Claims-based Measures Research Articles

Assessing 1 year Comorbidity Prevalence and Its Survival Implications in Medicare Beneficiaries Diagnosed with Cancer: Insights from a new SEER-Medicare Resource.

Almost half of Medicare beneficiaries diagnosed with cancer from 1992-2005 had at least one comorbid condition. Conditions impact a range of domains from clinical decision making to quality of life which are important to consider when conducting cancer research. We introduce a new SEER-Medicare resource to facilitate using claims data for cancer patients. We use the SEER-Medicare resource to estimate prevalence of comorbidities, 5-year survival rate by cancer site, stage, age and comorbidity severity, and prevalence of surgery by comorbidity for breast, prostate, colorectal and lung cancer. Overall, the most prevalent comorbidities in the year prior to cancer diagnosis were diabetes (27%), COPD (22%), peripheral vascular disease (14%), and congestive heart failure (12%). Comorbidity severity had a greater impact on the probability of dying from non-cancer causes than from dying from cancer. Severity of comorbidity and age consistently increased the probability of non-cancer death. The percentage of persons receiving surgery tended to be lower among those with severe comorbidity. This study demonstrates the utility of new SEER*stat databases that contain Medicare beneficiaries and claims-based measures of comorbidity. Our results demonstrate that comorbidity is common among older persons diagnosed with cancer and the impact of comorbidity on the probability of dying from cancer varies by cancer site, stage at diagnosis and age. Comorbidity is common among persons with cancer and impacts survival. Future research on the impact of comorbidity among cancer survivors is facilitated by new databases.

Unlocking success: Full year quality and utilization metrics of payer-provider OMH partnership.

41 Background: For more than a decade, CMS and commercial payers have sought to engage provider organizations in value-based cancer care payment models (VBPMs). Success of VBPMs requires payer/provider alignment on quality/utilization measures indicative of high-value care and the efficiency of data exchange between providers and payers. The following describes several key metrics from the first full year 2022 of the Tennessee Oncology (TO) and BlueCross BlueShield of Tennessee (BCBST) Oncology Medical Home (OMH) partnership. Methods: TO and BCBST implemented reporting methodologies for a combination of practice-reported quality and utilization measures and payer claims-based measures: Patient experience, Pathway adherence, Distress screening / interventions, Admissions/1000 patients, ED visits/1000 patients, Imaging utilization, and Observation stays. Bi-weekly meetings between practice and payer were necessary to review quality/utilization measure reporting, coordinate overall program administration, and monitor member attribution for care management and performance-based payments. Results: TO/BCBST co-developed a VBPM that included bi-directional reporting of practice-reported quality and utilization measures and payer claims-based measures. Technical challenges in data sharing and reporting occurred frequently and required regular troubleshooting. Questions regarding trends in observed data arose frequently for both TO and BCBST and required regular discussions between clinical/operational leaders of payer and practice (Table). Conclusions: These data suggest that payers and community oncology practices can co-develop value-based payment models that yield transparency in quality and utilization data reporting, and observed differences in performance and cost-efficiency among different practices in a market. Payers can gain increased transparency in quality and cost-efficiency of cancer care for their members. Practices can leverage care management payments and performance-based payments to support care delivery and quality/utilization data reporting improvements. Provider and payer capabilities to execute transparent QM/UM reporting is a major challenge in implementing OMH programs. 2022 OMH results. Measure % higher or lower than benchmarks Admissions /k -7.0 ED visit /k -8.4 Imaging utilization + 9.4 Observation stays -10.0 Patient experience NA Pathway adherence NA Distress screening / interventions NA Hospice reporting NA NA = not applicable, no benchmark data available/measured; /k = per 1000 patients; negative = lower than benchmarks; positive = higher than benchmarks.

Regional Variation in Deescalated Therapy in Older Adults With Early-Stage Breast Cancer

