Objective A non-negligible proportion of sub-Saharan African (SSA) households experience catastrophic costs accessing healthcare. This study aimed to systematically review the existing evidence to identify factors associated with catastrophic health expenditure (CHE) incidence in the region. Methods We searched PubMed, CINAHL, Scopus, CNKI, Africa Journal Online, SciELO, PsycINFO, and Web of Science, and supplemented these with search of grey literature, pre-publication server deposits, Google Scholar®, and citation tracking of included studies. We assessed methodological quality of included studies using the Appraisal tool for Cross-Sectional Studies for quantitative studies and the Critical Appraisal Skills Programme checklist for qualitative studies; and synthesized study findings according to the guidelines of the Economic and Social Research Council. Results We identified 82 quantitative, 3 qualitative, and 4 mixed-methods studies involving 3,112,322 individuals in 650,297 households in 29 SSA countries. Overall, we identified 29 population-level and 38 disease-specific factors associated with CHE incidence in the region. Significant population-level CHE-associated factors were rural residence, poor socioeconomic status, absent health insurance, large household size, unemployed household head, advanced age (elderly), hospitalization, chronic illness, utilization of specialist healthcare, and utilization of private healthcare providers. Significant distinct disease-specific factors were disability in a household member for NCDs; severe malaria, blood transfusion, neonatal intensive care, and distant facilities for maternal and child health services; emergency surgery for surgery/trauma patients; and low CD4-count, HIV and TB co-infection, and extra-pulmonary TB for HIV/TB patients. Conclusions Multiple household and health system level factors need to be addressed to improve financial risk protection and healthcare access and utilization in SSA. Protocol registration PROSPERO CRD42021274830
The protocol for this systematic review was registered on PROSPERO: CRD42021274830; and the findings reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [10]. We searched PubMed, CINAHL, CNKI, AJOL, African Index Medicus, PsycINFO, SciELO, Scopus, and Web of Science for studies published from 01 January 2000 to 31 December 2021 conducted in any of the 48 World Bank-defined SSA countries. Two authors (PE and LOL) independently searched the literature in February 2022 using search terms covering catastrophic health expenditure, financial catastrophe, risk factors, “factors associated with”, and sub-Sahara Africa–S1 Table. Boolean operators “AND” and “OR” were used to broaden the search. We also searched grey literature websites: New York Academy of Medicine Grey Literature and Open Grey; pre-publication server deposits: medRxIV and PrePubMed; Google Scholar®; and tracked references of included studies for relevant articles. We considered studies published in any of the six African Union languages: Arabic, English, French, Kiswahili, Portuguese, and Spanish; and translated non-English publications using a translation service. We underwent a moderation exercise to ensure uniformity; screened abstracts according to prior eligibility criteria (S2 Table); retrieved full texts for eligible studies; and resolved discrepancies by discussion. We used Mendeley Desktop® to identify and remove duplicates. At least two authors (PE, LOL, LUA, CAO, and UJA) independently extracted data from included studies using a template. We extracted the following data from each included study: authors names, publication status, study setting, publication year, study design, data source and authors’ description of the data representativeness, study period, sampling method, sample size (in households), and factors associated with CHE. We extracted reported adjusted odds ratio with the confidence interval at 5.0% statistical significance for each CHE-associated factor. Where two or more studies used the same secondary data to identify CHE-associated factors, we first assessed both studies for unique factors, but if similar factors were evaluated, we then considered the peer-review status of the studies; prioritizing peer-reviewed studies over non-peer-reviewed studies. Where a study described CHE-associated factor using more than one CHE definition, we extracted data for both definitions {10% total household expenditure (THE) and 40% non-food expenditure (NFE)}. For qualitative studies; we manually extracted all text under the headings ‘results/conclusions’. We cross-checked all extracted data for discrepancies which were resolved through discussion. At least two authors (PE, CAO, LUA, UJA, and LOL) independently assessed the quality of included quantitative studies using the Appraisal tool for Cross-Sectional Studies (AXIS tool) [11], and the Critical Appraisal Skills Programme (CASP) checklist for qualitative studies [12]. We resolved discrepancies in quality assessment scores by discussion until 100% agreement. We categorized the articles’ quality into high (studies met ≥ 70% of the quality criteria), moderate (between 40% and 69% of the quality criteria), and low (< 40% of the quality criteria). We used Microsoft Excel® to organize extracted data. We first summarized the included studies descriptively. To synthesize the evidence, we performed meta-analysis and narrative synthesis following the Cochrane Handbook for Systematic Reviews of Interventions and the Economic and Social Research Council (ESRC) Methods Programme [9, 13] guidelines. We pooled studies reporting quantitative estimates (odds ratios) from regression or matching analysis for CHE-associated factors in a random-effects meta-analysis to obtain pooled effect estimates. Random effects meta-analysis allows for differences in the treatment effect from study to study because of real differences in the treatment effect in each study as well as sampling variability [14]. Analyses were conducted using Stata version 16.1 (STATA Corp, College Station, TX). Where meta-analysis was not possible due to difference in the definition of CHE-associated factors, we analyzed the reported quantitative estimates narratively. For qualitative data, we independently performed line-by-line coding of text to group similar concepts and developed new codes when necessary. We organized free codes into descriptive major themes and sub-themes using an inductive approach as detailed by Thomas and Harden [15]. Each reviewer first did this independently and then as a group. Through discussion more abstract or analytical themes emerged and we resolved discrepancies between reviewers through discussion and consensus was achieved on all occasions. Finally, we globally assessed findings from both quantitative studies including meta-analysis for each CHE-associated factor–based of breadth of evaluation in included studies, consistency of an effect on CHE incidence, and methodological quality of included studies evaluating this factor–and when available, triangulated these with the participants’ lived experiences reported in qualitative studies to categorize each CHE-associated factor as either significant or marginal. We categorized a factor as “significant” if it was widely evaluated factors that consistently diminished or exaggerated the likelihood of CHE incidence. Otherwise, we categorized such factor as “marginal”. The original protocol was for a quantitative study. We decided to include qualitative studies to enrich our understanding of the key drivers of CHE based on individuals’ lived experiences, which population-based quantitative studies do not cover.