Objective III: Determine the role that HIV treatment and care programs play in the provision of family planning services, including client’s expectations of family planning and whether they feel
3.2 Candidate’s role in study design and conduct Overview
This section describes the candidate’s role in each phase of this study. Details of the methodology can be found in sections 3.3 and 3.4. My role in the conceptualisation of the study is described at the beginning of this thesis.
Once the study had been conceptualized and the research questions developed, I conducted a sequence of quantitative analyses, and used these results to inform the following qualitative study. I began with a preliminary analysis of demographic surveillance data, then designed and constructed a unique cohort from a compilation of existing datasets, performed data consistency checks and finally selected and applied survival analysis techniques to the cohort. The details of my role are presented below.
Role in the quantitative analysis
Upon arrival to the Africa Centre, I conducted descriptive analysis of crude birth rates in the demographic surveillance area to familiarise myself with the data available in the Africa Centre
81
Demographic Information System and to describe general fertility in the setting. A summary of the surveillance procedures and the data available are shown in Box 3.2.I.
Box 3.2.I An overview of demographic surveillance undertaken by the Africa Centre Since 2001 the Africa Centre has annually captured demographic information from a
defined surveillance area.
This surveillance includes 16 ‘core’ questionnaires and 5 additional ‘modules’
‘Core’ questionnaires capture household and individual information including births, deaths and migrations
A key informant, usually the head of household, is asked to give consent for participation and to respond to core surveillance questions.
Additional ‘modules’ include Women’s General Health and HIV surveillance.
Core questionnaires and modules are all administered at the household by trained fieldworkers.
Data capturers then enter information in to separate datasets that are each unique to a questionnaire round and compose the ‘Africa Centre Demographic Information System’, known as ACDIS.
I used these analyses to inform the design of a unique cohort of HIV positive and HIV negative female residents. In April 2010, I submitted a request for data to the Africa Centre data manager to access all ‘core’ and three of the five ‘module’ datasets: Women’s General Health, HIV, and
Household Socio-Economic. I requested every annual dataset since 2001 and later updated this request for new surveillance rounds until December 2012.
The 2007 HIV surveillance dataset served as the basis for my unique dataset, to which I first merged annual HIV surveillance datasets between 2008 and 2012. Merges were possible by linking
individuals through a unique identification number and I used this feature to add variables from other core and module datasets between 2007 and 2012. Variable selection is described in greater detail in section 3.3.6. With each merge I also cleaned the data; duplicate, missing and erroneous data were identified through cross-tabulations and resolved through consistency checks with other variables. I corrected any errors in the merge coding or discussed discrepancies with data managers and removed duplicate records. Once clean, I applied exclusion criteria to prepare this cohort
82
dataset for analysis, as detailed in section 3.3.4. This unique cohort dataset provided the basis of three longitudinal analyses.
I received training in techniques for longitudinal analysis during a short course on ‘advanced statistical methods in epidemiology’ at the London School of Hygiene and Tropical Medicine. I also received consistent informal support from staff and researchers at the Africa Centre to conduct complex analysis using STATA 11. The training helped to inform my decision to use Poisson and Cox statistical techniques for an open and a closed cohort analysis as presented in Chapter Four.
Role in qualitative study
To address this study’s qualitative research aims I decided to conduct semi-structured interviews with clients and health workers of the local HIV treatment and care programme. I was responsible for the design of this study and developed both client and health worker interview protocols. I decided a semi-structured tool was most appropriate because I had identified thematic areas of interest, but also hoped to capture emerging themes during the interviews. Before piloting, I sought and received ethical approval from University College London and the University of KwaZulu-Natal. I recruited an isiZulu speaking research assistant (RA) and prepared a training booklet, which I used during a week of training. I arranged and attended practice interviews between the RA and
researchers within the Africa Centre and a later pilot in a local clinic, so that I could provide further training and gain an understanding of the contexts of the interviews.
I observed three initial interviews conducted by the RA with clients in the clinics and provided feedback on how to improve interview conduct and probing. I waited outside of the room for the remaining interviews and arranged an immediate debrief with the research assistant, including a review of her field notes. The research assistant transcribed and translated all client interviews, with added details and interpretations of local or cultural nuances.
83
I conducted all health workers interviews in English. I piloted the health worker interview tool in a local clinic, conducted 12 interviews and made field notes during each interview. I transcribed electronic recordings of interviews on an on-going basis and used these results to inform areas that needed additional probing. When needed I sought the perspective of my research assistant to explain cultural nuances that I did not understand and to check consistency with client interviews.
Finally, I adapted a tool from Johns Hopkins to conduct standardised observations of family planning delivery in each HIV treatment and care facility where interviews were conducted.
I undertook the qualitative analysis of both client and health worker interviews. I first summarised data using an excel spreadsheet as interviews were conducted. I later completed a QSR two-day course to inform my conduct of a framework analysis using Nvivo software. Details of the qualitative methodology are provided in section 3.4 and the discussion from this analysis is provided in
Chapters Five and Six.
3.3 Quantitative Methodology