Hierarchical Condition Category Research Articles

Measuring substance use disorder (SUD) in commercial and Medicaid insured patients with cancer in Washington state.

177 Background: Studies have shown that SUDs not only impact cancer risk, treatment, and survivorship but also increase ED and hospital use. We sought to characterize SUD within Medicaid enrollees diagnosed with cancer in Washington state, hypothesizing thatMedicaid enrollees were more likely to be diagnosed with SUDs than commercial enrollees, and that patients with an SUD experienced higher rates of emergency department (ED) visits or inpatient (IP) stays during initial cancer treatment. Methods: We linked Washington State cancer registry records with claims data from two large commercial insurers and Medicaid. Eligibility criteria included patients under 65 diagnosed with a solid tumor between 2013-2021 who started systemic therapy within 6 months of diagnosis. SUD diagnoses were identified using codes from CMS’s hierarchical condition category for SUD in the 6 months following diagnosis. Chi-square tests were used to compare demographic characteristics of those with and without a SUD diagnosis. Multivariate logistic regressions were used to evaluate the association between SUD and likelihood of at least one ED visit or IP stay, after controlling for payer, sex, race, ethnicity, age, marital status, stage, cancer group, area deprivation index, and diagnosis year. Results: Among the 10,709 patients (6,680 Commercial, 3,718 Medicaid, 311 Multiple) who received systemic therapy, 7.2% had a SUD diagnosis. SUD patients were less likely to be Asian and more likely to be male, living without a partner, diagnosed at a later cancer stage, and living in more deprived areas. Medicaid patients had significantly higher rates of SUD than commercial-insured patients (14.3% vs 3.3%). Overall, 46.6% of cancer patients had an ED visit or IP stay. Patients with SUD had greater use of ED or IP visits than patients without SUD (70.1% and 44.8%, respectively). In the first 6 months of systemic therapy patients with SUD were significantly more likely to visit the ED or be admitted to IP (OR 2.12, 95% CI 1.79 - 2.50) than cancer patients without SUD. Conclusions: More than 1 in 7 Medicaid cancer patients in WA state has an identified SUD. SUD is associated with significant excess risk for ED/IP admissions during treatment. Accurate identification of SUD in patients is necessary prior to strategies aimed at mitigation of unplanned ED visits and hospitalizations during cancer care. Further research is needed on interventions that effectively reduce the risk during treatment.[Table: see text]

Expanding the Role of the Surgical Preoperative Evaluation Clinic: Impact on Risk and Quality Outcome Measures

ObjectiveTo prove that inpatient-adjusted surgical risk and quality outcome measures can be considerably impacted by interventions to improve documentation in the preoperative evaluation (POE) clinic. Patients and MethodsWe designed a quality improvement project with a multidisciplinary team in our POE clinic to more accurately reflect surgical risk and impact expected surgical quality outcomes through improved documentation. Interventions included an improved patient record acquisition process and extensive POE provider education regarding patient comorbidities’ documentation. For patients admitted after their planned operations, POE clinic comprehensive evaluation notes were linked to inpatient History and Physical notes. High complexity patients seen from October 1, 2018 to December 31, 2018 were the preintervention cohort, and the patients seen from January 1, 2019 to December 31, 2019 were the postintervention cohort. ResultsThe primary outcome measures included the total number of coded diagnoses per encounter and the number of coded hierarchical condition categories per encounter. The secondary outcomes included the calculated severity of illness, risk of mortality, case-mix index, and risk-adjustment factor. Postintervention results show statistically significant increases in all primary outcomes with a P<.05. All secondary outcome measures reported positive change. ConclusionOur interventions confirm that a comprehensive POE and thorough documentation provide a more accurate clinical depiction of the preoperative patient, which in turn impacts quality outcomes in inpatient surgical settings. These results are impactful for direct and indirect patient care and publicly reported hospital and provider level performance data.

Changes in Diagnoses and Site of Care for Patients Receiving Hospice Care From Agencies Acquired by Private Equity Firms and Publicly Traded Companies

Private equity firms and publicly traded companies have been acquiring US hospice agencies; an estimated 16% of US hospice agencies are owned by private equity (PE) firms or publicly traded companies (PTC). To examine the association of PE and PTC acquisitions of hospices with Medicare patients' site of care and clinical diagnoses. This cohort study of US hospice agencies used a novel national database of acquisitions merged with the Medicare Post-Acute Care and Hospice Public Use File for 2013 to 2020. Changes in sites of care and patient characteristics for hospice agencies acquired by PE or PTCs were compared with changes for patients in nonacquired for-profit hospice agencies. Private equity and publicly traded company acquisitions. This study used a difference-in-differences approach within an event-study framework to examine the association of PE and PTC acquisitions of hospice agencies with changes in patient diagnoses and sites of care. Dependent variables were annual hospice-level measures of the Hierarchical Condition Category (HCC) score and proportion of patients diagnosed with cancer or dementia. Sites of care included the proportion of patients receiving hospice care in their personal home, nursing home, or assisted living facility. A total of 158 hospice agencies acquired by PEs, 250 acquired by PTCs, and 1559 other for-profit hospice agencies were included. Preacquisition, hospice agencies that would later be acquired by PE or PTC served a mean (IQR) 30.1% (12.0%-44.0%) and 29.4% (13.0%-43.0%) of their patients in nursing homes respectively, a greater proportion compared with the 27.1% (8.0%-43.8%) served by for-profit hospices that were never acquired. Agencies acquired by PE between 2014 and 2019 saw a significant relative increase of 5.98% in dementia patients (1.38 percentage points; 95% CI, 0.35-2.40 percentage points; P = .008). In PTC-owned hospices, the proportion of patients receiving care at home increased by 5.26% (2.98 percentage points; 95% CI, 1.46-4.51 percentage points; P < .001), the proportion of dementia patients rose by 13.49% (3.11 percentage points; 95% CI, 2.14-4.09 percentage points; P < .001), and the HCC score decreased by 1.37% (-3.19 percentage points; 95% CI, -5.92 to -0.47 percentage points; P = .02). These findings suggest that PE and PTCs select patients and sites of care to maximize profits.

