Predictive modeling of inpatient mortality in departments of internal medicine

Machine Learning for Targeted Advance Care Planning in Cancer Patients: A Quality Improvement Study

CONTEXT

Prognostication challenges contribute to delays in advance care planning (ACP) for patients with cancer near the end of life (EOL).

OBJECTIVES

Examine a quality improvement mortality prediction algorithm intervention's impact on ACP documentation and EOL care.

METHODS

We implemented a validated mortality risk prediction machine learning model for solid malignancy patients admitted from the emergency department (ED) to a dedicated solid malignancy unit at Duke University Hospital. Clinicians received an email when a patient was identified as high-risk. We compared ACP documentation and EOL care outcomes before and after the notification intervention. We excluded patients with intensive care unit (ICU) admission in the first 24 hours. Comparisons involved chi-square/Fisher's exact tests and Wilcoxon rank sum tests; comparisons stratified by physician specialty employ Cochran-Mantel-Haenszel tests.

RESULTS

Preintervention and postintervention cohorts comprised 88 and 77 patients, respectively. Most were White, non-Hispanic/Latino, and married. ACP conversations were documented for 2.3% of hospitalizations preintervention vs. 80.5% postintervention (P<0.001), and if the attending physician notified was a palliative care specialist (4.1% vs. 84.6%) or oncologist (0% vs. 76.3%) (P<0.001). There were no differences between groups in length of stay (LOS), hospice referral, code status change, ICU admissions or LOS, 30-day readmissions, 30-day ED visits, and inpatient and 30-day deaths.

CONCLUSION

Identifying patients with cancer and high mortality risk via machine learning elicited a substantial increase in documented ACP conversations but did not impact EOL care. Our intervention showed promise in changing clinician behavior. Further integration of this model in clinical practice is ongoing.

Risk Prediction Models for Hospital Mortality in General Medical Patients: A Systematic Review

Objective

To systematically review contemporary prediction models for hospital mortality developed or validated in general medical patients.

Methods

We screened articles in five databases, from January 1, 2010, through April 7, 2022, and the bibliography of articles selected for final inclusion. We assessed the quality for risk of bias and applicability using the Prediction Model Risk of Bias Assessment Tool (PROBAST) and extracted data using the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist. Two investigators independently screened each article, assessed quality, and extracted data.

Results

From 20,424 unique articles, we identified 15 models in 8 studies across 10 countries. The studies included 280,793 general medical patients and 19,923 hospital deaths. Models included 7 early warning scores, 2 comorbidities indices, and 6 combination models. Ten models were studied in all general medical patients (general models) and 7 in general medical patients with infection (infection models). Of the 15 models, 13 were developed using logistic or Poisson regression and 2 using machine learning methods. Also, 4 of 15 models reported on handling of missing values. None of the infection models had high discrimination, whereas 4 of 10 general models had high discrimination (area under curve >0.8). Only 1 model appropriately assessed calibration. All models had high risk of bias; 4 of 10 general models and 5 of 7 infection models had low concern for applicability for general medical patients.

Conclusion

Mortality prediction models for general medical patients were sparse and differed in quality, applicability, and discrimination. These models require hospital-level validation and/or recalibration in general medical patients to guide mortality reduction interventions.

Do prior neurological comorbidities predict COVID-19 severity and death? A 25-month cross-sectional multicenter study on 7370 patients

BACKGROUND

The prognosis of COVID-19 cases that suffer from particular comorbidities is worse. The impact of chronic neurological disorders (CNDs) on the outcome of COVID-19 patients is not clear yet. This study aimed to assess whether CNDs can predict in-hospital mortality or severity in COVID-19 patients.

METHODS

Following a cross-sectional design, all consecutive hospitalized patients with PCR-confirmed COVID-19 who were hospitalized at three centers from February 20th, 2020 to March 20th, 2022, were studied. CND was defined as neurological conditions resulting in permanent disability. Data on demographic and clinical characteristics, COVID-19 severity, treatment, and laboratory findings were evaluated. A multivariate Cox-regression log-rank test was used to assess the primary outcome, which was in-hospital all-cause mortality. The relationship among CND, COVID-19 severity and abnormal laboratory findings was analyzed as a secondary endpoint.

RESULTS

We studied 7370 cases, 43.6% female, with a mean age of 58.7 years. 1654 (22.4%) patients had one or more CNDs. Patients with CNDs had higher age, were more disabled at baseline, and had more vascular risk factors and comorbidities. The ICU admission rate in CND patients with 59.7% was more frequent than the figure among non-CND patients with 20.3% (p = 0.044). Mortality of those with CND was 43.4%, in comparison with 12.8% in other participants (p = 0.005). Based on the Cox regression analysis, CND could independently predict death (HR 1.198, 95% CI 1.023-3.298, p = 0.003).

CONCLUSION

CNDs could independently predict the death and severity of COVID-19. Therefore, early diagnosis of COVID-19 should be considered in CND patients.

Comparing Outcomes of Critically Ill Patients in Intensive Care Units and General Wards: A Comprehensive Analysis

Background

The admission of critically ill patients to intensive care unit (ICU) plays a crucial role in reducing mortality. However, the scarcity of available ICU beds presents a significant challenge. In resource-limited settings, the outcomes of critically ill patients, particularly those who are not accepted for ICU admission, have been a topic of ongoing debate and contention.

