Health Care Reform, Accountable Care Organizations, Practice Guidelines and the Corporate Practice of Medicine

A century ago states began to codify prohibition of the corporate practice of medicine. Some states have statutes, some rely on attorney general opinions, and some rely on case law. In many states, combinations of these instruments and opinions have shaped the legal environment. The prohibition of “corporate practice” stemmed primarily from the ethical concern that corporate profits might influence physician clinical decision making. In fact the impetus seems to have been more aimed at restricting the types of business arrangements in which physicians could engage than at putting a check on excess corporate profits.

Now the health care landscape is far more complex. However the concerns over how external financial pressure might influence the doctor patient relationship should not disappear. Even the most sophistocated consumer of health care services could not be expected to anticipate where these pressures lie and how they might be manifested.

In Illinois, for example, important court cases of the past decade have dealt with the issue of hospital employment of physicians (they can) (Berlin v. Sarah Bush Lincoln Heath Center). This decision extablished the legality of hospital employment of physicians but left standing other prohibitions of the corporate practice of medicine. Most decisions of this type accross the country include the caveat that the employment must not directly influence a physicians clinical decision making.

Another case dealt with how physicians may share fees with corporate managed care (they may not, at least not on a sliding scale basis) (Vine Street Clinic, et al. v. HealthLink Inc). Both of these cases reflect the more general societal unease over how physicians might have their clinical judgement influences by external forces.

Ironically after the passage of health care reform, the next agenda item for congress is prohibition of certain behaviors of bankers who are widely held responsible for the last financial melt down. But has congress shifted its attention too soon? While they scrutinize the banking industry and legislate new ethical standards will the new health law lead to different, yet still huge, financial manipulation in the health care industry? The banks have long been regulated. The new health law will ultimately lead to enormous proliferation of state and federal regulations. Patients are in danger of becoming pawns in a very complex game.

No politician dares to utter the "R" word in the health care debate. Yet an unavoidable component of "controlling costs" will involve rationing of services ala the British system. Lamenting the enormous sums spent in the final weeks of life is one thing. Identifying the exact day each patient begins that count down is another. Prevention only postpones the judgment day. So rationing will occur disguised by euphemisms like "comparative effectiveness" (a research initiative created and funded in the new Act).

We are now just entering that new era of rationing of health care with the passage of the 2010 Healthcare Reform Act in Washington. Literally within minutes of the act being signed into law by the President several states filed legal challenges. Like virtually all legislation from Washington, a subtitle to the bill could read “The Lawyer, Accountant and Bureaucrat Relief Act of 2010”. Likely it will take a least a decade of promulgation of rules and adjudication of administrative law and civil court cases to know what the legal ramifications of the act are.

Among the many issues addressed or mandated in the bill is a pilot to test a prospective payment system with a much more global scope than the current Medicare “DRG” system for hospital  acute care episodes. The Act calls for the formation of Accountable Care Organizations (ACO’s) that will receive prospective payment for the care of certain groups of patient. This will require a structure that puts physicians in the same ethical conundrum that led to the various prohibitions of corporate practice a century ago.

Directly or indirectly (but always by a codified scheme that will have to pass legal muster in each state) physicians will share in the profits or losses of the ACO. This will be true whether the ACO is structured as a for-profit or a not-for-profit. To avoid the unbearable ethical dilemma of rationing health care one decision and one patient at a time, practice guidelines will be needed. Some method of corporate enforcement by carrot and stick will also be needed so that the guidelines are applied fairly and uniformly.

When, exactly, does the institutional, corporate implementation of “guidelines” become the corporate practice of medicine? For example, a complicated immune compromised hospital patient had a change in clinical status with fever. Her physician was contacted and she ordered an (expensive) antibiotic to be given immediately (“stat”) according to the FDA approved dosing scheme. But the hospital had a corporate guideline that automatically changed the order from a stat intravenous bolus dose to a slowly infused dose. The corporate guideline resulted in a 25% drug cost savings to the corporation since it results in 3 doses per day instead of the “standard” 4 doses.

