Sunday, January 18, 2009

The Distributed Knowledge of Medicine

The recent entry “A doctor’s view on diagnosis” reveals the challenges that physicians and their patients experience in the face of an unknown diagnosis. John’s situation is regrettably common: being sent to multiple doctors with none able to render a convincing or actionable diagnosis. This leaves the physician and patient emotionally drained and despondent, feeling as if there is no one who can help.

A counterbalance to complete despair is every physician’s first hand or anecdotal knowledge of a seemingly unsolvable case spanning months or years of evaluations, testing, and treatment that is suddenly and unexpectedly diagnosed. Frequently this change of fortune comes about by crossing paths with a single doctor. Fans of Fox’s House will recognize this as a weekly phenomenon. The opening chapter of Jerome Groopman’s How Doctors Think vividly outlines this occurrence, highlighting the weariness on the part of both the patient and the physician in perusing yet another referral and culminating in the diagnosis of celiac sprue, a prevalent but perpetually under-recognized illness. Recently, a young woman suffered from hearing loss, urticaria, and arthritis for decades until she walked into the office of my colleague who recognized the very rare Muckel Well’s syndrome. Relief from a lifetime of pain came within days by instituting treatment with an interleukin-1 receptor antagonist.

In the New England Journal of Medicine weekly case report, expert diagnostic discussions of challenging cases are presented. The point is made on multiple occasions (see June 26, July 3, and August 14, 2008) -- and with refreshing honesty -- that the case was solved on account of the consulting physician’s experience with the disease in question. After 9 years of thirst, insomnia, fatigue, arthralgias, and endocrine abnormalities followed by the development of radiographic abnormalities in the pituitary gland, cerebellum, and aorta, Dr. John A. Mills of the Rheumatology, Allergy, and Immunology Division of Massachusetts General Hospital made the very rare diagnosis of Erdheim–Chester disease (a type of histiocytosis) (

At the conclusion, the following exchange occurs between the physicians at the conference:

Dr. Harris: This patient saw internists, neurologists, and ophthalmologists in his home town, and at [Harvard University’s Massachusetts General Hospital] he saw two neuro-oncologists, a neuro-endocrinologist, and a hematologic oncologist. He also saw another rheumatologist after the biopsy results were available and before seeing Dr. Mills. Experienced pathologists did not consider this diagnosis when the biopsy specimen was first seen. In retrospect, the lesions seen on radiography that suggested the diagnosis to Dr. Mills had been present for several years, but their significance was not appreciated….This experience tells us that the diagnosis of histiocytoses can be remarkably difficult, even with all our available expertise and technology.

A Physician: It seems that this case has lessons for radiologists, pathologists, and internists. Dr. Mills, since most of us here are internists, what lessons should we draw from this case to make us better able to recognize this type of disorder sooner?

Dr. Mills: I was aware of this condition because I have been interested in vasculitis and particularly in aortitis…I have seen several patients with Erdheim–Chester disease who were referred to me with a diagnosis of aortitis, as this patient was. If you had not read literature on this disease, it is unlikely that you could have arrived at the diagnosis.

Whenever I observe this type of insight (and the circumstances that bring together the patient with that particular physician), I recall my challenging cases and think to myself that there must be somebody who recognizes the collection of signs and symptoms that evades diagnosis by me, my network of colleagues, and even Google. I am reminded that there is often a physician whose specific knowledge and experience -- which may derive from a single patient seen years ago or from their lifetime’s work -- holds the key for an undiagnosed patient.

The collective knowledge of human medicine and the experiences and insights of the worldwide medical community are vast. Medical science can be described as a distributed knowledge ( it is the union of all the knowledge of individuals in a community, which is far more than any one physician, hospital, university, or geographic locale can possess.

A great challenge is determining how to harness this knowledge and make it accessible for physicians who are faced with diagnostic dilemmas. Isabel ( and other diagnostic support systems are one way to connect to this distributed knowledge network. Physician interaction through the internet, essentially replicating the familiar curbside consultation many times over, is another promising way to connect a patient’s case with the appropriate physician. If these sorts of efforts were undertaken with regularity, it is possible that the solutions to difficult cases would come about not by serendipity, as in the instances described above, but rather by systematic and deliberate access to the global medical community’s knowledge and talent.

