How Data and Analytics Show the Value of Diagnostics in Health Care

It happens all the time: a patient doesn’t get the right diagnostic test at the right time, and her condition worsens—sometimes fatally. But new tools like data and analytics can have an impact by standardizing care protocols and ensuring that diagnostics are ordered when they should be.

Several experts at my firm, Boston Consulting Group, recently partnered with AdvaMedDx (an industry association of medical diagnostics companies) to test that concept. The findings—which we presented at a recent conference sponsored by AdvaMedDx—were both conclusive and compelling. Uneven adoption rates of a key diagnostic test that is clinically indicated means that some cancer patients didn’t receive a type of chemotherapy that could have improved their overall survival rate. At the same time, other patients received that chemotherapy—and thus exposed them to toxicities—without any evidence that it would help them.

To reach these findings, we looked at electronic health records for patients in the US who were diagnosed with metastatic non-small cell lung cancer (NSCLC) from 2011 through 2017. According to guidelines from the National Comprehensive Cancer Network, all patients with this disease should receive an epidermal growth factor receptor (EGFR) diagnostics test to determine if certain chemotherapy agents—namely, tyrosine kinase inhibitors (TKIs) such as Tarceva, Iressa, and Tagrisso—should be used.

 Our source was a data set of de-identified electronic health records from the health care services company Optum?. Ultimately, we examined 5,687 records. The analysis led to several critical findings: 

• Oncologists are increasingly testing patients. From 2011 through 2017, the percentage of metastatic NSCLC patients who received a test for EGFR mutations nearly doubled. Yet the fact that more than 30% of the patients did not receive the diagnostic means that some 1,700 patients in our dataset may have experienced delays in starting chemotherapy or received inappropriate treatment, possibly resulting in faster disease progression and death.
• The choice of an oncologist is the biggest factor behind testing disparities. Men were 40% less likely to receive the EFGR test than women. Patients with no insurance were 50% less likely to have an EGFR test than insured patients. However, the biggest factor in testing disparities was the managing oncologist, whose individual preference for ordering an indicated diagnostic often determines whether a patient receives that test.
• Variations in testing rates correspond to changes in survival rates. We segmented oncologists into quartiles on the basis of their EGFR testing rates. The oncologists in the top quartile were more than twice as likely to order the EGFR test as those in the bottom quartile. Even after adjusting for patient factors, patients treated by top-quartile oncologists saw an 11% decrease in mortality rate, compared with patients treated by bottom-quartile oncologists.
• Some patients received unnecessary chemotherapy, while others who needed treatment did not get it. Some 16% of patients received TKI therapies without EGFR testing. Of those patients, only 30% were likely to have an EGFR gene mutation and therefore benefit from these drugs. The remaining 70% received TKI treatments unnecessarily, putting them at risk from side effects—including immune system suppression, liver toxicity, and gastrointestinal problems—without any chance of benefit. A secondary but not insubstantial issue is cost; TKI medications typically run about $10,000 per month. 

Extrapolating from our findings to the broader US population, we estimate that of the 137,000 US patients with non-small cell lung cancer in 2017, about 30%, or 41,000 patients, were not tested for EGFR mutations. Of those, roughly 7,000 were still placed on TKI therapy, which exposed them to iatrogenic harm without the potential benefit. Additionally, by not testing 34,000 patients with this disease, more than 10,000 life-years were lost. 

 Sometimes progress in health care requires radical new advances in techniques and technology. Other times, it simply requires ensuring that providers and patients follow standard care protocols. By highlighting variations in care among providers and institutions—and the underlying causes of those variations—data and analytics can help physicians and clinicians cut through the fog that clouds decision making, and ultimately lead to better outcomes for patients.