Increasing the Use of Verification Results within Performance-Based Financing Programs

Abstract

Background: Within Performance-Based Financing (PBF) programs, independent verification is a known cornerstone to ensure the reported results are accurate. While verifications typically include reviews of documentation (quantity verification) and patient follow up (community verification), the literature suggests many programs focus payment calculations and data analysis on the quantity verification results, which measure internal record alignment but do not assess the validity of records (e.g. whether records represent delivered services). Further, to reduce the high cost of verifications, there has been a call for risk-based sampling whereby estimated probabilities of overreporting determine both the frequency and intensity of verification efforts. However, there is very little research on the factors that are associated with overreporting, which program planners need to focus sampling in high-risk areas. Objective: For a Voluntary Medical Male Circumcision (VMMC) PBF program in Zimbabwe that took place from 2016-2018, Aim 1 sought to understand the extent to which reported VMMCs could be verified in records and with patients, the paths and processes that led to the verification results, and if the two sources (records, patients) aligned at the facility-level. For the same program, Aims 2 and 3 both sought to understand the factors associated with overreporting.Methods: This study used programmatic secondary data from verifications activities, additional data gathered from Zimbabwe’s Ministry of Health and Child Care (MoHCC) for explanatory variables, as well as primary data collected through interviews with community verifiers. For Aim 1, descriptive statistics are presented as well as the reasons behind the results from the views of the verifiers’; the correlation between the quantity and community verification performance scores was also assessed. For Aim 2 and Aim 3, generalized mixed effects models were employed each with a different binary outcome to assess the relationship between health facility characteristics and overreporting. Aim 2 was at the facility-level and overreporting was defined based on the quantity verification results. Aim 3 was at the patient-level and overreporting was defined based on the community verification results. Results: For Aim 1, 25% (36,877/146,924) of reported VMMCs were reviewed and 3,676 patients were interviewed across 41% (144/355) of VMMC locations. We found that quantity verification results were not a good proxy for community verification results and that multiple parties, not just health facility staff, were responsible for data discrepancies—realities that point to the need for our proposed results to action framework. In Aim 2, we found that VMMC locations that were further away from the fixed health facility were less likely to overreport and that each additional staff at the VMMC location was associated with a 73% increase in the odds of overreporting. For Aim 3, we found that patients in the target age range, which were renumerated at a higher price point, were less likely to be interviewed and over two times more likely to be classified as overreported. Conclusions: Our findings are part of emerging evidence that suggest quantity verifications can portray a misleading and overly positive assessment of reported results tied to payment. Because tying payment to facility records alone risks overpaying for services and misreporting performance, programs should continue investing in community verifications and in-depth analysis of the results. The appropriate actions to take with community verification results often vary based on who was responsible for the issue and the underlying cause. To help navigate these complexities and increase the use of community verification findings, PBF programs should consider using and improving our proposed results to action framework.