Predictors of and Mortality from Resistant Infections in ICU Patients

Author(s):
Laura Stearns; Robert Sawyer

Background:

Resistant infections are a source of healthcare utilization and cost, and are especially problematic in the intensive care unit (ICU). Specific risk factors for these infections are still unclear.

Hypothesis:

We hypothesized that the risk factors for resistant Gram-negative rods (rGNR), resistant Gram-positive cocci (rGPC), and fungal infections (inherently resistant to empiric antibacterial therapy) would be similar in a cohort of ICU-acquired infections.

Methods:

Data were collected on patients requiring intensive care from 1997 to 2017 in a single university surgical-trauma ICU. Patients with ICU-acquired infections were analyzed, conditioned on the presence of rGNR, rGPC, or fungi.  Continuous variables were compared using Student’s t-test and categorical variables were compared using the chi-square test.  Independent predictors of the presence of a resistant pathogen and mortality were determined by logistic regression analysis.

Results:

4319 ICU-acquired infections were identified; 1998 were considered resistant and 2321 were considered non-resistant.  Identification of any resistant organism was significantly associated with female sex, non-trauma diagnosis, APACHE II score, liver disease, steroid use, history of any prior infection, and history of a resistant infection, but not days of prior antibiotic use (all p ≤ 0.02, C = 0.72, H-L test = 0.001).  Infections with rGNR were associated with days of therapeutic antimicrobials given for a previous infection but not total prior antimicrobial days during hospitalization, rGPC infections were associated with both days of therapeutic antimicrobials given for a previous infection and total prior antimicrobial days during hospitalization, and fungal infections were not associated with any measure of prior antimicrobial exposure.  Controlling for severity of illness and demographics, resistant infections were not associated with mortality compared to non-resistant infections (OR = 1.12, p = 0.17, C = 0.74, H-L test = 0.08).

Conclusions:

Demographic and treatment risk factors for resistance to antibacterial agents vary by resistant pathogen class.  The likelihood of rGNR infection appears to be most closely linked to recent antimicrobial exposure, while rGPC infection appears to be associated with totality of prior antimicrobial exposure.  Fungal infections may not be associated with prior exposure.  These findings suggest disparate mechanisms of dysbiosis for different classes of resistant pathogens.