Importance:Telemedicine is now widely used, stimulated by pandemic-era expansion rules and payment parity to in-person visits. Lawmakers continue to consider how to revise existing policies because of uncertainty about the potential for telemedicine to increase utilization and spending. Objective:To quantify the association between telemedicine adoption and visits and spending. Design, Setting, and Participants:This cohort study used multipayer medical claims data from MedInsight's research database for a national sample of adults continuously enrolled in Medicare fee-for-service, Medicare Advantage, dual-eligible, Medicaid, or commercial insurance, from January 1, 2019, to October 31, 2023. Data were analyzed from April 30, 2024, to February 10, 2026. Exposure:Regional telemedicine adoption (measured at the hospital referral region [HRR] level). Main Outcomes and Measures:The primary outcomes were (1) total combined telemedicine and in-person ambulatory visits (primary care, specialist, and preventive screening visits), and (2) total combined per-member-per-month spending on professional, inpatient, facility outpatient, prescription drug, and ancillary payments. By use of difference-in-differences analyses on visit-level and spending-level changes before (January 1, 2019, to December 31, 2019) vs after (January 1, 2021, to October 31, 2023) telemedicine expansion in high-telemedicine vs low-telemedicine adoption quintiles (measured at the HRR level), age-adjusted, sex-adjusted, and diagnosis-adjusted Poisson regressions were estimated, accounting for repeated measurements over time. Analysis was also stratified by urbanicity, payer, and Centers for Disease Control and Prevention Social Vulnerability Index quintiles to explore heterogeneous associations. Results:The sample included 3.04 million US individuals (mean [SD] age, 54.2 [17.2] years; 55.7% female) who utilized 120 million visits and incurred $178.4 billion in spending during 2019 to 2023. In 2019, the mean (SD) visit rate was 0.66 (0.035), and the mean (SD) spending rate was $774.59 ($36.78) per-member-per-month. Overall, point estimates suggested high-adopting areas had 2.4% (95% CI, -8.1% to 3.6%) fewer visits and 0.5% (95% CI, -13.1% to 13.9%) lower spending; however, 95% CIs crossed the null. Similarly, point estimates varied across subgroups but none achieved statistical significance: there were 4.4% (95% CI, -11.2% to 3.0%) fewer visits and 2.3% (95% CI, -18.9% to 17.8%) lower spending among urban populations, 2.5% (95% CI, -12.9% to 8.0%) lower spending for Medicaid-insured individuals, 5.3% (95% CI, -47.1% to 66.2%) lower spending for dual-eligible individuals, 3.0% (95% CI, -9.2% to 3.5%) lower spending for Medicare Advantage-insured individuals, and 1.5% (95% CI, -19.1% to 19.8%) lower spending among the most socially vulnerable populations. Conversely, point estimates suggested 3.4% (95% CI, -4.9% to 12.5%) greater visits and 3.8% (95% CI, -12.2% to 21.4%) higher spending in rural areas, 1.1% (95% CI, -12.8% to 17.3%) higher spending for commercially insured individuals, 1.0% (95% CI, -7.1% to 11.5%) higher spending for Medicare fee-for-service-insured individuals, and 4.5% (95% CI, -12.7% to 23.1%) higher spending among the least socially vulnerable groups. All 95% CIs crossed the null. Conclusions and Relevance:Nationwide telemedicine adoption was not significantly associated with changes in visits or spending, either overall or when stratified by urbanicity, payer type, or area-level social vulnerability, thus easing concerns about large utilization and spending increases from telemedicine expansion.

We combine 2021-2024 data on artificial intelligence (AI) adoption across U.S. shortterm general hospitals with national measures of hospital finances, volume, employment, and measured quality. Using synthetic difference-indifferences , we find that AI adoption is followed by approximately 3% higher net patient revenue, 3% higher total paid hours, and 7% higher patient volume. Total and clinical expenses also rise. By contrast, estimates for administrative expenses, administrative hours, and employee full-time equivalents are imprecise under inference clustered at the hospital-system level. Measured risk-adjusted mortality declines for several conditions, but unadjusted mortality and claims-based clinical-process measures do not show corresponding improvements, while documented severity increases. The results therefore point most clearly to operational expansion, throughput, and richer documentation; they do not establish administrative cost savings, per-unit productivity gains, or lower underlying mortality.