Trends and outcomes of outpatient total shoulder arthroplasty after its removal from CMS's inpatient-only list

In January 2021, the US Medicare program approved reimbursement of outpatient total shoulder arthroplasties (TSA), including anatomic and reverse TSAs. It remains unclear whether shifting TSAs from the inpatient to outpatient setting has affected clinical outcomes. Herein, we describe the rate of outpatient TSA growth and compare inpatient and outpatient TSA complications, readmissions, and mortality. Medicare fee-for-service claims for 2019-2022Q1 were analyzed to identify the trends in outpatient TSAs and to compare 90-day postoperative complications, all-cause hospital readmissions, and mortality between outpatients and inpatients. Outpatient cases were defined as those discharged on the same day of the surgery. To reduce the COVID-19 pandemic's impact and selection bias, we excluded 2020Q2-Q4 data and used propensity scores to match 2021-2022Q1 outpatients with inpatients from the same period (the primary analysis) and from 2019-2020Q1 (the secondary analysis), respectively. We performed both propensity score-matched and -weighted multivariate analyses to compare outcomes between the two groups. Covariates included sociodemographics, preoperative diagnosis, comorbid conditions, the Hierarchical Condition Category risk score, prior year hospital/skilled nursing home admissions, annual surgeon volume, and hospital characteristics. Nationally, the proportion of outpatient TSAs increased from 3% (619) in 2019Q1 to 22% (3456) in 2021Q1 and 38% (6778) in 2022Q1. A total of 55,166 cases were identified for the primary analysis (14,540 outpatients and 40,576 inpatients). Overall, glenohumeral osteoarthritis was the most common indication for surgery (70.8%), followed by rotator cuff pathology (14.6%). The unadjusted rates of complications (1.3 vs 2.4%, P<.001), readmissions (3.7 vs 6.1%, P<.001), and mortality (0.2 vs 0.4%, P=.024) were significantly lower among outpatient TSAs than inpatient TSAs. Using 1:1 nearest matching, 12,703 patient pairs were identified. Propensity score-matched multivariate analyses showed similar rates of postoperative complications, hospital readmissions, and mortality between outpatients and inpatients. Propensity score-weighted multivariate analyses resulted in similar conclusions. The secondary analysis showed a lower hospital readmission rate in outpatients (odds ratio: 0.8, P<.001). There has been accelerated growth in outpatient TSAs since 2019. Outpatient and inpatient TSAs have similar rates of postoperative complication, hospital readmission, and mortality.

Integrating Serious Illness Care into Primary Care Delivery for persons with dementia: CARIÑOS En Su Casa – Care for you at home

AbstractBackgroundThe patient‐centered medical home (PCMH) as a model of primary care transformation has the potential to improve the quality, safety, and efficiency of care and reduce costs, particularly during the COVID‐19 pandemic for persons with dementia. However, the effects of adopting the PCMH are lack of clinical and economic support, let alone the challenge of applying it in ethnically and culturally diverse patient populations.MethodTo address the limitations of the traditional PCMH model, the CARIÑOS approach, a patient‐driven, comprehensive, coordinated, and multi‐site care model, was developed by the University of Texas (UT) Health San Antonio Geriatrics and Supportive Care Practice to meet the healthcare needs of patients in an underserved, primarily Hispanic community, who are often frail and have complex medical and care needs between various healthcare settings. Early outcomes on costs were collected to demonstrate its economic effects.ResultOverall, the implementation of the CARIÑOS approach reduced total costs per patient per year by comparing six practices affiliated to the UT Health San Antonio Accountable Care Organization (ACO) – Regional Physicians Network. They serve similar underserved areas in the first and second Quarters of 2021. After adopting the CARIÑOS approach ($45,705), the costs were lower than 4 out of the 6 ACOs practices. In comparison, the Hierarchical Condition Category (HCC) score of the patients seen at the Geriatrics and Supportive Care Practice (HCC = 1.45) was higher than those seen at the other six practices (average = 1.15) during the same period of 2021.ConclusionThe CARIÑOS approach provides an innovative and comprehensive perspective delivering culturally congruent care for persons with dementia and support for their families.The main learning objectives of the session will be1. To enlist the elements of the CARINOS Approach for person centered care.2. To describe strategies that can be used to implement this approach in dementia care of underserved and Hispanic families.3. List the ways to engage promotores or community health workers as a bridge between the healthcare team and patient and family to improve patient experience.Key words – person centered, dementia care

Drivers of Individual and Regional Variation in CMS Hierarchical Condition Categories Among Florida Beneficiaries.

