rss
J Am Med Inform Assoc 10:494-503 doi:10.1197/jamia.M1330
  • Original Investigation

Creating a Text Classifier to Detect Radiology Reports Describing Mediastinal Findings Associated with Inhalational Anthrax and Other Disorders

Table 2

Performance of IPS Models

Classifier Sensitivity Specificity PPV FNR
IPS Initial 0.043 (74/1,729) 1.0 (77,305/77,305) 1.0 (74/74) 0.021 (1−77,305/78,960)
IPS Refined 0.351 (607/1,729) 0.999 (77,283/77,305) 0.965 (607/629) 0.014 (1−77,283/78,405)
IPS-M1 0.712 (1,230/1,729) 0.985 (76,161/77,305) 0.518 (1,230/2,374) 0.007 (1−76,161/76,660)
IPS-M2 0.856 (1,480/1,729) 0.972 (75,164/77,305) 0.409 (1,480/3,621) 0.003 (1−75,164/75,413)
  • NOTE. IPS Initial: Based on UPMC key word search; widening (“wide”) and mediastinal (“mediastinum” and “mediastinal”) concepts.

  • IPS Refined: IPS Initial model refined by authors—outcome measures obtained by applying the most sensitive classification threshold.

  • IPS-M1: IPS Refined with four additional terms in the widening concept (“mass,” “soft tissue,” “prominent,” and “prominence”).

  • IPS-M2: IPS-M1 with five additional terms in the widening concept (“opacity,” “opacification,” “opacifications,” “density,” and “densities”).

This Article

Access policy for JAMIA

All content published in JAMIA is deposited with PubMed Central by the publisher with a 12 month embargo. Authors/funders may pay an Unlocked fee of $2,000 to make the article free on the JAMIA website and PMC immediately on publication.

All content older than 12 months is freely available on this website.

AMIA members can log in with their JAMIA user name (email address) and password or via the AMIA website.