The COVID-19 vaccination campaign that made millions of Americans eligible for COVID-19 vaccines was primarily aimed at reducing COVID-19 transmission and mortality risks, but there may be important secondary benefits. Leveraging state-level exogenous variation in the timing of when people in different age groups became eligible for COVID-19 vaccination, we find that being eligible for vaccination leads to large reductions in self-reported anxiety and depression symptoms. These improvements in mental health stem from both self-vaccination and vaccination among near-age peers, suggesting both personal and external benefits of vaccine rollout policies. Furthermore, we find evidence that suggests that populations disproportionately impacted (i.e., those who had increased stressors due to the COVID-19 pandemic) experienced larger improvements in mental health because of vaccination. We estimate the economic benefit of reductions in anxiety and depression symptoms due to the COVID-19 vaccine campaign in the United States to be approximately $100 billion.

Introduction : Despite the existence of effective medications for opioid use disorder (MOUD), including methadone and buprenorphine, overdose deaths remain high. In 2020, Medicare began covering methadone for the treatment of opioid use disorder (OUD), and separate federal policy changes expanded MOUD access via telemedicine in response to the COVID-19 pandemic. However, little is known about national trends in methadone and buprenorphine dispensing across commercial and Medicare Advantage (MA) payers, especially by patient characteristics. Methods : A retrospective repeated cross-sectional study was conducted using 2019–2024 data from Optum’s de-identified Clinformatics® Data Mart Database to examine quarterly methadone and buprenorphine dispensing rates per 100,000 OUD-diagnosed enrollees. Rates were stratified by insurance type (commercial vs. Medicare Advantage), age, and dual eligibility status. Joinpoint regression identified statistically supported trend inflection points and annual percent changes (APCs) per segment. Results : Among OUD-diagnosed enrollees, MA methadone rates rose from 624 to 5,462 per 100,000 between 2020Q1 and 2024Q2 (AAPC +104.6%/year), while commercial methadone rates increased from 3,552 to 6,821 per 100,000 (AAPC +23.9%/year) during the same period. MA buprenorphine rates increased steadily (AAPC +20.7%/year), while commercial buprenorphine peaked in 2022 and declined at -20.0%/year after 2023Q2. Dually eligible MA enrollees under age 65 had the highest MOUD dispensing rates throughout. Conclusions : Methadone dispensing increased substantially among MA enrollees with OUD after 2020, particularly among dually eligible beneficiaries younger than 65 years. The post-2022 commercial buprenorphine decline in dispensing warrants further investigation. Tailored efforts may be needed to close payer-related treatment gaps and support broader MOUD access.

BACKGROUND:Private acquisition of public and nonprofit healthcare facilities is increasing throughout the United States (US). While privatization can be beneficial, growing evidence has demonstrated higher costs, larger patient volumes, and worsening outcomes. These trends are also apparent in substance use disorder (SUD) treatment facilities at a time when overdose and treatment demand are at an all time high. This study was designed to measure whether increases in private acquisitions of nonprofit substance use treatment facilities are happening faster in underrepresented and underresourced communities. METHODS:We used the Mental health and Addiction Treatment Tracking Repository (MATTR) to identify SUD treatment facilities that were nonprofit/public owned in 2019 (N = 2826 facilities). Our outcome was whether a facility became privatized and owned by a for-profit company by 2024. We linked MATTR to demographic census data and modeled privatization using a generalized estimating equation with a modified Poisson distribution, log link function, and robust standard errors. RESULTS:Twenty percent (n = 572) of public/nonprofit SUD treatment facilities were privatized between 2019 and 2024. Privatization of nonprofit/public facilities was more common in communities with lower household incomes (p < 0.01). CONCLUSION:Private acquisition of nonprofit/public SUD treatment facilities increased between 2019 and 2024. Acquisitions were disproportionately located in communities with higher rates of low-income households. Private acquisition can be beneficial for some facilities, but a growing evidence base is demonstrating how privatization is generally followed by worsening health outcomes in the process of restructuring and reselling at a profit. Treatment systems must not be extractive. Oversight and community involvement may help ensure mutual beneficence.