Objective

This study aimed to explore the outcomes and factors associated with ICU admission and mortality among critically ill patients in Thailand.

Methods

This prospective cohort study enrolled critically ill adults indicated for medical ICU admission. Patients were followed for 28 days regardless of whether they were admitted to an ICU. Data on mortality, hospital length of stay, duration of organ support, and factors associated with mortality and ICU admission were collected.

Results

Of the 180 patients enrolled, 72 were admitted to ICUs, and 108 were cared for in general wards. The ICU group had a higher 28-day mortality rate (44.4% vs 20.4%; P=0.001), but other outcomes of interest were comparable. Multivariate analysis identified alteration of consciousness, norepinephrine use, and epinephrine use as independent predictors of 28-day mortality. Higher body mass index (BMI), higher APACHE II score, and acute kidney injury were predictive factors associated with ICU acceptance.

Conclusion

Among patients indicated for ICU admission, those who were admitted had a higher 28-day mortality rate. Higher mortality was associated with alteration of consciousness and vasopressor use. Patients who were sicker and had higher BMI were more likely to be admitted to an ICU.

The Norton scale is an important predictor of in-hospital mortality in internal medicine patients

BACKGROUND

The Norton scale, a marker of patient frailty used to predict the risk of pressure ulcers, but the predictive value of the Norton scale for in-hospital mortality after adjustment for a wide range of demographic, and abnormal admission laboratory test results shown in themselves to have a high predictive value for in-hospital mortality is unclear.

AIM

The study aims to determine the value of the Norton scale and the presence of a urinary catheter in predicting in hospital mortality.

METHODS

The study population included all acutely admitted adult patients in 2020 through October 2021 to one of three internal medicine departments at the Laniado Hospital, a regional hospital with 400 beds in Israel. The main objective was to (a) identify the variables associated with the Norton Scale and (b) determine whether it predicts in-hospital mortality after adjustment for these variables.

RESULTS

The Norton scale was associated with an older age, female gender, presence of a urinary catheter, and abnormal laboratory tests. The odds of in-hospital mortality in those with intermediate, high, and very high Norton scale risk groups were 3.10 (2.23-3.56), 6.48 (4.02-10.46), and 12.27 (7.37-20.44), respectively, after adjustment for the remaining predictors. Adding the Norton scale and the presence of a urinary catheter to the prediction logistic regression model that included age, gender, and abnormal laboratory test results increased the c-statistic from 0.870 (0.864-0.876) to 0.908 (0.902-0.913).

CONCLUSIONS

The Norton scale and presence of a urinary catheter are important predictors of in-hospital mortality in acutely hospitalized adults in internal medicine departments.

Rate and risk factors of in-hospital and early post-discharge mortality in patients admitted to an internal medicine ward

BACKGROUND

We sought to quantify in-hospital and early post-discharge mortality rates in hospitalised patients.

METHODS

Consecutive adult patients admitted to an internal medicine ward were prospectively enrolled. The rates of in-hospital and 4-month post-discharge mortality and their possible associated sociodemographic and clinical factors (eg Cumulative Illness Rating Scale [CIRS], body mass index [BMI], polypharmacy, Barthel Index) were assessed.

RESULTS

1,451 patients (median age 80 years, IQR 69-86; 53% female) were included. Of these, 93 (6.4%) died in hospital, while 4-month post-discharge mortality was 15.9% (191/1,200). Age and high dependency were associated (p<0.01) with a higher risk of in-hospital (OR 1.04 and 2.15) and 4-month (HR 1.04 and 1.65) mortality, while malnutrition and length of stay were associated (p<0.01) with a higher risk of 4-month mortality (HR 2.13 and 1.59).

CONCLUSIONS

Several negative prognostic factors for early mortality were found. Interventions addressing dependency and malnutrition could potentially decrease early post-discharge mortality.

The presence of a urinary catheter is an important predictor of in-hospital mortality in internal medicine patients

RATIONALE AND OBJECTIVE

Mortality rates are used to assess the quality of hospital care after appropriate adjustment for case-mix. Urinary catheters are frequent in hospitalized adults and might be a marker of patient frailty and illness severity. However, we know of no attempts to estimate the predictive value of indwelling catheters for specific patient outcomes. The objective of the present study was to (a) identify the variables associated with the presence of a urinary catheter and (b) determine whether it predicts in-hospital mortality after adjustment for these variables.

METHODS

The study population included all acutely admitted adult patients in 2020 (exploratory cohort) and January-October 2021 (validation cohort) to internal medicine, cardiology and intensive care departments at the Laniado Hospital, a regional hospital with 400 beds in Israel. There were no exclusion criteria. The predictor variables were the presence of a urinary catheter on admission, age, gender, comorbidities and admission laboratory test results. We used bivariate and multivariate logistic regression to test the associations between the presence of a urinary catheter and mortality after adjustment for the remaining independent variables on admission.

RESULTS

The presence of a urinary catheter was associated with other independent variables. In 2020, the odds of in-hospital mortality in patients with a urinary catheter before and after adjustment for the remaining predictors were 14.3 (11.6-17.7) and 6.05 (4.78-7.65), respectively. Adding the presence of a urinary catheter to the prediction logistic regression model increased its c-statistic from 0.887 (0.880-0.894) to 0.907 (0.901-0.913). The results of the validation cohort reduplicated those of the exploratory cohort.

