A data-driven methodological framework for representative recruitment in psychiatric research: insights from the DOCUMENT study

Background

Slow participant recruitment is one of the predominant determinants of failure or delay across clinical research. Even when recruitment targets are met, study populations may be unrepresentative due to sampling biases introduced by recruitment pathways. However, the effectiveness and demographic consequences of recruitment strategies are frequently underreported, undermining the generalisability of clinical findings and contributing to research waste. This study provides a data-driven, quantitative evaluation of multimodal recruitment strategies in psychiatric research, leveraging insights from the DOCUMENT study to synthesise a methodological framework for effective and representative participant recruitment.

Methods

Between June 2022 and December 2024, the study utilised a multimodal strategy to recruit participants with major depressive disorder (MDD), schizophrenia (SZ), and healthy volunteers (HV) for a two-phase study to investigate cognitive deficits across groups. Recruitment strategies included NHS clinical services, electronic health records, research registries, primary care sites, online and social media advertising, printed material, institutional resources, and word of mouth. For each avenue, yield, proportion of diagnostic group, recruitment rate, eligibility fraction, labour and financial cost, and demographic skew were quantified.

Results

Across avenues, 194 participants were recruited (66 HV, 77 MDD and 51 SZ), with high retention (86–97%). Recruitment efficacy varied substantially by diagnostic group, with online advertising and research registries successfully recruiting MDD and HV participants but failing to recruit eligible people with SZ. Instead, SZ participants were primarily enrolled from labour-intensive clinical avenues (94%). Online recruitment showed higher accrual, but lower eligibility fraction compared to clinical pathways, revealing systematic sampling differences. Despite avenue-specific sampling biases, the multimodal approach yielded close demographic alignment to the 2021 UK Census for London in the study population. Data-driven adaptations, such as protocol amendments to eligibility criteria and online self-report triaging, improved study feasibility.

Conclusions

No single recruitment avenue was identified as sufficient for both efficient and representative psychiatric recruitment. Instead, multimodal strategies were necessary to dilute avenue sampling biases. Synthesising 30 months of data, we introduce a 10-point framework for enhancing recruitment effectiveness, feasibility, and representativeness. While grounded in UK-based psychiatric research, these principles are likely transferable to broader clinical research contexts to reduce research waste.