Objective: To measure the accuracy of automated tuberculosis case dete
ction. Setting: An inner-city medical center. Intervention: An electro
nic medical record and a clinical event monitor with a natural languag
e processor were used to detect tuberculosis cases according to Center
s for Disease Control criteria. Measurement: Cases identified by the a
utomated system were compared to the local health department's tubercu
losis registry, and positive predictive value and sensitivity were cal
culated. Results: The best automated rule was based on tuberculosis cu
ltures; it had a sensitivity of .89 (95% CI .75-.96) and a positive pr
edictive value of .96 (.89-.99). All other rules had a positive predic
tive value less than .20. A rule based on chest radiographs had a sens
itivity of .41 (.26-.57) and a positive predictive value of .03 (.02-.
05), and a rule that represented the overall Centers for Disease Contr
ol criteria had a sensitivity of .91 (.78-.97) and a positive predicti
ve value of .15 (.12-.18). The culture-based rule was the most useful
rule for automated case reporting to the health department, and the ch
est radiograph-based rule was the most useful rule for improving tuber
culosis respiratory isolation compliance. Conclusions: Automated tuber
culosis case detection is feasible and useful, although the predictive
value of most of the clinical rules was low. The usefulness of an ind
ividual rule depends on the context in which it is used. The major cha
llenge facing automated detection is the availability and accuracy of
electronic clinical data.