PROJECT ABSTRACT
Healthcare-associated infections (HAIs) are a leading cause of preventable harm in U.S. hospitals. Hospitals
are therefore required to report selected HAIs, including central line-associated bloodstream infections,
catheter-associated urinary tract infections, colon and abdominal hysterectomy surgical site infections,
methicillin-resistant Staphylococcus aureus bacteremia, and Clostridioides difficile infections, to the Centers for
Medicare & Medicaid Services (CMS) via CDC’s National Healthcare Safety Network. These data are used to
benchmark hospitals and to inform CMS’s pay-for performance programs. In many cases, however, conducting
surveillance for HAIs is complicated, resource intensive, and prone to high levels of interobserver variability.
Furthermore, many of the most common and morbid HAIs, such as hospital-acquired pneumonia and most
non-line associated bloodstream infections, are not routinely tracked or reported and are therefore largely
neglected by hospitals. Our underlying hypothesis is that electronic surveillance for hospital-onset sepsis
could provide a more complete, efficient, and objective method to identify a fuller array of the most serious
HAIs compared to current surveillance methods. In particular, CDC’s hospital-onset Adult Sepsis Event
definition uses routine electronic clinical data to identify patients with concurrent clinical indicators of presumed
serious infections (blood culture orders and antibiotic treatment) and concurrent organ dysfunction (initiation of
vasopressors or mechanical ventilation or significant changes in laboratory values). Preliminary data suggest
that CDC’s hospital-onset sepsis surveillance definition may identify many more HAIs than current CMS
reportable metrics and that hospital-onset sepsis is associated with very high mortality rates even when
reportable HAIs are absent. Before hospitals and policymakers consider routinely tracking hospital-onset
sepsis, however, data from diverse settings are needed to understand its incidence and variation across
hospitals, its impact on long-term outcomes, and the specific infections leading to sepsis. This proposal will
address these gaps through the following Specific Aims: 1) Characterize the incidence, in-hospital mortality,
and hospital-level variation for hospital-onset sepsis versus currently reportable HAIs, 2) Determine the impact
of hospital-onset sepsis versus reportable HAIs on patients’ long-term outcomes, and 3) Develop electronic
algorithms using clinical and administrative data to automatically identify the types of infections precipitating
hospital-onset sepsis. This work will be conducted using detailed clinical data from 145 HCA Healthcare
hospitals that collectively care for more than 1.7 million inpatients per year (approximately 5% of U.S. acute
care hospitalizations). In Aim 2, we will also link HCA data to CMS claims data to accurately capture post-
discharge outcomes for Medicare beneficiaries. This proposal directly addresses the goals of AHRQ and its
HAI Prevention Portfolio by providing the foundation for a new HAI surveillance paradigm that will help identify
new targets for prevention and quality improvement and thereby help catalyze better outcomes for patients.
