Large Networks, Big Data, And Healthcare
Zander posted this NY Times opinion piece to usv.com yesterday and it's been rattling around in my head since then. The author suggests that big data is coming to health care and bringing with it many issues that will have to be resolved. I am sure that is so. But I also think the intersection of big data and health care and our large networks thesis is likely to produce some interesting investment opportunities for us and some valuable health care services for consumers.
Jason Karlawish, the author of the NY Times opinion piece, writes:
This is a revolutionary shift. Once upon a time, medicine was a discipline based on the nuanced diagnosis and treatment of sick patients. Now, Big Data, networked computers and a culture obsessed with knowing its numbers have moved medicine from the bedside to the desktop (or laptop). The art of medicine is becoming the science of an insurance actuary.
The question is who will control the input of the patient data, the aggregated data sets, and the results the data science produces. If the answer is the current healthcare system; the insurance companies, the hospitals, and the doctors, then we will have missed a big opportunity to reshape healthcare. If, on the other hand, the data is entered by patients, controlled by patients, and benefits patients, then we would have something new, different, and disruptive.
Large networks of patients coming together to do this data science together and benefit together feels like its around the corner and coming fast. Maybe some enterprising entepreneur will take this "Omnibus Risk Calculator" put it into a clean and simple web service, allow us to connect our phones and connected devices to it, and peer produce a service that we can, together, use to manage our cardiovascular health. Maybe someone has already done that.
Comments (Archived):
It’s happening, but slowly. The industry and FDA haven’t caught up. The opportunity in the future is in healthcare. There are a lot of giants that can be toppled, and a lot of “world” that software can eat. The current big software players in medicine (McKesson, Cerner, etc) are excellent at billing and keeping track of costs. They are really crappy for diagnosis. EMR’s (electronic medical records) are horrible to try and use. Software developers have told me that trying to use the software tools inside the vertical silos is amazingly hard-like dealing with Microsoft.The government is the largest impediment to innovation. In many cases, investors have to look at innovation far differently. For example, “failing fast” might mean two or three years of trials. One cancer drug out of Northwestern took three years, and $30M dollars. Then failed. The FDA is a centralized monolith. No doubt, innovators should mind their p’s and q’s, cross their t’s and dot their i’s but it’s gone too far.I have seen a lot of truly amazing things out of the University of Illinois engineering school. Sensors where you never would imagine-creating big data; x-rays that almost read themselves (no need for 600k per year radiologists).Changing medicine from a one size fits all to a decentralized personally prescribed regimen for you based on your age, overall health, and genetic structure will happen one day.
This area interests me.But is it really big data that we are talking about or data input at the most beginning level?Exercise, nutrition, environmental–these need to start at the input level from the ground up. And today, even at a labeling level for nutrition in food we are in the Pleistocene era at best.
Here is a start: insurancenewsnet.com/oartic… Simply having quicker online access to information should lower costs and make care more efficient.
Exercise, nutrition, environmental–these need to start at the input level from the ground up.Mass of Walmart shoppers are the boat anchors dragging us all down. Or the people who use food to medicate for the hard life they lead. To them eating poorly (comfort food) is the way they get by every day.It’s easy for either of us to be “mr big balls” since we are at the point where our life is in balance. So we can get enough sleep, eat well and lead a less stressful life that requires less “bad lifestyle” medication.Otoh there are many others that are in particular predicaments (due to circumstances beyond their control or poor choices they made or simply where or to whom they were born) that it is not that simple.
Sorry- don’t agree that being healthy is a privilege of the we’ll to do.
It’s not a privilege. Anyone can follow a healthy lifestyle.But looking and acting like a jackass [1] just seems to manifest itself much more at places like Walmart than at Nordstrom.Not that I haven’t see tubs of lard exiting the deli in Cherry Hill NJ. I have. To me it’s pathetic. But the “tubs of lard parts per million” does appear to be lower. Also the incidence of smoking. And drinking to oblivion.[1] Of course that is just probably my bias of what I find comfortable and acceptable. They no doubt think that I don’t fit their perception of “normal”.
Tubs of lard (leaf lard) are better for you than Crisco and Vegetable oil.
I’ve passed your name on to the “tubs of lard” human rights movement š
But it may well be a privilege of the well informed ?There are poor people who are motivated and doggedly well informed on health and nutrition but they may be exceptions?
The largest group of people are neither rich nor poor.And information has no class, There is no privilege of the well informed. That’s pure myth.
“The largest group of people are neither rich nor poor.”OK – when you put it that way, I guess I have to agree and stand corrected on the economic class correlate.Still I think there is a subclass of people whom for whatever reason are by nature curiosity or cultural influence drive to pursue information and organized it into reusable frameworks.
There are endless subclasses, each important and unique.There are always corner cases that fall off the cliff. But you build for the center and the best include the most and the most needed by definition not exception.
There is a lot of truth in what you are saying.
That is where the innovation needs to be. It would also be great to have someone who can explain this occuring disruption to the public, say on the level of Kaku.
x-rays that almost read themselves (no need for 600k per year radiologists).Oh that’s total absolute BS. Even teleradiology scans are typically read two times. By the way algorithims can’t correlate a patients particular complaint (that brought them in for the scan) with the thing on the xray or ct. Judgement is needed for that. Someone will have to review. You could increase productivity for sure but to say “no need” is simply not correct.By the way the reason radiologists make (and they don’t make nearly that much “typically” I was involved in negotiating a contract for one a few years ago) is because a constraint on the supply of radiologists. Supply in specialties ebbs and flows. A few years ago cardiologists were in higher supply (don’t know now what the case is with that..)
It’s not BS. I have seen a heart x-ray startup that allows the reading of the x-ray to go from experienced high payed physician to a lab tech, same results.Your supply argument rings true. The AMA restricts supply.
Some counterpoints and deals with aiding not replacing:http://pubs.rsna.org/doi/fu…http://www.medscape.com/vie…I also remember about 9 or 10 years ago there was a big discussion about this technology and I can’t even find a link to the equipment that at the time was attempting to be a solution of just helping the radiologist but in no way replacing. In radiology false positives are already a problem.What is the name of the startup you are referring to? A search doesn’t reveal any obvious suspects.
epsilon-imaging.com is one.
Ok I just looked at that. And I don’t see anything at all on the site that comes even close to not needing a physician to read the imaging. (Not to say that someone might get involved prior to that who isn’t an MD but that wasn’t your point.)The “about” page says:Applications assist clinicians to enhance, standardize, and streamline interpretation and reporting of echo studies, ultimately assisting clinicians to improve patient care and management. Initial applications include cardio oncology, stress echo and RV.Definition of clinician is:a doctor having direct contact with and responsibility for patients, rather than one involved with theoretical or laboratory studies.Lastly don’t forget the trial attorneys who would have a field day with any test (this serious) not being “signed off” by a physician. I would assume that is part of the reason that even teleradiology has a 2nd sign off.
we are in stage one of development when it comes to health stuff. Rome wasn’t built in a day. I also saw a machine somewhere that replaced an anethesiologist (sp) in surgery.Besides, if the headline of my website was “never use a doctor again”, how far would anyone get with it!
Great Sunday Am post Fred!I watched an interview that Rohan did w/ Albert earlier this year. During the interview , I forget what the setup question was..but the answer that Albert gave was talking about a health care app modeled after Wikipedia – where users could publish content about different aliments, treatments and outcomes.I am hoping for some big changes in the health care space – it is very ripe and much over due for disruption. Plus it is way too important to our economy and our general well being.I can not wait to watch Human Dx develop.
Careful. I know this is a pro social network forum, but remember not so long ago all the people who would go to the web and click the first hit thinking they were going to get a diagnosis resulting in a lot of problems.
Health Diagnosis is not done with a spear but with a net.
that Albert gave was talking about a health care app modeled after Wikipedia – where users could publish content about different aliments, treatments and outcomes.Luckily most people won’t use that. And the ones that do will just end up needing more anxiety medication.Anyway that tool already exists. It’s called “do a google search and see the shit that people are saying and try to figure out the shit that matters and is correct from the stuff that isn’t because it’s biased and anecdotal”.While I can see leaving no stone unturned if you have some rare disorder or disease, the majority of people don’t have some rare disorder that they need help with. They have this or that and there is already a decent treatment algorithm for the 99% of this or that.So the ability to gather data on the internet just leads to the “worried well” (cybercondria) which leads to anxiety.Ditto for any tool. I mean the idea that Joe six pack or even any of us is going to both need and use a tool like that is absurd. (In mass anyway..) By the way look at the people streaming through Walmart on Black Friday and see if any of those people are close to being disciplined enough or care about their health enough to use any tools. You know to figure out that smoking or overeating or drinking is bad.
There are already sites like this “Patients Like Me” and “Ben’s Friends” and many new disease specific sites.
So where are the early innovators in this field? We need to see more startup examples that are solving various pieces of this puzzle.Hackpad of up and coming healthcare startups that are doing some amazing things by giving consumers more data and empowerment?The role of the doctor will have to change for sure. The challenge is that medicine is part science, part art. There are many situations when it’s not black and white, and outcomes aren’t always predictable to the T. But if the consumer can operate below some base level of certainty, and then go to the doctor with a joint decision, aided by the data, then this could open-up new possibilities.My Dad is a doctor, and so was his father, and my uncle. I spent 10 years selling and marketing medical equipment to cardiologists, emergency, OR and ICU doctors in the US and Canada, so I’ve also seen the role of technology and medical information systems up close. I configured systems and trained doctors and nurses on how to use the technology.Healthcare data in the cloud and in the hands of consumers will bring a lot of change. And yes, keep it away from the insurance companies, gatekeepers and the healthcare systems administrators. They will look after their interests, not the patients.
Where is this data that the consumers need access to?To me its the lack of data that is the holdup.What am I missing?
Some of it will come from smart devices or manually entered. I think data and the tools to easily manipulate that data and make some sense from it is lacking.
Super curious and interested but i’ll play the other side.Just not seeing how the cloud is a solution as you imply. The idea that doctors who are paid by insurance to take xrays or tests will put this in the cloud and not make that info available to those who paid for it is a non starter.The disruption has to be way more fundamental to my mind.
I think the opportunity might be in a) preventative medicine – before you see the doctor, b) chronic illnesses management where you’re self-managing your condition.You are correct that the middle part where the doctor has to see you or prescribe exams and tests is a tough nut to crack, although there are examples like mentioned before, i.e. sending infant ear infection data to the doctor who can offer a diagnosis remotely.
Agree.I’m really interested in how this is going to push down and disrupt the food industry.You can’t have a preventative culture without nutrition and source information about the food we ingest.In order to do this you need to whack certification, labeling, transparency, and public funding for school meals in the head.That is a going to be a bloody and interesting fight.
I’m really interested in how this is going to push down and disrupt the food industry.It’s about time as the food industry has been disrupting the health industry for many decades now!It’s like a disruption tug-of-war.
My favorite comment of yours ever!
Food industry is almost as heavily regulated as health. The USDA favors industrialized food.
The USDA favors industrialized food.I think you’ve just made the case”that the food industry is regulating the USDA” š
yes, it’s like banking and the SEC/CFTC
I hear this. I know you are right. It’s simply not the reality that I see at the street level in NY.
There’s no need for it. If the regulators weren’t there people might actually have to pay attention to what they ate instead of trusting in the hot dog because some regulator approved it.