To explore hierarchical condition categories (HCC) risk score variation among Florida Fee for Service (FFS) Medicare beneficiaries between 2016 and 2018. This study analyzed HCC risk score variation using Medicare claims data for Florida beneficiaries enrolled in Parts A & B between 2016 and 2018. The CMS methodology analyzed HCC risk score variation using annual mean county- and beneficiary-level risk score changes. The association between variation and beneficiary characteristics, diagnoses, and geographic location was characterized using mixed-effects negative binomial regression models. Not applicable. Counties in the Northeast [marginal effect (ME)=-0.003], Central (ME=-0.021), and Southwest (ME=-0.009) Florida have relatively lower mean risk scores. A higher number of lifetime (ME=0.246) and treatable (ME=0.288) conditions were associated with higher county-level risk scores, while more preventable conditions (ME=-0.249) were associated with lower risk scores. Counties with older beneficiaries (ME=0.015) and more Blacks (ME=0.070) have higher risk scores, while having female beneficiaries reduced risk scores (ME=-0.005). Individual risk scores did not vary by age (ME=0.000), but Blacks (ME=0.001) had higher rates of variation relative to Whites, while other races had comparatively lower variation (ME=-0.003). In addition, individuals diagnosed with more lifetime (ME=0.129), treatable (ME=0.235), and preventable (ME=0.001) conditions had higher risk score variation. Most condition-specific indicators showed small associations with risk score changes; however, metastatic cancer/acute leukemia, respirator dependence/tracheostomy, and pressure ulcers of the skin were significantly associated with both types of HCC risk score variation. Results showed demographics, HCC condition classifications (ie, lifetime, preventable, and treatable), and some specific conditions were associated with higher variation in mean county-level and individual risk scores. Results suggest consistent coding and reductions in the prevalence of certain treatable or preventable conditions could reduce the county and individual HCC risk score year-to-year change.

PO-03-205 ARRHYTHMIA BURDEN REVEALED BY AMBULATORY CARDIAC MONITORING IN A MEDICARE ADVANTAGE POPULATION

Medicare Advantage plans enroll an increasing proportion of Medicare beneficiaries annually. The capitation payment model for Medicare Advantage plans is dependent on prompt and accurate coding of Hierarchical Condition Categories (HCCs) that impact risk-adjusted payments. Reporting of disease is critical for clinics to estimate expenditures of Medicare Advantage populations with significant cardiac comorbidities. To quantify arrhythmia burden and potential risk adjustment impact in a Medicare Advantage population. We reviewed ambulatory electrocardiogram (ECG) monitors of 909 patients insured under a Medicare Advantage plan across 19 states. Each patient was enrolled in 14-day electrocardiogram monitoring according to clinical indication. Specialists evaluated recorded abnormal events and provided final interpretations. Standardized specialist-derived criteria were developed to assess for atrial arrhythmias, heart block, sick sinus syndrome, supraventricular tachycardia, and ventricular arrhythmias. Each patient with atrial fibrillation, supraventricular tachycardia, or sick sinus syndrome was assigned HCC96 (Specified Heart Arrhythmias). An additional monthly Medicare payment was calculated by multiplying the HCC96 coefficient of 0.268 by $750 (the approximate monthly Medicare payment corresponding to one point in risk adjustment). The average age of the sample population was 71.2 +- 0.3 (SE) years and 65.6% of patients were female. Common indications for ambulatory cardiac monitoring included palpitations (21.8%), syncope (15.4%), concern for atrial arrhythmia (9.5%), tachycardia (6.5%), and bradycardia (5.0%). Patients wore the ECG monitor for an average of 13 days. Abnormal findings included supraventricular tachycardia (59.4%), nonsustained ventricular tachycardia (6.7%), atrial fibrillation (3.0%), second- or third-degree heart block (1.2%), and sick sinus syndrome (2.6%). A majority (61.4%) had a diagnosis that could impact risk adjustment, leading to a potential increase of $1.35 million in annual Medicare reimbursement assuming this is the first recorded diagnosis of arrhythmia. Ambulatory ECG monitors reveal a high prevalence of arrhythmias that can have a significant risk adjustment impact for Medicare Advantage capitation plans. Additional clinical correlation is necessary to confirm the expected change in annual Medicare reimbursement.

Incremental Health Care Costs of Self-Reported Functional Impairments and Phenotypic Frailty in Community-Dwelling Older Adults : A Prospective Cohort Study.