Although trial data support the omission of axillary surgery and radiation therapy (RT) in women aged 70 years or older with T1N0 hormone receptor-positive (HR+) breast cancer, potential overtreatment in older adults with frailty persists. To determine how much geospatial variation in locoregional therapy may be attributed to region vs patient factors. This retrospective cross-sectional study included women aged 70 years or older who were diagnosed with HR+/ERBB2-negative (ERBB2-) breast cancer from January 1, 2013, to December 31, 2017. Data came from Surveillance, Epidemiology, and End Results-Medicare. Hierarchical multivariable modeling was used to evaluate the variance in deescalated care attributable to 4 domains, ie, (1) random, (2) region (health service area [HSA]), (3) patient factors, and (4) unexplained. Patient factors included age, frailty (validated claims-based measure), Charlson Comorbidity Index (CCI), and socioeconomic status (Yost index). Analyses were performed from January to October 2023. HSA. Deescalated care, defined as omission of axillary surgery, RT, or both. Standard therapy was defined as lumpectomy, axillary surgery, and RT or mastectomy with axillary surgery. Multivariable logistic regression was used to identify factors associated with deescalated care receipt. Of 9173 patients (mean [SD] age, 76.5 [5.2] years), 2363 (25.8%) were aged 80 years or older, 705 (7.7%) had frailty, and 419 (4.6%) had a CCI of 2 or greater. While 4499 (49.1%) underwent standard therapy, 4674 (50.9%) underwent deescalated therapy (1193 [13.0%] of the population omitted axillary surgery and 4342 [55.5%] of patients undergoing lumpectomy omitted RT). Of the total variance, random variation explained 27.3%, region/HSA explained 35.3%, patient factors explained 2.8%, and 34.5% was unexplained. In adjusted models, frailty and increased age were associated with a higher likelihood of undergoing deescalated therapy (frailty: odds ratio [OR], 1.70; 95% CI, 1.43-2.02; age, by 1-year increment: OR, 1.10; 95% CI, 1.09-1.11), but CCI was not. Patients in rural areas compared with those in urban areas (OR, 0.82; 95% CI, 0.68-0.99) and Asian and Pacific Islander patients compared with non-Hispanic White patients (OR, 0.68; 95% CI, 0.54-0.85) had a lower likelihood of undergoing deescalated therapy. In this retrospective cross-sectional study of women aged 70 years or older diagnosed with T1N0 HR+/ERBB2- breast cancer, region/HSA contributed more to the variation in deescalated therapy use than patient factors. Unexplained variation may be attributed to unmeasured characteristics, such as multidisciplinary environment and patient preference. Decision support efforts to address overtreatment should target regions with low rates of evidence-based deescalation.

Potential Overtreatment of DCIS in Patients with Limited Life Expectancy.

As the benefits of intensive locoregional therapy for ductal carcinoma in situ (DCIS) are realized over time in older adults, life expectancy may help to guide treatment decisions. We examined whether life expectancy was associated with extent of locoregional therapy in this population. Women ≥ 70 years old with < 5 cm of DCIS diagnosed 2010-2015 were identified in theSurveillance, Epidemiology, and End Results (SEER)-Medicare datasetand categorized by a life expectancy ≤ 5 or > 5 years, defined by a validated claims-based measure. Differences in locoregional therapy (mastectomy + axillary surgery, mastectomy-only, lumpectomy + radiation therapy (RT) + axillary surgery, lumpectomy + RT, lumpectomy-only, and no treatment) by life expectancy were assessed using Pearson chi-squared tests. Generalized linear mixed models were used to identify factors associated with receipt of lumpectomy-only. Of 5346 women (median age of 75 years, range 70-97 years), 927 (17.3%) had a life expectancy ≤ 5 years. Of the 4041 patients who underwent lumpectomy, 710 (13.3%) underwent axillary surgery. More patients with life expectancy ≤ 5 years underwent lumpectomy-only (39.4% versus 27%), mastectomy-only (8.1% versus 5.3%), or no treatment (5.8% versus 3.2%; p < 0.001). On multivariable analysis, women with life expectancy ≤ 5 years had a significantly greater likelihood of undergoing lumpectomy-only [OR 1.90, 95% CI (1.63-2.22)]. Life expectancy is associated with lower-intensity locoregional therapy for older women with DCIS, yet a large proportion of patients with a life expectancy ≤ 5 years received RT and axillary surgery, highlighting potential overtreatment and opportunities to de-escalate locoregional therapy in older adults.

Access, Utilization, and Quality of Behavioral Health Integration in Medicaid Managed Care

Many states have moved from models that carve out to those that carve in or integrate behavioral health in their Medicaid managed care organizations (MCOs), but little evidence exists about the effect of this change. To assess the association of the transition to integrated managed care (IMC) in Washington Medicaid with health services use, quality, health-related outcomes, and measures associated with social determinants of health. This cohort study used difference-in-differences analyses of Washington State's 2014 to 2019 staggered rollout of IMC on claims-based measures for enrollees in Washington's Medicaid MCO. It was supplemented with interviews of 24 behavioral health agency leaders, managed care administrators, and individuals who were participating in the IMC transition. The data were analyzed between February 1, 2023, and September 30, 2023. Claims-based measures of utilization (including specialty mental health visits and primary care visits); health-related outcomes (including self-harm events); rates of arrests, employment, and homelessness; and additional quality measures. This cohort study included 1 454 185 individuals ages 13 to 64 years (743 668 female [51.1%]; 14 306 American Indian and Alaska Native [1.0%], 132 804 Asian American and Pacific Islander [9.1%], 112 442 Black [7.7%], 258 389 Hispanic [17.8%], and 810 304 White [55.7%] individuals). Financial integration was not associated with changes in claims-based measures of utilization and quality. Most claims-based measures of outcomes were also unchanged, although enrollees with mild or moderate mental illness experienced a slight decrease in cardiac events (-0.8%; 95% CI, -1.4 to -0.2), while enrollees with serious mental illness experienced small decreases in employment (-1.2%; 95% CI -1.9 to -0.5) and small increases in arrests (0.5%; 95% CI, 0.1 to 1.0). Interviews with key informants suggested that financial integration was perceived as an administrative change and did not have substantial implications for how practices delivered care; behavioral health agencies lacked guidance on how to integrate care in behavioral health settings and struggled with new contracts and regulatory policies that may have inhibited the ability to provide integrated care. The results of this cohort study suggest that financial integration at the MCO level was not associated with significant changes in most measures of utilization, quality, outcomes, and social determinants of health. Additional support, including monitoring, training, and funding, may be necessary to drive delivery system changes to improve access, quality, and outcomes.

