The positive deviance/hearth approach to reducing child malnutrition: systematic review

Identification of studies

As this is the first systematic review of the Positive Deviance/Hearth program, a broad and inclusive approach to study sampling was used. Studies were identified in 20 databases, including CINAHL and MEDLINE, as well as grey literature from 26 sources including databases (e.g. Open System for Information on Gray Literature) and websites of organizations known to run programs (Appendix 1). The search was limited to articles published in English. There was no restriction on date of publication. The reference lists of all articles were reviewed to identify additional reports. The search term used for both peer reviewed studies and grey literature was ‘Positive Deviance’. ‘Hearth’ was not included in the search term as some programs were only referred to as ‘Positive Deviance’ rather than ‘Positive Deviance/Hearth’. The search term ‘Positive Deviance’ alone was able to identify programs titled ‘Positive Deviance’ and ‘Positive Deviance/Hearth’, and both were included in the review. Some very early programs were referred to only as ‘Hearth’ programs, without the term ‘Positive Deviance’, even though they used a positive deviance approach. These studies were included in the review if they were identified through searches of reference lists. It was not practical to search databases for the term ‘Hearth’ alone, as ‘Hearth’ has multiple meanings, which gave very large number of search results with only a few relevant studies.

To be included in the review, the study or report had to evaluate the effectiveness of a Positive Deviance/Hearth Program for child malnutrition that used similar steps to the CORE manual (Nutrition Working Group, Child Survival Collaborations and Resources Group [CORE] 2002a,Nutrition Working Group, Child Survival Collaborations and Resources Group [CORE] 2002b). No limitations were placed on study design as long as the study reported on program outcomes such as nutrition status or behaviour changes. Qualitative, quantitative and mixed-methods studies were included.

Study quality and data extraction

Studies were identified by a single researcher, first by examining the title and then the abstract, or in the case of grey literature, the Executive Summary or the first page where the term ‘Positive Deviance’ was mentioned. Data for all studies were extracted independently by two researchers onto a customized form. Initially, the intention was to use a previously validated quality appraisal checklist or form. However, the wide variation in report quality and study design in the grey literature made detailed checklists and appraisal forms impractical. As a result, a more basic form was developed that extracted high-level information on the study design, the completeness of the report sections on setting, population, intervention, study methodology and results, and whether any type of community participation was described for each step in the CORE process. Disagreements between the two researchers were resolved through discussion. Only a few such disagreements occurred, and almost all were because of one of the researchers overlooking a section of the report. This was particularly the case for grey literature reports, which tended to be long and included large amounts of content on other programs not relevant to this study. At the beginning of the review process, there were a small number of differences in opinion when assessing the level of completeness of the reports. These were resolved by better defining what was required to rate the report section as complete.

Because of variations in study design, quality and reporting, qualitative analysis rather than quantitative meta-analysis were used to draw conclusions on effectiveness. Although data were extracted by two researchers, results and analysis were interpreted only by the author.

Selection of studies

The search identified 267 peer reviewed articles and 611 grey literature documents containing the term ‘Positive Deviance’ (Figure 1). Of these, 10 peer reviewed studies and 14 grey literature reports met the inclusion criteria. Many documents excluded at the first stage of the search applied the Positive Deviance philosophy to other public health issues and programs, such as Female Genital Mutilation, child protection, hospital acquired infections, HIV/AIDS and cancer risk. Of those that were related to nutrition there were a large number of documents that described or mentioned the positive deviance approach in general terms, but did not report on the results of a positive deviance program.

In some cases, a single Positive Deviance/Hearth program had multiple studies and reports related to it. In total, 17 unique Positive Deviance/Hearth programs were identified in 12 countries. Six of the 10 peer reviewed studies were on one program. The decision of what counted as a single program was based on the number of evaluation reports. For example, USAID funded five NGOs to implement the program in Indonesia using a standardized approach. A final evaluation covered the work of all five NGOs, and so, all were treated as part of one USAID program.

One study in Bangladesh used the Positive Deviance approach to identify successful behaviours, but then used group peer education and individual counselling to disseminate the behaviours rather than a Hearth. While this program does not strictly adhere to the CORE manual, it was included in the review as one of only two randomized controlled trials (RCTs) identified.

Study quality

Table 1 summarizes the study designs used for each program. Nine of the programs, all grey literature reports, used a pre- and post-test design without a control. Three used a non-randomized trial with a comparison group and two used an RCT. Three programs included a non-randomized cross-sectional survey of younger siblings and a comparison group. These sibling studies involved identifying families who had previously had their elder children participate in the program. The studies compared the nutritional status of their younger siblings to similarly aged children who had not had an older sibling participate. The purpose of these studies was to determine whether the Positive Deviance/Hearth program was able to prevent malnutrition as the mother applied the behaviours she had learnt to her next children. Nine programs included specific qualitative methods, mainly interviews and/or focus groups with one or more stakeholder groups.

