Dataset Information

Robust Machine Learning Variable Importance Analyses of Medical Conditions for Health Care Spending.

ABSTRACT: OBJECTIVE:To propose nonparametric double robust machine learning in variable importance analyses of medical conditions for health spending. DATA SOURCES:2011-2012 Truven MarketScan database. STUDY DESIGN:I evaluate how much more, on average, commercially insured enrollees with each of 26 of the most prevalent medical conditions cost per year after controlling for demographics and other medical conditions. This is accomplished within the nonparametric targeted learning framework, which incorporates ensemble machine learning. Previous literature studying the impact of medical conditions on health care spending has almost exclusively focused on parametric risk adjustment; thus, I compare my approach to parametric regression. PRINCIPAL FINDINGS:My results demonstrate that multiple sclerosis, congestive heart failure, severe cancers, major depression and bipolar disorders, and chronic hepatitis are the most costly medical conditions on average per individual. These findings differed from those obtained using parametric regression. CONCLUSIONS:The literature may be underestimating the spending contributions of several medical conditions, which is a potentially critical oversight. If current methods are not capturing the true incremental effect of medical conditions, undesirable incentives related to care may remain. Further work is needed to directly study these issues in the context of federal formulas.

SUBMITTER: Rose S

PROVIDER: S-EPMC6153184 | biostudies-literature | 2018 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Robust Machine Learning Variable Importance Analyses of Medical Conditions for Health Care Spending.

Rose Sherri S

Health services research 20180311 5

<h4>Objective</h4>To propose nonparametric double robust machine learning in variable importance analyses of medical conditions for health spending.<h4>Data sources</h4>2011-2012 Truven MarketScan database.<h4>Study design</h4>I evaluate how much more, on average, commercially insured enrollees with each of 26 of the most prevalent medical conditions cost per year after controlling for demographics and other medical conditions. This is accomplished within the nonparametric targeted learning fram ...[more]

PMID: 29527659

Similar Datasets

Project description:ObjectiveTo understand what patterns of health care use are associated with higher post-hospitalization spending.Data sourcesMedicare hospital, skilled nursing, inpatient rehabilitation, and home health agency claims, and Medicare enrollment data from 2007 and 2008.Study designFor 10 common inpatient conditions, we calculated variation across hospitals in price-standardized and case mix-adjusted Medicare spending in the 30 days following hospital discharge. We estimated the fraction of spending differences between low- and high-spending hospitals attributable to readmissions versus post-acute care, and within post-acute care between inpatient rehabilitation facility (IRF) versus skilled nursing facility (SNF) use. For each service, we distinguished between differences in probability of use and spending conditional on use.Data extraction methodsWe identified index hospital claims and examined hospital and post-acute care occurring within a 30-day period following hospital discharge. For each Medicare Severity Diagnosis-Related Group (MS-DRG) at each hospital, we calculated average price-standardized Medicare payments for readmissions, SNFs, IRFs, and post-acute care overall (also including home health agencies and long-term care hospitals).Principal findingsThere was extensive variation across hospitals in Medicare spending in the 30 days following hospital discharge. For example, the interquartile range across hospitals ranged from $1,245 for chronic obstructive pulmonary disease to over $4,000 for myocardial infarction MS-DRGs. The proportion of differences attributable to readmissions versus post-acute care differed across conditions. For myocardial infarction, 74 to 93 percent of the variation was due to readmissions. For hip and femur procedures and joint replacement, 72 to 92 percent of the variation was due to differences in post-acute care spending. There was also variation in the relative importance of the type of post-acute spending. For hip and femur procedures, joint replacement, and stroke, whether patients received IRF was the key driver of variation in post-acute care spending In contrast, for pneumonia and heart failure, whether patients received SNF care was the key driver of variation in post-acute spending.ConclusionsThrough initiatives such as bundled payment, hospitals are financially responsible for spending in the post-hospitalization period. The key driver of variation in post-hospitalization spending varied greatly across conditions. For some conditions, the key driver was having a readmission, for others it was whether patients receive any post-acute care, and for others the key driver was the type of post-acute care. These findings may help hospitals implement strategies to reduce post-discharge spending.

Dataset Information

Robust Machine Learning Variable Importance Analyses of Medical Conditions for Health Care Spending.

Publications

Robust Machine Learning Variable Importance Analyses of Medical Conditions for Health Care Spending.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets