OBJECTIVES: Literature and analysis of real-world health costs show that a small group of patients consumes the large amount of health care resources. Proactive approach requires an accurate identification of high-cost users (HCUs) that can benefit from case management, stabilizing or reducing their costs before their health get worse.

METHODS: Cohort of 3,755 patients considered all the beneficiaries attended by the health plan during FY2017. Patients were listed in descending order of total expenses, and the top 5% were classified as HCUs spent 60% of the resources, than a binary variable was created to identify them. Potential factors that could influence HCUs classification includes demographic (age, sex), clinical (ICD-10) variables, with separation of specific chronic conditions (e.g. diabetes), rheumatologic diseases using immunobiologicals, cancer chemotherapy, number of visits, emergency admissions, avoidable hospitalizations and other causes, as well as participation in a Pharmacy Benefit Management (PBM) program. When necessary, continuous variables were converted into categorical. Thereafter, a number of variables were reduced using the variable clustering technique. Logistic regression was applied in the cohort of patients. Odds ratios and significance of the parameter (p values <0.05) were evaluated.

RESULTS: The applied model allowed to identify with high assertiveness the HCUs eligible for care coordination, as well who were not, for now, eligible and shall be excused from the program. Through this model, we identify which variables have the highest predictive power. Diagnosis information, health plan utilization and medicine consumption has significantly increased predictive power of demographic variables.

CONCLUSIONS: Care coordination using this model with different levels of health services, permitted personalized care interventions to high needs of HCUs. From the results found, we expect to develop an algorithm that produces a score capable to predict patients with high risk to become HCUs in the next two years.