CONCLUSIONS

The presence of a urinary catheter on admission is an important and independent predictor of in-hospital mortality in acutely hospitalized adults in internal medicine departments.

Relevant vs non-relevant subspecialist for patients hospitalised in internal medicine at a local hospital: which is better? A retrospective cohort study

Background

Studies of the treatment of patients in-hospital with a specific diagnosis show that physicians with a subspecialisation relevant to this diagnosis can provide a better quality of care. However, studies including patients with a range of diagnoses show a more negligible effect of being attended by a relevant subspecialist. This project aimed to study a more extensive set of patients and diagnoses in an environment where the subspecialist present could be controlled. Thus, this study investigated whether being attended by a physician with a subspeciality relevant to the patient’s primary diagnosis was prospectively associated with readmission, in-hospital mortality, or length of stay compared to a physician with a subspeciality not relevant to the patient’s primary diagnosis.

Methods

We have conducted a retrospective register-based study of 11,059 hospital admissions across 9 years at a local hospital in south-eastern Norway, where it was possible to identify the physician attending the patients at the beginning of the stay. The outcomes studied were emergency readmissions to the same ward within 30 days, any in-hospital mortality and the total length of stay. The patients admitted were matched with the consultant(s) responsible for their treatment. Then, the admissions were divided into two groups according to their primary diagnosis. Was their diagnosis within the subspeciality of the attending consultant (relevant subspecialist) or not (non-relevant subspecialist). The two groups were then compared using bivariable and multivariable models adjusted for patient characteristics, comorbidities, diagnostic group and physician sex.

Results

A relevant subspecialist was present during the first 3 days in 8058 (73%) of the 11,059 patient cases. Patients attended to by a relevant subspecialist had an odds ratio (OR) of 0.91 (95% confidence interval 0.76 to 1.09) for being readmitted and 0.71 (0.48 to 1.04) for dying in the hospital and had a length of stay that was 0.18 (− 0.07 to 0.42) days longer than for those attended to by a non-relevant subspecialist.

Conclusions

This study found that patients attended by a relevant subspecialist did not have a significantly different outcome to those attended by a non-relevant subspecialist.

Mortality prediction upon hospital admission - the value of clinical assessment: A retrospective, matched cohort study

Accurate prediction of mortality upon hospital admission is of great value, both for the sake of patients and appropriate resources' allocation. A myriad of assessment tools exists for this purpose. The evidence relating to the comparative value of clinical assessment versus established indexes are scarce. We analyzed the accuracy of a senior physician's clinical assessment in a retrospective cohort of patients in a crude, general patients' population and later on a propensity matched patients' population. In one department of internal medicine in a tertiary hospital, of 9891 admitted patients, 973 (10%) were categorized as prone to death in a 6-months' duration by a senior physician. The risk of death was significantly higher for these patients [73.1% vs 14.1% mortality within 180 days; hazard ratio (HR) = 7.58; confidence intervals (CI) 7.02-8.19, P < .001]. After accounting for multiple, other patients' variables associated with increased risk of mortality, the correlation remained significant (HR = 3.25; CI 2.85-3.71, P < .001). We further performed a propensity matching analysis (a subgroup of 710 patients, subdivided to two groups with 355 patients each): survival rates were as low as 45% for patients categorized as prone to death compared to 78% in patients who weren't categorized as such (P < .001). Reliance on clinical evaluation, done by an experienced senior physician, is an appropriate tool for mortality prediction upon hospital admission, achieving high accuracy rates.

Palliative Care Exposure Relative to Predicted Risk of Six-Month Mortality in Hospitalized Adults

CONTEXT

The optimal strategy for implementing mortality-predicting algorithms to facilitate clinical care, prognostic discussions, and palliative care interventions remains unknown.

OBJECTIVES

To develop and validate a real-time predictive model for 180 day mortality using routinely available clinical and laboratory admission data and determine if palliative care exposure varies with predicted mortality risk.

METHODS

Adult admissions between October 1, 2013 and October.1, 2017 were included for the model derivation. A separate cohort was collected between January 1, 2018 and July 31, 2020 for validation. Patients were followed for 180 days from discharge, and logistic regression with selected variables was used to estimate patients' risk for mortality.

RESULTS

In the model derivation cohort, 7963 events of 180 day mortality (4.5% event rate) were observed. Median age was 53.0 (IQR 24.0-66.0) with 92,734 females (52.5%). Variables with strongest association with 180 day mortality included: Braden Score (OR 0.83; 95% CI 0.82-0.84); admission Do Not Resuscitate orders (OR 2.61; 95% CI 2.43-2.79); admission service and admission status. The model yielded excellent discriminatory ability in both the derivation (c-statistic 0.873; 95% CI 0.870-0.877; Brier score 0.04) and validation cohorts (c-statistic 0.844; 95% CI 0.840-0.847; Brier score 0.072). Inpatient palliative care consultations increased from 3% of minimal-risk encounters to 41% of high-risk encounters (P < 0.01).

CONCLUSION

We developed and temporally validated a predictive mortality model for adults from a large retrospective cohort, which helps quantify the potential need for palliative care referrals based on risk strata. Machine learning algorithms for mortality require clinical interpretation, and additional studies are needed to design patient-centered and risk-specific interventions.