Health care systems need better strategies to identify older adults at risk for costly care to select target populations for interventions to reduce health care burden. To determine whether self-reported functional impairments and phenotypic frailty are associated with incremental health care costs after accounting for claims-based predictors. Prospective cohort study. Index examinations (2002 to 2011) of 4 prospective cohort studies linked with Medicare claims. 8165 community-dwelling fee-for-service beneficiaries (4318 women, 3847 men). Weighted (Centers for Medicare & Medicaid Services Hierarchical Condition Category index) and unweighted (count of conditions) multimorbidity and frailty indicators derived from claims. Self-reported functional impairments (difficulty performing 4 activities of daily living) and frailty phenotype (operationalized using 5 components) derived from cohort data. Health care costs ascertained for 36 months after index examinations. Average annualized costs (2020 U.S. dollars) were $13906 among women and $14598 among men. After accounting for claims-based indicators, average incremental costs of functional impairments versus no impairment in women (men) were $3328 ($2354) for 1 impairment increasing to $7330 ($11760) for 4 impairments; average incremental costs of phenotypic frailty versus robust in women (men) were $8532 ($6172). Mean predicted costs adjusted for claims-based indicators in women (men) varied by both functional impairments and the frailty phenotype ranging from $8124 ($11831) among robust persons without impairments to $18792 ($24713) among frail persons with 4 impairments. Compared with the model with claims-derived indicators alone, this model resulted in more accurate cost prediction for persons with multiple impairments or phenotypic frailty. Cost data limited to participants enrolled in the Medicare fee-for-service program. Self-reported functional impairments and phenotypic frailty are associated with higher subsequent health care expenditures in community-dwelling beneficiaries after accounting for several claims-based indicators of costs. National Institutes of Health.

Plan Selection, Enrollee Risk, and Health Spending on the Patient Protection and Affordable Care Act Individual Marketplaces, 2019

The Patient Protection and Affordable Care Act (ACA) individual marketplaces are a source of insurance for millions of residents in the US. However, the association between enrollee risk, health spending, and metal tier selection remains unclear. To describe individual marketplace enrollees' metal tier selections by risk score and assess enrollees' health spending by metal tier, risk score, and spending type. This retrospective, cross-sectional study analyzed claims data from the Wakely Consulting Group ACA database, a deidentified claims database built on data voluntarily submitted by insurers. Enrollees with continuous, full-year enrollment in on-exchange or off-exchange ACA-qualified health plans during the 2019 contract year were included. Data analysis was conducted from March 2021 to January 2023. Enrollment totals, total spending, and out-of-pocket cost were calculated, stratified by metal tier and the Department of Health and Human Services (HHS) Hierarchical Condition Category (HCC) risk score for 2019. Enrollment and claims data were obtained for 1 317 707 enrollees (53.5% female; mean [SD] age, 46.35 [13.43] years) across all census areas, age groups, and sexes. Of these, 34.6% were on plans with cost-sharing reductions (CSRs), 75.5% did not have an assigned HCC, and 84.0% submitted at least 1 claim. Compared with enrollees in bronze plans (17.2%), enrollees were more likely to be classified in the top HHS-HCC risk quartile if they selected platinum (42.0%), gold (34.4%), or silver (29.7%) plans. The highest share of enrollees with $0 total spending was noted with the catastrophic (26.4%) and bronze (22.7%) plans, while gold plans had the lowest share (8.1%). Median total spending was lower among bronze plan enrollees ($593; IQR, $28-$2100) vs platinum ($4111; IQR, $992-$15 821) or gold ($2675; IQR, $728-$9070). Within the top risk score decile, CSR enrollees had less average total spending than any other metal tier by more than 10%. In this cross-sectional study of the ACA individual marketplace, enrollees who selected plans with higher actuarial value also had greater mean HHS-HCC risk scores and health spending. The findings suggest these differences may be associated with variation in benefit generosity by metal tier, enrollee's perceptions of future health needs, or other barriers to care access.

Readmission Rates and Episode Costs for Alzheimer Disease and Related Dementias Across Hospitals in a Statewide Collaborative

There has been a paucity of research examining the risk and cost of readmission among patients with Alzheimer disease and related dementias (ADRD) after a planned hospitalization for a broad set of medical and surgical procedures. To examine 30-day readmission rates and episode costs, including readmission costs, for patients with ADRD compared with their counterparts without ADRD across Michigan hospitals. This retrospective cohort study used 2012 to 2017 Michigan Value Collaborative data across different medical and surgical services stratified by ADRD diagnosis. A total of 66 676 admission episodes of care that occurred between January 1, 2012, and June 31, 2017, were identified for patients with ADRD using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnostic codes for ADRD, along with 656 235 admission episodes in patients without ADRD. Using a generalized linear model framework, this study risk adjusted, price standardized, and performed episode payment winsorization. Payments were risk adjusted for age, sex, Hierarchical Condition Categories, insurance type, and prior 6-month payments. Selection bias was accounted for using multivariable logistic regression with propensity score matching without replacement using calipers. Data analysis was performed from January to December 2019. Presence of ADRD. Main outcomes were 30-day readmission rate at the patient and county levels, 30-day readmission cost, and 30-day total episode cost across 28 medical and surgical services. The study included 722 911 hospitalization episodes, of which 66 676 were related to patients with ADRD (mean [SD] age, 83.4 [8.6] years; 42 439 [63.6%] female) and 656 235 were related to patients without ADRD (mean [SD] age, 66.0 [15.4] years; 351 246 [53.5%] female). After propensity score matching, 58 629 hospitalization episodes were included for each group. Readmission rates were 21.5% (95% CI, 21.2%-21.8%) for patients with ADRD and 14.7% (95% CI, 14.4%-15.0%) for patients without ADRD (difference, 6.75 percentage points; 95% CI, 6.31-7.19 percentage points). Cost of 30-day readmission was $467 higher (95% CI of difference, $289-$645) among patients with ADRD ($8378; 95% CI, $8263-$8494) than those without ($7912; 95% CI, $7776-$8047). Across all 28 service lines examined, total 30-day episode costs were $2794 higher for patients with ADRD vs patients without ADRD ($22 371 vs $19 578; 95% CI of difference, $2668-$2919). In this cohort study, patients with ADRD had higher readmission rates and overall readmission and episode costs than their counterparts without ADRD. Hospitals may need to be better equipped to care for patients with ADRD, especially in the postdischarge period. Considering that any type of hospitalization may put patients with ADRD at a high risk of 30-day readmission, judicious preoperative assessment, postoperative discharge, and care planning are strongly advised for this vulnerable patient population.

