Menu
Menu
Menu
Menu
Background: The intended meaning behind responses to standard questions posed in largescale health surveys are not always well understood. Systematic follow-up studies, particularly those which pose a few repeated questions followed by open-ended discussions, are well positioned to gauge stability and consistency of data and to shed light on the intended meaning behind survey responses. Such follow-up studies require extensive coordination and face challenges in protecting respondent confidentiality during the process of recontacting and reinterviewing participants. Objectives: We describe practical field strategies for undertaking a mixed methods follow-up study during a large-scale health survey. Methods: The study was designed as a mixed methods follow-up study embedded within the 2014 Ghana Demographic and Health Survey (GDHS). The study was implemented in 13 clusters. Android tablets were used to import reference data from the parent survey and to administer the questionnaire, which asked a mixture of closed- and open-ended questions on reproductive intentions, decision-making, and family planning. Results: Despite a number of obstacles related to recontacting respondents and concern about respondent fatigue, over 92 percent of the selected sub-sample were successfully recontacted and reinterviewed; all consented to audio recording. A confidential linkage between GDHS data, follow-up tablet data, and audio transcripts was successfully created for the purpose of analysis. Conclusions: We summarize the challenges in follow-up study design, including ethical considerations, sample size, auditing, filtering, successful use of tablets, and share lessons learned for future such follow-up surveys.
The study was carried out in Ghana, West Africa, in three purposively selected administrative regions: Greater Accra, Northern, and Central. The study was a follow-up to the 2014 Ghana Demographic and Health Survey (GDHS). The follow-up study was designed as a mixed methods study that followed up with a subset of respondents from a nationally representative survey, a model characterized as an embedded sequential mixed methods study [12] and described in detail by Schatz [8]. The follow-up study was funded, planned, and fielded independently from the main GDHS, but respondents were selected systematically from among the original GDHS respondents. Mixed methods studies are well-positioned to provide important insights about demographic behaviors; such studies are particularly valuable when open-ended responses can be compared against findings from large-scale population studies [7,10,17]. The 2014 GDHS was a nationally representative survey of 9,396 women age 15–49 and 4,388 men age 15–59 residing in 11,835 interviewed households [18]. As with other Demographic and Health Surveys, the GDHS provides information on fertility, family planning, infant and child mortality, maternal and child health, nutrition, malaria, HIV, and non-communicable diseases in relation to respondents’ socioeconomic and demographic characteristics. Fieldwork for the 2014 GDHS was conducted by the Ghana Statistical Service and the Ghana Health Service, with technical assistance from ICF International through The Demographic and Health Surveys (DHS) Program, which is funded by the United States Agency for International Development. As is standard with DHS protocol, the GDHS used a two-stage sampling design with probability proportional to sample size [19]. In total, 427 clusters from across the country were selected, and within each cluster 30 households were randomly selected for inclusion. The GDHS attempted to interview all women of reproductive age in each selected household. The household response rate was 98.5 percent, and among women age 15–49 in selected households the response rate was 97.3 percent [18]. Three regions for follow-up were selected based on cultural and socioeconomic diversity, population size, and diversity in family planning use and fertility levels. These were Northern Region (NR), a very high-fertility region, Central Region (CR), a moderate-fertility region, and Greater Accra Region (GAR), with the lowest fertility. It was decided in advance of fieldwork that of the thirteen study clusters, five would be selected from NR, five from CR, and three from GAR. Within GAR all three clusters sampled would be urban, and within NR and CR there would be one urban and four rural clusters each. This ensured both a balance of urban and rural respondents and diversity among the urban population. Within each region, clusters for the follow-up study were selected as the GDHS was being fielded. A completely random subsample of GDHS clusters would not have been feasible; cluster selection needed to balance diversity with logistical practicality. The GDHS fieldwork extended over several months, but the follow-up study was fielded by six interviewers working full-time over the course of a single month. The aim was to visit clusters for the follow-up study in October 2014 within one to four weeks of the date of initial interview in the GDHS. Hence only a limited number of clusters were available for selection. Ghana Statistical Service shared fieldwork team itineraries and progress reports with the follow-up study team. Based on an examination of these schedules, and reflecting a desire for geographic and cultural diversity, the follow-up