Correlates of baseline performance do not predict results of an intervention to improve preventive care

Abstract Background Cross-sectional analyses of baseline performance often inform the development of interventions to improve care. An implicit assumption in these studies is that factors associated with better performance at baseline may also be useful in predicting change in performance over time. Methods We analyzed data collected from 1997–2002 at 77 practices in Northeast Ohio participating in an intervention to increase evidence-based preventive services delivery (PSD). Spearman's correlation coefficients and multivariable models assessed associations between practice-level characteristics (e.g., organizational structure, objectives, climate, and culture) and baseline PSD, and with final PSD controlling for baseline values. Patterns of associations for both outcomes were inspected for overlap. Results The mean PSD rate was 36.8% (+/− 8.8%) at baseline. This measure increased by an average of 4.9% (+/− 6.3%) by the end of the intervention. Of eight practice characteristics correlated with either baseline performance or change from baseline in PSD, only two were common to both: characteristics associated with baseline PSD did not predict final PSD in multivariable models. Conclusions Correlates of baseline performance differ from those related to change in performance. Practice assessments that focus on factors associated with change may be more useful in developing and implementing interventions to improve care.