Importance:Alcohol use disorder (AUD) is a major public health concern; medications for AUD (MAUD) are an effective form of treatment but remain underused. Identifying MAUD access trends and the characteristics of counties with limited availability can inform targeted efforts to expand treatment capacity. Objective:To examine trends in geographic availability of MAUD at US substance use disorder treatment facilities (SUDTFs) from 2017 to 2023 and assess county characteristics associated with SUDTFs offering MAUD. Design, Setting, and Participants:This nationwide cross-sectional study used data from the Mental Health and Addiction Treatment Tracking Repository, which includes longitudinal data on licensed SUDTFs and whether they offer MAUD (acamprosate, disulfiram, or naltrexone), to quantify trends in MAUD availability at SUDTFs from January 2017 to December 2023. Main Outcomes and Measures:The primary outcome was a county-year indicator for whether at least 1 SUDTF in the county offered MAUD. Explanatory county variables included rurality, percentage of traffic fatalities involving alcohol, percentage of the population that drank excessively, percentage of uninsured individuals, poverty rate, percentage of individuals over age 65 years, and percentage of non-Hispanic White individuals. Univariate logistic regressions with state and year fixed effects were used to explore associations between county characteristics and the probability that a county had any SUDTFs offering MAUD. Results:Across 22 000 county-years in a total of 3153 counties, the mean (SD) percentage of counties with at least 1 SUDTF offering MAUD increased from 34.12% (47.42%) in 2017 to 43.88% (49.63%) in 2021, but growth plateaued after 2021. Lower MAUD presence in a county was associated with rural-adjacent (difference, -22.40 percentage points [pp]; 95% CI, -24.43 to -20.38 pp) and rural-remote (-23.64 pp; 95% CI, -25.72 to -21.56 pp) relative to metropolitan county status as well as with a higher poverty rate (-0.66 pp; 95% CI, -0.93 to -0.38 pp), greater percentage of individuals aged 65 years or older (-2.33 pp; 95% CI, -3.02 to -1.65 pp), and higher proportion of non-Hispanic White individuals (-0.58 pp; 95% CI, -0.71 to -0.46 pp), whereas greater prevalence of binge drinking (difference, 1.90 pp; 95% CI, 1.26-2.54 pp) and a higher percentage of college-educated individuals (1.28 pp; 95% CI, 1.13-1.43) were associated with higher MAUD presence. Conclusions and Relevance:In this cross-sectional study, the proportion of SUDTFs offering MAUD increased from 2017 to 2021, but growth then plateaued. Policies supporting the expansion of MAUD-providing facilities, particularly in underserved counties, may be needed to address persistent gaps in access.

