Development and validation of prediction model using nursing notes on sentiment scores for prognosis of patients with severe acute kidney injury receiving continuous renal replacement therapy based on computational intelligence algorithms

Abstract

Background: Currently, the prediction values of models for the prognosis of acute kidney injury (AKI) receiving continuous renal replacement therapy (CRRT) were ordinary and establishing a better prediction model is necessary. Nursing notes are an important predictor of in-hospital mortality in intensive care unit (ICU) patients. This study established prognostic prediction models for AKI patients receiving CRRT especially using nursing notes.Methods: Totally, 682 AKI patients undergoing CRRT were included. AKI was diagnosed based on Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Four hundred and twelve patients lacking nursing notes data were excluded. Finally, 270 patients were included and randomly divided into a training set (n=189) and a testing set (n=81) at a ratio of 7:3. Univariate analysis explored the possible predictors of mortality in AKI patients receiving CRRT. Random forest models and broad learning system (BLS) models (with or without sentiment scores) were respectively constructed in the training set and verified in the testing set. The performances of the models were assessed by the sensitivity, specificity, and area under the curve (AUC).Results: For the random forest model including the sentiment scores, the AUC was 0.86 (95% CI: 0.81-0.91), the sensitivity was 0.72 (95% CI: 0.63-0.80), and the specificity was 0.87 (95% CI: 0.80-0.94) in the training set and the AUC was 0.78 (95% CI: 0.68-0.88), the sensitivity was 0.65 (95% CI: 0.49-0.80), and the specificity was 0.75 (95% CI: 0.62-0.88) in the testing set. For the BLS model including the sentiment scores, the AUC was 0.87 (95% CI: 0.82-0.92), the sensitivity was 0.95 (95% CI: 0.91-0.99) and the specificity was 0.48 (95% CI: 0.38-0.59) in the training set and the AUC was 0.82 (95% CI: 0.73-0.91), the sensitivity was 0.41 (95% CI: 0.25-0.56) and the specificity was 0.98 (95% CI: 0.93-1.00) in the testing set. Conclusions:The BLS models including the sentiment scores might offer a tool for quickly identifying patients AKI patients receiving CRRT with high risk of mortality and providing timely interventions to them for improving their prognosis.

Using artificial intelligence to predict mortality in AKI patients: a systematic review/meta-analysis

“…A total of 14 different ML/AI models were reported in these studies. These included: logistic regression (four studies) [ 3 , 6 , 8 , 11 ], RF (seven studies) [ 3 – 7 , 10 , 11 ], XGBoost (four studies) [ 3 , 5 , 10 , 11 ], SVM (four studies) [ 3 , 5 , 7 , 11 ], ANN/MLP (three studies) [ 3 , 5 , 7 ], BLS models (one study) [ 4 ], elastic net final (ENF) model fitted (one study) [ 10 ], PCM (which combined variables commonly selected with multiple ML methods plus clinical rationale of utility/feasibility assessment by the investigators) (one study) [ 11 ], κ-nearest neighbor (KNN) (one study) [ 5 ], multivariable adaptive regression splines (MARS) (one study) [ 5 ], customized Simplified Acute Physiology Score II (SAPS II) (one study) [ 7 ], mortality scoring system for acute kidney injury with CRRT (MOSAIC model) (one study) [ 5 ], Sequential Organ Failure Assessment (SOFA) (one study) [ 5 ] and Acute Physiologic Assessment and Chronic Health Evaluation II (APACHE II) (one study) [ 5 ].…”

“…AKI affects approximately 15% of hospitalized patients and 30%–60% of patients in the intensive care unit (ICU), with high morbidity and mortality risk in both settings. Continuous renal replacement therapy (CRRT) is commonly used to provide renal support in ICU patients with severe AKI, yet the mortality rate of AKI patients receiving CRRT remains high at 30%–70% [ 4 , 5 ]. Improvement in patient outcomes with AKI has proven to be complex due to the condition's wide array of etiologies and complex pathophysiology, which often leads to a delay in diagnosis [ 2 ].…”

“…It is anticipated that these AI models will be able to enhance prognostic accuracy and overcome shortcomings of previous models [ 6 ]. By efficiently utilizing large datasets such as the Medical Information Mart for Intensive Care III and patient electronic health records, these models can discriminately predict new, complex clinical scenarios, thus reducing mortality and poor patient outcomes in AKI patients [ 4 , 6 ]. Furthermore, even subjective factors such as sentiment analysis and nursing notes have shown to be highly useful in these advanced models [ 4 ].…”