Statistical methods for discrete data
were used to analyze disability onset. The data were considered discrete because disability was
monitored at 2-year intervals rather than continuously. Survival analysis for interval-censored
data28 provides survival curves for discrete data analogous to Kaplan-Meier curves for continuous data. For
each activity, an interval was determined for the onset of disability. For example, a 70-year-old
person free of walking disability at baseline but reporting disability 2 years later had an onset
interval of age 70 to 72 for walking disability. Similarly, a 70-year old person who entered the
study with walking disability was given an onset interval of age 0 to 70. Individuals who reported
no disability or who died before disability occurred contributed censored observations. The age
distributions of disability onset were estimated with the SURVIVAL module of SYSTAT.29