Importance:Asian language speakers with limited English proficiency (LEP) face significant barriers to accessing adequate mental health care. Despite worsening mental health outcomes for this population, there is limited research examining the availability of Asian language mental health treatment in the US. Objective:To quantify trends and analyze disparities in the geographic availability of Asian language mental health treatment from 2015 to 2024. Design, Setting, and Participants:This cross-sectional study of US mental health facilities from April 30, 2015, to December 9, 2024, used longitudinal data from the nationally representative Mental Health and Addiction Treatment Tracking Repository linked with county-level demographic data from the 2023 American Community Survey. Facilities were included if they completed the National Mental Health Services Survey or the National Substance Use and Mental Health Services Survey. Main Outcomes and Measures:Primary outcomes included the annual proportion of mental health facilities offering Asian language services and the proportion of counties with at least 1 such facility. For 2024, facility-level characteristics associated with Asian language services were assessed and geographic mismatches between service availability and the proportion of Asian language-speaking individuals with LEP were mapped. Results:The study included 3847 mental health facilities. Of these, 214 facilities (5.6%) offered services in at least 1 Asian language in 2024 (including Arabic, Chinese, Farsi, Hindi, Hmong, Japanese, Korean, Tagolog, and Vietnamese). The proportion peaked at 265 facilities (6.9%) in 2021, then declined from 2022 to 2024. The number of counties with at least 1 facility with Asian language services was 98 (6.3%) in 2024. Facilities offering Asian language services were concentrated in metropolitan areas (208 [97.2%]), particularly in California (57 [26.6%]) and the Northeast (52 [24.3%]). Rural areas lacked such services (3 of 485 rural facilities [0.6%] in 2024), even in counties with substantial populations of Asian language-speaking individuals with LEP (0 of 5 facilities). Conclusions and Relevance:This cross-sectional study found a persistent geographic mismatch between the mental health needs of Asian language-speaking individuals with LEP and the availability of appropriate linguistic services. The gap was pronounced in rural areas. The findings suggest that policies aimed at expanding the behavioral health workforce and increasing access to culturally and linguistically competent services to reduce ongoing disparities in mental health outcomes and access to care are urgently needed.

Importance The 988 Suicide & Crisis Lifeline (998 Lifeline) receives millions of contacts annually. Adequate staffing of 988 Lifeline centers may be important for timely, high-quality service, but little information exists on current staffing levels or difficulties. Objectives To describe 988 Lifeline center staffing and assess staffing-related difficulties. Design, Setting, and Participants In this cross-sectional study, a survey was fielded between May 6 and July 25, 2025, to all 206 centers in the 988 Lifeline network in the US and territories. Eligible respondents were individuals in leadership positions (eg, executive directors, vice presidents). Main Outcomes and Measures The survey measured staffing levels, shift coverage, modalities (telephone, text, and/or chat), operation of non-988 lines (eg, 211, local lines), proportions of paid and volunteer staff, remote and/or in-person work arrangements, and 4 domains of staffing difficulty: adequate staffing for the volume of contacts, acquiring funding to hire, recruiting staff, and retaining staff. Responses were linked with administrative data on location, presence of state 988 telecommunications fees, and subnetwork services (eg, national backup, Spanish language). Results Leaders at 159 of the 206 centers completed the survey (77% response rate), 71% (102 of 144) reported that their center was understaffed, and 89% (141 of 159) indicated difficulty acquiring resources to hire. Leaders at centers offering remote work reported greater difficulty in obtaining these resources compared with centers without remote work (94% [89 of 95] vs 81% [50 of 62]; odds ratio [OR], 3.40; 95% CI, 1.21-9.68; P = .02) but less difficulty recruiting staff (76% [72 of 95] vs 89% [55 of 62]; OR, 0.39; 95% CI, 0.14-0.98; P = .04). Respondents from centers with all paid staff reported greater difficulty recruiting compared with centers using at least some volunteers (86% [102 of 118] vs 66% [27 of 41]; OR, 3.28; 95% CI, 1.42-7.60; P = .006). Leaders at centers handling only 988 contacts reported less difficulty retaining staff than those also handling non-988 lines (63% [17 of 27] vs 83% [109 of 132]; OR, 0.36; 95% CI, 0.14-0.92; P = .03). Wide 95% CIs indicate uncertainty in the magnitude of these results. Conclusions and Relevance In this cross-sectional survey study of 988 Lifeline leaders, results suggested that most centers struggled to find resources to keep the center fully staffed. If staffing challenges persist, centers could face risks to staff well-being and service quality. Financing and operational strategies that support recruitment and retention will be critical to sustaining the quality and accessibility of 988 Lifeline centers.

