Diagnosing hospital bacteraemia in the framework of predictive, preventive and personalised medicine using electronic health records and machine learning classifiers

Leveraging explainable artificial intelligence for early prediction of bloodstream infections using historical electronic health records

Bloodstream infections (BSIs) are a severe public health threat due to their rapid progression into critical conditions like sepsis. This study presents a novel eXplainable Artificial Intelligence (XAI) framework to predict BSIs using historical electronic health records (EHRs). Leveraging a dataset from St. Olavs Hospital in Trondheim, Norway, encompassing 35,591 patients, the framework integrates demographic, laboratory, and comprehensive medical history data to classify patients into high-risk and low-risk BSI groups. By avoiding reliance on real-time clinical data, our model allows for enhanced scalability across various healthcare settings, including resource-limited environments. The XAI framework significantly outperformed traditional models, particularly with tree-based algorithms, demonstrating superior specificity and sensitivity in BSI prediction. This approach promises to optimize resource allocation and potentially reduce healthcare costs while providing interpretability for clinical decision-making, making it a valuable tool in hospital systems for early intervention and improved patient outcomes.

Artificial intelligence and prescription of antibiotic therapy: present and future

INTRODUCTION

In the past few years, the use of artificial intelligence in healthcare has grown exponentially. Prescription of antibiotics is not exempt from its rapid diffusion, and various machine learning (ML) techniques, from logistic regression to deep neural networks and large language models, have been explored in the literature to support decisions regarding antibiotic prescription.

AREAS COVERED

In this narrative review, we discuss promises and challenges of the application of ML-based clinical decision support systems (ML-CDSSs) for antibiotic prescription. A search was conducted in PubMed up to April 2024.

EXPERT OPINION

Prescribing antibiotics is a complex process involving various dynamic phases. In each of these phases, the support of ML-CDSSs has shown the potential, and also the actual ability in some studies, to favorably impacting relevant clinical outcomes. Nonetheless, before widely exploiting this massive potential, there are still crucial challenges ahead that are being intensively investigated, pertaining to the transparency of training data, the definition of the sufficient degree of prediction explanations when predictions are obtained through black box models, and the legal and ethical framework for decision responsibility whenever an antibiotic prescription is supported by ML-CDSSs.

Machine learning-derived blood culture classification with both predictive and prognostic values in the intensive care unit: A retrospective cohort study

OBJECTIVES

Diagnosis and management of intensive care unit (ICU)-acquired bloodstream infections are often based on positive blood culture results. This retrospective cohort study aimed to develop a classification model using data-driven characterisation to optimise the management of intensive care patients with blood cultures.

SETTING, METHODOLOGY/DESIGN

An unsupervised clustering model was developed based on the clinical characteristics of patients with blood cultures in the Medical Information Mart for Intensive Care (MIMIC)-IV database (n = 2451). It was tested using the data from the MIMIC-III database (n = 2047).

MAIN OUTCOME MEASURES

The prognosis, blood culture outcomes, antimicrobial interventions, and trajectories of infection indicators were compared between clusters.

RESULTS

Four clusters were identified using machine learning-based k-means clustering based on data obtained 48 h before the first blood culture sampling. Cluster γ was associated with the highest 28-day mortality rate, followed by clusters α, δ, and β. Cluster γ had a higher fungal isolation rate than cluster β (P < 0.05). Cluster δ was associated with a higher isolation rate of Gram-negative organisms and fungi (P < 0.05). Patients in clusters γ and δ underwent more femoral site vein catheter placements than those in cluster β (P < 0.001, all). Patients with a duration of antibiotics treatment of 4, 6, and 7 days in clusters α, δ, and γ, respectively, had the lowest 28-day mortality rate.

CONCLUSIONS

Machine learning identified four clusters of intensive care patients with blood cultures, which yielded different prognoses, blood culture outcomes, and optimal duration of antibiotic treatment. Such data-driven blood culture classifications suggest further investigation should be undertaken to optimise treatment and improve care.

IMPLICATIONS FOR CLINICAL PRACTICE

Intensive care unit-acquired bloodstream infections are heterogeneous. Meaningful classifications associated with outcomes should be described. Using machine learning and cluster analysis could help in understanding heterogeneity. Data-driven blood culture classification could identify distinct physiological states and prognoses before deciding on blood culture sampling, optimise treatment, and improve care.

