An Epic Dystopia

This appears in the October 2024 issue of The American Prospect magazine. Subscribe here .

I periodically see three wonderful doctors, my internist and two specialists. They all know that I'm a journalist who once wrote for . Every time I see them, even before they examine me, each one spends several minutes railing about something called Epic. That sort of thing tends to pique a journalist's curiosity.

Epic, known formally as Epic Systems, turns out to be a privately held company that provides the electronic database for patient records and billing used today in 39 percent of all U.S. hospitals. But that understates Epic's market dominance. The company has become the medical record database vendor for the vast majority of academic medical centers and other large hospital systems. Epic's software keeps track of about 260 million individual patient records, just under 80 percent of the entire U.S. population. Its revenue was $4.9 billion in 2023, according to a company spokesperson.

Epic has only one large competitor, Cerner, which has had several stumbles in recent years. In 2022, Cerner was purchased by Oracle Corporation, but it keeps losing hospital clients to Epic. In the past year, Cerner lost at least nine hospital systems to Epic.

In principle, electronic health records, or EHRs, should be a blessing for patients and clinicians alike. EHRs track every patient's medical history, warn against drug interactions, offer prompts when one condition might signal the risk of another, handle scheduling, facilitate communication among different doctors engaged with a patient, flag epidemiological patterns, and a great deal more. Epic boasts that in 2023 its system prevented 66 million potential adverse drug interactions and 250,000 potential surgical errors.

But in practice, clinical needs are often sacrificed to profit maximization. The demands by Epic's system on physician time have increasingly taken over the practice of medicine. Astonishingly, studies show that entering required data into the Epic system consumes about two hours of doctor time for every one hour spent providing hands-on patient care. The result has been an epidemic of doctor burnout and early retirements.

To its critics, Epic epitomizes everything perverse about the commercialized mess that American medicine has become. "Epic's clients are not doctors. They are the CEOs and CFOs who write the checks to Epic," says Dr. Bill Stead, who created pioneering electronic health record systems at Duke and then at Vanderbilt University Medical Center, both eventually supplanted by Epic.

Epic became the dominant vendor of databases because it was better than anyone else at combining regulatory compliance with maximizing hospital income. Epic enables the hospital to maximize the use of codes that determine the payment. "Before Epic, nobody was able to systematize upcoding," says an executive of one hospital system.

Epic's software can enable doctors and hospitals to overcharge patients, insurers, and Medicare and Medicaid.

There are about 10,000 possible billing codes that indicate conditions and complications. These Hierarchical Condition Category (HCC) codes allow increased payments based on risk. For instance, diabetes with no complications, HCC code 19, pays a capitation rate of $894.40, while diabetes combined with kidney failure can use 2 HCC codes, 18 and 136, which increases the capitation rate to $1,273.60. Doctors have the ability to also use codes for past patient conditions that have nothing to do with current presenting symptoms.

Assigning codes to each patient health malady is more of an art or artifice than a science, where doctors' judgment calls can bleed into Medicare fraud. Payments are based partly on time spent with a patient, on a scale from 1 to 5. One doctor told me that her supervisor, who gets reports from Epic on her billing practices, regularly contacts her and says things like "That appointment was a 2. Don't you think it might be a 3?"

Epic's software can thus enable doctors and hospitals to overcharge patients, insurers, and government agencies such as Medicare and Medicaid. Before a doctor can complete the record of a patient visit, they must respond to every question and check every required box. Hospitals have financial incentives to make patients look sicker so that they can maximize revenue, and Epic's software facilitates this.

Dr. David Bor, a senior pulmonologist who practices at the Cambridge (MA) Health Alliance and is a professor at Harvard Medical School, told me, "For billing and insurance purposes, the system requires a huge amount of material that we have to add into the note that is clinically irrelevant. It's a giant time sink." The irony is that the quality of care, one of the biggest original selling points of commercial database systems like Epic, is debased rather than enhanced. This detracts from patient care, while it serves as a money machine for hospitals that today are mostly for-profit, in fact if not in law.

An Epic spokesperson challenges this picture. "We take the integrity of health care billing and coding really seriously," he told me. "We have the tools to help hospitals code accurately, so that they can be fairly reimbursed, based on what the doctor or the nurse has documented in the chart. As a company, we are committed to supporting our healthcare providers in delivering care effectively and ethically without any incentive or support for practices like upcoding." In other words, if a hospital decides to use Epic's software to engage in upcoding, Epic's hands are clean.

