Can a community health worker administered postnatal checklist increase health-seeking behaviors and knowledge?: Evidence from a randomized trial with a private maternity facility in Kiambu County, Kenya

Background: Since the 2009 WHO and UNICEF recommendation that women receive home-based postnatal care within the first three days after birth, a growing number of low-income countries have explored integrating postnatal home visit interventions into their maternal and newborn health strategies. This randomized trial evaluates a pilot program in which community health workers (CHWs) visit or call new mothers three days after delivery in peri-urban Kiambu County, Kenya. Methods: Participants were individually randomized to one of three groups: 1) early postnatal care three days after delivery provided in-person with a CHW using a simple checklist, 2) care provided by phone with a CHW using the same checklist, or 3) a standard of care group. Surveys were conducted ten days and nine weeks postnatal to measure outcomes related to compliance with referrals, self-reported health problems for mother and baby, care-seeking behaviors, and postnatal knowledge and practices around the recognition of danger signs, feeding, nutrition, infant care and family planning. Results: The home visit administration of the checklist increased the likelihood that women recognized postnatal problems for themselves and their babies and increased the likelihood that they sought care to address those problems identified for the child. In both the home visit and mobile phone implementation of the checklist, actions taken for postnatal problems happened earlier, particularly for infants. Knowledge was found to be high across all groups, with limited evidence that the checklist impacted knowledge and postnatal practices around the recognition of danger signs, feeding, nutrition, infant care and family planning. Conclusion: We find evidence that CHW-administered postnatal checklists can lead to better recognition of postnatal problems and more timely care-seeking. Furthermore, our results suggest that CHWs can affordably deliver many of the benefits of postnatal checklists. Trial registration: ClinicalTrials.gov NCT02104635 ; registered April 2, 2014. The intervention was implemented with Jacaranda Health, a private-sector social enterprise located in Kiambu County, Kenya providing high quality and low-cost maternal and newborn healthcare to low-income women in peri-urban areas of Nairobi. These regions are densely populated with characteristic challenges of poverty including poor access to water and sanitation, food insecurity and safety concerns [6, 18]. Total fertility rates for women in urban poor regions are slightly higher than that of Nairobi (3.1 compared to 2.8) yet still lower than the national average of 4.6 [4, 6]. At Jacaranda Health’s 10-inpatient bed hospital in a peri-urban setting just outside Nairobi, women were approached for recruitment and written informed consent after a normal delivery just prior to their discharge home. Women were eligible if they had a complication-free delivery, their newborn experienced no visible complications, were over 18 years old, provided two phone numbers where they could be reached, and resided within 20 km of the Jacaranda Health facility. These eligibility criteria reflect Jacaranda Health’s referral guidelines during the study period. During the study, Jacaranda Health had the capacity to conduct normal deliveries, provide basic emergency obstetric care and refer women who were at increased risk of a complication in their pregnancy to higher-level hospitals. Jacaranda’s eligibility criteria were adapted from international guidelines including the World Health Organization and the U.K.’s National Institute for Clinical Excellence (NICE). The eligibility criteria excluded women with six or more past deliveries, advanced age (older than 35 years), abnormal values of ANC diagnostics (e.g. high blood pressure), history of medical complications, and history of obstetric complications. The CHWs participating in our study were employed exclusively as Jacaranda staff and managed by a nurse midwife dedicated to outreach and community efforts. CHWs in our study were recruited from the larger area pool of CHWs who are recruited from barazas (local community meetings) and who receive foundational training in basic areas and are supervised by Community Health Extension Workers (CHEWs). The recruitment and training of these CHWs varies significantly in practice across Kenya [19]. The randomized trial took place between April 2014 and October 2014. Patients were individually randomized prior to enrollment using numeric patient identifiers assigned by Jacaranda Health. A unique identifier is given to each Jacaranda Health client seeking any service (including antenatal care, delivery, postnatal care, and child wellness care) during the client’s first visit to Jacaranda. Randomization was conducted by assigning each of these unique identifiers to one of the three central treatment groups with equal probability: a CHW home visit three days after delivery, a phone call from a CHW three days after delivery, or a standard of care group that received a customer service phone call about their experience.1. At the start of study recruitment, all pre-existing assigned Jacaranda identifiers