4.1 Retrospective study
Table 2 shows a total of 1174 cases with more males (56.2%) than females. About 82%
of patients attended care at urban DTUs and 41% of total patients were tested by GeneXpert. More than 30% of patients traveled for more than 30 minutes and over 30 km to get to the DTUs. The study population mean age was 37.6 years (SD ±13.2) with a PLTFU proportion of 16.7%, Table 3. After excluding those who were known to have been referred to other centers, the proportion of PLTFU dropped from 16.7% to 10.0%.
Figure 5 shows most of the patients (36.2%) initiated TB treatment on the second day following diagnosis.
In multivariable logistic regression, DTUs located in urban setting had a nearly 3-fold increase in odds of PLTFU compared to DTUs in the rural setting, aOR: 2.51, 95% CI:
1.51–4.17. A travel time of more than 30 minutes from home to DTU versus ≤30.0 minutes showed an increase in the odds of PLTFU, aOR: 2.19, 95% CI: 1.56–3.09; and travel distance from home to DTU (>30 km versus ≤30.0 km, aOR: 2.31, 95% CI: 1.63–
3.27). (Table 4). In the non-linear spline regression analysis, there is a steep increase in the odds for not initiating treatment immediately after diagnosis as patient’s distance away from DTU increases (Figure 7).
In the multivariable Cox regression analysis, the DTUs in urban setting had a 28%
decrease in initiating treatment compared with those in the rural setting, aHR: 0.72, 95%
CI: 0.60–0.86. However, private health facilities had a 7% increase in initiating treatment compared to public facilities, aHR: 1.07, 95% CI: 0.93–1.24. In the multivariable Cox regression, significant determinants of time to treatment included location of DTUs ([urban versus rural], adjusted hazards ratio (aHR): 0.72, 95% CI:
0.60–0.86); travel time from home to DTUs (>30 minutes versus ≤30.0 minutes, aHR:
0.80, 95% CI: 0.69–0.92); and travel distance from home to DTU (>30 km versus ≤30.0 km, aHR: 0.76, 95% CI: 0.66–0.89). (Table 5).
4.2 Prospective study
A total of 1060 bacteriologically-confirmed pulmonary tuberculosis cases were recorded with a mean age (M = 39.3, SD ±11.9) years. About 60% of patients were males. More than 70% of patients utilized services at DTUs located at urban settings. A slightly equal proportion of patients visited both health care facilities, public (51%) and private (49%). About 60% of patients were diagnosed by smear microscopy and sixty percent of total patients were married. About 10% of total patients never had any form of formal education. The proportion of those employed in this study was 82.2% with just 23.6% making more than 60,000 CFA frs (USD120) as monthly income.
Approximately 20% of total patients spent more than 6 hours to-and-from the DTU for care. Majority of the patients (71.1%) had recently done their HIV test within the last one year. HIV infection was a major co-morbid condition among these patients, with a 78.1% proportion among those with co-morbidities. See Table 6.
There was a PLTFU rate of 10.6% (112 out of 1060 cases); 95% CI: 8.7–12.4). In the multivariable logistic regression, the South West Region had a 90% increase in odds of PLTFU compared to the North West Region (aOR: 1.90, 1.27–2.85). Patients with a monthly salary of ≥60,000 CFA frs had a significantly lower odds of PLTFU compared with those of a lower salary, aOR: 0.54, 95% CI: 0.32–0.92. Those who tested for HIV
≥1y ago had an increased risk for not initiating treatment compared with those who did their test less than a year, aOR: 2.05, 95% CI: 1.36–3.09. See Tables 7 and 8.
The median time from first positive result to initiating therapy was 2 (IQR: 1–3) days.
In the Cox regression analysis on time-to-treatment, starting treatment was significantly
24
lower (15%) in the South West region than in the North West, aHR: 0.85, 95% CI:
0.74–0.97. The private health facility initiated treatment 50% quicker than in the public facility aHR: 1.50, 95% CI: 1.32–1.71, See Table 9. Among those who did not return for treatment by day 7, 50% claimed in the phone interview that they never had money to return to hospital to initiate treatment among other reasons for PLTFU, with 12.2% of PLTFU had no willingness of returning for care. See Table 10.
4.3 Impact evaluation of patient counselling and phone reminder study The analyses of impact evaluation included 2160 bacteriologically confirmed pulmonary tuberculosis cases with a mean age of 39.2 years (SD ±12.0). A total of 57.8% (1249 out of 2160) were males. The proportion of those diagnosed by GeneXpert (37.9%) was lower compared to those diagnosed by microscopy. See Table 11.
The proportion of PLTFU during the total study period (Jul 2015 to August 2016, except for the break in Jan 2016) was 14.1%; 95% CI: 12.6–15.4. The baseline or control group had a PLTFU of 17.5% while the intervention group’s PLTFU was 10.6%. Just to note that the baseline PLTFU of 17.5% is different from the previous one of the original retrospective study (16.7%) because certain patients were excluded due to the age cut-off of ≥21 years as inclusion criterion for the impact evaluation study. In the logistic regression analysis, there was a statistically significant decrease in PLTFU in the intervention group both in the univariable analysis (crude OR: 0.56, 95% CI:
0.44–0.72) and the multivariable analysis (aOR: 0.61, 95% CI: 0.47–0.79) (Tables 12).
There was also a 79% increase in odds of PLTFU among those diagnosed by GeneXpert compared with those diagnosed by microscopy, (aOR: 1.79, 95% CI: 1.22–2.63). Also statistically significant were travel time and travel distance. A travel time of more than 30 minutes from home to DTU versus ≤30.0 minutes showed an increase in the odds of
PLTFU, aOR: 1.81, 95% CI: 1.38–2.38. And travel distance of more than 30 km from home to DTU versus ≤30 km showed an increase in the odds of PLTFU, aOR: 1.81, 95% CI: 1.37–2.38.
Table 13 shows an approximately 30% significant increase in initiating treatment of the intervention group compared with the control, both in the univariable (crude HR: 1.31, 95% 1.20–1.44) and multivariable (aHR: 1.30, 95% CI: 1.18–1.42) Cox regression analyses. The South West Region had a lower rate of initiating treatment than the North West Region (aHR: 0.85, 95% CI: 0.75–0.95). Getting on treatment was quicker in the private health facilities compared to the public (aHR: 1.31, 95% CI: 1.18–1.45).
Initiating treatment was faster with those who traveled less than 30 minutes to the DTU compared to those who travelled more than that, (aHR: 0.86, 95% CI: 0.78–0.95), see Table 13.
The proportion of PLTFU during the whole study period decreased over time (Figure 8). The proportion appeared lower in the intervention period compared to the historical control period (Figure 9), consistent with a significant association between the
intervention and PLTFU (Model 1, crude OR: 0.56, 95% CI: 0.44–0.72). In the logistic regression model with the addition of a linear term of time trend and an interaction term between the intervention and time (Model 2), none of the associations were statistically significant: the adjusted odds ratio of the intervention was 0.71 (95% CI: 0.25–2.01);
the adjusted odds ratio of the linear time trend was 0.999 (95% CI: 0.996–1.002), and the adjusted odds ratio of the interaction between the intervention and time trend was 1.0001; 95% CI: 0.996–1.005. See Figure 10.
26