Body mass index and hospitalization in the elderly

OBJECTIVES: To explore the association between body mass index (BMI) and hospital usage in the elderly. DESIGN: Retrospective cohort study. SETTING: Medicare Current Beneficiary Survey (1992-94). PARTICIPANTS: Eight thousand seven hundred fifty-four noninstitutionalized individuals aged 65 to 100 without cancer at baseline and with available data on height and weight. MEASUREMENTS: BMI categorized by quintiles and by the 1998 National Heart Lung and Blood Institute (NHLBI) BMI classification. Poisson regression was used for multivariate analyses relating BMI to number of hospitalizations, adjusting for sex, age, smoking status, and heart disease. RESULTS: During 20464 years of observation, 1199 individuals had 4096 hospitalizations and 34190 hospital days. Individuals in the lowest BMI quintile had a higher risk of hospitalization than those in the middle BMI quintile (RR=1.22; 95% confidence interval=1.1-1.4); stratified analyses by age revealed that this association remained for individuals aged 65 to 75. Using the NHLBI classification, underweight, overweight, mild obesity, and moderate to severe obesity were related to higher risk of hospitalizations than normal BMI in individuals aged 65 to 75. In individuals older than 75, underweight, overweight, and mild obesity were not related to a higher risk of hospitalizations. Moderate to severe obesity was related to a higher risk of hospitalization in individuals aged 75 to 89, which represented only 1.5% of the sample. CONCLUSION: BMI is not a predictor of hospitalization for most individuals aged 75 and older.