Electronic Letters to:
|
|
eLetters: Electronic letters published:
-
![[Read eLetter]](/icons/misc/sectionDOWN.gif)
Design issues in case-control studies
- K. S. Jacob (28 March 2007)
|
|
|||
|
K. S. Jacob, Professor of Psychiatry Christian Medical College, Vellore 632002 India
Send letter to journal:
ksjacob{at}cmcvellore.ac.in K. S. Jacob
|
The survey of case-control studies in psychiatry assessed selection and information bias in recently published investigations (Lee et al, 2007). Nevertheless, the principles for the valid selection of subjects are rarely highlighted in psychiatric literature. A case-control study can be viewed as a census of the base population to identify all cases or particular outcomes and ascertain determinants, modifiers, and confounders (Miettinen, 1985). A sample of the base population is then employed to also obtain information about such factors for comparison. Consequently, the design is also called the census-sample, case-base or case-referent approach. The base population can be defined a priori. Such a primary definition, as in a community based case-control study, would imply that the cases of interest are defined, secondarily, as the entire cases arising from the defined population base. Alternatively, the cases can be defined a priori as those appearing in a particular hospital over a particular span of time. The population base in such studies is secondary to the case selection. The challenge would then be to sample the population base of the particular hospital, often hypothetical, in order that the controls are representative of the base population from which the cases were selected. The overriding principle of subject selection in case -control studies is that the case series and the referent sample (controls) be representative of the same base experience (Miettinen, 1985). The failure to appreciate this principle results in the common error in the design that such studies often compare cases with non-cases. Such misunderstanding also results in choosing general population controls, normal/healthy controls, staff/students, patients’ relatives, etc. In addition to representativeness of the case and controls to the same population base is the issue of comparability. It implies the accuracy of information about the determinants under study. For example, if the setting and or illness per se influence the history/ providing information for cases, it is necessary to assure similar influences for “controls” so that they are comparable. This is done by careful choice of the control group which is comparable to the case group for the provision of such information (E.g. controls for studying a particular congenital malformation will necessarily have to be a series of another congenital malformation). The survey assessed published papers on selection and information bias (Lee et al, 2007). However, the criteria employed to assess selection and information bias fall far short of the concept of representativeness and of comparability. Ideas and words are two dimensions for issues related to case-control studies (Spitzer, 1985). Terminology should not stifle the concepts related to design. K. S. Jacob Department of Psychiatry, Christian Medical College, Vellore 632002 India ksjacob@cmcvellore.ac.in Conflict of interest: None References Lee, W., Bindman, J., Ford, T., et al. (2007) Bias in psychiatric case- control studies: Literature survey. British Journal of Psychiatry, 190, 204-209. Miettinen, O.S. (1985) The “case-control” study: valid selection of subjects. Journal of Chronic Disease, 38, 543-48. Spitzer, W.O. (1985) Ideas and words: Two dimensions for debates on case controlling. Journal of Chronic Disease, 38, 541-42. |
|||