SAN FRANCISCO — An analysis of records from nearly 200 hospitals showed that variations in the completeness and accuracy of claims data makes it difficult to compare sepsis rates and outcomes, according to findings presented at IDWeek.
Researchers said meaningful comparisons may require the use of objective clinical data to facilitate improved sepsis surveillance.
“Sepsis is a major cause of death in U.S. hospitals, yet timely and effective sepsis care can reduce the risk of death,” Chanu Rhee, MD, MPH, assistant professor of population medicine at Harvard Medical School and Harvard Pilgrim Health Care Institute, said during a presentation.
Previously, Infectious Disease News spoke with Konrad Reinhart, MD, chair of the Global Sepsis Alliance, about the global rise of sepsis. Although he said there have been improvements in coding standards in the last 5 years, before that “the medical system was not doing a good job of accounting for cases of sepsis.”
Rhee and colleagues found that the reliance on claims data may be hindering sepsis surveillance, research and quality improvement. Likewise, Rhee said variations in hospital diagnosis, documentation and coding practices may make it difficult to benchmark hospital sepsis outcomes using claims data.
“Administrative claims data have important limitations,” Rhee said. “We know they have low-to-moderate sensitivity when identifying sepsis and, more importantly, recent analyses have suggested that claims-based trends are biased by changing diagnosis and coding practices over time.”
Rhee and colleagues used the electronic health records of 193 hospitals in the United States — which included data on 4.3 million adult hospitalizations in 2013 or 2014 — to evaluate the sensitivity of claims data for sepsis and organ dysfunction. They defined clinical cases of sepsis using presumed infection and concurrent organ dysfunction as markers and tracked ICD-9-CM codes for severe sepsis or septic shock to determine sepsis incidence and mortality via hospital claims data.
According to Rhee and colleagues, hospitals’ claims data for sepsis and organ dysfunction exhibited low and variable sensitivity. For sepsis, the median sensitivity was 30%. According to Rhee’s presentation, median sensitivity for both acute kidney injury and shock was 66%. The median sensitivity for thrombocytopenia and hepatic injury was 39% and 36%, respectively.
The researchers observed only a moderate correlation between claims and clinical data for sepsis incidence and mortality rates and a substantial difference in the relative hospital rankings for sepsis mortality. Furthermore, Rhee and colleagues said 46% of hospitals ranked as having the lowest rate of sepsis mortality using claims data saw an increase in mortality rates when clinical data was used.
Rhee explained that varying claims data between hospitals limits its use when comparing sepsis rates and outcomes.
“I would be the first to acknowledge that there is no true gold standard for sepsis,” Rhee said. “However, the EHR clinical criteria, I believe, are more objective and consistent.” – Marley Ghizzone
Rhee C, et al. Abstract 1659. Presented at: IDWeek; Oct. 3-7, 2018; San Francisco.
Disclosures: The authors report no relevant financial disclosures.