Patterns of care for clinically distinct segments of high cost Medicare beneficiaries

Patterns of Morbidity Across the Lifespan: A Population Segmentation Framework for Classifying Health Care Needs for All Ages

BACKGROUND

Classification systems to segment such patients into subgroups for purposes of care management and population analytics should balance administrative simplicity with clinical meaning and measurement precision.

OBJECTIVE

To describe and empirically apply a new clinically relevant population segmentation framework applicable to all payers and all ages across the lifespan.

RESEARCH DESIGN AND SUBJECTS

Cross-sectional analyses using insurance claims database for 3.31 Million commercially insured and 1.05 Million Medicaid enrollees under 65 years old; and 5.27 Million Medicare fee-for-service beneficiaries aged 65 and older.

MEASURES

The "Patient Need Groups" (PNGs) framework, we developed, classifies each person within the entire 0-100+ aged population into one of 11 mutually exclusive need-based categories. For each PNG segment, we documented a range of clinical and resource endpoints, including health care resource use, avoidable emergency department visits, hospitalizations, behavioral health conditions, and social need factors.

RESULTS

The PNG categories included: (1) nonuser, (2) low-need child, (3) low-need adult, (4) low-complexity multimorbidity, (5) medium-complexity multimorbidity, (6) low-complexity pregnancy, (7) high-complexity pregnancy, (8) dominant psychiatric/behavioral condition, (9) dominant major chronic condition, (10) high-complexity multimorbidity, and (11) frailty. Each PNG evidenced a characteristic age-related trajectory across the full lifespan. In addition to offering clinically cogent groupings, large percentages (29%-62%) of patients in two pregnancy and high-complexity multimorbidity and frailty PNGs were in a high-risk subgroup (upper 10%) of potential future health care utilization.

CONCLUSIONS

The PNG population segmentation approach represents a comprehensive measurement framework that captures and categorizes available electronic health care data to characterize individuals of all ages based on their needs.

Primary Care Practice Structural Capabilities and Emergency Department Utilization Among High-Need High-Cost Patients

BACKGROUND

US primary care practices are actively identifying strategies to improve outcomes and reduce costs among high-need high-cost (HNHC) patients. HNHC patients are adults with high health care utilization who suffer from multiple chronic medical and behavioral health conditions such as depression or substance abuse. HNHC patients with behavioral health conditions face heightened challenges accessing timely primary care and managing their conditions, which is reflected by their high rates of emergency department (ED) utilization and preventable spending. Structural capabilities (i.e., care coordination, chronic disease registries, shared communication systems, and after-hours care) are key attributes of primary care practices which can enhance access and quality of chronic care delivery.

OBJECTIVE

The purpose of this study was to analyze the association between structural capabilities and ED utilization among HNHC patients with behavioral health conditions.

DESIGN AND MEASURES

We merged cross-sectional survey data on structural capabilities from 240 primary care practices in Arizona and Washington linked with Medicare claims data on 70,182 HNHC patients from 2019.

KEY RESULTS

Using multivariable Poisson models, we found shared communication systems were associated with lower rates of all-cause and preventable ED utilization among HNHC patients with alcohol use (all-cause: aRR 0.72, 95% CI: 0.62, 0.84; preventable: aRR 0.5, 95% CI: 0.40, 0.64) and HNHC patients with substance use disorders (all-cause: aRR 0.76, 95% CI: 0.68, 0.85; preventable: aRR 0.61, 95% CI: 0.52, 0.71). Care coordination was also associated with decreased rates of ED utilization among the overall HNHC population and those with alcohol use, but not among HNHC patients with depression or substance use disorders.

CONCLUSION

Shared communication systems and care coordination have the potential to increase the effectiveness of primary care delivery for specific HNHC patients.

Empiric segmentation of high-risk patients: a structured literature review

OBJECTIVES

Empiric segmentation is a rapidly growing, learning health system approach that uses large health care system data sets to identify groups of high-risk patients who may benefit from similar interventions. We aimed to review studies that used data-driven approaches to segment high-risk patient populations and describe how their designs and findings can inform health care leaders who are interested in applying similar techniques to their patient populations.

