The Good Doctor

by Kenneth Brigham

  How can you tell if your doctor is a good doctor, not just a nice, friendly person with good people skills? Well, there are two parts to that question: How does your doctor stack up in the larger world of doctoring? And how well-served is your health in this specific doctor-patient partnership?

  Driven largely by the needs of health care managers and funding sources to measure productivity, efficiency, and quality of care, a lot of effort is expended searching for metrics, things that can be measured and expressed in numbers that accurately reflect whether or not a doctor is good at what she does. This is harder than you might think. True, some things can be measured—blood sugar and related biochemicals, blood pressure, blood cholesterol levels, etc.—and such measurements in all of the patients in a doctor’s practice give some notion of success for those specific health outcomes in that group of patients. But if we must have metrics (numbers) then we’ll be judging how good a doctor is from the limited number of things that are readily measurable. For example, none of the doctor rating systems (there are several) include a measure of misdiagnosis, which is surely a critical measure of a doctor’s skills. The National Academy of Medicine says that misdiagnosis, arriving at the wrong conclusion for what is causing a person’s illness, is not a rare problem; somewhere between 10 and 25 percent of diagnoses are wrong. One study published in 2012 concludes that more than forty thousand patients die annually in this country’s intensive care units as a direct result of a wrong diagnosis.99 You doctor may have all the right boxes checked off on the list of measurable things and still be less effective than you would like.

  And a lot of undigitized things influence how well a doctor’s patients do—socioeconomics, severity of illness, mental and emotional capacity, etc.—so when the numbers are added up, they may or may not look good depending on some intangibles that are difficult to account for. It is even possible that the noblest doctors wind up with poor ratings based on the numbers because of whom they choose to serve. Those numbers are information and we should pay some attention to them, but we should not forget that information is not knowledge. If you really want to know how good your doctor is, you’ll need to look beyond the numbers.

  In spite of acknowledged flaws in any numbers-based rating system, the health care managers continue to insist that such ratings are essential for making doctors accountable and improving the value of health care. If you can’t measure it you can’t manage it goes the old business adage with an intuitive appeal that captivates the suits. The idea has even found its way into U.S. law! The Affordable Care Act requires the Center for Medicare and Medicaid Services (CMS) to “make publicly available . . . information on physician performance that provides comparable information on quality and patient experience measures.”100 Beginning with a web site called Physician Compare, the feds are well on their way to giving us a doctor rating system.101 Well, what’s wrong with that? Doctors should surely be accountable for what they do, and we need some way to know if we are getting what we are paying for medicine-wise.

  There’s nothing wrong with getting the available numbers and collating the data in a way that we can grasp. The big flaw in this approach is mistaking information for understanding. The conscientious doctor will pay close attention to her numerical scores, however they turn out to be developed and expressed. She will use those numbers to improve the things that she does that can be digitized. You ought to look at the numbers too; they are information. However, your doctor and you will value what she does, judge how good a doctor she is, from your personal experiences in this health care relationship—How healthy are you? When you get sick, how quickly do you return to health? How often and how badly does your doctor get it wrong and how does she deal with such mistakes? Don’t count on others’ opinions either; Press-Ganey scores, a commonly used measure of patient satisfaction, are, again, information, but they are not understanding. If a doctor refuses to prescribe antibiotics for a bad cold or some other unproven remedy hawked by a guy wearing surgical scrubs on TV, a disappointed patient may give her bad Press-Ganey scores precisely because she did the right thing—a bad score for a good doctor. Such scores may have little to do with your personal experience with your doctor and even less to do with how good she is at her chosen profession.

  But still, argue proponents of digitized medicine, if enough health-related things could be measured, it should be possible (given the robustness of increasingly muscular computers) to integrate all that information into numbers that accurately describe the quality of your care. The problem, these folks would say, is that too few things can be measured and so the major advances to come will be in measurement technology and computer programs for integrating information. Give computers enough information and clever enough algorithms and they can transform information into knowledge.

  The current boom in wearable sensors that translate what’s happening in your body into numbers is an example of this approach. And emerging technologies will expand the potential of such devices beyond what we can imagine now. According to Gary Wolf, author of Quantified Self, we may be able, before long, to monitor in real time, in quantitative terms, “sleep, exercise, sex, food, mood, location, alertness, productivity, even spiritual well-being.” New measurements are being added to the list at a dizzying pace. Wolf says, “more than thirty thousand new personal tracking projects are started by users every month.”102 It may well be that the coincidence of tiny electronic sensors, powerful smart phones, social media, and the cloud will enable construction of your quantified self.103 We will learn a lot about human health from such attempts. They may even shed some new light on the challenge of relating population data to your uniquely personal condition and evaluating your doctor’s performance.

