Editorial Commentary: Quasi-Experimental and Interrupted Time-Series Design.

Abstract

In this issue of the Journal of the Pediatric Infectious Diseases Society, Newland et al use a quasi-experimental study design. This study design, also known as a nonrandomized, postintervention design, is beneficial in infectious disease studies where complete control over the allocation and/or the timing of the intervention is not possible or is not ethical [1]. Scenarios in which randomization of the intervention is not possible include interventions implemented at a level (eg, physician, unit, hospital) other than the patient level, and those that require immediate implementation. In the first scenario, a physician who wants to study the safe transfer of patients from one unit to another would need to implement the intervention at multiple levels: the physician level, the unit level, and then at the hospital level. If the intervention in this scenario were an educational workshop on safe transfers, one would need to assume that this information would not be disseminated from the physicians who were officially “trained” and those that weren’t. This is not an appropriate assumption, as patient transfers affect multiple physicians in multiple units at multiple time points. In situations such as a meningococcal meningitis outbreak where an intervention such as antibiotic prophylaxis is needed quickly and an untreated control group (as in a randomized control trial) is unethical, a retrospective quasi-experimental design can be used to study the efficacy of the intervention. One of the most basic quasiexperimental designs includes a pretest and posttest with no control group. This design is depicted below; O1 represents the pretest observations, X represents the intervention, and O2 represents the posttest observations after the intervention has been implemented [2]: