Evaluation of the uptake of tuberculosis preventative therapy for people living with HIV in Namibia: a multiple methods analysis

The sample size for this evaluation was based on estimating the percentage of patients in HIV care receiving TPT services with sufficient precision, while accounting for between-facility variation. The true percentage of patients in HIV care (on ART and pre-ART) receiving TPT was assumed to be 50% (95% CI 45–55%). Assuming an intraclass correlation coefficient of 0.02, an average cluster size of 20, a design effect of 1.4, and taking into account the possibility of unavailable data or missing records (inflating the sample size by 5%), a sample size was estimated at 362–560 patient charts for review, corresponding to 20–50% of patients initiating TPT.

To calculate the sample size for each facility, a sampling fraction of 560 patients/163,179 total PLHIV (estimated total number of PLHIV enrolled in care in Namibia during July–September 2013) was applied to each facility included in the study. Sample size was distributed across facilities based on the proportion of PLHIV enrolled in each facility. A minimum number of 10 records were sampled per facility and, factoring in this minimum distribution, the total sample size increased to 905 patients across the 55 facilities.

At the time of the study, Namibia had 13 regions containing 34 health districts with a total of 35 hospitals, 36 health centers, and 143 clinics providing HIV testing and treatment and TPT. Sites participating in this study were purposively sampled with the aim of achieving adequate representation of different facility types, sizes, and geographic regions. Site selection was based on cumulative HIV care and treatment enrollment data from July–September 2013. In total, 55 facilities were selected, representing all 13 regions and 23 of 34 health districts. One facility, Orangemund Clinic, in the Luderitz Health District, was excluded because a permit for access could not be obtained. Because this clinic was small, ultimately representing 0.08% of the overall sample, it was considered to have a minimal effect of representativeness and not replaced with an alternate.

We performed a retrospective review of patient charts at each facility by evaluating the various paper and electronic data sources. We used standardized data abstraction tools to collect data on TB screening, TPT eligibility, TPT initiation and completion, as well as other clinical and demographic variables.

We summarized the demographic and clinical characteristics of people living with HIV along with select HIV-related clinical characteristics. Based on data associated with PLHIV at each facility, a TPT cascade was constructed using frequencies and percentages calculated from the number of patients reaching each of the three stages of the cascade (screened for TB, started TPT, completed TPT). The cascade was stratified by district, gender, age, training on cotrimoxazole and isoniazid (INH) prophylaxis, whether patient was on ART, and type of treatment facility where HIV services were provided.

After reviewing the sample of 54 facilities selected for the study, we determined that smaller facilities were over-represented and larger facilities were under-represented in the final sample. To rectify this, a sampling weight was calculated for each PLHIV on ART in the final sample by dividing the estimated total number of PLHIV on ART seen at each patient’s facility during July–September 2013, by the total number of patients sampled at that facility. Final TPT cascade percentages were then calculated using these sampling weights. Demographic and clinical information collected from patient records included: gender, age, marital status, ART status and start date, year of entry into HIV care, HIV care entry point, and current HIV care and treatment facility.

Qualitative surveys were conducted with HCWs (physicians, nurses and other health care staff) potentially involved with TPT provision. Ideally, five HCWs were selected and consented for survey participation in a private location within each facility. In cases where facilities had fewer than five HCWs, all individuals were surveyed. Teams consisting of staff from Namibia Ministry of Health and Social Services (MOHSS), U.S. Centers for Disease Control and Prevention (CDC) Namibia, and CDC Atlanta conducted surveys using a semi-structured questionnaire designed to assess knowledge and identify barriers and challenges associated with implementation of the TPT cascade in PLHIV. The survey was divided into three broad areas: practices and procedures used by HIV care and treatment facilities for implementing TPT programs; TB evaluation and diagnosis; and final comments. Surveys were recorded and then transcribed for analysis. The majority of questions were semi-structured, close-ended (multiple choice), as well as some open-ended questions defined to elicit short responses. A traditional qualitative analysis was not performed as all responses tended to be one word or a short phrase, based on the nature of the questions (designed to be short answer) and how data were collected (interviewer summarized the response). As a result, open-ended data lent itself to being quantified precisely.

We analyzed survey data using Nvivo 11.0 qualitative analysis software. Basic attributional data on HCWs (e.g., socio-demographic and occupational characteristics) were summarized, and an analytical framework was developed to facilitate coding of individual responses according to specific themes and sub-themes. Themes were reviewed and refined on a weekly basis via inter-group discussions that included representatives of the CDC, MOHSS, and public health experts from the University of Washington and Johns Hopkins University. Individuals from MOHSS were particularly important to this process, representing multiple years of frontline experience working with PLHIV and TB in health facilities across Namibia. This approach was meant to reduce investigator bias during the data analysis phase, particularly with respect to qualitative data and the various themes discussed. Relationships between attributional and thematic data were analyzed and summarized using basic descriptive statistics.

The final sample consisted of 861 PLHIV. Median age was 33 years, 59.5% were female, and most were on ART (82%), with more than half (53.7%) entering care in 2010 and 2011 (3 or 4 years prior to the study). Over half of the PLHIV in the study (54%) reported self-referring to HIV care, and a clear majority stated they received current HIV care and treatment in a hospital setting (81.9%) (Table 1).

Of 861 PLHIV in the sample population, 36 (4.2%) were excluded per National Guidelines (i.e. already on TB treatment prior to enrollment, had started TB treatment the same day as HIV enrollment, or because TB treatment had been started but the timing of treatment with respect to enrollment in HIV care was unknown) [27]. Of those sampled, 825 (95.8%) were eligible for TPT, though 15 were missing information on TB screening. Accounting for sampling weights, 87.1% (679/810) of PLHIV eligible for TB screening were actually screened, of which 16.8% (102/679) screened positive and were referred for diagnostic evaluation for TB disease (see Fig. 2). Of those referred, 25.8% (29/102) were started on TB treatment; the remaining individuals referred for evaluation did not start TB treatment (69.4%; 69/102) or treatment status was unknown (4.8%; 4/102). The total proportion of individuals eligible for preventative therapy (82.6%, 646/810) included individuals who screened negative and were not referred for TB testing, and individuals referred for TB testing but who did not start TB treatment. Of the total patients eligible for preventative therapy, 45.4% (294/646) started and 45.7% (139/294) of those who started completed a course of therapy. Overall, the weighted proportion of patients eligible for TPT who initiated and also completed therapy was 20.7% (139/646) (Figure 2). When the three steps of the TPT cascade were evaluated by district and key patient variables, TPT initiation and completion rates did not vary in terms of any patient variables, with the exception that patients on ART were more often screened for TB, compared to those not on ART (94.6% vs. 66.1%, p < 0.01), and that they were more often eligible for IPT (89.2% vs. 63.8%, p < 0.01). (Table 2).

Overall, 271 HCWs were surveyed, with 16.9% working at hospitals, 33.2% at health centers and 49.9% at clinics. The largest occupational group was nurses (61.3%, 166/271)), while physicians/doctors made up the fewest number of individuals surveyed (1.8%, 5/271). Not all participants answered all questions (answered ‘Don’t Know’ or answer was missing). Based on participant responses, several potential barriers to effective implementation of the TPT cascade were identified and are summarized into five broad categories below, with frequencies based on the number of respondents (n) for each question. Key findings are summarized below and in Table 3).