Inappropriate Laboratory Testing: Significant Waste Quantified by a Large-Scale Year-Long Study of Medicare and Commercial Payer Reimbursement Open Access

Medicare outpatient claims data were accessed through the CMS Virtual Research Data Center under institutional review board approval Diaceutics INC-DIA16-001. Claims from 146 012 commercial insurance plans were obtained through separate, confidential data purchase agreements with insurance claim clearinghouses. These included plans from all major insurance carriers including Aetna, Blue Cross Blue Shield, Cigna, Humana, Kaiser Permanente, United Healthcare, and others. Care was taken to exclude any data from the commercial database that also referenced Medicare or Medicaid derivatives as the payer. This step was essential to prevent duplication of patients across the 2 datasets. Current procedural terminology (CPT) codes were used to query the databases for vitamin D, PSA, lipid panel, and HbA1c (Table 1) tests with dates of service from January 1 to December 31, 2019. Data were queried and analyzed using SAS Enterprise Guide, KNIME (Knime Inc), and Microsoft Excel. All claims represented tests that were performed, not just ordered, and all data were from adults.

Source tables for both cohorts contained large amounts of data. The single-line library file for CMS 2019 data alone contained 1.84 billion rows of data across 83 columns. This volume of big data necessitated the use of aggregation functions to streamline computational resources and reduce analysis time. Even so, each alteration to the programming code resulted in run times spanning many hours to days. Hence, for this study we chose to study a limited number of variables.

There are at least 2 scenarios where test frequency (ie, the number of times a patient is tested within a given period) defines inappropriate testing: (1) when a patient need not be tested at all; for example, preoperative tests for healthy patients undergoing low-risk surgery^14,15  and (2) when a patient is tested at a frequency greater than what is required to inform their medical management. Scenario 1 can be considered a special case of scenario 2 in which the appropriate test frequency is 0. In this study, we focused on the second scenario, investigating tests used repeatedly for diagnosis or for monitoring health. Such an approach, inevitably, depends on selection of appropriate thresholds above which testing is considered “inappropriate.” We used a series of national guidelines from professional organizations and payers and reports of prevalence of conditions relevant to the tests being studied to estimate what we consider appropriate frequencies for repeat testing. These guidelines included those issued by the American Cancer Society, American College of Cardiology, American Diabetes Association, American Heart Association, the Centers for Disease Control and Prevention, the US Preventive Services Task Force, and others (see Table 1 for details). We also considered payer guidelines for coverage, including CMS and commercial payers (references included for each specific test). Of note, some tests are recommended at frequencies less than 1 per year; for instance, the American Urological Association prostate cancer guideline pertinent to the 2019 study period recommended a screening interval of 2 or more years.¹⁶ 

We generally followed published guidelines for setting appropriate testing thresholds when the recommendations were clear and unequivocal. When faced with multiple conflicting guidelines, or when the recommended testing frequency was too infrequent, we adopted a more tolerant approach to establishing thresholds. To clarify, a recommendation of “1 test every 2 years,” effectively implies 0.5 tests per year, and “1 test every 4–6 years” corresponds to 0.16 to 0.25 tests per year. Given that our study spanned 1 year, instances where the recommended frequency fell below 1 test per year were assigned a more tolerant threshold of 1 test per year. This adoption of more tolerant thresholds throughout our study resulted in fewer tests being identified as excessive, indicating that our overall estimates of unnecessary testing are likely conservative and not overestimates. To find the number of excess tests per patient (Table 2), we subtracted the threshold annual test frequency for each analyte or panel, given in Table 1, from total tests performed.

Costs of excessive testing were calculated by multiplying the number of excess tests by the reimbursement attached to the test CPT code in 19CLABQ4, the quarter 4 clinical laboratory fee schedule.¹⁷  This approach was applied to all claims sources, as the commercial payer data accessed did not include payment information. However, Medicare data did include payment details. Commercial reimbursement is generally higher than CMS reimbursement,¹⁸  therefore our excess expense calculations are likely significant underestimates.

