Hallucinogen-Psychosis Associations Are Confounded by Baseline Psychiatric History

Steinle and colleagues frame the paper around an important controversy in psychedelic policy and clinical research: whether hospital admissions involving hallucinogens reflect a direct causal risk of psychosis, or whether the association is driven by other factors. They note that earlier evidence from administrative and observational data has raised concern about subsequent psychosis or schizophrenia spectrum disorders after hallucinogen-related presentations, but those studies often did not capture outpatient psychiatric history well. The authors therefore emphasise that prior psychiatric illness, co-occurring substance use, and the prodromal phase of psychotic disorders may confound the apparent relationship between hallucinogen exposure and later psychosis. The study aims to test whether the observed association between hallucinogen-related admissions and later psychosis is explained by baseline psychiatric comorbidity. Using linked claims data that include both inpatient and outpatient diagnoses, the researchers examine whether people with hallucinogen-related admissions differ from other substance-related admissions in their pre-existing mental health burden and whether adjustment for those factors attenuates the association with subsequent psychosis. The paper is presented as relevant to both clinical assessment and policy debates about psychedelic safety and regulation.

This was a retrospective cohort study using the combined Merative MarketScan Commercial and Multi-State Medicaid Databases covering 2015-2019. The database includes emergency department visits and inpatient hospitalisations, which the paper refers to as admissions, and the study window ran from 1 January 2015 to 31 December 2019. The authors state that the data were de-identified and that the analysis was exempt from institutional review. The analytic cohort came from people with at least one substance-related admission during insurance enrolment and at least 180 days of continuous enrolment before the index admission so that premorbid psychiatric history could be assessed. Individuals with at least one hallucinogen-related admission were compared with those who had non-hallucinogen substance-related admissions, specifically opioid-, stimulant-, alcohol-, or cannabis-related admissions. After exclusions, the cohort included 6,184 people with hallucinogen-related admissions and 267,282 people with non-hallucinogen substance-related admissions. The main exposure was having at least one hallucinogen-related admission. For people with both hallucinogen and non-hallucinogen admissions, the index date was anchored to the first hallucinogen-related admission. The primary outcome was any psychotic disorder diagnosis occurring between 30 days and 6 months after the index substance-related admission, reflecting the DSM-5 convention that symptoms within 30 days of substance exposure may be substance-induced. Secondary analyses examined psychosis admissions as well as diagnosis claims. Covariates included age, sex, calendar year, insurance type, Charlson Comorbidity Index, and baseline psychiatric conditions, including mood disorders, anxiety disorders, psychotic disorders, and co-occurring substance use disorders before the index date. The authors first used descriptive statistics and unadjusted comparisons, then estimated Cox proportional hazards models. They reported unadjusted models followed by models adjusting for demographics and clinical history. Additional analyses were conducted in Medicaid enrollees to allow adjustment for race and ethnicity, plus sensitivity analyses addressing baseline psychosis, prodromal phase concerns, polysubstance use, age, and alternative outcome definitions. Analyses were performed in SAS 9.4, with two-sided significance testing.