This guideline was adopted according to corporate policy and procedures with the input of several hospital employed physicians. However, none of these physicians was directly involved in the patient’s care. The pharmacokinetic “greater area under the curve” rationale for the changed slower dose scheme has some merit. The rationale for immediately achieved high therapeutic antibiotic blood levels in a blood stream infection also has merit. No conclusive “evidence based” prospective clinical trial data exists to resolve the question of when (if ever) one dose scheme is superior to the other.

Even the term "guideline" is a euphemism. A guideline that automatically alters an attending physician's order, even over his written objection, is not a guideline.

Is this an example of the corporate practice of medicine?

Other examples are even more complicated. The hospital has a corporate directive that all patients being admitted have nasal swab screening for “MRSA” (a nasty, antibiotic resistant bacteria that can cause lethal infections but which is most often innocuous to asymptomatic nasal carriers). Patients are not offered the option of refusing the screening test. This example approximates the rationale for public health authority’s strategies for control of epidemics, which can include involuntarily quarantines, etc. But hospital corporations are not legally vested with the same statutory authority as public health officers.

Why would a patient care if he were screened or not? If he screens positive he will received “decolonization” of the nose and skin with antibiotic ointment and disinfecting chemicals. This has not been proven to be beneficial to the carrier (or for that matter to anyone) and is at best inconvenient and modestly uncomfortable. Exposure to the chemicals and antibiotic may also carry a very small risk to the carrier. The carrier will be put into isolation where visitors must don gowns and gloves. Health care providers must do the same. The net result may result in fewer contacts with friends, family and care givers. The carrier state may also lengthen hospital stay as many post hospital care facilities are reluctant to accept MRSA carriers. The carrier label will follow the patient and future hospital admissions will require automatic isolation per corporate guidelines.

Is this an example of the corporate practice of medicine?

These kinds of situations are not rare now and will become common place as guidelines are imposed by corporate authorities. Physicians who now are often self employed and relatively free to disagree with (and sometimes over ride) practice guidelines will become corporate employees. Ironically, corporate providers are hiring “patient care advocates” to smooth the path to implementation of guidelines. “Patient advocate” used to be part of the job description of physicians.

[To start at the beginning, follow the Roman numerals back to "I".] What are the variables that determine which individuals with thyroid cancer become diagnosed? In the last half-century an American physician who examined this question most famously was Alvan Feinstein. Feinstein was a mathematics major at the University of Chicago and he subsequently graduated from the medical school of that same institution. He went on to Yale University where he and a few others largely invented a new specialty that attempted to marry the disciplines of epidemiology and clinical practice.

Dr. Feinstein was fond of inventing, or at least rediscovering, words. One of his neologisms is the word “lanthanic”. A Yahoo search produced this definition from an online medical dictionary: “A descriptive term for a disease process that does not produce signs, symptoms, or clinical evidence of disease. The term is rarely used today. Lanthanic comes from the Greek word “lanthano” meaning “to lie hidden.” One might quarrel with this definition on several grounds. For example a process that does not produce signs, symptoms, or clinical evidence of disease might not be deemed a “disease process”.

A danger lurks here of falling into Alice's rabbit hole of convoluted semantics. Feinstein stratified diagnostic/prognostic categories according to how the diagnosis was discovered, not on the “snap shot” of the disease at time of discovery. Basically he invented a novel way of staging cancers. He provided evidence that seems intuitively robust. Obviously a patient who seeks medical attention because of shortness of breath caused by a thyroid cancer invading his windpipe is going to have a much poorer prognosis than our hypothetical man. Feinstein would have “staged” our hypothetical man as non-iatrotropic, lanthanic, asymptomatic.

This means our man did not seek out medical attention (his “iatrotropic stimulus”) because of thyroid cancer. Rather his “iatrotropic stimulus” was to seek vascular screening at the insistence of his spouse. However his thyroid cancer might still have been symptomatic so it might have been non-lanthanic. Our hypothetical man might have been ignoring symptoms of difficulty swallowing or he might have had a visible or palpable thyroid nodule. The point of all this confusing terminology is that diseases may behave very differently depending on how they are discovered, even within groups of patients with similar traditional prognostic factors.

