Service Availability and Readiness Assessment 2012

The sample for this survey comprised all districts in the Sentinel Panel of Districts (SPD). This is a panel of 23 districts, plus an additional four districts where demographic sentinel surveillance systems are in operation (Rufiji, Kilombero, Ulanga, Kigoma Urban). The SPD district sampling was conducted by the National Bureau of Statistics using a two-stage, population-weighted probability sample to assure a nationally representative sample of districts that also permitted stratification by zone and by urban/rural area. Sampling weights were included during statistical analysis to account for district's selection probability in a multistage sample design. Probability of a district to be included in a zone was calculated as number of districts selected over total number of districts in a zone. Sampling weights at first stage were calculated as a reciprocal of the probability of a district to be included in SAVVY sample. Since all health facilities in selected districts had equal chances of being included, no sampling weights were incorporated at second stage. It should be noted that results presented in the tables are the number of observations (unweighted counts) whereas results presented as percentages are based on weighted observations.
The overall sample of 27 districts had an estimated total population (2012) of 12.4 million, representing 27% of the total (estimated) Tanzania mainland 2012 population of 45.9 million. The total number of facilities (1908) in the SARA target sample represents approximately 27% of the estimated 7000 health facilities in Mainland Tanzania. Specialist, referral and national hospitals are omitted from the SPD facility sample and are therefore NOT represented in the results presented here, although regional hospitals are included. The total number of facilities in the sample districts is presented in Table 1.1. The target sample numbered 1908 health facilities. Data were collected at 1311 facilities, representing 60% of the target sample. Fourteen facilities had to be dropped in the final analysis due to inability to match it to facility identity, leaving a final total sample number of 1297. Of the 597 facilities where data were not collected, over half (310) were in the Dar es Salaam districts of Temeke,

Ilala and Kinondoni. A further 146 facilities were not covered in the districts of Sumbawanga, Kasulu and Mbozi. Reasons for lower response rate in these districts include the large number of facilities, highly dispersed (difficult to reach) facilities. In one case (Kasulu) illness of the enumerator precluded completion of data collection, while in another (Mbozi) time constraints meant that the deadline for commencing data analysis closed before data collection had been completed. The response completeness rate for this first SARA is certainly lower than ideal. However, we are of the view that the drop out will not seriously bias the results, except to say that private clinics are probably under-represented in our final sample. For future SARA exercises it will be important to have a realistic timeline for fieldwork, analysis and reporting so that data gaps can be followed up and higher reporting completeness can be attained.