Study area, villages involved in the trial and participants
The study was conducted in Bago, a division in the south of Myanmar with 12,941 sq km of forested land and a population of 5,313,613 inhabitants. With an annual rainfall of 3,291 mm and temperatures ranging from 20.9°C and 32.3°C, malaria is endemic all year round with a peak in transmission in June and July. Six townships with moderate endemicity (1–10 per 1,000 population) were selected for this study (Bago, Daik-U, KyaukTaGar, Oktwin, Taungoo and YeDarShay). In 2008, 21 malaria deaths were reported by the six township hospitals but the number of deaths in patients who did not arrive at the hospitals is unknown. No volunteers had previously been trained in Bago Division.
Farming and forest work are the most common occupations of villagers in the study area although seasonal work in plantations (teak, sugar cane and rubber) or construction sites is frequent (water dams, road construction sites and mines).
Sketch maps were available in some health centres but villages were not indicated on the maps, and Global Position System (GPS) points were not available. In successive consultations with approximately 200 midwives, information for the study was collected including village maps, distance between villages and information about villages under their care. Villages with fewer than 70 households and those where midwives were stationed were excluded from the trial, as were villages presenting serious security problems. Only those remaining villages classified as “Stratum 1a” were included in the study (highest level of endemicity as per malaria control programme classification to prioritize malaria control activities). Thirty-eight villages at least two hours distance away from the nearest selected villages were then identified.
The study population included villagers currently living in the endemic area or its surroundings for at least one month.
The study design
A cluster randomized controlled trial, with each village as the cluster, was chosen because the volunteer was accessible to all villagers in any given village. Eligible villages (clusters) without existing health staff were then randomized into intervention and comparison groups. The nature of the intervention allowed neither allocation concealment nor blinding. Villages with midwives (MW) in post were used as a second comparison group in the endline survey (see Figure 1). Village characteristics and mortality data were collected and house-to-house survey carried out in the two groups without health staff at baseline and in all three groups at endline. Information from the endline survey was supplemented by an analysis of volunteer logbooks, patient records completed by volunteers, midwives’ logbooks and field-based verbal autopsies.
The intervention
The intervention consisted of training unpaid village volunteers in the provision of RDT and ACT and supervising them. The VBDC programme supplied volunteers with Paracheck Pf® (Orchid Biomedical Systems, Goa, India), a RDT for the diagnosis of P. falciparum malaria with up to 100% sensitivity and a specificity of 75% to 98.3% [3, 31, 32]. Patients who tested positive were treated with a combination of 20 mg artemether-120 mg lumefantrine (Coartem®). Patients with fever (with or without chills) whose test results were negative were administered chloroquine to suppress P. vivax malaria, after excluding local infection. The programme did not provide paracetamol, an antipyretic commonly used and available locally to treat fever.
Training module and content preparation
A volunteer training module for malaria in Myanmar (see Additional file 1) and the Job Aid illustration on the use of RDT, recommended by the WHO were used [18, 33]. The training topics covered communication skills, the nature, symptoms and modes of transmission of malaria, fevers, vulnerable populations, illustrations of RDT, treatment regimens, patient referral, prevention and control measures, thermometer reading, record-keeping and reporting. Referral to basic health staff was recommended for infants and pregnant women, and in the presence of symptoms such as sore throat, difficult urination, ear discharge, cough, loose motion and skin ulcers. The training did not cover the treatment of non-malarial diseases.
Educational flip charts were available in local languages and distributed to volunteers to educate patients. Malaria supervisors also demonstrated the impregnation of mosquito nets. Practice on the use of flip charts and impregnation of mosquito nets was not included in the training.
Selection of community volunteers
For health volunteers to meet eligibility criteria, they had to reside in the village, be of either gender, 18 years and above, having completed primary school and easily accessible to villagers, willing to work on a volunteer basis, accepted and respected by villagers, and willing to work under the supervision of health staff. In each village, the leader selected a volunteer who was then approved by the health staff.
Training workshop
The training was held on 19–20 May, 2009 in the Bago Divisional Office. The Bago VBDC team included a malariologist, entomologist, laboratory technician, health education specialist, malaria supervisors; they facilitated the training with practical demonstrations and exercises. Information on volunteers’ background and willingness to serve was collected a day before the training and volunteers were provided with training modules. A self-administered, semi-structured questionnaire was used to evaluate the training.