Alternatively, two patients with identical diseases may look very different simply because the diseases are discovered at different time points in their natural histories. Feinstein's excellent but now largely forgotten 1967 book entitled “Clinical Judgment” explores many aspects of what today is referred to as “evidence-based medicine”. He pointed out that much of what we consider to be “evidence” is likely artifact. Some of the pit falls that Feinstein highlighted have become part of the medical lexicon such as "Berkson's fallacy" and "the Will Rogers phenomenon".

A mathematical thinker from an earlier era, Thomas Bayes, formalized a quantitative theorem that describes the posterior probability of a hypothesis H (i.e. after evidence E is observed) in terms of the prior probabilities of H and E, and the likelihood of E given H. The theorum implies that evidence has a stronger confirming effect if it was more unlikely before being observed. This is the basis for interpreting diagnostic tests, and it can yield some suprising insights that might seem at first to defy common sense.

Feinstein and Bayes basically were addressing the same problem. That is, how does a transition from the population in general to a subset of that population based upon a given characteristic change probabilities that an event will occur or a condition exists. Feinstein liked to use Venn diagrams to demonstrate the populations and probabilities. Bayes used a mathematical formula to calculate the probabilities.

Let us suppose, in Bayesian fashion, that there is a test that will identify all persons in the population with a given disease. Let us further suppose that this test is 95% specific. This describes a test with 100% sensitivity and 95% specificity. Most people, including many healthcare professionals, when presented with the hypothetical situation that their test is positive estimate that they almost certainly have the disease. The correct interpretation, of course, is that simply knowing the sensitivity and specificity of a positive test does not permit even a rough estimate of the probability that the disease is present in any given individual, unless the test is 100% sensitive and 100% specific.

A disease that is present in one person out of a million would result in 50,001 positive tests if the entire population were tested. This is because 95% specificity denotes a 5% false positive rate. The sensitivity of 100% means that the person with the disease will certainly be identified. If the test costs five dollars, an expenditure of $5 million creates the dilemma of identifying which person of the 50,001 people with a positive test actually has the disease.

If the next transition requires a test that costs $1000 and which has the same sensitivity and specificity as the first test, $50 million will be spent to create a smaller population of 2500 people, each with a 0.004 probability of having the disease. The probability for a person identified by the second test has been reduced from one in a million to about one in a thousand at a cost of $55 million.

The individual person needing treatment is still not specifically identified. Feinstein knew that clinicians intuitively understood the dilemma that rare diseases pose in the diagnostic process. Such clinical aphorisms as “common things occur commonly” or “if you hear hoof beats, don’t look for zebras” have been passed from generation to generation of doctors. Feinstein also knew that the more advanced a disease is, the more likely it would come to a physician’s attention (become an iatrotropic stimulus) and be correctly diagnosed.

Returning to our hypothetical man with the thyroid nodules we can perform a similar probabilistic analysis. However this time we will consider not the diagnostic transitions but rather the transitions related to deciding upon treatment. Here the dilemma is almost the opposite of the problem presented with the diagnosis of a rare disease.

Thyroid cancer is a very common disease. But how should it be treated, if at all? First, a caveat is needed. The numbers used in this exercise are estimates, possibly very poor estimates. Still, much of medical practice is based on poor (or controversial) estimates. Every year in the US roughly 2.5 million people die out of a population of roughly 300 million. Every year roughly 1500 people die from thyroid cancer. Of these, about half die from a very aggressive (“anaplastic”) form of cancer for which there is basically no treatment. That leaves about 750 deaths that might be preventable through a combination of early detection and effective treatment. Each year in the US about 40,000 cases of thyroid cancer are diagnosed. Autopsy studies of the thyroid glands of people who die from all causes show that about 20% had undiscovered “treatable” forms of thyroid cancer. Thus 750 people die from thyroid cancer and 500,000 die with undiscovered thyroid cancer.

