Patient-centered shared decision-making is one of the toughest tasks in clinical medicine. It involves a collaborative effort to collect and assess evidence, leavened with clinician experience, and framed within the context and values of the individual patient.
Guidelines can simplify the nearly impossible task of grounding decisions in evidence by synthesizing large volumes of incomplete and conflicting information into a usable roadmap accompanied by some generic best-care principles. Guidelines have become ubiquitous, which is a good thing. Perhaps too much of a good thing: at this writing, theNational Guidelines Clearinghouse lists 2592 disease-focused guidelines, including 692 for neurologic diseases and 526 for cardiovascular diseases.
Solutions are a common cause of problems, and guidelines are no exception. A good guideline properly used is a godsend. A bad guideline (or a good guideline improperly used) is a nightmare. When I talk of the potential pitfalls associated with guidelines, or point to errors in their use, I am often accused of rejecting evidence-based practice or evading my professional obligation to guide my patients to the ‘correct’ (evidence-based) answer.. These accusation reflect ignorance of the nature of medical evidence, unjustified optimism about the quality and applicability of population-based guidelines to individual patients, and a poor understanding of the trust-based patient-clinician relationship.
Lets be clear. I love evidence, depend on good guidelines, and I am committed to helping my patients make the best possible decisions. Patient-centric clinicians do not reject evidence or dodge their obligation to assess and explain medical information and then recommend courses of action to their patients. We do so all day long. The desire for a trusted source of both information and advice is a large part of why our patients come to see us. Although information and advice are freely available (inescapable) from friends and online, they seek our recommendations because they trust us, a trust based not just on our medical expertise, but also on confidence that we are THEIR advocate. They trust us to use the best and most current medical information, of course. They also believeand have the right to expect that we will put their interests ahead of their insurance company, the drug companies, their employer, our employer, and often above our own interests and the needs of our families. We regard that trust very highly, and take great pains to live up to it.
Guidelines are an important tool, but cannot be allowed to displace the patient as the center and driver of care. Here is what patient-centric clinicians are NOT comfortable with:
- Believing that guidelines represent 'Truth' or 'The Answer' rather than our current - and very provisional - best guess, and a good place to start the discussion rather than where the discussion ends.
- Believing that we know what our patients want, or that our understanding of what is best for our patients has more value than our patient's own understanding.
- Believing that population-validated evidence-based guidelines should be allowed to trump individual patient preferences, or interfere with the clinician's goal of helping the patient make their own individual and autonomous decisions.
The second and third items are philosophical issues. Too many policy makers and leaders in medicine appear poorly attuned to current trends, and especially to the changing concepts of patient autonomy and empowerment. I won't address these issues here, but would commend to those interested two fine recent books: 'Your Medical Mind: How to Decide What is Right for You' by Drs. Jerome Groopman and Pamela Hartzband (husband and wife at Harvard Medical School and Beth Israel Deaconess Medical Center) and especially 'Critical Decisions: How You and Your Doctor Can Make the Right Medical Choices Together' by Dr. Peter Ubel (physician, behavioral researcher, teacher and ethicist at Duke).
The first point is an argument for regarding guidelines, and the evidence upon which they rest, with humility and skepticism. I make this argument based on an understanding of how guidelines are created and on my 40 years of practical experience watching The Truth evolve and change. Good studies are very hard and very expensive to do. Many studies are poor, most are useful but limited, and only a few are truly good enough to be durable and widely applicable. Studies are done in limited population sets. Most show correlation rather than causation. Those that show causation are usually very narrowly focused and done in an academic or otherwise artificial setting. The goal is often to prove something rather than to learn something, and the studies are manipulated with this intent. (Medications are routinely tested against placebo rather than against other known and effective treatments. Antidepressant studies almost always exclude patients who are mildly or moderately depressed because it is known that in these groups the impact of medication cannot be distinguished from placebo. Patients with co-morbidities or complications are excluded from many studies of chronic disease to make statistical treatment easier and allow smaller study size.) Negative results are infrequently published and the existence of the conflicting data is kept secret by the commercial sponsor, so any data that challenges the efficacy of interventions is less likely to be seen. Only a minority of studies are replicated. There is extensive literature to show that a large number of studies are retracted, found to be flawed, or contradicted by later studies or new knowledge. Truth is refined or reversed on a regular basis. Examples abound. (For a riveting and non-medical treatment of this, read Ignorance: How it Drives Science by Stuart Firestein, neuroscientist and Chair of Biology at Columbia.) Beta-blockers were contra-indicated in heart failure until they were found to be the first (and, at the time, only) treatment for heart failure that actually improved outcomes rather than just reducing symptoms. Ulcer disease was a psychosomatic illness caused by stress until it was discovered to be related to an underlying infectious process. It was standard to patch eyes for corneal abrasions based on studies after eye surgery, until it was found that this increased pain and complications in outpatient abrasions. Post menopausal estrogen HRT was the standard of care until it was found to increase morbidity and mortality. Vaginal births after C-sections were forbidden until it was learned they could be done safely. Enemas, shaving, intravenous fluids, and internal real and uterine monitoring were standards of care for routine obstetrics. Standardized one-size-fits-all A1c targets remain in heavy use even though the literature and current recommendations (since 2011) say this is inappropriate. Numeric LDL targets are common, even though the latest recommendations are to treat global risk rather than a specific risk factor, and numeric lipid targets have largely been abandoned. Many point-of-care decision support systems and quality initiatives still use a target systolic BP < 130 in diabetics. (After several small studies and a reanalysis of old data called this into question almost a decade ago, a large study was published in 2010 showing that there was no reduction in CAD in diabetics by lowering the BP below 140, and that while there was a decrease in stroke, the adverse effects were unacceptable and mortality was not improved. Those who followed the evidence rather than simply kowtowing to guidelines had changed their clinical approach based on this, often damaging their ‘metrics’ and sometimes resulting in financial penalties.)
Evidence-based guidelines are often touted as a way to eliminate the subjectivity of the clinician’s experience or preferences from decisions. This can be true, but ignores the fact that the guidelines are built based on the subjective experience and preferences of the authors. The Europeans and the US use the same data to assess cardiovascular risk, but the Europeans recommend instituting statins if the 10 year risk is greater than 20%, and the Americans recommended a threshold of 10% until 2013 when they changed the threshold to 7.5% - all this using the same evidence. Urologists and the USPSTF used the same studies to come to starkly different recommendations about PSA testing. Multiple conflicting guidelines exist for mammography. Different professional groups support different treatment goals for hypothyroidism.
Multiple authors have noted that pay-for-performance with financial incentives or penalties tied to targets and guidelines amounts to using an intervention that is not evidence based (there is no evidence that P4P the way it is commonly done improves either quality or outcomes, though it has been shown to improve metrics) to incentivize clinical behavior that conflicts with current evidence-based recommendations and ignores patient preferences.
I am not alone in advocating for humility and skepticism rather than blind acceptance of evidence or robotic application of guidelines. This has become an area of active study in the EBM and quality literature. For example, consider a paper published September 2013 in the Journal of Clinical Endocrinology and Metabolism. The investigators reviewed endocrine and oncology guidelines. Among 357 recommendation in 17 guidelines issued by the Endocrine Society, they found 131 (34%) strong recommendations based on low-quality evidence and 33 (8%) strong recommendations with no supporting evidence. They also evaluated 169 cancer guidelines using published criteria from the Institute of Medicine and found that on average the guidelines met only 2.75 of the 8 standards. An August 2013 review of 18 asthma guidelines found that 9 of the 18 could only be recommended for use by the AGREE II instrument with modifications, and the other 9 were could not be recommended for clinical use. We routinely tell parents of adolescents (and the adolescents themselves) that the HPV vaccine is effective at preventing high risk infections with HPV and subsequent cervical cancer. We ignore (as do the guidelines) the fact that the testing was done in whites, that the available data indicates that blacks are infected by different strains not included in the vaccine, and that there is no outcome data to support the use of the vaccine in blacks or Hispanics.
How timely and mainstream is this issue?
- An excellent editorial in this week’s JAMA discusses the role that changed information and changed approaches to guidelines are having on the treatment of lipids, a discussion that makes it clear that Centricity's numeric target approach to LDL (and the Meridios Dashboard?) is not supported by either the data or the current guidelines.
- An October issue of JAMA had a commentary about the conflict between patient preferences and EBM guidelines.
- A recent brief but nice discussion of some of the problems with guideline issues.
- An important article showing that much of our published research, the information on which guidelines are based, turns out to be incorrect upon further study was published by Stanford researcher John Ioannidis. His 2005 article generated considerable discussion and the numeric conclusions he reached are not accepted as valid by all for methodological reasons, but the thrust of his point has not been seriously challenged and his work is frequently cited on this topic.
- One of my favorite commentaries on Ioannidis' article (because of the delightful and insightful xkcd cartoon and the corollaries he adds).
- I could go on (and on) but I won't.
