On Digital Healthcare
At USV, we are big believers in being public about our investment thesis and the work we do to arrive at them. We are also big believers in working with like minded VCs on our investments. A few years we decided to merge the two. My partners Albert and Andy started collaborating with Boris Wertz at Version Ventures on developing investment theses in the digital healthcare sector. They roped in two analysts, Zander and Angela and off they went. When Zander’s two year stint at USV ended, Jonathan replaced him on the effort.
For several years, this group has shared investment opportunities, research, and insights with each other. They have a shared database of startup companies and a shared market map. They have collaborated on one investment, Figure1, and thought seriously about a bunch more.
Yesterday, they started to share their learnings and, more importantly, their questions and concerns. It is called OnDigitalHealthCare and as it is currently conceived, it is a six part series on what they have learned and where they are going with all of those learnings. They are two parts into it right now.
I hope they don’t limit this to six parts and would love to see them continue to update this blog from time to time with additional learnings, questions, concerns, and insights. I think most everyone realizes that a computer in everyone’s pocket (and possibly elsewhere on their bodies) is going to massively impact healthcare over time. But how, when, and why is a lot less obvious to us, and I suspect everyone else. So we are trying to figure it out and sharing that process publicly in the belief that the more eyes and ears on our process the better outcomes for us and everyone else.
Comments (Archived):
Bookmaked to dive in later–thanks.There is little mystery about what needs to change and looking forward to seeing the thinking around the platforms from a business opportunity perspective.
It seems to me that anything empowering users and doctors makes a lot of sense, yet there are still few options that are enabling that in productive ways.Changing the healthcare system itself is a big puzzle. And it’s an opportunity.
[Long time lurker, first time poster]There’s a huge disconnect between the professionals (largely, nurses, care managers, patient access staff) who are end-users of these tools, the ones whom vendors seek to influence (largely, doctors), and the check-writers (CFOs, COOs of hospitals).This dynamic rewards the companies (Cerner, McKesson, et al) that can navigate the byzantine sales process (not unlike firms on the short list to bid on government contracts). The result: clunky, difficult-to-use tools, with limited inter-operability that would not last two seconds in a consumer retail market.One of my clients (a clinician) relayed that when she complained about how much the workflow slowed down her patient care, the CFO retorted, “this is about charge capture, not patient care.”And so that is where the investment (in terms of funding, and consequently, product development) dollars flow.
@disqus_QnXd3k2LTg:disqus Stay tuned for Part VI on Networks & Marketplaces where we talk about how frustrating it is to find networks that can truly flourish in the healthcare space because it’s all so permissioned (albeit often for good reasons).
@jonathan libov I’m intrigued. Looking forward to reading it.
We love you no matter what.I’ve noticed that when I was last at the dentist (and yet I still seem to be going back, damn wisdom teeth). The much more interesting thing: the software itself doesn’t interlock, nor does it interlock well with medical hardware for diagnostics. even odder: in some cases they didn’t even all lock in well with the same version of windows, let alone each other. It could lead a lot of doctors in debt if they need to force upgrade, or alternatively not offering best care. Which is why dental schools still teach how to develop silver halide x-rays, and why if I remember correctly my childhood dentist emulates versions of windows that were actually build on DOS to support certain versions of certain programs he won’t drop…..How much do you think everyone in the system being mandated to offer operating system independent, and to some degree software company independent, apis in order to facilitate information flow? Do you think there are real reasons to mandate facilitation for information flow that could disrupt these companies?
A mandate would be difficult to enforce, as much as the idea of APIs to facilitate information flow is right on.AthenaHealth may show other companies that there is money to be made through a more open ecosystem via their ‘More Disruption Please’ program (disclosure: my company is in the early stages of joining this program). They not only make APIs available, but they help companies use / implement them. In turn, they charge a fee.I think it has to come from within — legacy companies get disrupted by the ones offering openness, and race to catch up or get displaced.
