A Markov modeling framework for estimating the effects of asthma medication adherence on healthcare use and costs in pediatric asthma

Asthma medication adherence, measured by the asthma medication ratio (AMR), has been shown to be a reliable predictor of exacerbations among children with asthma. We developed a Markov model to estimate healthcare use, costs, and utility-associated outcomes across AMR-defined asthma states among children with asthma. We constructed an asthma Markov model using claims for a cohort of Medicaid beneficiaries aged 2-17 years (n = 214, 452) using 2013‐2014 MarketScan® data who had at least one claim for an inhaled corticosteroid. We calculated 3‐month AMRs as the basis of transition estimates in the model. The AMRs data were transformed to reflect a one‐month Markov cycle length. Our model included five states: controlled, sub-optimally controlled, inactive, IP admission, and ED. The Markov model used an overall 2‐year time horizon for 10, 000 beneficiaries. Patients entered the model in one of the five states as observed in the Medicaid data. Costs were based on 2017 Medicaid claims inflation adjusted to reflect 2020 costs. Medication-related costs were based on mean observed payments for controller and rescue medications, whereas costs for ED visits and inpatient admissions were based on median observed payments for those acute events. Utilities for each state were based on existing literature. Monthly costs for children with asthma were 235 for controlled, 35 for suboptimally controlled (attributable to fewer medication fills), 364 for months with ED visit, and 6, 229 with admission. Over a 2-year period, we estimated the 2‐year direct medical costs of asthma among Medicaid insured children to average 4, 552 per child. Children are on average 1. 6 months more in controlled state at a cost of 0. 89 more per child per month accounting for medication cost less the cost of ED visits and hospitalizations. In addition to the saved side effects of oral steroids and time hospitalized, there is one increased quality month per 13 months. This Markov model populated with Medicaid population exacerbation, ED, hospital and medication use and payment data provides a framework to estimate the potential clinical and economic impact of existing and novel targeted interventions aimed at improving medication adherence and reducing acute care use.