PURPOSE: To evaluate the value of ICD-9-CM code for identifying celiac disease (CD). METHODS: We searched administrative data to identify all adults with ICD-9-CM diagnosis code 579.0 (CD) at three teaching hospitals between 2000 and 2010. We then stratified patients according to the presence/absence of relevant serology and endoscopy codes into four groups: None, serology, endoscopy, and both. A diagnostic algorithm was applied to define CD status. RESULTS: Through random sampling and appropriate weighting, the 1200 reviewed patients represented a cohort of 8,122 cases. The overall positive predictive value (PPV) of the ICD-9-CM code was 15% (95% confidence interval [CI], 13%-17%). Case identification by a diagnosis code alone had a PPV of 4%, whereas the group with diagnosis code plus both serology and endoscopy testing had a PPV of 49%. Independent predictors of CD were non-Hispanic white, ICD-9-CM-coded patient group, total number of a diagnosis code, and receiving a diagnosis code by a gastroenterologist. The model had an area under the curve of 0.87 (95% CI, 0.84-0.89). CONCLUSIONS: The performance of ICD-9-CM 579.0 alone for identifying CD is extremely poor. Adding other readily available administrative data significantly improves CD case identification. The proposed case finding strategy via administrative databases may facilitate future research on CD.
PURPOSE: To evaluate the value of ICD-9-CM code for identifying celiac disease (CD). METHODS: We searched administrative data to identify all adults with ICD-9-CM diagnosis code 579.0 (CD) at three teaching hospitals between 2000 and 2010. We then stratified patients according to the presence/absence of relevant serology and endoscopy codes into four groups: None, serology, endoscopy, and both. A diagnostic algorithm was applied to define CD status. RESULTS: Through random sampling and appropriate weighting, the 1200 reviewed patients represented a cohort of 8,122 cases. The overall positive predictive value (PPV) of the ICD-9-CM code was 15% (95% confidence interval [CI], 13%-17%). Case identification by a diagnosis code alone had a PPV of 4%, whereas the group with diagnosis code plus both serology and endoscopy testing had a PPV of 49%. Independent predictors of CD were non-Hispanic white, ICD-9-CM-coded patient group, total number of a diagnosis code, and receiving a diagnosis code by a gastroenterologist. The model had an area under the curve of 0.87 (95% CI, 0.84-0.89). CONCLUSIONS: The performance of ICD-9-CM 579.0 alone for identifying CD is extremely poor. Adding other readily available administrative data significantly improves CD case identification. The proposed case finding strategy via administrative databases may facilitate future research on CD.
Authors: Elizabeth T Jensen; Suzanne F Cook; Jeffery K Allen; John Logie; Maurice Alan Brookhart; Michael D Kappelman; Evan S Dellon Journal: Ann Epidemiol Date: 2015-03-21 Impact factor: 3.797
Authors: Kathryn Peterson; Rafael Firszt; John Fang; Jathine Wong; Ken R Smith; Kristina A Brady Journal: Am J Gastroenterol Date: 2016-05-24 Impact factor: 10.864
Authors: Roby Abraham; John J Kelly; Jared M Newman; Qais Naziri; Nipun Sodhi; Anton Khlopas; Jaiben George; Neil V Shah; Assem A Sultan; Morad Chughtai; John W Barrington; Carl B Paulino; Michael A Mont Journal: Ann Transl Med Date: 2017-12
Authors: You Chen; Nancy M Lorenzi; Warren S Sandberg; Kelly Wolgast; Bradley A Malin Journal: J Am Med Inform Assoc Date: 2017-04-01 Impact factor: 4.497
Authors: Emilie K Johnson; Sarabeth Broder-Fingert; Pornthep Tanpowpong; Jonathan Bickel; Jenifer R Lightdale; Caleb P Nelson Journal: BMC Med Res Methodol Date: 2014-01-30 Impact factor: 4.615