Return to the question posed many paragraphs ago: what were hypothetical man’s odds of dying from thyroid cancer before the carotid screening? The answer by very rough approximation is 750 out of 2.5 million, or 3 in 10,000, or 0.03%. His chance of dying with thyroid cancer was 20%. What were his odds of dying from thyroid cancer after the carotid scan but before the biopsy? If 10% of small nodules harbor cancer and if all treatable thyroid cancers start as small nodules (and if all or very nearly all those dying with thyroid cancers had their cancers in a nodule by age 60) then his chance of being among the 750 thyroid deaths is 10% higher than it was before the nodule was discovered, or 0.033%. Using this kind of reasoning, a decision tree model can be constructed.

But the process is far more complicated than our current position in the tree (0.033%), which assumes treatment is always curative, that complications of treatment never occur, and that outcomes are not related to other variables such as co-morbid conditions, size of the thyroid nodule, invasion of capsule to name just a few potential effect modifying or confounding variables. Here is the 1996 recommendation of the USPSTF: “Screening asymptomatic adults or children for thyroid cancer using either neck palpation or ultrasonography is not recommended ("D" recommendation).” An update is underway, but not yet released.

My own conclusion after considering all this is to fall back on the dictum “Primum non Nocere” – first do no harm. The “accidental” discovery of a small thyroid nodule might best be ignored. Of course, I would ask about prior neck radiation, family history, and the like. And I’d advise follow up ultrasounds. The patient, on the other hand, might well only hear the word cancer and demand treatment.

II. The first phase of our man's journey is from the primary care physician to a specialist.  The specialist, an endocrinologist, examines the thyroid and detects no abnormality.  The endocrinologist asks for an ultrasound of the thyroid gland believing that better resolution will result from targeting the organ in question.  Indeed the thyroid scan is interpreted as showing three thyroid nodules, two the left lobe and one in the right lobe, all less than 10 mm. Next, the endocrinologist recommends that an interventional radiologist perform ultrasound guided cytology of the three nodules.  She does so because she believes that there is about a 10% probability for each nodule that thyroid cancer is present. The cytology for two of the nodules is interpreted as “atypical, suspicious for papillary thyroid cancer”. The slides are sent to Boston for review, and the reviewers agree. Now what?

Our man is referred to a surgeon who advises total thyroidectomy. This is skillfully performed without complications. Two of the nodules, including the one seen originally, did indeed contain papillary thyroid cancers of less than 10mm. The third “nodule” seen on ultrasound could not be found. (In 2005 about 60,000 thyroidectomies were done in the U.S. at a mean hospital charge of about $20,000 (weighted national estimates from HCUP Nationwide Inpatient Sample ((NIS), 2005.)) Several more visits to the endocrinologist are required during the next year to adjust thyroid replacement, and this will be done at least yearly in perpetuity. The $50 screening test for vascular disease led to a ~$30,000 thyroidectomy, all in.

The probability that our man will die from thyroid cancer is now very close to zero. The problem is that his chance of dying from thyroid cancer was pretty close to zero before the initial vascular screening. What were his chances of dying from thyroid cancer before the carotid screening? An examination of this question would require a consultation from Dr. Alvan Feinstein and Thomas Bayes. Unfortunately neither of these thinkers is still alive, but we can examine their ideas and attempt to apply them. First we need some estimates of actual risks.

People with a diagnosis of thyroid cancer are becoming more prevalent in the population of the United States.  The death rate from thyroid cancer appears to be quite stable.  A possible explanation for this can be found in the experience of our hypothetical man.  "Screening" for thyroid cancer is not being done in the general population.  However the frequency of imaging of the thyroid has increased dramatically.  Many, if not most, of these thyroid images are acquired incidentally in the course of screening for, or evaluation of, other diseases, such as our hypothetical man’s experience.