, who began the company in 1979 under the name Human Services Computing. She and her family own about 47 percent of all Epic stock, and at 81 she is still the company's CEO. In 2021, estimated her net worth at $6 billion , and it is probably double that today. Faulkner has a master's degree in computer science and no clinical degree, though early in her career she apprenticed to one of the most widely admired designers of a hospital-based electronic records system, Dr. Warner Slack, who created the system for Boston's Brigham and Women's Hospital based on regular input from doctors and nurses. But even Brigham and Women's well-loved system was eventually supplanted by Epic.

The company is based in the town of Verona, Wisconsin, where Epic has almost as many employees as the town has residents and a vibe that many critics describe as cultish. For instance, buildings on the Epic campus have fantasy themes drawn from Tolkien, Harry Potter, Roald Dahl, and other such writers. According to plans submitted to the Verona city Plan Commission, Epic's sixth campus on its 1,700-acre site will be called "Other Worlds," and has been described in the local press as having a "'floating forest' and plenty of references to elves." The fantasy theme is Faulkner's own, reflecting her long-standing interest in science fiction. It evokes a series of theme parks, as in the movie and TV series . Of Epic's 14,000 worldwide employees, about 12,500 work in Verona. Despite the fact that information technology is well suited to remote work, Epic executives encourage Epic employees to work at its home campus, so that they can bond as a team and absorb Epic values.

The development of electronic recordkeeping began with the best of intentions. Late in the George W. Bush administration, both parties collaborated on legislation called the HITECH Act , which stands for Health Information Technology for Economic and Clinical Health. The legislation was designed to provide financial incentives for hospitals and other health systems to adopt electronic recordkeeping systems, and to promote direct patient use of them. At the time, only about 10 percent of hospitals had any kind of electronic systems. The rest relied entirely on paper records.

All rich countries have electronic medical databases. But no other has engendered the doctor backlash we see in America.

After the financial collapse of 2009, Congress passed and Barack Obama signed the near-trillion-dollar American Recovery and Reinvestment Act. One goal was to get money out the door as fast as possible. The HITECH Act was folded into the Recovery Act, with added incentives to get electronic systems up and running. Some $27 billion in grants and bonus payments were available to hospitals via Medicare and Medicaid. This was ready-made for commercial vendors, since only a handful of large hospitals such as Boston's Beth Israel had the capacity to design their own systems. At the time, there were more than a hundred fledgling for-profit vendors of electronic health record systems.

There was a vogue in the Obama administration for combining electronic recordkeeping with pay-for-performance incentives, to improve both quality and efficiency. So a whole other set of metrics was layered on top of the basic electronic recordkeeping. This was also ready-made for one monopoly vendor to emerge, though few appreciated that at the time.

At the center of this grand strategy was Dr. David Blumenthal, a physician and former senior vice president at Boston's Partners HealthCare (now Mass General Brigham) in charge of information technology. Obama appointed Blumenthal to the newly created position of national coordinator for health information technology. It was a dream job and Blumenthal became a missionary for the idea of combining electronic records with quality incentives—with tens of billions of dollars available for rapid implementation.

Blumenthal and his colleagues added a system of metrics to optimize quality and efficiency, known as "meaningful use." It had about 140 data elements. Dr. Brent James served as chief quality officer at Intermountain Health, the largest nonprofit health system in the Mountain West, with 34 hospitals and 385 clinics. Dr. James told me, "I had a profound disagreement with the Obama health IT administration. They were bringing complex regulatory standards too soon, effectively locking in relatively immature IT technology ... The regulations raised compliance barriers that locked out smaller competitors. Epic wouldn't be where it is today if it weren't for the federal government."

As Blumenthal came to appreciate, the real problem was that electronic health records became part of a health system whose drivers were increasingly commercial. And the administration was content to let the market decide how hospitals would use the recordkeeping software. "Our system is a creature of the market economy," Blumenthal says today. "One thing few people understood at the time was how incentives would shape the product. The customers were the chief information officers and the chief executives of hospitals, not doctors. Their principal goal was to protect revenues. Systems like Epic were not designed to improve quality because there was no financial incentive to do so at the time."