study team proposed final cluster selections to ICF, which exercised some oversight for geographic diversity. The approximate locations of the final 13 clusters selected for the follow-up study are shown in Figure 1. Survey clusters. Cluster locations illustrated on the map have been randomly displaced up to ten kilometers from their actual location using standard DHS procedures to ensure respondent confidentiality [20]. The 2014 GDHS used computer-assisted field editing (CAFE). Initial data entry was done in the field before the GDHS team moved to the next cluster. Paper questionnaires were sent to the Ghana Statistical Service office in Accra for validation. As the data for each of the 13 selected follow-up clusters arrived in the Ghana Statistical Service home office, and after the initial data entry had been validated, ICF staff used a CSPro program to confidentially select eligible follow-up respondents from among those who had consented to be contacted for a follow-up interview: married or sexually active women age 15–44 who either met the standard DHS definition of unmet need (excluding postpartum amenorrheic), or were in a subset of GDHS family planning users in the cluster [16]. The CSPro program automatically selected and output the variables and identification fields necessary for reinterview into a spreadsheet format for each eligible respondent; these were uploaded to a secure server that could be accessed by the implementing agency and downloaded to follow-up study tablets. The follow-up study used three questionnaires, one for each of the eligible study groups: non-pregnant women classified by GDHS as having unmet need, pregnant women classified by GDHS as having unmet need, and a reference group of women currently using family planning. After an initial set of six identity verification questions, respondents were asked between one and three screening questions to confirm their eligibility for the assigned questionnaire (for example, ‘Are you currently doing something or using a method to delay or avoid getting pregnant?’). If the respondent indicated a different answer than was given to GDHS, for example because she had started or stopped using family planning since the last interview, interviewers were instructed to ask about the discrepancy and to switch to the correct questionnaire before proceeding. The questionnaires covered topics such as reproductive intentions, family planning use, attitudes toward family planning, the role of partner and extended family in decision-making, and barriers to access. Respondents were also re-asked a few key questions about pregnancy, fertility preferences, family planning use, and reasons for non-use: first to ascertain consistency of responses, and second to allow interviewers to probe further into their meaning. In total the questionnaires contained between 24 and 31 groups of questions. A typical group of questions was comprised of an open-ended and closed-ended question, followed by a prompt to explain the closed-ended answer. The actual number of questions respondents were asked varied depending on the questionnaire type and on the skip pattern followed based on their own responses. An 11-day translation and interviewer training was conducted at the University of Ghana-Legon for eight interviewer training candidates. Training provided an overview of the study research questions and design, the concept of unmet need for family planning, and principles of qualitative interviewing. Field procedures were discussed extensively; interviewers learned how to use the audio recorders and Android tablets and engaged in back-translations of the questionnaires and in extensive role plays of interviews. Role plays provided an opportunity to test and revise the tablet program. The eight interviewer candidates and three field supervisors practiced how to download survey cases, upload results to the secure server, and change from one questionnaire type to another. Interviewer training for the mixed methods follow-up survey was timed to coincide with GDHS fieldwork, such that the follow-up survey could conduct a pre-test in recently completed clusters in Accra toward the end of training. Pretesting in the two clusters involved running the selection program on the final data entry, collecting maps of the cluster and a field guide, and then seeing 12 respondents determined to be eligible. Each interviewer candidate conducted at least one pretest interview. The pretest proceeded through the follow-up survey process in full: requesting consent for interview and for audio recording, and administering a tablet-based interview with closed- and open-ended questions. Supervisors reviewed the downloaded data and audio and provided feedback to the interviewers; afterwards, slight revisions were made to the questionnaire and to the tablet program. At the end of training, all interviewers received copies of the final survey instruments and conducted full rehearsal interviews. Six of the eight interviewer candidates were hired for the study. Fieldwork for this study was conducted in October 2014. Three field teams, each consisting of two interviewers and a supervisor accompanied by a guide from the Ghana Statistical Service, tracked and identified the selected GDHS respondents, typically within one to four weeks of the original survey by using the household address, the name of the head of