Regulation always favors the entrenched players and large corporations. They trade some margin for market share, the little guy is squeezed out and innovation slows.If tech were regulated like our banking industry we wouldn’t have start ups. How many people with capital out there are thinking about starting a new bank or insurance company? P2P lending is the closest thing to that, but only in so far as the companies engaging in it are just facilitators.
The food industry responds to market demands. It is not disrupting the health industry. Every day Americans make their own decisions about what to eat and what to do in their spare time.
That is nice in theory but the shareholder-market-place for higher profits is the first-mover market-place here.Only at some point much later does the negative health/$ consequences associated with that first-mover shareholder-profit market-place behaviour become significant obvious/painful to consumers. Then and only then does the consumer-market-place demand send the corrective feedback signal.This is not the first or last time that profit-market-place forces have disappeared down the myopic short-sighted profit rabbit-hole.This is not meant to villainize corporations or corporate profits but more as a statement that something is seriously faulty in the social contract by which we manage and integrate the various market forces.Data-mining is all about amplifying beneficial social/economic synchronicities, not about us vs them.I thing in the long run Big-Data, taken to it’s logical conclusion, inevitably leads to some form of “Organic Capitalism”. Some form of networked win-win profit synchronization. Here the hell else is there to go with all this?This is going to tough on the right vs left polemic crowd.
I only meant to point the finger at the general public rather than the those offering the trade. It’s always easier to isolate bogeymen than tp indict the will of the people. Although I’m an optimist by nature I’m less bullinsh on a big data panacea. For instance, we know empirically that smoking tobacco is bad for our health (though I am skeptical of second hand smoke studies… As I digress ) yet people still smoke and chew tobacco. You could argue that the risk is more quantifiable now, but I don’t think the risk has much to do with it. Hopefully vaporized nicotine or other innovations can deliver the hit without the big c, but either way it’s a counter argument to the notion that data will lead us to healthier decisions.On balance I think big data will deliver benefits across all sectors, including food. But I think targeted diagnotics and therapeutics emanating from genetic data will be more successful in improving health than big data around food choices. mind that big data will also be a tool to help NestlĆ©, Pepsi etc. to optimize their delivery of refined sugar to our blood stream.
I see the fitbit idea in here somehow…morphing to fitfood.
FitFoods Market
I would not be too hopeful for disruption here. Research indicates that the more Americans know about their food the more calories they consume. When the do gooders passed regulation for caloric count labeling it caused people to eat more.If you want to get to prevention then start charging people more for insurance (or better yet just actual healthcare) for being fat or smoking or sedentary. Instead we are subsidizing bad behavior, from the farmers, the bloated food stamp beneficiaries, the types of food available on food stamps (double stuffed oreos and Pepsi at the convenience store) straight through to subsidized meds.
I see another side of this.Disruption on the food side is happening right before our eyes.Awareness of nutrition and issues around the source of your food is everywhere.Availability of healthy food from the most expensive to the down home restaurants is everywhere across New York.The power of the market to force even the recalcitrant sometimes posers like Whole Foods to adopt GMO labeling, done.Consumer level class action to clarify what terms like RAW mean, done.Ain’t nothing perfect of course. And yes in New York so I”m a bit of a huge bubble.But this has nothing to do with hope. It has to do with work to change the status quo and it’s happening.
There isn’t anything new about healthy food. It existed before unhealthy food. Are there more food options? Of course. But I wouldn’t call that disruption, and for every more widely available vegetable there are a dozen new junk foods.Just to keep semantics in line, there is nothing healthier about organic and non GMO food, at least as far as science is concerned. You could make an argument about processed vs. not processed but that has nothing to do with organic or non GMO. The whole GMO and gluten free things are just fads. There may be something to the whole beef and poultry supply when it comes to hormones and antibiotics. I get my meats at Whole Food’s corporate HQ in Austin but that’s also largely because I think they taste better. Certainly much more expensive but worth it.Getting hipsters to pay for pink “Himalayan” salt with dubious health claims doesn’t strike me as disruption.Whole Foods voluntarily asking suppliers to label GMO content is more about capturing the fad than a disruption.In summary I don’t think voluntary or (eeek) forced labeling of foods is disruptive and I certainly don’t see how big data is going to disrupt the food industry. If it will I bet it will be more about targeting consumers for their biases to get them to pay more for “special” foods or mountain dew or whatever.
More than one way to look at everything for certain.We disagree at the very core on this one.
Sure. And anyone should have a right to eat what they want. And if enough people buy into a fad then someone will see a way to profit from it, maybe by labeling their non GMO foods as such.To date there have been zero peer reviewed studies that indicate that non GMO foods are healthier than their round up ready equivalents. So trying to pass regulation to label them either was is costly for no public good.Just wait until this whole gluten think boils over. You can’t swing a dead cat in austin without hitting 10 people touting how great it is to be gluten free. Often in between sips of beer.
I am as certain that you are wrong as you are that you are right.Soā¦good time to end this one.Time and the market are always the winner.
We can certainly agree to disagree and I will vigorously defend your right to avoid GMO foods. There are plenty of legitimate reasons to abstain, not the least of which are the business practices of the likes of Monsanto. I will however not agree to disagree that science has not found any ill health effects from GMO foods relative to non GMO. The facts are the facts on that.
Manual data entry is one of gee biggest reasons Google Health failed. “Some” will be a very small percentage.
You’re in a position to know that, given what your startup is doing.Trick is how do you bootstrap acquiring that data if it’s not coming from a clean medical record to start with?
There is no bootstrapping right now. It’s all schlepping. http://www.paulgraham.com/s…
The most innovative insurance companies will see the value of devices that automatically send health data to a patient’s file and offer them free to subscribers as long as they are regularly used.Problem is, I’m having trouble finding any innovation in the health care system as a whole. Just slightly different ways of cramming together the same ‘ol stuff.Who will be the [your favorite innovative company here] that will fundamentally change the way we think about health insurance?
I’m not sure that health insurance is where the innovation will come from. I was focused more on the “provision of healthcare” by the the professionals and by how we, as patients or health consumers can better manage our own.
Pervasive-Health.com is one that comes to mind. I also know of a lot of medical companies based here in the midwest. Healthy-txt.com. If you make a trip to Jumpsimulation.org, you will find the most cutting edge medical simulator in the United States with a full at home testing facility. There are a lot of ways to attack this. The lab, the OR, the medical office, and the home. Many perspectives: researchers, practitioners, patient, administrator.Most of the innovation has been targeted toward administration–>operations and billing. It also has the lowest bar for FDA approval (or no bar at all)In Chicago, we have two active organizations dedicated to startups. One is ibio/propel, the other is Chicago Innovation Mentors. Insightpd.com just started doing things with startups and briteseed.com is incubating there. There is a lot starting to percolate here.
Super info, thanks!I consider myself lucky that I have a great doctor, all digital records, email contact with the practitioner and a believe in wellness at a core.I’m with you, this is a great segment to put some effort towards.
How are you able to have all digital records (other than getting your paper records and scanning)?
Doctors provide them to me on request.Odd private distribution system.Baby steps though.
Moving forward the US government has put a focus on patients being able to receive their records electronically. One version of this is the “Blue Button” promise – which is effectively a CDA/CCD document. All EHRs that are certified for meaningful use stage 2 (most of them) must be able to provide a patients records electronically.This trend will continue and Stage 3 of Meaningful Use is expected to have a number of new provisions around releasing/engaging patients with their data.
Thanksā¦news to me.Sounds like promises for the future.For now, I simply refuse to use doctors that don’t provide this. The best do.
great list. I know you have been interested in this area Jeff. thanks
Look at the startups out of Rock Health or at the Health 2.0 conferences, or Startup Health.
thanks.
Hidden because of happy. You’re not supposed to share that data.
I may be doing some work for one of the most amazing healthcare start-ups that i’ve seen (although not in the networks area – more in the future of medical technology.) What i find to exciting is about how much it impacts the world. Life and world changing. Makes me reinvigorated to work in marketing (also coming from someone with a father, step father, two uncles all doctors).
Name?
Here’s the lead investor — he revolutionized breast cancer detection last go around and he’s doing it again with his start up synaptive medical now http://www.rdmag.com/articl…
Great to hear.I’m convinced there is a surplus of science knowledge that isn’t being efficiently applied. Once it starts to get applied better, a lot of possibilities will open up.
Hi. http://stayinyourprime.com
“moving from bedside to desktop or laptop” This info is mobile, live and destined to be on our phones/tablets or something we’re wearing.
Side note – Zander is a pretty cool nick name for someone called Alexander
It would seem to me that a major prerequisite is that government is on-board. With the laws in place around medical data, it would appear that they hold the “keys to the castle” in this instance.
you would say that about education as well but the reform has come around them not with them
the opportunity to create networks of data independent of healthcare incumbents is before people become patients.
Amen. Very few things need disrupting as badly as our healthcare system. It’s exciting to see things like what the Freelancers Union is doing around both care and coverage. With the move towards the “gig economy,” along with the unbundling of everything, conditions could be shaping up to make major change possible (please oh please).A search on Angel List yields over 600 medical/health related startups. I scanned a couple hundred and only two or three seemed to be engaged in disrupting the current system of information and coverage.
We’re disrupting it and I can tell you there are many more than two or three but it’s very difficult to get access to the data which is why there aren’t more of us. Data is being opened up slowly and innovation will ramp up.
Unfortunately the first effects of data collection will be to LIMIT healthcare. In healthcare , there will always be the DATA JUDGE (who is not necessarily your doctor). This is already happening. Lets say you are a medicare patient and have been admitted to a rehab facility for post knee sugery physical therapy. You progress and are able to walk 150 feet with a walker. If the data and its deciders determine that you don’t require additional PT, you are going home. Data and healthcare is a complicated issue.
Absolutely correct.
http://youtu.be/NrTM5FyPzl8There is a role for the patient and data. Here is what I see (version 1) as a first step.
This is interesting, Rich. It certainly attempts to address one important area: getting history and symptom information to the doctor. Filling out a couple of sheets at the doctor’s office leaves enormous gaps, and this process could allow for a more thorough communication of conditions.As one with way too much experience (not for my own health) in the area of doctors, appointments, costs, insurance aggravation and battles, billing corruption, medical incompetency, and indifference, I have a huge personal investment in this topic. For healthy individuals the system is pretty easy to manage, but for chronically ill or those with difficult to treat conditions or diagnose issues we are now about to go from difficult to near impossible thanks to the ignorance of the Obama administration. For those who work with specialists, often traveling significant distances to see them after searching them out for years, the changes in access to healthcare in the U.S. is going to get much worse for people who are already suffering. As you have said above, adding in Big Data provides yet another tool for insurers to restrict access and deny coverage for treatment. We already experience every several weeks. It will get worse.Sorry to take this tack, and I do think technology can make healthcare better (and it already is), but steps which empower the insurers, whether commercial companies or the gov’t, remove power from individuals. Therefore it’s very important to approach this topic with a great deal of knowledge, experience, exploration, research and compassion.
You are spot on with regards to nonacute ambulatory vs acute and subacute in-patient healthcare. These are two different worlds.
Actually, acute in-patient can be much better off than those suffering chronic illnesses like late-stage lyme disease, CFIDS (chronic fatigue immunodeficiency syndrome) or other similar garbage collection diagnoses, type 1 diabetes (insulin dependent) with complications of other issues or disease, idiopathic severe headaches…So in a sense, we have three (at least) classes to consider. And ironically, those who are too ill to battle insurance companies are the ones most abused by the systematic process of denying important coverage for improving quality of life, if not achieving a cure.
