Study setting and context. Uganda first adopted the use of Xpert for TB diagnosis in patients living with HIV or requiring testing for possible multidrug-resistant (MDR) TB in 2011. In September 2017, Uganda adopted policy recommendations in line with WHO guidelines that recommends Xpert testing where available for all patients with possible TB [6]. With support from partners, Uganda has installed 249 Xpert devices in 227 of 1500 (15%) TB diagnostic units in the country [6] To date, Xpert devices have been deployed at all regional or district hospitals most of which are serving as hubs. In Uganda, sputum referral networks are facilitated by motorcyclists, locally called “boda riders.” Riders have been employed by Central Public Health Laboratories (CPHL) with support from implementing partners to carry sputum specimens to Xpert testing hubs and to return to the same health centers with results. With the establishment of this hub and spoke network, the Ministry of Health intended that all patients with signs or symptoms of TB presenting to community health facilities would have access to rapid, referral-based Xpert testing services. Each hub serves approximately 20-40 community health centers, and each rider’s daily route is expected to cover 4-8 facilities (NTRL, personal communication). Of the participating health centers in our study, 8 are level III characterized with a smaller human resource structure of about 18 staff, led by senior clinical officer, with an operating laboratory to provide TB health care services including TB screening, testing, diagnosis and treatment. The remaining 15 participating sites are level IV health centers, which provide all services the services of a level III center, and in addition have the physical infrastructure to admit patients for longer periods of observation and treatment, to conduct emergency operations and led by a senior medical officer and another doctor, and to support obstetrical services such as ante-natal care and labor and delivery.
Study design. This cross-sectional study combines several sources of triangulated quantitative and qualitative data to assess the coverage, penetration, and fidelity to implementation of Xpert testing in Uganda’s hub and spoke model from the perspective of level III-IV community health facilities.
We collected survey data from a sample of community health centers participating in hub-and spoke Xpert testing, supplemented by concurrently collected field notes from these sites and observations made by our research staff on days when they were present at these sites. Community health centers were included if they: 1) used sputum smear microscopy as the primary method for TB diagnosis; 2) participated in external quality assurance testing for sputum smear microscopy led by the National Tuberculosis and Leprosy Program (NTLP); and 3) were linked to an Xpert testing site/hub. Community health centers were excluded if they: 1) performed sputum smear examination on <150 patients per year based on 2015 NTLP data; 2) diagnosed <15 smear-positive TB cases per year; or 3) were located within Kampala District or >150 km from central Kampala City. Of 1105 level III-IV community health facilities in Uganda that reported to the NTLP, 27 met eligibility criteria based on review of 2015 NTLP TB testing and treatment data. Of those eligible sites, 23 were selected for the study after consultation with the NTLP.
Data collection. Data for this study was obtained in three ways: surveys, data abstraction from routine health registers, and field notes recorded by study staff from observing community health center staff, work processes and personal communication. This data was collected as part of a mapping exercise in preparation for studies conducted by the Uganda Tuberculosis Implementation Research Consortium (U-TIRC) aiming to assess and improve TB diagnostic evaluation in Uganda. Preparatory activities included data collection (as described below) as well as training for health center staff on National Tuberculosis Program Guidelines. A trained study Medical Officer provided a one day onsite TB guidelines training to all health center staff who were involved in TB evaluation and management. To ensure quality of data documentation, we also trained health center staff involved in TB care on how to complete the registers.
First, to assess characteristics of the community health centers, a research staff member administered a site assessment survey to clinic staff using a standardized tool developed to capture information on infection control procedures; laboratory procedures; TB testing and treatment procedures; and utilities, infrastructure and resources. Survey participants included (in order of priority based on seniority, in keeping with norms in Uganda) any of the following individuals available at the time of the study staff site visit: the health center director (also known as the facility in-charge), laboratory director, and/or TB focal person at each health center.
Second, to obtain patient-level data on utilization and outcomes of Xpert testing, we reviewed routine TB registers including NTLP laboratory registers, Xpert referral forms, an electronic reporting platform for GeneXpert results (GxAlert, SystemOne Northampton USA) and NTLP treatment registers, all for the period January – December 2016. We included data on all patients undergoing TB testing who had one or more of the following indications for Xpert testing per Uganda NTLP guidelines at the time the survey was completed: persons living with HIV, health care workers, contacts of drug-resistant (DR-TB) patients, pregnant women or breast-feeding mothers, prisoners, patients from refugee camps, and diabetics. To ensure that we captured of all Xpert testing, we also downloaded testing data directly from the Xpert devices at testing hubs serving each participating community health facility. We developed a standardized data extraction tool to match patients across data sources. Data abstracted included age, sex, date of initial sputum collection, method of sputum examination (smear microscopy, Xpert), sputum examination results, date of results, date of treatment initiation, and HIV status. All the data abstraction, collection, and entry was done by trained research assistants using secure mobile-based Research Electronic Data Capture (REDCap, Nashville USA)[7] platform hosted by the University of California San Francisco. To ensure data quality, data was subjected to a rigorous quality assurance process using REDCap generated reports on a bi-monthly basis by our data management team.
Finally, study staff trained as laboratory technologists and medical officers recorded observations about the TB diagnostic evaluation process in participating community health centers during site visits for trainings, surveys, and data abstraction. These study staff were also previously trained on National Tuberculosis Guidelines and specifically laboratory procedures for the diagnosis of patients with possible TB. Staff were asked to take notes related to the process of specimen collection, specimen transport, specimen testing, result reporting and patient linkage to treatment initiation if diagnosed with TB. Notes were shared between staff to assess comparability and 1-2 staff took notes at each site visit.
Data analysis.
We described health center characteristics including infrastructure and human resources, and process metrics for provision of TB diagnostic care survey data using proportions with 95% confidence intervals (CIs) for dichotomous outcomes and either medians with inter-quartile ranges (IQR) or means with standard deviations (SD) for continuous outcomes. We compared differences in proportions across sites using the χ2 test of proportions. Qualitative data from field notes were organized by health center, transcribed and reviewed for thematic interpretation using a framework analysis to identify key issues related to the hub-and-spoke model of TB diagnostic evaluation. We used qualitative inquiry to elucidate site specific and general health system barriers to Xpert testing. While formal coding was not conducted, data were summarized into categories corresponding to processes of care for TB diagnosis (Figure 1). We performed all quantitative analyses using Stata version 14 (Stata Corporation, College Station, TX). Qualitative data was categorized and reviewed in Excel (Microsoft, Redmond, WA).
The study was approved by the School of Medicine Research and Ethics Committee at Makerere University, the Uganda National Council for Science and Technology, and the University of California San Francisco Committee on Human Research.