Health and survival of young children in southern Tanzania

Nachingwea, Lindi Rural, Ruangwa, Tandahimba and Newala Districts are in Southern Tanzania and had a total population of about 900,000 people in 2002 (Figure 1) [12]. These districts are sub-divided into administrative areas called divisions, with 3 to 10 divisions in each district and a total of 24 divisions in the study area. Parts of Tandahimba and Newala are on the Makonde Plateau, up to 900 m above sea level. Lindi Rural, Ruangwa and Nachingwea have hilly areas as well as low lying plains. The major permanent rivers in the region are the Lukuledi, Matandu, Mbwemkulu and Mavuji. There are two main rainy seasons, November to December and February to May, but rain is not uncommon in any month. Malaria is endemic and transmission occurs all year round. The area had a wide mix of ethnic groups, including the Makonde, Mwera, Yao and many others. Although most people speak the language of their own ethnic group, Swahili is also widely spoken. The most common occupations are subsistence farming, fishing and small scale trading. Cashew nuts, sesame and groundnuts are the major cash crops while food crops are cassava, maize, sorghum and rice. Most people live in mud-walled and thatched-roof houses: a few houses have corrugated iron roofs. Common water supplies are hand-dug wells which rely on seasonal rain, communal boreholes, natural springs and river water. Most rural roads are unpaved: some are not passable during rainy seasons while others are too steep for vehicles to pass. The public health system comprises a network of dispensaries, health centres and hospitals offering a varying quality of care. Not all health facilities have a qualified prescriber (Medical Officer, Assistant Medical Officer, Clinical Officer or Assistant Clinical Officer) and sick children are not infrequently managed by nursing cadres (Nursing Officers, Nurse Midwives, Public Health Nurse 'B' or Maternal and Child Health Aides) even though these staff are, strictly speaking, not supposed to prescribe. Staff in these nursing cadres are generally responsible for preventive services such as antenatal care and well-child visits for weighing and vaccination. With the exception of Lindi Rural, each district has a District hospital. Some villages have village health workers. Children under five years of age are exempted from paying fees at any government health facility. In the five years prior to the 2002 National Census, the infant mortality rate estimated by indirect methods (probability of dying by the first birthday) was 129/1000 and 126/1000 in Lindi and Mtwara Regions respectively.

We did a survey in all 134 health facilities in the area in October 2004, including hospitals, health centres and dispensaries of the public health care delivery system, non-governmental not-for-profit organizations and the private sector. The overall focus was the availability and accessibility of maternal and child health care. Our aim was to undertake a structural and functional assessment by documenting the availability of vaccines, drugs, supplies and services essential for child health. We interviewed health workers, checked availability and functioning of equipment, and checked drug and vaccine stocks. The survey was conducted using an adapted World Health Organisation (WHO) health facility survey tool [18], without case-management observation or exit interviews. The tool included modules on health services, equipment and supplies. The survey was done by 16 interviewers working in groups of two, forming eight teams, with two supervisors who assisted the survey co-ordinator. Training was carried out over a period of five days and included interview technique, group work, role-plays and practical fieldwork including a pilot-test of the survey instruments. A letter of introduction from each Council Health Management Team, signed by the District Medical Officer and the District Executive Director, was given out at each facility before the interviews. To help assure the quality of data, at least one interview was accompanied by a supervisor each day. All forms completed each day were reviewed in the evening and feedback given to the interview teams.

We did a cross-sectional survey in a representative cluster sample of households from the five districts in July–August, 2004. Thirty clusters of 30 households were sampled from each division, giving a total of 720 clusters and approximately 21,600 households. Within any division, every household had an equal chance of being selected. The number of clusters selected from each ward (an administrative sub-unit of a division) was determined by probability proportional to the size of the ward, on the basis of the 2002 National Census population figures. The required number of clusters was then selected at random from a list of all sub-villages (vitongoji, singular kitongoji), as supplied by the district council. Within each sub-village, 30 households were selected at random using a modified Expanded Programme of Immunisation (EPI) – type sampling scheme that ensured an equal chance of any household being selected. Since there was no list of all households in the kitongoji, the supervisor went to the centre of the kitongoji and threw a pen to choose a random direction. He or she walked in the direction indicated until he reached the edge of the kitongoji, sketching a map of all the households passed, and numbering them as he went. For this purpose a household is a group of people who eat from the same cooking pot. One of these houses was selected at random as the starting point, or house 1 of the cluster. Throwing a pen to choose a random direction, the supervisor walked in that direction until he came to another household, which was the second house of the cluster, and so on. If there was a junction in the path, a pen was thrown again to select from the choices available. This procedure was repeated until 30 households were counted. All households were selected for an interview regardless of whether there were any women or children. Villages were visited one day before the survey interviewers arrived, and an invitation letter left in each of the selected households. Both village and kitongoji leaders were briefed about the visit. Interviewers went from house to house, and if there was nobody at home an interviewer returned later in the day. No substitute household was included if the owner was repeatedly absent or did not wish to participate. The supervisor visited all households where people were reportedly absent or did not wish to take part, as a quality control measure.

