In this retrospective study we examined patient and health facility factors associated with pretreatment LFU at public health facilities in Uganda; we found that about one in five patients diagnosed with TB experienced pretreatment LFU. Pretreatment LFU is a persistent problem in public health systems high TB burden settings(18–20). In India, one of the countries with the highest TB burden in the world, pretreatment LFU is estimated to be responsible for at least 8% (200,000) of all missing persons with TB annually(5). In our study, the observed proportion of patients experiencing pretreatment LFU would translate into 11% (10,000) of all missing persons with TB countrywide in that period.
Earlier studies examining pretreatment LFU among patients diagnosed with sputum microscopy showed that increased time and monetary costs associated with returning to health facilities to deliver a second sputum sample and/or collect sputum results were partly responsible for observed high rates of pretreatment LFU(9,11,21). Xpert® MTB/RIF testing, a near POC test that requires only one sputum sample and has relatively quick turnaround times held the promise of reducing pretreatment LFU. Results from one clinical trial conducted in South Africa showed a reduction in pretreatment LFU driven by the increased proportion of patients who received a same-day diagnosis(22). However, this finding has not been replicated in routine care settings both in South Africa and Uganda(19,23). Patients accessing Xpert®MTB/RIF testing in our setting still experience relatively long turnaround times(24). In our study, only one-third of patients received a same-day diagnosis.
High patient volumes (measured in our study by the number of Xpert® MTB/RIF tests run each day) likely prolong the turnaround time for Xpert® MTB/RIF testing, and result in more patients experiencing pretreatment LFU. This association between high patient volumes and pretreatment LFU has been shown in Asia and other parts of sub-Saharan Africa(12,13,25,26) and has been attributed to prolonged clinic waiting times, and increased laboratory turnaround times for sputum microscopy. In the Ugandan setting, high patient loads also make it harder to monitor treatment initiation among patients diagnosed with TB. The current system to monitor treatment initiation requires healthcare workers to manually reconcile laboratory registers with treatment registers, a task that may be difficult to perform regularly at health facilities with high patient volumes. At these health facilities, electronic systems that carry out real time monitoring of patient retention along the cascade of care could lead to reductions in pretreatment LFU. Although these kinds of electronic data innovations are commonplace in HIV care, they remain largely unutilized for TB care(27).
HIV-infected patients had higher rates of pretreatment LFU in our study consistent with data from other high TBHIV burden settings (20,28). Late presentation to care could partially account for these LFU patients. In Malawi, advanced HIV disease was shown to result in suboptimal linkage to TB treatment as patients were often too sick to return to the health facility for their results or died before treatment initiation(8). In Zimbabwe, close to 50% of pretreatment LFU was due to deaths before treatment initiation particularly among HIV-infected patients(20). In our study, late presentation to care was examined by analyzing the ART status of patients who were started on TB treatment. Among those patients whose ART status was available, about a quarter (27%) initiated ART after starting TB treatment. This is consistent with routine surveillance data from the AIDS Control Program that shows that despite the roll out of “test and treat” for HIV, about 30% of all newly diagnosed HIV patients still present with Stage III and IV disease(29). The introduction of additional point-of-care tests with shorter turnaround times e.g. lateral flow urine lipoarabinomannan (LF-LAM)(30) into the diagnostic algorithm for patients with Stage III and IV disease may improve linkage to treatment among this group of patients.
In our study, patients who did not have a phone number recorded was strongly associated with pretreatment LFU. Although patients may deliberately decline to divulge their phone numbers due to self-stigma related to TB(31), the proportion of patients with a recorded phone number in our study(63%) was comparable to the national phone coverage for rural areas (65.7%)(32) and is therefore likely to represent actual phone ownership. Patients without phone numbers may belong to a lower socio-economic class and may lack the financial means to return to health facilities to receive their results and start on TB treatment (33). The recently concluded patients’ costs survey in Uganda showed half of all TB patients incurred catastrophic TB care costs which were mainly driven by nonmedical expenditure such as travel(34). Interventions to reduce these costs for the most vulnerable patients e.g. prioritizing them for same-day diagnosis or provision of socioeconomic support may reduce pretreatment LFU. Similarly, community tracing interventions, where community healthcare workers conduct home visits to trace patients with no phones would also help reduce pretreatment LFU.
Consistent with studies from Ghana(12) and other settings in Uganda(35), there was no association between distance from the health facility and pretreatment LFU. This may be due to the decentralized nature of health service delivery in Uganda where patients access care at health facilities closest to their homes. In our study, nearly half of all patients resided within 20kms of the health facility. This also makes
Study Strengths and Limitations
Our study used data collected from different levels of the healthcare system. It is therefore likely that these findings are representative of and generalizable to the public healthcare system in Uganda. However, because data for this study was collected under routine programmatic conditions, missing data may have introduced bias into our study resulting in an overestimation of pretreatment LFU. This was minimized by triangulating many data sources within the healthcare facilities and at the district level.