Prevalence of Untreated Surgical Conditions in Rural Rwanda: A Population-Based Cross-sectional Study in Burera District

Key Points

Question  What is the prevalence of untreated index surgical conditions in a low-income country?

Finding  In this cross-sectional study, 2165 randomly selected individuals underwent physical examination in a rural district in Rwanda, which revealed a 12% prevalence of untreated surgical conditions.

Meaning  There is a significant burden of untreated surgical conditions in rural Rwanda, nearly double the previous estimates, that health systems planners must account for as surgical capacity develops.

Importance  In low- and middle-income countries, community-level surgical epidemiology is largely undefined. Accurate community-level surgical epidemiology is necessary for surgical health systems planning.

Objective  To determine the prevalence of surgical conditions in Burera District, Northern Province, Rwanda.

Design, Setting, and Participants  A cross-sectional study with a 2-stage cluster sample design (at village and household level) was carried out in Burera District in March and May 2012. A team of surgeons randomly sampled 30 villages with probability proportionate to village population size, then sampled 23 households within each village. All available household members were examined.

Main Outcomes and Measures  The presence of 10 index surgical conditions (injuries/wounds, hernias/hydroceles, breast masses, neck masses, obstetric fistulas, undescended testes, hypospadias, hydrocephalus, cleft lip/palate, and clubfoot) was determined by physical examination. Prevalence was estimated overall and for each condition. Multivariable logistic regression was performed to identify factors associated with surgical conditions, accounting for the complex survey design.

Results  Of the 2165 examined individuals, 1215 (56.2%) were female. The prevalence of any surgical condition among all examined individuals was 12% (95% CI, 9.2-14.9%). Half of conditions were hernias/hydroceles (49.6%), and 44% were injuries/wounds. In multivariable analysis, children 5 years or younger had twice the odds of having a surgical condition compared with married individuals 21 to 35 years of age (reference group) (odds ratio [OR], 2.2; 95% CI, 1.26-4.04; P = .01). The oldest group, people older than 50 years, also had twice the odds of having a surgical condition compared with the reference group (married, aged >50 years: OR, 2.3; 95% CI, 1.28-4.23; P = .01; unmarried, aged >50 years: OR, 2.38; 95% CI, 1.02-5.52; P = .06). Unmarried individuals 21 to 35 years of age and unmarried individuals aged 36 to 50 years had higher odds of a surgical condition compared with the reference group (aged 21-35 years: OR, 1.68; 95% CI, 0.74-3.82; P = .22; aged 36-50 years: OR, 3.35; 95% CI, 1.29-9.11; P = .02). There was no statistical difference in odds by sex, wealth, education, or travel time to the nearest hospital.

Conclusions and Relevance  The prevalence of surgically treatable conditions in northern Rwanda was considerably higher than previously estimated modeling and surveys in comparable low- and middle-income countries. This surgical backlog must be addressed in health system plans to increase surgical infrastructure and workforce in rural Africa.

Limited and inequitable access to surgical services in low- and middle-income countries (LMICs) is a growing concern in global health.¹ Of the 232.4 million major operations performed worldwide in 2009, only 3.5% occurred in the countries where 30% of the global population live.² In these largely impoverished countries, an inadequate surgical workforce^1,3,4 and dysfunctional community-based surgical referral systems^1,4 are among the barriers to providing adequate surgical care. However, the overall burden of specific surgical conditions in LMICs is largely unknown, which limits countries’ advocacy for appropriate interventions or evaluation of existing programs’ impacts.^5-7

Over the past decade, the global health community has recognized that it is important to accurately estimate surgical needs within communities.^1,2,8 To date, surgical literature relies primarily on mathematical modeling that uses data from operative logs and hospital records to estimate surgical needs in the community.^2,9,10 While this is a legitimate proxy in countries with adequate access to surgical care, the validation of the assumptions driving these models become weaker in LMICs where access to surgical care is poor.^11,12 In Rwanda, for example, only 1 operating room is available per 100 000 people,¹¹ and few people who need surgery present for care.¹¹

Recent population-based surveys have estimated the prevalence of surgical conditions in LMICs.^6,11,13 These surveys rely on self-reports of untreated possible surgical conditions. However, these methods are often not validated and are vulnerable to poor recall or limited understanding of surgically treatable conditions, resulting in low sensitivity.¹⁴ The goal of this study was to use physical examinations in a rural district in Rwanda to measure the prevalence of a set of conditions that the literature suggests compose the largest burden of nonobstetric surgical needs in LMICs.

