Publications

2026

AIMS: To track people with a cannabis-associated diagnosis during a hospital visit and examine rates of subsequent hospital visits over one year, including visits with mental health- and substance use-associated diagnoses.

DESIGN: Retrospective cohort study of Arizona, USA hospital discharge data. Patients were observed for one year before and after their 2018 index visit to capture hospital encounters with mental health- and substance use-associated diagnoses.

SETTING: Data were patient-level emergency department and inpatient hospital records from Arizona, USA 2017-2019.

PARTICIPANTS: Four groups of patients were identified based on diagnoses during their index visit in 2018. These included people who had 1) a cannabis-associated diagnosis and a co-occurring mental health-associated diagnosis (C+/MH+; n = 16 699), 2) a cannabis-associated diagnosis but not a mental health-associated diagnosis (C+/MH-; n = 21 406), 3) a mental health-associated diagnosis but not a cannabis-associated diagnosis (C-/MH+; n = 230 519), and 4) neither a cannabis-associated nor a mental health-associated diagnosis (C-/MH-; n = 1 096 939).

MEASUREMENTS: Records included diagnosis codes from the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM), which were used to identify cannabis-associated diagnoses, other substance use-associated diagnoses, and mental health-associated diagnoses. In analyses predicting subsequent hospital visits, covariates included demographic factors; index visit month; total number of past-year visits, visits with a mental health-associated diagnosis, and visits with a substance use-associated diagnosis; concurrent (to the index visit) non-cannabis substance use-associated diagnosis and the Elixhauser readmission index.

RESULTS: After adjusting for covariates, people with a cannabis-associated diagnosis during their index visit in 2018 had 1.23 times the rate of subsequent hospital visits as those without a cannabis-associated diagnosis (99.8% confidence interval (CI) [1.20, 1.26]), 2.74 times the rate of subsequent visits with a mental health-associated diagnosis (99.8% CI [2.64, 2.84]), and 1.76 times the rate of subsequent visits with a substance use-associated diagnosis (99.8% CI [1.69, 1.83]). Elevated visit rates were observed among individuals with a cannabis-associated diagnosis, regardless of whether they had a concurrent mental health-associated diagnosis, though associations were stronger among those without a concurrent mental health-associated diagnosis.

SUMMARY: A cannabis-associated diagnosis during a hospital visit appears to signal risk for higher rates of subsequent hospital use.

Childs B, Hudgens O, Livingstone J, Treacher A, Naeem J, Khan F, Goldman-Mellor S, Roaten K. Universal Suicide Risk Screening in a Safety-Net Health System: Findings From 5.2 Million Encounters.. Psychiatric services (Washington, D.C.). 2026:appips20260003. doi:10.1176/appi.ps.20260003

OBJECTIVE: The authors aimed to examine whether universal suicide risk screening was associated with increased identification of suicide risk across demographic and clinical subgroups.

METHODS: This retrospective observational study analyzed electronic health record data from 5,219,858 encounters at a safety-net health system (2010-2020). The primary outcome was encounter-level suicide risk identification. Multivariable logistic regression was used to estimate adjusted odds ratios (AORs) comparing the periods before and after implementation of universal screening.

RESULTS: Suicide risk was identified in 0.7% of preimplementation encounters and 3.5% of postimplementation encounters (AOR=5.8, 95% CI=5.7-5.9). Relative increases were greatest among encounters involving older adults, charity or self-pay status, and Black patients and encounters in outpatient settings.

CONCLUSIONS: Universal suicide risk screening substantially increased identification across settings, a finding that supports its use as a scalable strategy for equitable detection.

De Alba L, Margerison CE, Gemmill A, Goldman-Mellor S. Prenatal cannabis use disorder and postpartum psychiatric emergency department visits: A retrospective cohort study of trends and associations in California, 2016-2021.. Annals of epidemiology. 2026:110271. doi:10.1016/j.annepidem.2026.110271

PURPOSE: Cannabis is the most-used substance during pregnancy and has known associations with psychiatric disorders (e.g., psychosis, depression). No prospective studies, however, have investigated associations between prenatal cannabis use disorder (CUD) and psychiatric outcomes in perinatal populations.

METHODS: Annual cohorts of women with a live infant delivery in California from 2016 to 2021 were constructed using statewide linked hospital and emergency department (ED) discharge data. We examined temporal trends in prenatal CUD prevalence and compared risk of postpartum psychiatric ED visits between delivering women with prenatal CUD, prenatal alcohol use disorder (AUD; included as a benchmark exposure), and neither disorder.