Towards accurate prediction of patient length of stay at emergency department: a GAN-driven deep learning framework

Recently, the hospital systems face a high influx of patients generated by several events, such as seasonal flows or health crises related to epidemics (e.g., COVID'19). Despite the extent of the care demands, hospital establishments, particularly emergency departments (EDs), must admit patients for medical treatments. However, the high patient influx often increases patients' length of stay (LOS) and leads to overcrowding problems within the EDs. To mitigate this issue, hospital managers need to predict the patient's LOS, which is an essential indicator for assessing ED overcrowding and the use of the medical resources (allocation, planning, utilization rates). Thus, accurately predicting LOS is necessary to improve ED management. This paper proposes a deep learning-driven approach for predicting the patient LOS in ED using a generative adversarial network (GAN) model. The GAN-driven approach flexibly learns relevant information from linear and nonlinear processes without prior assumptions on data distribution and significantly enhances the prediction accuracy. Furthermore, we classified the predicted patients' LOS according to time spent at the pediatric emergency department (PED) to further help decision-making and prevent overcrowding. The experiments were conducted on actual data obtained from the PED in Lille regional hospital center, France. The GAN model results were compared with other deep learning models, including deep belief networks, convolutional neural network, stacked auto-encoder, and four machine learning models, namely support vector regression, random forests, adaboost, and decision tree. Results testify that deep learning models are suitable for predicting patient LOS and highlight GAN's superior performance than the other models.

Parameters of the complete blood count predict in hospital mortality

INTRODUCTION

Mortality rates are used to evaluate the quality of hospital care after adjusting for disease severity and, commonly also, for age, comorbidity, and laboratory data with only few parameters of the complete blood count (CBC).

OBJECTIVE

To identify the parameters of the CBC that predict independently in-hospital mortality of acutely admitted patients.

POPULATION

All patients were admitted to internal medicine, cardiology, and intensive care departments at the Laniado Hospital in Israel in 2018 and 2019.

VARIABLES

Independent variables were patients' age, sex, and parameters of the CBC. The outcome variable was in-hospital mortality.

ANALYSIS

Logistic regression. In 2018, we identified the variables that were associated with in-hospital mortality and validated this association in the 2019 cohort.

RESULTS

In the validation cohort, a model consisting of nine parameters that are commonly available in modern analyzers had a c-statistics (area under the receiver operator curve) of 0.86 and a 10%-90% risk gradient of 0%-21.4%. After including the proportions of large unstained cells, hypochromic, and macrocytic red cells, the c-statistic increased to 0.89, and the risk gradient to 0.1%-29.5%.

CONCLUSION

The commonly available parameters of the CBC predict in-hospital mortality. Addition of the proportions of hypochromic red cells, macrocytic red cells, and large unstained cells may improve the predictive value of the CBC.

An Optimized Machine Learning Model Accurately Predicts In-Hospital Outcomes at Admission to a Cardiac Unit

Risk stratification at the time of hospital admission is of paramount significance in triaging the patients and providing timely care. In the present study, we aim at predicting multiple clinical outcomes using the data recorded during admission to a cardiac care unit via an optimized machine learning method. This study involves a total of 11,498 patients admitted to a cardiac care unit over two years. Patient demographics, admission type (emergency or outpatient), patient history, lab tests, and comorbidities were used to predict various outcomes. We employed a fully connected neural network architecture and optimized the models for various subsets of input features. Using 10-fold cross-validation, our optimized machine learning model predicted mortality with a mean area under the receiver operating characteristic curve (AUC) of 0.967 (95% confidence interval (CI): 0.963–0.972), heart failure AUC of 0.838 (CI: 0.825–0.851), ST-segment elevation myocardial infarction AUC of 0.832 (CI: 0.821–0.842), pulmonary embolism AUC of 0.802 (CI: 0.764–0.84), and estimated the duration of stay (DOS) with a mean absolute error of 2.543 days (CI: 2.499–2.586) of data with a mean and median DOS of 6.35 and 5.0 days, respectively. Further, we objectively quantified the importance of each feature and its correlation with the clinical assessment of the corresponding outcome. The proposed method accurately predicts various cardiac outcomes and can be used as a clinical decision support system to provide timely care and optimize hospital resources.

Improving the In-Hospital Mortality Prediction of Diabetes ICU Patients Using a Process Mining/Deep Learning Architecture

Diabetes intensive care unit (ICU) patients are at increased risk of complications leading to in-hospital mortality. Assessing the likelihood of death is a challenging and time consuming task due to a large number of influencing factors. Healthcare providers are interested in the detection of ICU patients at higher risk, such that risk factors can possibly be mitigated. While such severity scoring methods exist, they are commonly based on a snapshot of the health conditions of a patient during the ICU stay and do not specifically consider a patient's prior medical history. In this paper, a process mining/deep learning architecture is proposed to improve established severity scoring methods by incorporating the medical history of diabetes patients. First, health records of past hospital encounters are converted to event logs suitable for process mining. The event logs are then used to discover a process model that describes the past hospital encounters of patients. An adaptation of Decay Replay Mining is proposed to combine medical and demographic information with established severity scores to predict the in hospital mortality of diabetes ICU patients. Significant performance improvements are demonstrated compared to established risk severity scoring methods and machine learning approaches using the Medical Information Mart for Intensive Care III dataset.