This population-based, cross-sectional survey was conducted in Burera District, Northern Province, Rwanda. There, the Rwandan Ministry of Health provides nearly all health care with technical and resource support from the international nongovernmental organization Partners In Health/Inshuti Mu Buzima. Most Rwandans are insured through a community-based insurance program that covers outpatient and inpatient services with minimal copays.¹⁵

Households in Burera District were chosen for the study using a 2-stage cluster-sampling design. Thirty villages throughout the district were randomly selected. The probability of selection was proportionate to village population size, based on 2010 population data from the National Institute of Statistics, Rwanda. In the selected villages, an initial household was randomly chosen, then a succession of the nearest households were visited until 23 households had been sampled. Individuals of all ages who resided in these households were eligible for inclusion.

The target sample size (n = 3000 individuals) was calculated for a parallel study to assess the validity of a survey tool to identify surgical disease.¹⁴ The predicted sensitivity of the survey tool determined the study sample size.¹⁴

In March and May 2012, 5 Rwandan physicians (3 senior surgical postgraduates and 2 general practitioners with surgical training) visited the selected households and conducted standardized physical examinations. Prior to these visits, 10 standard index conditions were chosen based on their high burden of disease (as ranked by disability-adjusted life-years) and/or anticipated high prevalence based on literature reviews of surgical epidemiology in LMICs, supplemented by articles from high-income countries when LMIC data were unavailable, as previously described.¹⁴ The contextual relevance of these conditions was verified and augmented through a focus group with Rwandan surgeons. Examinations focused on the study’s 10 index surgical conditions, which were injuries/wounds, hernias/hydroceles, breast masses, neck masses, obstetric fistulas, undescended testes, hypospadias, hydrocephalus, cleft lip/palate, and clubfoot. We included all injuries, regardless of severity or need for operation, when there was physical evidence of an injury/wound and the participant reported it had occurred within the last 12 months.

The study coordinator trained all physicians for uniformity prior to examinations. Household members were examined in private, with a family chaperone. Breast and limited pelvic examinations were included for female participants, as were urogenital examinations for male participants.

Demographic data were collected for each household member from the head of the household, including age, sex, marital status, education level, insurance status, household size, access to personal transportation (car, bike and/or motorcycle), and travel time to hospital (by usual transportation method). Examinations were done on weekends, when it was anticipated that household members were more likely to be home. If any household members were not available during the initial examination, the physicians attempted to locate them before leaving the village. Households in which nobody was available for examination were dropped from analysis.

Ethical approval for this study was obtained from the Rwanda National Ethics Committee (institutional review board 1497) and the institutional review board of Boston Children’s Hospital, Massachusetts (protocol No. IRB-P00002717). Written and verbal informed consents were obtained for all physical examinations. Illiterate household members consented via fingerprint. For children, parental consent and patient assent were obtained. All data were deidentified.

When a surgically correctable condition was identified on examination, the affected individual was instructed to seek a hospital referral at the local health center. This procedure was developed by the district and national surgical practitioners, and approved by all ethical bodies based on the high enrollment in Rwanda’s community-based health insurance, which requires proper referrals to cover costs.

We described the household and individual characteristics, and we reported the overall and condition-specific prevalence of surgical conditions with 95% CIs. We used multivariable logistic regression to identify factors significantly associated with the presence of a surgical condition. First, bivariate association between the variable and surgical condition status was assessed using logistic regression that controlled for clustering of observations at the household level. Any factor associated with the presence of a surgical condition at the α = .10 significance level was entered into a full multivariable model. Factors were removed individually using a backward stepwise procedure until all remaining variables were significant at α = .05. We used Stata version 12 (StataCorp LLC) for statistical analysis. For multivariable analysis, we did not consider marital status for individuals younger than 21 years, the Rwandan legal age of marriage, and we combined marital status and age ranges into single variables for each age group 21 years and older. All unmarried individuals (single, divorced, and widowed) were combined for final multivariable analysis.

Of the 690 sampled households, 5 were excluded because the head of household did not consent. An additional 68 households were excluded because no individuals within the household were available for examination (Figure). In the 617 analyzed households, most had 4 to 7 members (n = 323, 52.4%) (Table 1). Most households (n = 376, 60.9%) were within 2 hours of the hospital by their primary mode of transportation, including walking, public transportation, bicycles, motorcycles, or cars. Only 8.9% (n = 55) reported access to private, motorized transportation. In two-thirds of households (n = 408/604, 67.6%), at least 90% of household members were insured.

