Overall, 2,852 DR-TB adult patients, HIV+ on ART, initiated TB treatment between January 2012-December 2013 based on the DR-TB register data, and of these, 1,685 (59%) could be matched to the ART registers. Of the 1,526 patients who were reportedly not on ART (or the data was missing), 253 patients were found in the ART registers, having initiated ART. This brought our total matched up to 1,938 (62%) and a total of 3,105 DR-TB patients who were included in the subsequent analyses. While the rate of matching was high in Sites 1 and 2, with 76% and 80%, respectively, we could only match 59% of patients at Site 3. There were no substantial differences in age, gender, new treatment status, extra-pulmonary TB or TB outcome between those matched compared to those we couldn’t match to the ART register (Table 1), though the number of characteristics we could assess was limited.
In order to validate our matching algorithm, we randomly sampled 171 patients and compared the algorithm to our “gold standard” of manual matching. We found our algorithm to have a sensitivity of 87%, a specificity of 99%, a PPV of 98% and a NPV of 94%. The specificity was reduced by 1% and PPV by 2% due to an error in misclassification for one patient in the manual matching process.
Discrepancies between ART and DR-TB registers are presented in Table 2. Overall, 8.1% of patients had a discrepant date of birth, however the matching algorithm was more lenient on these if the name and surname were matched already; another 2.5% of patients had a date of birth missing in either register so these patients would have been matched on age, alongside gender and names. All patients in Site 1 had their HIV status recorded as HIV positive in the DR-TB register, but 6.2% and 4.5% of patients in Sites 2 and 3, respectively, had either a negative, unknown or missing HIV status in the DR-TB register, despite being initiated on ART. Overall, 8.7% of patients had no record of TB treatment (e.g. missing) in the ART register data and 2.5% of patients were recorded as not being on TB treatment (e.g. TB treatment incorrectly recorded as “No”) in the ART register record. Of the 217 patients whose TB treatment status was incorrectly recorded in the ART register, 82% had initiated ART prior to TB treatment and thus this is likely due to the records not being updated. However, of all patients who had initiated ART prior to TB treatment (61%, n=1,187), majority (85%) had TB treatment status recorded correctly, demonstrating that most records had been updated in real time.
Majority (87%) had ART status correctly recorded in the DR-TB register. While ART start date was reported in both ART and DR-TB registers for most patients at Site 1 (81.2%), 67.8% were discrepant, and three quarters of those discrepant differed by >1 month. At Site 1, 12% of patients had a missing ART start date in the ART register, though the data indicated they initiated treatment. Most patients at Sites 2 and 3 did not have ART start date recorded in the DR-TB register (99.7% and 80.5%, respectively). Of the 280 patients at Site 3 that did have ART start date recorded in both registers, 15.4% had discrepant dates with 68.7% of ART start dates differing by >1 month. Overall there was poor agreement of the reporting of the ART start date between the ART and DR-TB registers (Kappa=0.001). Sites 2 and 3 did not have ART regimen information recorded in the DR-TB register so we could not compare between the two datasets. At Site 1, where ART regimen was recorded in both DR-TB and ART registers for 44.3% of patients, it was in agreement for majority of them (86.4%). However, 11.4% of patients at Site 1 did not have ART regimen recorded in the ART register.
Overall, 7.3% of patients had a death reported in the DR-TB register but not the ART register. Of these 80.9% were reported in the ART register as a lost to follow-up, 14.9% as transferred out, and 2.8% of patients had no outcome reported. While these may be explained by the timing of when the information gets recorded, a large proportion of these deaths (45.4%) were reported as deaths in the DR-TB register less than 3 months after the non-death outcome in the ART register and 36.9% had dates which occurred prior to the non-death outcome date in the ART register.
Overall, 3.5% had a death reported in the ART register but not the DR-TB register. Of these, 56.7% were reported as lost to follow-up in the DR-TB register, 16.4% had no outcome reported, and one patient was reportedly transferred out. The remaining patients had TB treatment-specific outcomes reported: 11.9% cured, 9.0% completed treatment, 3.0% still on treatment and one patient had failed TB treatment. Of the 67 patients with death reported in the ART register and not the DR-TB register, 29.9% were reported as deaths in the ART register <3 months after the non-death outcome in the DR-TB register and 34.3% had their death dates prior to the non-death outcome date in the DR-TB register. Overall agreement between reporting of death was good between the DR-TB and ART registers (Kappa = 0.60).