"We digitized the health records over the past 20 years," another veteran of those efforts recalls, "but lost the hearts and minds of clinicians along the way."

An Epic spokesman also challenges this picture, telling me that from its earliest beginnings, Epic grew by digitizing clinical records with the goal of improving information accuracy and quality, and that its utility in billing came only later.

, which are run by insurance companies, are notorious abusers of upcoding, even as they sometimes deny necessary treatments. The government's Medicare Payment Advisory Commission's annual report to Congress projects about $83 billion in excess billing by Medicare Advantage insurers compared to conventional public Medicare in 2024 alone. Many MA insurers, like UnitedHealth, employ large numbers of physicians, and some have been found encouraging their doctors to upcode .

But this doesn't stop with privatized Medicare; one extensive research paper titled " Upcoding Medicare: Is Healthcare Fraud and Abuse Increasing? " documented a steady increase in the use of diagnostic codes that required higher payments by traditional Medicare.

For example, Diagnosis Related Groups (DRGs) were a widely touted reform in which Medicare pays hospitals a flat rate per patient stay, to discourage hospitals from profiting by keeping inpatients for prolonged periods beyond what is medically necessary. But even DRGs can be gamed, since they pay more based on the supposed intensity of the condition.

An extensive 2021 report by the Department of Health and Human Services' Office of Inspector General found that from 2014 through 2019, the number of hospital stays billed at the highest severity level increased almost 20 percent, and accounted for "nearly half of all Medicare spending on inpatient hospital stays." The OIG added that the billing patterns were indicative of upcoding, since almost 30 percent of the inpatient stays at the hospitals lasted a particularly short time, and "the number of stays billed at each of the other severity levels decreased."

The inspector general can and does prosecute Medicare fraud in extreme and flagrant cases. But neither HHS nor other insurers have anything close to the resources to go after the nation's more than 6,000 hospitals for what has become routine upcoding. One senior doctor recalls asking his hospital CEO why they spent over a billion dollars a year to convert to Epic. He was told that the hospital would make that back many times over.

All rich countries have electronic medical databases. But no other has engendered the doctor backlash we see in America. That's mainly because most foreign systems, even multi-payer systems, use fixed prices, and offer far fewer opportunities for gaming the system.

When I got a full workup at a French hospital in 2022 for a possible cardiac event that turned out to be nothing serious, I got a bill for 875.57 euros based on a schedule of fixed prices. Had I been French, the French social security system would have paid the entire cost. Profit maximization was not part of the equation. Electronic health records exist in France and most other Western countries, but their purpose is not profit maximization.

By contrast, several doctors across the U.S. have told me that Epic not only includes clinically trivial data entries, but long hours of time-consuming training sessions. The system is extremely arduous for clinicians. Just entering a prescription via Epic, according to one study , takes a doctor 18 separate keystrokes.

Dr. Bor of Cambridge Health Alliance told me, "Because you can use a standardized template, doctors tend to save time by using cut-and-paste from yesterday's notes, and then updating. That introduces mistakes. It promotes lying. I ran TB clinics. TB is not typically associated with pain. But the system would not let me get out of the chart without adding a pain score. So I would add zero."

Doctors don't begin to have the time to keep up with Epic demands during the usual clinical day, and find themselves catching up during their off-hours. When I contact my doctor electronically to ask a medical question, she typically responds around 2 a.m. Other doctors say they find themselves working long into the evening at home to keep up with Epic documentation demands.

Dr. Gordon Schiff of Boston's Brigham and Women's Hospital, who has been a leader in that hospital's quality program, quotes a colleague: "I am embarrassed to have others read my notes. They are so bad, but I have to keep cutting more and more corners to get them done."

Most of the dozens of people I interviewed were scathing in their description of the demands Epic puts on doctors' time, but would speak to me only on background. Hospitals have invested massively in Epic, and doctors have been ordered to get with the program. Clinicians have plenty of suggestions on how to make Epic more physician-friendly. But doctors who have made these suggestions tell me that, for the most part, these ideas fall on deaf ears.

An Epic spokesperson disputes this account, offering examples of how Epic systematically solicits physician feedback and pointing to Epic's flexibility in allowing individual hospitals to customize how the Epic database is used: "We design the software so that the customer can configure it. It's not an off-the-shelf system." If hospital management is not taking doctor complaints seriously, that's the fault of the hospital, not Epic.

