Development and evaluation of a subjective measurement of geographical hardness-to-reach: a survey of data collectors in a large-scale cluster randomised trial of azithromycin mass drug administration

Background: Distance from residence to the nearest health facility is a known predictor of child mortality. However, physical distance alone may not fully capture geographical hardness-to-reach. Subjective assessments from data collectors may offer additional insight into village-level accessibility and its relationship to health outcomes. Methods: This was a sub-study of Large-scale Assessment of the Key health-promoting Activities of two New Mass Drug Administration regimens with Azithromycin, a large-scale cluster-randomised trial of mass administration of azithromycin in Mali. We developed a 14-item survey and administered it to 25 data collectors to evaluate village geographical accessibility. We recorded survey completion time and respondent burden assessments to investigate the measurement's feasibility. We assessed convergent validity by correlation between two composite scores derived from survey items and objective measures (distance and travel time to the nearest health facility, the Malian national outreach programme status, and the average monthly rainfall) and predictive validity using under-five mortality as the criterion. Inter-rater reliability was evaluated by intra-cluster correlation coefficients (ICC). Results: The mean time of completing the survey was three minutes. Survey composite scores reflecting subjective village inaccessibility were associated with higher under-five mortality in the villages: one standard deviation change towards harder access associated with up to 8% higher under-five mortality rate, even after adjusting for distance to the nearest health centre. Internal consistency between the items ranged from moderate to high, with polychoric correlation coefficients ranging from 0.49 (items 'How long does it take for you to get to this village on rainy season' and 'Condition of the roads on non-rainy season') to 0.94 (items 'How hard this village is to reach' and 'How difficult it is for the residents to seek medical care'). Inter-rater reliability for all-item composite score and two-item composite score, was moderate to strong, with the average ICCs being 0.74 and 0.69, respectively). The composite scores showed expected associations with objective accessibility indicators, with Pearson correlation coefficients between all-item composite score and distance to the nearest health facility, average monthly rainfall, and travel time by car being, 0.36 (P < 0.001), 0.22 (P < 0.001), and 0.35 (P < 0.001), respectively. Pearson correlation coefficients between two-item composite score and distance to the nearest health facility, average monthly rainfall, and travel time by car were 0.34 (P < 0.001), 0.22 (P < 0.001), and 0.32 (P < 0.001), respectively. Conclusions: Our findings indicate acceptable feasibility and reliability, with moderate convergent and predictive validity. This simple survey can serve as a useful tool for evaluating hard-to-reach areas and guiding resource allocation in public health interventions, including mass drug administration campaigns. Registration: clinicaltrials.gov: NCT04424511 (parent trial).