[Quasi experimental evaluation of public health interventions (author's transl)]

Abstract

The classic experiment, the randomised controlled trial, is the best known and most revered of evaluation research methods. Randomization in community-based intervention trials, however, is not always possible because of ethical problems arising from with holding the experimental treatment from the control groups or the difficulties in conducting experiments in field settings which do not approach controlled laboratory conditions. In such circumstances, quasi-experimental or observational designs must be used. Two major principles are involved in using quasi-experimental methods: (1) the logic for establishing causality between treatment and effect is the same as that for randomised experiments, but the problems of assessing causality or internal validity are greater, and (2) assessment of the external validity or generalizability of quasi-experimental findings crucial to the interpretation of results. Selected quasi-experimental designs using time series and comparison groups are described with examples from public health intervention trials where threats to internal validity have been assessed by using different analytic techniques or gathering additional evidence. Quasi-experimental evaluations are most useful when opportunities exist for testing rival hypotheses concerning the internal and external validity, or the findings can be used to complement true experiments.