Monday, January 5, 2009

A doctor's view on diagnosis

I'm a family doctor. I worked for the past 6 years with one mid-level provider in a small office in an urban setting in New Jersey.

John was a patient I met soon after starting in my office. I saw his whole family, including his 6 children, his girlfriend and a sister. John was in his late 30's, generally healthy with only the occasional cold and mild, diet controlled high blood pressure. He was tall and thin, friendly, talkative. Honestly, I saw him more for his childrens' routine visits than for himself.

About a year later, John came in complaining of mild weakness in his left leg. After thoroughly examining him, concentrating on any neurological signs and symptoms, I called the oldest, most experienced neurologist in my health system, Dr. M. He and I agreed upon blood and imaging tests.

John and I agreed on the plan: we'd meet in the next week to review the results of all the testing. I had scheduled Dr. M's first available appointment which was in 6 weeks. This would be used either to confirm the diagnosis and continue or start treatment or to continue the workup if the first tests didn't lead to definitive diagnosis.

I was worried about multiple sclerosis.

The tests were all perfectly normal. John saw Dr. M, whose further testing was all normal. Dr. M sent John back to me. John's left leg got worse and his right leg became symptomatic. I sent him back to Dr. M. Dr. M repeated tests, failed to find any abnormalities and sent him back to me. I began to see John once a week. I took an hour (4 patient visits) to redo a complete history and physical. John saw 3 other neurologists, including a renowned physician at the University of Pennsylvania. Except for one, all doctors repeated the MRI at their institution. All physicians reported back that John was healthy - one suggested a psychiatrist.

I was beginning to dread our weekly meetings. My practice was thriving. Being a Spanish speaking female in an underserved, predominantly hispanic area, I was seeing 25-30 patients a day and working straight through my "lunch" and well into the evenings when I should have been home with my children. I was really busy. John would sit on the table and I would look over new MRIs and neurological tests I'd only heard of in medical school. I had nothing new to report. I couldn't be reassuring. I searched every resource I had. I called my residency faculty neurologists. I called the renowned physician to pick his brain (we never were able to speak, despite my numerous attempts.) I got curbside consultations at every Grand Rounds I attended. I was beginning to resent John. I wondered what he thought of my abilities as a doctor. As a person.

The vast majority of the typical family doctors' patients are accurately diagnosed instantaneously. I have taught medical students for years. One of the clinical skills that every student of mine hones is the process of creating a differential diagnosis based on the patient's age, gender, past medical history and chief complaint before even entering the room. Asthma exacerbation, diabetes, viral infection, hypertension, skin infection, vaginal infections, heart failure. These are the 'bread and butter' diagnoses we live and breathe.

Then there are the diagnoses we make after 2 visits, after waiting for some blood results or imaging studies. These are straightforward enough as well: renal failure, some heart failures, high cholesterol.

There are also some less common illnesses that take some creativity, the ability to see the whole picture, or the help of a specialist, but most of these patients return for management after only one or two visits with other physicians.

So, what was I to make of John? What was he supposed to do? I left my practice this past August when I moved to California. John still had no diagnosis. He walked with a cane, but at least his symptoms had stabilized. One of the neurologists had begun treatment for MS, although still refusing to formally give him the diagnosis due to lack of criteria.

As I think back on my thousands of patients, John was one of my most frustrating, disappointing encounters. I think this is because I, like many doctors, thrive on helping people. My pay is another person's relief, gratitude, happiness. Ultimately, I only superficially provided some of that for John because I care. I failed to truly help because I was unable to help him secure a diagnosis. Even if I could have facilitated finding the doctor who could have diagnosed him, that would have been success.

As doctors, we tend to diagnosis quickly and concentrate more on the details of treatment. Many of us see this as the interesting and rewarding part. But as I think back on cases like John's, I realize that diagnosis can be quite elusive. When you think about it, it can turn into the limiting factor to pursuing treatment. I do believe that we need better tools for complex diagnostic cases.

