Publications

Health system data incompletely capture the social risk factors for drug overdose. This study aimed to improve the accuracy of a machine-learning algorithm to predict opioid overdose risk by integrating human services and criminal justice data with health claims data to capture the social determinants of overdose risk. This prognostic study included Medicaid beneficiaries (n = 237,259) in Allegheny County, Pennsylvania enrolled between 2015 and 2018, randomly divided into training, testing, and validation samples. We measured 290 potential predictors (239 derived from Medicaid claims data) in 30-day periods, beginning with the first observed Medicaid enrollment date during the study period. Using a gradient boosting machine, we predicted a composite outcome (i.e., fatal or nonfatal opioid overdose constructed using medical examiner and claims data) in the subsequent month. We compared prediction performance between a Medicaid claims only model to one integrating human services and criminal justice data with Medicaid claims (i.e., integrated model) using several metrics (e.g., C-statistic, number needed to evaluate [NNE] to identify one overdose). Beneficiaries were stratified into risk-score decile subgroups. The samples (training = 79,087, testing = 79,086, validation = 79,086) had similar characteristics (age = 38±18 years, female = 56%, white = 48%, having at least one overdose = 1.7% during study period). Using the validation sample, the integrated model slightly improved on the Medicaid claims only model (C-statistic = 0.885; 95%CI = 0.877-0.892 vs. C-statistic = 0.871; 95%CI = 0.863-0.878), with small corresponding improvements in the NNE and positive predictive value. Nine of the top 30 most important predictors in the integrated model were human services and criminal justice variables. Using the integrated model, approximately 70% of individuals with overdoses were members of the top risk decile (overdose rates in the subsequent month = 47/10,000 beneficiaries). Few individuals in the bottom 9 deciles had overdose episodes (0-12/10,000). Machine-learning algorithms integrating claims and social service and criminal justice data modestly improved opioid overdose prediction among Medicaid beneficiaries for a large U.S. county heavily affected by the opioid crisis.

IMPORTANCE: There is limited information about trends in the treatment of opioid use disorder (OUD) among Medicaid enrollees.

OBJECTIVE: To examine the use of medications for OUD and potential indicators of quality of care in multiple states.

DESIGN, SETTING, AND PARTICIPANTS: Exploratory serial cross-sectional study of 1 024 301 Medicaid enrollees in 11 states aged 12 through 64 years (not eligible for Medicare) with International Classification of Diseases, Ninth Revision (ICD-9 or ICD-10) codes for OUD from 2014 through 2018. Each state used generalized estimating equations to estimate associations between enrollee characteristics and outcome measure prevalence, subsequently pooled to generate global estimates using random effects meta-analyses.

EXPOSURES: Calendar year, demographic characteristics, eligibility groups, and comorbidities.

MAIN OUTCOMES AND MEASURES: Use of medications for OUD (buprenorphine, methadone, or naltrexone); potential indicators of good quality (OUD medication continuity for 180 days, behavioral health counseling, urine drug tests); potential indicators of poor quality (prescribing of opioid analgesics and benzodiazepines).

RESULTS: In 2018, 41.7% of Medicaid enrollees with OUD were aged 21 through 34 years, 51.2% were female, 76.1% were non-Hispanic White, 50.7% were eligible through Medicaid expansion, and 50.6% had other substance use disorders. Prevalence of OUD increased in these 11 states from 3.3% (290 628 of 8 737 082) in 2014 to 5.0% (527 983 of 10 585 790) in 2018. The pooled prevalence of enrollees with OUD receiving medication treatment increased from 47.8% in 2014 (range across states, 35.3% to 74.5%) to 57.1% in 2018 (range, 45.7% to 71.7%). The overall prevalence of enrollees receiving 180 days of continuous medications for OUD did not significantly change from the 2014-2015 to 2017-2018 periods (-0.01 prevalence difference, 95% CI, -0.03 to 0.02) with state variability in trend (90% prediction interval, -0.08 to 0.06). Non-Hispanic Black enrollees had lower OUD medication use than White enrollees (prevalence ratio [PR], 0.72; 95% CI, 0.64 to 0.81; P < .001; 90% prediction interval, 0.52 to 1.00). Pregnant women had higher use of OUD medications (PR, 1.18; 95% CI, 1.11-1.25; P < .001; 90% prediction interval, 1.01-1.38) and medication continuity (PR, 1.14; 95% CI, 1.10-1.17, P < .001; 90% prediction interval, 1.06-1.22) than did other eligibility groups.