The Medicare dataset had approximately 64 million individuals enrolled, as determined by analysis of the Medicare master base summary file. The commercial health plan dataset had approximately 168.7 million insured individuals from 146 012 health plans across all major insurance carriers, representing more than 76% of all commercially insured patients in the United States. While calculation of the exact proportion of total US adults included in these datasets is complicated by the multiply insured or uninsured, we believe the number of individuals included in this study represents a significant fraction of the US adult population, estimated as 255 million in 2019.¹⁹ 

The age and sex distribution for both Medicare and commercially insured populations is shown in Figure 1, A and B. Table 1 shows the proportion of individuals in both datasets who were tested for each analyte. We assessed testing frequency for individuals across 4 tests in 2019 (Figure 2, A through D). Most individuals tested were tested at a frequency of once per year (see Table 1, dark bars in Figure 2, Athrough D). However, considerable numbers of individuals received tests at a greater frequency than given in Table 1. This applied to all tests in both Medicare and commercial health plan subscribers (Figure 2, Athrough D). The numbers of individuals receiving testing greater than threshold, and associated costs, are summarized in Table 2. In our study population of 232 million, we found that approximately 14.4 million of approximately 60.5 million (23.8%) people had excessive laboratory testing.

In 2015²⁰  and also in 2021 (after our main study period),²¹  the US Preventive Services Task Force found insufficient evidence to determine the balance of risks and benefits of screening asymptomatic adults for vitamin D deficiency. Some commercial payers have ceased covering routine vitamin D testing.^22,23  However, CMS guidelines recommend testing in individuals with 29 specified conditions associated with a high risk for vitamin D deficiency. The CMS guideline states that it is not reasonable and necessary to perform more than 3 tests per year even in this high-risk population.²⁴  Only approximately 10% of the population has vitamin D deficiency, although this can vary by race, age, season, and other factors.²⁵  Guidelines for the remaining 90% who are not vitamin D–deficient range from 0 to 1 test per year. Therefore, we considered 1 annual vitamin D test an appropriate test frequency. We found 18.3% (1 879 006 of 10 256 261 individuals) were tested twice or more per year (Table 2; Figure 2, A). On average, 2776 individuals were tested monthly and 4 patients were tested weekly. Total excess tests across all test frequencies for vitamin D in Medicare were 3.4 times those seen in the commercial data (2 209 389 versus 659 565).

Prostate cancer is the third leading cause of cancer deaths in the United States, with approximately 224 000 new cases in 2019. Guidelines for intervals of screening for prostate cancer vary widely. The American Cancer Society recommends PSA testing for all men older than age 50, and for men older than the ages of 40–45 in certain high-risk groups.²⁶  The US Preventive Services Task Force suggests annual testing for men aged 55–69 but not for men older than age 70 due to a higher risk of death from other causes.²⁷  CMS suggests annual PSA testing for men older than age 50.²⁸  The American Urology Association recommends PSA testing at most every 2 years.²⁶  Considering the different guidelines together, the recommended frequency of screening tests per year varied from 0 for men less than age 45 to 50, to 0, 0.5, or 1 for men older than 50. Consistent with our practice of setting tolerant thresholds, we set 1 test per year as an acceptable test frequency for most men. For monitoring certain groups of men diagnosed with intermediate and low-risk prostate cancer, the American Cancer Society recommends PSA testing every 6 months, or 2 tests a year—the highest recommended testing threshold.²⁹  Our study revealed that 20.1% (973 753 of 4 834 915 people tested) were tested twice or more, 7.1% (343 782 of 4 834 915) were tested more than 3 times, and 0.1% (5997 of 4 834 915) were tested monthly on average (Table 2; Figure 2, B). Total excess tests across all test frequencies for PSA in Medicare were 2.7 times those seen in the commercial data (1 240 256 versus 463 574).

