Study design
This open label randomized controlled trial used a parallel group design to individually randomize HIV positive women within 24 hours of delivery to receive either phone-based reminders on PMTCT messages or standard health care messages (no calls), with an allocation ratio of 1:1. The study was conducted from 19th September 2013 to 31st January 2014. Participants were recruited at three health facilities in Kisumu, Western Kenya. These study sites were Jaramogi Oginga Odinga Teaching and Referral Hospital (JOOTRH), Kisumu East District Hospital (KEDH) and Lumumba Health Centre.
Study eligibility
HIV infected women 18 years and older, who had a live birth, were eligible for enrolment if they owned a mobile phone on which they could receive calls, were willing to stay in the study area for at least 3 months after delivery and knew that they were HIV infected.
Enrollment and randomization
On delivery, eligible women were invited to participate in the study after being provided detailed study information by trial staff. Women who agreed to participate signed a written consent and were interviewed using a standard tool to collect demographic, socioeconomic and biomedical data. Enrolled mothers were randomized into either intervention or control groups using computer generated block randomization sequence generated using STATA 9.0 software.
Follow up procedures in the intervention arm
Every two weeks on a Monday morning, the researcher called each subject in the intervention arm every until the infant was 10 weeks old. Each call was aimed at reminding them and reinforcing key PMTCT messages (Nevirapine prophylaxis, exclusive breastfeeding of the HIV-exposed infants, early infant diagnosis, scheduled immunizations) as well as ascertainment of their overall health. Calling was inexpensive (at 4 Kenya shillings per minute) lasting approximately 2-5 minutes per call. Study participants were also allowed to call to ask questions and report concerns on infant health. All mobile phone communications between clinicians and study participants were recorded in a study log.
Follow up procedures in the Control group
Participants randomized to the control arm received their usual standard of care (SOC) clinic support but were not called by the researcher. They were however free to call the researcher at any time of their own initiative.
Follow-up activities in both arms
Study follow-up visits were designed to coincide with the scheduled well-baby follow-up clinics at six and ten weeks of life. Data collection was conducted during the scheduled 6- and 10-week clinic appointments, using a standard tool. During these visits, information was collected on infant adherence to Nevirapine, retention in care and early infant diagnosis; and at 10 weeks, additional data was collected on breastfeeding practice.
This study adheres to CONSORT guidelines.
Data management
Data from the questionnaires were coded and entered into Microsoft Access 2007 database. Data entry and cleaning were conducted concurrently with data collection. SPSS version 17.0 was used to analyze data.
Statistical analysis
The sample size determination showed that 75 participants per group allowed the study to detect a 50% relative difference in the primary outcome of retention in care at 6 weeks assuming a retention rate of 47% in the control arm, 95% level of confidence and 80% power.
Intervention and control arms of the study were compared using baseline characteristics. Similarities between the 2 groups was shown by comparing the baseline characteristics using Chi square/ Fishers’ exact tests and Student’s t / Mann Whitney U test for categorical and continuous variables respectively. Proportion of children that took their NVP as prescribed was used as the estimation of adherence. The prevailing guideline at the time of the study recommended infant NVP prophylaxis up to six weeks of age. The proportion of children seen at the well-baby clinic at six and ten weeks was the measure for retention in care. The proportion of HIV-exposed infants who had their HIV infection status determined by six and ten weeks was the estimate of EID. Infants in the two arms of the study were compared for adherence to NVP at six weeks, EID and retention in care at six and ten weeks, and prevalence exclusive breastfeeding at ten weeks. Logistic regression was used to estimate intervention effect using odds ratio (95% confidence intervals). All analyses were intention to treat and all statistical tests were based on a p value cut off of 0.05.