So why should we change our current practices? Our patients like us; our outcomes are generally good; we keep passing our board exams; and they keep telling us we have the best health care system in the world. Besides, change is hard work. Aren't we busy enough learning about how to survive in a managed care environment?
Unfortunately, there is good evidence that the quality of care we give our patients could be better. Such evidence comes from:
- clinical examples, in which lack of good evidence has led to harm for our patients
- common patterns of thinking that introduce bias ("heuristics")
- the wide variation in current clinical practice among physicians
- the difficulty of managing medical information, when results conflict and thousands of articles are published every month
- our knowledge declines over time, as we get further from medical school, and unfortunately traditional CME doesn't work
In this section, we'll discuss each of these in turn.
Practice without the best evidence
Much of what we do is not supported by reliable external evidence, be it from randomized trials, systematic evaluation of a diagnostic test, or careful follow-up of large numbers of patients. Consider the following examples:
Back to sleep
Physicians in western countries traditionally recommended that babies sleep on their stomachs. It was thought that by sleeping on their backs infants were at risk for regurgitation and aspiration, leading to sudden infant death syndrome (SIDS). In other words, the conventional wisdom was that infants behaved much like drunken rock stars! In the 1980's, some physicians asked the question, "Is there any evidence to support the practice of sleeping babies on their stomachs?" As it turned out, case-control and ecologic studies found a dramatic decrease in SIDS deaths among children sleeping on their backs, leading to the national "Back to Sleep" program. Had someone asked the question 20 years earlier, tens of thousands of lives might have been saved.
Eyepatches for corneal abrasion
The standard practice of both family physicians and ophthalmologists has always been to patch the eyes of patients who present with a simple, uncomplicated corneal abrasion, as well as provide a mydriatic agent and antibiotic eyedrops. However, no one ever asked if patching was beneficial - it was just "common sense". There have been at least five randomized controlled trials of patch vs no patch, and each has come up with the same answer: patches offer no benefit, and may even slow healing and increase patient discomfort.
Anti-arrhythmics agents
In the late 1980's, encainide and flecainide were marketed as anti-arrhythmic agents based on their ability to suppress venticular arrhythmias. However, a large randomized controlled trial (CAST) showed that mortality was considerably higher among treated patients than among controls, not an ideal outcome! (Echt, 1991) This especially demonstrates the problem of relying on "intermediate" or "disease-oriented" outcomes such as arrhythmia suppression, rather than looking at more important measures such as mortality, morbidity, and quality of life.
Steroids in prematurity
In 1973, a small study demonstrated that steroids given to women expected to deliver prematurely reduced the likelihood of death in their infants. Six further studies in the next 10 years had mixed results, primarily because they were all quite small. Had a meta-analysis been done in 1983, it would have shown that the overall results of all the trials combined supported a beneficial effect of steroids. However, it took another decade and seven more studies before these results were accepted and began to change practice. Had a systematic review of the literature been performed in 1983, it might have changed practice much sooner and saved thousands of lives.
|
|
The failure of common sense
There are many, many more examples of "obvious" interventions which fail to help our patients live longer or better lives. How can common sense fail us so badly? As human beings, we are "wired" to respond to cues from the environment in certain, predictable ways. In particular, it has been very successful from an evolutionary perspective to look for causality in our environment:
I shook the tree, and a coconut fell on my head. Hmm....maybe shaking the tree caused the coconut to fall on my head! Better not shake trees without first watching for falling coconuts...
Consider a similar example from the medical realm:
I gave my patient who has had bronchitis for 4 or 5 days an antibiotic, and 3 days later she felt better. Hmm...maybe giving the patient the antibiotic caused her to feel better! I better give all of my patients with brochitis an antibiotic...
This "causality heuristic" is driven by our belief in ourselves as physicians, our belief in the value of pharmaceutical interventions (which have been so successful for other conditions), and our patients' desire to see their decision to visit the physician validated. A lawyer (or my high school Latin teacher) would say: post hoc ergo propter hoc, or "after this, therefore because of this".
There are many other heuristics or "rules of thumb" which guide, and sometimes mislead, clinical decision-making. They have been described by Sox, and his book "Medical Decision Making" is an excellent source for further insights. For example, another common flaw in decision-making is the "availability heuristic". Certainly, we've all had the experience of diagnosing an unusual condition, then looking for it a little harder than usual in the next few patients.
Physicians can also be led astray by ignoring the prior probability of disease. For example, the likelihood that chest pain is caused by coronary artery disease is vanishingly small in an otherwise healthy 20 year old, but may exceed 50% in an older patient with multiple risk factors. The youth may require only a careful history and reassurance, while for the older patient catheterization may be an appropriate initial study. While having a single diagnostic strategy for all patients with chest pain would be convenient, it would not be good practice.
"Regression to the mean" is often unrecognized and can lead to inappropriate diagnoses and interventions. Consider a patient with a slightly elevated liver function test on the first measurement:
Clearly, it is much more likely, just given random variation, that the second measurement will be lower than higher. There is much more of the curve to the left than to the right of the initial measurement! For another example, pretend the above numbers represent rookie batting averages. Any baseball fan can tell you about the "sophomore jinx": A player who has a great rookie season (hitting .320, for example) is unlikely to do as well or better their second year. That is regression to the mean at work.
Variation in current practice
As new tests and therapies are developed, how do physicians decide which to adopt? Without a clear, consistent framework, these decisions are typically driven by the practice patterns of local "opinion leaders", advertising, pharmaceutical representatives, specialists who may see a different spectrum of patients, and other potentially biased sources. The result is a huge variation in practice patterns among regions, states, and even cities in the same state.(Wennberg, 1991) Consider the following graph which shows the rate of radical prostatectomy per 100,000 male Medicare beneficiaries, adjusted for age and race (Lu-Yao, 1993):
The range of surgery rates is incredible:
from 20/100,000 in Rhode Island to 429/100,000 in Alaska, a ratio
of 21 to 1! Since patients and prostates aren't that different
between those states, we must be considering different external
evidence. The result is that patients in some states undergo
too many prostatectomies, many of which leave the patient impotent
or incontinent. Conversely, patients in some states may
not be receiving the procedure often enough. The 
truth
is out there, and so is the most appropriate level of care,
which probably lies somewhere between the extremes. Shouldn't
all patients receive the best care?
Managing medical information
Each month, thousands of medical journals publish tens of thousands of articles. Even if you only consider the 90 or so clinically oriented journals of most interest to primary care physicians, they publish over 15,000 articles per year. If you read 40 articles every day of the year, you would still fall woefully behind!
Clearly, physicians can't read everything. In addition, much of the literature is in apparent conflict with a sort of "ping-pong" game played as issues are debated in the pages of journals. Results are often presented which are premature, use inappropriate outcomes, or represent communications to other researchers, not clinicians. Most importantly, many published studies suffer from serious flaws which invalidate their results. Unless specially trained in critical appraisal, physicians may be misled by these invalid results. A systematic, rational strategy is needed to deal with this information overload.
Knowledge declines over time
Although medical knowledge declines as physicians get further from medical school (Sackett, 1997), their ability to practice the "art" of medicine improves. Physicians get to know their patients, hone their diagnostic skills, and are exposed to an ever increasing number of patients and problems. Wouldn't it be great if our medical knowledge also improved as we moved through our careers?
In the next section, we will discuss how an evidence-based approach to practice, teaching, and research addresses these limitations of medical practice.