Bringing doctors together

or

A free, private network for verified physicians

Leading voices

Dr. Pugazhendhi Vijayaraman

EP · Geisinger

View

Dr. Robert Schaller

EP · HUP

View

Dr. Felix Yang

EP · NYU Langone

View

Dr. Gregory Marcus

EP · UCSF

View

Dr. Miguel Leal

EP · Emory University Hospital

View

Dr. Chirag Barbhaiya

EP · NYU

View

Dr. William Sauer

EP · BWH

View

Dr. Dorairaj Prabhakaran

General Cardiology

View

Dr. Trey Woods

General Cardiology · Emory Healthcare

View

Dr. Brahmajee Nallamothu

Interventional Cardiology · UMHS

View
  • Discover your community

  • Stay current

  • A profile that represents you

  • Find your people

About RoonPrivacyTerms

    Overview

    Relationship Between Insurance Status and Receipt of Cardiac Tests and Procedures During Hospitalization: A Cross-Sectional Study.

    Sep 30, 2024DOI: 10.1161/jaha.124.035797Journal ArticleAHA

    Written by: Michael I EllenbogenJoseph E MarineArmin Arbab-ZadehChathurangi H PathiravasanJenna SwannDaniel J Brotman

    Find these authors on Roon

    JM
    Dr. Joseph Marine
    EP · Johns Hopkins

    METHODS

    The requirement for institutional review board approval was waived, as this study used a limited data set for which institutional review board approval is not needed. Informed consent was not required. Data for this study cannot be shared due to a data use agreement. However, the statistical code will be shared upon request.We used data from the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality ‐ the State Emergency Department Databases for observation stays, and State Inpatient Databases for inpatient admissions.23 We obtained data for Kentucky, Maryland, New Jersey, and North Carolina from 2016 through 2018. These states were chosen on the basis of cost and availability of relevant data elements. We used the American Hospital Association Annual Survey24 for data on hospital characteristics.We included all encounters for adults (aged ≥18 years) with Medicaid or commercial insurance as the primary payer for an observation stay or inpatient hospitalization at an acute care hospital. Hospitalizations at pediatric, psychiatric, long‐term acute care, and specialty hospitals were excluded. Patients with diagnosis codes suggesting medication nonadherence or active substance use disorder were excluded as they more commonly have Medicaid and might be expected to undergo fewer tests and procedures (Data S1).The outcomes of interest were receipt of echocardiogram, stress test, cardiac catheterization (elective, in acute coronary syndrome [ACS], and in ST‐segment–elevation myocardial infarction [STEMI]), pacemaker, and subcutaneous cardiac rhythm monitor (SCRM; also known as implantable loop recorder). In addition to cardiac imaging and procedures, brain magnetic resonance imaging (MRI) served as a noncardiac control. These tests and procedures were identified using International Classification of Diseases, Tenth Revision (ICD‐10) procedure codes, Current Procedural Terminology codes, and revenue center codes (Table S1).For cardiac catheterization, ICD‐10 diagnosis codes were used to identify hospitalizations for ACS and STEMI (Data S1). Patients with codes consistent with ACS were excluded in the elective cardiac catheterization model. Patients with a primary or secondary diagnosis code signifying ACS or STEMI were included in the models for cardiac catheterization in ACS and STEMI, respectively (Data S1).
    This cross‐sectional study used propensity score matching to create cohorts of hospitalized Medicaid and commercially insured patients who were otherwise similar. The propensity score used age, sex, and the Elixhauser comorbidity index.25 Additionally, for the cardiac testing and procedures, 2 independent binary comorbidity indicators, for chronic congestive heart failure and chronic coronary artery disease, were included (Data S1).
