A Conceptual Framework to Predict Disease Progressions in Patients with Chronic Kidney Disease, Using Machine Learning and Process Mining.

Process Mining is a technique looking into the analysis and mining of existing process flow. On the other hand, Machine Learning is a data science field and a sub-branch of Artificial Intelligence with the main purpose of replicating human behavior through algorithms. The separate application of Process Mining and Machine Learning for healthcare purposes has been widely explored with a various number of published works discussing their use. However, the simultaneous application of Process Mining and Machine Learning algorithms is still a growing field with ongoing studies on its application. This paper proposes a feasible framework where Process Mining and Machine Learning can be used in combination within the healthcare environment.

Machine learning algorithms to uncover risk factors of breast cancer: insights from a large case-control study.

Introduction: This large case-control study explored the application of machine learning models to identify risk factors for primary invasive incident breast cancer (BC) in the Iranian population. This study serves as a bridge toward improved BC prevention, early detection, and management through the identification of modifiable and unmodifiable risk factors. Methods: The dataset includes 1,009 cases and 1,009 controls, with comprehensive data on lifestyle, health-behavior, reproductive and sociodemographic factors. Different machine learning models, namely Random Forest (RF), Neural Networks (NN), Bootstrap Aggregating Classification and Regression Trees (Bagged CART), and Extreme Gradient Boosting Tree (XGBoost), were employed to analyze the data. Results: The findings highlight the significance of a chest X-ray history, deliberate weight loss, abortion history, and post-menopausal status as predictors. Factors such as second-hand smoking, lower education, menarche age (>14), occupation (employed), first delivery age (18-23), and breastfeeding duration (>42 months) were also identified as important predictors in multiple models. The RF model exhibited the highest Area Under the Curve (AUC) value of 0.9, as indicated by the Receiver Operating Characteristic (ROC) curve. Following closely was the Bagged CART model with an AUC of 0.89, while the XGBoost model achieved a slightly lower AUC of 0.78. In contrast, the NN model demonstrated the lowest AUC of 0.74. On the other hand, the RF model achieved an accuracy of 83.9% and a Kappa coefficient of 67.8% and the XGBoost, achieved a lower accuracy of 82.5% and a lower Kappa coefficient of 0.6. Conclusion: This study could be beneficial for targeted preventive measures according to the main risk factors for BC among high-risk women.

Application of job shop scheduling approach in green patient flow optimization using a hybrid swarm intelligence.

With the increasing demand for hospital services amidst the COVID-19 pandemic, allocation of limited public resources and management of healthcare services are of paramount importance. In the field of patient flow scheduling, previous research primarily focused on classical-based objective functions, while ignoring environmental-based objective functions. This study presents a flexible job shop scheduling problem to optimize patient flow and, thereby, minimize the total carbon footprint, as the sustainability-based objective function. Since flexible job shop scheduling is an NP-hard problem, a metaheuristic optimization algorithm, called Chaotic Salp Swarm Algorithm Enhanced with Opposition-Based Learning and Sine Cosine (CSSAOS), was developed. The proposed algorithm integrates the Salp Swarm Algorithm (SSA) with chaotic maps to update the position of followers, the sine cosine algorithm to update the leader position, and opposition-based learning for a better exploration of the search space. generating more accurate solutions. The proposed method was successfully applied in a real-world case study and demonstrated better performance than other well-known metaheuristic algorithms, including differential evolution, genetic algorithm, grasshopper optimization algorithm, SSA based on opposition-based learning, quantum evolutionary SSA, and whale optimization algorithm. In addition, it was found that the proposed method is scalable to different sizes and complexities.

Development and Optimization of a Machine-Learning Prediction Model for Acute Desquamation After Breast Radiation Therapy in the Multicenter REQUITE Cohort.

Purpose: Some patients with breast cancer treated by surgery and radiation therapy experience clinically significant toxicity, which may adversely affect cosmesis and quality of life. There is a paucity of validated clinical prediction models for radiation toxicity. We used machine learning (ML) algorithms to develop and optimise a clinical prediction model for acute breast desquamation after whole breast external beam radiation therapy in the prospective multicenter REQUITE cohort study. Methods and Materials: Using demographic and treatment-related features (m = 122) from patients (n = 2058) at 26 centers, we trained 8 ML algorithms with 10-fold cross-validation in a 50:50 random-split data set with class stratification to predict acute breast desquamation. Based on performance in the validation data set, the logistic model tree, random forest, and naïve Bayes models were taken forward to cost-sensitive learning optimisation. Results: One hundred and ninety-two patients experienced acute desquamation. Resampling and cost-sensitive learning optimisation facilitated an improvement in classification performance. Based on maximising sensitivity (true positives), the "hero" model was the cost-sensitive random forest algorithm with a false-negative: false-positive misclassification penalty of 90:1 containing m = 114 predictive features. Model sensitivity and specificity were 0.77 and 0.66, respectively, with an area under the curve of 0.77 in the validation cohort. Conclusions: ML algorithms with resampling and cost-sensitive learning generated clinically valid prediction models for acute desquamation using patient demographic and treatment features. Further external validation and inclusion of genomic markers in ML prediction models are worthwhile, to identify patients at increased risk of toxicity who may benefit from supportive intervention or even a change in treatment plan.

