Healthcare delivery in Uganda is based on a tier system with primary, secondary and tertiary levels of care. Within the healthcare system, TB care services are provided at all secondary and tertiary levels of care, as well as selected primary care facilities. Over the past five years, sputum microscopy, previously the main diagnostic test for TB has been increasingly replaced by Xpert®MTB/RIF testing. To date, about 235 health facilities (15% of all health facilities which offer TB care services) are equipped with Xpert® MTB/RIF machines. At these health facilities, Xpert®MTB/RIF testing is the initial diagnostic test for all patients with signs and symptoms of TB14. Health facilities which do not have Xpert®MTB/RIF machines use sputum microscopy as the main stay of diagnosis but access Xpert® MTB/RIF testing for selected patient populations e.g. patients infected with HIV and those previously treated for TB through a specimen referral system. In this system, health facilities with Xpert®MTB/RIF testing act as “diagnostic hubs” for lower health facilities within a 20–30 km radius. Sputum samples are transported by motorcycle to the diagnostic hubs and results are returned to the referring health facilities by the same courier. All patients diagnosed with TB are registered in standardized paper based national registers
TB treatment is offered on an outpatient basis, unless there is an indication for hospital admission, and is free of cost to the patients. The Uganda National TB and Leprosy program (NTLP) recommends that all patients diagnosed with drug-susceptible TB are started on treatment as soon as possible, preferably within 24 hours9. For patients who prefer to complete TB treatment at a health facility other than the one where TB diagnosis is made, the NTLP still recommends that TB treatment is started and the patient is subsequently referred to the health facility of their choice.
The NTLP uses a national standardized paper based system for recording and reporting all patients with signs and symptoms of TB (TB presumptive registers); all patients with a bacteriological confirmation of TB (TB laboratory registers) and all patients started on TB treatment (TB treatment registers). These registers are present at all health facilities where TB services are offered. Data for all patients started on TB treatment within each district is collated into one district TB register periodically to allow for recording of treatment data for patients who may have started or completed TB treatment at a different health facility than the one they were diagnosed at.
To get a representative picture of the healthcare system, we selected for this study, health facilities from different levels of the healthcare system (three primary care facilities, four district hospitals and three tertiary referral hospitals) across ten districts in Uganda (Figure 1). All facilities in this study have Xpert® MTB/RIF testing available onsite and use this test as the initial diagnostic test for patients presenting with signs and symptoms of TB14. All facilities in this study also act as diagnostic hubs and receive additional samples for testing from primary care facilities within their catchment area.
At all participating health facilities, we used data from laboratory registers to identify patients aged ≥15 years who had a positive Xpert®MTB/RIF test and were rifampicin sensitive between January 1st and June 30th 2018. We excluded all patients had had an Xpert®MTB/RIF test done for treatment monitoring. We then compared diagnostic data with treatment initiation records in the health facility TB treatment registers and in the district TB registers (for patients who could have started on TB treatment at a different health facility within the same district). Patients were considered to have experienced pretreatment LFU if their names were in the laboratory register but no evidence of their names in the health facility clinic or district TB registers within two weeks of diagnosis.
Data on patient characteristics including age, sex, HIV status, ART status, and residence was collected from the laboratory registers. Patient’s age was categorized in ten-year age groups starting at 15 years in according to the practices on case notifications at the district and national level. Distance to the health facility was calculated as the linear distance, based on global positioning system coordinates (QGIS Desktop, Versions 2.12.0), from the patient’s recorded residence to the health facility.
Health facility records were used to obtain data on health facility characteristics including the number of Xpert®MTB/RIF tests done, cartridge and medicine stock outs. Data on cartridge malfunction was collected by interviewing laboratory healthcare workers at each health facility.
Baseline characteristics of the study population were described using frequencies and percentage. The proportion of patients who experienced pretreatment LFU was also described. A multilevel logistic regression model accounting for clustering by health facility was used to examine factors associated with pretreatment loss to follow up. A sensitivity analysis was performed using multiple imputation to examine the effect of missing data on our measures of association. For variables with missing data, we assumed data were missing at random and performed multivariate normal imputation12 using 100 imputations. We used age, sex and level of health facility as predictor variables. All data analyses were carried out using STATA® version 13