    Patients were matched within a hospital, using a 1:2 ratio of Medicaid to commercially insured patients with calipers of 0.05 times the SD of the propensity score. To generate the matched set for echocardiograms, stress tests, and brain MRI, encounters at critical access hospitals were excluded. To generate the matched set for catheterizations, encounters at hospitals that did not perform any catheterizations were excluded. To generate the matched set for pacemaker and SCRM placement, encounters at hospitals that did not perform any pacemaker implantations were excluded.
    We characterized patient demographics (age, sex, race, ethnicity, quartiled median income of patient zip code of residence), state, clinical characteristics (Elixhauser comorbidity index, frequency and rate of ACS, STEMI, and chronic congestive heart failure and coronary artery disease indicators), and frequency and rate of tests and procedures of the unmatched sample of Medicaid and commercially insured patients. Continuous variables were reported as mean and SD, and categorical variables were reported as counts and percentages.
    For each diagnostic test or procedure, we estimated univariate and multivariate generalized linear models on the matched set with a hospital‐specific indicator variable. For each model, the test or procedure of interest was the outcome, and insurance status was the predictor of interest, with additional covariates in the multivariate model for race (White, Black, Asian, other), ethnicity (Hispanic, non‐Hispanic), and zip code income quartile. We also estimated multivariate generalized linear models without a hospital indicator variable adding a covariate for state to calculate state‐specific coefficients (which was not possible in the multivariate model with a hospital indicator variable due to collinearity, since each hospital can only be in 1 state). Observations with missing data were excluded from regression analyses. We calculated unadjusted and adjusted odds ratios with 95% CIs. The Medicaid and commercially insured patients in each matched group were compared by age, Elixhauser comorbidity index, percentage women, and presence of congestive heart failure and coronary artery disease (except in the brain MRI‐matched sample).
    Additionally, for each diagnostic test or procedure, we performed a subanalysis of the Black population (controlling for zip code income) and a subanalysis of Maryland hospitalizations (controlling for race, ethnicity, and zip code income), done because the Maryland hospital payment model is unique and, in theory, lessens the financial incentives to perform tests and procedures in patients with commercial insurance relative to Medicaid.26
    To determine the influence of hospital characteristics on the relationship between primary payer and receipt of cardiac testing or procedures, for each diagnostic test or procedure we used a multivariate generalized linear model with a payer–hospital characteristic interaction term and a hospital‐specific indicator variable. The hospital characteristics of interest—teaching status (nonteaching, minor teaching, major teaching), geography (rural, urban), bed size (<200, 200–399, and ≥400), ownership (government, nonfederal; nongovernment, not‐for‐profit; for‐profit), payer mix quartile (ratio of patients with commercial insurance to Medicaid)—were each tested separately. For each test or procedure category, we reestimated the multivariate regression stratified by a given hospital characteristic for each hospital characteristic with a significant interaction term (P<0.05). Analyses were performed using RStudio version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria).

