This study was conducted in the Eastern Province of Rwanda at Kirehe District Hospital (KDH), a 233-bed facility operated by the Rwanda Ministry of Health and supported by Partners In Health, an international NGO. KDH serves a catchment area of 364,000 people including patients from Mahama Refugee Camp, which comprises over 50,000 people [17]. In Rwanda, over 91% of women deliver in health facilities [18]. Women in labor first present to their local health center where most vaginal deliveries take place. Complex cases and cases requiring surgical intervention are transferred to the district hospital to be assessed and managed by a general practitioner, who performs the c-section procedure, if indicated. After surgery, the woman is admitted to the postoperative ward for an average of 3 days for monitoring, medication administration, and wound checks. Before leaving the hospital, she receives postdischarge instructions directing her to the health center nearest to her home for follow-up and wound dressing changes.

This study included women aged ≥18 years who received a c-section at KDH between November 2, 2017, and September 4, 2018, and were residents of Kirehe District. Women who developed an SSI while being an inpatient or who remained inpatient at KDH past postoperative day (POD) 10 were excluded as they were not able to participate in a POD 10 home visit. Women who resided in Mahama Refugee Camp and were therefore not covered by the CHW network were also excluded.

We hired 4 study CHWs (sCHWs) using the Rwanda Ministry of Health criteria though a community-led process. The sCHWs received a 4-week training on implementation procedures, including the following: education on the Rwandan health sector; post–c-section follow-up; operating mHealth tools; best practices in wound photography; basic SSI physiopathology (including signs and symptoms of an infection); and how to examine surgical wounds and change wound dressings. Previously, we led a 7-month SSI protocol development study at KDH to identify a simple screening protocol with high accuracy to diagnose SSIs. In that study, three questions were found to have sufficient sensitivity and specificity for SSI diagnosis, which comprised the following: (1) fever since discharge from the hospital, (2) increasing pain since discharge, and (3) the presence of discolored wound discharge [19].

This randomized study included three arms—Arm 1, where the sCHW visited a participant’s home on POD 10 (SD 3 days) and administered the SSI screening protocol; Arm 2, where the sCHW called the participant and administered the SSI screening protocol over the phone on POD 10 (SD 3 days); and Arm 3, where the participant received the standard of care instructions to return to the health center for follow-up. In Arm 2, the sCHW attempted phone calls while sitting in the study office and made 3 call attempts before deeming a patient to be inaccessible. If a patient was deemed inaccessible, they were included in the evaluation of the feasibility of the intervention and classified as “not successfully assessed for SSI.” However, these individuals were not included in the evaluation of the presence of an SSI. If a participant in Arm 1 or 2 was suspected of having an SSI by the responses to the screening protocol, the sCHW was prompted to refer her to a nearby health center for additional medical care.

All data collected were entered and stored using REDCap (v8.10.20), a secure web application certified for medical research studies [20]. The study staff enrolled and randomized eligible participants at discharge, independent of any patient factors, to one of the three study arms. The study staff prepared study packets in sealed envelopes numbered consecutively. REDCap was then used to randomly generate arm assignments to each packet using simple randomization in a 1:1:1 ratio [16]. All consenting participants’ demographic and socioeconomic data were collected using a self-reported questionnaire administered by a trained study data collector. In addition, the study staff extracted clinical data from the patients’ medical files. Upon discharge, the patients received a packet with arm-specific follow-up and general discharge instructions.

Each health center in the catchment area and KDH had a study-specific patient registry to document return to care. Study staff entered the following details into REDCap: return to care status, SSI diagnosis (by nurse), treatment received, hospitalization, patient referral, and need for surgical procedures, if any. Data collectors contacted each woman on POD 30 to validate what was captured in the registry and to ensure that no follow-up visits were missed.

Data were analyzed using Stata (14.0 version, Stata Corp) statistical software. We characterized study participant demographics and clinical characteristics using descriptive statistics. For the primary outcome, a patient was classified as having returned to care if the return visit was documented in the health center registry or if the patient reported returning to the health center during the POD 30 follow-up call. In this primary analysis, we excluded anyone without information on return to care by POD 30 from analyses. Feasibility assessments across Arms 1 and 2 were reported as the percentage of visits where that specific task was completed. We used a Fisher exact test at α=.05 significance level to assess the association between patients’ return to care and interventions implemented in Arms 1 and 2 as compared to Arm 3 (standard of care). We used a logistic regression model to assess the impact of study interventions on return to care, controlling for potential confounders that were unbalanced at baseline. In this primary analysis, we excluded anyone without information on return to care by POD 30 from analyses. We used chi-squared tests to assess for differences in having information about return to care by study arm and by patient demographics. We also conducted a sensitivity analysis, whereby any individual missing information on return to care was presumed to have not returned to care.

The estimated sample size was 364 patients per arm, for a total of 1092 patients. We anticipated an SSI rate of 15%, which would result in 55 SSIs per arm. Assuming an 80% return to care rate in the two intervention arms and a 40% return to care rate in the standard of care arm, we would have 81% power to detect a difference between the two intervention arms as compared to the routine care arm. The trial was halted when 1166 patients were enrolled (in excess of the targeted sample size of 1092).