Volunteers were reimbursed for their travel costs (7.5 to 15 USD) and meals (17 USD) during their stay in Bago. Giveaways included a T- shirt, a cap and a handbag with a malaria volunteer logo.
Assessment of competence of the volunteers in using RDT and ACT
Following the training, a 16-item WHO RDT training evaluation instrument [33] was used as well as three sets of RDT sample results on a computer screen; this was done until each volunteer performed a RDT proficiently, correctly interpreted its result, and appropriately selected drug treatment and dosage.
Materials and the tasks to perform
After completing the training, two boxes of Paracheck Pf® RDT (50 packets), AL and chloroquine were provided to all volunteers. Other materials included disposable gloves, a lancet, cotton, spirit, and a timer-watch, thermometer, calendar, patient record forms and a logbook. Monthly calendar cards were issued to record the availability of volunteers and their meetings with health staff. On returning to their villages, volunteers were to inform village leaders, seek their support in endorsing and promoting the programme, display two posters announcing the availability of free RDT, and report to local health staff for future supervision.
Monitoring and supervision
Midwives in the nearest health facility routinely monitored and supervised volunteers during their immunization visits to villages (monthly). Malaria supervisors occasionally visited volunteers. The first author assessed the regularity of activities through monthly review of calendar cards and reports and regularly visited these townships to collect forms from malaria supervisors but did not supervise their activities.
Assessments
Data on village information and a listing of household members to screen for fever were collected. Subjects who had reported fever were interviewed by research assistants. Close relatives of the deceased were interviewed using the standard verbal autopsy guidelines [34]. Volunteers’ and midwives’ logbooks were also reviewed.
Village profile
Background characteristics, village population and malaria control activities were collected from village leaders.
House-to-house surveys
The baseline survey was conducted in the intervention and comparison villages between 8 March and 6 April, 2009. The endline survey was conducted in all villages visited for the baseline assessment, and in villages with midwives, between 21 November, 2009 and 17 February, 2010. Two teams consistently collected data for both baseline and endline surveys. They consisted of three research assistants from the Department of Medical Research - Lower Myanmar (DMR-LM) for the interviews, and a local health staff for the blood test.
Two pre-tested forms (a household form and an individual form) were used. The household form included village name, township name, a list of household members and their background characteristics such as age, sex, occupation, history of fever during the previous 30 days. The individual form included background characteristics, knowledge of malaria symptoms, number of febrile episodes, date of the episode, place of residence at the time of the episode, treatment type and place received, awareness of the availability and location where free RDT-based treatment was available, history of RDT and the main reason for use or non-use of RDT. The interval between the first symptom (fever) and blood test was assessed for timeliness of RDT use. Data in the endline survey were validated against information from volunteer logbooks.
For each survey, all households were visited and data collected for the household forms. Villagers with high- or middle-school education completed the household forms under the supervision of a research assistant. All household members above the age of six months, living in the village for over 30 days and having contracted a fever within the previous month were eligible for interview. Those with fever for less than 24 hours were excluded, as they could not be assessed for timeliness of RDT. Appointments were made with eligible subjects for a face-to-face interview. When eligible subjects were under the age of 15, the mother or a close household member, were interviewed.
Ascertainment of mortality for the intervention period
During the baseline survey, village leaders were asked to keep a death register and ascertain the date of death in the endline survey. When reviewed however, the death registers were unavailable or incomplete. Dates of deaths were identified in several ways during the endline survey. First, the list of the deceased after 21 May, 2009 was collected from midwives, village leaders and the informal social group registers. Second, the close relatives of the deceased were requested to recall events on the day of the death of a kin. Third, the dates of death were confirmed with the death notices distributed locally. When a death notice was not available, recall was facilitated by the use of the Myanmar calendar, which includes a festival day every month.