Preconception care utilization: Self-report versus claims-based measures among women with Medicaid.

The objective of this study is to compare self-reported preconception care utilization (PCU) among Medicaid-covered births to Medicaid claims. We identified all Medicaid-covered births to women ages 15-45 in 26 states in the year 2012 among the Pregnancy Risk Assessment and Monitoring System (PRAMS) survey and Medicaid Analytic eXtract (MAX) claims data, and identified preconception services in the latter using diagnosis codes published by Health and Human Services' Office of Population Affairs. We fit mixed-effects logistic regression models for the probability of PCU on sociodemographic factors (age, race, and ethnicity) and clinical diagnoses (depression, diabetes, or hypertension), separately for each dataset. Among 652,929 women delivering in MAX, 28.1% received at least one claims-based preconception service while an estimated 23.6% (95% CI 22.1-25.3) of PRAMS respondents reported receiving preconception care. Adjusting for age, chronic diseases, and state, PCU rates in both MAX and PRAMS were higher for non-Hispanic Black versus non-Hispanic White women (OR 1.51, 95% CI 1.49-1.54 and OR 2.05, 95% CI 1.60-2.62, respectively). Adjusting for differences in age, race and ethnicity, and state, PCU rates were higher for patients with diabetes (OR 1.34, 95% CI 1.29-1.40 and OR 1.82, 95% CI 1.16-2.85) or hypertension (OR 1.22, 95% CI 1.18-1.27 and OR 1.85, 95% CI 1.41-2.44). While Hispanic and Asian women were also more likely to report PCU than their non-Hispanic White counterparts (OR 2.07, 95% CI 1.53-2.80 and OR 3.37, 95% CI 2.28-4.98), they were less likely to have received it (OR 0.74, 95% CI 0.73-0.75 and OR 0.65, 95% CI 0.63-0.67). In conclusion, comparing self-report to claims measures of PCU, we found similar trends in the differences between non-Hispanic Black and White women, and between those with vs. without diabetes and hypertension. However, the two data sources differed in trends in other racial/ethnic groups (differences between Hispanic vs. non-Hispanic White women, and between Asian vs. non-Hispanic White women), and in those with vs. without depression. This suggests that while Medicaid claims can be a useful tool for studying preconception care, they may miss certain types of care among some sub-groups of the population or be subject to reporting differences that are hard to surmise. Both data sets have potential benefits and drawbacks as research tools.

Abstract 11771: Validity of an Administrative Claims-Based Measure of Low-Value Carotid Revascularization

Introduction: Low value care is defined as care where harms or costs outweigh benefits. Researchers, payers, and policymakers commonly use administrative claims data to identify low-value care, but the validity of many claims-based algorithms has not been assessed. Research Question: What is the validity of a commonly-used claims-based algorithm to identify low-value carotid artery revascularization (CAR) procedures, compared to a reference standard of chart review? Methods: This was a retrospective cohort study of Medicare fee-for-service beneficiaries attributed to a large accountable care organization. We used administrative claims data linked to electronic health record data and identified all CAR procedures (endarterectomy or stenting) from 01/01/2020-12/31/2021. We excluded patients without claims history in the year prior to CAR. Using methods previously described by policymakers and researchers, we classified CAR as low-value by claims if patients did not have diagnosis codes for stroke, transient ischemic attack, or other neurologic symptoms such as vision changes or hemiplegia in the year prior to CAR. We then reviewed linked electronic health records to determine if cases were misclassified by the claims-based measure. Per recent guidelines, we classified CAR as low-value by chart review if it was performed in patients with less than 50% symptomatic stenosis or greater than 80% asymptomatic stenosis. Results: A total of 127 CAR cases (mean age 77 years; 63% men; 87% white) with complete electronic health record data were identified, and 48 (38%) were classified as low-value by the claims-based algorithm. Only 11 of the 48 cases classified as low-value by claims were confirmed to be low-value by chart review. The positive predictive value of the claims-based algorithm was 23%. Conclusions: A commonly-used claims-based measure to identify low-value CAR procedures had poor accuracy with a positive predictive value of 23%. Using administrative claims to identify low-value CAR procedures might overestimate the number of procedures considered to be low-value. This has implications for the current system of quality reporting, particularly when quality is tied to financial incentives.

