Factors associated with overreporting based on community verification results in a performance-based financing program in Zimbabwe.

BACKGROUND

While most Performance Based Financing (PBF) programs perform community verifications to confirm patients received reported services, many focus analysis and payment calculations on facility record verification due to their lower cost. Risk-based sampling can reduce the cost of community verifications by targeting areas with the highest risk of overreporting but there is little research on the factors associated with risk to guide sampling decisions.

OBJECTIVE

This study explores facility-level and district-level factors associated with overreporting within a PBF setting.

METHODS

Using community verification data from a Voluntary Medical Male Circumcision (VMMC) program in Zimbabwe, we estimated two binary outcomes with generalized mixed effects models. Our primary outcome is a measure of overreporting, defined as when interviewed patients did not plausibly confirm receipt of the VMMC. Additionally, we assessed factors associated with patients who were selected but ultimately not interviewed. We employed inverse probability of treatment weighting to address non-response and bootstrapping-based multiple imputation to address missingness.

RESULTS

We found that patients in the target age range, which were compensated at a higher price point, were less likely to be interviewed and over two times more likely to be classified as overreported compared to patients outside this age range (OR: 2.92, 95% CI: 2.38-3.59). Patients from outside the fixed health facility were more likely to be interviewed and less likely to be classified as overreported. In-person interviews as opposed to phone interviews appeared to be a worthwhile investment (OR: 1.61, 95% CI: 1.20-2.16).

CONCLUSION

We identified various factors that were associated with unsubstantiated VMMCs to inform risk-based sampling; however, our findings also suggest potential data fabrication. Programs should consider employing similar methods to reduce costs and increase the use of community verification data.