(251) Dynamic containment strategy in critical ill patients with high risk of antibiotic drug resistance, a MIMIC-IV study

Background: Drug resistance results in high mortality in critical ill patients. Since there were very limited choice for certain pathogen, simply stop using those resisted antibiotics might not be the optimal strategy. Recent cell and stimulation studies found out that containment strategy would benefit patients by dynamically control pathogen density, which might start and stop treatment multiple times. This study is about to test this by using dynamic treatment regime model to see if optimal strategy would different from clinical practice.

Objectives: This study is to investigate if the adaptive strategy would be the optimal one in resisted antibiotics.

Methods:

Design: we used dynamic treatment strategy model to find out the optimal usage for resisted antibiotics.
Setting: we used MIMIC-IV database to establish our study. In total, 2,532 ICU admission (41,565 person-days) who administrated known intermediate/severe resisted antibiotics were included in this study, the index date was set as the day started the resisted antibiotics.
Exposures: the resisted antibiotics administration for each day in a week
Outcome: 28-day survival
Statistical analysis: We conducted dynamic weighted survival (DWSurv) modelling for analysis. We included age, gender, Charlson-index, SOFA and APSiii score as covariates and vital signs as dynamic variables (blip function) into model

Results: In clinical practice the resisted antibiotics administrated as a negative exponential growth manner (~20% off per day) for 87%, 69%, 53%, 42%, 36%, 31% of patients from day 2 to day 7, respectively. On the contrary, the model recommends a leap in a week, the optimal assignment was for 34%, 75%, 77%, 71%, 82%, 70% of patients from day 2 to day 7, respectively. The risk ratio of daily death for model recommendation vs clinical practice in 2-7 days were: 0.25, 0.67, 0.78, 1.08, 1.10, 0.89. A clear gap between clinical practice and optimal treatment strategy was observed.

Conclusions: Our study found out that adaptive treatment strategy in resisted antibiotic administration might have larger benefit than clinical practice in critical ill patients.