A Markov Transition Model for Estimating the Impact of Pediatric Asthma Medication Adherence on Healthcare Utilization and Cost

Research Objective Medication adherence, measured by the asthma medication ratio (AMR), has been shown to be a reliable predictor of exacerbations among asthmatic children. We constructed a Markov model examining the effects of optimal vs suboptimal control using a 3‐month rolling AMR on severe exacerbations resulting in an emergency department (ED) visit or inpatient (IP) admission. Study Design We constructed a Markov state transition model by analyzing pharmacy and medical claims for a cohort of children with asthma. After identifying children with at least one claim for an inhaled corticosteroid, we calculated rolling 3‐month period AMRs. AMR is defined as follows: [# controller medications/(# controller medications + # rescue medications)]. Medication adherence was classified as controlled (AMR ≥ 0.5), suboptimally controlled (AMR < 0.5), or inactive (no asthma claims in the preceding 3 months). Empirical monthly Markov state transitions were extracted for a 12‐month period, following the initial development of a 3‐month rolling AMR, and averaged to yield final transition estimates. Our model included five states: controlled, suboptimally controlled, inactive, IP admission, and ED visit. A one‐month cycle length was used with a 2‐year time horizon using 10,000 simulated patients. Patients entered the model in one of the five states using the empirical distribution at month 5 in the data. Costs for the model were extracted using 2017 Medicaid claims data. Asthma medication costs based on medication category (controller or rescue) along with median costs for an exacerbation resulting in an ED visit or inpatient admission were examined and used in the model. Utilities for each state were informed by prior literature. Population Studied Medicaid insured children aged 2‐17 years (n = 214,452) using 2013‐2014 IBM Truven Health Medicaid data who had asthma and at least one claim for an inhaled corticosteroid. Principal Findings Monthly costs for children with asthma were found to be $214 for controlled, $32 for suboptimally controlled (attributable to fewer medication fills), $565 for months where an ED visit occurred, and $5,883 when an admission occurred. We found over the 24‐month simulated cycle, children spent on average 20.86 months controlled, 0.43 suboptimally controlled, 0.02 in the ED, 0.01 IP, and 2.68 months in an inactive state. Children with inactive asthma had transition probabilities to the ED or IP somewhere between children who were controlled and suboptimally controlled. Suboptimally controlled children had much higher rates overall of ED or IP visits, yielding higher costs and opportunity for interventions. 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. Conclusions This research provides the groundwork for testing the cost effectiveness of potential interventions for suboptimally controlled asthmatic children—such as school‐based telehealth aimed at improving medication‐taking and symptom control, resulting in improved quality of life for children with asthma. Identification of these patients can be made using SureScripts or other real‐time pharmacy fill data. Implications for Policy or Practice These empirical Markov transition estimates and cost estimates provide a framework to test the cost effectiveness of existing and novel targeted interventions aimed at improving the quality of life of asthmatic children. Primary Funding Source National Institutes of Health.