Lab test under- and overuse
November 2024—In “Lab test use: what 1 billion claims tell us” (August issue, page 1), you detailed the conclusions of a research study that analyzed insurance claims from effectively the entire U.S. adult population. The study found what it had sought, that 14.4 million individuals tested had undergone excessive laboratory testing, which is low-value care that poses patient risks and wastes resources. In reviewing the study data and analyses, I found that the methodology did not accurately estimate overuse and that the study overlooked the conclusion the data was screaming for: We are screening far too few people for chronic diseases and need to invest in access to care.
As reported in CAP TODAY, the study started four years ago with the intent to find over-ordered tests using insurance claims data from 232 million U.S. adults with Medicare and commercial health insurance, with a focus on vitamin D, prostate-specific antigen, lipid panel, and hemoglobin A1c. The study is distinguished by its scale, which effectively includes claims for the entire population of 255 million U.S. adults when you consider that about eight percent of the population is uninsured.
Taking hemoglobin A1c and diabetes as the first example, the study finds that each year 19.6 million U.S. adults get one or more hemoglobin A1c tests. To put this figure in perspective, out of 255.2 million U.S. adults, just over half (135.7 million) have diabetes or prediabetes, yet only eight percent (19.6 million) get one or more HbA1c tests each year. The 92 percent of adults who went untested are the main story that the data tells, yet the study conclusions treat this as an afterthought.
The study continues by finding that of the 19.6 million U.S. adults who get HbA1c tests, 5.5 million get two or more HbA1c tests per year and 300,000 get five or more. The 5.2 million people who get two to four tests per year are categorized as having received inappropriate, low-value testing because diabetes screening guidelines recommend up to one HbA1c test per year. While the study does recognize that the American Diabetes Association recommends two to four HbA1c tests per year for people with diabetes, it counts all 5.2 million people who get two to four tests toward overuse as if we can assume they don’t have diabetes.
Who are the 5.2 million U.S. adults who get two to four HbA1c tests a year? Are these affluent people without diabetes who overtest out of privilege? Or are these people with diabetes who are testing two to four times a year like the ADA recommends? The study does not include a review of HbA1c results nor does it include an analysis of diagnosis codes, so we can’t say what the mix is. There are 38.1 million U.S. adults with diabetes. One study of Aetna members showed that 67.1 percent of people with controlled diabetes got the two tests per year recommended by the ADA, while 39.4 percent of people with uncontrolled diabetes got the four tests per year recommended by the ADA. Statistics may vary across studies, but it is incontrovertible that some people with diabetes get the two to four HbA1c tests they’re supposed to. Counting these 5.2 million people as people without diabetes who are committing wasteful overspending is a mischaracterization.
The study then quantifies the cost of HbA1c tests given to the 5.5 million people with two or more tests as $104 million of excessive laboratory testing. This $104 million is added with the other three conditions studied to get to $350 million, and then extrapolated to get a range of $2.54 to $4.28 billion across all laboratory tests. None of this math holds up when the $104 million is actually spent testing diabetics according to the guidelines.
The study’s conclusions follow the same pattern for lipid panels, PSA tests, and vitamin D tests. The study counts 6.1 million U.S. adults with two or more lipid panel tests as excessive testing because this exceeds annual testing guidelines for healthy people. Assuming these 6.1 million people are all healthy and counting their tests as excessive does not hold water when the CDC estimates that 86 million have high total cholesterol (above 200 mg/dL), 25 million have very high total cholesterol (above 240 mg/dL), 43 million have low HDL cholesterol (below 40 mg/dL), and 47 million are currently taking medication for cholesterol.
The study counts 974,000 U.S. adults with two or more PSA tests as excessive testing because this exceeds guidelines for healthy men. The American Cancer Society recommends two tests per year for people with intermediate and low-risk prostate cancer, and there are an estimated 3.4 million men living with prostate cancer in the U.S. Vitamin D is more difficult to quantify as it is linked to such a wide range of health conditions. But the main message I see in the study data is that 96 percent of U.S. adults don’t receive vitamin D screening.
I’m thankful that Ila Singh, MD, PhD, gives mention to the need to test the undertested, and I’m glad the research team is planning a follow-on publication to quantify undertesting. I hope the second publication corrects the estimates of overtesting by accounting for the testing needs of people who have health conditions. And I hope the second publication quantifies the extent of undertesting for the same set of tests.
The $350 million characterized as “wasteful” laboratory spending in this publication is some of the most valuable spending in our country’s trillion-dollar fight against chronic disease. The data in this study shows we are falling far short of delivering the tests that are medically necessary for screening the undiagnosed population and for managing the diagnosed population. Let’s ensure our policymakers understand the real story the data tells.
Eric Olson
Founder, Babson Diagnostics
Austin, Tex.
◆ Dave Smart, PhD, senior director of analytics, Diaceutics, Belfast, Northern Ireland; Jeff Schreier, BS, MBA, former senior director of partnerships, Diaceutics; and Ila R. Singh, MD, PhD, professor of pathology and immunology, Baylor College of Medicine, Houston, reply:
Your insights highlight important aspects of laboratory testing utilization, particularly the balance between overuse and underuse in chronic disease management.