Eligible women gave informed consent prior to participation. The study team members provided information in Kinyarwanda, including details of the three study arms and the right to withdraw from the study or refrain from giving information at any stage. Deidentified data were collected and managed using REDCap. This study was approved by the Rwanda National Ethics Committee (848/RNEC/2016) and Partners Human Research Committee (2016P001943/MGH). Seven months into the study, a Data and Safety Monitoring Board reviewed the study participants’ safety, data quality, and midterm outcomes, and deemed it appropriate to continue to study completion.

Surprisingly, nearly all patients in our study returned to care at least once by POD 30, with no significant difference in follow-up between arms. This contrasts with the findings in the Central African Republic, where a study reported that only 25% of surgical patients returned for a POD 30 follow-up visit [21]. A possible reason for this is that Rwanda, a small country with a strong functional decentralized public health system [22], offers greater access to follow-up care.

In this study, we observed that home-based follow-up care of participants allows the sCHW to enter the women’s homes, physically examine them, and take a photo of their wound. As close to 90% of participants in Arm 1 successfully visited and were assessed for SSI, we found that home visits are a feasible way to conduct post–c-section care. In rural Rwanda and many other low-resource settings, CHWs already provide in-home screening for child health [10], maternal health [23], and HIV care [24] referrals. The high feasibility of in-home screening could be due to the familiarity women have with the CHW system and how they value support from CHWs [24,25].

On the other hand, SSI screening by phone excluded close to 30% of women. Other studies in Tanzania [12] and Sudan [26] also demonstrated gaps in using telephone calls for postdischarge surveillance of SSIs. Currently, only 54% of households in Rwanda own a mobile phone [18]. Telephone-based interventions may be more feasible as phone access and network coverage expand. Two recent systematic review articles assessing the use of smartphones to identify SSI found that there are few articles in the literature, the majority are in high income settings, and they require smartphone ownership by patients. We have not found any other experience of CWH home follow-up for SSI identification and care [27,28]. Currently in the rural Rwandan setting, in-person sCHW visitation at the patient’s home provides greater follow-up coverage than phone calls alone.

Despite the null results in the difference between rates of return to care, this study has important implications linked to our understanding of the financial risks associated with health care seeking in this population. Undergoing a c-section is a financial burden for women in rural Rwanda [29]. This is true even for women covered by community-based health insurance. Our group has previously reported that the median out-of-pocket cost of transport for a single visit to the health center for women who received a c-section at KDH is up to 10% of their monthly income and that those who spent more money had increased risk of SSI [30]. Transportation cost was also self-reported by patients from Arm 3 of the study to be a barrier to health care seeking in the postoperative period [11]. These costs of transport were uniform across the arms of the study, as all three arms had equal rates of return to clinic.

In this setting, approximately 90% of patients do not develop an SSI. Treatment of an SSI is a principal reason for return to health center for care that cannot be provided by a CHW. Thus, accurate home rule-out of a post–c-section SSI can eliminate the need for 90% of women undergoing c-section to make the return journey and incur the out-of-pocket expense of transport to the health center. Given the financial burden of transport to the health center and the feasibility of in-person CHW SSI screening, leveraging the existing CHW system in Rwanda to bring postcesarean care to women’s homes could reduce both financial barriers to care and medical impoverishment. Further analysis is needed to determine the effect that home-based surgical wound monitoring would have on reducing unnecessary visits to the health center, health system cost savings, and workload on clinicians, though promising results have been shown in other settings [31,32]. This study’s findings will also be used to develop a supporting app to facilitate in-person SSI screening by CHWs.

This study had several limitations. Health center data may not have been consistent in quality due to variations in study patient tracking and data collection processes across sites. Targeted interventions, including calling health centers weekly with a list of expected enrolled patients and monthly in-person audits of each health center’s registry, were implemented to improve patient tracking. We also called all patients on POD 30 regarding their follow-up activities and SSI diagnoses. There was 100% agreement between data from registries and phone calls for the 228 patients from Arms 1 and 2 [19]. Secondly, we could not accurately or consistently capture dates of health center visits. As a result, we do not know when within-POD-30 women returned to care and whether there were differences between the intervention arms and the control arm. Additional research is needed to assess how CHW interventions affect the timeliness of return to care for post–c-section SSI evaluation.

We did not observe a difference in the rate of return to the health center between women who were visited at home, who called at home, and who asked to continue with standard of care visits. In fact, women in all groups demonstrated high levels of health seeking behavior. However, our study found that home-based post–c-section follow-up by CHWs facilitated by an mHealth app to identify and refer SSIs is feasible. Our previous studies have shown that health center visits can pose a significant financial burden on women following c-section. Therefore, use of home visits for postoperative care could greatly reduce the nonmedical costs related to transport for routine follow-up for women who do not develop SSIs. Home, mHealth-enhanced, visits were also found to be more effective than phone-based follow-up for connecting CHWs with patients. Thus, home visits have the potential to greatly reduce the patient’s economic burden of post–c-section care. Future studies to understand the acceptability of CHW home visits for patients and health care workers are needed before this can be adopted as a standard care protocol.