Sources of Low-Value Care Received by Medicare Beneficiaries and Associated Spending Within US Health Systems

This cross-sectional study examines referrals for low-value health care services and associated spending by ordering clinician among Medicare beneficiaries.

Beneficiary and Clinician Attribution to Health Systems: We used the 2018 AHRQ Compendium of US Health Systems to identify health systems. 3We used the CMS Medicare Shared Savings Program attribution methodology as follows: attributed beneficiaries were attributed to the health system that provided the plurality of primary care services across 2017 and 2018 when possible; beneficiaries not receiving primary care services were attributed to a health system via specialist physician services. 4We used the IQVIA OneKey database (which describes the relationships between providers, medical practices, and systems) to determine clinicians' health system affiliations. 5We excluded systems with fewer than 250 attributed beneficiaries and pediatric health systems (based on a Compendium indicator).

Identification of Low-Value Services:
We used 40 claims-based, low-value care definitions for services relevant to older adults, leveraging our prior work and the Milliman MedInsight Health Waste Calculator (version 8.0). 1,6Non-Waste Calculator-based measures were antipsychotics for patients with dementia, percutaneous coronary intervention (PCI) for asymptomatic patients, injection for low back pain, feeding tubes for patients with dementia, and short-interval repeat dual-energy X-ray absorptiometry (DEXA) scan.The Milliman MedInsight Waste Calculator is a propriety, stand-alone software that uses International Classification of Diseases, Ninth and Tenth Revision (ICD-9 and ICD-10) diagnosis codes; American Medical Association procedural codes; and National Drug Code entries to assign health care services provided within specific clinical scenarios to one of three categories: not wasteful, likely wasteful, or wasteful.We defined low-value services conservatively in this work by only including those tagged as wasteful (i.e., we required the Calculator to flag with Degree of Certainty = "W" (Wasteful) to categorize them as low-value).We also required the Sufficient History flag = "Y" for all services except cervical cancer screening, prostate-specific antigen testing (PSA) testing, and screening electrocardiograms (ECGs), which had maximum lookback periods exceeding 1 year.
To define an incidence of a low value service, we counted multiple instances received by the same beneficiary but required a certain time difference between claim dates for discrete services (see eTable 1 for details) and counted only the first observed instance of service measurement occurring over a prolonged timeframe (feeding tubes for patients with dementia and long-term medications).We used 2016-2018 data to identify beneficiaries eligible for each service and to measure receipt of each low-value service among these beneficiaries in 2017-2018.
Physician Identification for Claims: Based this on consultation with experts in CMS clinical coding, we ascribed each low-value service to its ordering clinician using the NPI on the claim as follows: "referring NPI" (for non-drug services identifiable using the professional claims file), "attending NPI" (for non-drug inpatient and outpatient facility claims), or "prescribing NPI" (for drugs).For Waste Calculator-based measures, we identified the ordering clinician based on the National Provider Identifier (NPI) on the "trigger" claim line flagged by the Waste Calculator as Wasteful.For non-Waste Calculator-based measures, we prioritized the referring NPI from the professional claim whenever possible.
To identify clinician type, we used the self-reported Provider Enrollment, Chain and Ownership System (PECOS) data (MD-PPAS file).PCPs were defined by specialty in general practice, family practice, internal medicine, pediatric medicine, or geriatric medicine.Specialist physicians were defined as those with any other specialty.Advanced practice clinicians included nurse practitioners, certified clinical nurse specialists, and physician assistants.The attributed PCP was the clinician providing the plurality of a patient's primary care services in 2017-2018.
Analyses: We calculated the share of each service originating from patients' attributed PCPs, other insystem PCPs, in-system advanced practice clinicians and in-system specialists.When calculating "All low-value services by clinician type," we multiplied Part D-based measures (which used 40% sample data) by 2.5 to ensure comparability across measures.
We used narrow and broad definitions to calculate a range of in-system Medicare spending on these lowvalue services from 2017-2018.When reporting spending comparisons in the results (i.e., between clinician types and for total vs in-system spending), we used the narrow definition only.For the narrow (claim-line) definition, we counted only payments associated with a claim line identified as low-value.For the broad (claim case) definition, we included the entire claim payment if a component claim line was identified as low-value. 7As charges are not itemized on claims for non-drug services provided in inpatient and skilled nursing facility settings, we excluded these claims from the spending calculations.We multiplied Part D-based measure estimates (which used 40% sample data) by 2.5 to ensure comparability across measures.
Following the Centers for Medicare and Medicaid Services data reporting rules, we suppressed results with 1-10 services and any values that would allow back-calculation of a result requiring suppression; when the latter occurred such that an additional cell required suppression, we chose the smallest cell to suppress.To construct the figure, we used a random number generator to assign values to suppressed results based on the difference between the unsuppressed row totals and the sum of the unsuppressed cells.

eTable. Low-Value Measure Definitions Measure Eligible Beneficiaries Flagged as low-value Calculator -Based? Counted Once per Beneficiary or Once per Service Measure Description
© 2023 Chant ED et al.JAMA Network Open.© 2023 Chant ED et al.JAMA Network Open.© 2023 Chant ED et al.JAMA Network Open.© 2023 Chant ED et al.JAMA Network Open.© 2023 Chant ED et al.JAMA Network Open.© 2023 Chant ED et al.JAMA Network Open.© 2023 Chant ED et al.JAMA Network Open.aSufficient history requirement waived for this measure.