Publications

Private equity (PE) acquisitions of opioid treatment programs (OTPs) are growing, with the potential to expand access to methadone, a critical yet underused medication that can cut the risk for overdose deaths by more than half. At the same time, PE's emphasis on short-term profitability has raised concerns from policy makers that PE acquisitions can consolidate ownership of OTPs among financial firms without expanding access to treatment. Using a difference-in-differences design with novel data on PE acquisitions of OTPs and methadone shipments to all OTPs during the period 2006-19, this study examined the effects of PE acquisitions on methadone supply. PE firms acquired 67 percent of the OTP market in the median county with any acquisition, often through multiple acquisitions within the same county. After acquisition, methadone shipments to PE-acquired OTPs increased by 13 percent relative to matched controls, but this was not statistically significant after adjustment for differential preacquisition trends, indicating that the increase was not driven by the acquisition itself. County-level methadone shipments and opioid mortality remained unchanged. Findings suggest that PE acquisitions of OTPs may consolidate ownership of OTPs among financial investors without changing methadone supply. Given policy makers' widespread call for increased supply, additional scrutiny of the impact of PE investment on patient access to methadone might be warranted.

BACKGROUND: Chronic diseases such as diabetes are a major burden to the US health care system. High medication adherence helps improve diabetes outcomes and reduce cost. Cost of medications can contribute to nonadherence. Use of a formulary decision support system with e-prescribing may be associated with greater use of generic medications, leading to lower costs and better adherence. A real-time prescription benefit (RTPB) solution provides patient-specific drug pricing, benefit information, and therapeutic options to choose the most cost-effective and clinically appropriate treatment.

OBJECTIVES: To examine whether RTPB is associated with increased adherence measured by proportion of days covered, higher utilization of generics, and generic dispensing rate? Is RTPB associated with lower plan and patient out-of-pocket (OOP) per-user per-month costs?

METHODS: This study used a retrospective, matched intervention-control analysis of commercial health plan members from a large pharmacy benefits manager. Members were eligible for inclusion if they initiated therapy between January and August 2021. Members were excluded if they were not continuously eligible for coverage over the study period. Members who initiated diabetes therapy with a prescriber using RTPB (intervention) were compared with those new to therapy with a prescriber not using RTPB (control). Index date for both samples was the first medication prescription in the index period. Members were matched on age and sex demographics. The evaluation period lasted 12 months after index date. Multivariable linear regression models were used to assess the impact of an RTPB program on adherence and proportion of prescriptions filled with a generic. A generalized linear model (gamma distribution, log link) estimated plan and OOP patient costs, whereas a generalized linear model model with the Poisson distribution was used to estimate the number of controlling for patient age, sex, social determinants of health score, and other patient- and plan-level covariates.

RESULTS: 1,302 matched pairs were included in the analysis. Findings show the proportion of days covered was 68.7% for control and 71.4% for RTPB members (P < 0.05). The average number of generic prescriptions for control and RTPB samples were 4.06 and 5.66, respectively (P < 0.05) and the generic dispensing rates were 44.9% and 60.1%, respectively (P < 0.05). The mean plan cost per member per month for diabetes medications, for the non-RTPB group, was 32.3% higher than the RTPB sample (a difference of $81.69, P < 0.0001) and the mean patient cost per month was 88.8% higher than the RTPB sample (a difference of $9.71, P < 0.0001).

CONCLUSIONS: Access to RTPB tools provides prescribers with formulary benefit and therapeutic options that allow them to provide the lowest-cost clinical treatment, thus improving adherence, increasing use of generic medications, and lowering plan and patient OOP costs.

BACKGROUND: Amid the opioid epidemic in the United States, hepatitis C virus (HCV) infections are rising, with one-third of individuals with infection unaware due to the asymptomatic nature. This study aimed to develop and validate a machine learning (ML)-based algorithm to screen individuals at high risk of HCV infection.

