Potential coverage of primary health care and performance indicators: differences between rural and non-rural municipalities of Tocantins, Brazil

Abstract Introduction: The paradox between the formal existence of a health service and effective access to it is a central issue in the primary health care (PHC) literature. Barriers to access combine geographical and sociocultural factors with the prevailing funding model, a problem amplified in regions of wide territorial dispersion. In Brazil, the National Primary Care Policy and the Previne Brasil programme sought to expand PHC coverage, but it remains unclear whether potential coverage is associated with care performance in the same way across different territorial contexts. This question is especially relevant in the state of Tocantins, in the Legal Amazon, where basic territorial coverage coexists with structural barriers to referral to more complex levels of care. This study assessed whether the association between potential coverage and performance differs between rural and non-rural municipalities. Methods: Ecological study including all 139 municipalities of Tocantins (107 rural and 32 non-rural), from 2022 to the first quadrimester of 2025 (10 four-month periods). Potential PHC coverage (e-Gestor) was correlated (Spearman, with 95% CI), using municipal means, with seven performance indicators of the Previne Brasil programme (Health Information System for Primary Care, SISAB) – prenatal care, gestational syphilis/HIV testing, dental care for pregnant women, cervical cytology, childhood vaccination, hypertension and diabetes – in the full sample and stratified by the rural–urban typology of the Brazilian Institute of Geography and Statistics (IBGE). The Benjamini–Hochberg correction for multiple comparisons (false discovery rate, FDR) was applied by stratum (main analysis, 7 tests) and jointly (14 tests, sensitivity analysis); a formal interaction test assessed moderation by rurality, and partial correlation controlled for population and socioeconomic level (GeoSES). An exploratory sensitivity analysis was restricted to coverage ≤150%. Results: In the overall sample (n=139), correlations were weak (rho 0.04 to 0.20) and none of the seven indicators maintained a significant correlation with coverage after correction for multiple comparisons (FDR-adjusted p ≥0.135); the partial correlation adjusted for population and GeoSES was also not significant for any indicator (partial rho −0.10 to 0.05; FDR-adjusted p=0.92). When stratified by rural–urban typology, non-rural municipalities (n=32) showed moderate positive correlations for six of the seven indicators (rho 0.40 to 0.62; FDR-adjusted p ≤0.027 with per-stratum correction); with joint correction, three remained significant (cervical cytology, hypertension and diabetes; FDR-adjusted p ≤0.005) and, after adjustment for population and GeoSES, only cervical cytology (partial rho=0.51; FDR-adjusted p=0.031 per stratum and 0.062 joint), whereas in rural municipalities (n=107) no correlation was significant (rho −0.16 to 0.00; upper 95% CI limit ≤0.19). The formal interaction test indicated moderation by rurality for six of the seven indicators (adjusted p 0.002 to 0.013 by Benjamini–Hochberg; result maintained with Holm and Benjamini–Yekutieli, p ≤0.033); the gestational syphilis/HIV indicator was not significant (p=0.063). With coverage ≤150% (n=43), the difference between strata was not significant. Discussion: The difference between strata, rather than a demonstrated conversion of coverage into access, is the main finding: the ecological design and indirect measures do not allow inference that the expansion of coverage has produced, or failed to produce, effective access. In the non-rural group the confidence intervals were wide; in the rural group the upper limits of the 95% CIs exclude correlations of the magnitude observed in the non-rural group. Potential coverage exceeded 150% in 69.1% of municipalities, which may reduce the informative variability of the measure. Travel, workforce provision, service organisation and funding, described in the literature, are hypotheses that were not assessed in this study. Conclusion: The relationship between potential PHC coverage and performance indicators differed between rural and non-rural municipalities of Tocantins: it was positive and moderate in non-rural municipalities and was not detected in rural ones. Because these are ecological associations between indirect measures, the results do not allow causal inference. Studies with direct measures of access, considering distance, workforce provision and work organisation in rural territories, are needed to inform PHC policies.