Giving control back to you the person the customer the patient — This is exactly why we’re building Prime: http://stayinyourprime.com
@tyler:disqus I tried to reply to your comment, but since your post was stuck in moderation I wasn’t allowed. I’ll paste it here.Thanks, Tyler. I’ll check it out. Good luck to you.—Wow, disqus is wacky today. Really weird behavior in Safari today during commenting.
I don’t see anything in here that deals with insurance. ZocDoc has that locked up. I’m not convinced that making a narrative is beneficial to me or my doctor. Seems like a lot of clicking around that can otherwise be addressed in a short conversation.
The research says otherwise. Everyone thinks that they do a great job communicating with their doctor, but its that drn doctor who just doesn’t listen. Turns out that the issue is the signal to noise ratio. Increase the signal and lessen the noise coming from the patient, only then can a doctor come up with a good and accurate diagnosis. There is some great research out there that has recorded Dr / Patient interaction during an office visit. Docs have just a few minutes to do fact finding. Their patients waste most of it. Garbage in, garbage out. Talking to your dr is like talking to a VC, you better have a 60 second PITCH.
I only have anecdotal evidence here so I’m going to take you at your word. Even if this is a challenge, how do you entice people to use this app? The only way I see it working is if you convince doctors to use it first, and to recommend their patients use it. It may be a problem and you may have a solution, but is it important enough for people to go use it?In my experience the biggest pain in the ass with the process is finding a good doctor, that is taking patients, that is near to me, and part of whatever arcane setup my insurance has around who I can see.As far as the standard interview, in my experience most people lie about their diet, sex life, alcohol, smoking, drug use etc. I wonder if the wearables space will help with that by providing direct and historic measurement .
What’s worse. Most of the medical studies are not sound from a statistical point of view. I’m hoping that the laws in Obamacare penalizing readmission would help this, but who knows
It’s medicare that’s calling the shots. You need to be in the weeds to make sense of things.
Right: We’re trying to move from not being able toget the statistics right one journal article at a timeto trying to get them right for one patient situationat a time. No chance. Totally messed up statistics.Maybe Karen Davis will be happy, but lots of people will die because of this nonsense.This is delicate surgery with dirty, rustytools. Maybe Nancy Pelosi dreams abouther ‘Obamacare’, but her dreams, in reality,will cause people to die. Nancy should belimited to organizing front lawn lemonadestands for charity drives, and even thenI wouldn’t drink her lemonade. Little Nancyhas her dreams to take over and ‘improve’20% of the US economy and one of themajor pillars in civilization with noreal background or understandingat all but just her dreams of security foreveryone from some fatherly Big Governmentin DC to “watch over me”. Did I mention,lots of people will die?Lenin, Hitler, Stalin, Mao and otherswith dreams of highly centralizedpower killed lots of people, butthe Obamacare crowd still hasn’t learned.That use of statistics is basically asking for a high form of ‘artificial intelligence’ and so faris just hopeless. We can’t yet see how toreplace a well trained and experienced physicianwith some ‘big data’, statistical methods, andsoftware, safely. Again, the promise is tokill people.E.g., “big data” itself is just some meaningless,media hype to have a ‘story’ to tell, right,from people who don’t know the differencebetween the weak and strong laws of largenumbers, believe in ‘the law of averages’,and don’t know about ‘uniformly minimumvariance unbiased estimators’, the Neyman-Pearson result, the Cramer-Rao lower bound, etc.Right: They can’t bring up a good Web site, butwe are going to turn over 20% of the economy andthe health of all of us to the incompetent fumble bumblers and their rusty tools in DC. Lots of dead people.Did I mention, they will kill many people?
So let me really go out on a limb here, Nancy Pelosi didn’t get your vote š Medicine is hardly an exact science and trying to make it so with statistics is fools gold. There are also many, many qualitative issues with medical care to rely too strongly on data sets, although I do believe that several countries strongly entrenched in social medicine do utilize statistics in prioritizing care for their overburdened system. (Hopefully that doesn’t begin to happen here in ’14 with so many citizens now eligible for affordable care). In sum, statistics are always open to interpretation and unquestionably they are most useful (and effective) when combined with sound judgment, which is critical for medical evaluation and care.
> So let me really go out on a limb here, NancyPelosi didn’t get your vote :)Now, now, of course I didn’t vote for Nancy. I’m inNYS! Of course, if I were in SF then why would Inot vote for ditsy Nancy? What would I have againstwildly overly emotional, irrational, dangerous,uninformed, misinformed, just plain wrong,irresponsible, confused, frustrated, blindlyambitious, did I mention ditsy?, bimboism? I mean,only her hairdresser and plastic surgeon know forsure, right? Can also toss in Karen Davis of theacademic, economic systems analysis, optimal healthcare planning research scam — I’ve been too closeto that junk, and it reeks. And we can count inDead Fish the Lesser, Ezekiel J. Emanuel, the oldSenator Kennedy health care planning staff (ah, allthose evenings smoking funny stuff with the hero ofChappaquiddick and of all fathers of gullible,naive, lonely, vulnerable daughters himself),associated scam artists, Hillary herself, etc.Fred might be able to find an investment that makesmoney in something related to health care and thatclaims to be using ‘big data’, etc. That whole’investment space’ looks to me like a muddy roadfull of potholes, rotting logs, and assorted deadanimals with a fetid swamp on either side. Goodplace to be, in the winter, with a shot gun withbird shot, so that at the end of the road can dosome duck hunting.Broadly the Obamacare insider crowd is dreaming ofgetting their hands on US health care, 20% of the USeconomy, and running it their way, right down to theone on one patient examining rooms, from big officebuildings around the DC beltway.In particular, one of their main dreams is to have’computers’ replace, dictate, guide, or constrainmuch of the work of physicians meeting patients oneon one in examining rooms. For this littleintellectual self-abusive wet daydream of theirs (amI making myself clear enough?) they want to use thegrand power, so far yet to pass safety, efficacy, orcommon sense, of ‘big data’ and ‘statistics’ wherethey don’t really know what either is. Of course,no one knows what ‘big data’ is except a lot of hypeon labels of snake oil, but the field of statisticsis well documented on the shelves of the researchlibraries. Documented? Yes. Widely understood?No!What the wet daydreams assume is that, with theunboundedly powerful magic of (the so far undefined)’big data’, that we might as well call the ‘bigscam’, and some ‘statistics’, replacing a lot of thework of those ‘overpaid’ physicians will be just ‘asmall matter of software’. Have we ever heard thislast before? Right: A few million lines ofhandmade, one line at a time, manual bench work!However, there are some problems here, somewhatbasic and unsolved. Broadly statistics is no moreup to the need than a three foot ash paddle willmove a US Navy aircraft carrier. Why? In a fewwords, the crucial information is buried in aforbiddingly astronomically huge combinatorialexplosion, that is, definitely one of the worstvarieties of ‘explosion’.Yes, yes, yes, I know; I know: There is a goodall-purpose solution — cross tabulation. It’spowerful stuff if only because it is the discreteversion of conditional distribution (from theRadon-Nikodym theorem — nice proof from von Neumannin Rudin’s ‘Real and Complex Analysis’) with niceapplications in any of the books on probability byJ. Neveu, L. Breiman, K. Chung, M._LoĆØve, among someothers and the non-linear, unbiased, minimumvariance estimator — simple proof, basicallymanipulate a quadratic much as in high school, leftas an exercise; let me know if you need hints. Themain thing you need to show is for the real numbersR and any Borel measurable f: R –> RE[(Y – E[Y|X])^2] <= E[(Y – f(X))^2].Do need to know that E[E{Y|X]] = E[Y] which is alsothe unbiased part.So, take a system with, say, 1,000,000 parts andmeasurements of each part. Now the uniform,universal way to apply cross-tabulation is toconsider all 2^1,000,000 subsets of the parts andtheir measurements.So, that’s our ‘big data’, right? We get that in ayear or so from typing in patient examining rooms,right? Right, Virginia, Nancy, Karen, and the restof the inside Obamacare crowd, right? Nope:Haven’t got even a few molecules in a small drop ofall the oceans in the galaxy — for a roughestimate. No bigger wet daydream.So, how do physicians do it? Well, they essentiallydon’t use statistics.Statistics can be great stuff in some circumstances– I’ve published peer-reviewed original research insome applicable mathematical statistics. But, did Imention “some circumstances”?The artificial intelligence community ran into muchthe same problem. Eventually they might haveconcluded that they were looking at writing separatecode for the combinatorial explosion of 2^1,000,000subsets. There they formulated the need for ‘deepknowledge’. They weren’t sure just what that was,but it was certainly ‘deeper’ than the ‘knowledge’they were working with.Yes, I’ve published peer-reviewed original researchin such artificial intelligence.So, how does anything work? Well, in ‘intelligence’there are some things that do work well, e.g., justin the kitty cat currently sleeping on my chest.He’s no dummy, that is, far, Far, FAR ahead of any’intelligent’ (cough, cough, upchuck) computer.My other kitty cat, the stray, learned in a bighurry that if I start to put on latex gloves, evenjust from the sound, he should head to the basementand hide and does. He gets ‘causality’ in a bighurry, without statistics.A better example is my auto mechanic or any goodauto mechanic. So, how do they do it? Sure: For afirst cut explanation, they have a lot of what theartificial intelligence community wanted to call’deep knowledge’; that is, in this case such amechanic actually understands how the heck a carworks. So there are some thousands of parts, andthey know a lot about the parts and, much moreimportantly, how they do and don’t interact.Heck, even computer nerd me could figure out thatwhy my 4 wheel drive SUV couldn’t climb out of mydriveway in the winter was a leak in a vacuum line.That vacuum was throttled by a little mechanicalswitch on the side of the transfer case and pulledon a vacuum diaphragm which pulled on a cable whichpulled on a pin in the front differential case,which is not really a differential since it has nospider gears. But the pin moved a sleeve whichconnected with a spline which made the front wheelsreceive power from the drive shaft from the transfercase. So, my temporary solution was to get aC-clamp and move the diaphragm, get out of thedriveway, remove the clamp, and then drive on. Thelesson here? Know how the heck the thing works.So, just think about it: Auto mechanics repair ahuge variety of problems, but most of these theyhave never seen before. So, ‘statistics’ wouldn’twork. So, how can the mechanics do the work?Because they know how the car works and use that.So, just program how the human body works, right?Thank you Nancy; by the way, how are you coming withthose front yard lemonade stands for your favoritecharities? Gotten any customers sick yet?No one but nobody knows how to program such a thing.And, whatever, it’s not ‘statistics’.Or, ‘statistics’ essentially wants the ‘big data’ tocontain close enough example cases with symptoms,treatments, and outcomes, etc. so that for one onone medicine can just look up a similar case andpick the treatment that worked well. Alas, evenauto mechanics can’t do that. Not even if GM, Ford,Toyota, etc. collected all the repair data theycouldn’t. Did I mention a combinatorial explosion?Or, with statistics, each model year that computerapproach would have to start collecting data allover again. Why? New cars. And the statisticsdoesn’t know what in the new cars changes oldknowledge and what doesn’t because the statisticsdoesn’t know how the car works. E.g., thestatistics doesn’t know that the new Bluetoothconnection has nothing to do with the line to therear emergency brake. All or nearly all aboutbrakes, ball joints, the oil filter is not affectedby the Bluetooth. And some new, fancy automatedmonitoring with a cell phone connection back to GMlikely really doesn’t change much in the ring andpinion setup or even the disk brakes.But, yes, I know, I know: The Obamacare crowd wantsto use ‘regression analysis’. Sorry, Nancy: That’sjust what try to do with a lot of assumptions andnot enough data to use cross tabulation. And, sorryNancy, you don’t have either the assumptions or thedata.What a bright, long experienced, Board certifiedphysician has is real ‘intelligence’, and gotta tellyou that it will be a very long way until softwarecan catch up with just my kitty cat and much lesswith such a physician.What’s going to happen is that Nancy, etc. will,from their ambition, ignorance, incompetence, etc.kill a lot of people, and then voters will seeObamacare for what it is — wet daydreams that killpeople from people who want their hands on 20% ofthe US economy.Then voters will slowly come to appreciate that USmedicine is a top, center crown jewel ofcivilization where computers have only meager rolesto play and no roles to play in the real, crucialcore of medicine where a physician works.Or, once we actually know how to do some work, andcan describe the work in precise steps, then oftenwe can program a computer to do the steps. But weneed to start with knowing how to do the work atleast in principle. Really, ‘computer science’isn’t much help for this work of the precise stepsbefore the software; that is, just because we wantthe work to be done by a computer does NOT mean thatcomputer science is the best way to say how to writethe software. That is, if we want to use computersto automate making pizza, we still have to come tothe project knowing how to make pizza, and computerscience won’t help here. Sometimes some statisticscan help. But when we don’t know, statistics isn’tmuch help, and we can’t yet really programintelligence, then we are, in the immortal words ofGroucho, “stuck-o”.Now, how many people will Nancy and the rest of theObamacare insiders kill before the public does a bigupchuck? We have several early warnings, a recentone being the failure of the Obamacare Web site.Other warnings are that everyone who can get out ofthe scope of Obamacare is rushing to do so.Likely Obamacare will die of its own weight fromvarious failings before it kills millions of people,although horror stories are coming already. Then,as soon as Obama is out of office, Obamacare can beput out of its misery; it will have served itspurpose, that is, political nonsense. That is, runfor elective office on the great promise ofObamacare, have it go slow while in office, and letit drop when out of office. We can call that a’scam’, right?