STUDY DESIGN

Structured literature review.

METHODS

We searched for original research articles published since 2000 that identified high-risk adult patient populations and applied data-driven analyses to segment the population. Two reviewers independently extracted study population source and criteria for high-risk designation, segmentation method, data types included, model selection criteria, and model results from the identified studies.

RESULTS

Our search identified 224 articles, 12 of which met criteria for full review. Of these, 8 segmented high-risk patients and 4 segmented diagnoses without assigning patients to unique groups. Studies segmenting patients more often had clinically interpretable results. Common groups were defined by high prevalence of diabetes, cardiovascular disease, psychiatric conditions including substance use disorders, and neurologic disease (eg, stroke). Few studies incorporated patients' functional or social factors. Resulting patient and diagnosis clusters varied in ways closely linked to the model inputs, patient population inclusion criteria, and health care system context.

CONCLUSIONS

Empiric segmentation can yield clinically relevant groups of patients with complex medical needs. Segmentation results are context dependent, suggesting the need for careful design and interpretation of segmentation models to ensure that results can inform clinical care and program design in the target setting.

Estimating Population Benefits of Prevention Approaches Using a Risk Tool: High Resource Users in Ontario, Canada

BACKGROUND

Healthcare spending is concentrated, with a minority of the population accounting for the majority of healthcare costs.

METHODS

The authors modelled the impact of high resource user (HRU) prevention strategies within five years using the validated High Resource User Population Risk Tool.

RESULTS

The authors estimated 758,000 new HRUs in Ontario from 2013-2014 to 2018-2019, resulting in $16.20 billion in healthcare costs (Canadian dollars 2016). The prevention approach that had the largest reduction in HRUs was targeting health-risk behaviours.

CONCLUSIONS

This study demonstrates the use of a policy tool by decision makers to support prevention approaches that consider the impact on HRUs and estimated healthcare costs.

A machine learning approach to identify distinct subgroups of veterans at risk for hospitalization or death using administrative and electronic health record data

Background

Identifying individuals at risk for future hospitalization or death has been a major priority of population health management strategies. High-risk individuals are a heterogeneous group, and existing studies describing heterogeneity in high-risk individuals have been limited by data focused on clinical comorbidities and not socioeconomic or behavioral factors. We used machine learning clustering methods and linked comorbidity-based, sociodemographic, and psychobehavioral data to identify subgroups of high-risk Veterans and study long-term outcomes, hypothesizing that factors other than comorbidities would characterize several subgroups.

Methods and findings

In this cross-sectional study, we used data from the VA Corporate Data Warehouse, a national repository of VA administrative claims and electronic health data. To identify high-risk Veterans, we used the Care Assessment Needs (CAN) score, a routinely-used VA model that predicts a patient’s percentile risk of hospitalization or death at one year. Our study population consisted of 110,000 Veterans who were randomly sampled from 1,920,436 Veterans with a CAN score≥75^th percentile in 2014. We categorized patient-level data into 119 independent variables based on demographics, comorbidities, pharmacy, vital signs, laboratories, and prior utilization. We used a previously validated density-based clustering algorithm to identify 30 subgroups of high-risk Veterans ranging in size from 50 to 2,446 patients. Mean CAN score ranged from 72.4 to 90.3 among subgroups. Two-year mortality ranged from 0.9% to 45.6% and was highest in the home-based care and metastatic cancer subgroups. Mean inpatient days ranged from 1.4 to 30.5 and were highest in the post-surgery and blood loss anemia subgroups. Mean emergency room visits ranged from 1.0 to 4.3 and were highest in the chronic sedative use and polysubstance use with amphetamine predominance subgroups. Five subgroups were distinguished by psychobehavioral factors and four subgroups were distinguished by sociodemographic factors.