A simple index predicting mortality in acutely hospitalized patients

BACKGROUND

Mortality rates used to evaluate and improve the quality of hospital care are adjusted for comorbidity and disease severity. Comorbidity, measured by International Classification of Diseases codes, do not reflect the severity of the medical condition, that requires clinical assessments not available in electronic databases, and/or laboratory data with clinically relevant ranges to permit extrapolation from one setting to the next.

AIM

To propose a simple index predicting mortality in acutely hospitalized patients.

DESIGN

Retrospective cohort study with internal and external validation.

METHODS

The study populations were all acutely admitted patients in 2015-16, and in January 2019-November 2019 to internal medicine, cardiology and intensive care departments at the Laniado Hospital in Israel, and in 2002-19, at St. James Hospital, Ireland. Predictor variables were age and admission laboratory tests. The outcome variable was in-hospital mortality. Using logistic regression of the data in the 2015-16 Israeli cohort, we derived an index that included age groups and significant laboratory data.

RESULTS

In the Israeli 2015-16 cohort, the index predicted mortality rates from 0.2% to 32.0% with a c-statistic (area under the receiver operator characteristic curve) of 0.86. In the Israeli 2019 validation cohort, the index predicted mortality rates from 0.3% to 38.9% with a c-statistic of 0.87. An abbreviated index performed similarly in the Irish 2002-19 cohort.

CONCLUSIONS

Hospital mortality can be predicted by age and selected admission laboratory data without acquiring information from the patient's medical records. This permits an inexpensive comparison of performance of hospital departments.

Hypoalbuminemia as Surrogate and Culprit of Infections

Hypoalbuminemia is associated with the acquisition and severity of infectious diseases, and intact innate and adaptive immune responses depend on albumin. Albumin oxidation and breakdown affect interactions with bioactive lipid mediators that play important roles in antimicrobial defense and repair. There is bio-mechanistic plausibility for a causal link between hypoalbuminemia and increased risks of primary and secondary infections. Serum albumin levels have prognostic value for complications in viral, bacterial and fungal infections, and for infectious complications of non-infective chronic conditions. Hypoalbuminemia predicts the development of healthcare-associated infections, particularly with Clostridium difficile. In coronavirus disease 2019, hypoalbuminemia correlates with viral load and degree of acute lung injury and organ dysfunction. Non-oncotic properties of albumin affect the pharmacokinetics and pharmacodynamics of antimicrobials. Low serum albumin is associated with inadequate antimicrobial treatment. Infusion of human albumin solution (HAS) supplements endogenous albumin in patients with cirrhosis of the liver and effectively supported antimicrobial therapy in randomized controlled trials (RCTs). Evidence of the beneficial effects of HAS on infections in hypoalbuminemic patients without cirrhosis is largely observational. Prospective RCTs are underway and, if hypotheses are confirmed, could lead to changes in clinical practice for the management of hypoalbuminemic patients with infections or at risk of infectious complications.

Predicting In-Hospital Mortality at Admission to the Medical Ward: A Big-Data Machine Learning Model

BACKGROUND

General medical wards admit high-risk patients. Artificial intelligence algorithms can use big data for developing models to assess patients' risk stratification. The aim of this study was to develop a mortality prediction machine learning model using data available at the time of admission to the medical ward.

METHODS

We included consecutive patients (ages 18-100) admitted to medical wards at a single medical center (January 1, 2013-December 31, 2018). We constructed a machine learning model using patient characteristics, comorbidities, laboratory tests, and patients' emergency department (ED) management. The model was trained on data from the years 2013 to 2017 and validated on data from the year 2018. The area under the curve (AUC) for mortality prediction was used as an outcome metric. Youden index was used to find an optimal sensitivity-specificity cutoff point.

RESULTS

Of the 118,262 patients admitted to the medical ward, 6311 died (5.3%). The single variables with the highest AUCs were medications administered in the ED (AUC = 0.74), ED diagnosis (AUC = 0.74), and albumin (AUC = 0.73). The machine learning model yielded an AUC of 0.924 (95% confidence interval [CI]: 0.917-0.930). For Youden index, a sensitivity of 0.88 (95% CI: 0.86-0.89) and specificity of 0.83 (95% CI: 0.83-0.83) were observed. This corresponds to a false-positive rate of 1:5.9 and negative predictive value of 0.99.

CONCLUSION

A machine learning model outperforms single variables predictions of in-hospital mortality at the time of admission to the medical ward. Such a decision support tool has the potential to augment clinical decision-making regarding level of care needed for admitted patients.

Factors Associated with In-Hospital Mortality in Acute Care Hospital Settings: A Prospective Observational Study

Background: In-hospital mortality is a key indicator of the quality of care. Studies so far have demonstrated the influence of patient and hospital-related factors on in-hospital mortality. Currently, new variables, such as nursing workload or the level of dependency, are being incorporated. We aimed to identify which individual, clinical and hospital characteristics are related to hospital mortality. Methods: A multicentre prospective observational study design was used. Sampling was conducted between February 2015 and October 2017. Patients over 16 years, admitted to medical or surgical units at 11 public hospitals in Andalusia (Spain), with a foreseeable stay of at least 48 h were included. Multivariate regression analyses were performed to analyse the data. Results: The sample consisted of 3821 assessments conducted in 1004 patients. The mean profile was that of a male (52%), mean age of 64.5 years old, admitted to a medical unit (56.5%), with an informal caregiver (60%). In-hospital mortality was 4%. The INICIARE (Inventario del Nivel de Cuidados Mediante Indicadores de Clasificación de Resultados de Enfermería) scale yielded an adjusted odds ratio [AOR] of 0.987 (95% confidence interval [CI]: 0.97–0.99) and the nurse staffing level (NSL) yielded an AOR of 1.197 (95% CI: 1.02–1.4). Conclusion: Nursing care dependency measured by INICIARE and nurse staffing level was associated with in-hospital mortality.