I wrote this for you to see, but it got moved, probably due to AVC’s setting for too deeply nested Disqus comments. You are absolutely correct here. I was chief statistician at AZ Dept Health Services, Office for Children with Special Healthcare Needs. My father was a cardiologist, uncle is a rheumatologist. I know what I’m talking about, when I say that you are correct! Who ARE you? I have followed you on Disqus since Jan 2011! Do you have a website or Twitter or such? I want to read more of what you write about. Please!
Thanks.I’m basically variety of applied mathematician. Attimes some of my math is for some problems incomputing in which case the work can be called’computer science’. And I need computing to do thedata manipulations specified by my work in math.Sadly, math didn’t help me understand girls, but afew times some victories I had in math, e.g., beatmy school’s “Most Intellectual” in a shootout at theboard in trig class, got me a little interest fromone very pretty and nice girl! So, in grades 9-12,I discovered that I could do math and did.On the Math SAT, three guys came in 1-2-3 with onlya few points between 1 and 3, with “MostIntellectual” 3, me at 2, and a guy that went toPurdue 1. Yes, I’m an agnostic (later anantagonistic), but 1 and 3 were both Jewish; theschool was public but by far the best in the cityand, apparently, okay or better on a national level– MIT came recruiting. So, right, the Jewish kidsknew what school to go to, and so did my dad — hehad an exceptionally good background in educationand looked for the school first and the housesecond.In college I wanted to major in physics but alsowanted to major in math as a background for the mathin physics. I ended up with just a math major andone course shy of a physics major.Sadly too much of the math was a bit too ‘pure’ forphysics, but there can be a point to such math. So,there was too much emphasis on abstract algebra;e.g., my honors paper was on group representations.Yes, E. Wigner used that in molecular spectroscopy,but all I did was the algebra.I did an NSF summer at Vanderbilt in, gads,axiomatic set theory — e.g., see the picture Iposted at http://www.avc.com/a_vc/201…where I show that there are infinitely manyinfinities, basically standard stuff in axiomaticset theory.As a math grad student at Indiana University, Istill wanted math for physics, but the departmentwasn’t much interested in that. In 2 out of 3courses they wanted me to take, I’d already coverednearly all the content in undergraduate school. Inthe third course, the prof got on my case, said mysolution was no good. I showed him that thenotation I’d used was standard in axiomatic settheory at which time he saw that my solution was onestep shorter than his and, thus, better. After fewmore such clashes I gave up on the department,started violin, met my wife, and left.Around DC I started a non-academic direction inapplied math. So I was drinking from a fire hose inoptimization, numerical linear algebra, essentiallyall of applied statistics and associated softwarepackages, a lot in signal processing, especiallysurrounding the fast Fourier transform, and more.E.g., there was a nice seafood bar in Silver Spring,and I took Blackman and Tukey, ‘The Measurement ofPower Spectra’ and read it while eating broiledflounder, French fries, coleslaw and drinking beerfrom small glasses. One scary dinner the guy nextto me saw the book and said, “I bet you work for theNavy”. Gads! I was being recruited by a Russianspy? So I didn’t answer him!I’d run into Tukey’s work in axiomatic set theory,uniformity in topology, stepwise regression,exploratory data analysis, the fast Fouriertransform, and measurement of power spectra. So,when I went for a Ph.D., I wrote Tukey at Princeton,mentioned that in my work I kept running into hiswork, and wrote him asking if there was a future insuch work and, maybe, such an education atPrinceton. G. Watson got my letter, said that theyhad the best program, and I got accepted toPrinceton, alas, only as a grad student!I went to Johns Hopkins instead because my wife wasalready there, we lived close, and Hopkins had someof what I wanted, e.g., optimization. Whyoptimization? Because I’d written software toschedule the fleet at FedEx and wanted somethingbetter, thus, was trying to attack some problems incombinatorial optimization.I got a Ph.D. in engineering from Johns Hopkins fromsome work in stochastic optimal control. The bestof the coursework at Hopkins was terrific stuff.When my wife’s Ph.D. was delayed, I dropped out andtook a job to support us until she finished, and thenI quickly finished. The job was more Navy work, inoptimization and stochastic processes. At onepoint, a rush project in two weeks, I wrote somesoftware and did an analysis of the survivability ofthe US SSBN fleet under a special scenario of globalnuclear war limited to sea. There was a continuoustime, discrete state space Markov processsubordinated to a Poisson process from some B.Koopman work. J. Keilson, ‘Green’s Function Methodsin Probability Theory’, did a technical review andpassed my work. I thought that my solution, neededin just two weeks, was something of a joke, but thework was taken seriously. Apparently a leading USintelligence agency bought it; I could name theagency, but then I’d have to …!The Navy specified the two weeks, and my wifeagreed: She had reservations for us for a vacationat Shenandoah. When I took some math books alongfor light reading, she was offended! No, mathhasn’t done much for my understanding of girls!Another time, there was a NASA question involvingtwo communications satellites and how to assignsignals to the satellites to minimize the worst caseof interference between signals. Later I noticedthat my solution was to the ‘bottleneck assignment’problem. A guy submitted the work to an integerlinear programming package with branch and bound andwas pleased at the fast solution. I had to tell himthe branch and bound didn’t get used because thatproblem was a network flow problem, and given aninteger solution each simplex iteration will returnan integer solution and eventually an optimalinteger solution.Another time there was a question in stochasticprocesses. Where the process is at time t might notbe defined because may with positive probabilityalready have gone to infinity at time t. So, I hadto tell them that they had a ‘non-regular’ Markovprocess. There is deep work there with the Martinboundary.My wife’s Ph.D. was in essentially mathematicalsociology, and there I got to discuss with herissues of ‘causality’, construction of statisticalmodels, statistical software, etc. Some of thediscussions were not on a Friday evening over apizza with a pitcher of beer! She encountered along, fatal illness.I’ve continued with applied math in various forms.At one point a guy had a marketing problem, hadformulated a 0-1 integer linear program with600,000+ variables and 40,000+ constraints, tried a’genetic algorithm’, ran for days, and gave up. Ifound a solution via Lagrangian relaxation, wrotesome software, and got a feasible solution withing0.025% of optimality in 905 seconds on a 90 MHz PC.The crucial core of my startup, a Web site, is someapplied math I derived; the site users will neversee the math, but the math is just crucial to theresults they will see, hopefully much better resultsthan available otherwise.I have no Twitter account (sorry Fred) and no blog,not even with Disqus (sorry Fred).
I’m sorry that your wife died. My husband died when he was only 45, then my father, who was my only relative who was close to us, who cared. I miss them both so much, it is hard to describe. We were only married a few years, no children. I wish I had spent my time more wisely.I first found Fred through his photographs on Flickr. His pictures are of things that I remember when I was growing up. I love IBM mainframes, system Z, APL, RACF. They make me feel young again, like Fred’s photographs! I worked for IBM, right after finishing school. I modelled DASD and other storage product performance. Queueing is fun! M/M/1.My background is similar to yours, except yours is better academically and you’ve accomplished more, and you’re probably brighter. I went to Swarthmore College for math (I didn’t know where to go to school, as I lived in New Mexico). I probably would have been happier at NMSU which has a wonderful engineering school. Then Stanford for Operations Research. I understand what you are referring to in your lengthy, kind message. Thank you, by the way! You would have done well with HPC. Fortran is still used a lot for that! Do you code in C?I read one of your old Disqus posts that described exactly what I did for United Airlines. It was an excellent description of airline inventory management using either integer or dynamic programming, I need to double check. I liked that a lot! I also recall a Disqus comment thread, maybe in December 2012 or even 2011, where Marc Andreessen was asking you and a few others about a “magic bullet”, a single web analytics metric. He owns Mixpanel (and lots more). Anyway, it was an amazing example of some exceptionally bright people and entrepreneurs coalescing in the comments section of an AllThingsDigital post. You did well!Do you like StackExchange a.k.a. StackOverflow? That is one of the very few examples of this soi disant sharing economy that actually has high quality content based on a meritocracy. Its not without its flaws, but overall, not much besides Microsoft Developer Network MSDN and StackExchange have demonstrated any genuine value, where none are exploited and there is a viable business model! I am worried that Fred is buying into the “total transparency” thing. No privacy in business or personal life is just as repressive as censorship! I am not a communist, nor fascist! I get called both, at times. I have an MBA from Wharton and was a proprietary trader for a few years, and happily worked at S&P until Sept 2001.I found a great article for Fred. I hope this doesn’t cause me to get banned from AVC. As I mentioned, I am worried about Fred and big data (not to mention bitcoin…). Big data is not the be all and end all solution! Anyway, this http://www.healthcatalyst.c… is perfect because it is written by a healthcare data services provider, but is remarkably honest about the (ir)relevancy of large amounts of data unless one knows what to do with it, or even has a practical application for it. There’s a great chart, which I’ll post here! It is hilarious, look at the labels on the x and y axes!http://www.healthcatalyst.c…Please consider getting a Twitter account, or giving me your email address so I can send you emails with math irony and humor?