Generative AI chatbots are frequently used for mental health-related questions, including suicide-related queries. Prior evaluations have focused on user interfaces (UI), which may include additional safety controls not present in direct application programming interface (API) access. To examine how frequently AI chatbot models provide direct responses to suicide-related prompts across access pathways and to determine whether safer behavior is intrinsic to the model or depends on UI-level safeguards. Observational cross-sectional study conducted in February and March 2026. We evaluated how five widely used consumer AI models (ChatGPT, Claude, Gemini, Grok, and Llama) responded to 30 previously vetted suicide-related prompts spanning five clinician-assigned risk levels. Each prompt was submitted 100 times through both public-facing UI and direct API access, yielding 30,000 total responses. The primary outcome was whether a response directly answered the suicide-related prompt. A direct response was defined as one that provided specific information or guidance related to the question asked, rather than refusing to answer, redirecting the user to a crisis resource, or responding only with general safety language. AI responses were categorized using a blinded large language model–based classifier. We estimated mixed-effects logistic regression models that predicted a direct response. The primary predictors were the AI model, access mode (UI vs. API), and prompt risk category. 69.8% of responses to the suicide prompts were direct. Direct responses were more common through APIs than UIs (77.4% vs. 62.2%). Differences in the likelihood of a direct response were most pronounced for higher-risk prompts: among very high-risk prompts, 24.8% of API responses were direct compared with 4.6% of UI responses; among high-risk prompts, 80.1% of API responses were direct compared with 48.4% of UI responses. Claude and Gemini had the highest direct response rates (78.1% and 73.9%), whereas ChatGPT, Grok, and Llama had lower rates (64.6%, 64.5%, and 67.8%). In mixed-effects models, UI access was associated with lower odds of a direct response than API access (odds ratio, 0.09; 95% CI, 0.08-0.10). Higher prompt risk was associated with lower direct response probability, and access-mode differences varied by risk level and model. AI safety behavior depends on how the user accesses the model. The safety observed in an AI chatbot’s UI should not be assumed to generalize direct model access through APIs or to downstream applications built on those APIs. Evaluations and policies must consider access channels to ensure comprehensive safety protections for all individuals interacting with AI chatbots.

This cross-sectional study assesses how Medicaid's role in financing inpatient psychiatric care has changed geographically across US states from 2014 to 2023.

This economic evaluation reports the number of therapeutic services for autism spectrum disorder that were acquired by private equity companies between 2015 and 2024.

Older adult opioid overdose deaths have increased over the past two decades in the United States. Methadone, one of three medications approved for opioid use disorder (OUD) treatment, was not covered by Medicare - the primary insurer of older Americans - for OUD until 2020. We study the response of opioid treatment programs (OTPs), the only healthcare providers that can dispense methadone for the treatment of OUD in the U.S., to this policy change using administrative data and a difference-in-differences framework. We find an increase in Medicare acceptance by OTPs and in the number of methadone treatment episodes post-policy. Further, we document spillovers to other insurance markets and changes in (non-methadone) treatment services offered by OTPs, suggesting that Medicare's entrance into this market has implications for all OUD patients receiving methadone treatment in OTPs.

Objective : To understand how 988 Lifeline centers are connected to the broader crisis care continuum—including 911, mobile crisis response, and community-based services. Methods : The authors conducted a cross-sectional survey of 988 Lifeline centers in the U.S. between May and July 2025. A total of 159 centers (78%) completed the survey, which assessed transfer to/from 911, coordination with mobile crisis teams, capacity for referrals, and tracking dispositions and outcomes. Exploratory bivariate analyses were conducted to understand the factors associated with transfers to/from 911 and mobile crisis response team dispatch capability. Results : Approximately 45% of centers reported bidirectional interoperability with 911. Among the 153 centers with a mobile crisis response team in their service region, 61% could dispatch the team directly. Exploratory bivariate analysis indicated that centers had higher odds of interoperating with 911 when mobile crisis services were present throughout their jurisdiction (OR = 2.10 [1.06, 4.20]) and could schedule appointments on callers’ behalf (OR = 2.88 [1.39, 6.23]). Conclusions : Variation remains in centers’ ability to interoperate with 911 and dispatch mobile crisis teams. Centers with these capabilities were associated with having multiple referral, dispatch, and coordination capacities. Improving integration between 988 and other crisis services requires strengthened partnerships, sustainable funding, and investments in technology and data-sharing infrastructure.