A reasonable estimate the number of newly diagnosed thyroid cancers in the United States per year is about 40,000.  A reasonable estimate of the number of deaths from thyroid cancer is about 1500 per year.  The prevalence of thyroid cancer at autopsy in people who died of other causes without a known diagnosis of thyroid cancer has been variously estimated at 5% to 35%. Given the generally slow rate of progression of thyroid cancers it is clear that there are many times more people living with undiagnosed thyroid cancer then there are people living with a diagnosis of thyroid cancer.

The practice of medicine has been likened to the art of probability.  The crucial word in this comparison is the word "art". The mathematical science of probability enables its practitioners to estimate uncertainty with great precision.  Medical practitioners, either formally or informally, use the tools of probability.  Most patients are vaguely aware of this, but neither practitioners nor patients have much formal training in either using or communicating probabilities.

Consider this hypothetical example.  A 60-year-old man, at the insistence of his spouse, participates in a low-cost vascular disease screening event for which he pays $50 out of pocket. His carotid arteries show only average atherosclerotic changes for his age.  Incidentally noted is a 9 mm left thyroid nodule. Because of this he visits his primary care physician.  In the course of the conversation the word cancer is used.

Tens of thousands of dollars later our hypothetical and here-to-fore only casually health conscious man no longer has a thyroid, will be taking thyroid pills for the rest of his life, and still is not 100% sure that he will not die from thyroid cancer. Should he lose his current health insurance or job he is worried that his age and now pre-existing diagnosis of cancer will make him either uninsurable, unemployable, or both.

The decisions that led from a screening test for vascular disease to a total thyroidectomy are rooted in probabilities.  Formal decision analysis requires many inputs. Traditionally medical problem solving first requires a diagnosis and then information about the value, safety and efficacy of treatments.

Conceptually, this is straightforward.  However, in practice, a diagnosis is only as certain as the accuracy of the findings upon which it is based. Very few single medical findings result in a 100% accurate diagnosis.  And even when many tests are considered together diagnostic probability rarely reaches 100%.  In our hypothetical man the incidental finding of a thyroid nodule in the course of testing for another disease probably does raise the probability that thyroid cancer exists.  But by how much and from what baseline risk is the probability raised?

If diagnostic conclusions are imperfect, these conclusions are usually much more robust than those related to the safety and efficacy of treatments. Further complicating the medical process is the fact that the diagnostic and the treatment decision making cannot be considered as unrelated, independent tasks.  For example, if our hypothetical man could be reliably informed that even if the nodule in his thyroid is cancer the probability that this cancer would lead to his discomfort or demise is 1 in 10 million both he and his physician might be comfortable with simply dismissing the finding without any further diagnostic evaluation.

Next, let us travel with our hypothetical man through the medical maze that he has entered through the thyroid nodule door.

[To start at the beginning of this series, follow the Roman numerals back to "I".]

Optimism is running high that the computerization of medical records can stop the escalation of health care costs. Many have even expressed the view that overall costs will fall when the full benefits of this computerization are realized.

Proponents of the "EMR" (electronic medical record) advance several propositions to support their position. As examples, they cite the frequent duplication of services that results from lack of communication.  Second, they believe that a reduction in medical errors most certainly will result from computerization of records and that these errors significantly add to total health care costs.  Third, digitalization of healthcare delivery will reduce costs by facilitating adherence to established protocols. Finally the concept that digitalization of processes increases productivity, like the concept of global warming, has achieved virtually universal acceptance. Skeptics of the the "EMR's" benefits soon will likely join  the exclusive club known as "denyers" whose current membership is restricted to denyers of global warming and the halocaust.

Each of the arguments in support of the electronic medical record can be countered by pointing out the potential for paradoxical effects.

A reduction in the duplication of medical services certainly would be laudable.  But where is there a universally agreed-upon definition of duplication?  Is a second opinion a duplication of services?  Is a lab test that is repeated to confirm its accuracy a duplication of services? Even assuming a perfect EMR that provides all caregivers with simultaneous real-time access to all data, for a universal electronic medical record to reduce wasteful duplication, some type of algorithm must be developed that actually prohibits, or at least sets up obstacles or penalties, when the system identifies duplication.

