Applying Stratified Random Sampling for Multi-Cadre Regulatory Compliance Research in Resource-Limited Tertiary Hospitals: Methodological Lessons from an NDPR Cross-Sectional Survey in Rivers State, Nigeria

Abstract

Background: Regulatory compliance surveys in healthcare require sampling strategies that achieve proportional representation across professionally diverse workforces while managing the operational constraints typical of low-resource hospital environments. Stratified random sampling is widely recommended for such contexts but its implementation in multi-facility, multi-cadre African healthcare research is rarely documented with sufficient methodological detail to support replication. This paper reports the methodological design, implementation, and performance of a stratified random sampling strategy used in a cross-sectional survey assessing Nigeria Data Protection Regulation (NDPR) compliance among health personnel across two tertiary hospitals in Rivers State, Nigeria—and evaluates the methodological lessons generated for survey researchers working in similar contexts.

Methods: A descriptive cross-sectional design was employed. The accessible population comprised approximately 2,400 health personnel across the University of Port Harcourt Teaching Hospital (UPTH) and the Rivers State University Teaching Hospital (RSUTH). Sample size was determined using the Cochran formula for proportions (z = 1.96; p = 0.48; e = 0.05), adjusted by 10% for anticipated non-response (final n = 383). A two-stage stratified random sampling procedure was applied: Stage 1 allocated the sample proportionally to each hospital by relative staff size; Stage 2 further allocated within each hospital proportionally to professional cadre. Simple random sampling was used within each stratum, using computerised random number generation from administrative staff registers. Post-hoc methodological evaluation assessed achieved representativeness through comparison of obtained cadre proportions with population proportions, and sampling performance through response rate, completeness, and non-response bias sensitivity analysis.

Results: Of 383 questionnaires administered, 353 were validly returned (92.0% overall; UPTH: 95.1%; RSUTH: 97.9%). Achieved stratum proportions closely matched population proportions across all cadres (maximum deviation: nurses, +2.1 percentage points). Non-response bias sensitivity analysis (Available Case vs. Multiple Imputation approaches) confirmed that findings were robust to missing data assumptions. Administrative register access required formal institutional permission and took 11 working days to obtain; once granted, register-based simple random sampling within strata was completed in three working days. Three practical methodological challenges were identified and resolved: incomplete register records for short-term contract staff (resolved by restricting sampling frame to permanent appointments meeting inclusion criteria); discrepancies between register-listed cadres and current deployment (resolved by cross-checking against department allocation lists); and low initial accessibility of some radiography staff (resolved by extended recruitment across two shift cycles).

Conclusions: Stratified random sampling from administrative registers is methodologically sound and operationally feasible in Nigerian tertiary hospital settings, yielding high response rates and representative cadre distributions when supported by early institutional access negotiation, cross-referencing of multiple administrative data sources, and flexible recruitment scheduling. This paper provides a transferable methodological blueprint for compliance survey researchers in similar low-resource, multi-cadre healthcare environments across West Africa and comparable settings