RESULTS: The study sample included 2,036,016 delivering women. Between 2016 and 2021, prenatal CUD prevalence increased 1.0% to 1.4%. Postpartum psychiatric ED visits occurred in 10.4% of women diagnosed with CUD, 15.6% with AUD, and 2.7% with neither diagnosis. Relative to women with neither, those with CUD had  40% higher risk of any postpartum psychiatric ED visit (RRadj=1.39, 95% CI: 1.32-1.47), while AUD was associated with 80% increased risk (RRadj=1.80, 95% CI: 1.56-2.09).

CONCLUSION: Elevated risk of adverse psychiatric outcomes associated with CUD underscore the need for consistent screening, diagnosis, and treatment among perinatal women with CUD diagnoses.

Drew LB, Goldman-Mellor S, Joachim G, Gemmill A, Margerison CE. Pregnancy-Associated Drug Overdose Mortality in the United States, 2018-2023.. Obstetrics and gynecology. 2026. doi:10.1097/AOG.0000000000006389

OBJECTIVE: To evaluate changes in pregnancy-associated drug overdose mortality in the United States from 2018 to 2023 and compare this with drug overdose mortality among reproductive-aged nonpregnant female decedents. Secondarily, to assess types and combinations of drugs that contribute to overdose mortality, regional patterns, and overdose timing.

METHODS: This was a cross-sectional study that used 2018-2023 National Center for Health Statistics death certificate data. Drug overdose deaths were identified using International Classification of Diseases, Tenth Revision codes, and specific drugs involved were identified using T codes. Among all female decedents of reproductive age (15-44 years), pregnancy-associated drug overdose deaths were identified from the pregnancy checkbox. We calculated absolute and relative changes in drug overdose mortality rates for pregnancy-associated deaths and non-pregnancy-associated deaths among reproductive-aged female individuals. The annual pregnancy-associated drug mortality rates were also calculated by drug type and U.S. region (Northeast, Midwest, South, and West).

RESULTS: There were 2,573 pregnancy-associated drug overdose deaths identified. Pregnancy-associated overdose mortality increased from 8.0 per 100,000 live births to 13.0 deaths per 100,000 live births (relative increase: 62.2%). The drug overdose mortality rate among female individuals of reproductive age increased from 16.8 per 100,000 live births to 22.3 per 100,000 live births across the same period (relative increase: 33.0%). The fentanyl-involved pregnancy-associated overdose mortality rate was higher than other drugs. The psychostimulant-involved rate increased across time (from 2.9 deaths/100,000 live births to 7.5 deaths/100,000 live births). Drug overdose mortality rates were highest in the South. Polysubstance-involved overdose deaths were also common, with fentanyl and psychostimulants accounting for 24.8% of deaths.

CONCLUSION: The relative increase in the drug overdose mortality rate among pregnant and postpartum individuals was nearly double that of nonpregnant female individuals of reproductive age, driven by fentanyl, increasing psychostimulant, and polysubstance overdose deaths.

Oyetunji TP, Chan-Golston AM, Ha S, Goldman-Mellor S. Cannabis Use Disorder and Suicidal Ideation in the United States: Persistent Risk and Stable Trends, 2014-2023.. Journal of studies on alcohol and drugs. 2026. doi:10.15288/jsad.25-00481

OBJECTIVE: Amid rising cannabis use, declining perceived harm, and policy liberalization over the past 15 years, we examined whether the association between cannabis use disorder (CUD) and past-year suicidal ideation among U.S. adults changed from 2014-2023 and varied by sex, age, and race/ethnicity.

METHOD: Using data from 415,861 adults in the 2014-2023 National Surveys on Drug Use and Health, we examined temporal association change between past-year suicidal ideation and CUD (≥2 of 9 harmonized DSM-5 criteria). Logistic regression models, adjusted for demographic, social, clinical, and geographic covariates, tested whether survey year moderated the CUD-suicidal ideation association overall and in strata by sex, age, and race/ethnicity.

RESULTS: Adults with cannabis use disorder had 57% higher odds of past-year suicidal ideation than those without (OR, 1.57; 95% CI, 1.44-1.72) across years. Joint Wald tests showed no evidence that the association changed over time overall (p = 0.12) nor within strata defined by sex or age. Although the joint interaction test suggested some variation among Hispanic respondents (p = 0.01), no individual year-specific interaction terms remained statistically significant after Bonferroni correction.

CONCLUSIONS: Adults with CUD had higher odds of suicidal ideation but, contrary to expectations, the strength of this association remained stable from 2014-2023 and did not differ significantly across sex, age, or racial/ethnic groups despite widespread legalization, increased potency, and shifting norms.