A Conceptual Framework to Predict Mental Health Patients' Zoning Classification.

Zoning classification is a rating mechanism, which uses a three-tier color coding to indicate perceived risk from the patients' conditions. It is a widely adopted manual system used across mental health settings, however it is time consuming and costly. We propose to automate classification, by adopting a hybrid approach, which combines Temporal Abstraction to capture the temporal relationship between symptoms and patients' behaviors, Natural Language Processing to quantify statistical information from patient notes, and Supervised Machine Learning Models to make a final prediction of zoning classification for mental health patients.

A data science approach for early-stage prediction of Patient's susceptibility to acute side effects of advanced radiotherapy.

The prediction by classification of side effects incidence in a given medical treatment is a common challenge in medical research. Machine Learning (ML) methods are widely used in the areas of risk prediction and classification. The primary objective of such algorithms is to use several features to predict dichotomous responses (e.g., disease positive/negative). Similar to statistical inference modelling, ML modelling is subject to the class imbalance problem and is affected by the majority class, increasing the false-negative rate. In this study, seventy-nine ML models were built and evaluated to classify approximately 2000 participants from 26 hospitals in eight different countries into two groups of radiotherapy (RT) side effects incidence based on recorded observations from the international study of RT related toxicity "REQUITE". We also examined the effect of sampling techniques and cost-sensitive learning methods on the models when dealing with class imbalance. The combinations of such techniques used had a significant impact on the classification. They resulted in an improvement in incidence status prediction by shifting classifiers' attention to the minority group. The best classification model for RT acute toxicity prediction was identified based on domain experts' success criteria. The Area Under Receiver Operator Characteristic curve of the models tested with an isolated dataset ranged from 0.50 to 0.77. The scale of improved results is promising and will guide further development of models to predict RT acute toxicities. One model was optimised and found to be beneficial to identify patients who are at risk of developing acute RT early-stage toxicities as a result of undergoing breast RT ensuring relevant treatment interventions can be appropriately targeted. The design of the approach presented in this paper resulted in producing a preclinical-valid prediction model. The study was developed by a multi-disciplinary collaboration of data scientists, medical physicists, oncologists and surgeons in the UK Radiotherapy Machine Learning Network.

Ensemble Risk Model of Emergency Readmissions (ERMER).

INTRODUCTION: About half of hospital readmissions can be avoided with preventive interventions. Developing decision support tools for identification of patients' emergency readmission risk is an important area of research. Because, it remains unclear how to design features and develop predictive models that can adjust continuously to a fast-changing healthcare system and population characteristics. The objective of this study was to develop a generic ensemble Bayesian risk model of emergency readmission. METHODS: We produced a decision support tool that predicts risk of emergency readmission using England's Hospital Episode Statistics inpatient database. Firstly, we used a framework to develop an optimal set of features. Then, a combination of Bayes Point Machine (BPM) models for different cohorts was considered to create an optimised ensemble model, which is stronger than the individual generative and non-linear classifications. The developed Ensemble Risk Model of Emergency Admissions (ERMER) was trained and tested using three time-frames: 1999-2004, 2000-05 and 2004-09, each of which includes about 20% of patients in England during the trigger year. RESULTS: Comparisons are made for different time-frames, sub-populations, risk cut-offs, risk bands and top risk segments. The precision was 71.6-73.9%, the specificity was 88.3-91.7% and the sensitivity was 42.1-49.2% across different time-frames. Moreover, the Area Under the Curve was 75.9-77.1%. CONCLUSIONS: The decision support tool performed considerably better than the previous modelling approaches, and it was robust and stable with high precision. Moreover, the framework and the Bayesian model allow the model to continuously adjust it to new significant features, different population characteristics and changes in the system.

Towards a threshold climate for emergency lower respiratory hospital admissions.

