Timely diagnosis of autism spectrum disorder (ASD) is essential for early intervention, family support, and equitable access to care. Persistent disparities in diagnostic access exist across socioeconomic, geographic, clinical, and health-system contexts. This systematic review and meta-analysis synthesized studies examining determinants of access across the ASD diagnostic pathway, including recognition and referral, diagnostic completion, diagnostic pathways, and timely diagnosis.
The review searched MEDLINE/PubMed, Embase, Scopus, Web of Science, Global Health, and grey-literature sources for studies published from January 2004 through December 2024. Records were screened and data extracted independently by two reviewers. Methodological quality of included studies was appraised with the Mixed Methods Appraisal Tool (MMAT).
Quantitatively comparable estimates were pooled using random-effects models with restricted maximum likelihood estimation. Heterogeneity was quantified by Cochran's Q, I2, tau2, and 95% prediction intervals. Pre-specified subgroup analyses and meta-regression were performed. Sensitivity analyses, funnel-plot assessments, trim-and-fill adjustment, Rosenthal fail-safe N, and a Bayesian random-effects model were also undertaken to evaluate robustness and potential small-study effects.
The electronic search identified 4,899 records. After removal of 537 records without associated data, 4,362 records were screened by title and abstract; 3,800 were excluded. Of 562 reports sought for retrieval, 112 could not be retrieved and 450 full-text reports were assessed. Ultimately, 22 unique studies met inclusion criteria. Nine studies supplied 23 quantitative effect estimates for meta-analysis; remaining studies contributed to the narrative synthesis.
The pooled overall diagnostic access outcome from the frequentist random-effects meta-analysis was 74.1% (95% CI 65.8–81.1%). Heterogeneity was substantial (Qe = 209.95, p < 0.001; I2 = 88.4%, 95% CI 79.1–94.4%; tau2 = 0.691). The 95% prediction interval was wide (32.8–94.4%), reflecting variability across settings.
A Bayesian random-effects analysis produced a highly concordant pooled estimate of 73.3% (95% credible interval 65.3–80.2%), with I2 = 87.5% and tau = 0.833. Markov chain Monte Carlo diagnostics indicated satisfactory convergence (R-hat = 1.000).
Studies covered determinants across socioeconomic, geographic, family, communication, screening, child developmental, provider, and health-system domains. When analyzed by outcome domain, pooled successful outcomes were:
Timely diagnosis exhibited particularly high heterogeneity (I2 = 91.2%), whereas diagnostic completion showed more moderate heterogeneity (I2 = 40.6%).
By determinant category, frequentist pooled estimates were:
Bayesian estimates were reported as 76.7% (BF = 53.76) for child developmental factors, 72.9% (BF = 226.32) for family/socioeconomic factors, 64.3% (BF = 25.60) for intervention/navigation, and 53.7% (BF = 0.684) for provider factors.
Meta-regression identified determinant category (Qm = 13.48, p = 0.004) and type of effect measure (Qm = 7.81, p = 0.020) as significant sources of between-study variation. Age group (p = 0.203) and geographic region (p = 0.453) did not significantly explain heterogeneity.
Family/socioeconomic factors had significantly larger effect sizes (B = 2.703, 95% CI 0.661–4.744; p = 0.009). Child developmental/neurobehavioral factors also had larger effect sizes (B = 1.516, 95% CI 0.047–2.985; p = 0.043).
Sensitivity analyses showed the pooled outcome was robust to exclusion of influential outlying estimates: excluding two outliers produced a pooled success of 77.1% (95% CI 71.6–81.9%). The Rosenthal fail-safe N was 1,723, indicating many additional null studies would be required to nullify the observed effect.
Two of three asymmetry tests suggested potential small-study effects. Trim-and-fill analysis identified seven potentially missing estimates and produced an adjusted pooled effect of 68.4% (95% CI 27.7–109.1%), a result with very wide confidence bounds reflecting uncertainty when imputing missing studies. The authors report these findings while noting the main pooled estimate's concordance with Bayesian analysis and sensitivity checks.
Across included studies, approximately three-quarters of observed ASD diagnostic outcomes represented successful access, but the high heterogeneity indicates access is highly context-dependent. Families were more likely to navigate diagnostic pathways successfully than to complete diagnostic assessment; timely diagnosis showed the most variability across settings.
Family and socioeconomic circumstances and child developmental characteristics emerged as especially important determinants of access. Provider-related determinants were more heterogeneous and less certain in pooled estimates.
The authors conclude that improving equitable ASD diagnosis requires interventions spanning the entire diagnostic pathway, including developmental surveillance and screening, referral coordination, family navigation and supports, enhanced provider capacity, increased specialist availability, and mechanisms that improve completion of diagnostic assessments.
The review included only 22 unique studies, with nine contributing quantitative estimates for pooling, and substantial heterogeneity across pooled estimates. Some reports could not be retrieved (112), which may limit comprehensiveness. Potential small-study effects and imputed missing studies introduced uncertainty in adjusted estimates.
The authors emphasize a need for greater longitudinal and implementation research, particularly in low- and middle-income countries where diagnostic infrastructure and specialist capacity remain limited. Because this report is a preprint, it has not been peer reviewed and should not be used alone to guide clinical practice.
All data produced in the work are reported as contained within the manuscript. The authors declared no competing interests and confirmed adherence to ethical guidelines and reporting standards. The preprint reports that appropriate ethical approvals and participant consents were obtained.