Over 10 years ago, two doctor friends (just graduated like me) asked me to build a social network to connect doctors and patients with them on the basis of, “You’re the only person we trust to do this.”It’s been fascinating to follow how the notions of trust between doctors, patients and diagnostic/prescriptive channels has changed since then. In particular, how Quantified Self and gamification via mobile fitness trackers has changed patients openness about sharing what had previously been considered private health data.
The blog would be great as monthly product.
@twaintwain:disqus how do you mean?
‘Health Affairs’ meets ‘Alt Assets’.Haha, tell Fred then AVC crew become Product people too as well as Nick and Brittany already being hackers with Quackpad!
There seems to be a lack of incentive for the parties in a position to utilize new technology to change. Tech in the healthcare space still seems to be viewed as a cost center, and if you were a doctor / nurse / hospital endlessly suffering the interfaces of the stove piped 90’s era tech incumbents you would probably be weary as well.It’s a shame that so much VC money is currently being burned up on hundreds of logistics companies hoping to win out, but not a lot is being spent to bridge the profit gap for quality software that could make the lives of patients and doctors significantly easier/ better.
Main problem is simply cost. Here is one of (and perhaps the biggest structural factor) https://danieljmitchell.fil… . End consumers only pay 12 cents on the dollar out of pocket. Both they have no clue “how expensive” a procedure is, and even if they wanted to, the system doesn’t have to respond to them like normal businesses because the end consumer out of pocket is a mere pittance and DOES NOT effect incentives of players in the business. Also, healthcare costs are ridiculously concentrated in a very sick group of people – meaning really 5-10% of people are driving the bulk of the cost. Have lots of ideas on healthcare, but figured these two charts would help for the unitiated on what the major structural factors are. Changing economic incentives is what is needed way more than new technology, although technology may help change them, but not on its own.http://theincidentaleconomi…
Yep, there´s a lot of talk about wearables + monitoring healthy people. Many have tried these things but have no use for them. Watching my healthy heart rate every day is boring. Digital Healthcare makes a lot of sense for people that are already sick, the 10%. And there it can save money. People that need at-home care or have to check blood pressure every four hours + get the doctors opinion instantly. Digital Healthcare can increase coordination between providers and lowers the costs of double treatment.
Thats a great point. The people who usually adopt the latest health tech often need it the least
@mattkruza:disqus Indeed that’s one of the things that makes this market so difficult to sort out. The converse is that the people who need health care/tech the most — babies and the elderly — are the least tech savvy. But maybe also an opportunity. Consider how an adult could help administer care to his elderly parent using something like Honor (https://www.joinhonor.com/) and a wearable or two. We’re starting to see a trickling of cases where wearables have started to have meaningful impact on clinical outcomes and care (http://medcitynews.com/2015….
Not only are they the least tech-savvy, they’re the least likely to be able to bear those costs.Them and the visually-impaired, hearing-impaired and motor-sensory -impaired.My team recently hacked a wearable for the visually impaired and kids which won 1st place and was selected by DevPost as “Staff Pick”.
To add to above, Startup Grind published an article I wrote on Machine Intelligence in which I make my views clear:”Now, before we drink the Kool-Aid about AI and wonder how to get a piece of the action, it’s worth asking if what’s being built will change the world for the better or for the worse, and what part we each play in making that future.An example of “for the better” would be creating machines capable of understanding the meanings in our natural language. It could also involve inventing “robots that care” so they’ll have the moral and emotional frameworks not to kill us, as Qualcomm proposed at the Tribeca ‘Imagination’ festival on bleeding edge technologies in April. It could also mean making sensors that can help kids cross the roads safely, giving blind people the tools to see and experience the world in better ways or something to stop spam emails forever.”So the wearable for visually-impaired and kids is an example of me “walking the talk” in the products I work on.
Yep the wearable trend will help many more seniors stay in their home
whats the objective function? Saving lives? Reducing pain? Increasing Rx useage? Decreasing Rx useage? Increasing dr office visits? Decreasing dr office visits? Longer hospital stays? Shorter hospital stays? More non emergency treatment? Less non emergency treatment? More lifestyle medicine? Less lifestyle medicine? Drugs that’make your current life style sustainable? Drugs that encourage you to change your lifestyle? More physical therapy? Less physical therapy?
