Reverse engineering Medicare, Inc.
By Christopher Weaver, The Wall Street Journal
One doesn’t just wake up one morning and decide to reverse engineer the federal government’s convoluted Medicare Advantage payment system. Taxpayers spend billions of dollars each year on excessive payments to private insurers in Medicare Advantage, but it is shrouded in secrecy and fueled by vast reams of patient data.
For The Wall Street Journal reporting team that cracked this system, the effort began more than a year before we published a single story on the topic. It was the fall of 2022 when a group of Journal reporters and editors began discussing what it would take to lift the veil on Medicare’s opaque and fast-growing move into privatization, along with other trends in the program.
We knew we’d need to get access to closely guarded, patient-level records of services and diagnoses in order to determine how the insurers were generating payments. We also wanted to be able to examine the care patients received in Medicare Advantage.
To get the data, the team rallied around an audacious idea: Propose a research project that could meet the government’s rigid criteria for an exemption to a law protecting patient privacy. The federal Medicare agency allows researchers to access confidential, patient-level information for institutional review board-approved research projects that stand to benefit Medicare.
The team proposed a project examining outliers in Medicare in which the reporters would do the research and pair it with traditional reporting, all the while taking steps to protect patients’ privacy. The idea ultimately won approval from Medicare, a first for any news organization.
Then came the hard part. The trove of data we uncovered consisted of records of 3,874,214,034 services in the most recent year alone. Some of our analyses spanned many years and each service could include dozens of detailed records. Over the course of the investigation, Journal reporter Tom McGinty learned the SAS programming language and wound up writing more than 40,000 lines of code. We let the data guide our reporting, setting up analyses to look at controversial areas of Medicare spending and digging into the outliers.
The reporting team, including Anna Wilde Mathews and Mark Maremont, interviewed dozens of doctors, nurses, and industry executives and government officials to make sense of the data and identify patients and medical providers whose experiences demonstrated the trends we were seeing. Reporters also turned up thousands of pages of company documents showing how Medicare insurers worked the system.
The Journal’s articles in this series found:
- Insurers received $50 billion in Medicare payments for submitting diagnoses — including for HIV — that no doctor or hospital treated from 2019 to 2021, with Medicare Advantage market leader UnitedHealth Group collecting about half that sum.
- Medicare Advantage insurers received more than $1,800 per visit in payments triggered by extra diagnoses recorded during hour-long home exams by nurse practitioners they dispatched to their members’ houses.
- The sickest Medicare patients — those in the last year of their lives — dropped out of Medicare Advantage plans at double the rate of other beneficiaries, netting insurers roughly $6 billion in savings as they avoided covering their care.
- Medicare Advantage plans were paid billions of dollars to cover Medicare benefits for veterans who received some or all of their care from the Department of Veterans Affairs, at no cost to the insurers.
- Doctors working directly for UnitedHealth Group diagnosed medical conditions that trigger extra payments in Medicare Advantage at far higher rates than other physicians, generating billions of dollars in extra revenue for the health care giant.
The project shined a light on one of taxpayers’ biggest expenses and the role of insurers in increasing those costs, generating significant impact.
After the Journal’s reports on insurer-driven diagnoses, the Office of Inspector General for the Department of Health and Human Services recommended for the first time that the Medicare agency stop paying insurers for diagnoses reported solely through home visits, a key source of the extra diagnoses. The Congressional Budget Office concluded that move, paired with another policy tweak, would save $124 billion over ten years.
House Ways and Means Committee staffers began an inquiry last year, led by Rep. David Schweikert, R-Ariz. In February, Sen. Charles Grassley sent a letter to UnitedHealth’s CEO citing the Journal’s findings and demanding that the company turn over extensive documents of its practices. Also that month, the Journal reported that the Department of Justice was investigating UnitedHealth for civil fraud, a probe that included interviews of medical providers named in our reports.
In a November speech, the Justice Department’s former antitrust director also cited the Journal’s findings, saying they showed how Medicare Advantage “insurers exploit the system to charge the government more.”
He added, “A recent Wall Street Journal report found that insurers charge Medicare about $50 billion from 2019 to 2021 for diseases, often unusual, that no doctor treated, none. Insurers earned almost $15 billion from one-hour home visits where nurses just inputted new diagnosis on a tablet often with scant diagnostic equipment.”
Some of the insurers whose practices we reported in detail contested our findings without specifying any specific errors. For instance, in a press release published on its website in December, UnitedHealth said the Journal’s articles “rely on often incomplete and inaccurate data to conduct flawed studies through a murky government ‘agreement’.”
The Journal has stood by its reporting. The Journal’s analyses used data UnitedHealth and other Medicare Advantage insurers themselves submitted to the federal government for payment purposes, and our data use agreement follows the government’s standard template for researchers.
The barriers to taking on this kind of project are steep. Getting the data and learning to use it was a long-term investment of human capital and financial resources. Members of the team had spent years reporting about Medicare and using other types of more readily available Medicare data, such as records of hospital claims and payments to physicians, before undertaking this effort.
But, there are infinitely more stories worth telling hidden in Medicare data. The data is a detailed window into one of the biggest line items of federal spending, and the program fuels the business models of all sorts of health care entities, ranging from local doctors offices to enormous insurers.
If you dive in, make sure you can line up a team with the right kind of expertise to find stories in this sprawling trove. You’ll want to make sure you are up to date on academic research and government audits and reports on the topics you want to examine. You can borrow heavily from the methodologies, often described in detail, of researchers who have covered that ground before you.
In addition, we found it was very valuable to be extremely transparent about our methods with the subjects of our reporting. Ahead of our first article, we provided insurers we wrote about with a detailed methodology that stretched a dozen pages long. Not every company agreed with our approach, and some provided valuable feedback we ultimately incorporated into our methods. It is better to hash out those differences before publishing an article.
Last, this type of data does have limitations. We explored those thoroughly and painstakingly, and backed away from story ideas we didn’t think the data was complete enough to bear out. Reporters will need to keep that in mind and turn to other methods for pursuing certain stories. The data also doesn’t show how the medical providers and insurers in our analyses wound up becoming outliers. In our reporting, we also filled in some gaps by obtaining company records and interviews with current and former employees, including medical providers, that shed light on the practices that led to the sometimes shocking results we documented.