were randomized. For clients who had attended antenatal care at Jacaranda, randomization was stratified using demographic and health variables collected in routine patient documentation. Identifiers were stratified by terciles of expected delivery date, distance from the client’s home to the facility, and primiparity. Random assignments were also given to all future Jacaranda IDs not yet in use; these assignments were made to each of the three treatment groups with equal probability. Random assignment of patient identifiers was done using a randomization sequence generated by the principal investigators with STATA 11, (StataCorp, College Station, TX). For women randomly assigned to the checklist groups, CHWs were trained to screen for maternal and newborn danger signs, to deliver targeted postnatal health education, and to refer mothers and their newborns to facility-based care if necessary using a checklist to guide them through the process that was available in English and Kiswahili (Additional file 1). Community health worker training was conducted by the research program manager and a designated nurse for the program. Prior to independent administration of the intervention, CHWs shadowed other Jacaranda Health nurses conducting postnatal health education counseling after delivery and observed a nurse-conducted home visit. CHWs contacted women either in their home or by mobile phone three days after delivery. The checklist was developed using international guidelines and academic publications and included components shown in Table 1 [12, 20, 21]. Newborn assessment characteristics such as poor feeding, fever, and jaundice also reflect the 2014 WHO guidelines.2 If a CHW detected a sign or symptom of maternal or neonatal illness through the assessment checklist on Day 3, the mother or child were referred to the nearest facility for curative care.3 The protocol specified that all complications, except for cracked nipples in isolation, were to be referred. For all referrals, a nurse conducted a phone follow-up the following day to ensure the woman and her child received appropriate care. Drafts of the checklist were locally pre-tested to ensure accuracy and comprehension among the communities and CHWs. Managers conducted regular audits of referral and checklist documentation to ensure appropriate completion and adherence to protocols. Topics covered by day 3 postnatal checklist CHWs completed a thorough four-day training to conduct screenings using the checklist and to counsel mothers and caregivers on essential postnatal health education. The training curriculum was adapted from international resources and modified to be relevant to specific programmatic contexts [20]. CHWs were evaluated through pre and post-test mechanisms to ensure thorough understanding and comprehension of essential information including: role-plays, observation of nurse-conducted visit and calls and, upon concluding the training, were signed-off by a clinical staff that their performance met quality standards. Trained research assistants obtained written informed consent for participation in the study prior to discharge from the Jacaranda Health hospital in Ruiru. Recruitment and survey administration occurred on a rolling basis throughout the study period based on the date that women delivered at the Jacaranda Health hospitals. Estimates of program impact were analyzed using administrative, programmatic, and survey data. Administrative data is routinely collected for all Jacaranda Health clients and uploaded into their medical records; these data include information on patient history and health services received, as well as demographic data. Electronic medical records data were available for all enrolled participants. The second data source was programmatic data collected by the CHW during administration of the checklist. The final source of data was follow-up participant surveys, administered by phone by trained research assistants at ten days and nine weeks after delivery. All participants were contacted at each data collection time point, regardless of whether they were previously reached (including those who were never reached by either a home visit or phone call on the third day after giving birth). Inability to reach participants for both the day 3 interventions and the follow-up phone surveys was assessed based on the following protocols: For the day 3 interventions, participants assigned to receive a customer service call (standard of care arm) or phone administration of the checklist were contacted up to four times by phone on the scheduled day (using all phone contact information provided by the participant at enrollment). The participant was considered unreachable if none of the four calls were answered. For the home visit arm, the CHW contacted the participant by phone on the morning of her scheduled visit to confirm the location of the residence, and subsequently knocked on the door or otherwise introduced herself at the home. If the participant was not reachable by phone while the CHW was in the field that day or was not available at the home for the entire day, she was considered unable to be reached by the checklist treatment. Regardless of whether the participants were reached on day 3, all were scheduled for follow-up mobile phone surveys at ten days and nine weeks after delivery. Attempts were