Do Patients Receive About Half of Recommended Health Care?

The idea that patients receive about half of recommended care when they see physicians is the conclusion of a RAND Corporation study [gated, but with abstract], spearheaded by Elizabeth McGlynn and colleagues. In fact, not a single outcome of care (like death, infection, ability to walk, relief of pain, return to work) was analyzed. Instead, the study focused only on 439 inputs called “indicators.” These inputs range from the serious and uncontroversial (giving a heart exam to patients with chest pain) to the unobjectionable but possibly trivial (counseling smokers to stop smoking, alcohol abusers to reduce their drinking, and patients with sexually transmitted diseases to practice safe sex).

Four questions immediately jump to mind: (1) What does it mean to say that a procedure is “recommended care?” (2) How do we know that the recommended care is actually good for patients? (3) How do we know whether the patients actually got the care? and (4) Does it make sense to combine the serious with the trivial in making a judgment about the overall quality of the US health care system?

Oh, the movie never ends
It goes on and on and on and on

How do we know the care is recommended?  The study assembled four, 9-member “expert panels” of physicians based on geographic location, practice setting and gender.  These panelists then used their expertise to vote on how likely an “indicator” is to help a patient (on a scale of 1-9). The methodology here is similar to that used years ago in a RAND study that concluded that about one-third of medical care is unnecessary. In that study, the initial views of the 9 expert panelists were all over the map. They were then assembled and group pressure was exerted to get a consensus. Even then, 7 of 9 experts could agree only half the time.

How do we know the care is desirable? Here is the bottom line: These indicators were artificially created, often without any actual study to prove they help patients.  This process is devised specifically because — as RAND states — there are few valid scientific studies to guide much of medical care.  While practicing doctors recognize that this is where the “art” of medicine lies, RAND attempts to artificially create a standard of care that has never been tested and — through a modified democratic process — turn it into profession-wide “recommended care.”  Having personally participated in such expert panels, I have witnessed the personal biases of experts guiding decision-making.  Strong personalities can sway the discussion and often an entire group will defer to one person if they don’t have enough knowledge of a subject.  Further, many will remain silent on one topic in order to prevail on another. 

How solid was the evidence used in the study?  RAND surveyed 13,275 people by phone in 12 cities over the course of about 13 minutes each to obtain a “health history.”  They then gathered available medical charts on only 37% of those interviewed. This led the authors to admit that the low numbers mean the “results are likely to be biased.”  Twenty registered nurses then “abstracted” the paper records to a computer, spending only 50 minutes each to do so, on the average. The problem here is that many of the medical issues are complicated. Whether a procedure should or should not have been followed depends in some cases on very difficult medical judgment. Furthermore, repeating the method of a nurse chart review only produced the same answer 80% of the time — a 20% error rate.

Of the 439 indicators, only 93 were based on the “gold standard” — medical studies that are well-designed, randomized, blinded and controlled.  Such studies are called “level I” and they are rare and only apply to a specific group of patients.  For instance, such a study might apply to the use of anti-cholesterol medication in diabetic men who have had heart attacks (but not to all people with high cholesterol).    

An analysis of the 93 indicators based on “level I” studies by this author reveal that in 42 cases there were fewer than 25 patients for that indicator.  For seven of these “level I” indicators only ONE patient was available for analysis.  In fact, 55% of these studies had fewer than 50 for each indicator!  Of the small group of patients for each of the indicators, there is no information on whether there were good reasons the patients did not receive a medication, surgery or test, or just lack of information in the medical record. In fact, the latter seems likely since immunization rates for influenza in the elderly was 85% based on phone interviews, but only 15% based on the nurses’ medical chart abstractions!

In some cases, the indicators are so complex that it would be difficult to determine whether they were followed or not.  For instance, one indicator (that had only 4 patients analyzed) states: “Patients admitted with the diagnosis of unstable angina who have angina longer than 5 minutes at rest associated with ischemic ST segment changes who do not have contraindications to heparin should receive heparin within 2 hours of the initial ECG that demonstrates ischemic changes, and continuous heparin infusion or subcutaneous LMW heparin for at least 24 hours (or until 26 hours after the ECG with ischemic changes).”  I added the emphasis to indicate all the complex items a registered nurse must find in a chart and interpret correctly for the study. This is a common type of RAND indicator.  In this case, 4 patients were evaluated using methods described above to conclude they received “appropriate” care only 41% of the time. 

Are the inferences about overall health care quality reasonable?  The nature of the indicators leaves much to be desired.  For instance, among the types of “failures” was the lack of counseling recorded in the medical chart (whether it occurred or not) to tell alcoholics to stop binge drinking or those with genital herpes on how to reduce risk of transmission. Many of the positive findings from the study were not reported in the press. For instance, of the in-patients with depression and “level I” indicator groups with more than 250 patients, 85% to 90% received appropriate medications.  And 94% to 99% of 489 back pain patients received “proper” care recommendations. The RAND Corporation concluded that more than 70% of patients received recommended care for cataracts, breast cancer and prenatal care, even when indicators supported by low quality studies were included. Over all, for two-thirds of the indicators, patients received recommended care more than 60% of the time.

However, while doctors try to achieve near perfect compliance with indicators like these in order to have good public report cards or obtain financial rewards from third-party payers, they may do more harm than good.  Many recent studies have shown that even 90%-plus compliance with higher quality indicators doesn’t help patients, and actually produces unintended consequences like avoiding high-risk patients, providing the wrong care in some cases to achieve better ratings and gaming of the system. They may also cause physicians and nurses to spend more time away from patients to enter data in computers. The harmful effects also disproportionately impact low-income and minority populations.

How do we ensure patients receive high quality care? Americans deserve access to the highest quality medical care available.  However, the only way to know whether these 439 indicators improves care is to spend hundreds of millions of dollars to conduct 439 high quality scientific studies with a minimum of hundreds to thousands of patients each over decades. Since resources are limited, a better way must be found.  That way is in the doctor-patient relationship using qualified and experienced doctors subject to valid review by their peers. It will be up to individual patients to ask their doctor – Do I need this test, drug or surgery and how often do you have complications? 

Perhaps a better way to get the high quality care is to demand that doctors spend more time talking to and examining patients.  Such time with doctors will shrink under the cookbook medicine protocols created by Congress and the White House using the RAND study and other pseudoscience.  It is estimated that to comply with these indicators, doctors would need to spend 15-20 hours per day. In these circumstances, patients will be subject to a “mill” mentality and treated like numbers that must be entered in a computer to satisfy a functionary sitting in a cubicle somewhere in Washington D.C.  A better solution is to empower patients financially to pay doctors for their time.  I find that when I look my patient in the eyes and spend 45 minutes with them, they are getting my best care.

Dr. David McKalip is a private practice brain and spine surgeon in St. Petersburg, Florida.  He serves as the President of the Florida Neurosurgical Society and Chairman of the FMA Council on Medical Economics.

Print