Abstract
Context: Death certificates are routinely used to estimate tuberculosis (TB) mortality rates. The validity of International Classification of Diseases, Tenth Revision (ICD-10) codes and text cause of death data for this purpose is uncertain.
Objective: To evaluate the accuracy of ICD-10 coded and text cause of death data in identifying TB-related deaths in Washington State.
Design: Cross-sectional descriptive study comparing TB-related deaths detected through Washington State death certificates to TB-related deaths identified in the Washington State TB registry during 2009-2010.
Main Outcome Measure(s): Sensitivity and positive predictive value of ICD-10 coded and text cause of death definitions in identifying TB-related deaths compared to the TB registry.
Results: All methods for identifying TB-related deaths using death certificate data overestimated the number of TB-related deaths compared to the tuberculosis registry. The positive predictive value ranged from 22% for a TB ICD-10 code as an underlying or multiple cause of death to 56% for TB listed in the direct cause of death text field. Seventeen (33%) of 51 subjects assigned with a TB ICD-10 code as an underlying or multiple cause of death had no evidence of TB on the death certificate and were not present in the TB registry.
Conclusions: Death certificates were not highly predictive of TB-related deaths. Use of the direct cause of death text field was the most accurate method to identify a TB-related death when using death certificates. Specific ICD-10 coding algorithms may misclassify subjects as having died from TB.