Conclusions

High-risk Veterans are a heterogeneous population consisting of multiple distinct subgroups–many of which are not defined by clinical comorbidities–with distinct utilization and outcome patterns. To our knowledge, this represents the largest application of ML clustering methods to subgroup a high-risk population. Further study is needed to determine whether distinct subgroups may benefit from individualized interventions.

No Association Between Pharmacogenomics Variants and Hospital and Emergency Department Utilization: A Mayo Clinic Biobank Retrospective Study

Background

The use of pharmacogenomics data is increasing in clinical practice. However, it is unknown if pharmacogenomics data can be used more broadly to predict outcomes like hospitalization or emergency department (ED) visit. We aim to determine the association between selected pharmacogenomics phenotypes and hospital utilization outcomes (hospitalization and ED visits).

Methods

This cohort study utilized 10,078 patients from the Mayo Clinic Biobank in the RIGHT protocol with sequence and interpreted phenotypes for 10 selected pharmacogenes including CYP2D6, CYP2C9, CYP2C19, CYP3A5, HLA B 5701, HLA B 5702, HLA B 5801, TPMT, SLCO1B1, and DPYD. The primary outcome was hospitalization with ED visits as a secondary outcome. We used Cox proportional hazards model to test the association between each pharmacogenomics phenotype and the risk of the outcomes.

Results

During the follow-up period (median [in years] = 7.3), 13% (n=1354) and 8% (n=813) of the subjects experienced hospitalization and ED visits, respectively. Compared to subjects who did not experience hospitalization, hospitalized patients were older (median age [in years]: 67 vs 65), poorer self-rated health (15% vs 4.7% for fair/poor), and higher disease burden (median number of chronic conditions: 7 vs 4) at baseline. There was no association of hospitalization with any of the pharmacogenomics phenotypes. The pharmacogenomics phenotypes were not associated with disease burden, a well-established risk factor for hospital utilization outcomes. Similar findings were observed for patients with ED visits during the follow-up period.

Conclusion

We found no association of 10 well-established pharmacogenomics phenotypes with either hospitalization or ED visits in this relatively large biobank population and outside the context of specific drug use related to these genes. Traditional risk factors for hospitalization like age and self-rated health were much more likely to predict hospitalization and/or ED visits than this pharmacogenomics information.

Describing the characteristics and healthcare use of high-cost acute care users at the end of life: a pan-Canadian population-based study

BACKGROUND

A minority of individuals use a large portion of health system resources, incurring considerable costs, especially in acute-care hospitals where a significant proportion of deaths occur. We sought to describe and contrast the characteristics, acute-care use and cost in the last year of life among high users and non-high users who died in hospitals across Canada.

METHODS

We conducted a population-based retrospective-cohort study of Canadian adults aged ≥18 who died in hospitals across Canada between fiscal years 2011/12-2014/15. High users were defined as patients within the top 10% of highest cumulative acute-care costs in each fiscal year. Patients were categorized as: persistent high users (high-cost in death year and year prior), non-persistent high users (high-cost in death year only) and non-high users (never high-cost). Discharge abstracts were used to measure characteristics and acute-care use, including number of hospitalizations, admissions to intensive-care-unit (ICU), and alternate-level-of-care (ALC).

RESULTS

We identified 191,310 decedents, among which 6% were persistent high users, 41% were non-persistent high users, and 46% were non-high users. A larger proportion of high users were male, younger, and had multimorbidity than non-high users. In the last year of life, persistent high users had multiple hospitalizations more often than other groups. Twenty-eight percent of persistent high users had ≥2 ICU admissions, compared to 8% of non-persistent high users and only 1% of non-high users. Eleven percent of persistent high users had ≥2 ALC admissions, compared to only 2% of non-persistent high users and < 1% of non-high users. High users received an in-hospital intervention more often than non-high users (36% vs. 19%). Despite representing only 47% of the cohort, persistent and non-persistent high users accounted for 83% of acute-care costs.

CONCLUSIONS

High users - persistent and non-persistent - are medically complex and use a disproportionate amount of acute-care resources at the end of life. A greater understanding of the characteristics and circumstances that lead to persistently high use of inpatient services may help inform strategies to prevent hospitalizations and off-set current healthcare costs while improving patient outcomes.