Life and health care are dynamic processes. Under some circumstances repetitive testing at very frequent intervals is essential.  For example, blood counts on an acutely bleeding patient may be required several times in a single day.  On the other hand a patient with iron deficiency anemia may not need blood counts more often than once in many days. Yet the clinical situations of acute blood loss and chronic blood loss represent a continuum where competent caregivers may make differing judgments about proper testing frequency.

So, of necessity, an algorithm designed to reduce duplicate blood counts must be flexible enough to allow for an extreme variety of clinical situations while still being strict enough to reduce health care costs. The practical consequence of such an algorithm may be to simply create another variant of the game of "mother may I" for healthcare providers.  One form of this contest already exists for third-party payer prior authorization of more expensive testing and treatment.

The game of “mother may I” as currently played by third-party payers probably does serve to reduce health care costs.  When an expensive test or procedure is desired by a healthcare practitioner that practitioner must know the criteria the third-party payer will employ to judge necessity.  Frequently these criteria are opaque and can only be discovered by trial and error, as in the game of “mother may I.”  Healthcare practitioners become frustrated, particularly since the criteria seem to shift from day to day or week to week or case by case.  This is healthcare cost containment by bureaucratic obstacle not logic.

On the other hand, a prior approval procedure that simply listed the exact criteria by which the test would be approved might result in a more modest or even no reduction in test utilization. A logical algorithm for the proper utilization and prevention of duplication of any test would by necessity be a work in progress. Creating and maintaining the enormous number of required algorithms would in themselves be a costly enterprises. Paradoxically, the possibility certainly exists that, at least in many cases, the costs of deploying a duplication reduction algorithm may exceed the dollars that could be saved. The game of utilization "mother may I" can be very expensive and can also have a number of unintended consequences. And the methodology for doing such cost accounting has not been perfected.

A second potential benefit from electronic medical records would be a reduction in medical errors.  No doubt medical errors can be catastrophic and costly.  The Institute of Medicine has suggested that as many as 100,000 lives are lost yearly in the United States alone through medical errors.  This subject has been discussed elsewhere on this blog.  Once again the methodology for studying this problem is in its infancy.  The available evidence does not consistently point to a reduction in errors after adoption of electronic medical records.  Certainly the electronic medical record has the potential for creating practice guidelines and checklists that could reduce errors. But, once again, the potential for paradoxical effects is evident.  As more time and attention is devoted to following the dictums of protocol less time may be available for the exercise of sound judgment.

But the biggest paradoxical effect of the electronic medical record is the tendency for such records to reduce the signal-to-noise ratio.  In the names of documentation and quality, boilerplate entries into the medical record are increasing exponentially.  Some anecdotal evidence can illustrate this.  The Mayo Clinic is world renowned  among healthcare providers as a center for consultation. Until just a few years ago, following a patient visit to the Mayo Clinic, a patient's physician could expect to receive a summary letter reflecting the considered judgment of a single physician who synthesized and summarized the findings.  Recently, however, that cover letter has devolved into a single sentence that indicates that the clinic's findings and recommendations are contained in the following pages.  Often these pages contain words and phrases that are obviously generated by some shorthand or "macro".  This leaves the reader to wonder what nuances may have been lost.  And more importantly how to judge the relative significance of these statements becomes impossible.  What once took a few minutes to understand now can take many times that amount of time.  The paradox of decreasing signal-to-noise ratio may well reduce productivity and increase miscommunication. This cannot reduce costs or errors.

A third benefit from the electronic medical record could be increased adherence to established protocols. Most clinicians view the plethora of existing protocols as useful starting points and as rough guidelines.  But once again for these protocols to be useful in cost containment they must carry some reasonable reward for adherence and penalty for deviation.  While this strategy is simple in concept, it can be very complex in application. Electronic drug prescription has a head start on the comprehensive electronic medical record.  Most electronic prescribing systems link individual patients to the individual formularies of their third-party payers.  The prescriber can be easily tracked to see how often he deviates from the suggested formulary, and an economic profile for each provider can be created.  Of course patients are not randomly assigned to prescribers. A potential consequence of this non-randomness is that prescribers who appear to be more expensive than their presumed peers will feel pressure to revert to the mean.

