Early clinical judgment in adult autism referrals: association with diagnostic outcome, verification bias and equity in an NHS triage pathway

Background United Kingdom demand for adult autism assessment exceeds capacity, and services rely on brief early contacts to decide who proceeds to full assessment. Whether a judgment formed at such a contact conveys diagnostic information, how closely it tracks progression to assessment, and whether it performs equitably are unknown. Methods In an NHS adult autism pathway, ADOS-2-trained Clinical Psychology Assistant Practitioners conducted a structured 30 min consultation to complete a quality referral for screening and triage and recorded an early clinical judgment (Unlikely, Neutral or Possible). The reference standard was the multidisciplinary diagnosis (developmental history, ADOS-2, clinical interview). In the assessed cohort ( n = 199), association was examined with Firth-penalized regression and discrimination with the area under the receiver operating characteristic curve (AUC). In a referral-level extract ( n = 1,134), progression to assessment by judgment category quantified partial verification. Variation by patient and assessor characteristics was examined in exploratory interaction analyses. Results Forty-nine of 199 assessed people (24.6%) were diagnosed: 0 of 19 (0%; 95% CI 0 to 17.6) rated Unlikely, 30 of 143 (21.0%) Neutral and 19 of 37 (51.4%) Possible. A Possible judgment was associated with higher odds of diagnosis than Neutral (Firth OR 3.92, 95% CI 1.85 to 8.39). Discrimination was modest (AUC 0.67, 95% CI 0.60 to 0.74); 71.9% of judgments were Neutral, and a possible judgment identified 38.8% of diagnosed people. Progression to assessment depended strongly on the judgment: 8 of 597 (1.3%) Unlikely, 111 of 437 (25.4%) Neutral and 43 of 55 (78.2%) Possible referrals proceeded. No interaction with patient characteristics or assessor gender was detected, although minority ethnic groups were small and confidence intervals wide. Conclusion Among assessed adults, early judgments recorded by graduate-level practitioners after a brief structured consultation were graded by diagnosis, but discrimination was modest. Triage and diagnostic clinicians knew the judgment, so incorporation bias cannot be excluded, and because Unlikely referrals were rarely assessed, the judgment cannot serve as a stand-alone rule-out. The clinical utility of the triage stage was not evaluated; prospective studies with concealed judgments and random verification of screened-out referrals are required.