We administered a modular questionnaire using handheld computers (Personal Digital Assistant or PDA)[19]. A household module included information on who lived in the household, dates of birth, education, and occupation, and ethnic group of the household head. We collected information on proxy markers of household socioeconomic status including ownership of a radio, a bicycle, animals, and poultry, whether the house had a corrugated iron roof, whether the house was owner-occupied or rented, and whether it was connected to the mains electricity supply. We also recorded the location of the household using a handheld Global Positioning System (GPS). In a birth history module, similar to that used in the Demographic and Health Surveys, we obtained information on survival of all children born in the five years prior to the survey with a view to getting complete and accurate information for the three years prior to the survey, July 2001 – June 2004. Mothers were asked about dates of birth, the sex of the baby, whether they were twins or singletons, and whether the child was still alive. For those who had died, we asked for the date of death. We also asked about antenatal care as well as the use of sulphadoxine-pyrimethamine (SP) for Intermittent Preventive Treatment for malaria in the most recent completed pregnancy (IPTp), and about mosquito net use for women who reported being pregnant at the time of the interview. For a sub-group of 8 in every 30 clusters (27% of households), stratified on division, we also included modules about the health of all children under two years. Mothers or carers (here we use the term 'mother' to denote the main carer) of all children under two years were interviewed about whether or not the child was currently breastfed and if so what other food or drink the child had received over the previous 24 hours. Information on routine vaccinations was documented directly from health cards or other written records. Where no such record was available we asked about the number of doses and type of vaccines received, but we did not ask for dates of each dose. It should be noted that estimates of vaccine coverage reported here are based on written information alone and are likely to be conservative. Mothers were asked whether the child had received vitamin A supplementation, and if so, how many months ago. Mothers were then asked about use of mosquito nets and their treatment, the name and location of their nearest health facility, and about any illness each child had during the 2 weeks prior to the survey, and what action had been taken. For children who had been sick, we asked about utilisation of appropriate (non-traditional) health care providers including village health workers, dispensaries, health centres, hospitals, or private doctors. Children under 2 years were invited to attend a measuring station set up in the middle of the village where they were weighed using hanging scales (Salter, UK). A finger-prick sample of blood was collected and haemoglobin (Hb) concentration measured using a battery-powered Hemocue photometer (HemoCue AB, Angelholm, Sweden). We tested for Plasmodium falciparum malaria parasitaemia using the Paracheck rapid diagnostic test device (Orchid Biomedical Systems, Goa, India). Children found to be anaemic (Hb < 11 g/dl) or parasitaemic were offered treatment according to Tanzanian national guidelines (14 day course of ferrous sulphate, SP, or both). Children with Hb < 5 g/dL were encouraged to attend a hospital urgently and their transportation facilitated.

The survey was carried out by 91 interviewers in 13 teams of 7, each with a supervisor, a driver and a car. Interviewers were trained for two weeks. All teams worked in all divisions. The training included interview technique and probing for dates using local event calendars. A detailed field manual was prepared, piloted during the training sessions and carried in the field by all interviewers. The trainees included participants from all study districts. At the end of the training period a pilot survey was carried out in part of Mtwara Rural district. During the survey, at least one interview was accompanied by a supervisor each day and he/she also repeated key aspects of further interviews independently at random. In these repeat interviews, the head of household and one woman aged 15–49 were asked a subset of questions. Any discrepancies between the initial and repeat interviews were discussed with the interviewer and appropriate action taken.

For the health facility survey, data were collected using conventional paper forms. Double data entry was done using DMSys software for clinical trials (SigmaSoft International), with range and consistency checks on entry and discrepancies between the two versions resolved with reference to the original forms. Data were summarised using tables. For the household survey, all data were entered into PDAs at the point of data collection. Standard range, consistency and completeness checks were carried out at that time. Particular attention was paid to the accuracy of dates of vital events because of the rapid change in the risk of death within the first year of life and the fact that age is typically prone to errors of recall. Analysis of household data was done in Stata (version 8, College Station, Texas, USA) in accordance with the design of the survey, using "svy" commands such as svytab to summarise and adjust for the clustered nature of the data. We created an index of household wealth based on ownership of assets. The household wealth index was the weighted sum of household characteristics including whether they owned the house they occupied, whether they owned consumer durables such as a bicycle, radio, or tin roof, whether they owned animals or poultry, whether they used wood for cooking, and whether they had mains electricity. The weights for the assets in the index were generated by Principal Components Analysis (PCA) on the correlation matrix [20]. Households were categorized into one of five equal sized groups from the most poor to the least poor. P-values for testing hypotheses concerning inequalities by sex, ethnic group and distance from the health facility were from design-based F-tests, equivalent to ordinary chi-squared tests adjusted for clustering. P-values for testing hypotheses concerning inequalities by socio-economic status quintile were based on a test for trend with socio-economic status from a logistic regression model adjusted for clustering. Mortality rates were derived from the birth histories from all births between July 2001 and June 2004, with time at risk calculated from the date of birth until the date of the first birthday or the end of June 2004, whichever was the earliest, or the date of death for those who died before their first birthday. Mortality rates were modelled using Poisson regression with robust standard errors to adjust for clustering. Weight for age Z-scores were calculated with reference to the US National Centers for Health Statistics standards using the EPINUT module of EPI-Info (CDC Atlanta, Georgia, US). Underweight was defined as weight-for-age z-score of less than -2, excluding outliers (z-score of < -5 or > 5). GPS data were cleaned by removing outliers in each cluster (households more than 10 km from the median location of households in that cluster) and straight-line distance to the nearest health facility was calculated, then categorised as over or under 5 km, which is roughly one hour's journey on foot.

The study was undertaken within the framework of the assessment of the community effectiveness of IPTi, part of the IPTi Consortium [17]. We received ethical approval from the local and national institutional review boards (Ifakara Health Research and Development Centre, Ifakara, and the National Tanzania Medical Research Co-coordinating Committee) through the Tanzania Commission for Science and Technology. In addition ethical and research clearance was also obtained from institutional review board of the London School of Hygiene and Tropical Medicine, UK, and Ethics Commission of the Cantons of Basel-Stadt and Basel-Land, Switzerland. During field work, information sheets in Swahili about the study were given out, explaining why it was being done, by whom, and what it would involve. In the household survey, written consent of all household heads was sought. Confidentiality of all study participants was assured.