household, and the woman’s relationship to the head of household. In rural areas a village leader was approached for permission before beginning fieldwork. Interviewers returned to households up to three times to complete the interview. The questionnaires were implemented in Mobile Data Studio software on Android Samsung Galaxy tablets. Closed-ended responses were entered into tablets, and open-ended responses were captured using audio. The use of Computer-Assisted Personal Interviewing (CAPI) enabled answers to be compared in real time against responses given to the GDHS, and respondents could be asked about any discrepancies. In order to confirm the identity of selected respondents and to enable the tablet to display appropriate GDHS data entry next to questions, a remote secure server had been set up to pre-populate data in follow-up questionnaires after selection of an eligible respondent and electronically signed verification (by the interviewer) that she had obtained the respondent’s consent to be interviewed. Respondents were asked six background questions to validate their identity: year of birth, month of birth, marital status, whether ever given birth, number of resident daughters, and number of resident sons. After the interview was completed and a field supervisor reviewed the tablet data entry, it was uploaded to the remote secure server and exported to a spreadsheet. Interviews were randomly audited by the Ghana Statistical Service guide to ensure that they were correctly completed. The entire process of fieldwork is summarized in Figure 2. Overview of study procedure. Out of 9,396 total female respondents age 15–49 in the 427 GDHS clusters, 99.6 percent gave consent to be contacted again for the follow-up study. In the 13 clusters selected for the follow-up study, a data processing program determined 142 respondents to be eligible for follow-up based on their permission and the qualifying criteria discussed earlier. Of these, 135 women were successfully contacted and reinterviewed. Two women refused at the time of follow-up, and five were away or unable to schedule a follow-up interview despite repeated attempts. Despite having matched on household address or structure number, name of head of household, and relationship to head of household, four of the 135 respondent identities could not be correctly verified during the data processing phase (for example there was a discrepancy in whether or not they had given birth). After excluding these unverified respondents, the final response rate was 92.3 percent. All respondents assented to audio recording of their interviews and the backup notebooks interviewers carried did not prove necessary. Two of the respondents were daytime visitors to their household and thus had not been fully interviewed by the GDHS. They were excluded from the qualitative sample on the grounds that no information about family planning use had been gathered about them by the GDHS. The remaining 129 interviews were transcribed and, if necessary, translated into English. These resulted in over 1,000 pages of transcripts. Transcripts were entered into an ATLAS.ti database and systematically coded according to question number and theme. Additionally, the quantitative tablet data was cleaned and checked against the interview transcripts. These data were then imported into Stata and confidentially linked to the final DHS dataset for analysis. Ethical clearance for the follow-up study was obtained in tandem with clearance for the GDHS from the ICF Institutional Review Board, which determined that both studies complied with all of the requirements of the US Code of Federal Regulations 45 CFR 46. Upon Institutional Review Board clearance, permission to share data between the GDHS and the follow-up study was obtained from the Ghana Statistical Service, which implemented the GDHS. As the GDHS was conducted on paper and cluster selection was not decided upon in advance, all female respondents age 15–49 were asked at the end of the GDHS questionnaire, in the language of their interview, if they would consent to be re-contacted for a follow-up study on family planning and childbearing. The follow-up survey obtained consent prior to reinterview, both for the interview itself and for audio recording of responses. In keeping with Institutional Review Board regulations, the confidentiality of the respondent’s information was maintained at all stages of the survey. Anonymous cluster and respondent identifiers were created and used for recordkeeping. The original information used to locate respondents for follow-up (name of household head, address, and all initial data entry from the GDHS) was destroyed by interviewers and supervisors at the conclusion of fieldwork, and only a new, anonymized identification number was maintained for correspondence with the DHS home office. Following the conclusion of the GDHS, cluster and household numbers for all respondents nationwide were scrambled according to established DHS protocol, and original records of cluster and household numbers were destroyed prior to linking respondents’ information with HIV test results. With permission from Ghana Statistical Service and the ICF Institutional Review Board, the follow-up study was able to maintain an internal, confidential linkage between the anonymously identified follow-up respondents and the final, scrambled GDHS dataset.
N/A