Development and validation of a machine learning-based prediction model for near-term in-hospital mortality among patients with COVID-19

OBJECTIVES

To develop and validate a model for prediction of near-term in-hospital mortality among patients with COVID-19 by application of a machine learning (ML) algorithm on time-series inpatient data from electronic health records.

METHODS

A cohort comprised of 567 patients with COVID-19 at a large acute care healthcare system between 10 February 2020 and 7 April 2020 observed until either death or discharge. Random forest (RF) model was developed on randomly drawn 70% of the cohort (training set) and its performance was evaluated on the rest of 30% (the test set). The outcome variable was in-hospital mortality within 20-84 hours from the time of prediction. Input features included patients' vital signs, laboratory data and ECG results.

RESULTS

Patients had a median age of 60.2 years (IQR 26.2 years); 54.1% were men. In-hospital mortality rate was 17.0% and overall median time to death was 6.5 days (range 1.3-23.0 days). In the test set, the RF classifier yielded a sensitivity of 87.8% (95% CI: 78.2% to 94.3%), specificity of 60.6% (95% CI: 55.2% to 65.8%), accuracy of 65.5% (95% CI: 60.7% to 70.0%), area under the receiver operating characteristic curve of 85.5% (95% CI: 80.8% to 90.2%) and area under the precision recall curve of 64.4% (95% CI: 53.5% to 75.3%).

CONCLUSIONS

Our ML-based approach can be used to analyse electronic health record data and reliably predict near-term mortality prediction. Using such a model in hospitals could help improve care, thereby better aligning clinical decisions with prognosis in critically ill patients with COVID-19.

Prediction of general medical admission length of stay with natural language processing and deep learning: a pilot study

Length of stay (LOS) and discharge destination predictions are key parts of the discharge planning process for general medical hospital inpatients. It is possible that machine learning, using natural language processing, may be able to assist with accurate LOS and discharge destination prediction for this patient group. Emergency department triage and doctor notes were retrospectively collected on consecutive general medical and acute medical unit admissions to a single tertiary hospital from a 2-month period in 2019. These data were used to assess the feasibility of predicting LOS and discharge destination using natural language processing and a variety of machine learning models. 313 patients were included in the study. The artificial neural network achieved the highest accuracy on the primary outcome of predicting whether a patient would remain in hospital for > 2 days (accuracy 0.82, area under the received operator curve 0.75, sensitivity 0.47 and specificity 0.97). When predicting LOS as an exact number of days, the artificial neural network achieved a mean absolute error of 2.9 and a mean squared error of 16.8 on the test set. For the prediction of home as a discharge destination (vs any non-home alternative), all models performed similarly with an accuracy of approximately 0.74. This study supports the feasibility of using natural language processing to predict general medical inpatient LOS and discharge destination. Further research is indicated with larger, more detailed, datasets from multiple centres to optimise and examine the accuracy that may be achieved with such predictions.

Padua prediction score and IMPROVE score do predict in-hospital mortality in Internal Medicine patients

Padua prediction score (PPS) and IMPROVE bleeding score are validated tools for venous thromboembolism (VTE) risk assessment recommended by guidelines, albeit not frequently used. Some data suggest that a positive PPS and IMPROVE score may be were associated with early mortality in Internal Medicine patients. Aim of the study was to characterize the predictive ability on mortality of the two scores using two different populations, respectively, as derivation and validation cohort. The derivation cohort consisted of 1956 Internal Medicine patients admitted to La Spezia Hospital in 2013. 399 Internal Medicine patients admitted to Carate Brianza Hospital in 2016 constituted the validation cohort. PPS and IMPROVE scores were applied to each patient using their validated cutoffs. Frequency of positive PPS and mortality were significantly higher in La Spezia patients. In the derivation cohort, the positivity of at least one of the two scores was associated with a significantly higher mortality compared to both negative scores. Similar results were observed in the validation cohort. In the derivation cohort, the sensitivity of a positive PPS score in predicting mortality was 0.97 (0.94, 0.98) but the specificity was 0.21 (0.19, 0.23), the negative likelihood ratio being 0.15. Sensitivity and specificity of a positive IMPROVE gave specular findings but the positive likelihood ratio was 2.19. The accuracy data in the validation cohort were in the same direction. Both PPS and IMPROVE are associated with in-hospital mortality but their additional predictive accuracy is modest. It is unlikely that both scores could be useful in clinical practice to predict death in hospitalized Internal Medicine patients.

Neurological Comorbidity Is a Predictor of Death in Covid-19 Disease: A Cohort Study on 576 Patients

Introduction: Prognosis of Coronavirus disease 2019 (Covid-19) patients with vascular risk factors, and certain comorbidities is worse. The impact of chronic neurological disorders (CND) on prognosis is unclear. We evaluated if the presence of CND in Covid-19 patients is a predictor of a higher in-hospital mortality. As secondary endpoints, we analyzed the association between CND, Covid-19 severity, and laboratory abnormalities during admission.