Abstract 14471: Can Administrative Claims Data Accurately Identify Low-Value Pre-Operative Cardiovascular Testing?

Background: Low-value care is defined as care where harms or costs outweigh the benefits. Low-value care services are often identified using administrative claims data, but the validity of claims-based measures may be limited due to the lack of clinical details. One example of a commonly-used claims-based measure of low-value care is pre-operative stress testing prior to low-risk surgery. Clinical practice guidelines recommend against routine testing since it provides no benefit, but testing may be clinically appropriate in patients with symptoms that warrant further investigation such as chest pain. AIM: To test the validity of a claims-based measure of low-value pre-operative stress testing against a gold standard of chart review. Methods: In this retrospective cohort study, we analyzed Medicare claims data linked to electronic health record data from a large accountable care organization. We applied a currently-used claims-based algorithm to identify patients who received a stress test within 60 days prior to a low-risk surgery (e.g., cataract surgery) in 2021. We then reviewed linked electronic health record data to assess the proportion of tests that were actually low-value or were misclassified (i.e., were clinically appropriate based on patient symptoms). Results: We identified 120 patients (median age 74, 50% female, 78% white) who underwent low-risk surgery and received a pre-operative stress test considered low-value by the claims algorithm. We excluded 35 patients who had insufficient data for chart review. Of the remaining 85patients, we found that 36 (42%) were confirmed low-value while 49 (58%) were actually clinically appropriate due to symptoms that prompted testing. The claims-based algorithm hada positive predictive value (PPV) of 42%. Conclusions: A commonly used claims-based measure for identifying low-value pre-operative CV testing misclassifies a substantial proportion of tests as low-value when they are clinically appropriate. This has significant implications for policy and practice, as it can erode trust in quality reporting systems and discourage clinically appropriate testing.

Abstract 14800: Alzheimer’s Disease and Related Dementias (ADRD) Limits Days at Home Among Medicare Beneficiaries Hospitalized for Heart Failure (HF)

Background: Alzheimer’s disease and related dementias (ADRD) are highly prevalent among adults with heart failure (HF). HF is associated with disability among older adults, especially those with ADRD. This may impact aging in place after HF hospitalizations, which is concerning given this is an important outcome for patients recovering from serious illnesses. We aimed to examine the association between ADRD and aging in place among Medicare beneficiaries hospitalized for HF. Methods: We identified 80,694 fee-for-service beneficiaries hospitalized for HF between 2017 and 2019 in a 20% random sample of Medicare claims. HF hospitalization and ADRD were identified using ICD codes. The primary outcome was restricted home time, a patient-centered claims-based measure of successful aging in place defined as spending ≥ 10 days not at home (e.g., hospitalized, nursing home, inpatient facility) over 6 months post-discharge. Multivariable logistic regression was used to examine the association between ADRD and restricted home time adjusting for demographic and clinical characteristics; secondary analyses evaluated the likelihood of returning to pre-hospitalization days at home 6 months post-discharge. Results: Eighteen percent (n= 14,503/80,694) of adults hospitalized for HF had concurrent ADRD. HF patients with ADRD were older (83 vs 77 years), more likely to be female (56% vs 48%), and have dual Medicaid/Medicare eligibility (27% vs 21%), but had similar hospital length of stays and days at home preceding HF hospitalization compared to those without ADRD. Of HF patients with ADRD, 63% had restricted home time (vs. 51% without ADRD) and 52% failed to return to baseline days at home (vs. 42% without ADRD). In adjusted models, older adults with ADRD had 34% greater odds of experiencing restricted home time (OR=1.34, 1.30-1.40), and 20% lower odds of returning to baseline days at home (OR=0.80, 0.77-0.83) 6 months post- discharge than those without ADRD. Conclusions: HF hospitalizations impact the ability of older adults to successfully age in place, especially among those with ADRD. Tailored post-discharge care solutions are urgently needed to address the challenges faced by HF patients with ADRD.

