Applications of data envelopment analysis in acute care hospitals: a systematic literature review, 1984-2022.

This study reviews scholarly publications on data envelopment analysis (DEA) studies on acute care hospital (ACH) efficiency published between 1984 and 2022 in scholarly peer-reviewed journals. We employ systematic literature review (SLR) method to identify and analyze pertinent past research using predetermined steps. The SLR offers a comprehensive resource that meticulously analyzes DEA methodology for practitioners and researchers focusing on ACH efficiency measurement. The articles reviewed in the SLR are analyzed and synthesized based on the nature of the DEA modelling process and the key findings from the DEA models. The key findings from the DEA models are presented under the following sections: effects of different ownership structures; impacts of specific healthcare reforms or other policy interventions; international and multi-state comparisons; effects of changes in competitive environment; impacts of new technology implementations; effects of hospital location; impacts of quality management interventions; impact of COVID-19 on hospital performance; impact of teaching status, and impact of merger. Furthermore, the nature of DEA modelling process focuses on use of sensitivity analysis; choice of inputs and outputs; comparison with Stochastic Frontier Analysis; use of congestion analysis; use of bootstrapping; imposition of weight restrictions; use of DEA window analysis; and exogenous factors. The findings demonstrate that, despite several innovative DEA extensions and hospital applications, over half of the research used the conventional DEA models. The findings also show that the most often used inputs in the DEA models were labor-oriented inputs and hospital beds, whereas the most frequently used outputs were outpatient visits, followed by surgeries, admissions, and inpatient days. Further research on the impact of healthcare reforms and health information technology (HIT) on hospital performance is required, given the number of reforms being implemented in many countries and the role HIT plays in enhancing care quality and lowering costs. We conclude by offering several new research directions for future studies.

Fall Risk Assessment in Stroke Survivors: A Machine Learning Model Using Detailed Motion Data from Common Clinical Tests and Motor-Cognitive Dual-Tasking.

BACKGROUND: Falls are common and dangerous for stroke survivors. Current fall risk assessment methods rely on subjective scales. Objective sensor-based methods could improve prediction accuracy. OBJECTIVE: Develop machine learning models using inertial sensors to objectively classify fall risk in stroke survivors. Determine optimal sensor configurations and clinical test protocols. METHODS: 21 stroke survivors performed balance, Timed Up and Go, 10 Meter Walk, and Sit-to-Stand tests with and without dual-tasking. A total of 8 motion sensors captured lower limb and trunk kinematics, and 92 spatiotemporal gait and clinical features were extracted. Supervised models-Support Vector Machine, Logistic Regression, and Random Forest-were implemented to classify high vs. low fall risk. Sensor setups and test combinations were evaluated. RESULTS: The Random Forest model achieved 91% accuracy using dual-task balance sway and Timed Up and Go walk time features. Single thorax sensor models performed similarly to multi-sensor models. Balance and Timed Up and Go best-predicted fall risk. CONCLUSION: Machine learning models using minimal inertial sensors during clinical assessments can accurately quantify fall risk in stroke survivors. Single thorax sensor setups are effective. Findings demonstrate a feasible objective fall screening approach to assist rehabilitation.

Using a Motion Sensor to Categorize Nonspecific Low Back Pain Patients: A Machine Learning Approach.

Nonspecific low back pain (NSLBP) constitutes a critical health challenge that impacts millions of people worldwide with devastating health and socioeconomic consequences. In today's clinical settings, practitioners continue to follow conventional guidelines to categorize NSLBP patients based on subjective approaches, such as the STarT Back Screening Tool (SBST). This study aimed to develop a sensor-based machine learning model to classify NSLBP patients into different subgroups according to quantitative kinematic data, i.e., trunk motion and balance-related measures, in conjunction with STarT output. Specifically, inertial measurement units (IMU) were attached to the trunks of ninety-four patients while they performed repetitive trunk flexion/extension movements on a balance board at self-selected pace. Machine learning algorithms (support vector machine (SVM) and multi-layer perceptron (MLP)) were implemented for model development, and SBST results were used as ground truth. The results demonstrated that kinematic data could successfully be used to categorize patients into two main groups: high vs. low-medium risk. Accuracy levels of ~75% and 60% were achieved for SVM and MLP, respectively. Additionally, among a range of variables detailed herein, time-scaled IMU signals yielded the highest accuracy levels (i.e., ~75%). Our findings support the improvement and use of wearable systems in developing diagnostic and prognostic tools for various healthcare applications. This can facilitate development of an improved, cost-effective quantitative NSLBP assessment tool in clinical and home settings towards effective personalized rehabilitation.