CONCLUSIONS AND RELEVANCE: Among US Medicaid enrollees in 11 states, the prevalence of medication use for treatment of opioid use disorder increased from 2014 through 2018. The pattern in other states requires further research.

STUDY OBJECTIVE: More than 17 million people have gained health insurance coverage through the Patient Protection and Affordable Care Act's Medicaid expansion. Few studies have examined heterogeneity within the Medicaid expansion population. We do so based on time-varying patterns of emergency department (ED) and ambulatory care use, and characterize diagnoses associated with ED and ambulatory care visits to evaluate whether certain diagnoses predominate in individual trajectories.

METHOD: We used group-based multitrajectory modeling to jointly estimate trajectories of ambulatory care and ED utilization in the first 12 months of enrollment among Pennsylvania Medicaid expansion enrollees from 2015 to 2017.

RESULTS: Among 601,877 expansion enrollees, we identified 6 distinct groups based on joint trajectories of ED and ambulatory care use. Mean ED use varied across groups from 3.4 to 48.7 visits per 100 enrollees in the first month and between 2.8 and 44.0 visits per 100 enrollees in month 12. Mean ambulatory visit rates varied from 0.0 to 179 visits per 100 enrollees in the first month and from 0.0 to 274 visits in month 12. Rates of ED visits did not change over time, but rates of ambulatory care visits increased by at least 50% among 4 groups during the study period. Groups varied on chronic condition diagnoses, including mental health and substance use disorders, as well as diagnoses associated with ambulatory care visits.

CONCLUSION: We found substantial variation in rates of ED and ambulatory care use across empirically defined subgroups of Medicaid expansion enrollees. We also identified heterogeneity among the diagnoses associated with these visits. This data-driven approach may be used to target resources to encourage efficient use of ED services and support engagement with ambulatory care clinicians.

Cost sharing in traditional Medicare can consume a substantial portion of the income of beneficiaries who do not have supplemental insurance from Medicaid, an employer, or a Medigap plan. Near-poor Medicare beneficiaries (with incomes more than 100 percent but less than 200 percent of the federal poverty level) are ineligible for Medicaid but frequently lack alternative supplemental coverage, resulting in a supplemental coverage "cliff" of 25.8 percentage points just above the eligibility threshold for Medicaid (100 percent of poverty). We estimated that beneficiaries affected by this supplemental coverage cliff incurred an additional $2,288 in out-of-pocket spending over the course of two years, used 55 percent fewer outpatient evaluation and management services per year, and filled fewer prescriptions. Lower prescription drug use was partly driven by low take-up of Part D subsidies, which Medicare beneficiaries automatically receive if they have Medicaid. Expanding eligibility for Medicaid supplemental coverage and increasing take-up of Part D subsidies would lessen cost-related barriers to health care among near-poor Medicare beneficiaries.

OBJECTIVES: To estimate rates and settings of low-value imaging among pediatric Medicaid beneficiaries and estimate the associated expenditures.

STUDY DESIGN: Retrospective longitudinal cohort study from 2014 to 2016 of children <18 years enrolled in Pennsylvania Medicaid. Outcomes were rates of low-value imaging for 5 conditions identified by diagnosis codes, healthcare settings of imaging performance, and cost based on paid amounts.

RESULTS: Of the 645 767 encounters for the 5 conditions, there were 37 525 (5.8%) low-value imaging services. Per 1000 encounters, there were 246.0 radiographs for bronchiolitis, 174.0 head computed tomography (CT) studies for minor head trauma, 155.0 and 33.3 neuroimaging studies for headache and simple febrile seizure, respectively, and 19.5 abdominal CT scans (without prior ultrasound examination) for abdominal pain. Rates of low-value imaging were highest in non-Hispanic White children and those in rural areas. In adjusted analysis, non-Hispanic White children were more likely to receive a CT scan for abdominal pain, and Black children were more likely to have imaging for bronchiolitis and minor head trauma. For individual conditions, up to 87.9% of low-value imaging (CT scan for minor head trauma) was in the emergency department (ED), with most imaging across all conditions occurring in nonpediatric EDs, up to 42.2% was in the outpatient setting (neuroimaging for headache), and up to 20.7% was during inpatient encounters (neuroimaging for febrile seizure). Outpatient and ED low-value imaging resulted in more than $7 million in Medicaid expenditures.

