Background: Ensuring that evidence based medicine reaches patients with diabetes in the US and internationally is challenging. The chronic care model includes evidence based management practices which support evidence based care. However, despite numerous studies, it is unclear which practices are most effective. Few studies assess the effect of simultaneous practices implemented to varying degrees. The present study evaluates the effect of fifteen practices applied concurrently and takes variation in implementation levels into account while assessing the impact of diabetes care management practices on glycemic and lipid monitoring.
Methods: Fifteen management practices were identified. Implementation levels of the practices in 41 medical centres caring for 553,556 adults with diabetes were assessed from structured interviews with key informants. Stepwise logistic regression models with management practices as explanatory variables and glycemic and lipid monitoring as outcome variables were used to identify the diabetes care practices most associated with high performance.
Results: Of the 15 practices studied, only provider alerts were significantly associated with higher glycemic and lipid monitoring rates. The odds ratio for glycemic monitoring was 4.07 (p < 0.00001); the odds ratio for lipid monitoring was 1.63 (p < 0.006). Weaker associations were found between action plans and glycemic monitoring (odds ratio = 1.44; p < 0.03) and between guideline distribution and training and lipid monitoring (odds ratio = 1.46; p < 0.03). The covariates of gender, age, cardiac disease and depression significantly affected monitoring rates.
Conclusions: Of fifteen diabetes care management practices, our data indicate that high performance is most associated with provider alerts and more weakly associated with action plans and with guideline distribution and training. Lack of convergence in the literature on effective care management practices suggests that factors contributing to high performance may be highly context-dependent or that the factors involved may be too numerous or their implementation too nuanced to be reliably identified in observational studies.