Population segments as a tool for health care performance reporting: an exploratory study in the Canadian province of British Columbia

Background

Primary care serves all age groups and individuals with health states ranging from those with no chronic conditions to those who are medically complex, or frail and approaching the end of life. For information to be actionable and guide planning, there must be some population disaggregation based on differences in expected needs for care. Promising approaches to segmentation in primary care reflect both the breadth and severity of health states, the types and amounts of health care utilization that are expected, and the roles of the primary care provider. The purpose of this study was to assess population segmentation as a tool to create distinct patient groups for use in primary care performance reporting.

Methods

This cross-sectional study used administrative data (patient characteristics, physician and hospital billings, prescription medicines data, emergency department visits) to classify the population of British Columbia (BC), Canada into one of four population segments: low need, multiple morbidities, medically complex, and frail. Each segment was further classified using socioeconomic status (SES) as a proxy for patient vulnerability. Regression analyses were used to examine predictors of health care use, costs and selected measures of primary care attributes (access, continuity, coordination) by segment.

Results

Average annual health care costs increased from the low need ($ 1460) to frail segment ($10,798). Differences in primary care cost by segment only emerged when attributes of primary care were included in regression models: accessing primary care outside business hours and discontinuous primary care (≥5 different GP’s in a given year) were associated with higher health care costs across all segments and higher continuity of care was associated with lower costs in the frail segment (cost ratio = 0.61). Additionally, low SES was associated with higher costs across all segments, but the difference was largest in the medically complex group (cost ratio = 1.11).

Conclusions

Population segments based on expected need for care can support primary care measurement and reporting by identifying nuances which may be lost when all patients are grouped together. Our findings demonstrate that variables such as SES and use of regression analyses can further enhance the usefulness of segments for performance measurement and reporting.

Persistence of High-Need Status Over Time Among Fee-for-Service Medicare Beneficiaries

Although administrative claims data can be used to identify high-need (HN) Medicare beneficiaries, persistence in HN status among beneficiaries and subsequent variation in outcomes are unknown. We use national-level claims data to classify Fee-for-Service (FFS) Medicare beneficiaries as HN annually among beneficiaries continuously enrolled between 2013 and 2015. To examine persistence of HN status over time, we categorize longitudinal patterns in HN status into being never, newly, transiently, and persistently HN and examine differences in patients' demographic characteristics and outcomes. Among survivors, 23% of beneficiaries were HN at any time-4% persistently HN, 13% transiently HN, and 6% newly HN. While beneficiaries who were persistently HN had higher mortality, utilization, and expenditures, classification as HN at any time was associated with poor outcomes. These findings demonstrate longitudinal variability of HN status among FFS beneficiaries and reveal the pervasiveness of poor outcomes associated with even transitory HN status over time.

Developing an actionable patient taxonomy to understand and characterize high-cost Medicare patients

BACKGROUND

Improving care for high-cost patients requires a better understanding of their characteristics and the ability to effectively target interventions. We developed an actionable taxonomy with clinically meaningful patient categories for high-cost Medicare patients-those in the top 10% of total costs.

METHODS

A cross-sectional study of a Medicare fee-for-service (FFS) patient cohort in the New York metropolitan area. We merged claims and neighborhood social determinants of health data to map patients into actionable categories.

RESULTS

Among 428,024 Medicare FFS patients, we mapped the 42,802 high-cost patients into ten overlapping categories, including: multiple chronic conditions, seriously ill, frail, serious mental illness, single condition with high pharmacy cost, chronic pain, end-stage renal disease (ESRD), single high-cost chronic condition, opioid use disorder, and socially vulnerable. Most high-cost patients had multiple chronic conditions (97.4%), followed by serious illness (53.7%) and frailty (48.9%). Patients with ESRD, who were seriously ill, and who were frail were more likely to be high-cost compared to patients in other categories. 72.7% of high-cost patients fell into multiple categories.

