This is my fourth in a series of imaginary lectures on remedial health policy for President Obama. My goal is to convince Obama that he relied on the wrong people for health policy advice. I am focusing on three people in particular: Elliott Fisher and his colleagues at the Dartmouth Institute, Peter Orszag, and Atul Gawande.
In my first comment , I criticized Obama for clinging to the belief that the Affordable Care Act has already reduced health care inflation and will continue to do so in the future. I devoted my second comment to explaining how influential the Dartmouth Institute has been. In my last comment I set forth the reasons why the Dartmouth group’s influence has declined since approximately 2010.
I devote this comment to a review of some of the evidence that indicates underuse (the failure of the health care system to deliver necessary care) is more prevalent than overuse. Knowing that fact is useful not just for understanding my criticism of Obama but for understanding how flimsy the justification is for accountable care organizations and other managed care nostrums. I have only enough space here to introduce you to the best of the under- and overuse literature. I believe it will be enough to convince you that underuse is more common than overuse. Once you comprehend that fact, you’ll also comprehend that it is neither logical or ethical to base health policy on the assumption that overuse is the only form of inappropriate use we must address. If we view underuse as an equally serious problem, then it makes no sense to promulgate managed care notions (such as shifting insurance risk to doctors) designed to address overuse.
Policy makers need to focus on inappropriate use – over- and underuse – and develop targeted solutions to both forms of inappropriate use. Shifting financial risk to doctors and micromanaging them is not a targeted solution. It’s a meat axe wielded by ideologues who see overuse everywhere and underuse nowhere.
How extensive is the research on over- and underuse?
Given how long and loudly managed care advocates have complained about overuse and the role that FFS allegedly played in it, you might think a large body of literature accumulated long ago demonstrating that specific services are overused, FFS is to blame, and underuse is trivial or non-existent. You would be wrong on all counts. The number of papers documenting overuse is small, studies comparing over- to underuse are extremely rare, the few studies that do compare over- to underuse find underuse greatly exceeds overuse, and no paper has ever been published in a peer-reviewed journal demonstrating that the FFS system causes the overuse that does exist.
After a Herculean search of 114,831 papers published over the period 1978-2009, Korenstein et al. reported finding only 172 studies documenting overuse. Moreover, “the majority of [those] studies focused on four interventions: antibiotics for URI [upper respiratory infection] and three cardiovascular procedures [carotid endarterectomy, coronary angiography, and bypass surgery].” Korenstein et al. concluded, “The robust evidence about overuse in the United States is limited to a few services.”
Because no comparable study of the underuse literature has been done, we can’t say with confidence that high-quality research on underuse is more voluminous than the overuse literature reported by Korenstein et al. My impression is that the underuse literature is larger but is simply ignored by the health policy elite and the media. In a 1997 paper on the underuse of angiography and revascularization (“revascularization” refers both to angioplasty and bypass surgery), Kravitz and Laouri observed, “Most health services research to date has been directed at identifying and reducing excessive utilization. Little attention has been given to underuse of care.” I’m not sure the first sentence is true; the actual number of good papers on underuse may greatly exceed those on overuse. What I’m sure of is that the health policy elite and the media just don’t care enough about the underuse research to talk about it anywhere near as often as they celebrate the limited overuse literature.
Underuse is more common than overuse
Papers that attempt to measure the incidence of both under- and overuse are rare, and papers that attempt to do so for multiple services are extremely rare. I am aware of only two that fit that last category. Both of those found that underuse occurred far more frequently than overuse.
The most comprehensive of those two studies is a paper published in the New England Journal of Medicine in 2003 by Elizabeth McGlynn et al. That study must have been very expensive. I doubt we’ll see another one as comprehensive for a long time.
McGlynn et al. interviewed 6,600 adults across 12 cities and examined their medical records to determine which medical services (related to 30 conditions and several preventive services) they received and whether those services were necessary or unnecessary. The authors found that underuse occurred four times as often as overuse. “[W]e found greater problems with underuse (46.3 percent of participants did not receive recommended care …) than with overuse (11.3 percent of participants received care that was not recommended …),” they reported.
The other very good study of both under- and overuse of multiple services appeared in the Journal of General Internal Medicine in 2005. Its authors, Rodney Hayward et al., studied the records of 621 patients seen in 12 VA systems in two states. They found an under-to-overuse ratio of 22 to one, far worse than the four-to-one ratio McGlynn et al. reported. “Of errors that could be classified ….,” they concluded, “95.7% … were classified as underuse and 4.3% … as overuse/misuse.” 
