[Prediction of 29-years mortality rate from initial and emerging risk factors during follow-up in a Warsaw population].

Searching for clues through cohort studies requires introducing initial values of predictors likely to forecast future morbid events. This may put the end points on unequal footing, due to different timing of preceding exposures. A long follow-up mortality study of an urban population with repeated measurements in the midst of the observation enabled splitting predictors of death into time-related components making comparisons more alike with regards to time. A random sample of 1898 Warsaw adult population was interviewed in 1970 on a variety of demographic, social, behavioural and morbid circumstances as well as the use of health services. The 29-year-long search for death events was instituted with 751 (39.5%) deaths ascertained till the end of 1999 and 166 persons not traced (8.7%). In 1979 the inquiry was repeated in 1366 respondents alive and ready to take part, with the same set of questions asked. Beyond base-line values of traits in question, the estimates of change (possible predictors both times; appearance or vanishing of a trait; no trait whatsoever; no answer) were arranged as dummy variables tested for creating an excess risk against double absence of exposure. Cox multiple regression models were used for assessing relative odds of death. The base-line '70 predictions limited the risk of death for women mainly to non-structured evidence of disease whereas among men they included social circumstances, smoking & bronchitis, and possession of a personal doctor. With the '79 base-line recent hospital discharge was added to the prediction of remaining ('80-'99) mortality experience of both sexes, and disability appeared as a risk factor in men. Using time-related variables disclosed the mortality-boosting influence of twice-confirmed smoking in both sexes, as well as the excess death following the so far absent protracted illness among men and women. The appearance of the excess risk of death due to addition of hitherto neutral exposures was more frequent among men than women. The same pattern of the new exposures emerged for recent hospital discharge vs. no hospitalization. It seems that arranging exposure in the time-related fashion discloses a newly evolving excess risk of death which may pave the way towards death in more vulnerable fraction of a population in largely non-specific fashion.