A machine learning model to predict heart failure readmission: toward optimal feature set.

Background: Hospital readmissions for heart failure patients remain high despite efforts to reduce them. Predictive modeling using big data provides opportunities to identify high-risk patients and inform care management. However, large datasets can constrain performance. Objective: This study aimed to develop a machine learning based prediction model leveraging a nationwide hospitalization database to predict 30-day heart failure readmissions. Another objective of this study is to find the optimal feature set that leads to the highest AUC value in the prediction model. Material and methods: Heart failure patient data was extracted from the 2020 Nationwide Readmissions Database. A heuristic feature selection process incrementally incorporated predictors into logistic regression and random forest models, which yields a maximum increase in the AUC metric. Discrimination was evaluated through accuracy, sensitivity, specificity and AUC. Results: A total of 566,019 discharges with heart failure diagnosis were recognized. Readmission rate was 8.9% for same-cause and 20.6% for all-cause diagnoses. Random forest outperformed logistic regression, achieving AUCs of 0.607 and 0.576 for same-cause and all-cause readmissions respectively. Heuristic feature selection resulted in the identification of optimal feature sets including 20 and 22 variables from a pool of 30 and 31 features for the same-cause and all-cause datasets. Key predictors included age, payment method, chronic kidney disease, disposition status, number of ICD-10-CM diagnoses, and post-care encounters. Conclusion: The proposed model attained discrimination comparable to prior analyses that used smaller datasets. However, reducing the sample enhanced performance, indicating big data complexity. Improved techniques like heuristic feature selection enabled effective leveraging of the nationwide data. This study provides meaningful insights into predictive modeling methodologies and influential features for forecasting heart failure readmissions.

Post-Stroke Functional Changes: In-Depth Analysis of Clinical Tests and Motor-Cognitive Dual-Tasking Using Wearable Sensors.

Clinical tests like Timed Up and Go (TUG) facilitate the assessment of post-stroke mobility, but they lack detailed measures. In this study, 21 stroke survivors and 20 control participants underwent TUG, sit-to-stand (STS), and the 10 Meter Walk Test (10MWT). Tests incorporated single tasks (STs) and motor-cognitive dual-task (DTs) involving reverse counting from 200 in decrements of 10. Eight wearable motion sensors were placed on feet, shanks, thighs, sacrum, and sternum to record kinematic data. These data were analyzed to investigate the effects of stroke and DT conditions on the extracted features across segmented portions of the tests. The findings showed that stroke survivors (SS) took 23% longer for total TUG (p < 0.001), with 31% longer turn time (p = 0.035). TUG time increased by 20% (p < 0.001) from STs to DTs. In DTs, turning time increased by 31% (p = 0.005). Specifically, SS showed 20% lower trunk angular velocity in sit-to-stand (p = 0.003), 21% longer 10-Meter Walk time (p = 0.010), and 18% slower gait speed (p = 0.012). As expected, turning was especially challenging and worsened with divided attention. The outcomes of our study demonstrate the benefits of instrumented clinical tests and DTs in effectively identifying motor deficits post-stroke across sitting, standing, walking, and turning activities, thereby indicating that quantitative motion analysis can optimize rehabilitation procedures.

Predicting the risk of hospital readmissions using a machine learning approach: a case study on patients undergoing skin procedures.

