During the period of 2006-2007, 65 neighborhoods of about 800 households each were identified in three Xhosa townships surrounding Cape Town, South Africa. The neighborhoods contained both formal settlements (government housing with an address and onsite water and sewage), site-and-service plots (plots of land where residents can build a home, with some access to water and sanitation facilities), and informal settlements (shacks or temporary structures that rarely have water or access to sanitation on the premises and are not on a specified plot of land). A mentor mother (MM) was recruited for each neighborhood using a variety of sources such as referrals by local community leaders, or by open application. Nominees were interviewed by supervisors for the Philani program. MMs chosen for the program had children who were thriving; they demonstrated strong communication and interpersonal skills, commitment to community service and showed an organized and disciplined approach to tasks. Thriving children and organizational skills were confirmed by home observations of the MM's household, where Philani supervisors ascertained whether the home was organized, children were monitored, and healthy food was available. MMs who were selected received training in child health, nutrition and other related topics. After training, one MM was assigned to each neighborhood, typically based on residential proximity. MMs received a stipend of $US 130/month from Philani to deliver home-based interventions and were expected to work four hours a day.
The flow of participants through the study is shown in Figure . MMs visited every home in their neighborhoods, identifying and weighing each child aged 5 or younger. The child's age and weight were plotted on a growth chart containing age-appropriate norms. Any child weighing less than 2 standard deviations below his or her weight-for-age norm (<-2SD) was classified as malnourished; this included all newborns weighing less than 2500 grams at birth. Any household with at least one malnourished or low birth weight child was invited to participate in the study. If there were multiple malnourished children in a household, all were followed but only one was randomly selected to be included in the analysis. Over 12 months, the 65 MMs recruited 788 mothers and their babies or children aged 5 or under. The mother-child unit is referred to as a dyad
. Assignment to the intervention condition was based on a sequence decided a priori
for every three dyads. Because the likelihood of benefits from the intervention far exceeded any chance of harm, the allocation ratio of intervention to control was 2:1 [23
]. Two out of three dyads (in random order) were assigned to the Philani intervention condition (n = 536). The third dyad became a control case (n = 252). The MMs were given randomly sequenced numbered folders marked I for intervention and C for controls and supervisors made sure the folders were allocated in the correct order. After data collection was completed, dyads assigned to the control condition were given the option of joining the Philani nutrition intervention program.
Flow of participants through the study.
This study was approved by the UCLA Office of the Human Research Protection Program (#G07-02-033) and the Stellenbosch University Health Research Ethics Committee (#N08/08/218).
Mentor Mothers received four phases of training: 1) watching experienced MMs implement the intervention in an inspiring manner, learning how to approach a family and build trust; 2) attending a month of training that covered nutrition; basic child health including HIV and TB, weighing of babies and completion of growth charts; how to recognize signs of abuse and crisis situations; and how to encourage depressed mothers to be more active and engaged with their children; 3) learning how to help mothers bond with their children and improve the consistency of healthy daily routines; and 4) implementing their first round of home visits independently in their neighborhoods.
An essential part of the intervention is for the MM to create a respectful and caring relationship with the mother/parent. Throughout the Philani experience, we have found that changing behavior is not possible without such trusting relationships. The MM who is a positive deviant has developed coping mechanisms which have made it possible for her to raise healthy children and a key component of the intervention is for her to share those coping mechanisms with other mothers. These might include initiating and maintaining breastfeeding; introducing solids correctly; feeding frequently but also creating good sleeping habits; providing organization, discipline and structure in the home; protecting the child from sources of infection, accidents and trauma; and seeking care when needed. A successful intervention will see the child gaining weight rapidly until fully rehabilitated, with a significant decrease in episodes of infection. With a healthy growing child the incidence of maternal depression will decrease and the bonding between mother and child will improve.
A supervisor accompanied each MM at least one day a month on a random schedule to ensure that implementation proceeded as planned. The supervisor collaborated with the MM in problem-solving and generating action plans when problems occurred in the field. Each MM and supervisor built a list of clinics and hospitals for referrals and a strategy for providing services relating to mental health issues (particularly post-partum depression), partner abuse, and legal problems. The quality of the implementation was monitored by reviewing the forms completed at each home visit, monitoring visitation patterns, collecting observations by supervisors, and brief ratings of home visits by the supervisors.