Richmond-Rakerd LS, Milne BJ, Houts RM, Davie G, D’Souza S, Goldman-Mellor S, Khalifeh L, Caspi A, Moffitt TE, Torvik FA. Mental health conditions are associated with increased risk of subsequent self-harm, assault and unintentional injuries in two nations.. Nature. Mental health. 2026;4(1):102–111. doi:10.1038/s44220-025-00553-w

Mental health conditions are associated with an increased risk of chronic physical diseases, but their implications for other physical health outcomes, including injuries, are less established. In this prospective cohort study, we tested whether mental health conditions antedate unintentional as well as self-harm and assault injuries, using administrative data from Norway (N = 2,753,646) and New Zealand (N = 2,238,813). In Norway, after accounting for pre-existing injuries, individuals with a primary care encounter for a mental health condition had an elevated risk of subsequent primary care-recorded injury. In New Zealand, as expected, individuals with a mental health-related inpatient hospital admission had an elevated risk of subsequent inpatient hospital-recorded self-harm injury, as well as assault injury. However, they also had an elevated risk of unintentional injuries. Associations extended to injury insurance claims. Associations were evident across mental health conditions, sex, age and after accounting for indicators of socioeconomic status. Risk was particularly increased for brain and head injuries. Patients with mental health conditions are an important group for injury prevention.

2025

Das A, Byrne CA, Oddo VM, Goldman-Mellor S, Kim SJ. Neighborhood Deprivation and Suicide Among Adolescent and Young Adult Cancer Patients.. Cancer medicine. 2025;14(19):e71247. doi:10.1002/cam4.71247

BACKGROUND: In 2024, approximately 84,100 adolescents and young adults (AYAs) between 15 and 39 years old received a cancer diagnosis.

AIMS: Given their unique psychosocial, economic, and clinical stressors, we examined whether AYA cancer patients living in deprived neighborhoods have a higher risk of suicide when compared to those in the least deprived neighborhoods.

MATERIALS & METHODS: Our sample comprised 486,374 AYA cancer patients from the Surveillance, Epidemiology, and End Results (SEER) dataset between 2006 and 2020. We use Cox proportional hazard models to test the relation between quintiles of neighborhood deprivation, from Q1 (most deprived) to Q5 (least deprived), and survival months until suicide mortality from the time of cancer diagnosis.

RESULTS: We find that AYA cancer patients living in more deprived neighborhoods (Q1, Q2, Q4) have a higher risk of suicide when compared to those in the least deprived neighborhoods (Q5) (Q1: HR-1.82 [1.14-2.90]; Q2: HR-1.95 [1.35-2.81]; Q4: HR-1.48 [1.05-2.07]).

DISCUSSION: Mental health services and monitoring from treatment through survivorship may support suicide prevention efforts for young cancer patients.

CONCLUSION: Early identification of high-risk AYA cancer patients living in deprived areas may help target suicide prevention interventions.

Goldman-Mellor S, Pathak J, Margerison C. Trends in prenatal opioid use disorder in California, 2010-2022.. American journal of preventive medicine. 2025:108086. doi:10.1016/j.amepre.2025.108086

INTRODUCTION: Opioid-related deaths among perinatal populations have increased sharply in the United States. Whether the recent ascendence of illicit fentanyl and other synthetic opioids in the drug supply translates to increasing prenatal opioid use disorder (OUD) remains unknown. This study tested whether California's comparatively late fentanyl influx, in 2019, was associated with a subsequent increase in OUD among pregnant people.

METHODS: This ecologic, repeated retrospective cohort study examined annual cohorts of individuals with an index live-birth delivery in California, 2010-2022. Analyses used statewide, all-payer linked hospitalization and emergency department (ED) records. Overall and subgroup-specific trends in prevalence of prenatal OUD, assessed using ICD-9/10-CM diagnoses recorded at delivery hospitalization and during prenatal ED visits, were examined using joinpoint and logistic regression analyses. Data analysis was conducted in 2025.

RESULTS: The study included 4,381,064 index deliveries. Between 2010 and 2022, prevalence of prenatal OUD doubled (as assessed at delivery: 0.15% to 0.30%; as assessed via prenatal OUD ED visits: 0.04% to 0.08%). Joinpoint analysis indicated that most of this increase occurred prior to 2018; annual percentage changes (APC) after 2018 were largely flat (at-delivery prevalence APC: 2.4%, 95% CI, -2.2% to 7.3%; ED prevalence APC: 1.4%, 95% CI, -7.0% to 3.8%). Differences in 2019-2022 vs. 2010-2018 prevalence of prenatal OUD were larger among non-Hispanic White and other/multiple-race individuals, and those covered by Medicaid.

CONCLUSIONS: The recent introduction of fentanyl to California was associated with only incremental increases in prevalence of prenatal OUD. Continued tracking of these trends is important for perinatal health.