made to contact participants for the day 10 follow-up survey beginning on day 10 after delivery, with daily contact attempts made for the subsequent seven days after the scheduled date. If the participant was unreachable or unable to complete the phone survey during this time, she was considered unreachable. Likewise, attempts were made to contact participants for the nine week survey beginning on the scheduled date, with daily contact attempts made for a subsequent 14 days. Postnatal maternal and newborn outcomes are captured at three time periods – three days, ten days, and nine weeks post-delivery. Day 3 outcomes were collected during administration of the checklist by the CHW and include maternal or newborn complications detected and referrals made for complications. These outcomes were collected in both treatment arms but not the standard of care arm. Day 10 and 9 week survey outcome measures include care-seeking behaviors for mother and newborns and both knowledge and practice of infant care, nutrition, feeding and recognition of danger signs. Table 2 provides definitions of all outcome measures. Definition of primary outcomes In order to avoid concerns about multiple outcome testing, levels of postnatal health knowledge were assessed by constructing summative indices of the number of items that participants were able to name in response to knowledge questions as suggested by O’Brien [22] and Kling and Liebman (2004) [23]. Knowledge outcomes were also converted into binary variables, indicating whether the participant was able to name a specified number of knowledge items. This study was approved by Institutional Review Boards at Harvard School of Public Health and the Ethical and Scientific Review Committee of African Medical Research Foundation (AMREF) in Nairobi, Kenya. The study design was registered at clinicaltrials.gov with identification number {“type”:”clinical-trial”,”attrs”:{“text”:”NCT02104635″,”term_id”:”NCT02104635″}}NCT02104635. Because of the small sample of this study, our study is powered to see only large changes in the primary care-seeking outcomes. With our original registered sample size of 109 individuals, we had 80 % power to detect a change of 30 percentage points in maternal care-seeking from a base of 8 % of women seeking care, using a two-sided test comparing any two treatment groups and a 5 % significance level threshold. Because child care seeking levels are so high, even with the standard of care (over 95 % of the sample seeks some care for their newborn prior to the 10 day survey), we were not powered on this outcome. Because of loss to follow-up our actual sample size for the survey conducted 10 days after delivery was 83. With our final sample, we have 80 % power to detect a change of 37 percentage points in maternal care-seeking compared to the standard of care, using a two-sided pairwise test and a 5 % significance level. Power calculations were conducted using STATA, version 12.1. We test for differences in outcomes across study arms on an intention-to-treat basis, with the treatment arms defined as participants’ randomized treatment assignment and the sample including all respondents where data is available (not just those reached by a day 3 intervention). We also provide Additional file 2: Tables A1-A3, which recreate all regression analyses restricted to the subset of participants who were successfully reached for the day 3 intervention. Program impacts were estimated using logistic regression for binary outcomes and ordinary least squares (OLS) regression for continuous outcomes. OLS and logistic estimates of program impact are presented both in unadjusted and adjusted models. Outcomes of interest are regressed on separate binary variables for assignment to one of the treatment arms. Adjusted estimates include controls for female participant’s age, marital status, employment status in the past 12 months, attendance at one or more antenatal care visits at a Jacaranda Health facility, and a binary indicator for whether the participant was enrolled in a postnatal contraceptive subsidy program as part of a concurrent randomized controlled trial. All analyses were performed using Stata software, version 12.1 (StataCorp, College Station, TX). Participant flow through enrollment, administration of the day 3 interventions, and follow-up surveying are presented in Fig. 1. Programmatic data were collected for the 77 women who were successfully reached at day 3 by the assigned intervention. The rate of reaching individuals at day 3 for the intervention was higher in the checklist call group (76 %) compared to the checklist home visit group (59 %). This difference may be due to difficulties finding participants either because they are traveling or because their households were difficult to locate. Participant survey data were collected for 83 participants at ten days post-delivery (24 in the standard of care arm, 32 in the phone-administered checklist arm, and 27 in the home visit-administered checklist arm). Participant surveys were collected for 59 participants at nine weeks after delivery (17 in the standard of care arm, 23 in the phone arm, and 19 in the home visit arm). Four participants were unable to be contacted at any data collection point after enrollment, due to errors in their contact details. Randomized trial study design and participant flow