Methods: Retrospective cohort study that included all the consecutive hospitalized patients with confirmed Covid-19 disease from March 8th to April 11th, 2020. The study setting was Hospital Clínico, tertiary academic hospital from Valladolid. CND was defined as those neurological conditions causing permanent disability. We assessed demography, clinical variables, Covid-19 severity, laboratory parameters and outcome. The primary endpoint was in-hospital all-cause mortality, evaluated by multivariate cox-regression log rank test. We analyzed the association between CND, covid-19 severity and laboratory abnormalities.

Results: We included 576 patients, 43.3% female, aged 67.2 years in mean. CND were present in 105 (18.3%) patients. Patients with CND were older, more disabled, had more vascular risk factors and comorbidities and fewer clinical symptoms of Covid-19. They presented 1.43 days earlier to the emergency department. Need of ventilation support was similar. Presence of CND was an independent predictor of death (HR 2.129, 95% CI: 1.382–3.280) but not a severer Covid-19 disease (OR: 1.75, 95% CI: 0.970–3.158). Frequency of laboratory abnormalities was similar, except for procalcitonin and INR.

Conclusions: The presence of CND is an independent predictor of mortality in hospitalized Covid-19 patients. That was not explained neither by a worse immune response to Covid-19 nor by differences in the level of care received by patients with CND.

Using structured pathology data to predict hospital-wide mortality at admission

Early prediction of in-hospital mortality can improve patient outcome. Current prediction models for in-hospital mortality focus mainly on specific pathologies. Structured pathology data is hospital-wide readily available and is primarily used for e.g. financing purposes. We aim to build a predictive model at admission using the International Classification of Diseases (ICD) codes as predictors and investigate the effect of the self-evident DNR (“Do Not Resuscitate”) diagnosis codes and palliative care codes. We compare the models using ICD-10-CM codes with Risk of Mortality (RoM) and Charlson Comorbidity Index (CCI) as predictors using the Random Forests modeling approach. We use the Present on Admission flag to distinguish which diagnoses are present on admission. The study is performed in a single center (Ghent University Hospital) with the inclusion of 36 368 patients, all discharged in 2017. Our model at admission using ICD-10-CM codes (AUCROC = 0.9477) outperforms the model using RoM (AUCROC = 0.8797 and CCI (AUCROC = 0.7435). We confirmed that DNR and palliative care codes have a strong impact on the model resulting in a decrease of 7% for the ICD model (AUCROC = 0.8791) at admission. We therefore conclude that a model with a sufficient predictive performance can be derived from structured pathology data, and if real-time available, can serve as a prerequisite to develop a practical clinical decision support system for physicians.

Prospective and External Evaluation of a Machine Learning Model to Predict In-Hospital Mortality of Adults at Time of Admission

IMPORTANCE

The ability to accurately predict in-hospital mortality for patients at the time of admission could improve clinical and operational decision-making and outcomes. Few of the machine learning models that have been developed to predict in-hospital death are both broadly applicable to all adult patients across a health system and readily implementable. Similarly, few have been implemented, and none have been evaluated prospectively and externally validated.

OBJECTIVES

To prospectively and externally validate a machine learning model that predicts in-hospital mortality for all adult patients at the time of hospital admission and to design the model using commonly available electronic health record data and accessible computational methods.

DESIGN, SETTING, AND PARTICIPANTS

In this prognostic study, electronic health record data from a total of 43 180 hospitalizations representing 31 003 unique adult patients admitted to a quaternary academic hospital (hospital A) from October 1, 2014, to December 31, 2015, formed a training and validation cohort. The model was further validated in additional cohorts spanning from March 1, 2018, to August 31, 2018, using 16 122 hospitalizations representing 13 094 unique adult patients admitted to hospital A, 6586 hospitalizations representing 5613 unique adult patients admitted to hospital B, and 4086 hospitalizations representing 3428 unique adult patients admitted to hospital C. The model was integrated into the production electronic health record system and prospectively validated on a cohort of 5273 hospitalizations representing 4525 unique adult patients admitted to hospital A between February 14, 2019, and April 15, 2019.

MAIN OUTCOMES AND MEASURES

The main outcome was in-hospital mortality. Model performance was quantified using the area under the receiver operating characteristic curve and area under the precision recall curve.

RESULTS

A total of 75 247 hospital admissions (median [interquartile range] patient age, 59.5 [29.0] years; 45.9% involving male patients) were included in the study. The in-hospital mortality rates for the training validation; retrospective validations at hospitals A, B, and C; and prospective validation cohorts were 3.0%, 2.7%, 1.8%, 2.1%, and 1.6%, respectively. The area under the receiver operating characteristic curves were 0.87 (95% CI, 0.83-0.89), 0.85 (95% CI, 0.83-0.87), 0.89 (95% CI, 0.86-0.92), 0.84 (95% CI, 0.80-0.89), and 0.86 (95% CI, 0.83-0.90), respectively. The area under the precision recall curves were 0.29 (95% CI, 0.25-0.37), 0.17 (95% CI, 0.13-0.22), 0.22 (95% CI, 0.14-0.31), 0.13 (95% CI, 0.08-0.21), and 0.14 (95% CI, 0.09-0.21), respectively.

