Charakterystyka, wykorzystanie zasobów opieki zdrowotnej i koszty w grupach pacjentów z epilepsją ubezpieczonych przez Medicaid: eksploracyjne podejście z wykorzystaniem uczenia maszynowego
Characterization, healthcare resource utilization, and costs of health equity clusters of Medicaid-insured patients with epilepsy: An exploratory machine learning approach
W skrócie
Badanie identyfikowało grupy pacjentów z epilepsją ubezpieczonych przez Medicaid, różniące się dostępem do nowoczesnych leków przeciwpadaczkowych w zależności od rasy, pochodzenia i regionu zamieszkania. Okazało się, że pacjenci z większymi barierami w dostępie do leków mieli więcej wizyt u lekarzy, więcej pobytów w szpitalu i wyższe koszty leczenia. Wyniki sugerują, że ograniczenia w dostępie do leków wpływają na koszty leczenia niezależnie od wieku, płci czy liczby przyjmowanych leków.
Oryginalny abstract (angielski)
OBJECTIVE: Identify health equity clusters among Medicaid enrollees and describe their characteristics and economic burden. METHODS: De-identified data of Medicaid-insured adults with epilepsy prescribed ≥ 1 antiseizure medication (ASM) on/after initial diagnosis (first ASM=index), with ≥ 12 months' continuous medical/pharmacy benefits pre/post index, were analyzed from an all-payer claims database (01/01/2014-06/30/2021). Patients were clustered using machine learning/K-prototypes by variables (number of third-generation ASMs with formulary restrictions, brivaracetam [BRV] coverage status, race, ethnicity) selected to assess formulary impacts/health equity. Euclidean distance/simple matching was used for continuous/categorical data; elbow plots identified the optimal cluster number. Demographics/characteristics and 12-month follow-up healthcare resource utilization (HCRU)/costs were examined. RESULTS: Five clusters were identified (N = 24,722): (1) mostly White, South region, easy access to third-generation ASMs (44.1%); (2) mostly Black, South region, easy access to third-generation ASMs (12.7%); (3) mostly White, North Central region, some third-generation ASM and high BRV access barriers (19.1%); (4) almost two-thirds Black, Northeast region, high third-generation ASM and some BRV access barriers (6.7%); (5) mostly White, North Central/Northeast region, high third-generation ASM and low BRV access barriers (17.3%). Clusters 1-2 were considered 'average' access barriers; 3-4 'high'; 5 'intermediate.' Clusters were generally similar in age/sex/ethnicity/ASM use. Over follow-up, cluster 3 had highest inpatient/outpatient HCRU and total costs. Prescription number/cost was highest in cluster 4, then cluster 3. Higher access restrictions patients generally had more prescriptions, outpatient/other visits, and higher costs. CONCLUSIONS: Results suggest a relationship between formulary restrictions and economic burden, seemingly independent of age, sex, geographic region, and treatment utilization.