Andy Kessler is a techie.
Engineer, financial analyst and fund manager.
He is rich, successful and semi-retired, by the looks of it.
He also ran into the US health care system, a fairly gentle bump I must say, and now he wants to see the current system ended, while preferably making another billion in the process.
The End of Medicine, is an easy read, anecdotal, bit choppy, reading almost like a printout of a blog.
Names are dropped, jargon flung (and to be fair, explained), and pharmaceutical companies and medical doctors are savaged.
It is in some ways a satisfying read, Kessler identifies a number of the problems with the current system, including the obsession with drugs, the low hit rate and the slow rate of progress. But the central theme, as it must be, is the insurance reimbursement structure, the medical monopoly (why is an MD required to do or authorize some things) and the cost structure.
Kessler wants to change this. To do so he focuses on two of the big three - heart attacks and cancer (the third being stroke), and he searches for tech solutions - economies of scale, innovation and rapid obsolescence.
The good news is that he identifies both some of the solutions, and some of the problems - and I think he is broadly right, diagnosis and early intervention with targeted therapy, including allowance for individual response to drugs and other therapies is coming, fast and will revolutionize the system.
This will include more need for patient initiative, better economies of scale, some ruthless database mining and lots of new toys. This is good, interesting and provides warm gee-whiz fuzzies.
The book though is curiously incomplete, almost blinkered. For example, it is true that medicare coding does restrict procedures available to patients and delays introduction of innovative treatment.
But, insurance companies don't have to follow there blindly, they could approve their own charge codes for new treatment, and a good company maximizing profit rationally ought to be seeking to authorize preventative and early treatment which saves them money in the long run (of course one of the problems is that if you pay now to save in a decade, the patient may not longer be your customer).
Secondly, he fixates on mass market tech (ok, so that is the point...), but a lot of prevention and early diagnosis is low tech: in fact he highlights one such, the manual prostate exam. It is not as funky as a PSA detector, or a prostate cancer specific biomarker with 3D scanner, but it is effective and quick.
He also ignores the entire soft prevention sector, of physical therapy, home nursing, and simple personal advice (such a pre and post natal assitance and counseling) - sectors which the US handles badly and which can be highly effective in lowering costs and extending life.
Finally, he doesn't look ahead, in asking what we die of if the Big Three are cured, he fails to consider degenerative (cf Alzheimers) and autoimmune diseases (I'm thinking arthritis, not AIDS) - I'm sure there is money to be made there.
There will still be plenty to do for doctors, and not just PhD-MDs and counselors on WebMD; someone must still set bones, lacerate stitches, pop out appendices, clean arteries and repair muscle, operate on some of those microtumors we'll be detecting (I'd be surprised if all are best treated chemically, or at all, even with targeted personalized treatment).
Evolution does not rest, infectious disease is not done with us, although some aspects of it are well under control for now, with prospects for improvement.
Another issue is quality of life: pain management, asthma, allergies, mobility - these are another field where pharma makes much money, or the system has high cost, because of the reliance on maintenance drugs and long term management, rather than prevention and early intervention.
Another interesting area he does not consider, is automation of diagnosis - not scanning x-rays with neural networks, but building AI expert systems for automated diagnosis (ok a friend of mine, whose name I will not drop, is building some).
A lot of diagnosis is reducable to binary tree decision branches: you have mild fever, vomiting, lower back pain, but no headache and you are not dehydrated - that is a lot of bits to diagnose a finite number of options. Couple of more and you're probably done. Could be competitive with mass screening on biochips. Or more likely complementary.
With some caveats: there is always the very rare thing, much beloved of TV drama script writers, which is not always well caught by such systems - although they may do better than most MDs, since by definition most doctors do not have experience of very rare syndromes, whereas an AI system can draw on all experiences. Secondly, some diagnosis is probably not explicitly deteministic - ie doctors are diagnosing using information without consciously processing it - maybe skin tone, or moisture in eyes, or some low amplitude twitching, detected by the doctor but not consciously factored into diagnosis - such things may be key to final differential diagnosis by experienced doctors and hard to capture by AI algorithms, of course that is where neural networks excel.