    RESULTS

    After removing encounters for patients with a history of medication nonadherence or an active substance use disorder, our unmatched sample included 4 291 471 encounters, with 1 357 994 (31.6%) Medicaid patients and 2 933 477 (68.4%) commercially insured patients (Table 1). The mean age was 47.6 years (SD, 17.7), and the sample was 65.3% women (2 804 128 encounters). The Medicaid group was younger and had a larger proportion of women. White patients comprised 48.0% of the Medicaid cohort and 68.4% of the commercial cohort, and Black patients comprised 31.7% of the Medicaid cohort and 19.1% of the commercially insured cohort.
    Patient and Hospital Descriptive Characteristics
    ACS indicates acute coronary syndrome; CAD, coronary artery disease; CHF, congestive heart failure; MRI, magnetic resonance imaging; SCRM, Subcutaneous cardiac rhythm monitor; and STEMI, ST‐segment–elevation myocardial infarction.
    Other includes American Indian patients, multiracial patients, and patients whose race was listed as other.
    Echocardiogram was the most frequently performed elective test (13.7% of encounters; 10.4% for Medicaid and 15.3% for commercially insured patients). Rates of all tests and procedures were higher in the commercially insured cohort. For both groups, the majority of encounters were at minor teaching (57.1%), urban (89.3%), nongovernment, not‐for‐profit (84.3%) hospitals. The only covariates with missing data were ethnicity (3.4% and 3.5% for Medicaid and commercial insurance, respectively) and zip code income (1.2% and 1.0%).
    In the adjusted models, commercially insured patients were significantly more likely to receive each test or procedure. The adjusted odds ratio ranged from 1.16 (95% CI, 1.00–1.34) for cardiac catheterization in STEMI to 1.40 (95% CI, 1.27–1.54) for pacemaker implantation, and was 1.12 (95% CI, 1.07–1.17) for brain MRI (Table 2; Tables S2 through S9). In the Black patient subanalysis, commercially insured patients were more likely to receive all tests or procedures except catheterization for STEMI (Table S10). The adjusted odds ratios ranged from 1.37 for both echocardiogram (95% CI, 1.29–1.45) and catheterization for ACS (95% CI, 1.21–1.54) to 1.95 (95% CI, 1.61–2.36) for pacemaker implantation. It was 1.23 (95% CI, 1.17–1.29) for brain MRI. For the Maryland subanalysis, commercially insured patients were more likely to receive all tests or procedures except catheterization for STEMI and SCRM (Table S11). The adjusted odds ratios ranged from 1.15 (95% CI, 1.08–1.22) for echocardiogram to 1.53 (95% CI, 1.25–1.87) for pacemaker. It was 1.15 (95% CI, 1.06–1.26) for brain MRI. In the matched groups, the Medicaid cohort was younger with more women (Table 3; Tables S12 and S13).
    Odds Ratio of a Commercially Insured Patient Compared With a Medicaid Patient Receiving a Test or Procedure
    For each model, the test or procedure of interest was the outcome, and insurance status was the predictor of interest, with additional covariates in the multivariate model for race (White, Black, Asian, other), ethnicity (Hispanic, non‐Hispanic), and quartile of median income for a patient's zip code of residence. We also included a hospital‐specific indicator variable in both the univariate and multivariate model.
    ACS indicates acute coronary syndrome; MRI, magnetic resonance imaging; SCRM, subcutaneous cardiac rhythm monitor; and STEMI, ST‐segment–elevation myocardial infarction.
    Indicates P‐values that met statistical significance (P < 0.05).
    Comparison of Propensity‐Matched Sets
    ACS indicates acute coronary syndrome; CAD, coronary artery disease; CHF, congestive heart failure; COM, commercial insurance; ECI, Elixhauser comorbidity index; M, Medicaid insurance; MRI, magnetic resonance imaging; SCRM, subcutaneous cardiac rhythm monitor; and STEMI, ST‐segment–elevation myocardial infarction.
    Cardiac comorbidities were not included in matching for MRI Brain.
    The hospital payer mix/payer and bed size/payer interaction terms were significant for the most tests and procedures. The hospital payer mix interaction term was significant for echocardiogram, stress test, elective catheterization, pacemaker, and brain MRI, with the payer–usage disparity highest for quartiles 2 and 3 in stratified analyses (Figure; Table S14).
    Cath indicates catheterization; CRM, cardiac rhythm monitor; Echo, echocardiogram; Gvt, government; and MRI, magnetic resonance imaging.
    The bed size interaction term was significant for echocardiogram, stress test, catheterization in ACS, SCRM, and brain MRI. For the stratified bed size analyses, the payer‐usage disparity was highest in small hospitals for echocardiogram and stress tests, in medium and large hospitals for cardiac catheterization in ACS, in large hospitals for SCRMs, and in small and large hospitals for brain MRI (Figure).