Cardiovascular disease is the leading cause of death in the United States, with 720 000 new and 335 000 recurrent coronary events in 2019. Testing for cholesterol and other lipids is commonly used to approximate the risk of cardiovascular disease. The 2018 American College of Cardiology/American Heart Association guidelines recommend checking cholesterol levels every 4 to 6 years for healthy adults; more frequently for those at higher cardiovascular risk.³⁰  CMS covers 1 lipid panel annually but allows more frequent testing of individual panel components for patients on lipid-lowering medications.³¹  Despite the recommendation for 1 panel every 4–6 years,³⁰  we took a more tolerant approach of considering 1 panel/y to be an acceptable frequency for most individuals, thereby resulting in fewer tests being flagged as excessive. Of the 25 784 504 individuals tested, 23.5% (6 059 605) were tested twice a year or more often, and 0.04% (11 408) were tested at least 12 times, averaging once a month (Table 2; Figure 2, C). People with Medicare were tested with lipid panels 2.6 times more than individuals covered by commercial insurance (5 696 534 versus 2 223 927).

An estimated 38.1 million adults in the United States have diabetes, causing an estimated 409 000 deaths in 2019. For nondiabetic individuals, the Centers for Disease Control and Prevention recommends 1 HbA1c test every 3 years.³²  For diabetics with stable glycemic control, the American Diabetes Association recommends measuring HbA1c twice a year, and for diabetics with changing therapy or not meeting glycemic goals, the American Diabetes Association recommends 4 tests a year, once each quarter.³³  Regarding Medicare fee-for-service beneficiaries, 27.5% had a diabetes diagnosis in 2019,³⁴  and it would be appropriate to test these patients 2–4 times each year, varying with the degree of glycemic control.³⁵  More than 1 test a year may be excessive for nondiabetics, and more than 4 tests for some diabetics, so our choice of 1/y for a majority nondiabetic population allows for greater tolerance in choice of appropriate test frequency. Therefore, we considered these 2 levels of excess testing (ie, >1 and >4) separately.

We observed that 28% (5 494 121 of 19 603 725 individuals tested) were tested twice or more, and 1.7% (327 016 of 19 603 725) received 5 or more tests in the year, with 0.03% (6756 of 19 603 725) getting tested at least once a month (Table 2; Figure 2, D). Medicare’s excess HbA1c tests were 4.8 times more than commercial payers (7 959 971 versus 1 643 474; Table 2).

Our analysis of 2019 claims data on CMS’ Virtual Research Data Center showed that Medicare spent approximately $7.04 billion on laboratory tests; that is, tests with CPT codes ranging from 80000 to 89999. The percentage of testing identified as excessive differed for each test analyzed, ranging from 27.7% for lipid panel tests to 46.6% for HbA1c tests (Table 2). By extrapolating these percentages to all laboratory testing, we estimated that excessive testing contributed between $1.95 billion and $3.28 billion to the total Medicare expenditure on laboratory tests in 2019. Reimbursement data for commercial claims are not publicly available due to antitrust constraints. Therefore, we estimated costs of testing to commercial payers by applying CMS payment levels¹⁷  to the commercial test volumes. Excess testing among individuals covered by commercial payers was 30.5% of that seen in CMS data on average, giving total cost estimates ranging from $594 million to $1 billion. Summing the estimated total costs of excess testing for the population covered by CMS and commercial payers gives $2.54 billion to $4.28 billion in 1 year.

This is the first report to our knowledge of a large-scale measurement of low-value testing across both Medicare and commercial payer populations for commonly used laboratory tests over a full year. The population studied included the entire Medicare enrollment for 2019, more than 64 million individuals, and more than 168 million individuals covered by commercial payers, with greater than 78.2 million tests included in the analysis. For comparison, this is nearly 50 times the number of tests examined in a previous large study.⁵  A considerable number of individuals who were tested underwent excessive testing, defined as receiving a greater number of tests than the thresholds we established using a combination of clinical and payer guidelines and the published prevalence of relevant conditions.