> I’m sorry that your wife died.Thanks. That she died has long been well known hereat AVC, and the community was very good about it.> Do you code in C?At times I’ve had to and did. But the language istoo primitive for me. Also I don’t like the’idiosyncratic’ syntax; e.g., Kernighan and Ritchiehave their little example parsing one of theirdeclare statements. I was able to write all thedeclare statements I needed, but I was never able toparse all that could be written.So that I can get my work done more quickly, I wanta language that has more features. So, on IBM’smainframes I was big on PL/I. Now I’m writing inMicrosoft’s Visual Basic .NET (VB) and there takingadvantage of the .NET Framework and the ‘commonlanguage runtime’ (CLR). VB looks nicer to me thanPL/I which, of course, is much older.I used IBM mainframes, but I didn’t like them. Myfirst complaint: They never bothered to have a goodhierarchical file system, one of my favorite things.Then there was the bureaucracy problem: Once theIBM people on DB/2 asked me how I liked theirsoftware, and I had to say “How can I like it? Dueto the bureaucracy I can’t get at it, touch it, useit and can’t create even one table with one row andone column.”, which was true. And I was in YorktownHeights designing a high performance, ‘active,object repository’ later shipped by IBM as part ofNetView.Once I was in a selling competition, with me pickingmy favorite super-mini computer from DEC, DG, andPrime against IBM’s best. The local IBM branchoffice worked hard, and they flew in theirsuper salesman Buck Rodgers twice. I won, with aPrime, which was a nice system.The Prime architecture borrowed a lot from Multics.Since the Intel 286 and 386 architectures alsoborrowed a lot from Prime and Multics, the Primesoftware could have been ported to 286 and 386 withsurprising ease which would have given Prime by farthe nicest OS on Intel chips years ahead of anyoneelse.I do like IBM’s XEDIT and Rexx. Significantly, bothwere done by IBM employees mostly on their own andwithout ‘official’ IBM management support. For VM,it was done at the IBM Cambridge Scientific Centeras an aid to writing operating systems; later itbecame a pillar of IBM’s product line and, now,nearly all high end server farms. Too often theofficial IBM management was poor at pickingprojects.By far my most heavily used software is KEdit whichis a PC version of IBM’s XEDIT. Yes KEdit uses aversion of Rexx as its macro language, and I haveabout 175 macros in my central collection and writenew ones as the need arises. E.g., last weekend Iwrote beidc, which abbreviates ‘begin, end, insertdated comment’ and results in, say,BEGIN Modified at 22:26:01 on Tuesday, December 3rd, 2013.END Modified at 22:26:01 on Tuesday, December 3rd, 2013.Last weekend I put a lot of those pairs in some codeand put code changes and/or comments between thepair. Maybe I will write another macro which givenone of these two on the ‘current line’ will find theother one and make it the current line.For spell checking, I use the amazing ASpell fromsome TeX distributions. That way I have only oneaddendum dictionary for essentially all my work.> integer or dynamic programmingNot many people even know what both are.Congratulations.> Marc Andreessen was asking you and a few othersabout a “magic bullet”, a single web analyticsmetric.Thanks. I found that in my file system directoryH:data05topicsmediawsjBTW, on my system there to find this directory don’tneed the H because there is an environment variablethat says that the subtree data05 is on drive letterH. So I could tree copy all of data05 to, say,drive letter J, tweak the environment variable, andmost of my software would still work. So, I get’drive letter independence’. I’ve been using thislittle trick back to PC/DOS although I understandthat Microsoft has a way to assign drive letters todisk partitions as one wishes; so far I’ve neverused that.Apparently Marc was not trying to do well thinkingabout ‘Web analytics’ but was just getting publicityfor one of his investments and attention for hisfirm to generate ‘deal flow’. I wasted timeresponding to him.Nerds, that is, people working in fields of highconceptual precision such as math, physics, andcomputing, can be too fast to take statementsliterally and at face value and too slow to detect ahidden agenda and a manipulation. When trying tosolve a math problem, the problem is not fightingback; humans do. So, when working with humans, haveto understand that there are two ‘sides’ to theproblem and not just one as got used to in, say,math class.> Do you like StackExchange a.k.a. StackOverflow?… MSDN?Yes, both. I mostly don’t try to contribute, butthey are amazingly good for answering questions.E.g., early using VB I had a ‘polymorphic’ sortroutine running for 600 seconds, put a question onMSDN or one of the Microsoft fora, and a Microsoftguy gave me enough on how to write polymorphic codefor me to get the execution time down to maybe 7seconds. In my first version, I was in effectasking VB to do me some extraordinary favors I wouldnever have asked had I known.As far as I can tell, VB is about as powerful as C#,say, at using the CLR and .NET. Since C# tries toborrow from the old C syntax, I prefer the moreverbose and easier to read VB syntax. Gee, my codehas declare statements, if-then-else, do-while,try-catch, the given VB statements, call a functionand get a return code, and little else. Simple, andso far in ‘computer science’ I see no great reasonfor more complicated programming language syntax orsemantics. Programming is still one line at a time,hand made, bench work, and, thus, for a largesoftware project like digging the Panama Canal witha tea spoon. Progress should be possible,Last night it appeared that i_partition isessentially a reserved variable name declared insome SQL Server DLL (‘namespace’); this is likely aMicrosoft bug; there’s a good chance I’ll find howto solve the problem at Stackoverflow or MSDN.To repeat myself, to me ‘big data’ is a lot of hypewithout even a clean definition. As I learned longago, usually don’t have to have terabytes of data toget powerful law of large numbers effects instatistical estimation.> There’s a great chart, which I’ll post here! It ishilarious, look at the labels on the x and y axes!The horizontal axis looks like years; the meanings ofthe vertical axis and the graph are mysteries.The colors have been selected so that I have a toughtime reading the page. Like about 25% of men, I’mpartially red-green color blind. More peopleputting graphs on the Internet need to understandthis 25% figure.When I do a graph, I like to make it easy to see,including on a black and white printer. E.g. theattached’Generic Detector Comparisons”Growth Scenarios’, one of which, at one time, keptFedEx from going out of business,The graph is solutions to the initial value problemfor the differential equationy'(t) = k y(t) (b – y(t))for b = 100, y(0) = 15, and various values of k.The vertical axis shows y(t), and the horizontalaxis shows time, t.Get the equation from assuming that growth isproportional to the number of current customerstalking about the business and the number ofpotential customers listening. So, it’s a fairlysimple but general ‘viral’ growth model andgenerates lazy S curves that rise asymtotically tob.afs_heap_sort02, performance of a sort routine.> Please consider getting a Twitter account, orgiving me your email address so I can send youemails with math irony and humor?I will likely have to get a Twitter accounteventually.For now, for e-mail can take ‘sigmaalgebra’ andchange ‘algebra’ to ‘waite’ and then append’@optonline.net’.
Thank you! Yes, I learned about dynamic programming, constrained optimization etc. when I got my M.S. in Operations Research at Stanford and worked at IBM GPD San Jose. Simplex method, traveling salesman, networks, none of that was of much interest to anyone except defense contractors e.g. TRW (not the consumer credit agency) in the early 1990’s. This is me http://linkedin.com/in/4lisa I made my URL, as 4 is my favorite number.The chart is humorous because it shows that the number of peer-reviewed journal articles or perhaps journals (about using quantitative methods to predict the rate of hospital readmissions) has increased enormously over the past 13 years. Yet they haven’t even come close to replacing retrospective reviews, sampled cases, by clinicians, i.e. registered nurses and M.D.’s The latter is how decisions are made now, about whether or not there are deficiencies in quality of care, as gauged by the number of readmissions for the same conditions. It is musing because it is showing that despite increases in computational power and human effort expended on the issue, we are not getting closer to replacing people for that critical role. Sadly, the ruling not-Democrats i.e. Hilary, want to get rid of all of that. Yet the article was written by a healthcare IT professional services provider, not a detractor, as a warning about simplistically placing faith in technology to solve problems such that operational and labor expenses approach zero.I wondered about what Marc Andreessen was up to, that day! I even tried to leave a comment on Disqus, as he was getting very valuable input from people that he should have paid for their time to give him feedback on his idea! I wondered why, not just you but many others, were so freely telling him so much, and in written form in a public venue. I even verified that it was genuinely Marc Andreessen. It was. That confirms a lot of the rumor based reputational cruft that I have run across about him. I am sorry he wasted your time. He should have paid good money to all of you. Several months later, it was erased. I am impressed how you save and can retrieve your writing and comments so well. I want, need to do that. Fred is the ONLY venture capital person that I would never criticize, well not harshly, even though he has said it is okay to speak freely, because he is the ONLY one of that crowd to dispense good advice to others, unsolicited, freely, that is directly relevant to individuals, not mere generic platitudes. That’s why I’m scared. Fred is sensible, savvy in business and a good man. He is like Morgan Guaranty Trust, well, like it was. I have said more than enough…Thank you for your email address. I am Mrs. Lisa E. Wells nee Kesselman. L.E. = Ellie. It is a pleasure to make your acquaintance!
Hello, Sigma! Yes, “a bright, experienced, Board-certified physician has real ‘intelligence'”. Obama-Hillary want to replace with a low cost, shoddy facsimile, for us (the plebescite masses). They think big data and statistical analysis are the means to realize the goal. They are wrong.
Ellie, “low cost, shoddy facsimile”. Howpejorative, provocative, critical, unfeeling!I’m “shocked, shocked to find that” shoddy work isgoing on there! “Here are your” obituaries, Sir.But, but, but, if Hillary, Nancy, Karen, and Obama,etc. could just take over 20% of the US economy, thehealth care of all of us, and one of the top centercrown jewels of civilization, and run everythingtheir way, and they KNOW that they could do a goodjob because they so much want a good job done, thenthink how much ‘better’ everything would be????I mean, to Nancy, Hillary, and Obama, the ACA justsounds and ‘feels’ so right!I mean, I mean, I mean, we have good evidence: Theydid actually bring up a Web site, right? What aboutUS health care could be harder than that?Maybe they might, I mean just for some of theskeptical, ACA detractors, do a few sample projects,say, run some lemonade stands, pass a test inundergraduate mathematical statistics, run an animalshelter, make a B or better in undergraduate organicchemistry, explain the difference between DNA andRNA, ‘help out’ in a trauma center, assist insurgery for colon cancer, read an EKG of a guyneeding a quadruple bypass, ride with an EMS for aweek, ‘help out’ in an ER with a 13 year old girlhemorrhaging from her attempt at an abortion with acoat hanger, I mean to allay fears that they are notjust charlatans, scam artists, fumble bumblers,incompetent boobs, dangerous, irresponsible,unscrupulous, political opportunists, etc., okay?
Healthcare program performance includes 1. fraud/abuse detection by providers, patients, claims processing; 2. clinical care evaluation.Big data, or better coding skills (SAS on a mainframe :o) makes 1, profiling for data anomalies and accounting irregularities, easier.@Shana You said that ACA would make things better by doing 2. That isn’t new though! We tracked hospital readmission frequency, stratified by doctor/caregiver, diagnosis ICD9, facility. And DME durable medical equipment e.g. is it poorly made, or patients don’t/can’t use it correctly. This isn’t innovative; Medicaid has done this for 5+ years. Of course ACA should do it!Big data and apps won’t help. To do 2., a clinician, usually a nurse w/10+ yrs work experience, must assess it, sometimes w/ input by an MD. That subject matter knowledge can’t be acquired with apps, nor big data e.g. trained machine learning.