Please don't misunderstand this as a rejection of either evidence of guidelines. All experienced clinicians need them and use them every day. We have them on our iPhones and iPads and they are downloaded or bookmarked on our laptops at home and desktops at work. We print copies and post selected summaries near our computers at work and sometimes in exam rooms. We often subscribe to online guideline services like NCCN or data collaboratives like Cochrane. We believe that evidence should inform medical decisions and that guidelines can help summarize and integrate huge amounts of clinical evidence and help make the cacophony of information usable. We love good guidelines because they help us do our jobs well.
The objection is to the MIS-use of guidelines. They are, by nature, flawed guestimates of what conflicting information suggests would be the most appropriate choice for a mythical typical or average patient in a limited setting. While they are invaluable for informing decisions, they are terrible at making decisions. It is intellectual laziness to turn the soft and conflicting data behind a guideline into a rigid protocol or a hard target, seductively freeing oneself in the process of the professional obligation to identify and then advocate for the best interests of individual patients. This is the medical version of a phrase I dread hearing when trying to resolve an issue with an institution or business:
"It isn't my fault. That's our company policy."
The skilled clinician is not someone who is good at picking and applying guidelines, but is someone who is skilled at determining when, why and how to deviate from guidelines. The honest clinician will not consider himself skilled because he has used the right guideline, and the skilled clinician will not consider himself honest when he tells the patient the best course is what the current guideline dictates without first assessing the validity of the guideline and incorporating the individuality of that patient. My job is not to get the patient to follow the guideline but to see how the patient and I can use the guideline to improve outcomes.
On a practical level, there is a very real difference between the two approaches, which I would characterize as follows:
- I prefer nuanced, patient-centered, shared decision-making based on evidence and patient preference, using guidelines only where appropriate, and using guidelines to inform rather than determine the decision. This approach can and sometimes should include my recommendation, which patients often request and usually appreciate. This approach focuses on quality and outcomes rather than on metrics or recipes.
- I object to standardized and therefore necessarily simplistic and paternalistic guideline-centered decision-making, where individual patient characteristics and preferences (and often the evidence itself) are secondary to achieving often poorly chosen targets, and where the focus moves from quality and patient outcomes to proxy metrics and their associated recipes.
Some stylized examples of conversations with patients follow. See if you can distinguish between the patient-centered and the guideline-centered approaches.
Pap smears:
- The current guidelines tell us we get the same benefit from a Pap smear every 3 years as from a Pap with co-testing for HPV every 5 years. If we do just the Pap smear and it stays normal, we can stop when you are 65. More patients who do the co-testing have false-positives that result in a procedure to biopsy the cervix, and if we do the HPV co-testing and the HPV is positive we may have to continue screening until you are 75. Do you have any questions about this or how it applies to you?
- It's been three years, so you are due for a Pap. (Or: It's been 5 years so you are due for a Pap and HPV.)
Pneumovax:
- The USPSTF and other groups recommend pneumovax. It is often called pneumonia vaccine but it does not actually protect against pneumonia. It is good protection against two uncommon but very serious kinds of infection with the pneumococcus bacteria: blood and brain infections. The evidence tells us that 1 out 4000 people who get it will avoid this very serious and often fatal illness, and about 1 out of 3 will have aggravating but not dangerous reactions to the vaccine. Before we decide whether or not you want this immunization, do you have any questions or concerns? (I don't volunteer it, but if they ask, I tell them that I have had my pneumovax, because I was willing to put up with the likelihood of an aggravating but temporary vaccine reaction to protect me against something that was very unlikely but could be devastating, sort of like buying homeowner's insurance.)
- You are 65 so you are due for the pneumovax. (Or worse, for the ‘pneumonia shot.’)
Colonoscopy:
- There is good evidence that we should be screening average risk patients for colon cancer at age 50 and periodically thereafter. Doing this actually prevents many colon cancers. The evidence is best for colonoscopy every ten years, but the recommended ways to screen also include sigmoidoscopy every 5 years or annual stool testing for blood. These alternatives are less intrusive and less expensive, but have lower effectiveness. Based on the evidence, I strongly encourage my patients to have a colonoscopy. Do you have questions/concerns, or should I set up a colonoscopy?
- You are 50 so you are due for a colonoscopy. (Note: there is some evidence that a substantial number patients who say no to this, or who say yes only to avoid an argument but then do not follow through, would say yes and follow through if offered sigmoidoscopy or stool testing. I have experienced this in my own patient population often enough to believe it.)
Diabetes and role of A1c in management:
- Patient A: The A1c is a test that is quite reliable at telling us what your average blood sugar has been over the last 100 days or roughly three months. This helps us see how well our treatment is working and that helps us decide how hard we should work to improve things. In patients without diabetes it is less than 6.0, and in groups of patients with diabetes, the higher the A1c, the greater the number of complications. Because you are recently diagnosed, young, without other complicating illnesses, and do not have complications of diabetes, the evidence-based recommendation is that we try to keep your A1c as close to 7.0 as possible, even though it may mean pushing hard for life style changes and using more medication. Let's talk about how we can get your A1c down to 7.
