PP, Crommelin MA, Coebergh JW. Prevalence of co-morbidity in lung cancer patients and its relationship with treatment. Lung Cancer 1998;21:105–13. [7] Coebergh JW, Janssen-Heijnen ML, Post PN, Razenberg PP. Serious co-morbidity among unselected cancer patients newly diagnosed in the southeastern part of the Netherlands in 1993–1996. J Clin Epidemiol 1999;52:1131–6. [8] van den Akker M, Buntinx F, Metsemakers JFM, Roos S, Knottnerus JA. Multimorbidity in general practice; prevalence, incidence, and determinants of co-occurring chronic and recurrent diseases. J Clin Epidemiol 1998;51:367–75. [9] Kaplan MH, Feinstein AR. The importance of clasifying initial comorbidity in evaluating the outcome of diabetes mellitus. Journal of Chronic Diseases 1974;27:387–404. [10] Charlson ME, Ales KL, Pompei P, Mac Enzie CR. A new method of classification of prognostic comorbidity for longitudinal studies: development and validation. Journal of Chronic Diseases 1987;40:373–83.
1147
[11] Flood AB, Scott WR, Ewy W. Does practice make perfect? Med Care 1984;22:98–114. [12] Greenfield S, Blanco DM, Elshoff RM, Ganz PA. Patterns of care related to age of breast cancer patients. JAMA 1987;257:2766–70. [13] Kahn KL, Park RE, Brook RH, et al. The effect of comorbidity on appropriateness ratings for two gastrointestinal procedures. J Clin Epidemiol 1988;41:115–22. [14] Brailer DJ, Kroch E, Pauly MV, Huang J. Comorbidity-afjusted complication risk. A new outcome quality measure. Med Care 1996;34:490–505. [15] Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol 1994;47:1245–51. [16] Kieszak SM, Flanders WD, Kosinski AS, Shipp CC, Karp H. A comparison of the Charlson comorbidity index derived from medical record data and administrative billing data. J Clin Epidemiol 1999;52:137–42. [17] Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613–9. [18] Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993;46:1075–9. [19] Deyo RA. Adapting a clinical comorbidity index for use with ICD9-CM administrative data: a response. J Clin Epidemiol 1993;46: 1081–2.