The 617 surveyed households provided 2827 individuals eligible for the study. Of these, 2165 (76.6%) were available for examination (Figure). All available participants consented. Slightly more than half of examined individuals were female (n = 1215, 56.2%) (Table 1). More than half of participants were younger than 21 years (n = 1164, 55.7%). Of these, 29.4% were 5 years or younger (n = 342). Of 928 adult participants, most were married (n = 745, 80.3%) and almost all reported their education stopped at primary school or earlier.

The overall prevalence of any index surgical condition was 12% (95% CI, 9.2%-14.9%) (Table 2). Hernias/hydroceles accounted for half of the overall prevalent conditions (6%; 95% CI, 4.1%-7.8%); injuries/wounds were the second most prevalent condition (5.3%; 95% CI, 3.3%-7.3%). Together, neck masses, undescended testes, breast masses, clubfoot, hypospadias, and hydrocephalus were present in only 1.2% (n = 31) of sampled individuals.

In the bivariate analysis, the presence of an index surgical condition was not significantly associated with access to transportation, travel time to the nearest hospital, self-perceived denial of health care or being insured (Table 3). Individuals younger than 5 years, 5 to 20 years, and older than 50 years had twice the odds of having a surgical condition compared with those aged 21 to 35 years (<5 years: OR, 1.96; 95% CI, 1.18-3.28; P = .01; aged 5-20 years: OR, 1.91; 95% CI, 1.25-2.92; P = .003; >50 years: OR, 2.03; 95% CI, 1.22-3.4; P = .01). Condition prevalence by age is in the eTable in the Supplement. Women had a significantly lower odds of having a surgical condition (OR, 0.75; 95% CI, 0.58-0.98; P = .04). Previously married individuals (widowed, divorced, or separated) had significantly higher odds of a surgical condition compared with married individuals (OR, 1.83; 95% CI, 1.02-3.31; P = .44).

In the multivariable model, age and marital status remained significantly associated with having an untreated surgical condition (Table 4). Individuals younger than 21 years had twice the odds of having an untreated surgical condition compared with the reference group of married individuals aged 21 to 35 years (<5 years: adjusted odds ratio [aOR], 2.18; 95% CI, 1.21-3.92; P = .01; aged 5-20 years: aOR, 2.14; P = .003). Both married and unmarried individuals older than 50 years had significantly increased odds of a surgical condition (aOR, 2.27; 95% CI, 1.11-6.01; P = .01 and aOR, 2.59; P = .03, respectively) compared with married individuals aged 21 to 35 years. Unmarried individuals aged between 36 and 50 years were also more likely to have an untreated surgical condition (aOR, 3.62; 95% CI, 1.33-9.85; P = .01); so were younger unmarried adults (aged 21-35), although the difference was not significant (OR, 1.69; 95% CI, 0.73-3.85; P = .22).

In the adjusted model, women’s lower odds of a surgical condition (aOR, 0.79; 95% CI, 0.6-1.04; P = .09) were not significant. The reduced model that excluded sex minimally changed the adjusted ORs for the age-marital status variables, but not their significance (Table 4).

This study is the first, to our knowledge, to estimate the prevalence of multiple targeted untreated index surgical conditions by conducting community-level physical examinations in a rural area in an LMIC. Our results show an overall prevalence of 12%, or approximately 12 000 surgical conditions per 100 000 people. This estimate is considerably higher than prior estimates, both from modeling (6.5%)^1,16 and self-reported surveys in Rwanda (6.4%).¹¹ Estimates suggest 6145 surgical procedures per 100 000 individuals are needed annually in Eastern and sub-Saharan Africa to meet development goals.¹⁶ The discrepancy between this estimate and our observed prevalence highlights the need for health system planners to address a significant backlog of untreated conditions in the community when expanding surgical services.

The most common surgically treatable conditions identified in our population were hernias/hydroceles and injuries/wounds. The prevalence of hernias/hydroceles in this study was similar to that in other studies in sub-Saharan Africa (3.2%-6.6%).^9,17,18 In addition, injuries are often the most common presenting surgical condition in hospitals in sub-Saharan Africa.^1,4,19,20 In Ghana, 9.3% of surveyed individuals reported a significant injury in the previous year.²¹ Thus, expansion of surgical service in Rwanda must address these 2 problems.