State- and County-Level Geographic Variation in Opioid Use Disorder, Medication Treatment, and Opioid-Related Overdose Among Medicaid Enrollees

The opioid crisis disproportionately affects Medicaid enrollees, yet little systematic evidence exists regarding how prevalence of and health care utilization for opioid use disorder (OUD) vary across geographical areas. To characterize state- and county-level variation in claims-based prevalence of OUD and rates of medication treatment for OUD and OUD-related nonfatal overdose among Medicaid enrollees. This cross-sectional study used data from the Transformed Medicaid Statistical Information System Analytic Files from January 1, 2016, to December 31, 2018. Participants were Medicaid enrollees with or without OUD in 46 states; Washington, DC; and Puerto Rico who were aged 18 to 64 years and not dually enrolled in Medicare. The analysis was conducted between September 2022 and April 2023. Calendar-year OUD prevalence. The main outcomes were claims-based measures of OUD prevalence and rates of medication treatment for OUD and opioid-related nonfatal overdose. Individual records were aggregated at the state and county level, and variation was assessed within and across states. Of the 76 390 817 Medicaid enrollee-year observations included in our study (mean [SD] enrollee age, 36.5 [1.6] years; 59.0% female), 2 280 272 (3.0%) had a claims-based OUD (mean [SD] age, 38.9 [3.6] years; 51.4% female). Of enrollees with OUD, 41.2% were eligible due to Medicaid expansion, 46.4% had other substance use disorders, 55.8% had mental health conditions, 55.2% had claims indicating some form of OUD medication, and 5.8% had claims indicating an overdose during a calendar year. Claims-based outcomes exhibited substantial variation across states: OUD prevalence ranged from 0.6% in Arkansas and Puerto Rico to 9.7% in Maryland, rates of OUD medication treatment ranged from 17.7% in Kansas to 82.8% in Maine, and rates of overdose ranged from 0.3% in Mississippi to 10.5% in Illinois. Pronounced variation was also found within states (eg, OUD prevalence in Maryland ranged from 2.2% in Prince George's County to 21.6% in Cecil County). In this cross-sectional study of Medicaid enrollees from 2016 to 2018, claims-based prevalence of OUD and rates of OUD medication treatment and opioid-related overdose varied substantially across and within states. Further research appears to be needed to identify important factors influencing this variation.

Accuracy of infection reporting in US nursing home ratings.

To assess the accuracy of nursing home-reported data on urinary tract infections (UTIs), which are publicly reported on Nursing Home Care Compare, and pneumonia, which are not publicly reported. We used secondary data for 100% of Medicare fee-for-service beneficiaries in the United States between 2011 and 2017. We identified Medicare fee-for-service beneficiaries who were nursing home residents between 2011 and 2017 and admitted to a hospital with a primary diagnosis of UTI or pneumonia. After linking these hospital claims to resident-level nursing home-reported assessment data in the Minimum Data Set, we calculated the percentages of infections that were appropriately reported and assessed variation by resident- and nursing home-level characteristics. We developed a claims-based nursing home-level measure of hospitalized infections and estimated correlations between this and publicly reported ratings. Medicare fee-for-service beneficiaries who were nursing home residents and hospitalized for UTI or pneumonia during the study period were included. Reporting rates were low for both infections (UTI: short-stay residents 29.1% and long-stay residents 19.2%; pneumonia: short-stay residents 66.0% and long-stay residents 70.6%). UTI reporting rates increased when counting additional assessments, but it is unclear whether these reports are for the same versus a newly developed UTI. Black residents had slightly lower reporting rates, as did nursing homes with more Black residents. Correlations between our claims-based measure and publicly reported ratings were poor. UTI and pneumonia were substantially underreported in data used for national public reporting. Alternative approaches are needed to improve surveillance of nursing home quality.

Longitudinal care continuity and avoidable hospitalization: the application of claims-based measures

BackgroundLongitudinal continuity between a patient and his/her primary care physician is an important aspect in measuring continuity of care (COC). The majority of previous studies employed questionnaire surveys to patients to measure the continual relationship between patients and their physicians. This study aimed to construct a provider duration continuity index (PDCI) by using longitudinal claims data and to examine its agreement with commonly used COC measures. Then, this study investigated the effects of the various types of COC measure on the likelihood of avoidable hospitalization while considering the level of comorbidity.MethodsThis study constructed a 4-year panel (from 2014 to 2017) of the nationwide health insurance claims data in Taiwan. In total, 328,044 randomly selected patients with 3 or more physician visits per year were analyzed. Two PDCIs were constructed to measure the duration of interaction between a patient and his/her physicians over time. The agreement between the PDCIs and three commonly used COC indicators, the Usual Provider of Care index, the Continuity of Care Index, and the Sequential Continuity Index, were examined. Generalized estimating equations were conducted to examine the association between COC and avoidable hospitalization by the level of comorbidity.ResultsThe results showed that the correlations among the three commonly used COC indicators were high (γ = 0.787 ~ 0.958) and the correlation between the two longitudinal continuity measures was moderate (γ = 0.577 ~ 0.579), but the correlations between the commonly used COC indicators and the two PDCIs were low (γ = 0.001 ~ 0.257). All COC measures, both the PDCIs and the three commonly used COC indicators, showed independent protective effects on the likelihood of avoidable hospitalization in three comorbidity groups.ConclusionThe duration of interaction between patients and physicians is an independent domain in measuring COC and has a significant effect on health care outcomes.