The prescriber could adopt several strategies in order to disappear back into the comfortable obscurity of the pack. To do this efficiently the prescriber once again would need to know the rules of the game.  Perhaps some measure of severity of illness impacts on the prescriber's profile.  A counterstrategy then would be to "upcode" by adding severity modifiers to the diagnostic codes. Perhaps a prescriber has a patient who is insistent upon receiving branded medication.  A counterstrategy would be to dismiss this patient from the practice.  Neither of these strategies would likely improve overall quality of care or reduce ultimate costs.

The final potential paradox of the electronic medical record to be discussed here is the nearly universal acceptance that digitalization must improve productivity.  This might be viewed as a modern-day corollary of the Curse of Kelvin. Lord Kelvin once famously observed: "If you can't express it in numbers you don't understand it." The curse of Kelvin immediately followed: "If you can't measure it, measure it anyway." In the age of digitalization the binary numbers equivalent of Kelvin's curse is "if you can't digitalize it, digitalize it anyway."

Reduced to simplest terms, the task of the frontline healthcare provider most often is to determine whether a patient on a given day with a given reason for visiting has no illness or a self-limited illness that will not impact on the health of the patient for more than a short time. That is to say, the task of the practitioner is to select from the very large number of patients in the denominator of the "worried well" the very small number of patients in the numerator of patients for whom skillful medical care could make a meaningful difference. This might be termed the "tyranny of the worried well denominator".

For purposes of illustration let us examine a hypothetical situation. Suppose there exists a clinician who can infallably segregate a worried well patient from the group of patients who require additional evaluation and treatment. Suppose this astute clinician could accomplish this 40 to 50 times a day based only on clinical findings. How could the electronic medical record improve this clinician's quality of care or cost effectiveness?

Obviously an electronic medical record, or for that matter, the keeping of any record at all, could not improve upon the performance of our hypothetical clinician. Since by definition the complaint of a worried well patient does not indicate a significant ongoing problem. Wasting the clinician's time in documentation of a unique, one time event could only reduce productivity and increase costs. 

One aspect of this hypothetical scenario that closely approximates reality is the need for frontline practitioners to assess many clinical situations in a day. Medical records must be recorded for a variety of reasons even if the very existence of these records cannot be conclusively shown to impact on important patient outcomes.  The practice of medicine has been described as the art of probability.  Few actual practitioners would claim that their actions rise to the level of the science of probability. At a minimum the medical record should reflect how the practitioner applied the art of medicine in each encounter.

The scientific approach to a problem is often described as a four step process.  First an observation or series of observations are made. Then a hypothesis is generated to explain the observations. The next step is to perform an experiment designed to either confirm or reject the predicate hypothesis. The results of this experiment are then used to draw final conclusion.

Every pediatrician knows that a child who presents with complaints of an upper respiratory infection has a small but finite risk of being in the early stages of potentially lethal meningitis. That pediatrician also knows that the overwhelming majority of such children have a self-limited illness. The medical record of such an encounter should reflect how the pediatrician grappled with the "tyranny of the worried well". A macro generated encounter note can be generated with a few key strokes that would include the pertinent positive and negative observations that support a diagnosis of uncomplicated upper respiratory infection.

Of the several million encounters for URI's in a year only a very few thousand will be complicated by meningitis  (and many cases of meningitis do not have a prodrome). A question for the advocates of the EMR is: Do the key strokes "URI cntl F3" that generates a full page detailed note guarantee a better outcome than the handwritten entry "URI"? The former record, if employed by the scrupulously honest, will provide a basis for assessing process (and a more robust defense should the gods of probability turn ugly). The latter record, if employed by the scrupulously contientious, will be less costly.