Even the parts of the Epic system that seem like time-savers can come at the expense of patient care, depending on how they are used. The latest innovation is the use of speech recognition software as enhanced by AI, to spare physicians the need to write notes during a patient visit. Epic uses two leading companies that pioneered this technology: Dragon, now rebranded as Nuance, which was bought by Microsoft, and Abridge, a startup created in 2020 at University of Pittsburgh Medical Center. The technology is known as "ambient listening."

During a patient exam, the doctor places a cellphone or other monitoring device nearby. The AI application then turns the conversation with the patient into something that looks just like a process note, including the patient's clinical history, medications, current condition, and the doctor's plan. But according to Dr. Schiff and others, the AI decides what to include from the conversation and what to leave out.

Some doctors love this and some hate it. It's a big time-saver, yet it also takes something away. As Dr. Schiff explains, "Writing the note yourself is a way of reflecting about the patient, thinking out loud about the assessment. It's the most important part of the note. But the opportunity to reflect is killed by automating the note. It's powerful and dangerous and makes errors. We are unfortunately falling for this because of the time squeeze."

Ironically, the same time squeeze that is substituting AI for the time-honored practice of writing notes is caused by all of Epic's other demands on the doctor's time. "Ambient listening is a solution to a self-created problem of requiring too much data entry by clinicians," one hospital executive told me. As Epic took over, according to another doctor I interviewed, something vital tended to get lost: cultures of local learning. Reliance on time-consuming automated communication crowds out hands-on communication between doctors and patients and among doctors.

Epic's rejoinder is that doctors just need to get more proficient and more comfortable using the many tools that Epic offers. For instance, the ambient listening app allows the doctor to review and revise the process note created by AI. But to Epic's many critics, all this looks like Epic increasingly taking over the practice of medicine.

, the Epic software's clinical prompts are sometimes rudimentary and irrelevant. Quality metrics have been a huge priority for the Department of Health and Human Services, and hospitals are financially rewarded for improved outcomes. But Epic turns this goal into an automated system of alerts that may have little to do with a patient's actual condition and risks. That in turn causes busy doctors to routinely override the alerts. The problem is known as alert fatigue.

For example, HHS concluded that hospitals were insufficiently mindful of the risks in hospitalized patients of sepsis, an infection that invades the entire body. Various studies estimate that sepsis is present in 30 to 50 percent of hospitalizations that culminate in death. So HHS created a system of financial rewards and punishments based on the hospital's performance. In turn, Epic created a set of automated computer alerts of supposed sepsis risks, based on correlations in its massive database. But expert analysis shows it paints with far too broad a brush.

One extensive study of Epic's sepsis alert protocol, after reviewing 27,697 patients, found a vast number of false positives and false negatives. The study, published in , concluded that "the Epic Sepsis Model poorly predicts sepsis; its widespread adoption despite poor performance raises fundamental concerns about sepsis management on a national level," and that the model creates "a large burden of alert fatigue," in which overwhelmed doctors just ignore alerts entirely. (An Epic spokesperson takes issue with this criticism, sending me a PowerPoint quoting testimonials from six hospital systems which reported that use of the Epic sepsis model resulted in reduced incidence of sepsis.)

Another general study of alert fatigue found that close to 100 percent of alerts were routinely overridden by harassed doctors. In this respect, Epic has become like the boy who cried wolf. By excessively broadcasting poorly designed alerts, Epic cultivates doctor skepticism of the system's clinical veracity and reliability.

Epic has many of the hallmarks of a monopoly. It wields immense market power.

Dr. Lara Goitein, a practicing pulmonologist, directs a quality improvement program at Dartmouth Health designed to empower frontline physicians and nurses to make improvements in systems. She and a colleague were interested in tracking and reducing the overuse of sedatives in the ICU that resulted in more patients staying for prolonged periods on mechanical ventilators, which added risks of longer ICU stays and increased infections.

"The EHR [electronic health record system] allowed us to do some simple things, like identify all the patients on mechanical ventilators and find the amount of time they were in the ICU," Goitein said. "But we needed another level of detail for a quality improvement project—for example, the types of sedatives, the amounts given, and patient sedation levels, especially during efforts to remove the ventilator." For that, the electronic system proved useless, even though the hospital had invested heavily in specialized consultants to extract the data.