After conducting an extensive review of both clinical and payer guidelines, we established test frequency thresholds and applied them to the entire populations under investigation. We need to specifically note that for a significant number of individuals within these populations, any testing at all may be inappropriate testing. For example, consider guidelines that recommend testing only once every 2 years; in such cases, the appropriate number of tests within a single year might indeed be 0 for patients who were tested in the previous year. Our overarching approach has been one of greater tolerance in setting thresholds to identify overuse of tests. Opting for a threshold of 1 test per year in all cases where the recommendation was for less than 1 test a year resulted in fewer tests being flagged as excessive. Despite this accommodating stance, our study found that the extent of excessive testing was considerable.

The same lab test can be utilized for multiple indications, with correspondingly different testing thresholds. The PSA test is used both for screening potential cases of prostate cancer and also for monitoring prostate cancer in specific situations where immediate treatment is not essential. It is reasonable to assume that in a population-wide study, the majority of PSA tests are conducted for screening purposes. A much smaller proportion of tests are administered to individuals with intermediate and low-risk prostate cancer undergoing active surveillance. The protocol for active surveillance typically involves a PSA test every 6 months, thus justifying a frequency of 2 tests per year for this subgroup. To our knowledge, widespread implementation of PSA testing occurring more than twice a year, as observed in a significant portion of our study’s patient population, is not common. Commercial data exhibited lower rates of excess testing, both in total number of tests and the proportion of tests, compared to Medicare data. This difference could be attributed to Medicare patients being older and requiring certain tests, such as PSA, more frequently. However, it is important to note that a higher volume of testing does not necessarily imply a higher rate of excessive testing. It may, therefore, suggest that commercial payers scrutinize test orders more rigorously, along with higher denial rates, compared to Medicare, which had 3.4-fold more excess tests than the commercial payer group, with the highest test frequency averaging 1 test per week for all 4 analytes. Commercial payers had similar or higher excess test volumes to Medicare but only at lower levels (eg, <20 tests per patient per year for vitamin D, <13 for PSA, <20 for lipid panel, and <14 for HbA1c). Frequencies higher than 20 tests per patient per year were only seen for Medicare (see Figure 2, Athrough D). Further analysis of denial rates, actual payments, and responsibility for denied claims may shed light on the differences in excess testing rates between Medicare and commercial insurers, although this is outside the scope of the current study.

Vitamin D, PSA, lipid panel, and HbA1c were considered suitable for our study because they are commonly used tests with known indications and accessible testing frequency recommendations, and are in the top 25 by volume of all laboratory tests used in the United States.¹³  Almost all of these top-25 tests are procedures that individuals may undergo 1 or more times within a specific timeframe, potentially leading to overuse. They are typically ordered in outpatient settings and have a clear fee-for-service structure, allowing for more accurate measurement of test frequency from billing data. These features enabled us to define overuse for these tests, something not possible for all laboratory tests.³⁶  In contrast, reimbursement for inpatient testing is usually bundled with other charges associated with a diagnosis-related group, and is, therefore, almost impossible to isolate and analyze. The tests we examined closely resemble most other tests in the top 25 Medicare tests in terms of test volumes, reimbursement, and total Medicare expenditure. However, there were some differences in the degree of overuse among the 4 tests we focused on. Consequently, some variance should be anticipated when estimating overuse across all tests. Our analysis enabled us to project the magnitude of overuse, although the precise dollar amount can only be determined by individually examining each of the approximately 5000 tests.