Age and marital status were associated with having an untreated surgical condition, with young and elderly individuals at highest risk. This is consistent with previously published literature, which also found high surgical needs in young children^11,22 and elderly patients.^6,7,23 Further, our data suggest that unmarried individuals have an increased risk of untreated surgical conditions. To our knowledge, this has not been explicitly included in other surgical studies; however, a meta-analysis of factors associated with seeking medical care during labor in LMICs found that being married increased health care seeking in some studies, but not consistently across studies.²⁴ The link between marital status and access to surgical care should be studied. Other studies in LMICs have found that rural residence^7,11 and poor literacy^7,23 increased the risk of an untreated surgical condition. While education, transportation access, and travel time to hospital were not associated with untreated surgical conditions in our study, this may be owing to the limited heterogeneity of these variables within our population. Travel time may have limited some care-seeking, but most participants (60%) lived 2 hours or less from the hospital and most households (70%) reported always going to the hospital when needed. We suspect that some participants did not seek care because they did not realize they had a treatable surgical condition. In our parallel study, only 50% of the surgical conditions identified on examination were known by the surveyed head of household.¹⁴ Other studies also demonstrate that, when individuals perceive they are healthy, they do not seek care for surgical conditions.^6,7

The high prevalence of untreated surgical conditions highlights 3 areas to improve: surgical capacity, awareness of surgically correctable conditions, and access to surgical care. First, the health care system’s capacity to address these surgical conditions, both electively and on an emergency basis, must improve, particularly at district hospitals. In Rwanda’s decentralized health care system, the district hospital should provide basic and emergent surgical interventions. Hernia repair and injury care, the population’s most common conditions, should be available at the district hospital level, according to recommendations from the World Health Organization, the Lancet Commission on Global Surgery, and others.^1,8,25 Access to elective hernia repair can decrease morbidity and mortality, as emergency repairs have higher rates of bowel resection and mortality.²⁶ Two recent studies of surgical care at 3 district hospitals in Rwanda, including Burera District, demonstrated some challenges in providing timely access to surgical care at district hospitals. The first study found that only 25.3% of patients presenting for nonobstetric surgical diseases underwent operation prior to discharge, and 21% were referred to a tertiary hospital for surgical care.¹⁹ The second study found that 23% of injured patients were referred to the tertiary center, and nearly half of these (46.8%) waited more than 2 days to transfer.²⁷ Although most surgeons in Rwanda practice at tertiary referral centers,¹² most general surgery operations at the main referral center in the capital are emergencies (70%), and only 50% of surgical patients come from districts outside the capital.²⁸ The high referral rate, delayed transfers, and high emergency case load at referral centers suggest that capacity and infrastructure must improve at both the district and tertiary levels to address surgical backlogs.

Improving access to surgical care at the district hospital requires more functional operating theaters, equipment, and trained clinicians. Task-sharing can increase surgical output, including hernia repairs and injury care, with good results.^1,29 Additional training for general practitioners to repair hernias and care for injuries could help as additional surgeons are trained.³⁰ Rwanda is engaged in the Rwanda Human Resources for Health Program,³¹ which addresses the surgical workforce shortage by engaging US surgeons to work in Rwanda as partners to Rwandese surgical faculty to increase the number of surgeons who are trained annually. The program has tripled the number of trainees in the past 4 years. With more surgeons, the Rwandan Ministry of Health aims to decentralize specialists, including posting surgeons at provincial and district hospitals by 2020,³¹ a process already underway. Given that most procedures performed at district hospitals are emergency cesarean deliveries,¹² more surgeons for elective procedures must be accompanied with more infrastructure. Without additional functional theaters, neither district nor referral hospital staff can address the nonemergency surgical backlog identified in this study.

Additional challenges highlighted by our results include facilitating awareness of untreated surgical conditions and access to care for recognized conditions. To increase awareness, educational efforts should target the general public and health care workers, especially nurses and community health workers who contact patients at health centers and in the community. Increased awareness must be accompanied by increased access to care. The Lancet Commission on Global Surgery highlighted the importance to timely access to care by making the primary indicator the percentage of the population who could access key emergency procedures (cesarean deliveries, open fracture treatment, and laparotomy) within 2 hours.¹ Almost one-third of households in our study did not go to the hospital when they thought care was needed. This is consistent with a Ghanaian study where half of all injured patients did not seek care, including a quarter of the most severely injured patients.³² There are many reasons people in LMICs may not seek care, including the cost of care, cost of travel, distance to health centers, lack of awareness about the condition, and lack of trust in the health care system.^4,6,7,11,13 Expanding the limited ambulance transport system at district hospitals, and/or reimbursing patients or taxi drivers for travel costs, could potentially improve access to surgical care. Future studies should further elucidate barriers to seeking surgical care in Rwanda and evaluate programs that address those barriers.