I don’t think changing incentives can/will make a difference at all. The consumer is not the customer, they are the user. Many if not most people are deferring to their doctors about treatments. Some people may be a little more DIY (just like some people fix their own cars) but for the most part you walk in and pay an expert some money and they use or recommend what they think works.
You are correct.
the car analogy is informative. most aren’t diy, but don’t you still somewhat price shop for car repairs? I don’t look for the cheapest, but if someone quotes me 500 for a brake job and someone else 200, i will probably get a third quote and if they come in at 250 then i know not to take the first… i feel like you are almost proving my point, because that is how most shop for repairs but almost no one could do that for healthcare right now
You can do it for a health procedure but it requires quite a bit more effort. If you were paying cash it’d be worthwhile (Google medical tourism for some more extreme examples) but many people are also constrained by what healthcare providers are in their insurance network in addition to geography.
There are a few startups that are aiming to bring price transparency to the market: Medycation (https://www.medycation.com/), RealSelf (http://www.realself.com/), GoodRx (http://www.goodrx.com/).This is also an interesting case study (http://kfor.com/2013/07/08/… in what happens when pricing in healthcare is brought into the light.
A bit old but there’s a section on our periodic table dedicated to cost transparencyhttps://www.cbinsights.com/…
Agreed that is possible. Best current way is on prescriptions if you have poor prescription coverage. Goodrx and pharmacy checker for international prescription (some people have saved up to 95% on prescriptions, and while technically illegal to import for money, the govt. has made clear they won’t target individuals who buy less than 3 months supply). Agreed the issue is massively complex, and part of the problem is a “fast solution” here is measured in years not weeks or months which makes it very hard for entrepreneurs
depends on the field. It may require a lot more effort.I know I can definitely get cheaper radiology services. I also know I am much more likely to get a misreading in a statistically significant way if I went cheaper, sine oddly enough, the US government decided to compare outcomes for reads by radiologists.
And how is said expert incentivized?
With money i would suppose 🙂 like the rest of our capitalist system
Ha — good answer! Was hoping to get a bit more into the weeds but love the response.
Happy to attempt to get into the weeds. Not sure exactly what type of expertise you are talking about, but i assume you are responding to the other matt’s comment about listening to a medical expert on what to do. I mean, i think the best way to incentivize them is truly in the long-term with a bigger monetary reward. The biggest way to do this is to help educate the individual on what good care looks like, what might be the range of reasonable prices, and to have 3rd party trusted opinions. In something as complex as healtcare i think both quality advice (somewhat similar to a consumer reports) is important for the average person, and then for middle and upper middle class they could hire consultants and/or tech websites and concierges to help give an unbiased opinion on best providers. The consumer reports style advice coulld be very cheap and / or provided by insurance companies. The others probably can work in the $25-100 a month range (totally spitballing, but i think fundamentally that is a reasonable price estimate). One company sort of in that space now is https://www.getbetter.com/. Haven’t used them so no clue if they are actually implementing right but the idea is very sound i think
The current system, as I understand it, pays the expert every time s/he makes an assessment or performs an action. It does not, as I understand it, pay them for the results. Go to a doctor 5x for the same thing and you are paying them 5x as much as you would if they provided the solution the first time.Further, insurance covers costs of all drugs at the negotiated price (which may or may not include a margin for the provider…not sure). Doctors knows that the patient is not the payer and that the patient wants to live so there is carte blanche attitude (which from a consumer utility thing is great but from a cost/efficiency standpoint might not be).The hypothesis becomes that this payer-provider-patient paradigm where patients’ costs are fairly fixed to premiums + deductibles (some of which is picked up by employers in applicable situations) leads to what amounts to a producer surplus situation (where provider gets surplus from transaction and payer forecasts surplus into their risk pool which affects premiums).