Introduction: Even with modern advancements in medical care, one of the persistent challenges hospitals face is the frequent readmission of patients. These recurrent admissions not only escalate healthcare expenses but also amplify mental and emotional strain on patients. Methods: This research delved into two primary areas: unraveling the pivotal factors causing the readmissions, specifically targeting patients who underwent dermatological treatments, and determining the optimal machine learning algorithms that can foresee potential readmissions with higher accuracy. Results: Among the multitude of algorithms tested, including logistic regression (LR), support vector machine (SVM), random forest (RF), Naïve Bayesian (NB), artificial neural network (ANN), xgboost (XG), and k-nearest neighbor (KNN), it was noted that two models-XG and RF-stood out in their prediction prowess. A closer inspection of the data brought to light certain patterns. For instance, male patients and those between the ages of 21 and 40 had a propensity to be readmitted more frequently. Moreover, the months of March and April witnessed a spike in these readmissions, with ~6% of the patients returning within just a month after their first admission. Discussion: Upon further analysis, specific determinants such as the patient's age and the specific hospital where they were treated emerged as key indicators influencing the likelihood of their readmission.

Predicting risk of sepsis, comparison between machine learning methods: a case study of a Virginia hospital.

Sepsis is an inflammation caused by the body's systemic response to an infection. The infection could be a result of many diseases, such as pneumonia, urinary tract infection, and other illnesses. Some of its symptoms are fever, tachycardia, tachypnea, etc. Unfortunately, sepsis remains a critical problem at the hospitals and leads to many issues, such as increasing mortality rate, health care costs, and health care utilization. Early detection of sepsis in patients can help respond quickly, take preventive actions, and prevent major issues. The main aim of this study is to predict the risk of sepsis by utilizing the patient's demographic and clinical information, i.e., patient's gender, age, severity level, mortality risk, admit type along with hospital length of stay. Six machine learning approaches, Logistic Regression (LR), Naïve Bayes, Support Vector Machine (SVM), Boosted Tree, Classification and Regression Tree (CART), and Bootstrap Forest are used to predict the risk of sepsis. The results showed that different machine learning methods have other performances in terms of various measures. For instance, the Bootstrap Forest machine learning method exhibited the highest performance in AUC and R-square or SVM and Boosted Tree showed the highest performance in terms of misclassification rate. The Bootstrap Forest can be considered the best machine learning method in predicting sepsis regarding applied features in this research, mainly because it showed superior performance and efficiency in two performance measures: AUC and R-square.

The Analysis of Hospital Readmission Rates After the Implementation of Hospital Readmissions Reduction Program.

OBJECTIVES: This study aims to analyze the impact of Hospital Readmissions Reduction Program (HRRP) on the nationwide optimization efforts of length of stay (LOS) and readmissions in the United States. METHODS: We use the Nationwide Readmission Database between 2010 and 2016 provided in the Healthcare Cost and Utilization Project by the Agency for Healthcare Research and Quality. The study focuses on acute myocardial infarction, chronic obstructive pulmonary disease, congestive heart failure (CHF), pneumonia monitored by the HRRP and 2 conditions, septicemia, and mood disorders that were not monitored by the HRRP but had among the highest readmissions. Patient demographics and readmissions were analyzed based on insurance type, LOS, and Charlson Comorbidity Index. RESULTS: The readmissions vary by conditions, LOS, and insurance types. Congestive heart failure has the highest readmissions among the 6 analyzed conditions at approximately 25%. The readmission rate of CHF rises to 30% for the Medicaid patients and varies between 30% and 35% by LOS. Patients with CHF with higher Charlson Comorbidity Index demonstrates the highest readmissions among 6 conditions. The patients with longer LOSs had higher readmissions, and Medicare patients have a higher reduction in readmissions in acute myocardial infarction and mood disorders compared with the other forms of payments. CONCLUSIONS: Our figures show that targeted programs, such as HRRP, may have a positive impact on readmission rates. We, however, observe some graphical evidence that nontargeted conditions could exhibit similar trends. Because of heterogeneity in hospital and patient characteristics, it is pivotal for researcher to consider them in formal analyses.