Frailty predicts referral for elder abuse evaluation in a nationwide healthcare system-Results from a case-control study.

Elder abuse (EA) is common and has devastating health impacts. Frailty may increase susceptibility to and consequences of EA for older adults, making healthcare system detection more likely, but this relationship has been difficult to study. We examined the association between a recently validated frailty index and referral to social work (SW) for EA evaluation in the Veterans Administration (VA) healthcare system. We conducted a case-control study of veterans aged ≥60 years evaluated by SW for suspected EA between 2010 and 2018 (n=14,723) and controls receiving VA primary care services in the same 60-day window (n=58,369). We used VA and Medicare claims data to measure frailty (VA Frailty Index) and comorbidity burden (the Elixhauser Comorbidity Index) in the 2 years prior to the index. We used adjusted logistic regression models to examine the association of frailty or comorbidity burden with referral to SW for EA evaluation. We used Akaike Information Criterion (AIC) values to evaluate model fit and likelihood ratio (LR) tests to assess the statistical significance of including frailty and comorbidity in the same model. The sample (n=73,092) had a mean age 72 years; 14% were Black, and 6% were Hispanic. More cases (67%) than controls (36%) were frail. LR tests comparing the nested models were highly significant (p < 0.001), and AIC values indicated superior model fit when including both frailty and comorbidity in the same model. In a model adjusting for comorbidity and all covariates, pre-frailty (aOR vs. robust 1.7; 95% CI 1.5-1.8) and frailty (aOR vs. robust 3.6; 95% CI 3.3-3.9) were independently associated with referral for EA evaluation. A claims-based measure of frailty predicted referral to SW for EA evaluation in a national healthcare system, independent of comorbidity burden. Electronic health record measures of frailty may facilitate EA risk assessment and detection for this important but under-recognized phenomenon.

Medicare beneficiaries with more comprehensive primary care physicians report better primary care.

To examine whether primary care physician (PCP) comprehensiveness is associated with Medicare beneficiaries' overall rating of care from their PCP and staff. We linked Medicare claims with survey data from Medicare beneficiaries attributed to Comprehensive Primary Care Plus (CPC+) physicians and practices. We performed regression analyses of the associations between two claims-based measures of PCP comprehensiveness in 2017 and beneficiaries' rating of care from their PCP and practice staff in 2018. The analytic sample included 6228 beneficiaries cared for by 3898 PCPs. Regressions controlled for beneficiary, physician, practice, and market characteristics. Beneficiaries with more comprehensive PCPs rated care from their PCP and practice staff higher than did those with less comprehensive PCPs. For each comprehensiveness measure, beneficiaries whose PCP was in the 75th percentile were more likely than beneficiaries whose PCP was in the 25th percentile to rate their care highly (2 percentage point difference, p=0.02). Medicare beneficiaries with more comprehensive PCPs rate overall care from their PCPs and staff higher than those with less comprehensive PCPs.

Association of Family Income With Morbidity and Mortality Among US Lower-Income Children and Adolescents

Family income is known to be associated with children's health; the association may be particularly pronounced among lower-income children in the US, who tend to have more limited access to health resources than their higher-income peers. To investigate the association of family income with claims-based measures of morbidity and mortality among children and adolescents in lower-income families in the US enrolled in Medicaid or the Children's Health Insurance Program. This cross-sectional analysis included 795 000 participants aged 5 to 17 years enrolled in Medicaid (Medicaid Analytic eXtract claims, 2011-2012) living in families with income below 200% of the federal poverty threshold (American Community Survey, 2008-2013). Follow-up ended in December 2021. Family income relative to the federal poverty threshold. Record of International Classification of Diseases, Ninth Revision codes for an infection, mental health disorder, injury, asthma, anemia, or substance use disorder and death record within 10 years of observation (Social Security Administration death records through 2021). Among 795 000 individuals in the sample (all statistics weighted: mean [SD] income-to-poverty ratio, 90% [53%]; mean [SD] age, 10.6 [3.9] years; 56% aged 10 to 17 years), 33% had a diagnosed infection, 13% had a mental health disorder, 6% had an injury, 5% had asthma, 2% had anemia, 1% had a substance use disorder, and 0.6% died between 2011 and 2021, with the mean (SD) age at death of 19.8 (4.2) years. For those aged 5 to 9 years, higher family income was associated with lower adjusted prevalence of all outcomes, except mortality: children in families with an additional 100% income relative to the federal poverty threshold had 2.3 (95% CI, 1.8-2.9) percentage points fewer infections, 1.9 (95% CI, 1.5-2.2) percentage points fewer mental health diagnoses, 0.7 (95% CI, 0.5-0.8) percentage points fewer injuries, 0.3 (95% CI, 0.09-0.5) percentage points less asthma, 0.2 (95% CI, 0.08-0.3) percentage points less anemia, and 0.06 (95% CI, 0.03-0.09) percentage points fewer substance use disorder diagnoses. Except for injury and anemia, the associations were more pronounced among those aged 10 to 17 years than those 5 to 9 years (P for interaction <.05). For those aged 10 to 17 years, an additional 100% income relative to the federal poverty threshold was associated with a lower 10-year mortality rate by 0.18 (95% CI, 0.12-0.25) percentage points. Among children and adolescents in the US aged 5 to 17 years with family income under 200% of the federal poverty threshold who accessed health care through Medicaid or the Children's Health Insurance Program, higher family income was significantly associated with a lower prevalence of diagnosed infections, mental health disorders, injury, asthma, anemia, and substance use disorders and lower 10-year mortality. Further research is needed to understand whether these associations are causal.