In terms of study quality, only a small number of studies used highly rigorous designs with a control group and only some of those used an appropriate matching strategy for selecting the control group. However, nine programs used a combination of both qualitative and quantitative evaluation techniques, which improves validity and provides a more accurate picture of program effectiveness. In many cases, assessing the quality of the study was difficult because of the large variation in the completeness of reports. Only two reports, one peer reviewed and one grey literature, provided a complete description of the setting, population, intervention, study methodology and results. Seven reports had incomplete information in all of these areas.

Community participation

Only one program described community participation during all stages of the process (Maslowsky et al. 2008). All programs except one involved community members in the delivery of the Hearth sessions. However, community involvement in defining the problem, determining existing behaviours, identifying positive deviants and monitoring the program was usually either not reported or limited to the involvement of a small number of community volunteers or representatives.

Quantitative results

The outcome variables reported by studies varied widely and included weight gain, nutritional status, weight for age Z scores (WAZ), feeding practices, hygiene practices and breastfeeding rates. Even when the same outcome variable was used, such as nutritional status, the reference standards used to determine a child’s status were often not reported. Some grey literature reports failed to report statistical significance. Because of the heterogeneity in outcome variables and report quality, a quantitative meta-analysis could not be performed. Table 2 summarizes the quantitative results from the more robust study designs. Table 3 shows the total number of studies with positive and negative results for each study design.

All studies that used a pre- and post-test design without a control showed that the malnourished children who participated in the program gained weight between day 0 and the end of the Hearth sessions. The amount of weight gained and the proportion of children rehabilitated varied significantly between and within programs. For example, the percentage of children gaining at least 400 g in 1 month in the USAID funded program in Indonesia range from 36 to 54% between NGOs and 21–92% between program sites (McNulty & Pambudi 2008). The results from these studies must be treated with caution in the context of a systematic review. While the study designs were appropriate for their intended use in the field, they have many internal threats to validity, which make them inappropriate for drawing overall conclusions on the effectiveness of the Positive Deviance/Hearth approach.

Of the three non-randomized trials, one reported a positive result: substantially fewer children in the intervention community suffered from underweight, wasting or stunting 6 months after program implementation; one reported a positive result 1 month after program implementation, and one reported no significant difference between the intervention and comparison groups, although multivariate analysis suggested that this may have been due to confounding effects. Sample sizes ranged from 50 to 192 children in the intervention group. Only one study had a clear matching strategy for the comparison group.

Three studies used a non-randomized cross-sectional survey of younger siblings and a comparison group to assess the potential for the program to prevent malnutrition. Of these, two reported that younger siblings in the intervention group had substantially lower rates of malnutrition than the comparison group, although one of these studies also showed that wasting rates were higher in the intervention group. One reported no significant difference. One reported improvements in behavioural practices among mothers in the comparison group. All three studies matched the comparison communities on one or more variable. Sample sizes ranged from 46 to 128 children and four to five communities in the intervention group.

Two studies used an RCT design and both used appropriate tests for statistical significance. One reported no statistically significant difference in terms of nutrition status, although intervention children who were younger (up to 15 months) and more malnourished (less than −2 Z) at baseline, deteriorated less than comparable children in the comparison communities. The study also showed a reduced rate of respiratory illness in the intervention group compared with the comparison group, but no difference in the rate of diarrhoeal disease. The sample size for the study was six intervention and six comparison communities. The other study did not measure nutritional status or weight as an outcome. Both studies showed statistically significant improvements in child feeding practices in the intervention group compared with the control group.

Qualitative results

Studies using qualitative methods unanimously reported that the program was considered success by mothers and program staff. Most also reported a high level of community enthusiasm and engagement. The managers of the program in Gujarat, India, said it was the most popular program being run by Integrated Child Development Services (Jeevan Daan Maternal and Child Survival Program, Ahmedabad n.d.). Some studies reported that participant mothers were sharing the information they had learnt with their neighbours. In the Indonesian program, this appeared to be more common in rural rather than urban areas (McNulty & Pambudi 2008). Both mothers and fathers reported visible improvements in the children who participated in the program: they were more active, looked healthier and were cleaner (Burkhalter & Northrup 1997; Berggren 2004; Maslowsky et al. 2008). This visible change appeared to motivate the mothers and volunteers to continue with the program. One study found that mothers in the intervention group reported a greater increase in empowerment than comparison mothers (Hendrickson et al. 2002).