Verbal autopsy survey
After identifying the deceased during the intervention period, the international standard verbal autopsy guidelines were used to identify causes of death in the post-intervention assessment, using questionnaires appropriate for each age groups [34]. Two doctors and one non-medical assistant interviewed the close relatives of the deceased. The hospital record forms and investigation results of the deceased were also checked with relatives, when available. An interviewer transcribed the information into the verbal autopsy questionnaire. Being responsible for the hospital and public health performance, the township medical officer is responsible for reporting malaria-related deaths. In the verbal autopsies, local health staff and unlicensed practitioners were asked about the symptoms of the deceased such as fever, blood pressure, oedema and tightness of chest.
Two physicians, blinded to the intervention status, identified the causes of death from the verbal autopsy questionnaires. When a discrepancy between their conclusions arose, the final diagnosis was made by a senior doctor from the study area. All were independent from the control programme. Using the standard WHO case-definitions for malaria [35, 36], diagnoses were categorized into deaths from probable malaria, non-malaria and fever of unknown origin. Malaria was recorded as a cause of death if complications similar to those observed for malaria had occurred before death. Non-malaria deaths were recorded as diseases other than malaria. Fever of unknown origin was recorded for deaths from fever with no complications mimicking serious malaria, where blood test results were unknown.
Other reviews
Volunteers’ and midwives’ (18) logbooks and patient record forms were reviewed from 21 May, 2009 until the end of the survey to determine trends in caseload and use of RDT.
Main outcomes of interest
The primary outcome measures were: (i) whether villagers had received blood test-based treatment for fever, (ii) interval between onset of fever and the blood test, (iii) mortality. The secondary outcome variable was knowledge of malaria mode of transmission.
Sample size
The proportion of people anticipated to have received treatment within 24 hours after the intervention period was 50% in the comparison arm and 70% in the intervention arm, based on the 2005 Roll Back Malaria target of at least 60% of malaria cases receiving appropriate treatment within 24 hours of the onset of symptoms and the unavailability of information on proportion receiving malaria diagnosis within 24 hours from onset of fever information in Myanmar [35]. The initial screening of household members for fever during the pre-test in two villages and additional qualitative information from local malaria assistants was used to draw an expected mean cluster size of 15, with cluster size ranging from five to 20 and an intra-cluster correlation 0.1. A total of 17 clusters with 257 individuals each were required for the intervention arm and the comparison arm at 5% level of significance and 80% power [37].
Data analysis
Descriptive statistics including percentages, mean and standard deviations were derived from volunteer background data, logbooks and patient record forms.
The impact of the training was assessed using intention-to-treat analysis method, where two villages in the intervention group were included in the intervention arm although these villages did not receive training of volunteers. “R” software [38] and installed packages were used for the analysis (epicalc [39], survey [40], MASS [41], lme4 [42]). Statistical comparisons were made between two groups (intervention and comparison) and between the two surveys to test the main hypothesis. All three groups (intervention, comparison and MW groups) were included in computations for the modelling.
Survey analysis for house-to-house survey data
The surveys were analysed using sampling weights to compute estimates and standard errors for outcomes. Sampling weights were calculated using the inverse of the probability of selection, or the estimated total village population divided by the number of villagers screened for fever.
The analysis of knowledge in interviewed subjects who had contracted fever used a scale from zero to eight for eight means of transmission, with a correct answer for each mode of malaria transmission scoring one.
To establish significant differences between the groups (intervention vs comparison) and surveys (baseline vs endline), the first and second-order Rao-Scott corrections to the Pearson chi-squared test for proportions and two sample t-test for continuous variables were used.
Multivariable analysis
Due to limitations in the survey package in model fitting for hierarchical data, generalized, linear mixed model with random intercepts was employed to predict the binary outcome variable (diagnostic blood test for malaria) with the individual (subjects with febrile illness) as the first level and the cluster (village) as the second level. Akaike’s information criterion (AIC) value was used in the selection of the best-fit model.
Other analyses
The mortality rates were computed as the ratio of the number of deaths and the mid-survey period population.
Ethical approval
The Ethical Review Committee, Department of Medical Research (Lower Myanmar) - No. 4/Ethics/2007 and the Ethics Committee, Prince of Songkla University, Thailand - EC 52-054-18-6-2 approved this study.
Trial registration
Australian New Zealand Clinical Trials Registry ACTRN12610000706077.