Health Services Research最新文献

Objective: To identify state strategies to increase access to medications for opioid use disorder (MOUD) through Section 1115 Substance Use Disorder waivers.

Study setting and design: We conducted a qualitative analysis of 27 waiver applications that were implemented between 2015 and 2020. We identified barriers and proposed strategies for expanding MOUD access and utilization.

Data sources and analytic sample: After excluding five states due to insufficient information, we analyzed 22 applications.

Principal findings: We identified six barriers and eight corresponding strategies. Barriers included care delays, limited MOUD facilities, lack of care transition support, limited MOUD access in residential treatment, insufficient care coordination, and prescriber shortages. Commonly proposed strategies were requiring access to MOUD in residential treatment, which was stipulated by the Centers for Medicare & Medicaid Services, addressing prescriber shortages through education and technical assistance, campaigns to address stigma, and increased reimbursement. Other strategies included changes to prior authorization requirements, efforts to increase the number of facilities that offer MOUD, and changes to improve care transitions.

Conclusions: States proposed a variety of strategies to expand access to and use of MOUD. Future research could investigate how these approaches, implemented individually or in combination, are associated with outcome change and impact.

Objective: To assess the impact of pandemic-related changes in state scope of practice law (SOPL) for certified registered nurse anesthetists (CRNAs) on the utilization of anesthesia services during the COVID-19 pandemic, which served as a natural experiment.

Study setting and design: We used a difference-in-differences approach to compare changes in the utilization of anesthesia services during the COVID-19 pandemic in areas that expanded SOPL (removed requirements for supervision or direction) to those that did not. Additionally, we examined if the impact of changes in SOPL on patient access differed by underserved status.

Data sources and analytic sample: To understand patient access to anesthesia care, we used a large, national administrative claims database of privately insured and Medicare Advantage enrollees to measure utilization rates per 1000 members from 2018 through 2022. We used the county-level density of anesthesia providers to identify underserved areas. We used data on the changes in SOPL at the state level and assessed changes in utilization prior to and during COVID-19.

Principal findings: In the areas that changed SOPL, removing requirements for supervision or direction, utilization of anesthesia procedures increased by 18 procedures per 1000 members over the study period (17% increase; p-value 0.066) compared with an increase of 9 procedures per 1000 members (7% increase; p-value 0.031) in areas that maintained SOPL requiring supervision. However, increases in utilization in underserved and not underserved areas were similar across SOPL statuses.

Conclusions: This study provides evidence that the SOPL that allows CRNAs to practice without the requirement of supervision or direction results in greater access to anesthesia services compared with a more restrictive SOPL requiring supervision.

Objective: To examine the impact of private equity (PE) hospital acquisitions on maternal health for Medicaid patients.

Study setting and design: This quasi-experimental study focuses on 66 PE acquisitions of hospitals between 2014 and 2018, analyzing national Medicaid claims data from 2011 to 2020. Using a difference-in-differences (DiD) framework, the study compares labor and delivery (L&D) outcomes at PE-acquired hospitals with matched control hospitals to evaluate the effects on patient volume, process of care, and quality outcomes for Medicaid patients.

Data sources and analytic sample: The analysis uses data from the Transformed Medicaid Statistical Information System (T-MSIS) and Medicaid Analytic eXtract (MAX), including over 1 million L&D hospitalizations. The analytic sample comprises 66 PE hospitals and 290 matched control hospitals.

Principal findings: PE acquisition was associated with a significant 12% decrease in Medicaid L&D market share (p < 0.05). The reduction was more pronounced in states with larger Medicaid-to-commercial payment gaps (-15.8% vs. -7.2%). However, no significant changes were observed in low-risk cesarean rates, number of procedures, length of stay, or severe maternal morbidity.

Conclusions: PE acquisitions of hospitals are associated with reduced Medicaid market share, particularly in states with lower Medicaid reimbursement relative to commercial insurance. Policymakers should consider addressing these issues by adjusting Medicaid payment rates to support vulnerable populations in PE-acquired hospitals.

Objective: To compare outcomes for Department of Veterans Affairs (VA) enrollees' behavioral health (BH) hospitalizations by source (VA-direct, VA-purchased community care (CC), Medicaid, Medicare, private insurance, and other payers).

Study setting and design: We conducted a retrospective, longitudinal study with VA enrollees from 2015 to 2017 to examine differences in BH hospitalization outcomes by source. We used generalized linear models with clustered standard errors to predict length of stay (LOS), cost, and 30-day readmission.

Data sources and analytic sample: We studied 124,609 BH hospitalizations of 77,299 VA enrollees in 11 geographically diverse states.

