Validated methods for identifying tuberculosis patients in health administrative databases: systematic review.

TLDR: To summarise validated methods for identifying TB in health administrative databases, a systematic literature search in two databases found diagnostic accuracy measures of algorithms using out-patient, in-patient and/or laboratory data vary widely across studies.

Abstract: Background An increasing number of studies are using health administrative databases for tuberculosis (TB) research. However, there are limitations to using such databases for identifying patients with TB. Objective To summarise validated methods for identifying TB in health administrative databases. Methods We conducted a systematic literature search in two databases (Ovid Medline and Embase, January 1980-January 2016). We limited the search to diagnostic accuracy studies assessing algorithms derived from drug prescription, International Classification of Diseases (ICD) diagnostic code and/or laboratory data for identifying patients with TB in health administrative databases. Results The search identified 2413 unique citations. Of the 40 full-text articles reviewed, we included 14 in our review. Algorithms and diagnostic accuracy outcomes to identify TB varied widely across studies, with positive predictive value ranging from 1.3% to 100% and sensitivity ranging from 20% to 100%. Conclusions Diagnostic accuracy measures of algorithms using out-patient, in-patient and/or laboratory data to identify patients with TB in health administrative databases vary widely across studies. Use solely of ICD diagnostic codes to identify TB, particularly when using out-patient records, is likely to lead to incorrect estimates of case numbers, given the current limitations of ICD systems in coding TB.