"After many months of trying and hundreds of expensive person-hours, we finally decided to cut our losses," she said. "I hired a nurse to do a chart audit—that is, to read through a bunch of charts and enter data into an Excel spreadsheet. It took less than a week. We made some changes and ended up decreasing the average length of mechanical ventilation by a full day. With this and other changes, our rate of ventilator-associated pneumonia fell by two-thirds."

The EHR in that study was Cerner, but Dr. Goitein finds similar frustrations trying to use Epic for systematic quality improvements. Epic's rejoinder is that if doctors just took advantage of everything Epic offers, they could extract all the data they needed.

Dr. Goitein is preparing a paper for publication titled "The EHR Ate My Quality System."

less of an issue in a database designed primarily for clinical purposes. The best locally designed electronic medical record systems that predated Epic, built by teams of information technology specialists and medical leaders at some of the nation's best hospitals, such as Beth Israel and Brigham and Women's in Boston and Vanderbilt in Tennessee, did exactly that early in this century. But by about 2015, they all found themselves overwhelmed by the increasingly complex pay-for-performance metrics designed by HHS, combined with the financial imperatives of an increasingly commercial system that corrupted even nonprofits. "The process of internally keeping up with the regulatory changes and technical changes became impractical," says Vanderbilt's Dr. Stead.

Epic is simply better than anyone else at meeting the ever-changing demands of HHS aimed at grafting quality and efficiency incentives onto a for-profit system. Epic has also been a bonanza for consultants. Every time Epic adds a new complication or optional feature, tech consultants spring up to advise hospitals how to use it, further adding to outsourcing and middleman costs.

With great reluctance, the largest homegrown systems gave up and shifted to Epic, shouldering transition costs of at least a billion dollars , plus costly annual licensing fees. "They had not only economies of scale," says a former senior official at HHS, "but economies of learning."

Just this past June, Boston's Beth Israel finally completed converting its excellent homegrown and doctor-friendly electronic records system to Epic. The hospital's convenient patient access portal, called PatientSite, was replaced by Epic's clunkier universal system called MyChart. Beth Israel, now merged with the Lahey Hospital and Medical Center, was one of the last holdouts. The transition threw the hospital into chaos. I happened to have a routine medical visit during the first week of the changeover, and an appointment that should have taken 20 minutes took close to two hours because of new Epic demands on clinicians.

I sent a message to the hospital's manager of media relations, Kristina Murray, explaining my research for this and requesting an interview with a senior executive. In response to her questions about what I'd be asking, I explained that I was interested in why the hospital had switched to Epic, what it had cost, and how management proposed to address the massive doctor backlash and reduce the toll on clinicians. After a few days of silence, she replied, "I am unable to help you with this inquiry. Best of luck with the story."

of electronic health records has been what data wonks call interoperability: the ability of a doctor at one hospital to access a patient's records at another hospital, and the ability of the system as a whole to produce epidemiological and public-health data. With a single near-monopoly vendor of databases, interoperability should be close to automatic. But some hospitals have been proprietary about their patient records. This is known as data hoarding or information blocking, and HHS has issued regulations making it illegal. Lately, interoperability has become another Epic selling point, as long as you play by Epic's rules. Epic, with far more patient records than any other system, now helps doctors in hospitals that use its system to exchange patient data with non-Epic hospitals, and also works on data sharing with federal agencies. This is a public service that helps entrench Epic's market dominance.

There's another twist that reflects Epic's market power. To qualify as a preferred vendor within the Epic system, your company can join what Epic initially branded as its Partners and Pals program. This is now called Showroom. For a fee to Epic, vendors of specialized subsystems gain privileged access to Epic's hospital customers.

A set of for-profit vendors have developed systems that tap into the Epic database but create a separate clinician interface that is more user-friendly for doctors. One such vendor, called SaVia , allows a hospital to create and modify clinical support apps. The problem is that such vendors add yet another layer of expense. Epic continues to make refinements to keep hospitals tethered to its system, further adding to the complexity, and effectively increasing the demands that doctors modify their practice patterns by getting with the program.

Epic's massive patient database is also a treasure trove for research. Because access to the data is far from intuitive or automatic, researchers have the choice of partnering with Epic or paying Epic charges for gaining access to the desired data in usable form.

