Cross-Validation of Nebraska's Inpatient Syndromic Surveillance System for Chronic and Infectious Diseases

Tuesday, June 24, 2014: 2:30 PM
108, Nashville Convention Center
Sandra Gonzalez , Nebraska Department of Health and Human Services, Lincoln, NE
Ming Qu , Nebraska Department of Health and Human Services, Lincoln, NE
Thomas Safranek , Nebraska Department of Health and Human Services, Lincoln, NE

BACKGROUND: Surveillance is a critical component of national efforts for prevention of adverse health events.  The Nebraska Department of Health and Human Services (NDHHS) has traditionally defined the burden of such events by retrospectively analyzing hospital discharge data (HDD). However, these data are limited by the lack of immediate availability and the limited number of data elements present.  Timeliness of data analysis and event detection could potentially be improved by using syndromic surveillance data.  Although traditionally used for detection of communicable diseases, studies have demonstrated the use of near-real time syndromic surveillance data for monitoring and analysis of non-communicable diseases. Accordingly, NDHHS has expanded surveillance by implementing near-real–time inpatient syndromic surveillance (IPSS). ICD9-CM diagnostic codes are reported for all records reported to NDHHS IPSS, along with other demographic and clinical information. Therefore, we hypothesized that IPSS data might be consistent with HDD. Our objective was to evaluate NDHHS IPSS data quality by cross-validating the reported primary diagnostic codes.   

METHODS: Data quality was assessed by comparing 2012 NDHHS IPSS and HDD 2012 data from Hospital A.  Cross-validation focused on 4 chronic diseases which represent leading causes of death and disability in the US (myocardial infarction, stroke, diabetes, and osteoarthritis); 4 types of cancer with highest incidence rates in Nebraska (lung or bronchus, colon or rectum, breast, and prostate); and influenza. Corresponding ICD9-CM primary diagnostic codes are as follows: 410 (myocardial infarction), 433-434 and 436 (ischemic stroke), 430-432 (hemorrhagic stroke), 250 (diabetes), 715 (osteoarthritis), 162 (lung or bronchus cancer), 153-154 (colon or rectum cancer), 174 (breast cancer), 185 (prostate cancer), and 487-488 (influenza). Monthly counts of these specified IPSS and HDD diagnosis code frequency distributions were compared using Pearson correlation coefficients.

RESULTS: Completeness for NDHHS IPSS and HDD primary diagnosis codes was 100%. Correlations were 0.97(p<0.001) for myocardial infarction, 0.94(p<0.001) for ischemic stroke, 0.89(p<0.001) for hemorrhagic stroke, 0.96(p<0.001) for diabetes, 0.95(p<0.001) for osteoarthritis, 0.48(p=0.118) for lung or bronchus cancer, 0.90 (p<0.001) for colon or rectum cancer, 0.73(p=0.007) for breast cancer, 0.87(p<0.001) for prostate cancer, and 0.98(p<0.001) for influenza.

CONCLUSIONS: Significantly high correlations (≥ 0.89) were observed between 2012 Hospital A NDHHS IPSS and HDD for 80% (8/10) of conditions analyzed. Results of this study suggest that near-real–time NDHHS IPSS data hold promise to enhance timeliness of health event surveillance. By facilitating near-real–time monitoring of morbidity and mortality trends, NDHHS IPSS might aid more timely identification of at-risk populations and corresponding guidance for prevention strategies.