INTRODUCTION:

Specialty medications constitutes a large portion of healthcare spend and it is important for payers to understand, measure and manage it. We used time series analysis to predict specialty drug spending.

OBJECTIVES

To forecast specialty drug spend for 3-year horizon of a prospective large employer group with 1 year of specialty drug spend history

METHODS

Univariate time series models (UTM) were used to predict future medical and pharmacy spend per member per year (PMPY). Specialty spend (both medical and pharmacy) data of members enrolled under commercial plans was analyzed from 2005 to 2019. Employers providing 15 years of continuous medical and pharmacy coverage to their members were included in the study. Average cost PMPY at employer level were used to calculate yearly rates to forecast future spend of employers in 2020, 2021, 2022. Cost PMPY was deflated using Construct index of cyclically sensitive inflation (CSI) values taken from US Bureau of Labor Statistics.

RESULTS

UTM were fitted using 2005 to 2016 data as training dataset and data from 2017 to 2019 as validation dataset. Model resulted in 6.66% and 4.34% Mean Absolute Percentage Error (MAPE) for Medical and Pharmacy cost PMPY series, respectively. Around 88% of Employers with more than 10,000 members resulted with less than 40% (3-year average) MAPE in case of Medical and 100% in case of pharmacy. 1-year MAPE was less than 40% for about 86% of employer groups for medical data and 100% for pharmacy. The average yearly forecasted rate change was 8.5% in case of medical spend and 11.7% in case of pharmacy spend.

CONCLUSIONS

Employers with greater than 10,000 employees enrolled under commercial plan performed well when yearly rates from the UTM were applied to the cost PMPY series. Due to non-static nature of employees under smaller employer groups, high volatility was observed.