Our study analyzed laboratory test utilization patterns on a national scale, examining more than 1 billion claims from approximately 232 million individuals, effectively covering the entire U.S. population, as you accurately noted. The unprecedented scale of our data set presented significant challenges in incorporating detailed clinical information, such as diagnostic codes, which would help determine testing needs more precisely. While these data were available, analyzing them for such a vast population was a significant computational challenge. Many patients had multiple comorbidities, and considering a matrix of diagnoses alongside tests would result in dozens or even hundreds of combinations per individual. The Centers for Medicare and Medicaid Services 2019 data alone consisted of 1.84 billion rows across hundreds of columns located in different tables. The commercial payer data from more than 140,000 plans had even greater complexity. Despite using aggregation functions to minimize computational load, each alteration to the programming code added significant processing time.
Our study represents both an exploration of the feasibility of this innovative big data approach and the substantial findings it yielded. Now that we have demonstrated the capability to handle such large data sets, we anticipate that we as well as other investigators will undertake more complex studies incorporating diagnostic codes and additional variables to calculate the correct number of tests needed at the individual level. This will allow us to refine our assessments and provide a more comprehensive understanding of laboratory test utilization in the future.
We recognize that establishing appropriate test frequency thresholds is challenging for all utilization studies, especially at such a large scale, and we stressed this in our paper. After an extensive review of clinical and payer guidelines, we established test frequency thresholds and applied them to the populations under investigation. In cases where guidelines recommended testing frequencies less than once per year (e.g. one test every two years, or even four to six years as in the case of lipid panels), we adopted a more tolerant threshold of one test per year. This approach resulted in fewer tests being flagged as excessive, indicating that our estimates of unnecessary testing are likely conservative rather than overestimates. Despite this accommodating stance, our study found that inappropriate testing remains considerable.
Regarding HbA1c testing, we acknowledge that patients with diabetes require more frequent monitoring—typically two to four times per year—as recommended by the American Diabetes Association. Without diagnostic codes, we could not precisely determine how many individuals receiving two to four tests were diabetic. However, given that more than 70 percent of the Medicare population is not diabetic, it is likely that a substantial portion of these tests were performed on nondiabetic individuals. Additionally, we found that a significant number of individuals received HbA1c testing at frequencies exceeding four times per year, with some tested monthly or even weekly. Such frequencies surpass clinical guidelines even for patients with uncontrolled diabetes and are physiologically unnecessary.
For lipid panels, guidelines generally recommend one panel every two to six years for risk assessment in healthy individuals, with annual testing for those at higher risk. Monitoring lipid-lowering therapy should focus on individual panel components, such as LDL-cholesterol, rather than repeating the full lipid panel. In our study, we focused exclusively on the lipid panel and considered one test per year as an acceptable frequency. We observed that more than 6 million individuals received two or more lipid panels in a single year, with some tested monthly. Repeating full lipid panels at such frequencies is generally not recommended and does not provide additional clinical benefit.
Clinicians may order familiar test panels even when individual tests would be more appropriate for monitoring treatment, as is the case with lipid panels versus individual components. We hope that laboratories can facilitate change by enhancing their systems to emphasize the ordering of appropriate individual tests over comprehensive panels. This, coupled with increased clinician awareness, can promote more precise and judicious use of tests.
It is reasonable to assume that the majority of prostate-specific antigen tests in a populationwide study are conducted for screening purposes. We considered one PSA test per year as appropriate for most men—a tolerant threshold when the American Urological Association suggests most men be tested once every two years. We acknowledge that patients under active surveillance for prostate cancer may require more frequent monitoring, typically every six months, justifying two tests per year. However, our findings showed that a significant portion of individuals received PSA testing more than twice per year, with some tested monthly. Such frequent testing is unlikely to be clinically justified for the majority of patients.
We share your concern about undertesting, representing missed opportunities for early detection and intervention. While our study focused on test overuse, we recognize the importance of addressing both overuse and underuse to optimize patient outcomes.
The tests selected for our study were chosen based on their presence in the list of top 25 Medicare tests, and the availability of clear guidelines and coverage determinations, which allowed us to estimate appropriate testing frequencies. This approach is not feasible for other tests, such as blood counts or metabolic panels, due to the lack of such guidelines. The tests we evaluated serve as reasonable proxies for other high-volume tests within Medicare’s top 25, providing a solid foundation for our projections. While some variation is anticipated when estimating overutilization across all tests, our analysis indicates that Medicare may be incurring direct excess expenses ranging from $1.95 to $3.28 billion annually due to excessive laboratory testing. Accurately determining the exact dollar amount would require a detailed review of each of the approximately 5,000 tests, incorporating specific clinical information at the individual level.
Our intention is not to undermine the value of necessary laboratory testing but to highlight areas where test utilization may not align with clinical guidelines, potentially leading to unnecessary health care costs and patient burden. We appreciate your insights and agree it is essential to balance the reduction of low-value testing with ensuring adequate screening and monitoring for those who need it. We would hope that greater stewardship of finite resources would enable a realignment of those resources from areas offering little or no value in patient management to areas potentially offering significant value to patients, such as earlier cancer diagnosis, staging, and molecular profiling.
By fostering collaboration among laboratories, clinicians, and payers, we can promote the appropriate ordering of tests and enhance patient care. Laboratories can help by improving systems to facilitate the identification and ordering of appropriate tests. Payers could exercise greater scrutiny in reimbursing tests at inappropriate frequencies, though measures must be implemented to prevent cost transfers to patients.
Your thoughtful engagement with our study contributes to the important dialogue on optimizing laboratory test utilization for the benefit of patients and the health care system.