Medicine · 20 Nov 2026
An interpretable AutoML framework for predicting carbapenem resistance and ICU outcomes in patients with Pseudomonas aeruginosa-positive cultures: A single-center retrospective cohort study
Xiaochun Lv, Qi Ren, Lihong Zhu, Kun Chen, Jianbing Wang, Fang Chen, Kailing Jin, Fanghua Fan, Kai Linh-index 1
Citations by database
Each database indexes a different set of journals, preprints and books, so they are shown side by side rather than blended.
- OpenCitations0
- Crossref0
- Europe PMC0
Citations per year
The current year is still counting.
- 2026 so far0
Abstract
Carbapenem-resistant Pseudomonas aeruginosa remains an important threat in intensive care units (ICUs), but few studies have jointly evaluated carbapenem resistance and conventional ICU outcomes using leakage-controlled validation. We conducted a retrospective single-center cohort study of 847 adults with P aeruginosa -positive cultures from 2018 to 2023. A fixed index time was defined at 48 hours after ICU admission, and predictors were restricted to information available before that time. Nine classifiers and 8 regressors were evaluated using patient-grouped, temporally blocked 5-fold outer cross-validation. Preprocessing and variance threshold filtering–mutual information–L1 regularization feature selection were fitted within each outer training fold. Fold-specific classification thresholds were derived from inner out-of-fold predictions to target a sensitivity of 0.80. Performance, calibration, and Shapley additive explanations analyses used held-out outer-fold predictions. Pairwise comparisons were adjusted using the Benjamini–Hochberg false-discovery-rate procedure. Resistance analyses included 846 patients. Logistic regression had the highest pooled out-of-fold receiver operating characteristic area under the curve for carbapenem resistance of the first post-index isolate (0.581; 95% confidence interval [CI] = 0.540–0.619). Random forest had the highest receiver operating characteristic area under the curve for resistance of the last post-index isolate (0.656; 95% CI = 0.614–0.696) and in-hospital mortality (0.856; 95% CI = 0.830–0.880). At the training-derived thresholds, sensitivity/specificity was 0.828/0.308, 0.794/0.409, and 0.785/0.753, respectively. After false-discovery-rate adjustment, the leading classifiers significantly outperformed 1, 6, and 2 of 8 comparators, respectively. Stacking regression had the lowest root mean squared error (RMSE) for ICU length of stay (46.126 days; 95% CI = 39.598–52.583; R 2 = 0.711), whereas bagging regression had the lowest RMSE for total hospitalization expenses (38.949 units of 10,000 Chinese yuan; 95% CI = 33.646–44.058; R 2 = 0.788). No pairwise difference in RMSE or mean absolute error among regressors remained significant after adjustment. Held-out Shapley additive explanations analyses identified endpoint-specific associations involving antimicrobial exposure, inflammatory markers, and host characteristics. Leakage-controlled internal validation showed high discrimination for in-hospital mortality, limited-to-modest discrimination for the resistance outcomes, and substantial internally estimated explained variance for the continuous outcomes. No algorithm was uniformly superior. External multicenter validation and prospective evaluation are required before clinical use.
What the data shows
- No database records a citation yet, as is usual for a paper this new.
- Kai Lin has the highest h-index among the authors (1).
Graph
This paper in the middle, its authors and their most-cited papers below, the papers it cites to the left and the papers citing it to the right. The timeline lays every paper out by year.
Select a node for its details. Drag to pan, scroll to zoom, drag a node to move it.
Start with
Cited by 0 papers
Who cites this paper, from open citation data, with Crossref's record of each; the 100 newest when there are more.
References (24)
What it cites, from open citation data, with Crossref's record of each.
Sign in to see 4 more references
A free account opens the rest of the citing papers and references on every paper’s page.