Higher Surgeon Volume is Associated With a Lower Rate of Subsequent Revision Procedures After Total Shoulder Arthroplasty: A National Analysis.

Studies assessing the relationship between surgeon volume and outcomes have shown mixed results, depending on the specific procedure analyzed. This volume relationship has not been well studied in patients undergoing total shoulder arthroplasty (TSA), but it should be, because this procedure is common, expensive, and potentially morbid. We performed this study to assess the association between increasing surgeon volume and decreasing rate of revision at 2 years for (1) anatomic TSA (aTSA) and (2) reverse TSA (rTSA) in the United States. In this retrospective study, we used Centers for Medicare and Medicaid Services (CMS) fee-for-service inpatient and outpatient data from 2015 to 2021 to study the association between annual surgeon aTSA and rTSA volume and 2-year revision shoulder procedures after the initial surgery. The CMS database was chosen for this study because it is a national sample and can be used to follow patients over time. We included patients with Diagnosis-related Group code 483 and Current Procedural Terminology code 23472 for TSA (these codes include both aTSA and rTSA). We used International Classification of Diseases, Tenth Revision, procedural codes. Patients who underwent shoulder arthroplasty for fracture (10% [17,524 of 173,242]) were excluded. We studied the variables associated with the subsequent procedure rate through a generalized linear model, controlling for confounders such as patient age, comorbidity risk score, surgeon and hospital volume, surgeon graduation year, hospital size and teaching status, assuming a binomial distribution with the dependent variable being whether an episode had at least one subsequent procedure within 2 years. The regression was fitted with standard errors clustered at the hospital level, combining all TSAs and within the aTSA and rTSA groups, respectively. Hospital and surgeon yearly volumes were calculated by including all TSAs, primary procedure and subsequent, during the study period. Other hospital-level and surgeon-level characteristics were obtained through public files from the CMS. The CMS Hierarchical Condition Category risk score was controlled because it is a measure reflecting the expected future health costs for each patient based on the patient's demographics and chronic illnesses. We then converted regression coefficients to the percentage change in the odds of having a subsequent procedure. After controlling for confounding variables including patient age, comorbidity risk score, surgeon and hospital volume, surgeon graduation year, and hospital size and teaching status, we found that an annual surgeon volume of ≥ 10 aTSAs was associated with a 27% decreased odds of revision within 2 years (95% confidence interval 13% to 39%; p < 0.001), while surgeon volume of ≥ 29 aTSAs was associated with a 33% decreased odds of revision within 2 years (95% CI 18% to 45%; p < 0.001) compared with a volume of fewer than four aTSAs per year. Annual surgeon volume of ≥ 29 rTSAs was associated with a 26% decreased odds of revision within 2 years (95% CI 9% to 39%; p < 0.001). Surgeons should consider modalities such as virtual planning software, templating, or enhanced surgeon training to aid lower-volume surgeons who perform aTSA and rTSA. More research is needed to assess the value of these modalities and their relationship with the rates of subsequent revision. Level III, therapeutic study.

Emergency Department and Hospital Utilization Among Older Adults Before and After Identification of Elder Mistreatment