Principal findings: Predicted mean LOS (9.03 days, 95% CI 8.92-9.14 days; p < 0.001) and cost ($17,608, 95% CI $17,347-$17,870; p < 0.001) were highest for VA-direct hospitalizations, while the mean readmission rate was lowest for VA-direct hospitalizations (17.36%, 95% CI 17.03%-17.69%; p < 0.001). Average marginal effects for each non-VA hospitalization source were statistically significantly different from VA-direct hospitalizations (p < 0.001): between 2.13 and 2.90 days less for LOS, $11,141 to $12,144 less for cost, and 2.71% to 5.18% higher for readmission rate.

Conclusions: The majority of BH hospitalizations were in VA-direct care (56%), with 44% provided in locations outside VA hospitals: Medicare (19%), CC (7%), private insurance (7%), other payers (6%), and Medicaid (5%). There are trade-offs between BH hospitalizations provided in VA-direct care (lowest readmission rate, highest LOS and costs) and other sources.

Objective: The objectives of this study were to evaluate associations of social deprivation with acute care utilization among patients with cancer, and to examine potential effect modification by physician network vulnerability.

Study setting and design: For this retrospective cohort study, the primary exposure variable was neighborhood-level socioeconomic disadvantage, operationalized through the social deprivation index (SDI). We assembled physician patient-sharing networks and calculated a measure of network vulnerability for each referral region to capture specialist scarcity. The two outcomes of interest were counts of emergency department (ED) visits and non-elective hospitalizations during the 12 months following cancer diagnosis. We conducted hurdle regressions, with logistic and negative binomial mixed-effects models for the zero and positive, non-zero parts of the outcome distribution, respectively, and stratified by physician network vulnerability.

Data sources and analytic sample: We analyzed 2016-2020 Surveillance, Epidemiology, and End Results (SEER)-Medicare linked data for Medicare beneficiaries diagnosed with breast, colorectal, or lung cancer.

Principal findings: The study cohort comprised 47,756 patients with breast, colorectal or lung cancer. Patients in high SDI neighborhoods (vs. low) had a higher probability of at least one ED visit across all physician network vulnerability strata (low network vulnerability-average marginal effect (AME) [95% CI]: 0.03 [0.01-0.05]; medium network vulnerability-AME [95% CI]: 0.03 [0.01-0.04]; high network vulnerability-AME [95% CI]: 0.05 [0.02-0.08]). Conditional on at least one ED visit, patients in high SDI neighborhoods (vs. low) had a greater relative risk of additional ED visits when their region was characterized by low physician network vulnerability (RR [95% CI]: 1.25 [1.09-1.43]).

Conclusions: Our findings suggest that SDI and physician network vulnerability interact to increase the probability and likelihood of ED visits, but the interaction was minimal for non-elective hospitalizations. More research is needed to better understand how social drivers of health and oncology workforce scarcity affect care utilization and outcomes in patients with cancer.

Objective: To examine whether the new Medicare inflation rebate policy was associated with changes in manufacturer pricing behavior.

Study setting and design: In this cross-sectional study of 156 top-selling brand-name drugs, we used linear regression to evaluate whether there was an association between drugs' exposure to the policy (i.e., Medicare's share of net US sales) and differences in year-over-year price changes before (2021-2022) versus after (2022-2023, 2023-2024) the policy took effect.

Data sources and analytic sample: The study used Medicare spending data and average sales prices from the Centers for Medicare and Medicaid Services, wholesale acquisition costs from Eversana NAVLIN's Price & Access database, and sales revenue and estimated rebates from SSR Health. Vaccines, biosimilars, drugs approved after 2020, and those with generic or biosimilar competition before 2023 were excluded. Drugs were stratified by whether they derived most sales from Medicare Part B or Part D.

Principal findings: The median Medicare share of net sales was 28% (IQR: 18%-37%) for 50 Part B drugs and 32% (IQR: 16%-49%) for 106 Part D drugs. Median year-over-year price changes in 2021-2022, 2022-2023, and 2023-2024 were 3.2%, 2.9%, and 3.4% for Part B drugs and 5.0%, 5.9%, and 4.9% for Part D drugs. There was no association between drugs' Medicare share of net sales and differences in price changes pre- vs. post-policy for Part B drugs (2023: p = 0.99; 2024: p = 0.09). For Part D drugs, each 10% increase in drugs' share of Medicare sales was associated with a 0.18% (95% CI, 0.01%-0.35%, p = 0.04) higher price change in the first year after policy implementation; there was no significant association in the second year (p = 0.17).