CONCLUSIONS: Among the studied conditions, more than 1 in 20 encounters included low-value imaging, mostly in nonpediatric EDs and for bronchiolitis, head trauma, and headache. Interventions are needed to decrease the future performance of these low-value services.

BACKGROUND: There is growing interest in financing housing and supportive services for homeless individuals through Medicaid. Permanent Supportive Housing (PSH), which integrates non-time-limited housing with supportive services for people who are disabled and chronically homeless, has seen rapid growth in the last decade, but clear evidence on the long-term impacts of PSH, needed to guide state efforts to finance some PSH services through Medicaid, is lacking.

OBJECTIVE: Assess changes in Medicaid expenditures and utilization associated with receiving PSH.

DESIGN: Cohort study using a difference-in-differences approach.

PARTICIPANTS: A total of 1226 PA Medicaid enrollees who entered PSH 2011-2016 and remained in PSH for 180 days or more, and a matched comparison cohort of 970 enrollees experiencing housing instability who did not receive PSH.

MAIN MEASURES: Medicaid spending in aggregate, and on behavioral and physical health services; emergency department (ED) visits and inpatient hospital stays.

KEY RESULTS: Three years after PSH entry, spending decreased by an average of $145/month in the PSH cohort relative to changes in the comparison cohort (p = 0.046), with the greatest relative spending reductions occurring for residential behavioral health ($64, p < 0.001) and inpatient non-behavioral health services ($89, p = 0.001). We also found relative reductions in ED use (4.7 visits/100 person-months, p = 0.010) and inpatient hospital stays (1.6 visits/100 person-months, p < 0.001).

CONCLUSIONS: These results can inform emerging state efforts to finance PSH services through Medicaid. Additional state expenditures to expand financing for PSH services could be partially offset by reductions in Medicaid spending, in part by facilitating a shift in treatment to outpatient from acute care settings.

IMPORTANCE: State Medicaid programs have reported concerns about rising drug prices and spending, particularly regarding drugs entering the market through the accelerated approval program under the US Food and Drug Administration (FDA). The accelerated approval program enables the FDA to approve drugs on the basis of unverified surrogate end points, meaning that clinical benefits for these products are uncertain at the time of approval. However, state Medicaid programs are legally required to cover these drugs. Little is known about the set of products with accelerated approval over time, their use among Medicaid beneficiaries, or the magnitude of their financial influence on state Medicaid programs.

OBJECTIVE: To identify the number and class of drugs approved through the FDA's accelerated approval pathway and analyze state Medicaid programs' use and spending on these drugs from 2015 through 2019.

DESIGN SETTING AND PARTICIPANTS: In this cross-sectional study, biannual FDA reports were used to identify products granted accelerated approval and their associated indications approved between December 1992 and December 2020. State Medicaid Drug Utilization Data files available for 1992 through 2019 were used to estimate national totals for spending and use of outpatient drugs.

MAIN OUTCOMES AND MEASURES: National Medicaid use and gross and net spending on drugs with accelerated approval from 2015 through 2019.

RESULTS: Since the inception of the FDA's accelerated approval pathway in 1992 through 2020, 216 product-indication pairs granted accelerated approval were identified, comprising 149 unique products. The composition of drugs approved through the pathway has changed over time, with 28 of 30 (93.3%) product-indication pairs receiving accelerated approval in 2020 being indicated for cancer. Relative to all outpatient prescription drugs paid for by Medicaid, products with accelerated approval ranged from 0.2% to 0.4% of use (1.3-2.4 million prescriptions annually). Despite their infrequent use, drugs with accelerated approval represented a minimum annual net spending on all drugs covered by Medicaid of 6.4% ($2.2 billion of $34.6 billion) in 2015 and a maximum of 9.1% ($2.5 billion of $27.6 billion) in 2018. Estimated annual gross spending on drugs with accelerated approval ranged from $4.2 billion to $4.9 billion over 2015 through 2019, and estimated net spending from $2.2 billion to $2.6 billion.

CONCLUSIONS AND RELEVANCE: In this cross-sectional study of 216 drugs granted accelerated approval, state spending on drugs approved through the FDA's growing accelerated approval program represented an outsized amount of spending relative to use. Because drugs with accelerated approval have come to market on the basis of trials using surrogate end points, considerable amounts of this spending may have been attributable to products with unproven clinical benefits.