Limiting has been the scenario for he last 50 years.Disruption is already underway and starting to show lots of non-limiting value. Rock Health is a great example.
It is, see my app below.
Actually, having some experience working in such a setting, Id have to agree with Fred. More often than not the patient is able to walk 150′ with a walker and little assistance within 10 days but the facility is able to get 20 days of care from the patient and will document whatever they need to to maximize length of stay. Don’t be surprised, it happens a lot. If patients had data on the care received from these facilities, from other patients, and if hospitals provided that data to patients prior to their discharge, that would help enormously. Also, on a side note Fred, the folks at Symcat have taken a stab at the CV risk calculator, worth checking out: http://blog.symcat.com/post…
Not sure what your point is? Are there patients trying to get out of rehab facilities after day 10? Of course not. Are there rehab centers trying to make the patient data fit into the requirenmemts for a 20 day stay? Of course they are. Problemwith subacute nursing assited rehab is that what data is useful for the patient? They feed you, care for you and provide physical and occupational therapy. Is getting out quickly a good data point or a bad data point?
Actually, my point is that there are many patients who do want to get out as soon as possible – they want to be at home, not in a rehab hospital. And there are surgeons who want their patients home as soon as possible as well (recently they have been pushing for 5 day stays). The longer you stay in such a facility, which will be mixed with patients recovering from pneumonia, patients with CDiff, etc the more likely you as a patient are going to catch something eventually. Getting home sooner, for joint replacements at least, is usually the best possible outcome. It’s what patients want and it’s what their surgeons want and there may be exceptions where that is not an optimal outcome, but by and large it is.
that’s awesome. i just took the risk calculator. i have a 5.7% chance of a cardiovascular even in the next ten years
This is a good point — isn’t this what happened with HMOs in the 1990s? At the beginning of the decade, health care cost inflation was a major concern, and HMOs successfully slowed that down, by limiting healthcare (perhaps most famously by limiting the time mothers could stay in hospitals post delivery). Then came the backlash against HMOs, restrictions were relaxed, and health care cost inflation rose again.There’s another point worth considering about that history. If the data judge is a private health insurance company, patients have some recourse — insurance companies are sensitive to public, political, and regulatory pressure. But if we end up with a single-payer system because ACA blows up the non-government health insurance market, then the sole data judge is the government. And patients will have no recourse.
That horse has just about (about 40%) left the barn:Medicare: Medicare spending, which represented 21 percent of national health spending in 2011, grew 6.2 percent to $554.3 billion, an acceleration from growth of 4.3 percent in 2010.Contributing to the increase in 2011 was faster growth in spending for skilled nursing facilities and physicianā services, as well as an increase in Medicare Advantage spending growth.ā¢ Medicaid: Total Medicaid spending grew 2.5 percent in 2011 to $407.7 billion, a deceleration from 5.9-percent growth in 2010. This was partly due to slower growth in Medicaid enrollment of 3.2 percent in 2011 compared to 4.9 percent-growth in 2010. Federal Medicaid expenditures decreased 7.1 percent in 2011, while state Medicaid expenditures grew 22.2 percentāa result of the expiration of enhanced federal aid to states in June 2011.
Rich, didn’t realize you were plugging away in healthcare. I’d enjoy digging deeper at some point to hear more. Got to support one another in this crazy industry –
Random, we’re planning an online widget/service for the risk calculator – not a monumental task. Our business is web services for healthcare. If/when we get around to this I’ll let folks know.”Big data” has been all the hype in healthcare for the last few years. The government even started a conference around it led by the U.S. CTO Todd Park. The challenges to realizing “data driven care” are numerous. Quality of available data, lack of semantic interoperability, missing data standards, medical culture, patient’s real relationship with their health, etc.I have no doubt it will come. As others have mentioned, data acquisition is still part of the hold up. Today’s EHRs started as databases and later slapped a terrible UI on top. Quality data acquisition is painful.There are a ton of interesting startups in this arena – and a few big guys. Accretive health (public) recently decided to make a big push into “big data” and started a data team. While Accretive is based in Chicago the data team is run by a former LinkedIn data scientist out in SF. What’s really interesting? The team of 6-7 Accretive engineers is renting extra space from, and co-locating with, a pre-launch health tech startup. A scrappy startup in its own right.In medicine, tech is only a very small part of the problem/solution. Realizing “big data” in healthcare is more about changing culture and reimbursement than anything. It will happen, slowly.
please let us know when the widget launchesdon’t you think patients will take things into their own hands and take advantage of big data regardless of the culture and reimbursement practices of the legacy industry?
It always comes down to incentives. Usually that’s reimbursement based. That said, we’re already seeing patients take things into their own hands when the system fails them. It’s usually a desperation play – not the first stop.PatientsLikeMe, CrowdMed, etc are used by patients who the traditional establishment has failed to help. Patient’s who clammer for pricing transparency are ones who insurance has failed and are stuck paying out of pocket. This is a fringe (but a growing) group.Anticipating consumer behavior in healthcare is hard. People aren’t making purchase decisions based on convenience. The “best” service doesn’t win. Big data in healthcare is hard because health data isn’t the same as other data and the idea of offering up something for free and monitizing the data later can be a barrier to spread. See anything written by Deborah Peel for the extreme patient privacy perspective. (Side note: Data Ownership needs to be “figured out”. You and I know it’s the patient’s data. The medical establishment/government is less clear and this hurts consumerization of healthcare).In essence, the patients most incentivized to take things into their own hands are the marginalized and forgotten. A very difficult group to reach and educate. Even harder to find a business model that works. Omada Health stands out here as a group having initial success reaching consumers – their business model is still dependent on employers and insurance though.I have little doubt consumerization of healthcare takes place. I’m just very uncertain of the timeline. Any business with goals of doing this in the future needs to have two other business models that work. My prediction: We still have a few years of fee for service, then outcomes based reimbursement before consumer driven healthcare starts to spread. I hope I’m wrong and it comes sooner.——–Appendix——-Practicefusion is the obvious counterpoint to the selling data argument. Few with deep knowledge of the medical industry likes their business model/thinks it works long term. Somewhere in there is a model that works but it doesn’t seem to be patients paying directly (yet) or selling data to pharma. Most consumer plays in health still end up falling back to the employer or insurer to cover – limited ability to “disrupt”.
I will let you know about the widget. We’d been planning it as a provider tool but I’ll take some additional time to figure out how this looks as something directly for consumers. Good thought exercise non the less.A major focus of ours is to help put patient sourced data in front of clinicians. I can think of a few easy ways to do that with this tool. Thanks for pushing me to think about this in a different way.
no one has mentioned Flatiron Health yet?http://www.flatiron.com/
it looks provider centric not patient centric
it is, but it’s a start, and it’s probably a necessary step to organize an industry that’s so backwards…- reece
If the answer is the current healthcare system; the insurance companies, the hospitals, and the doctors, then we will have missed a big opportunity to reshape healthcare. If, on the other hand, the data is entered by patients, controlled by patients, and benefits patients, then we would have something new, different, and disruptive.Why is new, different and disruptive, “better”?Patients “people” that is, are notoriously bad with things like this.They have a hard enough time taking medicine to the point where they need pill minders organized by day with multiple chambers. And can’t remember the movie they saw the night before. Or to pickup something on the way home.I mean would you say that having the gas station keep track of the gas that you pumped is a slightly better idea than you having to keep track of the same data point?
.Relying on sick people to enter data is a stretch.Better that all this info be digitized from the beginning.JLM.
.At the end of the day regardless of what direction all of this — actionable intelligence — takes it MUST ultimately translate into “better” healthcare that results in lower mortality rates, less pain and suffering, more effective treatment, longer life, higher quality of life and lower costs.Much of this could be effected immediately by some very simple changes many of which require no or very little technology or additional costs.We need to produce more doctors — not brilliant doctors but docs who can provide immediate, location insensitive, palliative care. This simply requires the AMA to get its thumb off the scale of med schools. Not every doc is going to become a millionaire specialist.Medical education has to be reduced to a cost-benefit relationship that evaporates the debt burden of becoming a doctor. Education for service model of investment. The military does this very effectively right now.We need to develop prescribing nurses whose credentials are garnered by on the job training and continuing education who are drifting toward the capabilities of GPs. Able to prescribe drugs only for the most basic of maladies — green v white snot and the flu would be a damn good start. But able to take that costly screen of what really ails the patient.We need to move medicine back to the neighborhoods. The largest providers already gets this “neighborhood policing” service model. Also a great model for indigent care as the indigents — drum roll — live in identifiable, concentrated locations. Clinics. The charity world gets this now.We need to use just a bit of technology — digital lab results in a central repository (already happening in earnest but not nearly as widespread as necessary), digital screenings of common ailments, digital Skype type contact with remote patients and the ability for persons to carry their medical records with them on a flash drive or cloud access. Break your arm in Ft Lauderdale, the emergency room can look at your drugs and allergies immediately, post your new records for your docs back home.Tort reform. Get the Irish Sweepstakes out of medicine.The list could go on forever. This is low hanging fruit.Quick story — back in the day I used to always demand a copy of my lab work. The labs used to think I was nuts. I used to like to graph some of my test results over a longer period of time. I almost had a fist fight with a lab doc who told me he could NOT release my records “under the law”.I got my attorney to give him a shout and he agreed he could and not only that but that I was the owner of the records, not the physician who ordered the tests.It saved a lot of time on “lost records” as I would bring my copy of the tests and my graphs. It saved a lot of time.Funny thing is that now my doc asks me for my graphs before I arrive for my appointment. I send them to him and he has someone doublecheck the numbers.We have gotten to the ultimate relationship — we are partners in my health. BTW, in response to Obamacare he has stopped taking any Medicare patients even ones who have been patients for over 30 years. Going on Medicare is not an insurance change.JLM.
cost-benefit relationship that evaporates the debt burden of becoming a doctor.As someone who knows the real numbers of this [1] I don’t feel this is a real problem. You end up with, say 200k debt but a great career that is about as solid and secure as it gets. We are not talking about law school here. Or a history degree. We are talking about medical school. The debt is paid out over many many years. Debt service is very reasonable relative to earnings.I have no problem with paying out over time something that you will use over time. After all you don’t pay cash for your house typically.In the 1980’s I had to buy a machine that was +-150k and had no such guarantees of long term income. Or anything close to that. What kind of guarantee did you get when building buildings?To me the “burden” argument is simply not valid.[1] My wife.
.Good point.I guess my argument goes more to the notion of not having a medical degree being an economic “sure thing” wherein the prospect of riches is assured.The military has done this very effectively and has attracted top notch talent.My nephew — VMI grad who just blossomed — went to Bowman Gray and is headed to the Army with all of his bills having been paid.Son of a doctor, he is someone who will excel as a military surgeon for over 20 years.Damn good trade for him and the Nation.One size does not fit all but this is a good outcome.JLM.
I know a few people who had the opportunity to do that (Military).Non military you can also get the medical tuition paid if you agree to practice in a less desirable place.For that matter every doctor will regularly get postcards allowing them to make way more money than they make in their current community by practicing in some out of the way place where nobody wants to live and work. They regularly will try and negotiate with their current employer who will say “ok then go practice there!”.Many are actually nice places but they are off the beaten path and not “home”.the is someone who will excel as a military surgeon for over 20 years.Sure but doesn’t that mean you have to go and be a surgeon where the military wants you to go? That seems quite a handcuff. Not to mention I’m curious if anyone ran the numbers on the pay differential with the debt load and lifestyle compromises.