Yeah i would pretty much agree with that as you describe. One way that changes now is with HSA’s that make individuals responsible for everything but annual physical up to a deductible betweeen $1,300 and $6,300 which serves to make the patient have skin in the game. I think HSA’s are the ansewr for the 80% of healthy people, and would lead to substanital cost changes. For the very sick though, other methods are needed. But HSA’s deal with the majority of the issues you bring up for the healthy population, which is still a meaningful bit of the cost.
Skin in the game might be the answer from a cost perspective, but I am unsure if it leads to a situation where we are providing the best coverage for all. There are many goals we should look for when it comes to healthcare and it is difficult to prioritize them.This is why healthcare is hard. Personally I like the kaiser model where payer and provider are one but the provider network has to be huge to make it work requiring huge startup costs.
Yeah many networks are moving more towards kaiser, and i am decently ok with that approach conceptually, but can really limit options for a patient if they get a few doctors they disagree with / don’t like. Having sold heallth insurance for a bit kaiser was pretty much either “i love how its all integrated / in one place” or “i will NEVER have kaiser again, not enough choices, everything is too regimented”. There are merits to both approaches, but consumer push back would be intense if that became the only game in town.
pretty much. Obamacare has provisions to penalize insurers and particularly hospitals for bad outcomes in an effort to stem costs.We’ll see
Save your comments for Part IV on networks and marketplaces, where we dig into some of these issues
Looking forward to reading Part IV.FYI, it looks like there’s a typo you might want to fix.
Good catch, thanks!
The incentives change drastically depending on the situation. Halfway through a flight, if the pilot walks down the aisle and asks, “how much for a safe landing?” almost every passenger’s willingness to pay goes way up.More specific to healthcare, even with perfect incentives, catastrophic care has catastrophic prices. A kidney transplant (my dad had a transplant in ’99) costs $200K+. Not only did he not have time to shop, the difference between a $175K and $200K is negligible. So we have insurance. Great. Almost everyone agrees this is a good use of insurance. No consumerism here, ut perhaps that is OK.On the other end, creating incentives for low cost, predictable procedures sounds good. Make patients consumers, they will comparison shop, providers will compete on value. We’ll get Wal-marts and Saks for healthcare. In theory. However, in reality, enormous life pressures mean the population segment that can least afford to postpone care will do so when they are paying their own dollars. The cumulative effect = public health issue because then we (taxpayers) pick up the tab for catastrophic care of long-term chronic illness (diabetes is a good example). Now we are back to someone picking up the tab for low cost, predictable procedures…which chips away at consumer-oriented incentives.
Will the marketplace price be set by Google in the same way that ad prices are?* http://www.wired.com/2015/0…* http://www.forbes.com/sites…
This will prove to be a myth. First spend Deductables (out of pocket) may have the highest inflation rate than any consumer purchase over the last 15 years.
For some individuals, yes. But not for all big companies, and we need the supply side reforms of more choice / lower costs for consumer powered choice to make a meaningful difference.
News flash !Sick people use up most of the healthcare budget, just like it always was everywhere.. . . . .In all other developed nations with single payer systems users have even less awareness of the underlying cost but those countries still have a much smaller % of GDP consumed by healthcare.Most single payer systems have more bargaining power to pushback against powerful healthcare and pharmaceutical company excesses. In the USA the healthcare and pharmaceutical companies do as they please on pricing using lobbyists to make sure the government is not allowed to pushback with volume pricing even when a government healthcare program is the central buyer/payer.Then Obamacare gets whittled down, before passage, to a system where those that cannot afford to pay the industry’s inflated pricing get the tax payer to pickup those inflated costs on their behalf without ever addressing why the American healthcare industry is entitled to a disproportionate share of American GDP relative to other developed nation.
Wait times are much longer in canada and many single payer. America (those with money) will not tolerate that, so unlikely to have single payer. Also, the us does most of the medical innovation, subsidizing these other systems. So besides being centrally planned we need then to actually follow laws of supply and demand and increase the supply of doctors and give individuals incentives to be healtheir since north of 50% of costs are related to life style choices.
for many services, they are longer here…
which services? Would be curious to see detailed links … not saying there mightn ot be a few… but in general much shorter wait times in US from all data i have seen.