Small numbers of individuals, including 11 for vitamin D, 24 for PSA, 66 for lipid panel, and 31 for HbA1c, were tested on average every 2 weeks. The highest testing frequency observed was about 1 test per week for all analytes. Such frequent testing has no clinical or physiologic relevance for these biomarkers. Testing frequency must align with the physiologic activity and lifespan of the molecule being tested. For example, HbA1c is formed by the irreversible binding of glucose to hemoglobin A, which occurs within red blood cells that have a lifespan of about 120 days. Consequently, measuring HbA1c levels weekly (31 patients), as we observed, would not be useful even if resources were unlimited. Since HbA1c levels are related to the 120-day lifespan of red cells, measurement more than 4 times per year represents a testing frequency higher than can be justified on the basis of the physiology of the biomarker, let alone clinical considerations.

A meta-analysis of laboratory test utilization studies ending in 2012 estimated 10%–30% of lab tests were unnecessary,⁵  which extrapolates to a waste of $700 million to $2.113 billion based on our figure for Medicare laboratory testing in 2019. Results from our study, which examined at least an order of magnitude greater number of patients and claims, produced a lower limit of the estimate that was more than 2.8 times greater, and an upper limit of the estimate that was more than 1.5 times greater than those of the 2012 study.⁵  We attribute these significant differences at least in part to the difference in scale of the 2 studies. Our study examined only 4 lab tests, but the excess Medicare testing costs, calculated at approximately $269 million (from Table 2), represent 3.8% of Medicare’s total spending on laboratory testing. The tests were not chosen based on a priori knowledge of high rates of excess testing, so the overall excess testing costs across all laboratory testing codes may well be higher. Our calculations suggest that Medicare may be spending billions of dollars on excess testing based on the observed rate for the chosen 4 tests, which comprise approximately 14% of all CMS laboratory test volume.

Reducing LVC could have a transformative impact on healthcare.³⁶  These savings could be used to increase testing for individuals who are not being tested at recommended frequencies, such as screening to check for diabetes covered by Medicare Part B, a benefit only 4% of Medicare beneficiaries took advantage of, in 2020.³⁷  Savings could also support the reimbursement of higher-cost, underreimbursed tests, such as molecular tests, biomarker tests, or next-generation sequencing, which are critical for patients with cancer and genetic disorders. A study on non–small cell lung cancer patients found that not ordering biomarker tests for targeted therapies limits the benefits of personalized medicine in clinical settings.^38–40  Underreimbursement of these tests could be a reason for the failure to order appropriate tests.^41,42  This study only considered reimbursement to the laboratory based on the 2019 Clinical Laboratory Fee Schedule.¹⁷  Thus, the costs of excess testing paid for by commercial payers is likely to be significantly underestimated as commercial insurers have been shown to pay between 116% to 173% more than Medicare for the same services.¹⁸  Additional costs related to excess testing not included in this study include the labs’ costs of following up on denied claims to cover their testing costs, costs incurred by the ordering practice due to denied claims, and costs to the individual being tested, the last often being overlooked in excess testing considerations. Approximately 14.9% of Medicare claims were not paid, and labs performing these tests may have absorbed the loss to maintain their relationships with those who are ordering tests. Patients have copays and coinsurance payments, and in some cases, self-pay costs. Patients may also incur travel and time off work costs. There are additional costs for unnecessary blood draws, doctor visits, and pretests and posttests for both payers and patients. Lab tests are generally inexpensive: the average total reimbursement for the top 90% of test volume for Medicare laboratory testing being $23.37 per test.¹³  However, small errors or variations in testing can lead to unnecessary downstream investigations, hospitalizations, and delayed hospital discharges, resulting in higher costs. More frequent testing will proportionally increase errors and variations. Additionally, excess testing may cause overdiagnosis, reclassifying healthy people with mild problems or at low risk as sick, which could result in overtreatment and unnecessary care, leading to patient harm while utilizing resources that could be “better spent treating and preventing genuine illness.”⁴³  Finally, frequent testing may cause patient anxiety.⁴⁴  Many unnecessary tests are ordered despite widely recognized initiatives like “Choosing Wisely,”⁴⁵  or the “Less is More” series.⁴⁶  Likely reasons are complex and multifactorial and include the use of more familiar panels when individual tests might suffice, or ordering systems that bundle tests (eg, bundling fasting glucose and HbA1c, which results in excessive testing for the latter). Other reasons for excessive testing include poor communication between different electronic health record systems, lack of laboratory medicine education, defensive medicine, patient requests, and fraud.^36,38,47,48 