The full substance-related sample contained 273,466 individuals. Of these, 6,184 had at least one hallucinogen-related admission and 267,282 had only non-hallucinogen substance-related admissions. About 10% of the hallucinogen cohort (695/6,184) had hallucinogen-related admissions only, without any non-hallucinogen substance-related admissions during enrolment. The hallucinogen cohort was younger than the comparison group, with a median age of 24 years versus 34 years. Medicaid enrolment was more common in the hallucinogen group than the comparison group, and among Medicaid enrollees the hallucinogen group had a lower proportion of non-Hispanic White individuals. Baseline substance use and psychiatric comorbidity were more common before hallucinogen-related admissions than before other substance-related admissions. For example, prior opioid use disorder was present in 22.5% of the hallucinogen cohort, prior cannabis use disorder in 36.6%, prior stimulant use disorder in 16.2%, and prior alcohol use disorder in 28.1%, all higher than the corresponding rates in the non-hallucinogen cohort. Baseline psychotic disorders were also more common in the hallucinogen group, at 19.2% versus 6.3% in the comparison group. After the index admission, 16.4% of the hallucinogen group received at least one psychosis diagnosis between 30 days and 6 months, compared with 6.6% of the non-hallucinogen group. Psychosis admissions occurred in 9.3% versus 3.3%, respectively. In unadjusted Cox models, hallucinogen-related admissions were associated with a higher hazard of post-index psychosis diagnosis (HR 1.22, 95% CI 1.19-1.25) and psychosis admission (HR 1.16, 95% CI 1.13-1.19). After adjusting for baseline psychotic disorder alone, the association with psychosis diagnosis was attenuated but still present (HR 1.12, 95% CI 1.09-1.15). After further adjustment for demographics and baseline clinical characteristics, the association was no longer increased and was essentially null (HR 0.97, 95% CI 0.95-1.00). In the Medicaid subgroup, the same general pattern was seen: the unadjusted association weakened after adjustment, including when race and ethnicity were controlled for. Across adjusted models, baseline psychiatric disorders were strong predictors of later psychosis, especially prior psychotic admissions (HRs 1.28-1.35), stimulant use disorder (HRs 1.27-1.32), and opioid use disorder (HRs 1.28-1.32). Sensitivity analyses restricted to people without baseline psychosis, stratified by baseline psychiatric status, extended follow-up to 12 and 24 months, focused on hallucinogen-only users, examined those younger than 26 years, and used alternative outcome definitions. These analyses were reported as broadly consistent with the main finding that the apparent hallucinogen-psychosis association was substantially reduced after accounting for baseline vulnerability.

The authors argue that the higher rate of later psychosis seen after hallucinogen-related admissions is largely explained by pre-existing psychiatric comorbidity rather than by hallucinogen exposure itself. They emphasise that once baseline clinical characteristics were taken into account, hallucinogen-related admissions were not associated with a significantly increased risk of later psychosis. In contrast, baseline psychotic disorders and other psychiatric or substance use disorders were strongly associated with later psychosis. They place these findings in the context of earlier research that has suggested a link between hallucinogens and psychosis, but often without detailed outpatient psychiatric history. The authors suggest that such earlier associations may reflect residual confounding by indication, meaning that people who present with hallucinogen-related admissions may already differ in important ways from other substance-using patients. They also note that the median age of hallucinogen-related admissions overlaps with the age at which schizophrenia spectrum disorders often emerge, making causal ordering difficult to establish from observational claims data. The paper further argues that population-level administrative data cannot reliably determine whether substance exposure preceded, coincided with, or followed the onset of prodromal psychotic symptoms. The authors say this limits causal inference and helps explain why the hallucinogen-psychosis relationship may be overstated if baseline vulnerability is not adequately measured. They also note that most people in the hallucinogen cohort had polysubstance involvement, which complicates interpretation. Several limitations are acknowledged. The database could not distinguish between specific hallucinogens such as psilocybin, LSD, MDMA, PCP, ayahuasca, and ketamine, each of which may have different risk profiles. Hallucinogen detection in admissions may also be less reliable than detection for other substances, so some hallucinogen-involved psychosis cases may have been misclassified. The sample is limited to people with substance-related admissions and therefore does not generalise to most recreational users. The 6-month outcome window may miss longer-term effects, and some individuals may have been in the prodromal phase of psychosis before their first recorded hallucinogen-related admission. The authors also note the possibility of under-ascertainment in both directions because some psychosis diagnoses may predate undocumented hallucinogen use. Despite these limitations, the authors highlight the strength of linked inpatient and outpatient claims, which allowed them to assess psychiatric history more comprehensively than studies relying on admissions alone. They suggest that the findings have implications for policy discussions about psychedelic medicalisation, decriminalisation, and regulation, because epidemiological associations do not necessarily imply direct causation. They also indicate that clinical screening for people seeking psychedelic therapy should extend beyond hospitalisation history to include outpatient mental health diagnoses, and that harm reduction strategies may need to target people with pre-existing psychiatric vulnerability.

The authors conclude that the apparent association between hallucinogen-related admissions and later psychosis is substantially confounded by baseline psychiatric comorbidity. They state that this should be taken into account in policy discussions, clinical assessment, and harm reduction approaches, and that psychiatric screening for prospective psychedelic therapy users should include outpatient mental health history rather than relying only on prior admissions.