This article is more than 5 years old. Information may no longer be current.
Coding data may be unreliable to estimate HAI incidence
Click Here to Manage Email Alerts
The use of administrative code data may not be accurate enough for surveillance and reporting of health care-associated infections, according to study results published in Clinical Infectious Diseases.
“We found the accuracy of administrative code data for surgical site infections and Clostridium difficile colitis is moderate at best, although the specificities were high,” Michihiko Goto, MD, fellow in the division of infectious diseases, at University of Iowa Carver College of Medicine, told Infectious Disease News. “This means we can potentially miss a substantial number of those infections, if we are heavily relying on ICD codes.”
Michihiko Goto
Goto and colleagues conducted a systematic review and meta-analysis to evaluate the sensitivity and specificity of administrative code data (ACD) to detect health care–associated infections (HAIs). The analysis included 19 studies of various HAIs, including catheter-associated urinary tract infection, central line–associated bloodstream infection, ventilator-associated pneumonia and other ventilator-associated events, C. difficile infection, methicillin-resistant Staphylococcus aureus and post-procedure pneumonia.
For C. difficile infections, there were seven studies available. The pooled analysis showed that ACD have moderate sensitivity (76%; 95% CI, 56.2-88.7) and high specificity (99.9%; 95% CI, 99.6-100). There were nine studies focused on surgical site infections. Data from the pooled analysis of these studies indicated that ACD have moderate sensitivity (80.7%; 95% CI, 59.1-92.3) and high specificity (97.1%; 95% CI, 93.9-98.7). For the remaining infections, there were limited studies.
The limitation of ACD is especially important when HAI surveillance is linked to hospital comparison or financial incentives, Goto said, because if those evaluations are conducted based on inaccurate information, high-performance hospitals can be erroneously published.
“I believe our findings will have an important impact in that they highlight the few HAIs that can be detected using administrative claims and the majority that can’t,” Goto said. “By continuing to use inaccurate ICD9 code-based measures, financial incentive programs for HAIs could potentially hinder the honest efforts of hospitals targeting the prevention of HAIs.”
Goto said that they were surprised that the accuracy data for some of the important HAIs are sparse, and that it is important to continue accumulating data, especially on conditions that have not been well-studied. In addition, ACD is constantly changing, especially with the introduction of ICD-10M in 2014, he said. Validation studies will need to be repeated with the new system.
Michihiko Goto, MD, can be reached at: Division of Infectious Diseases, Department of Internal Medicine, University of Iowa Carver College of Medicine, 200 Hawkins Dr., SW54-GH, Iowa City, IA 52242. Email: michihiko-goto@uiowa.edu.
Click Here to Manage Email Alerts