Public Expenditure Tracking Survey in Health 2007

Zambia PETS in Health 2007 adopted a multistage sampling frame involving provinces, districts, and health facilities, and within health facilities, health workers and patients.

1. During the first stage of sample selection, five provinces were chosen. Two urban provinces - Lusaka Province, and Copperbelt Province were purposively selected. One rural province - Southern Province - was deliberately included in the sample on the advice of the Ministry of Health. (It was the most highly resourced rural province in terms of the number of health facilities, therefore it took a disproportionately larger share of Zambian government health funds). The two other rural provinces - Northern Province and Western Province - were randomly chosen from the list of remaining six rural provinces.

2. Districts were selected during the second stage of the sampling process using purposive and random sampling. Lusaka province had only four districts, one of which was selected for questionnaire pre-testing. Initially, to save on costs and time, a fixed number of four districts were selected for each province. It is important to note that of the three districts selected by default in Lusaka province, Lusaka Urban is the provincial headquarters and is the most urbanized district in the country; Kafue represents a mix of urban and rural areas; and Luangwa is typically a rural district. In the rest of provinces, it was intended that the provincial capital be purposively selected, plus three other districts selected randomly. Another exception was that given their size, Northern and Southern Provinces were granted five districts and Lusaka was granted three instead of four. Overall, 21 districts were selected, accounting for 29 percent of all districts in the country. The following districts were chosen:

• Lusaka Province: Lusaka Urban, Kafue, and Luangwa,
• Copperbelt Province: Ndola, Mpongwe, Mufulira, and Chililabombwe,
• Southern Province: Livingstone, Siavonga, Namwala, Sinazongwe, and Kalomo,
• Western Province: Mongu, Shangombo, Sesheke,
• Northern Province: Kasama, Mpika, Nakonde, Chinsali, and Chilubi Island.

3. Facilities were selected during the third stage using the simple random sampling without replacement technique. The complete list of health facilities was drawn from the inventory made by the Central Board of Health (CBOH) in 2002 and published as "Health Institutions in Zambia: A Listing of Health Facilities According to Levels and Locations". The survey aimed to capture a number of facilities in each district commensurate with the district population, with 50 percent lying within 10 kilometers of the central business district and the other 50 percent outside the 10 kilometers radius. Given the distribution of hospital facilities, it was expected (and later observed) that the sampling frame would include the district hospital or a higher-level hospital, whichever existed in the respective districts. The total number of facilities selected represented 13 percent of all health facilities in Zambia.

• Sampling of hospitals (1st and 2nd level, 18 in total): The distribution of hospitals in Zambia is such that there is typically one hospital in each district. Provincial centers, which tend to host second level (regional) hospitals, do not have level one (district) hospitals. A few districts like Shangombo and Nakonde may not have any hospitals at all. 19 hospitals (across all three levels of care) were selected by default through the random selection of the districts, as discussed above. The final sample of hospitals consisted of 18 facilities.

• Sampling of health centers (132 in total): With the respective district serving as the sampling cluster for health centers, health centers were randomly selected within each district. The sample size of health centers per district within each province was weighted by the total number of public (government and mission) health centers in the district relative to centers in the other districts.

Patient exit interviews will be conducted on a sample of patients visiting the sample facility during the survey. The sampling procedure will involve picking every 4th-7th patient on the queue, depending on the utilization level at each facility. Prior appointment and consent will be sought while the patient is on the queue. Five patients will be chosen per facility as the budget could not accommodate interviewing a larger sample. Thus, a total of 750 patients will be interviewed.

At least two health workers from each health facility will also be interviewed. Where possible, a simple random sampling procedure will be used in selecting the sample of staff from the authorized establishment data obtainable at the Ministry of Health headquarters. However, data about staff establishment available centrally are often hampered by transfers, resignations, long leave, long term illnesses, and deaths. Thus, only staff present at the time of the survey will be potential interviewees. The in-charge of the health facility will also be interviewed.