Predicting High Health Care Resource Utilization in a Single-payer Public Health Care System

Rosella, Laura; Kornas, Kathy; Yao, Zhan; Dujovny, Manuel; Bornbaum, Catherine; Fransoo, Randy; Stukel, Thérèse A.

doi:10.1097/mlr.0000000000000837

Cited by 33 publications

(51 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Effectively predicting healthcare service utilization has multiple clinical implications and can help to improve delivery, population health, and resource allocation to support the Quadruple Aim and support transitions towards value-based care delivery systems [ 36 , 37 ]. Predictive models have previously been developed and validated to predict healthcare resource utilization using patient-level EHR data [ 37 – 39 ]. Further, they have demonstrated outperformance of existing clinical prediction rules, with machine learning models demonstrating small performance benefits (with limited clinical differences) over statistical models [ 38 ].…”

Section: Discussionmentioning

confidence: 99%

Exploring supervised machine learning approaches to predicting Veterans Health Administration chiropractic service utilization

et al. 2020

View full text Add to dashboard Cite

Background: Chronic spinal pain conditions affect millions of US adults and carry a high healthcare cost burden, both direct and indirect. Conservative interventions for spinal pain conditions, including chiropractic care, have been associated with lower healthcare costs and improvements in pain status in different clinical populations, including veterans. Little is currently known about predicting healthcare service utilization in the domain of conservative interventions for spinal pain conditions, including the frequency of use of chiropractic services. The purpose of this retrospective cohort study was to explore the use of supervised machine learning approaches to predicting one-year chiropractic service utilization by veterans receiving VA chiropractic care. Methods: We included 19,946 veterans who entered the Musculoskeletal Diagnosis Cohort between October 1, 2003 and September 30, 2013 and utilized VA chiropractic services within one year of cohort entry. The primary outcome was one-year chiropractic service utilization following index chiropractic visit, split into quartiles represented by the following classes: 1 visit, 2 to 3 visits, 4 to 6 visits, and 7 or greater visits. We compared the performance of four multiclass classification algorithms (gradient boosted classifier, stochastic gradient descent classifier, support vector classifier, and artificial neural network) in predicting visit quartile using 158 sociodemographic and clinical features. Results: The selected algorithms demonstrated poor prediction capabilities. Subset accuracy was 42.1% for the gradient boosted classifier, 38.6% for the stochastic gradient descent classifier, 41.4% for the support vector classifier, and 40.3% for the artificial neural network. The micro-averaged area under the precision-recall curve for each oneversus-rest classifier was 0.43 for the gradient boosted classifier, 0.38 for the stochastic gradient descent classifier, 0.43 for the support vector classifier, and 0.42 for the artificial neural network. Performance of each model yielded only a small positive shift in prediction probability (approximately 15%) compared to naïve classification.

show abstract

Section: Discussionmentioning

confidence: 99%

Exploring supervised machine learning approaches to predicting Veterans Health Administration chiropractic service utilization

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The “healthy patients” group was characterised by few prescriptions (including at least 1 prescription for oral iron supplements) and a high deductible that did not change in the next year, indicating a self-assessment of very good health status. Self-reported general health has been found to be an important indicator of future health care utilisation in previous studies [18, 37]. Claims data do not include information on self-reported health, so changes in the deductible may serve as an indicator of patients’ individual expectations regarding upcoming health expenditures.…”

Section: Discussionmentioning

confidence: 99%

Prediction of health care expenditure increase: how does pharmacotherapy contribute?