Elder mistreatment is common and has serious health consequences. Little is known, however, about patterns of health care utilization among older adults experiencing elder mistreatment. To examine emergency department (ED) and hospital utilization of older adults experiencing elder mistreatment in the period surrounding initial mistreatment identification compared with other older adults. This retrospective case-control study used Medicare insurance claims to examine older adults experiencing elder mistreatment initially identified between January 1, 2003, and December 31, 2012, and control participants matched on age, sex, race and ethnicity, and zip code. Statistical analysis was performed in April 2022. We used multiple measures of ED and hospital utilization patterns (eg, new and return visits, frequency, urgency, and hospitalizations) in the 12 months before and after mistreatment identification. Data were adjusted using US Centers for Medicare and Medicaid Services Hierarchical Condition Categories risk scores. Chi-squared tests and conditional logistic regression models were used for data analyses. This study included 114 case patients and 410 control participants. Their median age was 72 years (IQR, 68-78 years), and 340 (64.9%) were women. Race and ethnicity were reported as racial or ethnic minority (114 [21.8%]), White (408 [77.9%]), or unknown (2 [0.4%]). During the 24 months surrounding identification of elder mistreatment, older adults experiencing mistreatment were more likely to have had an ED visit (77 [67.5%] vs 179 [43.7%]; adjusted odds ratio [AOR], 2.95 [95% CI, 1.78-4.91]; P < .001) and a hospitalization (44 [38.6%] vs 108 [26.3%]; AOR, 1.90 [95% CI, 1.13-3.21]; P = .02) compared with other older adults. In addition, multiple ED visits, at least 1 ED visit for injury, visits to multiple EDs, high-frequency ED use, return ED visits within 7 days, ED visits for low-urgency issues, multiple hospitalizations, at least 1 hospitalization for injury, hospitalization at multiple hospitals, and hospitalization for ambulatory care sensitive conditions were substantially more likely for individuals experiencing elder mistreatment. The rate of ED and hospital utilization for older adults experiencing elder mistreatment was much higher in the 12 months after identification than before, leading to more pronounced differences between case patients and control participants in postidentification utilization. During the 12 months after identification of elder mistreatment, older adults experiencing mistreatment were particularly more likely to have had high-frequency ED use (12 [10.5%] vs 8 [2.0%]; AOR, 8.23 [95% CI, 2.56-26.49]; P < .001) and to have visited the ED for low-urgency issues (12 [10.5%] vs 8 [2.0%]; AOR, 7.33 [95% CI, 2.54-21.18]; P < .001). In this case-control study of health care utilization, older adults experiencing mistreatment used EDs and hospitals more frequently and with different patterns during the period surrounding mistreatment identification than other older adults. Additional research is needed to better characterize these patterns, which may be helpful in informing early identification, intervention, and prevention of elder mistreatment.

Alternative Predictive Models for Medicare Patient Cost

As health care expenditures increase, patient cost mitigation becomes more essential. Cost mitigation through intervention programs such as accountable care organizations relies on the ability to accurately predict patient risk, which is notoriously difficult because of highly skewed data. We examine the Medicare Limited dataset (a 5% sample of Medicare claims) that includes demographics, costs, and health conditions. We first consider the Centers for Medicare and Medicaid Services (CMS) currently used Hierarchical Condition Category (HCC) linear model and then implement more complex two-part generalized additive and random forest models to predict patient costs in a future year based on current-year data. We find that the latter models more accurately predict the entire distribution of Medicare patient costs and can better support the existing cost mitigation frameworks. The two-part lognormal generalized additive model is chosen as the optimal model for its robust performance and reasonable interpretability when the data have extreme values.

Validation of the self-reported domains of the Edmonton Frail Scale in patients 65 years of age and older

IntroductionIn the era of virtual care, self-reported tools are beneficial for preoperative assessments and facilitating postoperative planning. We have previously reported the use of the Edmonton Frailty Scale (EFS) as a valid preoperative assessment tool.ObjectiveWe wished to validate the self-reported domains of the EFS (srEFS) by examining its association with loss of independence (LOI) and mortality.MethodsThis is a post-hoc analysis of a single-institution observational study of patients 65 years of age or older undergoing multi-specialty surgical procedures and assessed with the EFS in the preoperative setting. Exploratory data analysis was used to determine the threshold for identifying frailty using the srEFS. Procedures were classified using the Operative Stress Score (OSS) scored 1 to 5 (lowest to highest). Hierarchical Condition Category (HCC) was utilized to risk-adjust. LOI was described as requiring more support at discharge and mortality was defined as death occurring up to 30 days following surgery. Receiver operating characteristic (ROC) curves were used to determine the ability of the srEFS to predict the outcomes of interest in relation to the EFS.ResultsFive hundred thirty-five patients were included. Exploratory analysis confirmed best positive predictive value for srEFS was greater or equal to 5. Overall, 113 (21 percent) patients were considered high risk for frailty (HRF) and 179 (33 percent) patients had an OSS greater or equal to 5. LOI occurred in 7 percent (38 patients) and the mortality rate was 4 percent (21 patients). ROC analysis showed that the srEFS performed similar to the standard EFS with no difference in discriminatory thresholds for predicting LOI and mortality. Examination of the domains of the EFS not included in the srEFS demonstrated a lack of association between cognitive decline and the outcomes of interest. However, functional status assessed with either the Get up and Go (EFS only) or self-reported ADLs was independently associated with increased risk for LOI.ConclusionThis study shows that self-reported EFS may be an optional preoperative tool that can be used in the virtual setting to identify patients at HRF. Early identification of patients at risk for LOI and mortality provides an opportunity to implement targeted strategies to improve patient care.

Gender differences in Medicare payments to cardiothoracic surgeons in 2019.

