EFFECTIVENESS OF MULTIDISCIPLINARY CARE FOR PATIENTS WITH DIABETES IN MEXICO: AN APPLICATION OF QUASI-EXPERIMENTAL METHODS

Purpose:

   The Ministry of Health recently implemented a set of facilities of multidisciplinary care for patients with diabetes mellitus (DM) nationwide in response to the poor results in glycemic control that their traditional model of care had produced. However, the effectiveness of these facilities on has not been evaluated. Our aim is to evaluate the short-term impact of these facilities on controlling DM patients in Mexico.

Method:

   We used longitudinal data from two datasets of patients with diabetes, one from medical records of patients receiving care at multidisciplinary health facilities, and a survey from MoH primary-level units as counterfactual. We applied a quasi-experimental approach to estimate the average treatment effects (ATE). Sample size of patients exposed to multidisciplinary care was 961 patients and 418 controls. The outcome of interest was glycated hemoglobin (HbA1c), the exposure time was 12 months, and we applied nearest-neighbor propensity score matching methods (PSM) to reduce selection bias given the non-random assignment to the treatment and control groups. The vector of variables that were included to balance the two groups at baseline were gender, age, years with diabetes, body mass index (BMI), systolic blood pressure (SBP), density of lipoprotein (HDL) and total cholesterol. We tested different matching algorithms and the main results did not vary.

Result:

   Average baseline levels of HbA1c were 9.13% for the treated and 9.28% for the controls, with no differences between groups (t-stat=1.03). However, in the raw sample the treated individuals were younger (54.3 vs 57.1 yrs), had more time with DM (10.5 vs 7.5 yrs), larger BMI (31.6 vs 30.5), lower SBP (124.6 vs 139.1 mm Hg), larger HDL concentrations (43.5 vs 41.5 mmol/mol) and total cholesterol (195.1 vs 186.9 mmol/mol). At follow-up, we observed a reduction of 1.6 pp of HbA1c on the unmatched sample (t-stat=-12.95), in contrast we found an ATE reduction of 2.1 pp (z=-13.11) on the matched sample. 

Conclusion:

   We found evidence of causal effects on glycated hemoglobin in patients exposed to a multidisciplinary disease-management program. The magnitude of this reduction can yield substantial health benefits. In order to inform policy makers on the value-for-money of a novel intervention experience in Mexico aimed at reducing diabetes-related complications and mortality, our results should be complemented with information on costs and long-term outcomes.