
BACKGROUND Medication therapy management (MTM) is a core component of managed care pharmacy that improves medication safety, adherence, and clinical outcomes in high-risk populations. However, traditional eligibility criteria lack precision, limiting scalability and impact. OBJECTIVE To provide real-world evidence of how machine learning (ML) integration into pharmacy workflows can improve MTM delivery by identifying patients most likely to engage. METHODS We conducted a retrospective cohort study using a de-identified pharmacy dataset. MTM-eligible patients in 2025 were assigned to centralized or community pharmacists via an ML-based precision-targeting model, whereas a 2024 cohort was assigned using traditional eligibility-based assignment. The propensity score matching model incorporated demographics, medication patterns, social determinants of health, engagement history, and plan-level/store-level characteristics. The primary outcome was the call favorability rate (acceptance of a pharmacist consultation). RESULTS After propensity score matching, 183,888 matched patient pairs were analyzed. The ML cohort showed a 21.2% call favorability rate vs 12.6% in the historical cohort, an 8.6% absolute increase and 68% relative improvement (risk ratio = 1.68). Results were robust to moderate unmeasured confounding (E value = 2.75). Results were statistically significant (P < 0.001) and remained stable with checks for bias. CONCLUSIONS ML-based precision targeting was associated with higher MTM call favorability compared with traditional eligibility-based approaches. These findings support integrating ML into pharmacists' workflows to improve patient engagement and resource use amid evolving reimbursement structures. The results were robust to potential hidden bias, suggesting that this approach could be scaled for use in managed care MTM programs.
BACKGROUND Schizophrenia is a disabling mental illness affecting more than 2.8 million people in the United States. Long-acting injectable antipsychotics (LAIs) require less frequent dosing and may provide a more favorable treatment option over short-acting oral antipsychotics for patients with schizophrenia with suboptimal adherence or for those who desire a more convenient dosing regimen. There are 2 ways in which payers can cover LAIs: they can be covered under the medical benefit for administration by a health care provider or under the pharmacy benefit for self-administration by the patient. Existing literature is limited to general industry commentaries related to differences in health plan billing procedures under the pharmacy and medical benefit and the impact of these reimbursement pathways to health care institutions. However, the impact of various health plan coverage models on total cost of care (TCOC) and health care resource utilization (HCRU) among commercially insured patients receiving LAIs through pharmacy vs medical benefit is unknown. OBJECTIVE To evaluate differences in TCOC and HCRU among commercially insured patients with schizophrenia who received LAIs through pharmacy vs medical benefit. METHODS Retrospective study of Blue Cross Blue Shield of Michigan and Blue Care Network commercially insured members who received LAIs from January 1, 2021, to December 31, 2024. Two cohorts were included based on the benefit billed for LAI claim(s): (A) members with at least 1 LAI claim under pharmacy, or those with the first claim billed under medical with all subsequent claim(s) billed only under pharmacy, and (B) members with LAI claims billed only under medical. Members were continuously enrolled in both benefits and had an International Classification of Diseases, Tenth Revision code for schizophrenia on any claim during the pre-index period. Members were excluded if they switched post-index from the medical benefit to the pharmacy benefit cohort, or vice versa. RESULTS After applying eligibility criteria and propensity score matching, there were 61 members in each cohort. Members in cohort B (medical benefit) had significantly higher mean direct-cause medical TCOC compared with cohort A (pharmacy benefit) (+$7,569; P < 0.023). However, after accounting for LAI spend for cohort B ($1,672) vs cohort A ($12,915), the total spend was not significantly different between the 2 groups (-$3,673; P < 0.34). Cohort B demonstrated significantly higher values in the following secondary endpoints: mean all-cause inpatient cost (+$12,633; P < 0.01), mean all-cause total medical cost (+$18,582; P < 0.02), and mean all-cause inpatient utilization (8.90 more claims per member; P < 0.006). However, similar results were observed for both groups across several key secondary endpoints, including mean all-cause pharmacy benefit costs (excluding LAI cost), mean emergency department costs and utilization, mean outpatient costs and utilization, and mean total utilization. CONCLUSIONS Higher mean direct-cause medical TCOC spend was observed for cohort B, mainly driven by inpatient spend. However, similar outcomes were demonstrated across most other endpoints, which highlights the utility of cross-benefit LAI coverage.
BACKGROUND Cardiovascular disease (CVD) accounts for the highest health care expenditures in the United States, with persistent disparities among racial and ethnic minorities. OBJECTIVE To evaluate factors contributing to differences in total health care expenditure among Hispanic, non-Hispanic Black (hereafter Black), and non-Hispanic White (hereafter White) adults aged 65 years and older with CVD. METHODS This study had a repeated cross-sectional design and used data from the Medical Expenditure Panel Survey (2014-2021). The primary outcome was the total health care direct expenditure (inpatient, outpatient, and medical and pharmacy cost). Inclusion criteria were being aged at least 65 years and having at least 1 lifetime diagnosis of CVD. Survey-weighted descriptive analyses were conducted, followed by 3 hierarchical log-linear regression models. Model 1 included race and ethnicity only. Model 2 adjusted for clinical factors (sex, age at first CVD diagnosis, CVD severity, and Elixhauser comorbidity score). Model 3 additionally incorporated socioeconomic, functional, and contextual factors, including education, income, insurance status, physical and cognitive limitations, geographic region, and survey year. RESULTS The final sample included 8,585 participants (weighted N = 11,412,569). In the unadjusted model, Hispanic and Black participants had 9% (P = 0.03) and 11% (P = 0.005) lower health care expenditures, respectively, compared with White participants. After adjustment for clinical factors, Hispanic participants had 26% lower expenditures (P < 0.001), and the Black-White difference was no longer statistically significant. In the fully adjusted model, race and ethnicity were not independently associated with expenditure; however, significant interactions were identified between race and ethnicity and sex (Hispanic women: +32%, P = 0.002), physical limitations (Hispanic: +131%, P = 0.005; Black: +118%, P = 0.04), cognitive limitations (Black: +129%, P = 0.008), and geographic region, alongside independent associations with education, income, insurance status, and physical and cognitive limitations. CONCLUSIONS Differences in health care expenditure among older adults with CVD appear to be shaped by a combination of clinical, socioeconomic, and contextual factors rather than race and ethnicity alone. These findings highlight the importance of equity-focused, value-based strategies that address underlying social determinants of health to improve access, reduce disparities in health care utilization, and potentially lower overall health care costs. Importantly, lower expenditures may reflect underutilization of necessary care rather than improved efficiency.