Questioning the status quo: sickness absence research so far claims more than it should!

This article is in the nature of a cautionary tale for those working alongside any group of medics who are required to collect epidemiological data, or plan based upon data collected by medical experts without having access to that data. The field of work is occupational health and the specialised area is the relationship between socio-economic variables and absence from work due to sickness. The over-riding criticism of the sickness absence literature to date is that the use of analytical statistics is poor or in most cases almost non-existent and that there is very little understanding of the epidemiological concepts of 'confounding' and 'effect modification'. An example is given of one particular statistical procedure, multi-variate hierarchical log-linear modelling, which if used by future researchers could prevent the recurrence of some of the previous problems.