| Postintervention | Variance in end points due to baseline differences will yield an imprecise result unless the number of participants is large | | Risk that comparisons will be confounded by differences between departments or organisations |
| Example: RCT in which nurses were randomly assigned to act as dedicated medication nurses or continue as general nurses, with a comparison of error rates between the two groups26 No baseline measurements of error rates were made | | Example: National Evaluation of Sure Start27 Due to the timing of the Sure Start evaluation, baseline measurements of key outcomes were not possible. Comparisons of the effect of the intervention are being made between Sure Start and “Sure Start to be” communities, although group assignment was not made randomly |
| | | |
| Before and after | Allows for specific comparison of change net of any baseline differences. Enables comparisons to be made between sites that change most or least | | Controls for baseline difference possible—see text |
| Example: Evaluation of a targeted risk factor reduction plan to prevent falls in older in-patients28 A pre–post design proved to be essential as differences in the number of falls and injuries from falls between control and intervention wards were statistically significant in the preintervention period | | Example: Evaluation of nationally mandated drug use reviews in nursing homes29 Levels of inappropriate medication use in nursing homes before and after the review policy were compared to assisted living facilities. The results identified that the fall in inappropriate medication use could not be attributed to the national drug use reviews, as there was a reduction in inappropriate use in both nursing homes and assisted living facilities |