Statistical problems associated with cohort studies of workers in an industry

Risk factors in a working environment are studied mainly by case control studies or cohort studies of workers in several occupational settings. While the cohort studies are by far more powerful in detecting risk factors there are many problems associated with these studies, apart from the high cost and long duration. Some of these problems include the large number of variables involved, the evaluation of missing observations and the resulting bias, problems in selecting appropriate statistical models and analyses as well as identifying the disease subgroup (including loss to follow up). The problems are considered under four groupings: (1) identifying outcome, (ii) missing data, (iii) estimation of work exposure and (iv) statistical analysis. The apparent advantages of retrospective cohort study should be weighed against these problems when deciding on a study design. More appropriate statistical methods, other than SMR and PMR analysis, should be considered while analysing data from such studies.