Background

Diarrheal disease is the second leading cause of death among children under five years in Kenya [1]. In 2011, there were approximately 19 million episodes of diarrhea and an estimated 9500 deaths in children under five due to diarrhea [2]. Oral rehydration solution (ORS) and zinc are recommended by the World Health Organization (WHO) and United Nations Children’s Fund (UNICEF) to treat diarrhea in children [3]. ORS can prevent up to 93% of diarrhea deaths in children and zinc has been shown to reduce the duration, severity, and reoccurrence of diarrhea [4,5]. Both treatments cost less than $0.50. The Government of Kenya (GOK) adopted ORS into the national guidelines for pediatric diarrhea in the 1980s and updated the guidelines to include zinc in 2007. [6] However, the 2008-09 Kenya Demographic and Health Survey (KDHS) found that only 39% of children received ORS and only 0.2% received zinc for treatment of diarrhea [7]. The low coverages were driven by several key barriers, including low knowledge of recommended treatment by caregivers and providers, limited availability of zinc and ORS, expensive products in private outlets, and policies restricting zinc to patients with a prescription. To accelerate progress, the Ministry of Public Health and Sanitation (MOPHS), with the support from the Clinton Health Access Initiative (CHAI) and with participation of implementing partners such as the United Nations Children’s Fund (UNICEF), PATH, and Micronutrient Initiative (MI), launched a 5-year national plan to increase uptake of ORS and zinc to 90% between 2011 and 2016 [6].

Since the start of the program implementation, results from the KDHS conducted in 2014 found that ORS and zinc coverage had increased to 8% (from less than 1% in 2008-09) [8]. In this study, we aim to update coverage estimates for 2016 in order to evaluate changes in ORS and zinc coverage during the program period.

Study objective

The primary objective of this study is to measure ORS and zinc coverage in 2016 and to evaluate coverage changes since 2011. Second, the study aims to estimate the number of diarrheal deaths averted due to coverage changes between 2011 and 2016.

Program description

Kenya’s efforts to improve pediatric diarrhea treatment rates focused on four key domains: (1) developing supportive policy, which included allowing zinc to be dispensed over-the-counter and shifting government procurement to co-packaged ORS and zinc; (2) creating a sustainable and competitive market for ORS and zinc to improve availability and affordability; (3) increasing consumer demand for ORS and zinc; and (4) improving health provider practices in the management of pediatric diarrhea. Table 1 summarizes the activities by each domain.

One of the steps to developing a supportive policy environment was to increase access to zinc in private retail outlets by allowing zinc to be dispensed over-the-counter and enabling sales alongside ORS. The policy change also allowed for zinc to be marketed directly to caregivers and patients, which is prohibited for prescription medicines. This policy change was achieved in 2012. Second, to ensure ORS and zinc were administered together, the GOK recommended adoption of co-packaged ORS and zinc to all health facilities. Starting in 2015, the Kenya Medical Supplies Authority (KEMSA), a state corporation which oversees procurement and distribution of medical supplies, switched their procurement of pediatric diarrhea treatments from individually packaged units of ORS and zinc to co-packaged ORS and zinc. KEMSA also updated their product specifications to improve ease of administration and meet patient preferences. KEMSA’s new specifications required that zinc tablets were dispersible and taste-masked and that ORS was flavored. These technical updates [9,10], in addition to the revision of the national policy on management of diarrhea to include zinc, were broadly disseminated.

To develop a self-sustaining and competitive market for ORS and zinc, the program engaged with three local manufacturers of ORS and zinc – Cosmos, Universal, and Laboratory & Allied – to introduce co-packaged ORS and zinc starting in 2015 and to negotiate affordable prices. The development of a co-pack and price negotiations led to a reduction in the retail price of a full treatment course (four 500mL ORS sachets and 10 zinc dispersible tablets) from US$1.55 to US$0.67 (Figure 1). In addition, the program worked with marketing agencies to develop promotional materials on ORS and zinc and freely provided the promotional materials to suppliers for distribution by their sales teams during routine visits to private medicine outlets. Program monitoring data showed that the percent of private medicine outlets stocking both ORS and zinc increased from 30% in 2014 to 78% in 2015 (Figure 2).