CONCLUSIONS

High-cost patients are highly heterogeneous. A patient taxonomy incorporating medical, behavioral, and social characteristics may help providers better understand their characteristics and health needs.

IMPLICATIONS

Mapping high-cost patients into clinically meaningful and actionable categories that incorporate medical, behavioral, and social factors could help health systems target interventions. Integrated approaches, including medical care, behavioral health, and social services may be needed to effectively and efficiently care for high-cost patients.

Unplanned index hospital admissions among new older high-cost health care users in Ontario: a population-based matched cohort study

BACKGROUND

Most health care spending is concentrated within a small group of high-cost health care users. To inform health policies, we examined the characteristics of index hospital admissions and their predictors among incident older high-cost users compared to older non-high-cost users in Ontario.

METHODS

Using Ontario administrative data, we identified incident high-cost users aged 66 years or more and matched them 1:3 on age, gender and Local Health Integration Network with non-high-cost users aged 66 years or more. We defined high-cost users as patients within the top 5% most costly high-cost users during fiscal year 2013/14 but not during 2012/13. An index hospital admission, the main outcome, was defined as the first unplanned hospital admission during 2013/14, with no hospital admissions in the preceding 12 months. Descriptively, we analyzed the attributes of index hospital admissions, including costs. We identified predictors of index hospital admissions using stratified logistic regression.

RESULTS

Over half (95 375/175 847 [54.2%]) of all high-cost users had an unplanned index hospital admission, compared to 8838/527 541 (1.7%) of non-high-cost users. High-cost users had a poorer health status, longer acute length of stay (mean 7.5 d v. 2.9 d) and more frequent designation as alternate level of care before discharge (20.8% v. 1.7%) than did non-high-cost users. Ten diagnosis codes accounted for roughly one-third of the index hospital admission costs in both cohorts. Although many predictors were similar between the cohorts, a lower risk of an index hospital admission was associated with residence in long-term care, attachment to a primary care provider and recent consultation by a geriatrician among high-cost users.

INTERPRETATION

The high prevalence of index hospital admissions and the corresponding costs are a distinctive feature of incident older high-cost users. Improved access to specialist outpatient care, home-based social care and long-term care when required are worth further investigation.

Subgroups of High-Cost Medicare Advantage Patients: an Observational Study

BACKGROUND

There is a growing focus on improving the quality and value of health care delivery for high-cost patients. Compared to fee-for-service Medicare, less is known about the clinical composition of high-cost Medicare Advantage populations.

OBJECTIVE

To describe a high-cost Medicare Advantage population and identify clinically and operationally significant subgroups of patients.

DESIGN

We used a density-based clustering algorithm to group high-cost patients (top 10% of spending) according to 161 distinct demographic, clinical, and claims-based variables. We then examined rates of utilization, spending, and mortality among subgroups.

PARTICIPANTS

Sixty-one thousand five hundred forty-six Medicare Advantage beneficiaries.

MAIN MEASURES

Spending, utilization, and mortality.

KEY RESULTS

High-cost patients (n = 6154) accounted for 55% of total spending. High-cost patients were more likely to be younger, male, and have higher rates of comorbid illnesses. We identified ten subgroups of high-cost patients: acute exacerbations of chronic disease (mixed); end-stage renal disease (ESRD); recurrent gastrointestinal bleed (GIB); orthopedic trauma (trauma); vascular disease (vascular); surgical infections and other complications (complications); cirrhosis with hepatitis C (liver); ESRD with increased medical and behavioral comorbidity (ESRD+); cancer with high-cost imaging and radiation therapy (oncology); and neurologic disorders (neurologic). The average number of inpatient days ranged from 3.25 (oncology) to 26.09 (trauma). Preventable spending (as a percentage of total spending) ranged from 0.8% (oncology) to 9.5% (complications) and the percentage of spending attributable to prescription medications ranged from 7.9% (trauma and oncology) to 77.0% (liver). The percentage of patients who were persistently high-cost ranged from 11.8% (trauma) to 100.0% (ESRD+). One-year mortality ranged from 0.0% (liver) to 25.8% (ESRD+).

