Retrospective study:

This study brings out the fact that PLTFU of bacteriologically-confirmed pulmonary TB patients in Cameroon is still a crucial issue that requires urgent measures. The high PLTFU rate implied that these infectious TB patients could still transmit the disease in the community even after diagnosis. The analyses also revealed that many patients had to travel long distances to be diagnosed at a health facility, and poor geographical access to the facility was a significant risk factor for PLTFU.

From this study, 16.7% of bacteriologically-confirmed pulmonary TB patients did not initiate TB treatment within 7 days at the DTUs of diagnosis. In a previous systematic review, the proportion of PLTFU in some African countries ranged from 6% to 38%.

However, the estimates from different studies could not be directly comparable with that from this study since different methods were used to define and measure PLTFU.

The temporal definition of PLTFU ranged from two days to three months in prior studies ⁷. For example, patients might be classified as having PLTFU in other studies if they did not receive treatment within two days of diagnosis. The temporal definition of PLTFU in the present study was 7 days and the follow-up duration was at least 60 days.

When the two-days’ criterion used by Uchenna and colleagues in Nigeria was applied, the proportion of PLTFU increased from 16.7% to 47.5%.

After excluding those who were known to have been referred to other centers, the proportion of PLTFU dropped from 16.7% to 10.0%. Noteworthy is that the 10.0%

figure might be an underestimate since we do not know whether those referred to start treatment elsewhere eventually received treatment. Fewer studies attempted to trace the true outcome of those who were initially lost to follow-up. In several Asia studies, a

common reason for PLTFU was receiving treatment at a private clinic. However, true outcome and reason for PLTFU remained largely under-studied in Africa.

A key and unique finding from this study was that longer travel distance from patient’s home to DTU was associated with PLTFU. Compared to the patients who traveled 30 km or less to the DTU where they were diagnosed, those who travelled over 30 km to the DTU had a 2.4-fold increased odds of PLTFU. A similar pattern was also observed for travel time and PLTFU. A possible reason for this finding could be that, living far away from the DTU caused some patients to give up on receiving treatment. Another possibility is that living far away caused some of the patients to be referred to the DTU nearer their homes. Few studies have equally examined geographical access as a potential reason for PLTFU. In a Vietnam study, the DTU being far away from the household was one of the reasons given by patients who had not initiated treatment. On the other hand, the distance between patient’s residence and the health facility was not associated with PLTFU in a single-hospital study from Ghana.

About 32% of the patients in our study had to travel for more than 30 km and 36%

traveled for more than 30 minutes from their homes to the DTUs where they were diagnosed. One possibility for this long travel could be that the DTUs may not be evenly distributed within these two regions where the patients lived. It could also be a matter of patients’ preferences of health facility. In other words, patients may not seek TB diagnosis at the nearest DTU with regard to their residence. Indeed, we did observe this phenomenon in our study when we visually inspected the distribution of TB patients by DTUs. Further explorative analyses on patients’ care seeking pathway will generate critical information on how to overcome the patients’ barrier to TB diagnosis

28

and treatment. Our future study will also look at identifying these TB hotspots of PLTFU after adjusting for population size.

The Xpert MTB/RIF assay is an automated, nucleic acid amplification test with a turn-around time of two hours. With the possibility of “same-day” TB diagnosis, use of Xpert MTB/RIF might lead to a decrease in PLTFU compared to smear microscopy, which requires multiple visits to the clinic. In a previous randomized controlled trial in southern Africa, the use of Xpert MTB/RIF at the point of care was associated with a lower proportion of PLTFU (2/182, 1%) when compared with smear microscopy (3/114, 3%), although the difference was not statistically significant.³⁸ In our study, being diagnosed by Xpert MTB/RIF was however not associated with a decreased odds of PLTFU when compared to sputum microscopy in the multivariable analysis. In fact, in our third analysis where the retrospective and prospective data were combined, diagnosis by Xpert MTB/RIF was associated with a significantly increased odds of PLTFU (crude OR: 1.78, 95% CI: 1.29–2.47), which should be further investigated in future studies.

The DTUs located in urban settings need special attention in the sense that people being diagnosed at DTUs located in urban settings had a nearly 3-fold increased odds of not initiating treatment as compared with those who were diagnosed at DTUs located at rural areas (adjusted OR= 2.51, 95% CI: 1.51–4.17). One reason could be that some patients diagnosed at DTUs in urban areas may decide to travel to their localities of origin, usually at rural areas to be treated at DTUs closest to their family members for better family support. Another reason could be that those in urban settings may not want to be absent from work so as not to lose their daily wages. The rural/urban difference was also seen as a significant factor in a study conducted in Vietnam.

This study has some limitations. As a retrospective cohort study, it lacked certain key information about the patients such as marital status, income level, level of education, knowledge about TB, and medical co-morbidities. Although we collected information about referral status (i.e. those who were sent to initiate treatment at other centres) among those who did not initiate treatment, we were not able to conduct an active follow-up to ascertain the true outcome of those who had not initiated treatment as well as the reasons for PLTFU. Lastly, the geolocation of patient’s residence was estimated using an online geodatabase based on self-reported address, instead of onsite measurement.