This paper does not explore the underlying reasons for excess test ordering, and further research involving various stakeholders is needed to fully understand this issue. Hospitals and laboratories play a role in facilitating excess testing. Inpatient testing is typically reimbursed based on diagnosis-related group, with built-in penalties for overtesting, encouraging utilization management or stewardship programs. However, outpatient procedures, reimbursed on a fee-for-service basis, lack incentives for hospitals, laboratories, and providers to reduce testing frequency. Laboratories could enhance their systems to facilitate the identification and ordering of appropriate tests, emphasizing the importance of individual tests over comprehensive panels. This, coupled with greater awareness among clinicians regarding the availability of individual tests as an option for monitoring treatment, will promote more precise and judicious use of tests. Payers could exercise greater scrutiny in reimbursing tests⁴⁹  at inappropriate frequencies, although this might inadvertently shift costs to patients; measures must be implemented to prevent this cost transfer. Finally, there is a need for greater awareness on the part of patients⁵⁰  who are paying for these tests with not just their time and money but also their blood.

The study focuses on 4 laboratory tests, but more can be added in the future for a more comprehensive analysis. The definition of excess testing is complex, especially for analytes without defined guidelines.³⁶  Appropriate testing frequency may vary for each patient; for instance, this study did not consider how many patients tested for HbA1c had uncontrolled diabetes. It is important to note that no analysis was done on the diagnosis codes applied to each patient for excess tests, which could refine the test frequency threshold used to define excess. Additionally, it may be important to determine the time interval between tests and use a larger time window for analysis, indicating that some patients are receiving tests more frequently than recommended. By analyzing intervals between tests, it may be possible to identify periods of higher-than-normal testing related to the patient’s condition, which may be obscured by averaging over a year. Of interest are artificial intelligence studies that predict if testing a given patient for a given analyte at a given time is useful.^51,52  While such personalized analysis could become more common, clinical guidelines are expected to remain the primary method for determining appropriate test frequencies in the foreseeable future.

The year 2019 was chosen to avoid confounding effects of the COVID-19 pandemic on test utilization.¹¹  Our analysis of test utilization for CMS and commercial payers using 2022 claims data for HbA1c testing indicated that there was minimal change in the proportion of excess testing at each of the cutoff-point values. The proportion of more than 12 tests for both payers has slightly increased and is slightly lower for more than 2 and more than 5 tests (Table 3), with similar distributions for CMS and commercial claims in both years (not shown). The maximum number of tests for a given patient for CMS was 54 in 2019 and 53 in 2022, while for commercial payers, it was 19 in 2019 and 16 in 2022. These findings suggest that there may not have been a significant change in excess testing between 2019 and 2022. The LVC problem likely persists, with significant resources wasted on tests that offer little or no value to the individual receiving them. The 2019 data serve as a baseline both for studying the effects of the COVID-19 pandemic and future developments in the landscape of LVC laboratory testing. This study, like most LVC studies, focuses on “volume rather than value.”³⁶  While calculating LVC expenses is feasible with claims data, determining the negative impact on patients and healthcare systems is more complex, and difficult to measure on a national scale. In conclusion, this study shows that 4 commonly ordered laboratory tests are significantly overused. This overuse provides no value to patients and wastes billions of healthcare dollars that could be better used.

The authors gratefully acknowledge Karen Prince for her expertise in preparation of the figures; Susanne Munksted, MS, for helpful discussions; and Brian Jackson, MD, PhD, and Harsh Thaker, MD, PhD, for comments on the manuscript.