Identification of 'cut-points' or thresholds of climate factors would play a crucial role in alerting risks of climate change and providing guidance to policymakers. This study investigated a 'Climate Threshold' for emergency hospital admissions of chronic lower respiratory diseases by using a distributed lag non-linear model (DLNM). We analysed a unique longitudinal dataset (10 years, 2000-2009) on emergency hospital admissions, climate, and pollution factors for the Greater London. Our study extends existing work on this topic by considering non-linearity, lag effects between climate factors and disease exposure within the DLNM model considering B-spline as smoothing technique. The final model also considered natural cubic splines of time since exposure and 'day of the week' as confounding factors. The results of DLNM indicated a significant improvement in model fitting compared to a typical GLM model. The final model identified the thresholds of several climate factors including: high temperature (≥27°C), low relative humidity (≤ 40%), high Pm10 level (≥70-µg/m3), low wind speed (≤ 2 knots) and high rainfall (≥30mm). Beyond the threshold values, a significantly higher number of emergency admissions due to lower respiratory problems would be expected within the following 2-3 days after the climate shift in the Greater London. The approach will be useful to initiate 'region and disease specific' climate mitigation plans. It will help identify spatial hot spots and the most sensitive areas and population due to climate change, and will eventually lead towards a diversified health warning system tailored to specific climate zones and populations.

The 7th IMA international conference on quantitative modelling in the management of health and social care: IMA health 2013 special issue.

Health and social care systems are facing major challenges worldwide, due in part to changes in demography and advances in technology and in part to changes in the structure and organisation of care delivery. The IMA Health 2013 conference brought together health care managers, clinicians, management consultants, and mathematicians, operational and health service researchers, statisticians and health economists from across the world with a view to bridging the gap between the respective communities, to exploring recent developments and identifying opportunities for further research. The eight selected papers of this special issue have been grouped into two broad categories. First, there are five papers that report on studies conducted in or relevant to care provision within hospitals. The three remaining papers concern studies aimed at problems related to care provided outside the hospital including long-term care, community based care services and public health. A key learning point arising from these papers and the discussions that took place during the conference is that the systems modelling community need not only to focus their efforts in developing new and improving the performance of existing algorithms, but also in achieving better integration with qualitative research methods and with various relevant strands of the social sciences (ethnography, organisation behaviour etc.). In any case, collaborative projects which engage directly with those involved both in delivering and receiving health care is key if modelling is to make a difference in tackling the messy and complex problems of health and social care.

A structured review of long-term care demand modelling.

Long-term care (LTC) represents a significant and substantial proportion of healthcare spends across the globe. Its main aim is to assist individuals suffering with more or more chronic illnesses, disabilities or cognitive impairments, to carry out activities associated with daily living. Shifts in several economic, demographic and social factors have raised concerns surrounding the sustainability of current systems of LTC. Substantial effort has been put into modelling the LTC demand process itself so as to increase understanding of the factors driving demand for LTC and its related services. Furthermore, such modeling efforts have also been used to plan the operation and future composition of the LTC system itself. The main aim of this paper is to provide a structured review of the literature surrounding LTC demand modeling and any such industrial application, whilst highlighting any potential direction for future researchers.

Profiling hospitals based on emergency readmission: a multilevel transition modelling approach.

Emergency readmission is seen as an important part of the United Kingdom government policy to improve the quality of care that patients receive. In this context, patients and the public have the right to know how well different health organizations are performing. Most methods for profiling estimate the expected numbers of adverse outcomes (e.g. readmission, mortality) for each organization. A number of statistical concerns have been raised, such as the differences in hospital sizes and the unavailability of relevant data for risk adjustment. Having recognized these statistical concerns, a new framework known as the multilevel transition model is developed. Hospital specific propensities of the first, second and further readmissions are considered to be measures of performance, where these measures are used to define a new performance index. During the period 1997 and 2004, the national (English) hospital episodes statistics dataset comprise more than 5 million patient readmissions. Implementing a multilevel model using the complete population dataset could possibly take weeks to estimate the parameters. To resolve the problem, we extract 1000 random samples from the original data, where each random sample is likely to lead to differing hospital performance measures. For computational efficiency a Grid implementation of the model is developed. Analysing the output from the full 1000 sample, we noticed that 4 out of the 5 worst performing hospitals treating cancer patients were in London. These hospitals are known to be the leading NHS Trusts in England, providing diverse range of services to complex patients, and therefore it is inevitable to expect higher numbers of emergency readmissions.

A decision support tool for health service re-design.