- Patient B: The A1c is a test that is quite reliable at telling us what your average blood sugar has been over the last 100 days or roughly three months. This helps us see how well our treatment is working and that helps us decide how hard we should work to improve things. In patients without diabetes it is less than 6.0, and in groups of patients with diabetes, the higher the A1c, the greater the number of complications. Because you (are older, have other significant medical problems, are on other medications, already have diabetic complications, have limited resources) the current evidence based recommendation is that we individualize your A1c target to (somewhere between 7.5 and 9.0 for most patients), to balance the benefits of better sugar control against the problems associated with more aggressive treatment. Lets talk about what changes we can make to improve your sugar control.
- The goal for your A1c is 7.0 so we need to check it every 3-4 months and see you to adjust your dose or add medications until we get there. Then we'll check it every 6 months.
LDL and statins in the setting of diabetes:
- The old guidelines set a target LDL of < 100 in diabetics. The current more evidence-based guidelines have eliminated numeric targets and simply aim to have diabetics (and patients with CAD or stroke) on an effective statin in the dose that they tolerate. Because there is no evidence that other cholesterol medications help improve outcomes (even though they make the numbers better) we can stop your niacin, but we should consider seeing if you tolerate a higher dose of your statin.
- Because you have diabetes, we need to get your LDL under 100 with an ideal goal of 70, using a combination of statins and other medications like niacin.
Influenza vaccine:
- It's influenza vaccine season again. Influenza is a common and potentially serious illness which kills more than 20,000 Americans every year. We have a safe vaccine which can prevent between 1/2 and 2/3 of cases of influenza in a population and prevent thousands of deaths. There are three reasons I recommend it to my patients. One, it is safe. Two, it prevents many cases of a serious illness that at best causes several days to a week of missed work and play, and at worst is fatal. Three, the best way to protect the very young, the old, and those with chronic disease (all of whom have immune systems that don't always make good antibodies even if they get the vaccine) is to vaccinate as many of the rest of us as possible. That's why my wife and I always get our flu shots. Even though it often makes our arm sore or we have a headache or mild fever the next day. We get it every year, without exception. Do you have any questions or concerns about getting the flu shot? (Notice that I ask about concerns before I ask for an answer. Behavioral data in and out of medicine suggests that, once people have committed to a position, factual information contradicting that position is more likely to strengthen than change their position.)
- It's influenza vaccine season again. If you haven't already gotten it, we should give it today.
Guidelines are a necessary component of shared patient-centered decision-making. They can benefit both patients and clinicians. The patient-centered approach is compatible with clinician advocacy: it can include an unsolicited recommendation, as long as it is coupled with balanced information and an opportunity to answer questions and explore the patient's perspective. Patient-centered decision-making is harder and it takes longer, but the evidence suggests that less is done, outcomes are better, costs are lower, and patients are more satisfied with both the process and the outcomes.
Guidelines are tools. They can be good tools or bad tools. They can be sharp tools or dull. They can be the right tool for a specific job, or useless for that job. But they are only tools. They are not the job. We are clinicians with a sacred obligation: our job is to keep our patients’ welfare as the highest priority. Study and use the guidelines, but stay focused on the patient. Practice patient-centered care, not guideline-centered care.
*********************
(Note: Because this post originated from discussions about QI and P4P at my institutions, the examples here were selected as pertinent to typical QI and P4P programs. However, this is an issue that permeates every clinician's day. Choices among equally valid options (or among equally unstudied options) abound: antibiotics versus surgery for appendicitis, antibiotics or watching for spontaneous resolution of an uncomplicated otitis media, duration of treatment for a simple cystitis, when to treat for bacterial sinusitis or bacterial pulmonary complication of a lower respiratory viral illness (bronchitis), what tests to do when a patient complains of fatigue, how often to repeat the evaluation of fatigue when prior evaluations have yielded nothing, whether to treat a blood pressure of 148/86, whether to treat mild or moderate depression with medication, Hepatitis A vaccine and rotavirus vaccine in infants, when to switch to preventive treatments from acute flare treatments in remitting illnesses like migraine and gout, appropriate treatment of poison ivy. Most of what we (PCPs) do during our clinical day resides in the gray zone where there is either no evidence or poor evidence, so we must use common sense, basic sciences like physiology, microbiology and pathology, custom, and anecdotal experience to select and present options to patients. We should be honest about what we are doing and stop pretending that our scientific training and access to knowledge gives us the ability to determine the correct answer for others.)