METHODS: We conducted prognostic modeling using the 2016-2023 OneFlorida+ database of all-payer electronic health records. The study included individuals aged ≥18 years who were tested for HCV antibodies, RNA, or genotype. We identified 275 features of HCV, including sociodemographic and clinical characteristics, during a 6-month period before the test result date. Four ML algorithms-elastic net (EN), random forest (RF), gradient boosting machine (GBM), and deep neural network (DNN)-were developed and validated to predict HCV infection. We stratified patients into deciles based on predicted risk.

RESULTS: Among 445 624 individuals, 11 823 (2.65%) tested positive for HCV. Training (75%) and validation (25%) samples had similar characteristics (mean, standard deviation age, 45 [16] years; 62.86% female; 54.43% White). The GBM model (C statistic, 0.916 [95% confidence interval = .911-.921]) outperformed the EN (0.885 [.879-.891]), RF (0.854 [.847-.861]), and DNN (0.908 [.903-.913]) models (P < .0001). Using the Youden index, GBM achieved 79.39% sensitivity and 89.08% specificity, identifying 1 positive HCV case per 6 tests. Among patients with HCV, 75.63% and 90.25% were captured in the top first and first to third risk deciles, respectively.

CONCLUSIONS: ML algorithms effectively predicted and stratified HCV infection risk, offering a promising targeted screening tool for clinical settings.

BACKGROUND: Cross-sectional studies have shown that antidepressants (ADs) and benzodiazepines (BZDs) are commonly co-prescribed for depression, potentially increasing the risk of falls and related injuries (FRI) compared to monotherapies. However, little is known about the longitudinal dosing patterns (i.e., trajectory) of ADs and BZDs and their associated FRI risk.

METHODS: This retrospective cohort study used group-based multi-trajectory models to identify AD-BZD trajectories among older Medicare fee-for-service beneficiaries with depression initiating ADs with/without BZDs. We measured the standardized daily doses of AD and BZD within 84 days after AD initiation and categorized them into negligible, very-low, low, moderate, high, or very-high levels with a discontinuing, declining, increasing, or stable trend. Then, we assessed the subsequent 12-month FRI risk associated with each trajectory.

RESULTS: Among 102,750 eligible beneficiaries, the mean age was 75.5 years (SD = 7.5); 67.0% were female, 81.2% were White, and 4.9% experienced an FRI. We identified 12 distinct AD/BZD trajectories, of which 79,424 patients received AD monotherapy, and 23,326 patients received both ADs and BZDs. Compared with Group A (low discontinuing AD; 17.3% of the cohort; FRI crude incidence rate = 99.7/1000 person-year), trajectories with a higher dose or a longer duration of AD use were associated with an increased FRI risk, regardless of BZD use. The hazard ratios (HR) and 95% confidence intervals (CI) for Groups B (low declining AD; 31.0% of the cohort), C (moderate increasing AD; 23.5%), and D (high increasing AD; 5.4%) were 1.11 (1.04-1.19), 1.24 (1.16-1.32), and 1.29 (1.16-1.42), respectively. Combining ADs and BZDs at very-low doses or with declining trends did not significantly alter FRI risk compared to AD monotherapy. However, FRI risk increased when BZDs were used at low doses (either with stable or increasing trends). The HR and 95%CI for Groups J (moderate increasing AD/low stable BZD, 1.3%) and L (very-high increasing AD/low-dose increasing BZD) were 1.71 (1.41, 2.08) and 1.96 (1.53, 2.49), respectively.

CONCLUSIONS: We observed a dose-response relationship between AD use and FRI risk, independent of BZD use, highlighting the importance of initiating ADs at the lowest effective dose and closely monitoring to prevent FRI.