At 65 years of age I and my wife have had our share of healthcare consumption and wait times have never been significant for us or any of our friend and family. American politicians (GOP) are always trumping up that and other BS about Canadian healthcare.Wait times may be better in the USA for wealthy patients but for the poor not so much. And Canadian health outcome are better than American outcomes as measured by the UN.
Wait times statistics are way higher in canada per pretty much all evidence. Not even going to take the time to cite sources, as that is disingenuous to argue otherwise. Maybe people on medicaid are worse off, but medicare and private insurance (70%+ of the market) have shorter wait times. Disingenuous to make it sound like only the “wealthy” have shorter wait times
Wait times are not a one dimensional race. The combination of effective/reasonable wait times in combination with effective/affordable access(without personal bankruptcy) are the more important metric that leds to the overall better outcomes by Canada and most other advanced nations as measured by the UN.There are always trade-offs to be balanced!Cherry picking myopic metrics is just an ongoing rationalization by American ideologues to prop up their denial that their healthcare system is far too profit-centred to be universally effective/accessible to all its citizens.
My wife watches Grey’s Anatomy. I happen to overhear and oversee it from time to time. That disclaimer out the way – there was a whole lot more mobile tech in the last season of Greys – in particular i think it was a MS Surface. All Hospital based of course, but interesting nonetheless.
Going to the VA this morning. wish technology could improve the claims process so I wouldn’t have to go to a third party to get what I earned/need to take care of my son/ self
.The country’s failure as evidenced by the VA is a mortal sin. The VA has to try to be as fucked up as it really is and the people at the top should all be executed.The VA should be disbanded and every vet should get a Humana card.The worst thing is how much money the VA puts into “non-medical” expenses.JLMwww.themusingsofthebigredca…
The military should include healthcare basic training for new recruits. The problem with healthcare is that you often have to learn about the ecosystem while your are sick.
.I can’t vouch for what they do today but in the late 1960s and the early 1970s they certainly did.Personal hygiene is as important as treating battlefield wounds. If I had a guy with VD or a sunburn — they were not available to lug a rifle or do their job as certainly as if they were wounded.A man who is hurt actually costs more than one man — the caregivers and transporters are not available for duty also.In my day, it was nothing for a platoon leader or company commander to check every man’s feet in the unit, force them to change socks, make sure they were carrying soap, and making them bathe/shave every day.We used to have “short arm” inspection overseas — particularly in Korea where the VD was incredibly virulent and powerful — resistant to antibiotics.Once a year, soldiers would get marched over to the dispensary for a physical. They all had to take flu shots and get an entire regimen of shots to be ready to go overseas.The on post medical facilities of the regular army were pretty damn good but we had the draft in those days and there were a lot of good draftee docs.JLMwww.themusingsofthebigredca…
Yep, Healthcare ought to be engrained at the same level as caring for your weapon. Kids at 18, what a missed opportunity
.Haha. In the army when you call your “rifle” your “gun” the DIs would make you hold your rifle in your right hand and your manhood in your left hand and say,”This is my rifle; this is my gun.This is for fighting; this is for fun.”Just a little bit of perspective from back in the day.JLMwww.themusingsofthebigredca…
Our lot still marched you to the dispensary 20 odd years ago: 4 stations, 2 nurses, one in each arm at each station, and if the CSM was feeling sympathetic no pushups that afternoon.