  The numbers game is enticing to doctors as well as patients. Like Thomas Gradgrind in the Dickens novel Hard Times, when faced with the ambiguity that is inevitable in human conditions, both we and our doctors are tempted to plead, “Fact, fact, fact!” The American Heart Association’s website encourages you to “Know Your Health Numbers”—weight, body mass index, blood sugar, blood pressure, cholesterol, maybe even hemoglobin A1c (a diabetes index) and C reactive protein (a measure of inflammation).104

  Numbers sound like something solid that you can rely on, but it’s not so simple. If you Google know your numbers you’ll get ten pages of hits luring you to a variety of programs with cleverly designed web pages. Some of these have been accused of “non-evidence-based fear mongering.” And you don’t want to lay awake nights worrying about your hemoglobin A1c or your C-reactive protein unless those numbers have a personal meaning to you. Losing sleep will likely make the numbers worse, threatening a spiral of mutually reinforcing unhealthy results.

  So you’ll need help sorting out the best numbers-based program for you. And you’ll also need help prioritizing the numbers, evaluating their accuracy, and understanding how your behavior influences them. If you decide to get into the numbers game by yourself, you’re likely to run into trouble. A good doctor can help.

  But the good doctor also knows that what’s best for you isn’t always determined by the numbers. Information is not understanding! If you fall down the entranceway stairs in your condo and break your leg, this doctor knows that you’ll want to get it taken care of as fast as possible. She’ll see that you get a proper cast, adequate pain medication, a walking cast as soon as possible to free you from the crutches, and the cast removed after the minimum required number of days. However, she also knows that if you are a cross-country skier and break your leg, you may want something quite different. You’ll want to be free of pain of course. You’ll want to get rid of the crutches and the sooner you can get rid of that pesky cast, the happier you’ll be. But, what you really, really want is to be back on your skies schussing across a pristine snowfield inhaling the brisk winter morning air. As far as you’re concerned, the value of your attempts to be healthy is tied directly to how much they advance you toward what you really, r
eally want. If leaving the cast on a few more days and continuing with the crutches, or even dealing with some additional pain, will get you back on your skis faster, then you’ll probably go for it. And you’ll go for it with this doctor’s blessing. She understands.

  The people who study behavior talk about value driven planning.105 That’s where you plan to use your resources to do the things you value most, starting at the top of the list. Some people do a better job of taking care of their health if they approach it that way. While tracking numbers may make a lot of us behave in healthier ways, for some of us, what feels important can be a more potent motivator than numbers. If you can realize the things you value most about your health, the numbers may take care of themselves. You might practice yoga or meditation. The idea of mindfulness might appeal to you and mindful eating may be the most effective way for you to control your diet. You might find the time for long walks in the park, or time in the early mornings to sit in the garden and listen to the birds awaken. Your doctor should keep track of your numbers and see that you pay attention to the important ones. But she will realize that you may be more value driven to do whatever you need to do to be healthy. The thoughtful doctor knows your numbers, she has the information, but she also understands what they mean to you.

  Phillip Peterson, Get Inside Your Doctor’s Head (Baltimore: The Johns Hopkins University Press, 2013).

  M. Chung, R. Oden, B. Joyner, A. Sims, and R. Moon, “Safe Infant Sleep Recommendations on the Internet: Let’s Google It,” Journal of Pediatrics 161 (2012):1080-1084.

  F. North, W. Ward, P. Varkey, and S. Tulledge-Scheitel, “Should You Search the Internet for Information about Your Acute Symptom?,” Telemedicine Journal and E-Health 18 (2012): 213-218.

  Todd Rose, The End of Average (San Francisco: HarperOne, 2015).

  Bradford Winters, Jason Custer, Samuel Galvagno Jr., Elizabeth Colantuoni, Shruti Kapoor, HeeWon Lee, Victoria Goode, Karen Robinson, Atul Nakhasi, Peter Pronovost, David Newman-Toker. “Systematic Review: Diagnostic Errors in the Intensive Care Unit: A Systematic Review of Autopsy Studies,” BMJ Quality & Safety 10 (2012): 1136.

  Section 10331(a) of the Patient Protection and Affordable Care Act (ACA).

  Physician Compare Initiative, www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/physician-compare-initiative/Physician-Compare-Overview.html.

  Gary Wolf, “The Data-Driven Life,” The New York Times Magazine, April 28, 2010.

  Gary Wolf, “The Quantified Self,” TED, https://www.ted.com/talks/gary_wolf_the_quantified_self?language=en.

  “Know Your Health Numbers,” American Heart Association, http://www.heart.org/HEARTORG/Conditions/Diabetes/PreventionTreatmentofDiabetes/Know-Your-Health-Numbers_UCM_313882_Article.jsp#.V9LSWZgrIhc.

  Matthew Leitch, “Value Driven versus Target Driven Planning,” Dynamic Management for an Uncertain World, July 5, 2006.

  CHAPTER 10

  The Good, the Bad, and the Ugly of Statistics

  When we use single numbers to estimate uncertain future outcomes . . . we are not just usually wrong, but are consistently wrong.

  —from Harry Markowitz’s foreword to The Flaw of Averages: Why We

  Underestimate Risk in the Face of Uncertainty by Sam L. Savage106

  Here’s the good doctor’s problem with statistics. Statistical significance is the virtually universal criterion that medical doctors and scientists use to determine what is true. But the good doctor knows that statistics are perfectly capable of lying. Statisticians tacitly admit that. The American writer Darrel Huff even wrote a DIY manual on the subject that he titled How to Lie With Statistics,107 the best-selling statistics book of the second half of the twentieth century.