Orthostatic hypotension (OH) is a common inpatient condition associated with falls, syncope, and mortality. However, standardized approaches for inpatient management of OH are lacking and may vary across clinical specialties. In this retrospective observational cohort study, we reviewed the electronic medical records of patients admitted to Beth Israel Deaconess Medical Center between April 1, 2015 and June 1, 2021 with a diagnosis of OH or medication-related hypotension. Variables of interest included admitting service, presenting symptoms, suspected etiology, and management. Among the 400 inpatients with OH, one-third had OH documented on admission. Dizziness and lightheadedness were the most common symptoms; medical patients experienced dizziness, falls, and other symptoms more frequently than surgical patients. Volume depletion and medications were the leading suspected causes of OH. Surgical patients were less likely to have medication-related OH and were more likely to lack an identified etiology. Cardiovascular disease was more frequently implicated in cardiology patients. Volume depletion, neurodegenerative disease, and other conditions were more often suspected among medical patients. Management commonly involved volume resuscitation and medication adjustment, though medication changes were less frequent in surgical patients. Nonpharmacologic interventions were more common among medical patients. By discharge, OH had resolved in only one-third of patients. In summary, inpatient OH was most often identified after admission, attributed to hypovolemia, treated with fluids, and unresolved at discharge, with differences in symptoms, etiology, and management between specialties. Prospective studies are needed to formalize diagnostic and treatment strategies for OH in the hospital setting.

OBJECTIVE: Determine whether a Medicare 7-day limit on initial opioid prescriptions (effective January 1, 2019) was associated with reductions in duration, dosage, and subsequent opioid fills among post-operative adults with disabilities.

BACKGROUND: Post-operative adults with disabilities are at increased risk for uncontrolled pain and extended opioid use.

METHODS: We identified adults with disability entitlement aged 18-64 from national Medicare Advantage health plan claims data who underwent common surgeries between July 2016-June 2021. We used a repeated cross-sectional interrupted time series design to examine changes in opioid prescribing associated with the 2019 7-day limit.

RESULTS: Sample included 24,910 member-index months (mean age [SD] 55.5 [8.0] years; 14,413 [57.9%] female) representing 24,283 members. The 7-day limit was associated with an 11.8 percentage point (pp) reduction (95% CI -13.3, -10.2) in the likelihood of an initial fill >7-day supply that remained lower than predicted (-5.7 pp, 95% CI -7.6, -3.7) by June 2021. Cumulative 30-day MME was lower than predicted (-35.9 MME, 95% CI -53.7, -18.1) in January 2019 but higher than predicted by June 2021 (95.2 MME, 95% CI 56.9, 133.6). The likelihood of ≥1 fills within 30-days of the initial fill was not significantly different in January 2019 but was 6.9 pp higher (95% CI 3.1, 10.7) than predicted by June 2021.

CONCLUSION: The 2019 Medicare 7-day limit was associated with immediate large reductions in initial opioid duration and dosage. Increased dosage observed in June 2021 may have been mediated by an increase in subsequent opioid prescriptions after the initial fill.

INTRODUCTION: Early detection and intervention are crucial for reducing the impacts of depression and associated healthcare costs. Few studies have used electronic health records (EHR) and machine learning (ML) with a longitudinal design to predict depression onset. We developed and validated ML algorithms using EHR to identify patients at high risk for the onset of diagnosis-based major depressive disorder (MDD) in primary care settings.

METHODS: Using a prognostic modeling approach with retrospective cohort study design, we identified patient visits in primary care settings for individuals aged ≥18 years from the Accelerating Data Value Across a National Community Health Center Network Clinical Research Network 2015-2021 data. We measured 267 features at six-month intervals starting six months prior to the first encounter. We developed algorithms using Least Absolute Shrinkage and Selection Operator (LASSO), random forest, and XGBoost with 10-fold cross validation. Using hold-out testing data, we measured prediction performance (e.g., C-statistics), stratified patients into decile risk subgroups, and assessed model biases.

RESULTS: Among eligible 1,965,399 individuals (mean age = 43.52 ± 16.04 years; male = 35%; African American = 20%) with 4,985,280 person-periods, the MDD onset rate was 1% during the study period. XGBoost performed similarly to other models and had the fewest predictors, (C-statistic = 0.763, 95% CI = [0.760, 0.767]). XGBoost had a 66.78% sensitivity, 74.19% specificity, and 2.55% positive predictive value at the balanced threshold identified using Youdan Index. The top three risk decile subgroups captured ∼70% of MDD cases, without significant racial or sex biases.

CONCLUSIONS: An ML algorithm using EHR data can effectively identify individuals at high risk of depression onset within the subsequent six months, without exacerbating racial or sex biases, providing a valuable tool for targeted early interventions.