Oh, and these things scale.
Anyway, I digress. So what about the book?
I'm glad I read it, it was interesting, thought provoking.
The author makes some good points, some are probably right, some are probably irrelevant, a couple may be completely wrong. We'll know in a few years which are which.
It will be interesting to see what happens when the techies from the Valley really focus their minds and energies on Other World problems, we're starting to see that happen, some good will come out of it, some will fall flat. Either way I want to see it coming to know what to look for.
- Log in to post comments
interesting to read your review. i too read this book, and really tried to bring myself to review it, but i found myself so angered by it i couldn't. it was insulting to me a doctor, and envisioned medicine through some Dr. Who, cold techie lenses as a world where everyone commits malpractice upon themselves and 'scanners' somehow are omniscient. sure, some medical illnesses are structural, but most are physiologic, hormonal, neoplastic, or psychosocial. his portrayal of doctors was beyond "savaging," to the point of unbelievable characatures. people should write about what they know instead of trying to foist their silicon valley experiences onto the behemoth of complexity that is healthcare.
dang it, there i go reviewing it... i'll stop.
With respect to expert systems for diagnosis, see Lawrence Weed's work (problem-knowledge couplers).
Apparently he is outspoken, and expert systems have met with resistance in the medical community. I suppose some doctors view the concept as threatening or insulting to their expertise.
But, as you say, the real purpose of these tools is to complement the diagnoses doctors, not compete with them. A computer system can suggest alternatives that the doctor may not have considered; the true human expert can take those suggestions into account, and is free to utilize or dismiss them as he/she sees fit.
Even in light of this, doctors are probably afraid of malpractice suits if the AI system suggests a possibility that turns out to be correct, and they ignore it. If they are solely responsible for making recommendations, second-guessing is more difficult. It's a shame, because systems like Weed's PKCs would most likely save lives.
There have been a few limited trials giving physicians handheld computers similar to Palm PDAs that monitor certain obvious situations, such as prescribing a medication when it is contraindicated by the patient's condition, but the problem with computerizing medicine is that it would require computers to be far more complex and human disease far simpler than either are in reality.
With regard to both Abi's comment and this book, it never ceases to amaze me how so many people seem to make the same fallacy in this arena: they actively presume that everyone involved in the American healthcare system must be a complete idiot.
They presume that if physicians would only do things their way, they could save more lives; if policy analysts would only consider their crackpot proposal, we could fix the entire system; they say that Medicare coding is too restrictive (never mind that there are almost 100,000 CPT codes), but they fail to offer a better method for reporting a procedure which may be performed millions of times across the country in a given year in such a way that the physicians can be reimbursed within a few weeks.
If a non-physicist claimed to have figured out GUT, people would be (rightly) skeptical, but when someone outside the healthcare field claims to be able to solve the major problems in that field, there is wayyyy too much credulity.
Hyperion:
Lawrence Weed is neither a crackpot nor a health care outsider. (From his bio: "Lawrence L. Weed received his MD from Columbia University. He has held positions at Yale, the University of Pennsylvania, Case Western Reserve, the Columbia branch of Bellevue Hospital, Walter Reed Army Medical Center, Johns Hopkins Hospital, and Eastern Maine General Hospital. In 1969, he joined the faculty at the University of Vermont. His research focus is primarily on information gathering in medicine, but he has also spent considerable time researching nucleic acids.") Most notably, he's the inventor of the popular POMR format for medical record keeping (and also the SOAP note).
The point of PKCs is not to "solve the major problems of medical diagnosis", but to provide a tool to counteract the documented underdiagnosis of rare disorders. As I mentioned, there are real social disincentives for adopting such tools. PKCs don't attempt to replace human doctors for diagnosis; they are only intended to suggest less common alternatives that the physician may not have considered. There is a literature which indicates that more common disorders are diagnosed in place of their less common alternatives more frequently than the actual relative prevalence of those disorders warrants, i.e., many physicians have an unconscious bias in favor of diagnosing the most common possibility. This does not imply that "everyone in the American healthcare system is an idiot".