Figure 1.  Retail price of full treatment course* before and after price negotiation. *Full treatment course = four 500mL oral rehydration solution sachets and 10 zinc tablets

Figure 2.  Percent of private pharmacies with both ORS and zinc in-stock on the day of the visit in 2014 and 2015. ORS – oral rehydration solution

To increase demand for ORS and zinc among caregivers, the program launched radio spots in the local language promoting ORS and zinc. The spots were designed to improve diarrhea prevention and treatment practices and to encourage prompt care seeking. To reinforce the mass media messages, interpersonal outreach was conducted through clinic-based “health talks” that targeted women attending routine antenatal care (ANC), immunization and well-baby visits. Standardized, illustrated “health talk kits” that included a pictorial chart booklet covering key child health topics, such as diarrhea management, vaccines, nutrition, and breastfeeding, were developed and provided to health workers. Health workers used these communication materials to provide daily informational sessions to large groups of caregivers who were waiting for appointments.

The MOPHS also re-introduced oral rehydration therapy (ORT) corners within health facilities. ORT corners aim to provide on-site rehydration for sick children and, at the same time, give caregivers hands-on experience to learn how to correctly prepare and administer ORS to their sick children. It has been a successful strategy in previous programs throughout the world [11,12].

Lastly, to build provider capacity, the MOPHS expanded trainings on Integrated Management of Childhood Illnesses (IMCI) through platforms such as didactic trainings, continuous medical education (CME) programs, mentorship and supportive supervisions, and on-the-job trainings. After updating the diarrhea management policy to include zinc, the MOPHS conducted refresher trainings across the country. The program streamlined IMCI trainings to one week of clinical and didactic trainings, with a cascading peer-to-peer model where trained clinicians were responsible for training all other clinicians who treat pediatric patients at their facility. IMCI certification was dependent on an assessment of the facility’s compliance, not just the trained clinician’s compliance, with the IMCI protocol six weeks after the training. The model also included routine supportive supervision by county and sub-county leadership to confirm ongoing compliance and help address any gaps, particularly challenges related to staff turnover. The program also created an IMCI mobile app to make the guidelines readily available to anyone using a smartphone.

Evaluation design and setting

The 2016 study was designed as a cross-sectional household survey to measure coverage of ORS and zinc. We also used Spectrum v5.753 Lives Saved Tool (LiST) (Avenir Health, Glastonbury, CT USA), with ORS and zinc coverage results from existing national household surveys and this study’s cross-sectional household survey to model the estimated number of diarrheal deaths averted due to changes in coverage between 2011 and 2016. LiST is a deterministic mathematical modelling software that uses intervention coverage inputs and literature on the effect of those interventions on cause-specific mortality reduction in order to attribute mortality reductions due to changes in intervention coverage. LiST has been used to help guide policymakers in setting priorities for investments in interventions and to evaluate the impact of programs aiming to scale up proven interventions. The theoretical approach, modeling structure, and specific methods for LiST are published elsewhere [13,14].

Survey study population

The study population was children age 0-59 months who had diarrhea in the two weeks preceding the survey.

Survey sampling design

The survey sample was designed to be nationally representative. The sample was drawn from the Fifth National Sample Survey and Evaluation Programme (NASSEP V). This is a frame that the Kenya National Bureau of Statistics (KNBS) uses to conduct household-based surveys throughout Kenya. The NASSEP V contains 96 251 clusters and is stratified into 47 counties, and within each county, further stratified into urban and rural strata. KNBS used two-stage cluster randomized sampling to select households for inclusion in the study. In the first stage, clusters were stratified by county and urban/rural areas and KNBS randomly selected 560 clusters (279 urban clusters and 281 rural clusters) with probability proportional to size. In the second stage, 25 households were randomly selected from each cluster using the latest household listing available for the cluster.