This cross-sectional study identifies patterns of telehealth utilization and spending for anxiety, depression, bipolar disorder, schizophrenia, and posttraumatic stress disorder before, during, and after the COVID-19 pandemic among Medicare fee-for-service beneficiaries.

To combat the ongoing opioid crisis, policy makers and public health officials are developing novel policies to increase the availability of medications for opioid use disorder (OUD). An important question is to what extent geographic availability of opioid treatment programs (OTPs) is associated with treatment receipt. Understanding this association may help with developing additional policies to increase medication for OUD dispensing and improve population health outcomes. To quantify trends in dispensing methadone to Medicare beneficiaries based on proximity to an OTP. This cross-sectional study analyzed 2020 Medicare fee-for-service claims for methadone for beneficiaries with a recent diagnosis of OUD merged with drive times to OTP locations. Medicare beneficiaries enrolled in a Part D prescription plan and diagnosed with OUD in any of the 3 quarters before and during the 2020 quarter of interest were examined. Drive time between the centroid of a beneficiary's zip code and the closest OTP. Quarterly methadone receipt among Medicare beneficiaries with a recent OUD diagnosis was assessed using logistic regression models. In 2020, there were 640 706 Medicare beneficiaries with a recent OUD diagnosis (mean [SD] age, 62.5 [13.5] years; 55.6% female; 65.5% residing in an urban locality at the time of diagnosis). Of these beneficiaries, 9.6% lacked an OTP within a 60-minute drive time. The probability of a beneficiary receiving methadone decreased as the drive time from an OTP increased. Specifically, in urban areas, the likelihood of methadone receipt decreased by a relative 54% from a mean of 5.29% (national interval, 4.27%-6.52%) for beneficiaries who lived within a 5-minute drive of an OTP to 2.39% (national interval, 1.92%-2.98%) for those who lived within a 15-minute drive from an OTP. For rural beneficiaries, the likelihood of methadone receipt decreased by a relative 27% from a mean of 3.42% (national interval, 2.73%-4.28%) for a 5-minute drive time to 2.39% (national interval, 1.92%-2.98%) for a 15-minute drive time. Evidence of a threshold effect at a drive time of 20 minutes was observed for methadone receipt, after which the rate slowed and was similar between urban and rural beneficiaries. These findings suggest that the likelihood of methadone receipt may vary based on proximity to OTP facilities. Where OTPs are located may be a contributor to whether an individual receives methadone treatment.

This cross-sectional study evaluates whether telehealth utilization is associated with changes in payer-based and rural-urban disparities in substance use disorder treatment.

This study characterizes trends in physician participation in Medicare between 2013 and 2023 and whether physician- and county-level factors were associated with program exit.

Objectives This study explores trends in buprenorphine availability at substance use disorder treatment facilities (SUDTFs) and by waivered clinicians during the pandemic. We also examined whether there were differences in access based on a county's metropolitan status and annual fatal drug poisoning rate. Methods Data from the Substance Abuse and Mental Health Services Administration' Behavioral Health Treatment Locator between July 2019 and May 2021 were used to calculate trends in SUDTFs offering buprenorphine and the number of waivered clinicians per 10,000 population. We calculated unadjusted trends over time, stratified by whether a county was above or below the annual median age-adjusted fatal drug overdose rate in that year and the county's metropolitan status. Results Results showed an increase in SUDTFs and waivered clinicians offering buprenorphine before the pandemic, but the rate leveled off during the pandemic. On average, the increase in facilities was about 8 percentage points per year, and the increase in waivered clinicians was 0.29 per year. The percentage of SUDTFs offering buprenorphine peaked at 47%, and the number of waivered clinicians leveled off at 1.61 per 10,000 population. There were more SUDTFs and clinicians offering buprenorphine in metropolitan versus nonmetropolitan counties. There were also more SUDTFs and clinicians offering buprenorphine in counties above versus below median poisoning rates. Conclusions This study provides insights into how buprenorphine availability changed during the COVID-19 pandemic and before the removal of the X-waiver in 2023. More outreach will be needed to encourage the offering of buprenorphine by SUDTFs and office-based clinicians.