Comparing outcomes following a first episode of psychosis in autistic and non-autistic people: a clinical retrospective cohort study

Problem Statement
There is a recognized clinical overlap between autism spectrum conditions and psychotic disorders, with meta-analytic evidence suggesting autistic individuals face a 3–4-fold increased risk of developing psychosis. While preliminary evidence indicates that autistic traits may be associated with poorer outcomes following a first episode of psychosis (FEP), there is a lack of large-scale clinical data characterizing the specific service use patterns and outcomes for autistic people compared to non-autistic people within the context of early intervention and inpatient care. This study addresses the gap in understanding how autism influences the trajectory of care, service utilization, and hospitalization outcomes in individuals with a first episode of psychosis.

Methodology
This research employed a retrospective cohort study design using data from the Cambridgeshire and Peterborough NHS Foundation Trust (CPFT) Research Database.

• Participants: The study analyzed data from 7,409 individuals with a recorded psychotic disorder diagnosis. The cohort included 224 autistic individuals (185 with a confirmed diagnosis and 39 awaiting assessment) and 7,185 non-autistic individuals.
• Data Extraction: Clinical records from 2013 onwards were extracted. Covariates included gender, ethnicity, Index of Multiple Deprivation (IMD) decile, and age at diagnosis.
• Service Use Metrics: The study examined community service use (Early Intervention in Psychosis [EIP] referrals, liaison psychiatry, Section 136 police detentions, crisis team referrals, and telephone crisis services) and inpatient service use (voluntary and involuntary admissions, length of stay, psychiatric intensive care unit [PICU] admissions, and out-of-area admissions).
• Medication Analysis: A subset of the data (6473 non-autistic and 207 autistic participants) was analyzed for antipsychotic prescribing patterns, including total lifetime antipsychotics, long-acting injectables (LAIs), and clozapine use, utilizing natural language processing (NLP) on free-text records.
• Statistical Analysis: Descriptive statistics were used for initial examination. Inferential analyses included Cox proportional hazard models for time-to-admission, negative binomial regression for count data (e.g., number of contacts, duration of stay), and binomial logistic regression for binary outcomes. Models were adjusted for age, ethnicity, gender, IMD, and diagnosis. An exploratory analysis subdivided the cohort into four groups: control (neither autism nor learning disability [LD]), autism only, LD only, and autism with LD.

Key Results

• Demographics: The autistic group was significantly younger at the time of psychosis diagnosis (mean 34.4 years) compared to the non-autistic group (mean 48 years). The autistic group had a higher proportion of males (66.5% vs. 49.5%) and lower rates of missing ethnicity data.
• Community Service Use:
  • Autistic individuals were significantly more likely to have been subject to Section 136 police detentions (8.9% vs. 4.9%) and had a higher median number of contacts with the telephone crisis "First Response Service" (median 3 vs. 2).
  • Conversely, autistic individuals were less likely to have been referred to Crisis Resolution and Home Treatment Teams (CRHTT) (44.6% vs. 60.9%).
  • Referral rates to EIP services and treatment completion rates were similar between groups.
• Inpatient Service Use:
  • Autism was associated with a higher likelihood of psychiatric hospital admission (adjusted hazard ratio [aHR] 1.42, 95% CI 1.08–1.88).
  • Autistic individuals experienced significantly longer median total inpatient stays (112 days vs. 49 days, p<0.0001).
  • Autistic individuals were more likely to have been admitted out-of-area (7.4% vs. 3.5%).
  • There was no significant difference in the proportion of males admitted to PICU or the length of PICU stays between groups.
• Medication: No significant differences were found between groups regarding the total number of lifetime antipsychotics prescribed, the use of LAIs, or the use of clozapine.
• Role of Learning Disability (LD): LD was a strong predictor of service use patterns. The "autism and LD" and "LD only" groups showed distinct patterns, particularly regarding longer inpatient durations and lower CRHTT usage. In adjusted regression models, LD remained a significant predictor of community and inpatient service use, whereas autism alone was not always an independent predictor after adjustment for LD and other covariates.

Significance and Claims
The paper claims to demonstrate clear differences in the patterns of community and inpatient service use between autistic and non-autistic people following a first episode of psychosis. The authors highlight that autistic individuals are at greater risk of hospitalization and utilize emergency and resource-intensive services (such as police detentions and crisis phone lines) more frequently.

The study posits that these findings suggest unmet needs in community care for autistic people with psychosis, potentially leading to a reliance on acute and involuntary interventions. The authors note that while the "Transforming Care" initiative aims to move care to the community, the reduction in residential placements without a corresponding increase in community capacity may be contributing to the observed longer inpatient stays for vulnerable populations, including those with autism.

The authors conclude that while the study identifies these differing patterns, it remains unclear how best to support autistic people with psychosis to prevent deterioration and hospitalization. They emphasize the need for future research to invest in care models specifically designed for this population, with a core goal of admission avoidance and effective community support. The paper modestly acknowledges limitations, including potential data missingness, classification biases, and the relatively small sample size of the autistic group, which may have contributed to Type II errors in some adjusted analyses.