  Statistics are about probabilities (how likely something is to be true or not) and the magic number is 0.05 (or 5 percent). If results have less than a five percent chance of being wrong (p<0.05; statistically significant in the jargon) then they are deemed likely to be true. But a probability of 0.05 means that there is a one in twenty chance that the results are totally random, not true at all. And how does your doctor know if you’re among the one in twenty for whom the truth is different than what the statistics say? Of course she doesn’t know until she learns everything she possibly can about you and maybe not until after some diagnostic and therapeutic trial and error. The problem is that statistics are meaningless for an N of 1; and you, I, and the other seven or so billion of our kind are each an N of 1.

  One of us (MMEJ) practiced for many years as a cancer surgeon. His conversation with a patient newly diagnosed with a serious cancer might have gone something like this:

  Patient: Well, doc, what are my chances of beating this thing?

  MMEJ: Your chances of surviving this are either 0 or 100 percent.

  Patient: What do you mean? Can’t you give me a number, some kind of odds?

  MMEJ: If you want to know what percent of a large group of people with this cancer will survive it, I will of course give you a number. But for you personally, it’s all or none. You will either survive it or you won’t. So let’s be positive and assume that you’re a survivor.

  Patient: I’m not sure whether I ought to feel optimistic or depressed.

  MMEJ: Given the only alternative, don’t you think it makes the most sense to assume you’re going to beat it?

  Patient: Maybe so, doc. Maybe so.

  What this imaginary patient was told was true, but having convinced him that he and his doctor should set out a course of treatment assuming the best possible outcome, then what? None of the three major therapies for cancer—surgery, radiation, or drugs either separately or in some combination—is without serious collateral damage. So how do the doctor and the patient decide on a treatment that has the best chance of succeeding while doing the least harm?

  Well, in spite of the argument that statistical analysis of population studies may not be relevant to an individual patient, we have to start somewhere, and so we start with the one size suit, the statistical results of rigorously conducted and analyzed studies of groups of patients with a similar disease. A one size suit is, after all, better than no suit at all.

  That’s a starting place, but where do we go from there? Sorry, but a few more words about statistics are necessary since that is how most folks go about measuring the value of evidence.

  Most studies of effects of interventions in groups of patients are analyzed using frequency-based statistics—means, medians, confidence limits—all of the numbers from individual patients condensed into uber-numbers that describe the whole population and carry the weight of statistical authenticity; p<0.05=probably true, p>0.05=maybe not. But these numbers don’t apply directly to individuals even in the study group, much less to the patient sitting in a doctor’s office examining room anxiously awaiting advice about what to do and feeling very much alone. Some statisticians have taken on this problem and tried to expand their craft to deal with how information from groups can be used to inform a specific circumstance. Those efforts started around the middle of the eighteenth century.

  In addition to coming to grips with her reservations about statistics mentioned above, somewhere along her way, the good doctor also met the elusive statistician, philosopher, and, judging from an alleged portrait, somewhat overweight and dour Presbyterian minister, Rev. Thomas Bayes of Tunbridge Wells, Kent, England.108 Rev. Bayes was born circa 1701 and died at the age of fifty-nine having never published the work that is his major legacy. His notes on what came to be called Bayes theorem were published two years after his death by his friend Richard Price in the Philosophical Transactions of the Royal Society of London. The article was titled, “An Essay towards solving a Problem in the Doctrine of Chances” and the ideas described in the article gave rise to an entire field of what is still
known as Bayesian statistics.109

  The basic idea of Bayesian statistics is that the likelihood of a specific thing happening, the statisticians’ probability, is influenced by a number of factors in addition to previous group experiences and that probability changes as more information becomes available. There is a school of thought that this way of interpreting medically related information is closer to being accurate and is more useful in practice than the more commonly used frequency-based approach.

  Without dealing with the math, here are a couple of illustrations of how it works. Computer scientist, engineer, and educator Kevin Boone uses the simple illustration of picking a probable winner of a two horse race110—he calls the horses Fleetfoot and Dogmeat. The horses have raced head to head twelve times in the past; Fleetfoot won seven of those races. Seems easy, bet on Fleetfoot. But it turns out that on four of those occasions it was raining and on the rainy days Dogmeat won three of the four races. So if it’s raining on the day of the race, bet on Dogmeat (ignoring the poor animal’s unfortunate name). If it’s sunny, put your money on Fleetfoot. When additional factors are taken into account the predicted outcome may differ from that predicted by the overall data from the entire experience.

  But the implications of additional information are not always so obvious. Consider the Monty Hall problem.111 You are a contestant on the quiz show Let’s Make a Deal and Mr. Hall, the show’s host, tells you that two of three closed doors hide a goat and the third hides a car. You are to guess which one has the car and you choose door number one. Before finalizing your choice, Mr. Hall opens door number two revealing a goat and gives you the opportunity to either stick with your original guess or change it to door number three. What choice gives you the better chance of revealing the car?