This study has limitations to consider when interpreting the results. First, it focused on 1 district in 1 low-income country. Untreated surgical conditions could be higher in Burera District for unidentified reasons of climate, terrain, or culture. We suspect that other Rwandan districts and LMICs where most of the population are rural subsistence farmers have similar rates of surgical disease. We recommend that similar studies be done in these areas to support data-directed surgical system planning.

A second limitation was that only 72% of eligible participants were examined. Engagement in labor appears not to have been the cause of the nonparticipation of eligible individuals. While it is possible that individuals with surgical conditions were less likely to work outside the home and thus more likely to be examined for this study, only 2 people reported that they were unable to work. In addition, visits occurred on Saturdays and Sundays to minimize the number of individuals who were at work or school. Therefore, we suspect that the failure to be examined was randomly distributed in the study population.

A third limitation is that this study only evaluated select, nonemergency surgical conditions and therefore underestimates the true prevalence of surgically treatable conditions. We excluded emergency surgical conditions, such as acute abdomen and obstructed labor, and none were identified on examination. We also excluded variables related to cesarean deliveries, which in many LMICs compose most of the surgical procedures performed.^10,12,33 Future studies could expand the list of conditions assessed and include longitudinal and/or hospital data for more comprehensive estimates of the burden of surgical disease. Our study also included all injuries/wounds and did not distinguish between ones needing consultation from ones needing an operation. Assessing injury in a cross-sectional study is challenging, because injuries often heal, even without proper treatment. While this affects planning for operating theater capacity, knowing overall surgical disease prevalence is essential for surgical workforce planning.

Based on physical examinations in the community, there is a high prevalence of unmet surgical need in Rwanda. This highlights the need to expand district hospital surgical capacity in LMICs. With expanded capacity, efforts must focus on increasing community awareness about surgically treatable conditions and improving patient access to services. Our results highlight that current methods to estimate surgical burden in LMICs potentially underestimate the need, undermining resource mobilization. We recommend that other groups use physical examinations to assess surgical need and formally study barriers to surgical care seeking. These efforts will be invaluable to health system planning and to the monitoring and evaluation of surgical programs.

Accepted for Publication: July 23, 2017.

Corresponding Author: Rebecca G. Maine, MD, 4008 Burnett-Womack Building, Campus Box 7228, Chapel Hill, NC 27599-7228 (rebecca_maine@med.unc.edu).

Published Online: October 25, 2017. doi:10.1001/jamasurg.2017.4013

Author Contributions: Dr Maine had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Maine and Linden shared first author status.

Study concept and design: Maine, Linden, Riviello, Kamanzi, Mody, Niyonkuru, Mpunga, Meara, Hedt-Gauthier.

Acquisition, analysis, or interpretation of data: Maine, Linden, Riviello, Mody, Ntakiyiruta, Kansayisa, Mubiligi, Ntaganda, Hedt-Gauthier.

Drafting of the manuscript: Maine, Linden, Kamanzi, Hedt-Gauthier.

Critical revision of the manuscript for important intellectual content: Maine, Linden, Riviello, Mody, Ntakiyiruta, Kansayisa, Niyonkuru, Mubiligi, Mpunga, Ntaganda, Meara, Hedt-Gauthier.

Statistical analysis: Maine, Kamanzi, Hedt-Gauthier.

Obtained funding: Linden.

Administrative, technical, or material support: Maine, Riviello, Mody, Kansayisa, Mubiligi, Mpunga, Ntaganda.

Study supervision: Linden, Riviello, Mpunga, Hedt-Gauthier.

Conflict of Interest Disclosures: None reported.

Funding/Support: The Harvard Sheldon Traveling Fellowship helped support data collection. This study also received support from the Department of Global Health and Social Medicine Research Core at Harvard Medical School (Dr Hedt-Gauthier).

Role of the Funder/Sponsor: The Harvard Sheldon Traveling Fellowship had no role in the design and conduct of the study; management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.