Many of the outpatient services are currently only available in hospitals, however there are plans to provide some of these services alongside with General Practitioners. Consequently, General Practitioners could soon be based at polyclinics. These changes have caused a number of concerns to Hounslow Primary Care Trust (PCT). For example, which of the outpatient services are to be shifted from the hospital to the polyclinic? What are the current and expected future demands for these services? To tackle some of these concerns, the first phase of this project explores the set of specialties that are frequently visited in a sequence (using sequential association rules). The second phase develops an Excel based spreadsheet tool to compute the current and expected future demands for the selected specialties. From the sequential association rule algorithm, endocrinology and ophthalmology were found to be highly associated (i.e. frequently visited in a sequence), which means that these two specialties could easily be shifted from the hospital environment to the polyclinic. We illustrated the Excel based spreadsheet tool for endocrinology and ophthalmology, however, the model is generic enough to cope with other specialties, provided that the data are available.

Measuring and modelling occupancy time in NHS continuing healthcare.

BACKGROUND: Due to increasing demand and financial constraints, NHS continuing healthcare systems seek to find better ways of forecasting demand and budgeting for care. This paper investigates two areas of concern, namely, how long existing patients stay in service and the number of patients that are likely to be still in care after a period of time. METHODS: An anonymised dataset containing information for all funded admissions to placement and home care in the NHS continuing healthcare system was provided by 26 (out of 31) London primary care trusts. The data related to 11289 patients staying in placement and home care between 1 April 2005 and 31 May 2008 were first analysed. Using a methodology based on length of stay (LoS) modelling, we captured the distribution of LoS of patients to estimate the probability of a patient staying in care over a period of time. Using the estimated probabilities we forecasted the number of patients that are likely to be still in care after a period of time (e.g. monthly). RESULTS: We noticed that within the NHS continuing healthcare system there are three main categories of patients. Some patients are discharged after a short stay (few days), some others staying for few months and the third category of patients staying for a long period of time (years). Some variations in proportions of discharge and transition between types of care as well as between care groups (e.g. palliative, functional mental health) were observed. A close agreement of the observed and the expected numbers of patients suggests a good prediction model. CONCLUSIONS: The model was tested for care groups within the NHS continuing healthcare system in London to support Primary Care Trusts in budget planning and improve their responsiveness to meet the increasing demand under limited availability of resources. Its applicability can be extended to other types of care, such as hospital care and re-ablement. Further work will be geared towards updating the dataset and refining the results.

Towards effective capacity planning in a perinatal network centre.

OBJECTIVE: To study the arrival pattern and length of stay (LoS) in a neonatal intensive care/high dependency unit (NICU/HDU) and special care baby unit (SCBU) and the impact of capacity shortage in a perinatal network centre, and to provide an analytical model for improving capacity planning. METHODS: The data used in this study have been collected through the South England Neonatal Database (SEND) and the North Central London Perinatal Network Transfer Audit between 1 January and 31 December 2006 for neonates admitted and refused from the neonatal unit at University College London Hospital (UCLH). Exploratory data analysis was performed. A queuing model is proposed for capacity planning of a perinatal network centre. OUTCOME MEASURES: Predicted number of cots required with existing arrival and discharge patterns; impact of reducing LoS. RESULTS: In 2006, 1002 neonates were admitted to the neonatal unit at UCLH, 144 neonates were refused admission to the NICU and 35 to the SCBU. The model shows the NICU requires seven more cots to accept 90% of neonates into the NICU. The model also shows admission acceptance can be increased by 8% if LoS can be reduced by 2 days. CONCLUSIONS: The arrival, LoS and discharge of neonates having gestational ages of <27 weeks were the key determinants of capacity. The queuing model can be used to determine the cot capacity required for a given tolerance level of admission rejection.

Emergency readmission criterion: a technique for determining the emergency readmission time window.

A frequently chosen time window in defining readmission is 28 days after discharge. Yet in the literature, shorter and longer periods such as 14 days or 90-180 days have also been suggested. In this paper, we develop a modeling approach that systematically tackles the issue surrounding the appropriate choice of a time window as a definition of readmission. The approach is based on the intuitive idea that patients who are discharged from hospital can be broadly divided in to two groups-a group that is at high risk of readmission and a group that is at low risk. Using the national data (England), we demonstrate the usefulness of the approach in the case of chronic obstructive pulmonary disease (COPD), stroke, and congestive heart failure (CHF) patients, which are known to be the leading causes of early readmission. Our findings suggest that there are marked differences in the optimal width of the time window for COPD, stroke, and CHF patients. Furthermore, time windows and the probabilities of being in the high-risk group for COPD, stroke, and CHF patients for each of the 29 acute and specialist trusts in the London area indicate wide variability between hospitals. The novelty of this modeling approach lies in its ability to define an appropriate time window based on evidence objectively derived from operational data. Therefore, it can separately provide a unique approach in examining variability between hospitals, and potentially contribute to a better definition of readmission as a performance indicator.