.You don’t want to be the first guy in line when the shot gun is cold. It tears a hole in your arm.JLMwww.themusingsofthebigredca…
Also the AMA is the biggest (and most destructive) union / quasi union in the US. Always has baffled me that conservatives pretty much despise unions, but don’t mind that the AMA (and through there proxy at lcme for licensing colleges) has MASSIVELY restricted both supply of new doctors and limits what other providers (mainly NP’s and PA’s) can do. Also, they are responsible directly and indirectly for fighting against so many tehcnologies (like many states banning video being included in teladoc and other remote healthcare statutes. Just crazy). And i know only around 20% of doctors are in it now, but it doesn’t change that the AMA blocks almost all innovation and most importantly restricts the supply… its really an area the conservatives should have a field day with, but i guess the counterpoint is a lot of doctors are wealthy so they buy conservative influence with that wealth and country club connections?
but it doesn’t change that the AMA blocks almost all innovation and most importantly restricts the supply…As far as blocking supply what makes you think that the rest of the system can in fact accept an increase in supply in the way you are making it sound? Building refineries doesn’t help if there are not enough trucks and gas stations to pump the gas. (My wife is somewhat involved in this process by the way in the training and education..)As far as blocking innovation medicine should be a slow moving process with through review. (Same with automobiles). The safety (of which not rushing into new things is part) needs to be baked into the process because lives and health are at stake. Not some stupid photo sharing app. Right now having the governor that you describe could very well be a benefit not a detriment.Not interested in a system that moves at the speed of internet and other software development where shitty is the norm and there are daily hacks, breakdowns and things that simply don’t work where the problem is foisted on the end user to deal with the aggravation. And you know this happens. [1]Take disqus. No big deal of course, no lives at stake but has anybody noticed that the number of comments listed on the first page never matches the number of comments listed on the comments page? Can’t work like that with medicine and healthcare. In case you are wondering why medical devices cost so much failure must be tightly controlled or there are big consequences.[1] This all started from what I can tell with the PC revolutions where the race to meet the competition meant pushing products out to the channel before they were ready for prime time. Didn’t matter if they worked or not. And it got much much worse once software people didn’t need floppy disks can cd’s to fix things as well…..
Only time for a quick response right now, but i am not mainly talking about “photo-sharing” app type innovations. I am taking econ 101. Almost 50k apply to medical school each year (almost all with the intelligence needed and only 18-20k get in, because the LCME limits the number of students accepted and schools accredited. Both directly and indirectly). Even better example is dental schools. Right now we graduate around 5500. We graduated over 6200 back in the early 80’s. Dentist wages went down and there was a purge. if we had not limited the number we would graduate around 9000 a year. Over the 30 year or so workforce it means we would have around 270k vs 160-180k dentists. this would drive price decreases. Whether 15% or 50% (i think closer to 50%), this would occur with ZERO extra risk in terms of new models / treatment. simply more supply for a very inelastic demand (health care services). Of course there are issues throughout the system so these changes would take 5-10 years to fully take effect, but that is definitely needed.
Unfortunately you are saying “would drive price decreases”. That assumes that the same quality of applicants would continue if the pay and money to be made was not the same. The way I look at it if there is less money to be made then you will have people who decide not to be doctors or dentists since money has definitely driven many people into the more difficult parts of medicine or even the easier parts. By the way somewhat well known I believe that psych doesn’t pay that well and tends to be an easier part of the industry.I am not sure that you are aware (I mean you might be of course) the intensity and work involved in a typical medical practice. Personally I want the highest level of person applying for and treating me. Not someone who can just get good grades or pass the course. You have to understand that it is definitely more than just being able to pass tests and do the work.
if you want to pay $200 to see a md and i want to pay $80 to see a nurse practiontioner then i am ok with that. That is what a market is. Yes its complicated, but those choices should be allowed. NP’s are still pretty regulated
Will be highly interested to read what they are producing. One of the really big problems in health care is the sales cycle. It’s also a difficult sale, with a lot of moving parts. The big health care tech providers, Cerner-McKesson-Epic also don’t make it easy to innovate inside the silo. Often, the CFO is the driver of innovation in a hospital-and it’s all about costs. That’s why when you hear a lot of medical pitches a good portion of the pitch focuses on the lifetime value of customers and the way the software, device, procedure can drive down costs and increase care. Being able to push things to nurses with a short check by a doctor is one example.I think eventually Bitcoin/Blockchain might integrate itself usefully into the med tech system. One of the problems with medicine is lack of a transparent price. You never know what you are paying from place to place, or state to state for the exact same procedure. Blockchain apps might decentralize it and turn pricing into a transparent price that allows market dynamics to work in the sector. Right now they don’t.