Conclusions: Medicare inflation rebates were not associated with smaller price increases among the top-selling drugs most affected by the policy. Additional measures are needed to prevent drug manufacturers from raising prices each year, such as extending inflation rebates to commercially insured patients.

Objective: To investigate the effect of the unwinding of the pandemic-related continuous Medicaid enrollment provision on health insurance coverage and economic hardship.

Study setting and design: The termination of the continuous Medicaid enrollment provision during early 2023 and the subsequent state-level resumption of the standard renewal process prompted large-scale Medicaid disenrollments nationwide. Using state-month variation in the incidence of the first round of disenrollments, we estimate the effects of the unwinding process on health insurance coverage, including Medicaid enrollment, and the likelihood of experiencing economic hardship for the adult population.

Data sources and analytic sample: We use state-level monthly Medicaid enrollment data from the Centers for Medicare and Medicaid Services and self-reported individual-level indicators of Medicaid coverage, being uninsured, and economic hardship from the U.S. Census Bureau's Household Pulse Survey. Our key findings are substantiated by evidence drawn from recent annual data from the Current Population Survey and the Survey of Household Economics and Decisionmaking.

Principal findings: States' unwinding of the continuous Medicaid enrollment provision reduced state-level Medicaid enrollment by 4% [-0.071-0.004]. We do not, however, find statistically significant effects on changes in the probability of being without any health coverage and experiencing economic hardship for the overall adult population. However, further evidence reveals that the effects can be heterogeneous depending on demographic and educational characteristics.

Conclusions: The unwinding of the continuous Medicaid enrollment provision reduced overall Medicaid enrollments. However, there is no evidence that these provisions changed the probability of being uninsured and experiencing economic hardship for the general adult population. This study opens an important research scope for investigating the long-term implications of unwinding large-scale pandemic-related relief measures.

Objective: To evaluate the association between enrollment in the Bundled Payments for Care Improvement -Advanced (BPCI-A) program and changes in utilization of minimally invasive surgery and 90-day episode spending for patients undergoing major gastrointestinal surgery.

Study setting and design: We compared hospitals that voluntarily enrolled in BPCI-A to control hospitals that did not participate. We used entropy balancing to reweight controls to match the BPCI-A cohort based on observable patient and hospital characteristics. We then used a difference-in-differences approach to estimate the association between surgical approach and 90-day episode payments.

Data sources and analytic sample: We used Medicare claims and American Hospital Association data between 2013 and 2021 to evaluate whether hospital enrollment in the BPCI-A program was associated with changes in 90-day episode spending and utilization of minimally invasive surgical approaches. Using entropy balancing, we reweighted the control group to achieve covariate balance with beneficiaries who obtained care at BPCI-A program hospitals. We performed a difference-in-differences analysis using multivariable linear and generalized linear models, adjusting for patient demographics, comorbidities, and hospital characteristics, with standard errors clustered at the hospital-year level to evaluate these outcomes.

Principal findings: Changes in 90-day episode payments at BPCI-A program hospitals versus non-program hospitals were not significantly different (-$172, 95% CI: -$1104 to $760). In comparing trends at BPCI-A program and control hospitals, we identified no significant differences in utilization trends for minimally invasive surgical approaches (relative risk difference: -0.003, 95% CI: -0.10 to 0.04). The similarity in utilization trends between BPCI-A program and control hospitals was observed in the context of increasing overall utilization of MIS approaches from 40.3 to 38.4 to 43.9 to 42.9 during the study period, respectively.

Conclusions: We found no evidence that hospitals participating in BPCI-A's major bowel surgery episodes led to differences in episode spending or utilization of minimally invasive surgical approaches.

Objective: To define measures of Medicare diagnosis coding intensity that capture the dynamics of changes in coding practices.

Study setting and design: Retrospective analysis of coding for risk adjustment using observational claims data from Medicare beneficiaries.

Data sources: Enrollment and claims data from 2017 and 2018 of a random 20% sample of Medicare beneficiaries were subset to those assigned to an Accountable Care Organization in 2018.

Principal findings: We decompose the prevalence of a diagnosis code into incidence (proportion of beneficiaries that newly have the code) and persistence (proportion of beneficiaries who previously had the code and continue to do so). Together these define steady-state prevalence, the hypothetical long-run prevalence implied by no changes in current rates of incidence and persistence of coding. Steady-state prevalence can help explain why observed prevalence tends to grow over time without continued behavioral change. For example, our measures suggest that the prevalence of the Specified Heart Arrhythmias diagnosis would continue to rise from 18.7% in 2018 to 28.0% without changes in coding practices.

Conclusions: Researchers and policymakers can better understand why changes in coding practices can take years to be fully reflected in data and monitor coding behavior by using our proposed measures.