Characterizing Subgroups of High-Need, High-Cost Patients Based on Their Clinical Conditions: a Machine Learning-Based Analysis of Medicaid Claims Data

Nuti, Sudhakar V.; Doupé, Patrick; Villanueva, Blanca; Scarpa, Joseph R.; Bruzelius, Emilie; Baum, Aaron

doi:10.1007/s11606-019-04941-8

Cited by 4 publications

(3 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… 33 , 34 , 35 Moreover, studies relying largely on diagnosis codes, while able to identify which medical conditions cooccur, may lack the ability to distinguish patients based on severity of illness, functional status, medication needs, or nonmedical factors that are associated with health outcomes and therefore are limited in their ability to suggest specific care innovations. 36 , 37 , 38 , 39 Results from diagnosis code analyses are useful for identifying which conditions are enriched or cooccur in high-cost conditions (eg, end-stage renal disease, diabetes with multiple comorbidities, and acute on chronic illness) but are less helpful for designing care around patient-centered phenotypes (eg, frail elderly adults, chronic pain management). 40 In addition, we linked our analytic results to clinical interpretation.…”

Section: Discussionmentioning

confidence: 99%

“…Other large studies in this area have been limited by reliance on administrative data for large general populations (leading to self-evident divisions between healthier and less healthy categories), and studies with greater depth of clinical data have been limited to smaller populations . Moreover, studies relying largely on diagnosis codes, while able to identify which medical conditions cooccur, may lack the ability to distinguish patients based on severity of illness, functional status, medication needs, or nonmedical factors that are associated with health outcomes and therefore are limited in their ability to suggest specific care innovations . Results from diagnosis code analyses are useful for identifying which conditions are enriched or cooccur in high-cost conditions (eg, end-stage renal disease, diabetes with multiple comorbidities, and acute on chronic illness) but are less helpful for designing care around patient-centered phenotypes (eg, frail elderly adults, chronic pain management) .…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

et al. 2020

View full text Add to dashboard Cite

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

et al. 2020

View full text Add to dashboard Cite

show abstract

“…14 Hayes et al segmented individuals into groups based on chronic diseases and functional limitations and found that higher needs (defined as three or more chronic diseases and functional limitations) were linked to greater healthcare spending and out-of-pocket costs. 17 Previous studies have focused on segmenting individuals using a single time window; 8,14,15,18 less work has focused on the multi-year temporal persistence of high-cost individuals. The present study segments the top cost decile of MA enrollees based on spending patterns and clinical criteria derived from existing literature.…”

Section: Introductionmentioning

confidence: 99%

Temporal Patterns of High-Spend Subgroups Can Inform Service Strategy for Medicare Advantage Enrollees

et al. 2021

View full text Add to dashboard Cite

Background Most healthcare costs are concentrated in a small proportion of individuals with complex social, medical, behavioral, and clinical needs that are poorly met by a fee-for-service healthcare system. Efforts to reduce cost in the top decile have shown limited effectiveness. Understanding patient subgroups within the top decile is a first step toward designing more effective and targeted interventions. Objective Segment the top decile based on spending and clinical characteristics and examine the temporal movement of individuals in and out of the top decile. Design Retrospective claims data analysis. Participants UnitedHealthcare Medicare Advantage (MA) enrollees (N = 1,504,091) continuously enrolled from 2016 to 2019. Main Measures Medical (physician, inpatient, outpatient) and pharmacy claims for services submitted for third-party reimbursement under Medicare Advantage, available as International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) and National Drug Codes (NDC) claims. Key Results The top decile was segmented into three distinct subgroups characterized by different drivers of cost: (1) Catastrophic: acute events (acute myocardial infarction and hip/pelvic fracture), (2) persistent: medications, and (3) semi-persistent chronic conditions and frailty indicators. These groups show different patterns of spending across time. Each year, 79% of the catastrophic group dropped out of the top decile. In contrast, 68–70% of the persistent group and 36–37% of the semi-persistent group remained in the top decile year over year. These groups also show different 1-year mortality rates, which are highest among semi-persistent members at 17.5–18.5%, compared to 12% and 13–14% for catastrophic and persistent members, respectively. Conclusions The top decile consists of subgroups with different needs and spending patterns. Interventions to reduce utilization and expenditures may show more effectiveness if they account for the different characteristics and care needs of these subgroups.

show abstract

Caring for high-need patients

Hempel,

Bolshakova,

Hochman

et al. 2023

BMC Health Serv Res

View full text Add to dashboard Cite

Objective We aimed to explore the construct of “high need” and identify common need domains among high-need patients, their care professionals, and healthcare organizations; and to describe the interventions that health care systems use to address these needs, including exploring the potential unintended consequences of interventions. Methods We conducted a modified Delphi panel informed by an environmental scan. Expert stakeholders included patients, interdisciplinary healthcare practitioners (physicians, social workers, peer navigators), implementation scientists, and policy makers. The environmental scan used a rapid literature review and semi-structured interviews with key informants who provide healthcare for high-need patients. We convened a day-long virtual panel meeting, preceded and followed by online surveys to establish consensus. Results The environmental scan identified 46 systematic reviews on high-need patients, 19 empirical studies documenting needs, 14 intervention taxonomies, and 9 studies providing construct validity for the concept “high need.” Panelists explored the construct and terminology and established that individual patients’ needs are unique, but areas of commonality exist across all high-need patients. Panelists agreed on 11 domains describing patient (e.g., social circumstances), 5 care professional (e.g., communication), and 8 organizational (e.g., staffing arrangements) needs. Panelists developed a taxonomy of interventions with 15 categories (e.g., care navigation, care coordination, identification and monitoring) directed at patients, care professionals, or the organization. The project identified potentially unintended consequences of interventions for high-need patients, including high costs incurred for patients, increased time and effort for care professionals, and identification of needs without resources to respond appropriately. Conclusions Care for high-need patients requires a thoughtful approach; differentiating need domains provides multiple entry points for interventions directed at patients, care professionals, and organizations. Implementation efforts should consider outlined intended and unintended downstream effects on patients, care professionals, and organizations.

show abstract

Characterizing Subgroups of High-Need, High-Cost Patients Based on Their Clinical Conditions: a Machine Learning-Based Analysis of Medicaid Claims Data

Cited by 4 publications

References 4 publications

Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

Use of Latent Class Analysis and k-Means Clustering to Identify Complex Patient Profiles

Temporal Patterns of High-Spend Subgroups Can Inform Service Strategy for Medicare Advantage Enrollees

Caring for high-need patients

Contact Info

Product

Resources

About