To examine the impact of changes in comorbidity--as measured by the Charlson comorbidity index--on self-rated health in a large sample of community-dwelling elderly over a 1-year period, and to examine the differential effects of changes in specific Charlson diagnostic categories. Longitudinal survey data on self-rated health were linked with Medicare inpatient, outpatient, and physician visit data for 30,535 U.S. elderly residing in Pennsylvania. Multivariate logistic regression with fractional polynomials was used to model relationships involving baseline and changing Charlson comorbidity with self-rated health decline, and to evaluate covariate interactions. Comorbidity change was associated with greater likelihood of worsened self-rated health, but the relationship was nonlinear and was moderated by age and baseline comorbidity. The impact of comorbidity change appeared to be less among older individuals and those with higher baseline comorbidity. Declines in self-rated health were most likely following new diagnoses for metastatic tumors, paralysis, and dementia. Self-rated health is responsive to changes in Charlson comorbidity, but nonlinearity and interactions suggest complexity in how elderly respond to comorbidity change. Younger individuals and those with initially low comorbidity are more likely to reduce self-ratings of health following new diagnoses for chronic conditions.