Jödicke

Zellweger

Tomka

et al. 2019

BMC Health Serv Res

View full text Add to dashboard Cite

BackgroundRising health care costs are a major public health issue. Thus, accurately predicting future costs and understanding which factors contribute to increases in health care expenditures are important. The objective of this project was to predict patients healthcare costs development in the subsequent year and to identify factors contributing to this prediction, with a particular focus on the role of pharmacotherapy.MethodsWe used 2014–2015 Swiss health insurance claims data on 373′264 adult patients to classify individuals’ changes in health care costs. We performed extensive feature generation and developed predictive models using logistic regression, boosted decision trees and neural networks. Based on the decision tree model, we performed a detailed feature importance analysis and subgroup analysis, with an emphasis on drug classes.ResultsThe boosted decision tree model achieved an overall accuracy of 67.6% and an area under the curve-score of 0.74; the neural network and logistic regression models performed 0.4 and 1.9% worse, respectively. Feature engineering played a key role in capturing temporal patterns in the data. The number of features was reduced from 747 to 36 with only a 0.5% loss in the accuracy. In addition to hospitalisation and outpatient physician visits, 6 drug classes and the mode of drug administration were among the most important features. Patient subgroups with a high probability of increase (up to 88%) and decrease (up to 92%) were identified.ConclusionsPharmacotherapy provides important information for predicting cost increases in the total population. Moreover, its relative importance increases in combination with other features, including health care utilisation.

show abstract

“…Each year, the total costs for each participant will be categorised as the bottom 0%–50%, top 11%–50%, top 6%–10%, top 2%–5% and top 1% of healthcare users in Ontario according to predetermined cut-offs from a general population sample. 14 34 …”

Section: Methods and Analysismentioning

confidence: 99%

“…For instance, the Canadian Community Health Survey (CCHS) has been linked to administrative data to predict high-cost users in Ontario, Canada. 14 However, the CCHS is a household survey that does not represent homeless individuals. By excluding a group of individuals who are considered frequent users of health services, these models are likely to underestimate future healthcare resources and system costs.…”

Section: Introductionmentioning

confidence: 99%

Patterns and predictors of high-cost users of the health system: a data linkage protocol to combine a cohort study and randomised controlled trial of adults with a history of homelessness

et al. 2020

View full text Add to dashboard Cite

IntroductionHomelessness is a global issue with a detrimental impact on health. Individuals who experience homelessness are often labelled as frequent healthcare users; yet it is a small group of individuals who disproportionately use the majority of services. This protocol outlines the approach to combine survey data from a prospective cohort study and randomised controlled trial with administrative healthcare data to characterise patterns and predictors of healthcare utilisation among a group of adults with a history of homelessness.Methods and analysisThis cohort study will apply survey data from the Health and Housing in Transition study and the At Home/Chez Soi study linked with administrative healthcare databases in Ontario, Canada. We will use count models to quantify the associations between baseline predisposing, enabling, and need factors and hospitalisations, emergency department visits and physician visits in the following year. Subsequently, we will identify individuals who are high-cost users of the health system (top 5%) and characterise their patterns of healthcare utilisation. Logistic regression will be applied to develop a set of models to predict who will be high-cost users over the next 5 years based on predisposing, enabling and need factors. Calibration and discrimination will be estimated with bootstrapped optimism (bootstrap performance—test performance) to ensure the model performance is not overestimated.Ethics and disseminationThis study is approved by the St Michael’s Hospital Research Ethics Board and the University of Toronto Research Ethics Board. Findings will be disseminated through publication in peer-reviewed journals, presentations at research conferences and brief reports made available to healthcare professionals and the general public.Trial Registration NumberThis is a secondary data analysis of a cohort study and randomized trial. The At Home/Chez Soi study has been registered with the International Standard Randomised Control Trial Number Register and assigned ISRCTN42520374.

show abstract

Predicting High Health Care Resource Utilization in a Single-payer Public Health Care System

Abstract: Supplemental Digital Content is available in the text.

Cited by 33 publications

References 35 publications

Exploring supervised machine learning approaches to predicting Veterans Health Administration chiropractic service utilization

Exploring supervised machine learning approaches to predicting Veterans Health Administration chiropractic service utilization

Prediction of health care expenditure increase: how does pharmacotherapy contribute?

Patterns and predictors of high-cost users of the health system: a data linkage protocol to combine a cohort study and randomised controlled trial of adults with a history of homelessness

Contact Info

Product

Resources

About