Underuse of cardiac and other services
As I noted above, Korenstein et al. found that the limited research on overuse tends to focus on antibiotics and cardiovascular interventions. The underuse literature includes several papers on coronary angiography, angioplasty, and coronary artery bypass graft (CABG) surgery. Thus, it is possible to set the studies of cardiovascular-service underuse next to those that studied overuse and compare rates.
Korenstein et al. reported that “overuse rates of CABG were generally lower than 15 percent, and overuse rates for CA [coronary angiography] were generally lower than 20 percent.” 
The underuse research indicates that underuse rates of cardiac services are much higher than the overuse rates (under-15 and 20 percent) reported by Korenstein et al. Here are underuse rates for angiography reported by six papers:
- 41% (Kravitz and Laouri 1997 );
- 44% (Laouri et al. 1997);
- 22% (Carlisle et al. 1999);
- 54% Medicare FFS, 65% Medicare Advantage (Guadagnoli et al. 2000);
- 42% (Garg et al. 2002);
- 49% Medicare FFS, 56% VA (Petersen et al. 2003).
With one exception (Carlisle et al.) the six studies listed above report underuse rates for angiography above 40 percent, which is far higher than the under-20-percent overuse rates Korenstein et al. found. The one exception (the Carlisle paper) was also the only one of the six that also examined overuse. That paper reported a mere 4 percent overuse rate – almost one-sixth of the reported 22-percent underuse rate. If the true overuse rate of angiography is around 10 percent and the true underuse rate is 40 percent or higher, then the ratio of under- to overuse of this valuable diagnostic test is at least 4 to 1.
Similarly, the few studies that have estimated underuse of revascularization find rates much higher than the under-15-percent rate for CABG overuse reported by Korenstein et al. Here are revascularization underuse rates from four studies:
- 23% (Kravitz and Laouri 1997);
- 25% (Laouri et al. 1997);
- 26% (Leape et al. 1999);
- 25% men, 22-24% women (Epstein et al. 2003).
The last study listed above, Epstein et al., also reported overuse rates for men and women. In men overuse was 14 percent (far below the 25 percent underuse rate Epstein et al. found) and in women 5 percent (way below the underuse rate of 22-24 percent).
Studies of medical care for non-cardiovascular conditions also reveal very high rates of underuse. Asch et al. found that “underuse of necessary care is widespread for the 15 … conditions [studied] … in the relatively well-insured Medicare population. For almost half of the indicators, less than two thirds of beneficiaries received needed care.” For example, 59 percent of seniors who were told they should have their gall bladders removed did not have it done.
Underuse is underestimated
As bad as the underuse rates I just reported are, they are underestimated. That’s because estimating true underuse rates requires that we know who in a given population needed a service and didn’t get it. All the research I discussed above studied patients who made contact with the medical system.  But many people who need a service don’t get it because they never visited a clinic or hospital, and therefore left no record of underuse. Note that the overuse research does not suffer from this limitation. By definition you can’t get a medical service, including an unnecessary one, without seeking medical care. There is, in other words, no large quantity of undetected overuse lurking out there as there is with underuse.
Determining true underuse rates requires field work – surveys and interviews and diagnostic tests – outside of the usual clinical or hospital setting to determine the prevalence of a condition or need. That research is expensive and there is, therefore, not a lot of it. What research of this type we have indicates unmeasured underuse is enormous. The CDC, for example, reports that 67 million adults in the US have high blood pressure but 30 percent of them are not being treated for it (14.1 million don’t know they have high blood pressure, and another 5.7 million know but aren’t being treated). Research on other chronic conditions, such as tooth decay, depression, and diabetes, reveals similar results.
To sum up what I’ve presented so far: The research indicates underuse occurs at epidemic levels, it occurs at very high rates even among the insured, it is common even for expensive medical services (where we would expect the FFS incentive would reduce underuse to zero if in fact that incentive were as powerful as managed care advocates make it out to be), and occurs more often than overuse.
Causes of underuse
Patient aversion to visiting doctors and hospitals, and doctors and hospitals not having enough time or resources, are the main causes of underuse.  I’ll discuss one study documenting each of these factors.
One of the best articles demonstrating how much Americans dislike seeking medical care was a paper generated by the well known RAND Health Insurance Experiment. The paper , by Shapiro et al., reported that 80 percent of participants refused to see a doctor when they suffered “serious symptoms” such as unexplained bleeding, shortness of breath from climbing stairs, and unexplained loss of consciousness. Insurance status had no influence on the 80 percent rate; even among participants who had “free care” (their medical care would have been totally paid for by insurance), the rate was 80 percent. The vast, vast majority of Americans don’t have 100 percent coverage. The unaffordability of medical care is, in addition to widespread aversion to medical treatment, another significant reason why many people avoid doctors and, in the case of patients who did see doctors, refuse to follow physician recommendations. 