The STS Thoracic Surgery Practice and Access Task Force - 2019 Workforce Report noted gender-based differences in the income of cardiothoracic surgeons in the United States. We analysed the 2019 Medicare payment data for thoracic and cardiac surgeons to investigate the gender-based payment gap among cardiothoracic surgeons. The 2019 Medicare Physician and Other Practitioners by Provider and Services data set merged with the Doctors and Clinicians National Downloadable File was utilized to conduct a cross-sectional analysis of gender differences in Medicare payments, numbers of services, unique billing codes, years in practice, Medicare beneficiary age, regional population density (rural-urban commuting area code) and patient panel complexity (hierarchical condition category) for providers. The providers' self-reported gender (women or men) and provider type (thoracic surgery or cardiac surgery) were binarily set according to the Center for Medicare and Medicaid Services standards. Independent analyses were performed with thoracic and cardiac surgeons. We also used the 2013 and 2016 Medicare Physician and Other Practitioners by Provider and Services data sets to analyse the trends in adjusted gender-based payment differences across 2013, 2016 and 2019. After controlling for the covariates, women thoracic surgeons received $25,183.50 [95% confidence interval (CI) $16,307.60, $34,059.40] less than the mean Medicare payment than men thoracic surgeons. Likewise, women cardiac surgeons received $20,960 [95% confidence interval (CI) $1,014.80, $40,902.80] less than the mean adjusted Medicare payment than their men counterparts. In 2019, women cardiothoracic surgeons received a significantly lower mean Medicare payment than men cardiothoracic surgeons after controlling for the number of services, unique billing codes, the complexity of the patient panel, years in practice and regional population density. The payment gap between women and men exhibited no statistically significant change over 2013, 2016 and 2019.Future studies are warranted to understand the association between gender representation and the pay gap.

PREDICTORS OF PARTICIPATION IN A TELEPHONE-BASED SOCIAL CONNECTEDNESS INTERVENTION FOR OLDER ADULTS

Abstract Background Social isolation is a well-documented contributor to poor mental and physical health, and interventions promoting social connectedness have been shown to be protective. This study examined predictors of participating in a telephone-based social connectedness intervention for socially isolated older adults. Methods Data was obtained from a social connectedness intervention that paired students with older adults. Eligible participants included Houston area older adults, 65 years or older, enrolled in Medicare Advantage plans. Eligible participants were contacted telephonically and asked to complete the 3-item UCLA Loneliness Scale. Those who screened positive for loneliness were invited to participate in the social connectedness intervention. Logistic regression models, that accounted for sociodemographic, clinical and functional indices, were used to identify predictors of participation. Results Females (OR: 1.23, 95% CI: 1.04-1.45), and racial/ethnic minorities (African American OR: 1.39, 95%CI: 1.16-1.68; Hispanic OR: 1.43, 95%CI: 1.04-1.99) were more likely to participate in a telephone-based social connectedness intervention. Older adults with a disability (OR: 2.37, 95%CI: 1.96-2.87), higher CMS Hierarchical Condition Category risk scores (OR: 1.15, 95%CI: 1.06-1.25) and those who reported having 1+ social needs (OR: 1.38, 95%CI: 1.14-1.66) were more likely to participate. Of all Charlson Comorbidity Index flags examined, diabetes was the single strongest predictor of participation (OR=2.49, p=0.02). Older adults who reported functional comorbidities for anxiety and depression were also more likely to participate. Conclusion Telephone-based social connectedness interventions can reach vulnerable older adults with clinical and social needs, and can be useful in addressing racial/ethnic health equity gaps in socially isolated older adults.

Abstract 14878: Natural Language Processing of Unstructured Medical Record Can Predict Cardiac Arrhythmia by Quantifying Chronic Disease Burden

Introduction: Prediction of atrial fibrillation (AF) and sick sinus syndrome (SSS) is difficult. CMS Hierarchical Condition Categories scores (CMS-HCC) is a risk adjustment score used by CMS to account to risk adjust quality and cost metrics by quantifying the burden of chronic disease and patient complexity. Its calculation is based on patient’s demographics and ICD10 diagnostic codes. We have created an algorithm (MYTH) that can determine ICD10 diagnostic codes from unstructured medical chart using Natural Language Processing through syntax analysis, knowledge graph and regular expressions. The ICD-10 codes can then be used to calculate CMS-HCC. Because chronic disease burden may increase the probability of cardiac arrhythmia, we hypothesize that CMS-HCC can predict the onset of atrial fibrillation and sick sinus syndrome. Methods: We used MYTH to automate ICD10 coding and HCC score calculation based on textual outpatient cardiology records of 926 patients. The algorithm was additionally applied to calculate CHA2DS2-VASc scores. These were applied to predict the diagnosis of AF and SSS on clinical follow-up retrospectively. Results: CMS-HCC scores outperforms CHA2DS2-VASc scores in the prediction of onset of AF: (AUC: CHA2DS2-VASc 0.703, CMS-HCC 0.778, difference P=0.0006, individually P&lt;0.0001). While HCC score is predictive of the onset of SSS (P&lt;0.001), its improvement over CHA2DS2-VASc score does not reach statistical significance (AUC: CHA2DS2-VASc 0.702; CMS-HCC 0.744, difference: P=0.297, individually P&lt;0.001). Conclusions: Natural Language Processing algorithms can understand unstructured medical text, automate ICD coding and automate CMS-HCC score calculation. The burden of chronic diseases, as reflected by the CMS-HCC score, is highly predictive of future atrial fibrillation and modestly predictive of sick sinus syndrome.

The impact of seasonality with hierarchical condition category coding on Oncology Care Model performance for the U.S. Oncology Network.

