OBJECTIVE: To examine the association between declining serum cholesterol and mortality in a sample of older nursing home residents.DESIGN: A retrospective cohort study.SETTING: A 203‐bed nursing home.PARTICIPANTS: Persons aged 65 and older, resident in the nursing home on January 1, 1988, or admitted through December 31, 1989, were eligible (n = 185) for the study. Follow‐up for mortality was conducted until June 30, 1991. Fifty‐five survivors with two or more cholesterol levels recorded before January 1, 1990, and the 76 decedents with two or more recorded cholesterol levels constituted the analytic sample (71% of eligible subjects).OUTCOME MEASURE: Mortality of the nursing home residents.RESULTS: Cholesterol declined 31.1 mg/dL/yr (95% confidence interval [CI], 19.7 to 42.6) among decedents, versus 4.2 mg/dL/yr (95% CI, −4.9 to 13.2) among survivors. The association between cholesterol decline (absolute or relative rates) and mortality was examined using logistic regression controlling for age, sex, and tube feeding. Compared with a referrent group with no change or increase, declining cholesterol greater than 45 mg/dL/yr was accompanied by an adjusted relative odds for death of 6.2 (95% CI, 2.1 to 18.4); declining cholesterol greater than 20% per year was accompanied by an adjusted relative odds for death of 7.3 (95% CI, 2.4 to 22.2). Extreme declines greater than 20% per year occurred in 47% of decedents but in only 15% of survivors.CONCLUSION: Precipitously declining cholesterol appeared to be a marker for mortality in the sample and may help explain the low cholesterol‐mortality association in older nursing home residents.J Am Geriatr Soc 44:31