    DISCUSSION

    Financial incentives impact decisions at numerous points in the health care system. However, our understanding of what drives physicians to overuse tests and procedures is limited. The association between a patient's insurance status and diagnostic and procedural intensity (or low‐value care) has primarily been studied for outpatients, and the evidence is mixed.16, 17, 18, 19, 20, 21, 22 It has not been well‐characterized for hospitalized patients.
    Prior work found that hospital payer mix is associated with intensity and quality of care of hospitalized patients. More favorable payer mix was associated with higher‐quality and lower‐cost care27 and higher diagnostic intensity (measured at the hospital level).14 Additionally, greater managed care penetration was associated with less intense care.28 We sought to describe the relationship between a patient's insurance status and cardiac diagnostic and procedural intensity using propensity score matching to mitigate the effect of differences between Medicaid and commercially insured populations.
    We hypothesized that commercially insured patients would receive more intense care due, in part, to higher reimbursement rates, including professional fees. A recent Congressional Budget Office analysis found that professional fees paid by commercial insurers were on average 29% above Medicare fee‐for‐service rates,29 and Medicaid professional fees varied by state but were generally lower than Medicare (and were lower for all states in our sample).30 The majority of Medicaid patients (69% in 2018) are in a managed care plans, creating additional incentives to constrain usage.31
    We also expected this commercial/Medicaid usage differential to be wider for tests and procedures without well‐defined indications or clear guidelines and for tests and procedures with higher levels of reimbursement. We anticipated a smaller differential for urgent or emergent situations and for tests or procedures that are often ordered by physicians who may not financially benefit from reviewing or performing them.
    We found that patients with commercial insurance were more likely to receive all the tests and procedures we studied. These findings were robust and generally consistent in the subanalyses. For cardiac catheterization across clinical scenarios (elective, ACS, and STEMI), the effect size of payer status was consistent with our hypothesis: its magnitude decreased with increasing urgency. Additionally, the effect size was larger for procedures with greater reimbursement (elective catheterization and pacemaker implantation compared with echocardiogram and stress test), though the magnitude of these differences was small. Finally, tests and procedures less susceptible to physician‐induced demand32 (echocardiogram and brain MRI) had lower effect sizes, though still higher rates in the commercially insured population.
    Our findings are consistent with the limited research on the relationship between insurance status and cardiac testing and procedures. Medicaid (relative to commercially insured) patients had an adjusted odds ratio of 0.59 (95% CI, 0.41–0.84) of their physician ordering a stress test during an ambulatory visit.33 Additionally, Medicaid patients with STEMI had lower rates of revascularization than commercially insured patients.34 More generally, patients who saw fee‐for‐service cardiologists received more imaging than those who saw salaried cardiologists, and patients seeing fee‐for‐service cardiologists who were able to bill for image interpretation received more imaging than those who saw fee‐for‐service cardiologists who could not.35
    We do not believe that most physicians' clinical decisions are highly motivated by financial concerns, but instead that in situations in which the need for a test or procedure is equivocal, a financial incentive (or disincentive) related to insurance status may have a measurable impact at the margin on a subset of physicians. Hospitals attempting to maximize revenue could make patients' insurance status more obvious in the electronic medical record or enhance incentives for physicians to perform more tests or procedures in patients with more highly remunerative insurance through gainsharing arrangements. Additionally, it is possible that patients who may have more difficulty accessing outpatient specialty care might be less likely to receive tests and procedures requiring outpatient follow‐up.
    In our analysis, there was little consistency across types of imaging and procedures in hospital characteristics associated with a greater insurance usage differential. Moreover, the strata of a given hospital characteristic associated with a larger differential varied across tests and procedures. The exception was hospital payer mix, which had a positive interaction term for the majority of tests and procedures, and in the stratified analyses, intermediate payer mix hospitals (quartiles 2 and 3) consistently had the highest differential. This suggests that hospitals with more diverse payer populations might be more likely to incentivize higher diagnostic or procedural intensity on patients with more remunerative insurance.
    While not our primary focus, several findings related to covariates are noteworthy. Consistent with prior literature,36, 37, 38, 39, 40, 41 Black patients were less likely to undergo cardiac catheterization in all 3 scenarios (though for STEMI, the P value was 0.05) and pacemaker implantation. Black patients were more likely to undergo stress testing, suggesting the possibility that stress tests were performed in lieu of elective cardiac catheterization.