A typical MM visit at each home lasted 20 to 60 minutes. During the visits, the MM weighed the participating child and discussed his progress with the mother. The MM also made sure the mother had the social grants she might be entitled to, and that the mother understood proper nutrition and hygiene. MMs stressed the importance of breastfeeding, the proper time to introduce solids, frequent feeding, and a mixed diet including vegetables and fruit. She checked to see if immunizations were up to date and that the child was dewormed. In each MM's caseload there was likely to be one emergency a week (e.g., a child would be ill with high fever, have difficulty breathing, or appear severely dehydrated). These cases were brought to the Philani Health Clinic or the local public health clinic to receive immediate attention. MMs did not distribute food supplements.
The following data was collected by MMs at the time of recruitment, from dyads in both the intervention and control arms of the study:
○ Children's background characteristics. Mothers reported on several characteristics of the child enrolled in the study: age, gender, birth weight, and whether or not the child was already enrolled in a nutrition program.
○ Mothers' background characteristics. Mothers also reported whether any of their children had died, whether they were employed, whether they were receiving any government grants, where they were born, and number of years living in the Cape Town area.
○ Housing/living situation. Mothers reported housing conditions (classified as formal, site and service, or informal), number of adults living in the household, whether there was water onsite at the home, whether there was a flush toilet, and whether they had had to reduce or skip meals due to lack of money. MMs reported two subjective measures for the mothers' living conditions: overall smell (pleasant, neutral, or poor) and hygiene (good, average, or poor).
Children in the control condition were weighed by MMs at baseline, and at 3, 6, 9, and 12-month follow-ups. Children in the intervention condition were also weighed by MMs at baseline, and at each intervention home visit. Rehabilitation
to an acceptable weight was indicated by achieving a weight-for-age Z-score (WAZ) [25
] that was above the cutoff for study eligibility (>-2SD), i.e., above the third percentile of weight-for-age norms. Time to rehabilitation was noted at the first assessment at which the child reached the target weight.
We compared demographic and household characteristics of the dyads across intervention conditions at recruitment. We also compared dyads followed over time vs. those with no follow-ups. Chi-square tests and t tests were conducted for categorical and continuous measures, respectively. Where appropriate, Fisher's exact test was conducted on categorical measures with sparse cell counts and the Wilcoxon two-sample test was conducted on continuous measures with skewed distributions.
We compared the time to rehabilitation between the intervention and control conditions using discrete time survival analysis models [26
]. Discrete-time versus continuous-time models were used to capture the discrete nature of the follow-up intervals. That is, when normal weight was achieved, we knew that it had occurred between the current and previous assessment, but we did not know the exact time. Therefore, a child rehabilitated any time between the baseline and 3-month assessments was considered to be rehabilitated at 3 months and a child rehabilitated after 3 months but before the 6-month assessment was considered to be rehabilitated at 6 months. Rehabilitation times were coded similarly at the 9 and 12 month assessments. Using this technique, the probability of rehabilitation at a given assessment is conditioned on the child surviving and not being rehabilitated prior to that period. A logistic framework is used, modeling the log odds of rehabilitation as a function of time period and other covariates in the model.
There were three possible outcomes for children during the study period: 1) rehabilitation (i.e., achieving an acceptable weight); 2) death, at which point they no longer contributed to person-period time (i.e., observations beyond death were censored); or 3) the child did not reach normal weight by the final follow-up and was censored at 12 months. Because of the possibility of intermittency in the follow-up weighings (i.e., missing an earlier weight but not missing later ones), there was no censoring for loss to follow-up. For time periods where there was missing data (i.e., a weight was not recorded), it was assumed that rehabilitation had not occurred. To account for dyads for whom no follow-ups were obtained, models were run two ways. In the first, it was assumed that rehabilitation did not occur, and the child was censored at 12 months. In the second, these dyads were removed from the analysis. Results were nearly identical under both approaches, and the results from the analyses that assumed no rehabilitation are presented here. Discrete-time survival models were implemented in SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA) [26
Models included covariates for randomized intervention condition assignment, classified as intervention or control; discrete time interval, classified as 3, 6, 9, or 12-months; intervention-by-time interval interactions to test the intervention effect; and background characteristics, including child age, gender, and living conditions.