For some, I’m sure such choices are not dictated by monetary measures alone, and I’m thankful for that. š
.Military surgeons get a lot of trauma work and you are going to be pretty close to the battlefield and then Germany these days.You will be using the best equipment and operating every single day.If you want to excel in your profession you are going to have to go where the action is.This young man comes from a tradition of military service and he wants to serve. You don’t go to VMI because you wanted to be a poet.JLM.
Maybe some enterprising entepreneur (sic) will take this “Omnibus Risk Calculator” put it into a clean and simple web service, allow us to connect our phones and connected devices to it, and peer produce a service that we can, together, use to manage our cardiovascular health.I’m sure someone will do that and post it to USV or HN for sure. If (as you say) it’s not been done already.But why do we need to use it to “manage our cardiovascular health”. From what I can tell that’s just one item that has to do with cardiovascular health. Of many.Not to mention that you encourage a group of people to be the “worried well”. It’s no substitute to visiting your doctor to get (presumably) sound medical advice and opinions. As William points out (and I agree 100%) medicine is art and science. You are not going to get me anywhere near to believing that good tools in the hands of consumers is going to replace judgement.
Live your life as of you have heart disease and you’ll likely never get heart disease.
I think a similar opportunity exists for the real estate industry: http://brandondonnelly.com/…
At Lively, I am happy to say that we are in the midst of distributing a hardware platform (necessary evil) that is simple, inexpensive, beautifully designed and social. That’s what customers say and the press has been saying as well.For us, it’s simply about finding the most effective way to gather ADLs on elders, that combined with genomic data, nutritional data (think Proteus or other ingestable techs), vitals measured through smartphone apps and EMR to complete the perfect functional form so taht each individual can optimize their health.We’ve started with seniors since it’s the hardest area to crack and the most need, since we spend 40-50% of our health care dollars in the last 18 months of our lives.The potential speaks for itself.
The data has been there forever. The hard part is getting providers to actually pay attention to the data. My friend Dave Newman has been evangelizing getting doctors to pay attention to the data for years http://fora.tv/2013/11/06/t… . It is surprisingly difficult to do this. I have helped him integrate the data into EMR systems, this seems to be fairly effective but adoption is an uphill battle in non ACO systems.In other news the snake oil sellers are riding hard into the consumer medical device with a pretty-smartphone-display market and the horse they are riding in on is kickstarter.
with a pretty-smartphone-display market and the horse they are riding in on is kickstarter.Link?
Here’s one. I have seen many others. I have not been keeping an active list.http://www.kickstarter.com/…I vetted this one with friends in the Stanford Neuroscience department. They are dubious.Airo was another one.
The founders:http://www.pomona.edu/news/…(Deserve credit for being 2013 graduates and doing this I will give them that).Would I buy it? No. I’m not looking for “the answer”. To me the answer is to work hard and to be pragmatic. I don’t need any of those exercise helpers (I recognize some people get benefit) I just exercise every day. I actually just discovered that music makes exercise easier (for real until this year I never listened to music while excercising..)Here is a company I looked into years ago for a friend that had a child that had adhd. It was a “ymmv” for sure. Apparently helped some people didn’t help others.http://www.achievewell.net/…What’s interesting is that with the above “center” they are dealing with desperate people and parents who have tried others things and will pay for another chance at a fix to a problem.With “melon” the target appears to be the entertainment value of the product and buying into perhaps some quick fix or answer or solution to some “problem”. It’s a “party in your brain” type product. Which is fine. Many things are bs or just something we perceive as having value that really is just an emotion or need invented by marketing. But that does not mean that the person parting with the money doesn’t get benefit. They do. I like things I buy that are invented and they give me pleasure. All of us do.In any case this is just packaged marketing for sure. If you changed the founders and tweaked the presentation they wouldn’t have come close to raising that kind of money. The photography oddly was really bad looks like they used a smart phone to take the pics.
I was an organizer last year of a world IA day event (information architecture) and we had a mix of academic and entrepreneurial approaches to big data and healthFriends from http://www.stratasan.com came; one of the higher profile startupsThere were a lot of familiar examples — how Google maps / Twitter predict fluAnd there were public health examples — map lifestyle diseases and income by zip code and plan outreach and clinics accordinglyBut the biggest block that everyone saw was the lack of public aggregated data because of fears of insurance and or employment denied or higher rates charged because of illness.Universal healthcare unlinked from employment would end those concerns and information could flow more freely
I recently did some project work for a start-up in the healthcare industry. There’s a lot going on in this space. One interesting company is Practice Fusion, a VC-backed aggregator of medical records that already connects the medical community w/ over 75 million patients. Their software is provided free to all medical providers in exchange for non-intrusive ads placed on their network. Electronic healthcare records systems will also benefit from Obamacare and Meaningful Use protocols beginning in 2014. There are both gov’t requirements and incentives for this type of reporting. Separately, the industry needs a Kayak type service for informing consumers about the cost of medical care and procedures, which is not only out of control, but varies far too widely from market-to-market and facility-to-facility. The problem is, sadly, it’s not in everyone’s best interest to serve such a need.
This is already happening. There are lots of startups tackling big data and health, patient reported outcomes, prevention etc. There are also over 500 accountable care organizations that are moving to a model where they are paid on outcomes, not procedures. Google and Watson will be big players in this space in the future as well. 23 and Me was on a gamble to get enough data before the FDA noticed so they could prove their results without a big trial.I was recently at FutureMed where the CMIO of Kaiser said they’d been doing big data sine before they had an EMR. There was also a lively debate about whether doctors were going to have to be data scientists in the future or whether computers would do that and doctors need to get back to the caring part of healthcare.I founded an m-health startup less than a year ago and have noticed that the general tech community is quite unaware of all the disruption that is actually happening.
Plus, they are wary of investing in it. It’s also a non starter to talk to potential fund investors and mention “health care”.
Yep! That came as a bit of a shock to us early on. š
The idea of a Personal Health Record (PHR) is exactly what you are proposing: patient-controlled medical data. There are many versions, but they already exist. http://en.wikipedia.org/wik…There is a tremendous opportunity to crunch the data that we get from a lot of patients to not only improve their care but also the care of others. We already have the opportunity to aggregate data and anonymize it. Anonymization is necessary for Safe Harbor: http://en.wikipedia.org/wik…. Even though Safe Harbor is an EU thing, it’s something with which any company that might go international has to be compliant.Right now, I believe that the hospitals own the aggregated data and they can do with it as they like within the bounds of HIPAA. What you’re asking for, I think, is patients opting in to let their data be used for the good of all.What we actually need is for hospitals to allow patients access to their PHRs. Patient portals are the first step, but there’s more access that has not yet been given. Of the customers of the major healthcare IT company where I work, ~7% have PHRs turned on, even though patient portals being enabled is part of Meaningful Use. Hospitals have to be pushed to enable PHRs on top of just giving patients viewing access to their medical information.
As disruption historically works its way up from the bottom I think you are correctly pointing towards the best low-hanging-fruit entry-point into this oligarchic money/data tree. You have correctly identified the substrateundertow entry point :-)It seems analogous to the consumerization of computing.First you legally force all healthcare entities to provide all their customers with the information those customers have(directly or indirectly) purchased from said healthcare entities.Then it become a much simpler exercise in convincing most healthcare consumer to provide their anonymized healthcare information into a common public data-silo for everyone’s mutual benefit.The key here is the fact that there is almost no credible argument to be made by any of the incumbent healthcare entities regarding their customer right to own the information they have paid for.
i wasn’t thinking of a personal health record actually. i was thinking of a personal health graph
Then you might be talking about something that’s already part of most patient portals. In ours, you can see your glucose readings over time, for example. Those are automatically filed and there’s no data entry required.
I think privacy is a big concern with cloud healthcare, particularly as startups tend to move fast and break things.It’s one thing to have scammers use your Facebook photos to make bots, or your compromised server get indexed by the shodan search engine but having your medical data get indexed by the dark web seems comparable to losing your SSN.It would be interesting if there was a way patient records could be transferred with the bitcoin protocol using something like mastercoin.
Just stuff as simple as password reset procedures is atrocious in many modern health care and maintenance apps. This includes apps geared toward rather sensitive health conditions. Too many store passwords in plain text, emailing them to the user in the clear, etc. It’s ridiculous. Proper security is a must, is relatively easy to accomplish if developers care enough, and yet Adobe, LinkedIn, Sony, and on and on, too often don’t get it right either.(minor edit for clarity)
Great post, Fred, and It is an area I have been thinking about a lot as well – 2014 might be the year that consumerization hits the healthcare space. HumanAPI (http://humanapi.co/) is an interesting company that is collecting patient data from different sources and unifying it.
have you invested in humanapi boris?
I haven’t but I like the theme.
> some valuable health care services for consumers.While killing a lot of other consumers.> The art of medicine is becoming the science of an insurance actuary.Nonsense. Incompetent, uninformed, misinformed, just plain wrong,smoking funny stuff nonsense.
I have no idea why Prof. Karlawish said that other than to get more attention to this opinion piece. The 2nd to the last paragraph says:Calculators like the Omnibus Risk Estimator are simply tools, or devices, akin to the hip prostheses and pacemakers doctors implant in their patients. They are designed well, or they arenāt.But then he gives a ray of hope (and support) by saying this:But unlike other medical devices, which must undergo standardized testing and unbiased review and monitoring by the Food and Drug Administration, these risk calculators are developed with little regulatory authority over their design and use. There needs to be better oversight. I have no idea why he would attach his name to a piece with an obviously (agreeing with you) ridiculous sound bite.He could have said ” The art of medicine is becoming the science of an insurance actuary. Or so some people would have you believe..”
“The question is who will control the input of the patient data, the aggregated data sets, and the results the data science produces. If the answer is the current healthcare system; the insurance companies, the hospitals, and the doctors, then we will have missed a big opportunity to reshape healthcare. If, on the other hand, the data is entered by patients, controlled by patients, and benefits patients, then we would have something new, different, and disruptive.”IMHO this statement is a bit simplistic and flawed. A while back you posted numbers about user engagement metrics on social networks such as Foursquare in which you shred that only 10% are active participants with another 20% being semi-active.If customer’s need to actively share the data I believe that you will find the same lack of participation by most customers (just ask the Google Health team).I believe that all the data that’s already collected by doctors and medical providers need to act as a universal base of data, with smart devices and apps which doctors prescribe being an additional source of automated data. And only then can you have data that “patients” add in which will be a smaller sample but can be layered on top to extract additional information.This kinda conforms to how studies are done today, with a group of patients being asked to provide additional information, from new tests to keeping journals.This actually reminds me accounting where you receive a base of general data from your financial institution, but then users need to manually extend the data to match categories and accounts.
10% of all users creates a large panel that can and often is statistically relevant
Like medical studies 10% is a phenomenal participation, but as is also the case with these studies the base data is collected by professionals and not simply a questionnaire filled in by the user. Also health is different as the need for outliers is far greater, because if you don’t get the right AdWords placement nobody died which can’t be said for Health applications.There’s another important point which is the fact that most people are not proactive when it comes to their health which can be seen from the type of consumers buying wearables, so applications will need to be more like a Google Now which monitors their health data in the background and presents application interactions akin to a mobile push message or a Google Now card.What this really means is that the cloud, Internet of things health sensors, wearables, big data and every other vertical buzzword morph into a single big capability and opportunity. So we need to stop thinking in terms of a given way of doing things that we’re used to, but start cherry picking models from different verticals.That said how health tech ultimately plays out will be very interesting.