CONCLUSIONS AND RELEVANCE

Prospective and multisite retrospective evaluations of a machine learning model demonstrated good discrimination of in-hospital mortality for adult patients at the time of admission. The data elements, methods, and patient selection make the model implementable at a system level.

May nutritional status worsen during hospital stay? A sub-group analysis from a cross-sectional study

Hospital malnutrition is a detrimental prognostic factor regarding hospital mortality, complications, and length of stay. However, the role of hospitalization itself on nutritional status has not been fully elucidated. We report the results of a secondary analysis from the dataset of a recent cross-sectional study at Fondazione Policlinico Universitario A. Gemelli IRCCS, Rome, Italy. Data from patients evaluated at admission and discharge were collected and compared. One hundred thirty-nine patients were evaluated. Mean length of stay was 13.6 (± 7.7) days. Patients at risk of malnutrition, according to NRS-2002, were 75 (53.9%), while 63 (45.3%) were malnourished according to ESPEN Criteria. Compared to admission, at discharge, patients reported a significant decrease in Mid-Upper Arm Circumference (MUAC)-from 26.5 cm (± 3.6) to 25.9 cm (± 3.7) (p = 0.016), a reduction in Phase angle (PhA)-from 4.25° (± 1.20) to 4.01° (± 1.15) (p = 0.005), fat-free mass (FFM)-from 47.5 kg (± 9.19) to 44.9 kg (± 9.4) (p = 0.03) and fat-free mass index (FFMI)-from 16.9 kg/m² (± 2.3) to 15.8 kg/m² (± 2.7) (p = 0.04). Laboratory data showed a reduction of albumin-from 29.2 (± 5.7) to 28.0 (± 5.9) (p = 0.01) and Onodera's PNI- from 29.1 (± 5.6) to 27.6 kg (± 5.6) (p = 0.039). At the multivariate linear regression analysis, the variables significantly associated with a worsening of PhA at discharge are the PhA value at admission and the diagnosis of malnutrition according to ESPEN Criteria. Hospitalization leads to significative changes in nutritional status. A clinical concern should be raised about the quality of hospital food and meal times and on the need for a clinical nutritionist on the ward.

Risk, prevalence, and impact of hospital malnutrition in a Tertiary Care Referral University Hospital: a cross-sectional study

Hospital malnutrition is still underestimated among physicians, even in internal medicine settings. This is a cross-sectional study, aiming to estimate the risk, the prevalence and the impact of malnutrition in an Internal Medicine and Gastroenterology Department of a large Italian hospital (Fondazione Policlinico Universitario A. Gemelli IRCCS, Rome). Patients were evaluated within 72 h from admission according to Nutritional Risk Screening-2002 (NRS-2002) and European Society for Clinical Nutrition and Metabolism (ESPEN) Criteria. Anthropometric, laboratory tests and Bioelectrical Impedance Analysis (BIA) derived phase angle were also performed. Length of hospital stay (LOS) and in-hospital mortality were collected. Univariate and multivariate analyses were conducted to correlate nutritional status with LOS and hospital mortality. In 10 months, 300 patients were enrolled: male patients were 172 (57.3%); mean age was 63.7 (± 17.6). At admission, 157 (52.3%) patients were at risk of malnutrition; 116 (38.7%) were malnourished. Malnourished patients had a mean LOS of 11.5 (± 8.0) days, not-malnourished 9.4 (± 6.2) days (p < 0.05). In-hospital mortality did not significantly differ between the two groups. Multivariate analysis shows that both malnutrition (p = 0.04; 95% CI 0.03-3.41) and phase angle (p = 0.004; 95% CI - 1.92 to - 0.37) independently correlate with LOS. In an Internal Medicine and Gastroenterology Department, over half (52.3%) of the patients were found at risk of malnutrition, and over a third (38.7%) were malnourished at hospital admission. Malnutrition and BIA-derived phase angle are independently associated with LOS. ESPEN Criteria and phase angle could be performed at admission to identify patients deserving specific nutritional support.

Incorporating machine learning approaches to assess putative environmental risk factors for multiple sclerosis

BACKGROUND

Multiple sclerosis (MS) incidence has increased recently, particularly in women, suggesting a possible role of one or more environmental exposures in MS risk. The study objective was to determine if animal, dietary, recreational, or occupational exposures are associated with MS risk.

METHODS

Least absolute shrinkage and selection operator (LASSO) regression was used to identify a subset of exposures with potential relevance to disease in a large population-based (Kaiser Permanente Northern California [KPNC]) case-control study. Variables with non-zero coefficients were analyzed in matched conditional logistic regression analyses, adjusted for established environmental risk factors and socioeconomic status (if relevant in univariate screening),± genetic risk factors, in the KPNC cohort and, for purposes of replication, separately in the Swedish Epidemiological Investigation of MS cohort. These variables were also assessed in models stratified by HLA-DRB1*15:01 status since interactions between risk factors and that haplotype have been described.

RESULTS

There was a suggestive association of pesticide exposure with having MS among men, but only in those who were positive for HLA-DRB1*15:01 (OR pooled = 3.11, 95% CI 0.87, 11.16, p = 0.08).

CONCLUSIONS

While this finding requires confirmation, it is interesting given the association between pesticide exposure and other neurological diseases. The study also demonstrates the application of LASSO to identify environmental exposures with reduced multiple statistical testing penalty. Machine learning approaches may be useful for future investigations of concomitant MS risk or prognostic factors.