Health 2.0 tools are fostering great connections between individuals in health care. Nonetheless, most of these tools are focused on treatment and disease management. It's time we use them to help solve these diagnostic challenges.

Saturday, January 3, 2009

Online Diagnostic Tools

The first thing many patients do upon experiencing symptoms is to seek information online to try to understand what may be causing their symptoms, how serious it is, and what they can do about it. This time spent searching online increases exponentially the longer a patient goes without being able to find a diagnosis, with an inflection point once a patient realizes that his/her condition may be serious. But while there has been a proliferation of online health tools over the past several years, very few of these are actually intended for, and effective at, assisting with the process of diagnosis. This leaves these patients short on options and represents a significant opportunity for web-based health tools to deliver value.

With the onset of any condition more substantial than a flu or cold, most patients will Google their symptoms to see what they may have. Unfortunately, this is largely ineffective for most conditions. Patients often have difficulty identifying and describing symptoms and understanding the information they read about a condition. Cyberchondria, a Microsoft Research study of consumer use of the internet for diagnosis, concludes that patients frequently misinterpret health information online and that trying to self diagnose can even be detrimental. Additionally, significant symptom overlap between conditions and high prevalence of nonspecific symptoms such as “abdominal pain” or “headache” mean that using search tools to try to assist with diagnosis is largely ineffective for those untrained at interpreting complex medical information.

These limitations also extend to health information sites such as WebMD or Healthline. While these services can be good sources for general health information, they are largely limited to basic information about treatments or conditions and their diagnostic tools are highly limited. Informational sites from the Mayo Clinic, Cleveland Clinic, and Harvard Medical School provide more sophisticated condition overviews, but expect to address the issue of diagnosis in their clinics.

Yahoo and Google Groups, other condition-specific message boards, and patient social networks can be valuable ways for patients to share information. There are groups for nearly every condition and many provide helpful insight into the aggregated experiences of members, accelerating the learning curve for new patients. At the same time, patients must be cautious of much of the information presented through these sources. Beware individuals with agendas (e.g. pushing “miracle” cures) and focus on the posts that provide links to reputable sources to back up claims. Also, patients should be aware of some of the psychological challenges associated with their status as a patient. Patients tend to want to believe those doctors (or other sources) who provide good news and disregard the ones who do not. This is natural, but not always accurate or productive. In a similar manner, patients may choose to attribute negative side effects of a new medication to evidence that it is working (e.g. a “Herxheimer reaction”). Patients are in a difficult and vulnerable situation, but keeping these challenges in mind will help them to accurately interpret information to give them the best chance of securing a diagnosis.

Cleveland Clinic and Partners Online (a collaboration between Harvard, Mass General, Brigham and Women’s, and Dana-Farber) have developed valuable remote second opinion services for patients (see Diagnosis Tools section). These services involve patients (and/or their doctors) sending in their medical record with a specific question, whether it is to establish or confirm a diagnosis or to determine the best treatment. For a fee, specialists at these medical centers then review the files and issue an opinion. In a retrospective study of the first year of this service, Kedar et al found that 5% of patients using Partners Online received a change in diagnosis and noted that the implications of this for patient prognosis and cost of care were profound. This demonstrates not only that remote diagnosis is possible once sufficient health information is collected, but also that it can be extremely valuable in terms of health and economic outcomes. These services are currently limited to a handful of conditions, tend to be more focused on treatment than diagnosis, and are generally not covered by insurance, but they have helped a large number of patients and represent an important step in the right direction by allowing individuals to access top quality care for their condition, regardless of location.

While online tools can be extremely useful adjuncts and information sources, there are few effective resources aimed at resolving complex diagnoses. Nonetheless, the early success of online second opinion services points to great potential to use the internet to expand these applications. As use of digital health information and acceptance of telemedicine continue to grow, I expect that we will see the provision of a broader array of health services online, including those to assist with complex diagnoses.