Survey data collection procedures

KNBS hired data collection personnel and took them through five days of training. The data collection personnel were selected based on their experience in conducting previous household-based surveys. Seventy enumerators and 24 supervisors were engaged to implement the study. The training covered the sampling approach, survey tools, use of the electronic data collection platform SurveyCTO (Dobility Inc, Boston MA USA), research ethics, and practice in the field. The data collection commenced on November 2, 2016 and was completed on December 7, 2016.

The survey data collection tool was adapted from the KDHS, specifically the household members listing section and the child illness section. The enumerators listed all the de facto members in the household including their ages. They then asked whether any of the children under five had diarrhea in the two weeks preceding the survey. If the child did have diarrhea in the last two weeks, the enumerator asked questions regarding the care and treatment for the child, including where care was sought and what treatments were given. The tool was piloted and revised based on feedback from the field data collection personnel. A team of three supervisors and nine enumerators were engaged in the pilot. In addition, to aid in memory recall, picture cards of local ORS, zinc, antibiotics, anti-diarrheal medicines, and traditional medicines were developed, tested, and deployed during fieldwork.

Upon reaching the selected clusters, data collection teams met with the area village elders to explain the purpose of the study and seek their support in locating the selected households. Once the selected households were identified, data collectors requested to speak with the main caretaker of children in the household, explained the purpose of the study, requested written informed consent from the respondent, and if consent was given, conducted the interview.

Data analyses

Descriptive analyses were conducted on three primary outcomes:

The outcomes were disaggregated by socioeconomic status (SES) and location of the household (urban or rural), and we conducted two-sample t tests to examine whether the outcomes differed by SES or location. SES was represented by wealth quintiles. To construct the wealth quintiles, principle component analysis using household assets were conducted [15]. The urban and rural definitions provided by the NASSEP V sampling frame were used to categorize households.

Sampling design weights and non-response weights were calculated based on the NASSEP V sampling frame and incorporated into all analyses. The analysis was conducted using Stata 14 (StataCorp, College Station TX, USA).

LiST statistical analyses

The study used KDHS 2003, KDHS 2008-09, KDHS 2014, and the 2016 survey described above as direct inputs into the LiST model for ORS and zinc coverage in 2003, 2008, 2014, and 2016, respectively. For 2011, the year the program initiatives began, ORS coverage was estimated by linearly projecting forward the average annual coverage rate increases between 2003 and 2008. Zinc was not included in the 2003 KDHS and was not part of the national guidelines until 2007, therefore, we assumed zinc coverage was 0% in 2007. We used the rate of increase between 2007 and 2008 to estimate zinc coverage for 2011. We estimated ORS and zinc coverage for all years without direct survey data (eg, 2012, 2013, and 2015) by linearly projecting forward the average annual coverage rate increases between survey periods.

We also estimated coverages of other interventions included in the LiST software (eg, prevalence of exclusive breastfeeding, vaccine coverage, etc.) using the KDHS 2008-09 and KDHS 2014 results. As above, we used linear interpolation to estimate coverage of these additional interventions for all years in between 2008 and 2014. We did not gather coverage of other interventions in the 2016 study’s survey. To obtain 2016 coverage estimates, we projected forward the annual linear change in coverage from 2008 to 2014 to estimate the coverage for 2015 and 2016. The aim of projecting forward coverage estimate trends to 2016 was to produce conservative estimates of the diarrhea deaths averted attributable to ORS and zinc coverage changes. The LiST model first attributes deaths averted to preventative and protective interventions (eg, exclusive breastfeeding, vaccines, etc.) and the remaining deaths averted are then attributed to treatment interventions (eg, ORS and zinc).

Using LiST, we calculated the number of additional diarrheal deaths averted attributable to each child health intervention, compared to a counterfactual in which coverage of all interventions were kept constant at 2011 values. We also calculated lower and upper sensitivity bounds, which are the number of diarrheal deaths averted using the lowest and estimates of intervention effectiveness used by LiST.