The application of artificial intelligence in the management of sepsis

Sepsis is a complex and heterogeneous syndrome that remains a serious challenge to healthcare worldwide. Patients afflicted by severe sepsis or septic shock are customarily placed under intensive care unit (ICU) supervision, where a multitude of apparatus is poised to produce high-granularity data. This reservoir of high-quality data forms the cornerstone for the integration of AI into clinical practice. However, existing reviews currently lack the inclusion of the latest advancements. This review examines the evolving integration of artificial intelligence (AI) in sepsis management. Applications of artificial intelligence include early detection, subtyping analysis, precise treatment and prognosis assessment. AI-driven early warning systems provide enhanced recognition and intervention capabilities, while profiling analyzes elucidate distinct sepsis manifestations for targeted therapy. Precision medicine harnesses the potential of artificial intelligence for pathogen identification, antibiotic selection, and fluid optimization. In conclusion, the seamless amalgamation of artificial intelligence into the domain of sepsis management heralds a transformative shift, ushering in novel prospects to elevate diagnostic precision, therapeutic efficacy, and prognostic acumen. As AI technologies develop, their impact on shaping the future of sepsis care warrants ongoing research and thoughtful implementation.

Methodological issues of the electronic health records' use in the context of epidemiological investigations, in light of missing data: a review of the recent literature

BACKGROUND

Electronic health records (EHRs) are widely accepted to enhance the health care quality, patient monitoring, and early prevention of various diseases, even when there is incomplete or missing information in them.

AIM

The present review sought to investigate the impact of EHR implementation on healthcare quality and medical decision in the context of epidemiological investigations, considering missing or incomplete data.

METHODS

Google scholar, Medline (via PubMed) and Scopus databases were searched for studies investigating the impact of EHR implementation on healthcare quality and medical decision, as well as for studies investigating the way of dealing with missing data, and their impact on medical decision and the development process of prediction models. Electronic searches were carried out up to 2022.

RESULTS

EHRs were shown that they constitute an increasingly important tool for both physicians, decision makers and patients, which can improve national healthcare systems both for the convenience of patients and doctors, while they improve the quality of health care as well as they can also be used in order to save money. As far as the missing data handling techniques is concerned, several investigators have already tried to propose the best possible methodology, yet there is no wide consensus and acceptance in the scientific community, while there are also crucial gaps which should be addressed.

CONCLUSIONS

Through the present thorough investigation, the importance of the EHRs' implementation in clinical practice was established, while at the same time the gap of knowledge regarding the missing data handling techniques was also pointed out.

Radiomics based on readout-segmented echo-planar imaging (RS-EPI) diffusion-weighted imaging (DWI) for prognostic risk stratification of patients with rectal cancer: a two-centre, machine learning study using the framework of predictive, preventive, and personalized medicine

Background

Currently, the rate of recurrence or metastasis (ROM) remains high in rectal cancer (RC) patients treated with the standard regimen. The potential of diffusion-weighted imaging (DWI) in predicting ROM risk has been reported, but the efficacy is insufficient.

Aims

This study investigated the potential of a new sequence called readout-segmented echo-planar imaging (RS-EPI) DWI in predicting the ROM risk of patients with RC using machine learning methods to achieve the principle of predictive, preventive, and personalized medicine (PPPM) application in RC treatment.

Methods

A total of 195 RC patients from two centres who directly received total mesorectal excision were retrospectively enrolled in our study. Machine learning methods, including recursive feature elimination (RFE), the synthetic minority oversampling technique (SMOTE), and the support vector machine (SVM) classifier, were used to construct models based on clinical-pathological factors (clinical model), radiomic features from RS-EPI DWI (radiomics model), and their combination (merged model). The Harrell concordance index (C-index) and the area under the time-dependent receiver operating characteristic curve (AUC) were calculated to evaluate the predictive performance at 1 year, 3 years, and 5 years. Kaplan‒Meier analysis was performed to evaluate the ability to stratify patients according to the risk of ROM.

Findings

The merged model performed well in predicting tumour ROM in patients with RC at 1 year, 3 years, and 5 years in both cohorts (AUC = 0.887/0.813/0.794; 0.819/0.795/0.783) and was significantly superior to the clinical model (AUC = 0.87 [95% CI: 0.80-0.93] vs. 0.71 [95% CI: 0.59-0.81], p = 0.009; C-index = 0.83 [95% CI: 0.76-0.90] vs. 0.68 [95% CI: 0.56-0.79], p = 0.002). It also had a significant ability to differentiate patients with a high and low risk of ROM (HR = 12.189 [95% CI: 4.976-29.853], p < 0.001; HR = 6.427 [95% CI: 2.265-13.036], p = 0.002).

Conclusion

Our developed merged model based on RS-EPI DWI accurately predicted and effectively stratified patients with RC according to the ROM risk at an early stage with an individualized profile, which may be able to assist physicians in individualizing the treatment protocols and promote a meaningful paradigm shift in RC treatment from traditional reactive medicine to PPPM.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13167-022-00303-3.