66 Background: The Oncology Care Model (OCM) is a 6-month, episode-based, Medicare value-based care program, which rewards practices for decreasing the total cost of care (TCOC) compared to a trend adjusted predicted baseline called the benchmark price. The predicted baseline and trend factor are a function of 14 covariates in a generalized linear model with a log link and gamma distribution. Select non-cancer comorbidities, represented by a subset of Hierarchical Condition Category (HCC) flags assigned to the episode in the calendar year when the episode initiates, is a major covariate of the linear model. Patient episodes with one or more HCC flags are expected to have higher episode expenditures and receive a higher adjustment to the benchmark. Here, we seek to describe the seasonality of HCC flags and its impact on the benchmark for OCM episodes in The US Oncology Network (The Network). Methods: All eligible OCM episodes data from 14 practices in The Network participating in the OCM for performance periods (PP) 3-9 were analyzed to measure the average number of HCC flags per episode. The relative contribution of HCC flags to the benchmark was calculated by unraveling the linear model. The difference of the average HCC flags, benchmark, and relative contribution of HCC flags to the benchmark for episodes starting in different quarters of the calendar year were evaluated. Results: Average HCC flags for episodes showed a seasonal decline during each calendar year, with episodes initiating during the first quarter of a calendar year having 16.25% higher HCC flags, compared to those in the last quarter (1.93 vs 1.66 flags). The benchmark and the relative contribution of the HCC flags to the episode benchmark were lower in the last quarter of the year (4% and 16.5% respectively) compared to the first quarter. Episode expenditures did not show a similar seasonality pattern. Conclusions: The assignment of HCC flags based on the episode initiation date, leads to a seasonality effect on the average HCC flags and benchmark for episodes initiating in different parts of the calendar year. The seasonality results from a progressively abbreviated period available to assign HCC flags for episodes initiating later in the calendar year. We also hypothesize that the annual reporting requirement for HCC flags, and risk adjustment coding by professionals at the start of each new calendar year, contributes to this seasonality. The financial impact of seasonality on episodic value-based care model benchmarks necessitates a modified, non-seasonal approach to comorbidity-based risk adjustment.

Multiscale dimensions of county-level disparities in opioid use disorder rates among older Medicare beneficiaries.

BackgroundOpioid use disorder (OUD) among older adults (age ≥ 65) is a growing yet underexplored public health concern and previous research has mainly assumed that the spatial process underlying geographic patterns of population health outcomes is constant across space. This study is among the first to apply a local modeling perspective to examine the geographic disparity in county-level OUD rates among older Medicare beneficiaries and the spatial non-stationarity in the relationships between determinants and OUD rates.MethodsData are from a variety of national sources including the Centers for Medicare & Medicaid Services beneficiary-level data from 2020 aggregated to the county-level and county-equivalents, and the 2016–2020 American Community Survey (ACS) 5-year estimates for 3,108 contiguous US counties. We use multiscale geographically weighted regression to investigate three dimensions of spatial process, namely “level of influence” (the percentage of older Medicare beneficiaries affected by a certain determinant), “scalability” (the spatial process of a determinant as global, regional, or local), and “specificity” (the determinant that has the strongest association with the OUD rate).ResultsThe results indicate great spatial heterogeneity in the distribution of OUD rates. Beneficiaries' characteristics, including the average age, racial/ethnic composition, and the average hierarchical condition categories (HCC) score, play important roles in shaping OUD rates as they are identified as primary influencers (impacting more than 50% of the population) and the most dominant determinants in US counties. Moreover, the percentage of non-Hispanic white beneficiaries, average number of mental health conditions, and the average HCC score demonstrate spatial non-stationarity in their associations with the OUD rates, suggesting that these variables are more important in some counties than others.ConclusionsOur findings highlight the importance of a local perspective in addressing the geographic disparity in OUD rates among older adults. Interventions that aim to reduce OUD rates in US counties may adopt a place-based approach, which could consider the local needs and differential scales of spatial process.

Evaluating The Accuracy Of Medicare Risk Adjustment For Alzheimer's Disease And Related Dementias.

In 2020 Medicare reintroduced Alzheimer's disease and related dementias (ADRD) Hierarchical Condition Categories (HCCs) to risk-adjust Medicare Advantage and accountable care organization (ACO) payments. The potential for Medicare spending increases from this policy change are not well understood because the baseline accuracy of ADRD HCCs is uncertain. Using linked 2016-18 claims and electronic health record data from a large ACO, we evaluated the accuracy of claims-based ADRD HCCs against a reference standard of clinician-adjudicated disease. An estimated 7.5percent of beneficiaries had clinician-adjudicated ADRD. Among those with ADRD HCCs, 34percent did not have clinician-adjudicated disease. The false-negative and false-positive rates were 22.7percent and 3.2percent, respectively. Medicare spending for those with false-negative ADRD HCCs exceeded that of true positives by $14,619 per beneficiary. If, after the reintroduction of risk adjustment for ADRD, all false negatives were coded as having ADRD, expenditure benchmarks for beneficiaries with ADRD would increase by 9percent. Monitoring ADRD coding could become challenging in the setting of concurrent incentives to decrease false-negative rates and increase false-positive rates.