Individuals with incidentally detected lung nodules can benefit from automated detection methods, according to an analysis involving nearly 500 individuals. Moreover, such approaches can be implemented across multiple health systems.

Investigators reviewed charts at 2 institutions to determine if a lung nodule was documented. They used a natural language processing (NLP) algorithm using only knowledge of the key words, qualifiers, excluding terms, and the logic linking these parameters to determine the NLP’s detection ability. Among the results:

• 3 in every 10 patients had an incidentally detected lung nodule.
• NLP identified 4 in every 10 with such a nodule.
• NLP sensitivity was similar at both institutions (90% and 96%); specificity was identical (86%).