CLAIMS-BASED FRAILTY INDEX (CFI) AS A MEASURE OF DEMENTIA SEVERITY IN MEDICARE BENEFICIARIES WITH ADRD

Abstract Lack of a dementia severity measure in administrative claims data is a major barrier in Alzheimer’s Disease and Related Dementias (ADRD) research using such data sources. A claims-based frailty index (CFI; range 0 to 1, higher scores indicating greater frailty) is a validated measure of frailty that can be calculated from administrative claims data and is correlated with functional status. We conducted a cross-sectional study to examine whether CFI can be a claims-based measure to differentiate dementia severity in 814 participants with ADRD with linked to Medicare claims in the National Health and Aging Trends Study (NHATS). We estimated the Functional Assessment Staging Test (FAST) scale (3: mild cognitive impairment; 4: mild dementia; 5: moderate dementia; 6–7 moderate-to-severe dementia) using NHATS variables and calculated CFI from Medicare claims in the prior 12 months. The prevalence of FAST stage 5 or higher was 244 (weighted percentage, 25.9%). The C-statistic of CFI to identify FAST stage 5 or higher was 0.78 [95% CI:0.73–0.82], with a CFI cut-point of 0.3 achieving the maximum sensitivity 72.5% and specificity 67.5%. The equipercentile linking procedure showed that CFI scores of 0.30, 0.35, and 0.40 corresponded to FAST stages 4, 5, and 6, respectively. Our results support the utility of a CFI as a proxy of dementia severity in administrative claims data.

Can markers of disease severity improve the predictive power of claims-based multimorbidity indices?

Claims-based measures of multimorbidity, which evaluate the presence of a defined list of diseases, are limited in their ability to predict future outcomes. We evaluated whether claims-based markers of disease severity could improve assessments of multimorbid burden. We developed 7 dichotomous markers of disease severity which could be applied to a range of diseases using claims data. These markers were based on the number of disease-associated outpatient visits, emergency department visits, and hospitalizations made by an individual over a defined interval; whether an individual with a given disease had outpatient visits to a specialist who typically treats that disease; and ICD-9 codes which connote more versus less advanced or symptomatic manifestations of a disease. Using Medicare claims linked with Health and Retirement Study data, we tested whether including these markers improved ability to predict ADL decline, IADL decline, hospitalization, and death compared to equivalent models which only included the presence or absence of diseases. Of 5012 subjects, median age was 76 years and 58% were female. For a majority of diseases tested individually, adding each of the 7 severity markers yielded minimal increase in c-statistic (≤0.002) for outcomes of ADL decline and mortality compared to models considering only the presence versus absence of disease. Gains in predictive power were more substantial for a small number of individual diseases. Inclusion of the most promising marker in multi-disease multimorbidity indices yielded minimal gains in c-statistics (<0.001-0.007) for predicting ADL decline, IADL decline, hospitalization, and death compared to indices without these markers. Claims-based markers of disease severity did not contribute meaningfully to the ability of multimorbidity indices to predict ADL decline, mortality, and other important outcomes.

Meaningful measures for advanced cancer: Building a measure set to promote goal-concordant care.

