1065. Accuracy of Administrative Code Data for Surgical Site Infections: A Meta-Analysis
Session: Poster Abstract Session: Surveillance of HAIs: Evaluating National Strategy
Friday, October 4, 2013
Room: The Moscone Center: Poster Hall C
  • SSI poster upload.png (339.7 kB)
  • Background: Administrative code data (ACD), such as ICD-9CM, are widely used for research and quality-assessment purposes.  The National Healthcare Safety Network (NHSN) currently requires manual data abstraction for surveillance of surgical site infection (SSI): ACD are a potentially attractive tool to automate this time-consuming process.   Our aim is to summarize the evidence for accuracy of ACD for the diagnosis of SSI, compared to the current surveillance methods.

    Methods: Data Sources: Electronic literature search of PubMed, EMBASE, Web of Science, and CINAHL (from 1/1/1979 to 12/31/2012). Study Selection: Observational studies that: (1) Reported accuracy of ACD to detect SSI; (2) Used ICD-9CM; and (3) Used NHSN SSI surveillance criteria (or equivalent) as the reference standard. Data Synthesis: We screened 1548 published studies and included nine in our final analysis.  Two reviewers independently extracted information on study design, populations, ACD, reference standard method, and accuracy of diagnosis.  Statistical calculations were performed by R version 3.0 with mada package.

    Results: There was uniformly high specificity (0.90-1.00), but sensitivity varied widely from 0.21 to 0.98, with pooled sensitivity 0.807 (95CI: 0.59-0.92) and pooled specificity 0.97 (95%CI: 0.94-0.99).  Positive predictive values (PPV) ranged from 0.11 to 0.77.   Three studies only used ICD-9 code 998.5 (postoperative infection).   Sensitivity was somewhat higher in six studies that included ICD-9 codes 996.6 (infection and inflammatory reaction due to internal prosthetic device, implant, and graft) and 998.5, sensitivity ranged from 0.61 to 0.979 with pooled sensitivity 0.87 (95%CI: 0.72-0.95) and pooled specificity 0.98 (95%CI: 0.94-0.99).   Four studies evaluated SSI in orthopedic procedures, with pooled sensitivity 0.89 (95%CI: 0.66-0.97) and pooled specificity 0.98 (95%CI: 0.90-0.99).  There was wide heterogeneity among studies in settings and usage of ACD.

    Conclusion: We estimate ACD has moderate sensitivity and high specificity for SSI, when at least both of ICD-9 codes 996.6 and 998.5 are included.   ACD may perform better for orthopedic procedures.  However, PPV can still be low, considering low incidence of SSI.   These results suggest that ACD may be incorporated as a part of algorithmic automated surveillance, but not as a primary screening tool.

    Michihiko Goto, MD1,2, Michael Ohl, MD, MSPH1,3, Eli Perencevich, MD, MS, FIDSA, FSHEA3,4 and Marin Schweizer, PhD3,5, (1)Division of Infectious Diseases, Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City, IA, (2)VA National Quality Scholars Program, Iowa City VA Health Care System, Iowa City, IA, (3)Center for Comprehensive Access and Delivery Research and Evaluation (CADRE), Iowa City VA Health Care System, Iowa City, IA, (4)Department of Internal Medicine, University of Iowa Carver College of Medicine, Iowa City, IA, (5)Internal Medicine, University of Iowa Carver College of Medicine, Iowa City, IA


    M. Goto, None

    M. Ohl, None

    E. Perencevich, None

    M. Schweizer, None

    Findings in the abstracts are embargoed until 12:01 a.m. PST, Oct. 2nd with the exception of research findings presented at the IDWeek press conferences.