Willing but Not Ready? Documentation Quality as a Barrier to Artificial Intelligence Adoption in Nigerian Healthcare

Abstract

Nigeria has demonstrated a commitment towards nationwide integration of artificial intelligence products into healthcare. However, concerns remain regarding feasibility due to historic challenges with data quality. Currently, there are no guidelines for scrotal ultrasound documentation—a prerequisite for generating high-quality data, and robust models. This study was carried out to assess routine scrotal ultrasound documentation quality as a proxy measure for AI readiness in Nigerian healthcare. To achieve this, we conducted a retrospective, descriptive cross-sectional study of scrotal ultrasonographic reports retrieved from health institutions in South Eastern Nigeria. Three hundred reports, generated between 2020 and 2025 were randomly selected and assessed for documentation quality across four domains using a de-novo structured checklist. Overall and domain specific compliance scores were then computed. Overall documentation quality was suboptimal, with a mean compliance score of 56·41 ± 8·45%. Removing the demographic elements of the reports resulted in a notable decline in mean compliance scores (49·05%), suggesting that overall completeness is inflated by administrative fields rather than clinically informative content. Tertiary institutions demonstrated higher compliance than secondary institutions (61·72% vs 53·76%; 95% CI: 6·24–9·75), though deficiencies persisted across all domains. Documentation quality was highest in the demographic domain. Our findings suggest that current documentation practices may undermine robust model performance and equitable deployment. Addressing this through standardized reporting and regulatory alignments are prerequisites for producing usable, trustworthy evidence in the digital health era.

Ähnliche Arbeiten