Predicting Health Care Cost Transitions Using a Multidimensional Adaptive Prediction Process.

Abstract

Managing population health requires meeting individual care needs while striving for increased efficiency and quality of care. Predictive models can integrate diverse data to provide objective assessment of individual prospective risk to identify individuals requiring more intensive health management in the present. The purpose of this research was to develop and test a predictive modeling approach, Multidimensional Adaptive Prediction Process (MAPP). MAPP is predicated on dividing the population into cost cohorts and then utilizing a collection of models and covariates to optimize future cost prediction for individuals in each cohort. MAPP was tested on 3 years of administrative health care claims starting in 2009 for health plan members (average n=25,143) with evidence of coronary heart disease. A "status quo" reference modeling methodology applied to the total annual population was established for comparative purposes. Results showed that members identified by MAPP contributed $7.9 million and $9.7 million more in 2011 health care costs than the reference model for cohorts increasing in cost or remaining high cost, respectively. Across all cohorts, the additional accurate cost capture of MAPP translated to an annual difference of $1882 per member, a 21% improvement, relative to the reference model. The results demonstrate that improved future cost prediction is achievable using a novel adaptive multiple model approach. Through accurate prospective identification of individuals whose costs are expected to increase, MAPP can help health care entities achieve efficient resource allocation while improving care quality for emergent need individuals who are intermixed among a diverse set of health care consumers.