The effect of recall period on reported out-of-pocket health expenditure in Ghana

Background: Out-of-pocket health payments (OOPs) are a key indicator of health financing systems’ performance. Measuring OOPs through household surveys is challenging and yet it is the primary source of information in the absence of comprehensive data on user charges in the public sector and market data from the private sector. The choice of the recall period has been identified as a source of bias in previous studies. This study investigates the effect of two different types of recall periods on the agreement between OOPs reported by households and providers. Methods: Households were sampled for our community survey from the Navrongo Health and Demographic Surveillance System, Ghana. Two versions of a health expenditure module were developed differing only in the recall periods, “shorter recall periods” 2 weeks for medicines and outpatient care, 3 months for preventive care and 6 months for inpatient care and medical products. The longer recall periods were 4 weeks, 6 months and 12 months. Households from both community and provider sampling were randomly assigned to the two questionnaires. The providers included the hospital, one clinic and, health facilities and drug shops in the area. We estimated the ratio between the overall mean household OOPs and overall mean provider OOPs. We assessed agreement between the individual matched household-provider OOPs using Bland-Altman analysis. Findings: The short and long-recall period versions of the questionnaires were administered to 746 and 480 households with matching success to provider records of 72% and 84%, respectively. The most common spending categories were inpatient care and medicines in this sample. The overall mean OOPs reported by the households were higher than provider records for both recall periods. For matched household-provider data, there was no evidence of a difference in the agreement between the household and provider OOPs for inpatient care, the ratio of household to provider for the 12 months recall was estimated to be 0.74 (95% CI 0.45, 1.19; p = 0.22) that of the ratio of household to provider for the 6-month period, where less than 1 would indicate better agreement. For medicines, the ratio of 4 weeks to 2 weeks was 1.26 (0.93, 1.39; p = 0.39). Conclusion: There were considerable challenges in using provider data to assess the accuracy of reported OOPs in this setting. There was no evidence from this study that the agreement between household and provider data differed by recall period, however the confidence intervals of the effect were wide, and an effect cannot be ruled out. These findings underscore the need for caution in relying solely on provider-reported data and highlight the importance of considering alternative data collection methods to improve the accuracy of healthcare cost assessments.