Does big data mean small privacy? Canadian woman refused entry to US due to medical history (hospitalization for depression). .
Having been involved in healthcare for over 25 years as an employer (Plan Administrator for an over 1,000 lives self funded plan) I can say that getting doctors to share data with each other and with their patients would be a dramatic first step in making the patient the actual “consumer” of the healthcare decision process.Obamacare was to create an Electronic Medical Records mandate, which was to achieve this goal; sadly its a disaster. Its too bad that government purchasing follows a 20th Century way to procure services. I think with Obamacare the government would have been much further ahead had they gone through VC firms and invested in start ups to achieve their goals rather than procure their services from stodgy blue chip firms with more political clout than ability.50% of our nation’s annual expenditures on healthcare are tied to just 5% of our population.http://www.nihcm.org/images…Studies have also shown that 80% of an individuals lifetime medical expenses are incurred in the last 2 years of their life of which over 30% is for care that made little to no difference in their eventual outcome (there was no improvement in quality of or quantity of life). “WHAT IF:”Patients had access to their lab work results? What if they could track their results historically for blood work, EKG’s, their blood pressure, so on and so forth. I would say that 90% of Americans, even those who go to the Doctor regularly, couldn’t tell you what their weight was a year ago, their blood pressure, or their cholesterol.Then patients could review their own data and assist their doctors in deciding what their annual physical should involve. Then they could become knowledgeable and search the web for drug interactions that may effect them based upon their personal medical history.Then patients could actually sit down with their doctor, when a condition arises, and be prepared with the latest research about promising solutions for their condition.Most Americans don’t have access to doctors who are true specialists who are current on changes within their specialty.Patients would be empowered to become an active participant in managing their health rather than dependent on a doctor who has about 15 to 20 minutes budgeted for the office visit and who expects a nurse to follow through.Its a first step in a very long process.
The advances in data collection, storage, and processing are incredibly exciting but I agree with Karlawish that there is significant risk. Many analytical processes have limitations or are not appropriate for certain use cases. Technology has made these tools available to the average business user, who is likely not trained in these limitations. Trained, experienced, data scientists understand this “art” of how to apply the appropriate techniques to real-world use cases. The scaling of this “art” is what will transition us to Analytics 3.0, in my humble opinion.
This is most definitely a space that I am interested in, working right at the intersection of healthcare and consumer technology. Looking forward to see how this space evolves.
First, in full-disclosure and to provide some framework on how I look at the issue. My father and grandfather are both physicians. My father is a Primary Care (Internal Medicine/Geriatrics/Family Practice) doctor, my grandfather a Cardiologist. My mother has worked for several of the large health insurance companies. As a result, she saw how they often delayed payment to providers in order to review claims to find reasons to deny the claims. Finally, I have a pre-existing condition ā juvenile diabetes (insulin dependent). For those unfamiliar with it, it is a genetic disease. I have been a juvenile diabetic for well over 20+ years. When I became old enough to no longer be on my parents insurance and before I found my first job after college, I was denied by every health insurance company I applied to because of my pre-existing condition. So yes, the healthcare system needs to be reformed, but it needs to be done in a way where patients and doctors come together as a team. With that said, we need to be really clearabout what specifically we are talking about when talking about healthcare. There are some diseases that really do require someone other than the patient to take control. I am thinking of mental health issues (bipolar, depression, etc), down syndrome, autism, etc. Iāve had enough friends who were bipolar and decided on their own to stop taking their medicine. They are now dead.In terms of ābig dataā, as others have already mentioned, that is already happening. As I said, I am a juvenile diabetic. I use a blood glucose monitor and insulin pump. I have the option of using sensors as well, but choose not to. My blood glucose monitor beams my blood sugar reading to my insulin pump. This allows me to do one of three things: 1) give an extra dose of insulin if needed, 2) determine how much insulin to take based on my carb ratio/food intake, or 3) put the system on suspend so that I donāt receive any insulin if my sugar is low. I am also able to save my blood glucose readings and email it to my doctor (if I choose) so she and I can work out the best way to handle my diabetes. As someone who also suffers from low blood sugars and seizures, I value my doctorās opinion.Regarding patients taking more control over their healthcare that really depends on the individual. As I said, Iāve been diabetic for over 20+ years and I sometimes suffer from seizures brought on by low blood sugars. (The seizures donāt happen all the time though). I have the choice of using sensors to read my blood sugar and tell my pump that I either need more insulin or should put it on suspend. I choose not to use the sensors in part because I would have to in essence stick myself twice and wear two different ādevicesā in two different spots on my body. (The sensors and the pump currently canāt be worn near each other). There is nothing at the moment that would motivate me to change my mind on this. And I am not alone. As someone mentioned, you may make a lot of tools to help manage various chronic diseases, but at the end of the day, you have to get the patient to use it. Thereās where the problem comes in.Finally, in terms of healthcare data, people like me who have pre-existing conditions and who have been denied health insurance coverage tend to be very protective of our data. We know what itās like to be denied care because of it. In my case, juvenile diabetes is genetic. So it wasnāt something I did, it wasnāt my environment, etc, but I can still be penalized for it. Also, each personās healthcare data is unique. This is why we must be careful on how we use these tools. My data is different from my brotherās and cousinās data ā both of whom are also juvenile diabetics. So a one solution fits all approach is not the best approach. Assomeone said, medicine is part science and part art.
Isn’t Google trying to help us live longer? I would like that! š
We are planning to do some of these heavy lifting big data / healthcare tasks . Our initial focus will be a massive dataset of radiology files so we will try to sort/clean/make sense out of it and see if there are interesting correlations that will benefit patients. Will keep you postedEyal
Can you say “tort reform”? ;)What you say above is absolutely correct, and fixing this one issue would have done much more for health care in the U.S. than what the ill-conceived, ill-constructed, ill-executed, AHA could ever have been hoped to achieve. It’s sad, really.
The Lawyers will shoot them before the government regulators. Prior to getting shot, they will get audited.
Torts are the secondary problem. When you are compensated more by doing more stuff rather than outcome then you tend do more stuff.
Baloney Sandwich
i am not a billionaire but i am game. can you suggest how to do it?
I don’t have the link handy, but there is an allowance in the Medicare rules that provides for doctors to refer a patient to themselves for certain very costly and profitable procedures. The practice involves setting up another entity and referring the patient to yourself and in so doing, Medicare pays many times the “normal” rate for the procedure. I’ll see if I can find the reference, but perhaps you already know of it.Edit to a add link: http://www.montereyherald.c…Many articles on Medscape (subscription required) and in medical journals, but the above link provides a decent overview.
If I had the data on outcomes of doctors performance, then I could make an educated decision as a consumer on what matters to me. Otherwise I have to make a choice based on an irrelevant factor (I liked the doc, or her office,etc.) or on statistically insignificant sample on loosely coupled point data (I ask my colleagues, and they tell me who they like). I don’t have enough relevant outcome information on all the doctors in the area that are relevant to me.Even if I had all of the data on provider outcomes, this fails to cover overall satisfaction regarding the total experience (wait time, billing, etc.) and depending on the reason for the visit, that non-medical factor has a different weighting to my overall satisfaction.
As a long time healthcare VC investor, I couldn’t agree with you more when you talk about the fact that the normal market forces do not apply in healthcare – where the government pays an ever increasing portion of the bills. There is a fundamental disconnect between the payer and the consumer of the service unlike all other areas that most tech focused VCs invest in.This is not to say that insurance benefit design changes like high deductible plans and reference based pricing wont help consumers become more savvy….but the vast majority of the excess cost in healthcare is not coming from the audience that uses web services to manage their health. It’s convenient to think so but just not true.The vast majority of excess costs come from sick, poor, elderly called dual eligibles (people who qualify for both medicare + medicaid). The sickest of this group have multiple chronic diseases and at least presently, the elderly wont use the technologies the rest of us use to manage those conditions.See recent data from Pew:http://pewinternet.org/Repo…Among those with chronic diseases, the two most popular ways to track one’s condition:1. In my head 43%2. Pencil and paper 41%….way down the list: App (4%), Website (1%)And these numbers are for those who actually bother to track their conditions – which the same data shows is a vanishingly small percentage of those with chronic disease:”28% of trackers living with two or more conditions [track] daily”So – the sad math is that 50% of 41.4m US adults over age 65 have 2+ chronic diseases (CDC, Census #s) x 28% of them “track” their disease daily x 5% use an app/website= 289,000 elderly adults with multiple conditions who use some form of technology today to manage their chronic disease.Put another way, 1.4% of the elderly with multiple chronic conditions are tracking today using technology.Meanwhile, Pew said in May that 43% of adults over 65 are using online social networks. So the elderly embrace the use of Facebook but not something as important as using tools to manage their own health and possibly bend the mortality curve so they can be around longer to see the grandkids on facebook.There are two responses to this startlingly low penetration for healthcare apps among those who need them the most:1. It reflects the fact that 98.6% of the possible users of these services who could benefit most are a wide open opportunity set for a great tech focused team applying their skills in healthcare and that the truly great products have not yet been built in this sector2. For whatever reason, elderly people with multiple diseases will not use technology for their healthcare. That hasn’t changed to date and probably wont change going forward -reasons could include the fact that few if any apps have shown any mortality or morbidity benefit or that the medical community hasn’t embraced most of this innovation and wont until there is hard data (refer to the recent “23andMe problem”)To my core, I strongly support innovation and am as much an optimist about the future of tech+healthcare as I have ever been. Big data and other technologies will help healthcare immensely and I am involved in a few companies oriented in this way – but all enterprise (hospital, insurance co) focused ones.Unfortunately for all of us, that’s where the money and customers for IT derived products are and will be for many many years to come.The critical factor that traditionally tech investors and entrepreneurial teams must not forget is that adoption cycles in healthcare take much longer than those in tech.A study a few years back showed that the average medical innovation (bypass surgery, new med categories) take 17 years (!) to reach 80% of the eligible treatment population.This miscalculation of adoption cycles in healthcare-oriented innovation is the “third rail” of IT investing in healthcare.I’ve written about this fact extensively – for those who care to read. http://www.pehub.com/2013/1…By the way – kudos to Fred and Albert for planting a flag here – I am rooting for them
Agreed.By the way, here’s my favorite study on using patient satisfaction as measure:Patient satisfaction linked to higher health-care expenses and mortality http://www.ucdmc.ucdavis.ed…From the journal publication: “In a nationally representative sample, higher patient satisfaction was associated…with greater inpatient use, higher overall health care and prescription drug expenditures, and increased mortality.”Lesson: the things we often take for granted as being good factors in healthcare do not prove themselves out in big studies.
Doctors don’t fiercely protect their keys to the castle bc they’re bad people.I’m sure that is true but the road to heaven is paved with good intentions.Conflicts of interest are, more often than not, caused by subliminal forces undetected or under appreciated by the participants.When you drill down on the organic relationship between the pharmaceutical industry and the medical community this becomes all too obvious.The over use of Statins is a classic example!
i just did it herehttp://www.cvriskcalculator…no big deal
i mean it