Availability and accuracy of cancer and smoking data obtained from next of kin for decedents in a retrospective cohort study.

We report on the availability and accuracy of next-of-kin data on 372 decedents gathered in the conduct of a retrospective cancer incidence study. Two mailed questionnaires were followed by phone calls for nonrespondents. The overall response rate for the next of kin of decedents was 59%, markedly less than the 79% response rate for living cohort members. Validity of data provided by next of kin regarding cancer incidence and smoking habits of the decedents was assessed via a comparison with medical records, which were considered accurate for the purposes of this comparison. For the 117 men for whom we had both types of records, next of kin were generally able to report accurately both whether the decedent ever had cancer (91% agreement) and the cancer type (84% agreement). Next-of-kin data generally agreed with medical record data for number of years smoked or for number of years since quitting, whereas next of kin underreported the amount smoked by decedents. The difficulty in contacting next of kin and their frequent inability to provide detailed and accurate information on smoking may make it difficult to adjust for confounding by smoking in nested case-control studies.