Pilot phase and implications for study design

The pilot phase involved visits to 10 health centres in Uganda and Tanzania in October - November 2010 to assess the feasibility of quantitative data collection. I visited a variety of HIV care health centres including those eventually selected for inclusion in the study and I examined how data about CPT use were recorded.

The health centres for the pilot phase visits aimed to meet the following site eligibility criteria, which were the criteria used to select sites for the main study:

sites conducted HIV testing,

113 site provided HIV care to both adults and children,

sites provided CPT through the national MoH system, and

the provision of CPT was not dependent upon external support, such as a research study.

Sites for the pilot visits and main study were selected in collaboration with the national MoH and the local EfA partner and aimed to provide a mix of peri-urban and rural sites and to include at least one large hospital and one smaller health centre in each country. I sought to include the greatest number of sites which were feasible to visit within the study time-line. In Uganda, included sites were in the Wakiso district and in Tanzania sites were in the Mwanza and Shinyanga districts.

The districts were selected due to their proximity to the base of the local study EfA study partner. The provision of CPT in two sites from the pilot visits was dependent upon external support as sites were involved in research studies and these sites were not included in the main study.

The main study, therefore, was conducted in 8 HIV outpatient clinics, 4 in each of Uganda and Tanzania. In Uganda, the study was conducted in 2 peri-urban hospitals (1 specialising but not limited to paediatric care) and 2 rural health-centres. The Ugandan sites were Entebbe district hospital, MildMay International, Kasanje Health Centre II and Kigungu Health Centre III. All children who are diagnosed with HIV within the Wakiso district of Uganda are referred to one of the Joint Clinical Research Centre (JCRC), The AIDS Support Organisation (TASO) or MildMay (all non-governmental organisations (NGOs) working in collaboration with the national MoH), and as such, few other MoH facilities have a normal distribution of adults and children accessing HIV care. MildMay International was included as

114 Pansiansi Dispensary and Buzuruga Health Centre. The Hindu Union Hospital was included because much HIV care in Tanzania is provided through the MoH with support from faith-based organisations. A map of the location of the study sites can be found in Appendix 3.1. The selected health centres were similar to other peri-urban hospitals and rural health centres in terms of distance from major cities, number and cadre of staff in clinic, facilities available at each clinic and populations served.

In each health centre visited during the pilot phase, I obtained spare blank copies of the HIV diagnostic registers, pre-ART and ART registers, pharmacy records, monitoring and evaluation (M&E) reports and forms used in patient records. I compared the data tools used and their degree of completion across the sites and across the countries and established which data items were feasible to collect in the main study. The following issues were identified which influenced study design.

Firstly, all patients who enrol into HIV care (pre-ART patients as well as ART-eligible patients) are enrolled in the pre-ART register for pre-ART care, where pre-ART and ART-eligible patients receive CPT. ART-eligible patients enrol firstly in the pre-ART register because they spend a short period in pre-ART care receiving ART

adherence counselling before commencing ART. Therefore, the pre-ART register is useful to gather information on all patients enrolling into care. Only clinics where the diagnostic and pre-ART registers and the submitted clinic monitoring reports were available for the same time-period were included in the main study. As the

115 ability to retrieve historical diagnostic and pre-ART enrolment registers and

monitoring reports varied between sites, I selected the first quarter of 2009 as the entry point for diagnosis.

Secondly, in larger health centres, different diagnostic registers are used in different clinics (e.g. within tuberculosis clinic as well as the HIV clinic), and patient

identification numbers are not always unique, or moreover, linked, between the registers. Furthermore, data are sometimes copied over from one register into the

‘main’  diagnostic  register  meaning  that  any  individual  patient’s  information  could   appear twice. In order to avoid double-counting, the names and initials of each patient were, therefore, temporarily recorded. After establishing which register the health facility considered the  ‘main’  diagnostic register, the main register was searched for the names of all patients recorded in the secondary registers (e.g. the tuberculosis register), and, recognising that names of patients may be the same for family members, additionally matched on sex, age and date of diagnosis.

Immediately after cross-checking, the full names of patients were destroyed.

Thirdly, as the date of CPT initiation was rarely recorded in the pre-ART register, I used the date of enrolment in the pre-ART register as a proxy for CPT initiation.

Fourthly, to gather longitudinal data on those who initiated CPT, I matched patients manually while in clinic on sex, age and date of diagnosis, between the diagnostic and pre-ART register, for patients diagnosed/born January – March 2009. This was necessary as no identification number linking records exists, and information on patients transferring-in was rarely recorded. Similarly, in order to assess time from HIV diagnosis to CPT initiation in the absence of a unique patient identifier, I

116 clinic, but allowed the following variation on any one variable: age by up to 1 year difference, date of diagnosis by up to 1 month difference, and allowing variation in sex (imperfect matching). This matching was used to derive a minimum estimate of the proportion initiating CPT. I further matched the records of patients in the pre-ART register without a date of HIV diagnosis on clinic, age and sex. I also assumed that the number of patients who could not be matched in the minimum estimate had transferred-in for care unofficially and I assumed the same number of patients per clinic would transfer-out unofficially and initiate CPT elsewhere. These patients were not included in the main analyses but can be seen to provide a

maximum estimate of the proportion initiating CPT.