One of Epic's latest innovations is called Cosmos , a database for research purposes of more than 100 million patient records purged of individual identifiers. According to Epic's pitch on its website, "Cosmos integrates both inpatient and outpatient charts into a single, comprehensive record—including as patients move between health systems. In addition to diagnoses and medications, it includes patient-generated data and specialty-specific data." This will give Epic another huge profit center to sell such data to researchers.

of a monopoly. It wields immense market power. Under the Sherman Antitrust Act, being a monopoly is not a "per se" antitrust violation if you got to be a monopoly by building a better mousetrap. That surely describes Epic. It got better than anyone else at using data to allow hospitals to maximize their income. As payment systems and reporting requirements became ever more complex, most large hospital systems had nowhere else to turn.

But the other test under the Sherman Act is whether a monopoly abuses its market power. How might Epic be abusing its power? Because hospitals have few other options, it could be overcharging. It's very expensive to shift to Epic, and it would be expensive to dump Epic in favor of a competitor or a local system. That lock-in also increases Epic's pricing power.

Epic's profit margins are not a matter of public record because it is a privately held company, so it would take an investigation to reveal whether they reflect monopoly pricing power. By enabling hospitals to overbill, Epic's systems might also be contributing to the increasing costs of the entire health system.

Epic's program for vendor partners also raises antitrust questions about regarding Epic's role as a monopsony. To the extent that vendors pay Epic to gain privileged access to its hospitals as potential customers, does that suppress innovation and competition among other vendors in the sector? Does it raise prices? Is it analogous to Google giving preference to some sellers?

The Federal Trade Commission handles antitrust cases in the area of health care. I queried FTC media relations on whether, hypothetically, abuses by a near-monopoly provider of health data systems might warrant an antitrust investigation. I was told, "We have no comment."

In many respects, a national health database truly is a natural monopoly. But to be efficient, natural monopolies must be either regulated as a public utility or placed in the public domain. A private, unregulated for-profit monopoly is the worst of all possible worlds. It can be opaque, arrogant, nonresponsive to complaints, and still make a fortune.

As Dr. Stead puts it, "The frustrations with Epic can't be fixed by Epic alone, because each fix requires a coordinated change in health system policy, clinical workflow, and software logic. They need to be fixed locally, but most hospitals got rid of their own clinical informatics experts when they put in Epic."

Ideally, our national health database would be in the public domain and treated like a universal social asset. Local users would be free to customize it. The fact that Epic's logic is hyper-commercial is a reflection of the fact that our entire health system is hyper-commercial.

"There's a need for broader reform," says Dr. Adam Gaffney, a critical care physician at the Cambridge Health Alliance who was previous president of Physicians for a National Health Program. "Handing the keys over to a private monopoly doesn't make sense. You need to think of a national medical record system as a public service."

But that fundamental reform is likely to happen only as part of a shift to a national health insurance system or perhaps in the aftermath of a major antitrust action against Epic. In the meantime, the epidemic of doctor burnout and early retirements caused by digital demands might compel some reforms. The American Medical Association projects a shortage of between 37,000 and 100,000 doctors over the next decade. For now, hospital CEOs are big fans of Epic's revenue enhancement. That will change only when lots more exhausted doctors quit.

POSTSCRIPT: After this went to press for our October print issue, P Health, a fledgling data platform company, filed suit in the Southern District of New York on September 23, alleging that Epic violates the Sherman Antitrust Act by using its market power to block P's access to Epic's Carequality network for data sharing. Epic has responded by raising security concerns and accusing P of potentially violating the HIPAA law, which protects patient privacy.

A "payer platform" of the sort offered by P allows health plans to retrieve and store medical records from master electronic health record systems such as Epic. From Epic's perspective, such retrieval is one key part of what Epic itself offers via its own Epic Payer Platform (EPP).

From P's perspective, this is a legitimate area for vendor competition. But because of Epic's practices, according to the allegations of the lawsuit, "no competitors emerged to challenge EPP during the first few years of its existence. That was because of a calculated move by Epic to stifle competition. No competitors could challenge EPP because Epic made it commercially impossible for any payer platform other than EPP to access records stored in Epic's EHR software."

Whatever the facts and merits of this particular lawsuit, which will be judged in court, it could serve as a template for other potential legal challenges to Epic's immense market power.