Work

Completeness, consistency and analytical usability of routine data for preterm and small newborn care in Western Kenya: a multicentre mixed-methods study

Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Objectives To assess the completeness, consistency, plausibility, uniqueness and analytical usability of routine maternity and newborn records used for preterm and small newborn care, and how facility health-information-system readiness and working conditions explain these patterns. Design Multicentre cross-sectional mixed-methods study of routine health data with embedded explanatory qualitative interviews. Setting Twelve health facilities (levels 3–5) in Bungoma and one neighbouring county, Western Kenya. Participants 349 maternity-register and 291 newborn records for deliveries from July 2024 to May 2026; 10 facility readiness assessments; two provider key informant interviews. Primary and secondary outcome measures Variable-and record-level completeness from original uncleaned abstractions; prespecified consistency and plausibility rule violations; duplicates; agreement between recorded and rule-derived preterm, low-birth-weight (LBW) and very-low-birth-weight classifications. Results Median record completeness was 100% (IQR 94.7–100) for register and 93.5% (IQR 84.8–100) for newborn records; a complete core record was present in 175/349 (50.1%, 95% CI 44.9 to 55.4) and 77/291 (26.5%, 95% CI 21.7 to 31.8) respectively. No maternal age, gestational age, birth weight or Apgar value was implausible, but chronological errors were common: recorded and date-derived length of stay differed by >1 day in 92/225 (40.9%) and admission preceded birth in 35/257 (13.6%). Agreement with rule-derived classification was high for very low birth weight (kappa 0.89, 95% CI 0.83 to 0.95) but moderate for LBW (kappa 0.53, 95% CI 0.32 to 0.73; McNemar p=0.001). Complete documentation was more likely with three or more record sources (adjusted prevalence ratio 1.80, 95% CI 1.05 to 3.08) and less likely for newborns who died (0.46, 95% CI 0.25 to 0.84). Providers described documentation competing with workload. Conclusions Identification, gestational age, birth weight and disposition were documented sufficiently for classification and mortality analysis; timing fields, maternal history and transport variables were not. Preterm-birth prevalence was not estimable without an all-delivery denominator. Strengths and limitations of this study All completeness, consistency and plausibility estimates were computed from the original, unedited abstractions, without imputation or correction, so that the records were assessed as they exist in routine practice. Three routine data sources from twelve facilities were assessed against prespecified rules, and every dataset was verified cell by cell against the original electronic export before analysis. Recalculated classifications served only as a rules-based consistency reference; no independent clinical gold standard (such as direct observation or re-measurement) was available, so accuracy could not be assessed. Only ten facilities contributed readiness data, so facility-level correlations are exploratory and imprecise, and no complex facility-level model was fitted. The explanatory qualitative component comprised two provider interviews at the county referral hospital and therefore illustrates rather than represents documentation practice across all facility levels.
Source: medRxiv
Language: English
Open Access status: green
OpenAlex ID: W7236036540
Cites: 40
Cited by:
Related to: 10