The 2009 baseline survey covered households in 5 rural districts in 5 provinces: Farza (Kabul Province), Shahfoladi (Bamyan Province), Ghorian (Herat Province), Farkhar (Takhar Province), and Qurqin (Jawzjan Province). These household surveys used the same sampling method as that of the annual household surveys of the MOPH’s USAID-funded Partnership Contracts for Health Services through nongovernmental organizations (NGOs) in those five provinces. The standard procedures for informing respondents of the purpose of the assessments and the guarantee for anonymity used in the annual household surveys were applied and the survey did not introduce new interventions, nor prevented access to interventions nor exposed individuals to possible harm.
As part of a lot quality assurance sampling (LQAS) method applied to sick children under 2 years of age in 5 districts, we first selected at least 130 households to be surveyed in each district, with the hope of yielding 100 households per district in which there had been a sick child within the previous 2 weeks. The planned total sample involved 100 households drawn from each of the 5 districts, for a total of 500. We used the listings of all the households in those 5 districts to identify the households belonging to 5 supervisory areas in each district. A supervisory area is a defined part of a district in which the NGO responsible for delivering health services and the MOPH regularly oversee all health activities. Within each supervisory area, at least 19 households were selected randomly.
In total, 492 children of 2 years of age or less who had been ill with acute respiratory infection (ARI), diarrhea, or fever within the previous 2 weeks were identified and included in the study. The parent or caretaker was interviewed only if there had been a sick child in the household within the previous 2 weeks. If there had been no sick child in any of the randomly selected households, the surveyor went to the nearest household seeking the presence of a sick child within the previous 2 weeks. The surveyor continued moving to the nearest household until a household with a sick child was identified in place of the initially randomly selected household. This is why more households were sampled than the intended sample of 100 households per district. In households in which a child under 2 years of age had been sick in the previous 2 weeks, the surveyor sought verbal consent from the household member for participating in the survey, as recommended in the procedures of the standard LQAS HHS in Afghanistan.
At each household where a sick child was identified, the surveyor used a structured questionnaire to ask the parent or caretaker a series of questions about the nature of their child’s illness; the nature of the illness; whom they had consulted outside the home for the illness; if they were referred, how they complied with a referral; and any real or perceived problems in accessing the next-level health facility that affected compliance with the referral, including geography, distance, transportation, and costs related to compliance with the referral. The data that were obtained differentiated between children who had been referred outside the home and those who had not been referred outside the home for their illness.
The survey fieldwork was carried out by data collectors and supervisors overseeing their sampling work. The surveyors were staff of the NGO providing services in the district. The staffs received 2 days of training and were checked to obtain more than 90% inter- and intra-surveyor reliability in using the survey questionnaires. Before leaving a household, the surveyor checked that all the questions had been completely answered. After a district was sampled, the survey supervisors ensured that all surveys were checked for completeness. If there were any missing responses, the surveyor would return to that household and complete the remaining questions. A second surveyor performed a 5% re-survey of the sampled households to check the reliability of the survey results. Upon completion of all the surveys, the data were reviewed for completeness and coded for entry into a database. When all the data were available, we held a workshop to analyze the data and review the results with the nongovernmental organizations and seek solutions to problems identified.
Statistical significance was tested by the two-tailed Fisher exact test for 2x2 contingency tables and the chi-square test for independence for larger contingency tables, using GraphPad InStat version 3.1, 32 bit for Windows, GraphPad Software, San Diego California USA, http://www.graphpad.com