    Interesting patterns emerged with respect to state variation in use of cardiac tests and procedures. Hospitalized patients in Kentucky were more likely to receive all cardiac tests and procedures. A potential explanation is higher levels of cardiac disease in Kentucky. Centers for Disease Control and Prevention data show that Kentucky had between the seventh and ninth highest cardiac morbidity between 2016 and 2018 (the other states in our sample were closer to the national mean).42 Of note, Kentucky was not an outlier in our positive‐control analysis for brain MRI.
    We also conducted a subanalysis of Maryland given its unique hospital payment model: All insurers (public and private) pay the same rate for the same hospital service at a given hospital, set by the Maryland Health Services Cost Review Commission.26 However, professional fees are not included under this so‐called Maryland waiver. Given that professional fees can make up a significant portion of a cardiologist's income, it is not surprising that the discordance in testing and procedure rates by insurance status persisted in the Maryland subanalysis.
    However, Maryland had the lowest usage for all cardiac tests and procedures (except echocardiogram), which may be related to its regulatory environment. The Health Services Cost Review Commission initiated the Maryland All‐Payer Medicare Model Contract in 2013 to begin transitioning from a fee‐for‐service system to a total cost‐of‐care model.26 The rates of elective catheterizations and catheterizations in ACS were markedly lower in Maryland. In addition to its payment model, this may represent the impact of several high‐profile Maryland lawsuits43, 44, 45 for inappropriate catheterizations and stent placement, which led to implementation in 2012 of a catheterization‐specific regulatory apparatus in Maryland.46
    Our study has several limitations. First, its design does not allow determination of a causal relationship between insurance and usage. However, an advantage of studying this relationship for hospitalized patients is that there is an element of randomization of patients to physicians. The selection bias from patients choosing physicians, which commonly occurs in the ambulatory setting, is constrained in the hospital. Additionally, use of propensity score matching, including matching patients within a hospital, led to well‐matched samples for the main analyses and subanalyses.
    Given the inability to determine a causal relationship, we do not know if the different rates of cardiac testing and procedures were the result of physician preference, patient preference, or both. We suspect it was the result of both factors: physicians responding to financial incentives and commercially insured patients, who may be more well‐educated and engaged with their health care, preferring more intensive care. We controlled for patients' zip code income quartile, but this is an imperfect measure of their education and actual income.
    A second limitation is that we were unable to determine whether the higher rates of cardiac tests and procedures represented overuse for commercially insured patients or underprovision of care for Medicaid patients. Previous work has quantified overuse of stress tests,33 cardiac catheterizations,47 and SCRM implantations,48 and regions with greater numbers of cardiologists per capita have higher rates of low‐value cardiac tests.49, 50, 51 A third limitation is that we did not have data on the frequency of these tests and procedures done in the ambulatory setting. However, patients with commercial insurance likely had better access to outpatient tests and procedures, so we doubt that our findings could be explained from shifting some hospital‐based procedures to the outpatient setting. Finally, it is beyond the scope of this study to assess the relationship between greater diagnostic and procedural intensity and outcomes. However, prior work found only a weak negative relationship between risk‐adjusted spending for heart failure and acute myocardial infarction hospitalizations and risk‐adjusted mortality rates for these conditions.52
    Commercially insured patients were more likely to undergo a range of cardiac tests and procedures during a hospitalization. This may reflect, in part, financial incentives impacting physician decisions and potentially leading to low‐value care. Our findings support the need for the development of value‐based hospital payment models. Future research should use provider‐level details including compensation models to better understand the impact of compensation on clinical decisions.

    Sources of Funding

    Dr Ellenbogen is supported by the Agency for Healthcare Research and Quality (K08HS028673). The funding agent had no role in the study design, conduct, drafting, or decision to publish the article.

    Disclosures

    Dr Arbab‐Zadeh receives research support from Canon Medical Systems. Dr. Ellenbogen received personal fees for consulting work with Garner Health outside the submitted work. The remaining authors have no disclosures to report.

    Supporting information

    Data S1
    Tables S1–S14
    JMDr. Joseph Marine

    Dr. Joseph Marine

    Cardiology · Cardiac Electrophysiology
    Johns Hopkins Medicine
    josephm