One issue with that is that CPT codes are owned by the AMA and it is LITERALLY ILLEGAL to show cost comparisons of them… unbelievable incompetence on our govts part to allow protection of those. If we are going to pay so much for healthcare on a govt level i am PERFECTLY ok with them creating their own coding system for a few hundred million and open sourcing so there is more transparency. Again, another point conservative leaders should lead on in healtcare. Any senators or house members you know (or presidential candidates 😉 ) send them my way, happy for them to share thoughts free of charge and share my comprehensive proposals. 🙂
your own health datashould be proteced by a blockchain technology.
You never know what you are paying from place to place, or state to state for the exact same procedure.I don’t really care what others are paying elsewhere. I care about whether I need the treatment and avoiding having to go to the doctor. But that is me.Anyway, free market system. Prices may be high in one place (NYC) vs. another place (Wisconsin), both arbitrary examples, however ultimately there are costs structures in place (doctors, nurses, janitors, billing department) that have developed over time based on how providers are and were getting paid in the past and it doesn’t matter at this point why that happened. Right now it is the way it is. If there was more money paid for a procedure or hospital stay that money didn’t necessarily all flow to the bottom line. It went in people’s pockets (and not just doctors) as salary.My point is if you take what a gastroenterologist or brain surgeon or internist is getting paid in one area and slash it drastically “you are getting to much for that GI bleed treatment”, they will just go and practice in another geographic area and if that is not possible they will simply retire in extreme cases. Ditto for the guy mopping the floors who maybe is paid more in one hospital than another.
you don’t know what you are paying in two different hospitals down the block from each other in Philly.
I don’t care. I want the care that is appropriate for what I am getting treated for. I am not choosing hospitals or healthcare by cost. I already know that the best healthcare in Philly will probably come from either Penn, Jeff or Chop (just a few examples) except if there is a particular expert at another hospital or perhaps in NYC.By the way, maybe, just maybe, the fucking overweight losers who have smoked their entire lives, drink to excess and eat shitty food [1] should reign it in a bit so that I don’t have to pay for their medications (diabetes, heart and so on) so they can stop being a drain on the healthcare system and costs. (Later we can then discuss end of life care whereby people are kept alive at tremendous expense and at all possible costs…)[1] Supplied by the food industrial complex. Have you ever seen some of the wadlers who buy those huge sugary drinks at Starbucks? And the people walking around in supermarkets addicted to all of the flavorful and well marketed tasty food? Maybe a large part of the problem is actually with the food supply. Actually not maybe. It is the root of much of this problem that we have now. That along with medications that are expensive and that allow people to eat anything they want and take a pill to keep their blood pressure down or diabetes in check.
Wondering how much of this is also a function of the modern work environment. I’ve had many friends who live in suburban houses start jobs and immediately gain about 20 lbs after they started their first job because they work 8-7 and then by the time they get home (they drive there instead of walk) they’re too exhausted to do anything but make a meal and then relax and sleep. Some of the bigger companies are better about this (and have good tax lawyers) so they provide gyms on site so you can expend energy during your lunch hour.For better or worse people seem to react to their environment and incentives, and for many living healthy is very much rowing against the stream. Cities seem to be much more conducive to it because you naturally walk places to get there and get that extra exercise in which has a compound impact.