Turning now to the second major cause of underuse: Kimberly Yarnall and her colleagues at Duke are among the handful of researchers who have sought to determine whether doctors have enough time in the day to treat every issue presented to them according to existing guidelines. In a paper published in 2009, Yarnall et al. determined that a primary care doctor would need 21.7 hours per day to provide “care for prevention, chronic care, and acute care to an average patient panel.” Managed care proponents can cram doctors into ACOs and medical homes to their hearts’ content, but if doctors just don’t have enough time and resources, underuse will continue.
Implications of rampant underuse
John Wennberg, Elliot Fisher and their colleagues at the Dartmouth Institute have been telling policy makers and reporters for almost a half century now that regional variation is due to overuse. It is never due to underuse. They have stated repeatedly that if the regions of the country where Medicare spending is the highest could be forced to reduce spending to the levels of the lowest-spending regions, US health care spending could be cut by 30 percent. That claim implies (a) that all regional variation can be blamed on overuse in the allegedly high-cost regions and, conversely, that little or no underuse is occurring in the low-spending regions and (b) that underuse is not a problem in the allegedly high-cost regions and, therefore, would not be aggravated by meat-axe approaches to reducing overuse.
Here is a particularly astonishing example of the claim that overuse is the problem and underuse can be ignored. In a 2003 editorial in the New England Journal of Medicine, Fisher wrote: “If all regions could safely adopt the practice patterns of the conservative regions, Medicare spending (and perhaps health care spending overall) would fall by about 30 percent. But achieving such savings in the short term is most likely a practical and political impossibility (a third of the health care workforce would have to find new lines of work).” (p. 1666) You see? According to Fisher, overuse is so bad and underuse so trivial that one-third of the health care workforce can be given pink slips if we just crack down on overuse.
I have a very different assessment. Because underuse is so common, we will need to maintain or expand the medical work force, not chop it down by a third, if we intend to address both over- and underuse (not just overuse). The funds to pay for additional workers, if they are needed, and many of the additional workers themselves, should come out of the bloated administrative apparatus that has grown up around doctors and patients over the last half-century thanks to the ascendance of managed care ideology and its obsession with overuse.
In the final installments of this series I will examine the writings and statements of Peter Orszag and Atul Gawande, two men who influenced Obama who were in turn deeply influenced by the lopsided attention to overuse Fisher et al. promoted.
 Hayward et al. used “overuse” and “error of commission” interchangeably and, similarly, “underuse” and “error of omission” interchangeably. The title of their paper signifies what they found: “Sins of omission: Getting too little medical care may be the greatest threat to patient safety.” Unlike Fisher et al.’s many papers on overuse, this paper got zero attention from the media.
 I suspect the average overuse rates Korenstein et al. reported for CABG and angiography (“lower than 15 percent … and 20 percent” respectively) are lower today than they were in the 1970s and 1980s when angiography and CABG were still relatively new. If, for example, Korenstein et al. had based their overuse estimates on research done since 1990 rather than 1978, their estimates would have been lower. For example, a 1993 paper by Bernstein et al. reported only a 4 percent overuse rate for angiography, and a 2011 paper by Chan et al. reported a mere 1 percent overuse rate for angioplasty “for acute indications” and a 12 percent rate for “non-acute indications,” for a weighted average rate of 5 percent.
 The McGlynn and Hayward studies underestimated underuse because they studied only patients who had contact with doctors and hospitals. McGlynn et al. excluded the 7 percent of their original sample that had “no visits to a health care provider during the previous two years.” Hayward et al. excluded any VA patient who hadn’t had at least two doctor visits in each of the preceding two years. Similarly, Asch et al. limited their denominator to Medicare FFS enrollees who had already received a diagnosis of one of the 15 conditions studied.
 Another memorable paper revealing widespread distaste for medical care was published in 2001 by Hawker et al. The authors reported enormous underuse of hip and knee replacements among adults over 54 years of age diagnosed with hip or knee arthritis. In one county, underuse was 85 percent, in another 91 percent. What makes these findings even more amazing is that the subjects were Canadians living in the province of Ontario, which meant their surgery would have been free. One other interesting finding: The willingness of patients to undergo surgery varied by almost two-fold between the counties. Fisher et al. would have us believe that differences in physician behavior, not differences in patient demand for services, explains regional variation.
 The impossibility of achieving 100-percent agreement on the appropriateness of every medical good and service given to every patient is also a factor in both over- and underuse. Researchers who examine medical records and claim forms after the fact will inevitably classify some services as inappropriate that doctors and patients might argue were appropriate.