Ethical approval

Ethical approval for the study was received from Kenya Medical Research Institute (Protocol #: KEMRI/RES/7/3/1).

Survey results

Figure 3 presents the results of the sampling and listing exercise. Of the 14 000 households sampled from the NASSEP V sampling frame, we were able to find the household structures for 12 352 households. Of that, 2022 respondents were unavailable at the time of the visit, 62 respondents declined to participate, and 6366 did not have a child under five living in the household. We completed interviews with 3902 households and found 5233 children living in those households.

Figure 3.  Sampling results.

Of the children under five, we found 14% (95% confidence intervals (CI) = 13%, 15%) had diarrhea in the last two weeks. Table 2 presents outcomes and selected characteristics of children who had diarrhea in the last two weeks. Among children with diarrhea, we found 42% (95% CI = 38%, 47%) received ORS, 18% (95% CI = 15%, 23%) received zinc, and 15% (95% CI = 12%, 19%) received both ORS and zinc. Most caregivers of the children sought advice or care for the diarrhea case outside the home – 63% (95% CI = 58%, 68%). Of those that did seek care, 70% (95% CI = 64%, 76%) sought care from a public health care source and 23% (95% CI = 18%, 29%) sought care from a private health care source. A small proportion – 4% (95% CI = 2%, 7%) – sought care from a non-medical source (eg, friend, family, or traditional healer), and only 3% (95% CI = 1%, 7%) sought care from multiple sources.

Table 3 presents ORS and combined ORS and zinc coverage by wealth quintile and urban/rural area. ORS coverage was 43% (95% CI = 37%, 49%) in rural areas and 41% (95% CI = 35%, 47%) in urban areas. For the poorest and wealthiest quintiles, ORS coverage was 48% (95% CI = 40%, 57%) and 31% (95% CI = 21%, 42%), respectively. Combined ORS and zinc coverage was 15% (95% CI = 11%, 21%) in rural areas and 15% (95% CI = 10%, 21%) in urban areas. Among the poorest and wealthiest quintiles, combined ORS and zinc coverage was 21% (95% CI = 14%, 33%) and 14% (95% CI = 7%, 26%), respectively. We did not find a statistically significant difference in the odds of using ORS (P = 0.61) or combined ORS and zinc (P = 0.93) by urban or rural locations. We did find the odds of using ORS was 0.47 (95% CI = 0.25, 0.87) times lower among the wealthiest quintile compared to the poorest quintile (P = 0.02). We did not find any statistically significant difference in the odds of using combined ORS and zinc by wealth quintile (P = 0.28).

LiST model results

Table S1 in Online Supplementary Document presents the estimated coverages for child health interventions used in the LiST model. We estimated that in 2011, 45% of children with diarrhea received ORS and 0.8% received zinc. Using these estimates, the 2014 KDHS, and our own survey as inputs for the LiST model, we estimated that, compared to the counterfactual, an additional 3340 (sensitivity bounds: 2670, 3920) diarrheal deaths were averted between 2011 and 2016 due to changes in ORS and zinc coverage. The number of diarrheal deaths averted was greatest in 2014 when ORS coverage reached 54% and zinc coverage was 8%, and prior to the full implementation of devolution (see Discussion). Compared to if coverage remained at 2011 values, we estimated an additional 1330 (sensitivity bounds: 1090, 1520) diarrheal deaths were averted in 2014.

Rotavirus vaccination, which was introduced in Kenya in 2014, was estimated to avert 1190 (sensitivity bounds: 550, 1590) additional diarrheal deaths. The estimated coverage of rotavirus vaccine was 0% in 2011 and 74% in 2016. Age-appropriate breastfeeding was estimated to avert an additional 730 (sensitivity bounds: 300, 1010) diarrheal deaths with coverage of exclusive breastfeeding for children less than 1 month having increased from 68% in 2011 to 76% by 2016.