Disagreement on estimating cause of death under tuberculosis treatment between the national tuberculosis registry and statistics data in South Korea, 2011-2020

Abstract

Background: Disagreement persists regarding the cause of death in patients with tuberculosis (TB) between the national TB registry and vital registration statistics. This study investigated the disagreement and contributing factors between TB-related and non-TB-related deaths using an integrated national TB database in South Korea.

Methods: We identified a sub-set cohort of 29,033 patients with drug-susceptible TB registered between 2011 and 2020 who died during TB treatment. The cause of death was identified by the Korean National Tuberculosis Surveillance System (KNTSS) and Statistics Korea database. We performed a multinomial logistic regression to identify factors associated with the cause of death.

Results: The participants' median age was 79 years (interquartile range, 70-85 years), with males comprising 63.2 %. Among the 29,033 deaths, 12,937 (44.6 %) and 6028 (20.8 %) were classified as TB-related deaths in Statistics Korea and KNTSS, respectively. The overall agreement rate for the cause of death between the two databases was 0.72 (95 % confidence interval, 0.71-0.72), increasing from 0.69 in 2011-0.77 in 2020. Among discrepant cases, 92.2 % (7545/8181) were classified as non-TB-related in KNTSS but as TB-related in Statistics Korea. Over the study period, the proportion of individuals classified as non-TB-related deaths in both databases and the agreement rate increased. In the multinomial analysis, age, sputum acid-fast bacilli smear, Charlson comorbidity index, and comorbidities, such as cancer and end-stage renal disease, showed distinguishing features across the cause of death groups.

Conclusions: Substantial disagreement on the cause of death persists between the national TB registry and vital registration statistics, though this decreased during the study period. Most disagreement likely indicates an underestimation of TB-related deaths in the TB registry. Age, sputum smear, and comorbidities were identifiable characteristics across groups that may influence the cause of death coding process.