CONCLUSIONS

We identified clinically distinct subgroups of patients within a heterogeneous high-cost Medicare Advantage population using cluster analysis. These subgroups, defined by condition-specific profiles and illness trajectories, had markedly different patterns of utilization, spending, and mortality, holding important implications for clinical strategy.

Applying Machine Learning Algorithms to Segment High-Cost Patient Populations

BACKGROUND

Efforts to improve the value of care for high-cost patients may benefit from care management strategies targeted at clinically distinct subgroups of patients.

OBJECTIVE

To evaluate the performance of three different machine learning algorithms for identifying subgroups of high-cost patients.

DESIGN

We applied three different clustering algorithms-connectivity-based clustering using agglomerative hierarchical clustering, centroid-based clustering with the k-medoids algorithm, and density-based clustering with the OPTICS algorithm-to a clinical and administrative dataset. We then examined the extent to which each algorithm identified subgroups of patients that were (1) clinically distinct and (2) associated with meaningful differences in relevant utilization metrics.

PARTICIPANTS

Patients enrolled in a national Medicare Advantage plan, categorized in the top decile of spending (n = 6154).

MAIN MEASURES

Post hoc discriminative models comparing the importance of variables for distinguishing observations in one cluster from the rest. Variance in utilization and spending measures.

KEY RESULTS

Connectivity-based, centroid-based, and density-based clustering identified eight, five, and ten subgroups of high-cost patients, respectively. Post hoc discriminative models indicated that density-based clustering subgroups were the most clinically distinct. The variance of utilization and spending measures was the greatest among the subgroups identified through density-based clustering.

CONCLUSIONS

Machine learning algorithms can be used to segment a high-cost patient population into subgroups of patients that are clinically distinct and associated with meaningful differences in utilization and spending measures. For these purposes, density-based clustering with the OPTICS algorithm outperformed connectivity-based and centroid-based clustering algorithms.

Segmentation of High-Cost Adults in an Integrated Healthcare System Based on Empirical Clustering of Acute and Chronic Conditions

BACKGROUND

High-cost patients are a frequent focus of improvement projects based on primary care and other settings. Efforts to characterize high-cost, high-need patients are needed to inform care planning, but such efforts often rely on a priori assumptions, masking underlying complexities of a heterogenous population.

OBJECTIVE

To define recognizable subgroups of patients among high-cost adults based on clinical conditions, and describe their survival and future spending.

DESIGN

Retrospective observational cohort study.

PARTICIPANTS

Within a large integrated delivery system with 2.7 million adult members, we selected the top 1% of continuously enrolled adults with respect to total healthcare expenditures during 2010.

MAIN MEASURES

We used latent class analysis to identify clusters of alike patients based on 53 hierarchical condition categories. Prognosis as measured by healthcare spending and survival was assessed through 2014 for the resulting classes of patients.

RESULTS

Among 21,183 high-cost adults, seven clinically distinctive subgroups of patients emerged. Classes included end-stage renal disease (12% of high-cost population), cardiopulmonary conditions (17%), diabetes with multiple comorbidities (8%), acute illness superimposed on chronic conditions (11%), conditions requiring highly specialized care (14%), neurologic and catastrophic conditions (5%), and patients with few comorbidities (the largest class, 33%). Over 4 years of follow-up, 6566 (31%) patients died, and survival in the classes ranged from 43 to 88%. Spending regressed to the mean in all classes except the ESRD and diabetes with multiple comorbidities groups.

CONCLUSIONS

Data-driven characterization of high-cost adults yielded clinically intuitive classes that were associated with survival and reflected markedly different healthcare needs. Relatively few high-cost patients remain persistently high cost over 4 years. Our results suggest that high-cost patients, while not a monolithic group, can be segmented into few subgroups. These subgroups may be the focus of future work to understand appropriateness of care and design interventions accordingly.