Monitoring length of stay through control charts: a comparative study of diabetic patients.

Objective: Monitoring length of stay (LOS) can help medical decision makers identify areas of potential improvements and improve resource management, which results in better quality of care for patients. This study aims to monitor process performance at hospitals by implementing a statistical process control (SPC) approach on LOS.Methods: The study focuses on diabetic inpatients admitted to hospitals in two national healthcare systems. The data used in this study were collected from two hospitals: (1) a 500-bed teaching hospital in Southwest Virginia in the U.S., and (2) a 1100-bed teaching and research hospital located in Ankara, Turkey. I-MR charts were used to analyze the datasets and monitor the variations of LOS.Results: The results of I-MR charts showed that LOS was longer in Turkey than the U.S. LOS was skewed toward minimum values in the U.S. whereas it was spread out in Turkey. The average LOS was 3.27 days (STD = 2.30) in the U.S. while it was 7.28 days (STD = 4.56) in Turkey. The differences in two national healthcare systems may be reflected in the LOS variable.Conclusion: This study implements a control chart-based approach to monitor LOS and detect prolonged hospitalization for diabetic patients. As presented in I-MR charts, there are abnormal LOS observations in each data set. The decision makers and caregivers must analyze I-MR charts to identify either common or special causes of variation. Each abnormal LOS requires a detailed patient-centric analysis. Care providers and decision makers can investigate the root causes of abnormal LOS for each patient by further exploring the characteristics of diabetic patients who had abnormal LOS at hospitals, such as age, preexisting conditions, or the type of medical procedure conducted on each patient.

Proactive Risk Assessment of Surgical Site Infections in Ambulatory Surgery Centers.

BACKGROUND: The Socio-Technical Probabilistic Risk Assessment, a proactive risk assessment tool imported from high-risk industries, was used to identify risks for surgical site infections (SSIs) associated with the ambulatory surgery center setting and to guide improvement efforts. OBJECTIVES: This study had 2 primary objectives: (1) to identify the critical risk factors associated with SSIs resulting from procedures performed at ambulatory surgery centers and (2) to design an intervention to mitigate the probability of SSI for the highest risk factors identified. METHODS: Inputs included quantitative and qualitative data sources from the evidence-based literature and from health care providers. The Socio-Technical Probabilistic Risk Assessment ranked the failure points (events) on the basis of their contribution to an SSI. The event, entitled "Failure to protect the patient effectively," which included several failure points, was the most critical unique event with the highest contribution to SSI risk. RESULTS: A total of 51.87% of SSIs in this setting were caused by this failure. Consequently, we proposed an intervention aimed at all 5 major components of this failure. CONCLUSIONS: The intervention targets improvements in skin preparation; proper administration of antibiotics; staff training in infection control principles, including practices for the prevention of glove punctures; and procedures to ensure the removal of watches, jewelry, and artificial nails.

Can a Patient's In-Hospital Length of Stay and Mortality Be Explained by Early-Risk Assessments?

OBJECTIVE: To assess whether a patient's in-hospital length of stay (LOS) and mortality can be explained by early objective and/or physicians' subjective-risk assessments. DATA SOURCES/STUDY SETTING: Analysis of a detailed dataset of 1,021 patients admitted to a large U.S. hospital between January and September 2014. STUDY DESIGN: We empirically test the explanatory power of objective and subjective early-risk assessments using various linear and logistic regression models. PRINCIPAL FINDINGS: The objective measures of early warning can only weakly explain LOS and mortality. When controlled for various vital signs and demographics, objective signs lose their explanatory power. LOS and death are more associated with physicians' early subjective risk assessments than the objective measures. CONCLUSIONS: Explaining LOS and mortality require variables beyond patients' initial medical risk measures. LOS and in-hospital mortality are more associated with the way in which the human element of healthcare service (e.g., physicians) perceives and reacts to the risks.