Perfomance Score Extraction from Panel Data for Multi-Criteria Decision Analysis (Mcda) Using a Regression-Based Approach

Objectives: Key to effectively implementing MCDA is to ensure the performance matrix offers a valid comparison across alternatives. We propose to bring econometric methods into the field of MCDA by estimating performance scores from patient-level panel data while matching on observable patient characteristics between different treatment alternatives. The aim of this study is to demonstrate the application of these methods to a case study using outcomes of disease management programs for Cardio-Vascular Risk Management (CVRM: n= 9) and Chronic Obstructive Pulmonary Disease (COPD: n= 4), prospectively monitored over a twoyear period.
Methods: Performance scores were grouped according to the triple aim framework for the three aims of integrated care: 1) improving population health outcomes, 2) improving patient experience and 3) reducing costs. Included indicators were the EuroQol-5D, Short-Form-36, smoking and physical activity levels, the PACIC (Patient Assessment of Chronic Illness Care) and various cost measurements. Estimation was done by means of the average predicted outcomes from a generalised linear model. To increase comparability between programs a multinomial generalisation of propensity score matching (PSM) was applied.
Results: Differences between the estimated treatment effects were expected based on the comprehensiveness of their interventions, e.g. estimated smoking rate was 28% in our most comprehensive COPD program, compared to 38% in our least significant program. PSM influenced the results, especially for costs and to a lesser extent for different dimensions of the EuroQol-5D. Overall, CVRM programs were more susceptible to changes resulting from PSM, which may be attributed to the higher number of programs.
Conclusions: The proposed econometric methods offer a novel way to estimate performance scores from outcomes data in disease management programs. The estimated performance matrix offers useful distinctiveness between criteria and programs as an input for follow-up studies which further explore the performance matrix or attach relative weights to each performance indicator to allow a formal MCDA.