Routine health information system data quality in a hill and mountain district of Nepal: a cross-sectional desk review of four years of DHIS2 data from Sankhuwasabha, 2021-2025

Introduction. Nepal's routine health information system runs on DHIS2, but no published data-quality assessment covers its mountainous districts. The method raises a second issue: the WHO Data Quality Review (DQR) tests internal consistency partly through related-indicator pairs, assuming each indicator is recorded independently. DHIS2 validation rules act on paired values at entry, so a rule-governed pair may agree for reasons unrelated to recording quality. Methods. In this study, a WHO DQR Module 1 desk review was conducted on aggregated DHIS2 data from all 132 facilities in Sankhuwasabha district, analysed at district and local level (palika) resolution: seven datasets for completeness and timeliness, and 17 tracer indicators for internal consistency, for fiscal year 2081/82 (2024/25) against the three preceding years (2021/22 to 2023/24). Pairs were classified in advance as rule-governed or free, and three denominator-free ratios compared with Nepal Demographic and Health Survey (NDHS) 2022 estimates for Koshi Province. Results. Completeness was at least 90.6% in every dataset and unit. Timeliness fell below 90% in 22 of 70 unit-dataset combinations, concentrated in four remote local levels. Of four rule-governed pairs only iron and deworming changed materially, 1.42 to 1.00; two never agreed, indicating rules that query rather than block entry. The free pairs diverged: Penta-1 to Penta-3 dropout fell from +3.6% to −5.4%, negative in 9 of 10 local levels, and BCG was below Penta-1 in 8 of 10. District totals were inverted at both steps (BCG 1,537, Penta-1 1,666, Penta-3 1,756), which redistribution cannot produce. Conclusion. Reporting is nearly complete, but the data carry a persistent immunization anomaly that completeness and timeliness miss, pointing to a systematic recording process. Because a rule-governed pair may agree by configuration, DHIS2 assessments should classify pairs against the active validation-rule set before interpretation.