New machine learning scoring system for predicting postoperative mortality in gastroduodenal ulcer perforation: A study using a Japanese nationwide inpatient database

Application of Machine Learning Based on Structured Medical Data in Gastroenterology

The era of big data has led to the necessity of artificial intelligence models to effectively handle the vast amount of clinical data available. These data have become indispensable resources for machine learning. Among the artificial intelligence models, deep learning has gained prominence and is widely used for analyzing unstructured data. Despite the recent advancement in deep learning, traditional machine learning models still hold significant potential for enhancing healthcare efficiency, especially for structured data. In the field of medicine, machine learning models have been applied to predict diagnoses and prognoses for various diseases. However, the adoption of machine learning models in gastroenterology has been relatively limited compared to traditional statistical models or deep learning approaches. This narrative review provides an overview of the current status of machine learning adoption in gastroenterology and discusses future directions. Additionally, it briefly summarizes recent advances in large language models.

Long-Term Risk of Being Bedridden in Elderly Patients Who Underwent Oncologic Surgery: A Retrospective Study Using a Japanese Claims Database

BACKGROUND

Although functional outcomes are important in surgery for elderly patients, the long-term functional prognosis following oncologic surgery is unclear. We retrospectively investigated the long-term, functional and survival prognosis following major oncologic surgery according to age among elderly patients.

METHODS

We used a Japanese administrative database to identify 11,896 patients aged ≥ 65 years who underwent major oncological surgery between June 2014 and February 2019. We investigated the association between age at surgery and the postoperative incidence of bedridden status and mortality. Using the Fine-Gray model and restricted cubic spline functions, we conducted a multivariable, survival analysis with adjustments for patient background characteristics and treatment courses to estimate hazard ratios for the outcomes.

RESULTS

During a median follow-up of 588 (interquartile range, 267-997) days, 657 patients (5.5%) became bedridden and 1540 (13%) died. Patients aged ≥ 70 years had a significantly higher incidence of being bedridden than those aged 65-69 years; the subdistribution hazard ratios of the age groups of 70-74, 75-79, 80-84, and ≥ 85 years were 3.20 (95% confidence interval [CI], 1.53-6.71), 3.86 (95% CI 1.89-7.89), 6.26 (95% CI 3.06-12.8), and 8.60 (95% CI 4.19-17.7), respectively. Restricted cubic spline analysis demonstrated an increase in the incidence of bedridden status in patients aged ≥ 65 years, whereas mortality increased in patients aged ≥ 75 years.

CONCLUSIONS

This large-scale, observational study revealed that older age at oncological surgery was associated with poorer functional outcomes and higher mortality among patients aged ≥ 65 years.

Risk Factors for Postoperative Bleeding Following Breast Cancer Surgery: A Nationwide Database Study of 477,108 Cases in Japan

BACKGROUND

Although postoperative bleeding is a common and serious complication in breast cancer surgery, the risk factors remain unclear. Therefore, we examined the risk factors using a Japanese nationwide database.

METHODS

Patients who underwent breast cancer surgery between July 2010 and March 2020 were identified from a Japanese nationwide database. Multivariable analyses for 47 candidate risk factors (4 patient characteristics, 32 comorbidities, 5 tumor characteristics, 3 preoperative drug uses, and 3 surgical procedures) were conducted to investigate risk factors associated with postoperative bleeding requiring reoperation. Two sensitivity analyses were conducted: an analysis for postoperative bleeding with or without reoperation and an analysis for patients who underwent total mastectomy without breast reconstruction.

RESULTS

Among the 477,108 patients included, 7048 (1.5%) developed postoperative bleeding and 2357 (0.5%) underwent reoperation for postoperative bleeding. Male sex, old age, body mass index ≥ 25.0 kg/m², several comorbidities (deficiency anemia, cardiac arrhythmias, hypertension, liver disease, psychoses, and valvular disease), preoperative heparin use, and several procedures were identified as risk factors. Deficiency anemia showed the highest odds ratio among the risk factors (4.41 [95% confidence interval, 3.63-5.36]). High odds ratios were also observed in total mastectomy (2.32 [2.10-2.56]), flap reconstruction (1.93 [1.55-2.40]), and preoperative heparin use (1.64 [1.26-2.14]). The results corresponded with the sensitivity analyses.

CONCLUSIONS

This study identified several risk factors for postoperative bleeding in breast cancer surgery, such as high body mass index, anemia, cardiovascular diseases, liver diseases, psychoses, preoperative heparin use, and surgical procedures.

Predicting Postoperative Mortality With Deep Neural Networks and Natural Language Processing: Model Development and Validation

Background

Machine learning (ML) achieves better predictions of postoperative mortality than previous prediction tools. Free-text descriptions of the preoperative diagnosis and the planned procedure are available preoperatively. Because reading these descriptions helps anesthesiologists evaluate the risk of the surgery, we hypothesized that deep learning (DL) models with unstructured text could improve postoperative mortality prediction. However, it is challenging to extract meaningful concept embeddings from this unstructured clinical text.

Objective

This study aims to develop a fusion DL model containing structured and unstructured features to predict the in-hospital 30-day postoperative mortality before surgery. ML models for predicting postoperative mortality using preoperative data with or without free clinical text were assessed.

Methods

We retrospectively collected preoperative anesthesia assessments, surgical information, and discharge summaries of patients undergoing general and neuraxial anesthesia from electronic health records (EHRs) from 2016 to 2020. We first compared the deep neural network (DNN) with other models using the same input features to demonstrate effectiveness. Then, we combined the DNN model with bidirectional encoder representations from transformers (BERT) to extract information from clinical texts. The effects of adding text information on the model performance were compared using the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). Statistical significance was evaluated using P<.05.

Results

The final cohort contained 121,313 patients who underwent surgeries. A total of 1562 (1.29%) patients died within 30 days of surgery. Our BERT-DNN model achieved the highest AUROC (0.964, 95% CI 0.961-0.967) and AUPRC (0.336, 95% CI 0.276-0.402). The AUROC of the BERT-DNN was significantly higher compared to logistic regression (AUROC=0.952, 95% CI 0.949-0.955) and the American Society of Anesthesiologist Physical Status (ASAPS AUROC=0.892, 95% CI 0.887-0.896) but not significantly higher compared to the DNN (AUROC=0.959, 95% CI 0.956-0.962) and the random forest (AUROC=0.961, 95% CI 0.958-0.964). The AUPRC of the BERT-DNN was significantly higher compared to the DNN (AUPRC=0.319, 95% CI 0.260-0.384), the random forest (AUPRC=0.296, 95% CI 0.239-0.360), logistic regression (AUPRC=0.276, 95% CI 0.220-0.339), and the ASAPS (AUPRC=0.149, 95% CI 0.107-0.203).

Conclusions

Our BERT-DNN model has an AUPRC significantly higher compared to previously proposed models using no text and an AUROC significantly higher compared to logistic regression and the ASAPS. This technique helps identify patients with higher risk from the surgical description text in EHRs.