Surveillance of hospital-acquired infections (HAIs) is the foundation of infection control. Machine learning (ML) has been demonstrated to be a valuable tool for HAI surveillance. We compared manual surveillance with a supervised, semiautomated, ML method, and we explored the types of infection and features of importance depicted by the model.

From July 2021 to December 2021, a semiautomated surveillance method based on the ML random forest algorithm, was implemented in a Brazilian hospital. Inpatient records were independently manually searched by the local team, and a panel of independent experts reviewed the ML semiautomated results for confirmation of HAI.

Among 6,296 patients, manual surveillance classified 183 HAI cases (2.9%), and a semiautomated method found 299 HAI cases (4.7%). The semiautomated method added 77 respiratory infections, which comprised 93.9% of the additional HAIs. The ML model considered 447 features for HAI classification. Among them, 148 features (33.1%) were related to infection signs and symptoms; 101 (22.6%) were related to patient severity status, 51 features (11.4%) were related to bacterial laboratory results; 40 features (8.9%) were related to invasive procedures; 34 (7.6%) were related to antibiotic use; and 31 features (6.9%) were related to patient comorbidities. Among these 447 features, 229 (51.2%) were similar to those proposed by NHSN as criteria for HAI classification.

The ML algorithm, which included most NHSN criteria and >200 features, augmented the human capacity for HAI classification. Well-documented algorithm performances may facilitate the incorporation of AI tools in clinical or epidemiological practice and overcome the drawbacks of traditional HAI surveillance.

To evaluate the impact of ertapenem use in Pseudomonas aeruginosa carbapenem resistance, taking into account the volume of antimicrobial consumption, the consumption by the entire hospital of alcohol-based antiseptic hand rub, and the density rate of invasive practices.

Before-and-after trial.

A tertiary care university hospital in southern Brazil.

Ertapenem was first added to the hospital formulary in June 2006, and it was excluded in February 2009. We evaluated Pseudomonas aeruginosa resistance rates through 3 study periods: period 1, before ertapenem use (17 months); period 2, during ertapenem use (33 months); and period 3, after exclusion of ertapenem (15 months).

After introduction of ertapenem, there was a significant decrease in median consumption of imipenem or meropenem, from 2.6 to 2.2 defined daily doses (DDDs) per 100 patient-days (level change from 0.04 to -1.08; P < .01), and an increase in the use of these medications after ertapenem exclusion, from 2.2 to 3.3 DDDs per 100 patient-days (level change from -0.14 to 0.91; P < .01), by segmented regression analysis. There was no difference in the incidence density of carbapenem-resistant P. aeruginosa infection related to ertapenem use throughout the study periods. However, by multiple regression analysis, the reduction in the rate of carbapenem-resistant P. aeruginosa infection correlated significantly with the increase in the volume of alcohol used as hand sanitizer, which was from 660.7 mL per 100 patient-days in period 1 to 2,955.1 mL per 100 patient-days in period 3 (P = .04). Ertapenem use did not impact the rate of carbapenem-resistant P. aeruginosa infection.

Use of alcohol-based hand gel, rather than ertapenem, was associated with a reduction in the rates of carbapenem-resistant P. aeruginosa infection. Measures to reduce resistance must include factors other than just antimicrobial stewardship programs alone.