Oculomics for sarcopenia prediction: a machine learning approach toward predictive, preventive, and personalized medicine

Aims

Sarcopenia is characterized by a gradual loss of skeletal muscle mass and strength with increased adverse outcomes. Recently, large-scale epidemiological studies have demonstrated a relationship between several chronic disorders and ocular pathological conditions using an oculomics approach. We hypothesized that sarcopenia can be predicted through eye examinations, without invasive tests or radiologic evaluations in the context of predictive, preventive, and personalized medicine (PPPM/3PM).

Methods

We analyzed data from the Korean National Health and Nutrition Examination Survey (KNHANES). The training set (80%, randomly selected from 2008 to 2010) data were used to construct the machine learning models. Internal (20%, randomly selected from 2008 to 2010) and external (from the KNHANES 2011) validation sets were used to assess the ability to predict sarcopenia. We included 8092 participants in the final dataset. Machine learning models (XGBoost) were trained on ophthalmological examinations and demographic factors to detect sarcopenia.

Results

In the exploratory analysis, decreased levator function (odds ratio [OR], 1.41; P value <0.001), cataracts (OR, 1.31; P value = 0.013), and age-related macular degeneration (OR, 1.38; P value = 0.026) were associated with an increased risk of sarcopenia in men. In women, an increased risk of sarcopenia was associated with blepharoptosis (OR, 1.23; P value = 0.038) and cataracts (OR, 1.29; P value = 0.010). The XGBoost technique showed areas under the receiver operating characteristic curves (AUCs) of 0.746 and 0.762 in men and women, respectively. The external validation achieved AUCs of 0.751 and 0.785 for men and women, respectively. For practical and fast hands-on experience with the predictive model for practitioners who may be willing to test the whole idea of sarcopenia prediction based on oculomics data, we developed a simple web-based calculator application (https://knhanesoculomics.github.io/sarcopenia) to predict the risk of sarcopenia and facilitate screening, based on the model established in this study.

Conclusion

Sarcopenia is treatable before the vicious cycle of sarcopenia-related deterioration begins. Therefore, early identification of individuals at a high risk of sarcopenia is essential in the context of PPPM. Our oculomics-based approach provides an effective strategy for sarcopenia prediction. The proposed method shows promise in significantly increasing the number of patients diagnosed with sarcopenia, potentially facilitating earlier intervention. Through patient oculometric monitoring, various pathological factors related to sarcopenia can be simultaneously analyzed, and doctors can provide personalized medical services according to each cause. Further studies are needed to confirm whether such a prediction algorithm can be used in real-world clinical settings to improve the diagnosis of sarcopenia.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13167-022-00292-3.

Predicting acupuncture efficacy for functional dyspepsia based on routine clinical features: a machine learning study in the framework of predictive, preventive, and personalized medicine

Background

Acupuncture is safe and effective for functional dyspepsia (FD), while its efficacy varies among individuals. Predicting the response of different FD patients to acupuncture treatment in advance and therefore administering the tailored treatment to the individual is consistent with the principle of predictive, preventive, and personalized medicine (PPPM/3PM). In the current study, the individual efficacy prediction models were developed based on the support vector machine (SVM) algorithm and routine clinical features, aiming to predict the efficacy of acupuncture in treating FD and identify the FD patients who were appropriate to acupuncture treatment.

Methods

A total of 745 FD patients were collected from two clinical trials. All the patients received a 4-week acupuncture treatment. Based on the demographic and baseline clinical features of 80% of patients in trial 1, the SVM models were established to predict the acupuncture response and improvements of symptoms and quality of life (QoL) at the end of treatment. Then, the left 20% of patients in trial 1 and 193 patients in trial 2 were respectively applied to evaluate the internal and external generalizations of these models.

Results

These models could predict the efficacy of acupuncture successfully. In the internal test set, models achieved an accuracy of 0.773 in predicting acupuncture response and an R ² of 0.446 and 0.413 in the prediction of QoL and symptoms improvements, respectively. Additionally, these models had well generalization in the independent validation set and could also predict, to a certain extent, the long-term efficacy of acupuncture at the 12-week follow-up. The gender, subtype of disease, and education level were finally identified as the critical predicting features.

Conclusion

Based on the SVM algorithm and routine clinical features, this study established the models to predict acupuncture efficacy for FD patients. The prediction models developed accordingly are promising to assist doctors in judging patients' responses to acupuncture in advance, so that they could tailor and adjust acupuncture treatment plans for different patients in a prospective rather than the reactive manner, which could greatly improve the clinical efficacy of acupuncture treatment for FD and save medical expenditures.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13167-022-00271-8.