210 Background: Research suggests that patients with advanced cancer do not reliably receive care consistent with their goals and values, indicating an important gap. While experts promote measurement of ‘goal concordant care’ this construct is difficult to measure due to definitional and data complexity. The 10 cancer hospitals of the Alliance of Dedicated Cancer Centers (ADCC) have collaboratively implemented interventions to improve goal concordant care and create a wholistic measure set through the Improving Goal Concordant Care (IGCC) Initiative. Methods: ADCC launched the 3 year initiative in 2020. Oncology, palliative care and quality experts completed conceptual work prior to launch including identifying core intervention components, specific aims and potential measures. Following an environmental scan, a workgroup brainstormed relevant measures, then applied feasibility and priority ratings. Patient/family advisors provided guidance. In 2020, a measure panel initiated technical development, using existing specifications as feasible. Claims-based measures were modeled. Following alpha testing, quarterly data submission of selected measures began in 2022. Results: During the conceptual phase, the workgroup identified 60 measures. 19 measures remained following the prioritization and feasibility rating; 14 have been tested and implemented to date (Table). A stakeholder (e.g., patients, physicians, administrators) assessment regarding each measure’s value in assessing quality and usefulness for improvement is in progress. Conclusions: In support of IGCC, we created a set of measures to assess progress towards achieving goal-concordant care. This measure set is intended to be useful in any cancer setting.[Table: see text]

Comparison of Low-Value Services Among Medicare Advantage and Traditional Medicare Beneficiaries

Low-value care in the Medicare program is prevalent, costly, potentially harmful, and persistent. Although Medicare Advantage (MA) plans can use managed care strategies not available in traditional Medicare (TM), it is not clear whether this flexibility is associated with lower rates of low-value care. To compare rates of low-value services between MA and TM beneficiaries and explore how elements of insurance design present in MA are associated with the delivery of low-value care. This cross-sectional study analyzed beneficiaries enrolled in MA and TM using claims data from a large, national MA insurer and a random 5% sample of TM beneficiaries. The study period was January 1, 2017, through December 31, 2019. All analyses were conducted from July 2021 to March 2022. Enrollment in MA vs TM. Low-value care was assessed using 26 claims-based measures. Regression models were used to estimate the association between MA enrollment and rates of low-value services while controlling for beneficiary characteristics. Stratified analyses explored whether network design, product design, value-based payment, or utilization management moderated differences in low-value care between MA and TM beneficiaries and among MA beneficiaries. Among a study population of 2 470 199 Medicare beneficiaries (mean [SD] age, 75.6 [7.0] years; 1 346 777 [54.5%] female; 229 107 [9.3%] Black and 2 126 353 [86.1%] White individuals), 1 527 763 (61.8%) were enrolled in MA and 942 436 (38.2%) were enrolled in TM. Beneficiaries enrolled in MA received 9.2% (95% CI, 8.5%-9.8%) fewer low-value services in 2019 than TM beneficiaries (23.1 vs 25.4 total low-value services per 100 beneficiaries). Although MA beneficiaries enrolled in health management organization and preferred provider organization products received fewer low-value services than TM beneficiaries, the difference was largest for those enrolled in health management organization products (2.6 fewer [95% CI, 2.4-2.8] vs 2.1 fewer [95% CI, 1.9-2.3] services per 100 beneficiaries, respectively). Across primary care payment arrangements, MA beneficiaries received fewer low-value services than TM beneficiaries, with the largest difference observed for MA beneficiaries whose primary care physicians were reimbursed within 2-sided risk arrangements. In this cross-sectional study of Medicare beneficiaries, those enrolled in MA had lower rates of low-value care than those enrolled in TM; elements of insurance design present in the MA program and absent in TM were associated with reduction in low-value care.

Assessing patient, physician, and practice characteristics predicting the use of low-value services.

To examine characteristics of beneficiaries, physicians, and their practice sites associated with greater use of low-value services (LVS) using LVS measures that reflect current care practices. This study was conducted in the context of a large, nationwide primary care redesign initiative (Comprehensive Primary Care Plus), using Medicare claims data in 2018. We examined beneficiary-level total counts of LVS based on the existing 31 claims-based measures updated by excluding three services provided with diminishing frequency to Medicare beneficiaries and by replacing these with more recently identified LVS. We estimated hierarchical linear models with an extensive list of beneficiary, physician, and practice site characteristics to examine the contribution of characteristics at each level in predicting greater use of LVS. We also examined the proportion of variation in LVS use attributable to the set of characteristics at each level. The study included 5,074,642 Medicare fee-for-service beneficiaries attributed to 32,406 primary care physicians in 11,009 primary care practice sites. Patients with disabilities, end-stage renal disease, and those in regions with higher poverty rates receive 10 (standard error [SE]=3.0), 80 (SE=14.0), and 10 (SE=1.0) more LVS per 1000 beneficiaries across all 31 measures combined than patients without such attributes, respectively. Greater physician comprehensiveness and an increase in the number of primary care practitioners at a practice were associated with 40 (SE=20.0) and 20 (SE=6.0) fewer LVS per 1000 beneficiaries, respectively. Yet, the explanatory variables we examined only account for 11 percent of the variation in LVS use, with most of the variation (87 percent) being due to unobserved differences at the beneficiary level. Unexplained residual variation, from underlying patient preferences and behavior of non-primary care providers, could be important determinants of LVS use.