I’ve had many friends who live in suburban houses start jobs and immediately gain about 20 lbs after they started their first job because they work 8-7 and then by the time they get home (they drive there instead of walk) they’re too exhausted to do anything but make a meal and then relax and sleep. There are a few reasons for that actually. Exercise and looking good no longer a top priority. Job security and money is more important at that time.So if they are married (and if they live in suburban houses I can assume they probably are) then keep in mind that they also no longer have to “mate” and as such they don’t have the same need to look good in order to find a partner. So keeping fit to look good is less important. [1] This is in addition to the time that you mentioned. Also their peers will tend to go down the same road so you have the opposite of “hipsters in brooklyn” going on for lack of a better way to put it.Of course what you are saying is true but really at the core is the concept which I have used to describe many things which is “you can only be as honest as your competition”. [2] So if others are working their ass off you have to keep up with them even at the expense of your health.[1] And I think anecdotal when people are either getting divorced or are cheating they tend to take their appearance more seriously because they need to. You see this with women quite frequently or at least I have noticed it.[2] See ETSY attached as a good example….
“One hospital charges $8,000 — another, $38,000″Title says it all. http://www.washingtonpost.c…That said, it’s a pretty significant challenge in any industry to figure out pricing for an unknown thing in advance, especially if you don’t have signifiant data to help with finding average costs and significant volume to cover outliers (read, the market consolidation is going to continue). You’ve got a few rare organizations moving to fixed cost procedures (Geisinger and Coronary artery bypass surgery http://www.fastcompany.com/…, but that’s definitely the exception right now.Just look at Castlight’s effort to bring pricing transparency. Great team, investors, lots of cash and they’ve hit enough walls to market themselves as the generic “Enterprise Healthcare Cloud”, not the “Complete pricing transparency solution” (http://rockhealth.com/2014/…
These numbers are often window dressings.
As long as you have insurance…see what kind of deal you get if you go in and try to negotiate on your own behalf.
True, but that horse has left the barn and really isn’t the issue.
yeah, the reality is that there’s no “the” issue…more facets to this monster than you can shake a stick at. I could nerd out on this all day. I do come to think of it.The fundamental problems in health care are systems (people and process) based, not technology. I’m long on any solution that improves/enforces good process. I’m cold on most anything that promises a technological solution. Even with the greatest NLP engine in the world scanning every scrap of text, if that isn’t presented to a provider in a way that causes them to change their workflow, it’s not worth a whole lot. At best it’s a component of a solution.
Hmm, I wonder what parts IV and V will be…
It’s right there on the website. Part IV is “Networks & Marketplaces”. There’s no Part V but there is a brief concluding section.
Better double check those roman numerals then. (╭ರ_⊙)
Fred, Albert et al should talk with James Dias, the CEO of Wellbe.me. Apart from running our 22 person startup that’s defining connected care, he’s a big thinker in the space and gets pulled in by a lot of hospital ceos to help them innovate.We’ve built the last mile in healthcare to knit disparate entities together into holistic carepaths for the patient, all accessed by phone or computer. We produce better outcomes, less risk, better patient experience and satisfaction.
Ton of producer surplus in and created by system. Hope consumerification alleviates that.
I think most everyone realizes that a computer in everyone’s pocket (and possibly elsewhere on their bodies) is going to massively impact health care over time. So, in health care, a computer becomes yet another example of the long ridiculed solution looking for a problem. Yes, that approach is backwards, but it doesn’t always fail! And, computers are (A) in some surprising ways darned powerful and getting more powerful rapidly and (B) have not yet been fully exploited. So, maybe there will be opportunities for the theme of computers in health care.But, there stand to be some problems:(1) For anything very serious in health care, it commonly takes a lot in time and money to satisfy important regulatory bodies and to establish safety and efficacy.(2) Health care data can be sensitive, and much of end user computing is awash in data security problems.(3) Since end user computing devices and the last mile of the Internet are not very reliable, they cannot be used where high reliability is important.(4) We’re still looking for the important problems in health care general purpose, end user computing will solve. It would appear that, for such a problem with the solution to be embraced by the medical profession, that profession should already be concerned about the problem and looking for a solution. What are such problems?(5) The medical profession is not highly interested in computers and, thus, will be slow bring forward problems and embrace computer based solutions.(6) The computer industry is really good at the computing for bookkeeping, billing, inventory, word processing, and taxes and at data base, data communications, etc. but not at health care.(7) Health care is awash in serious issues of legal liability.
Wondering where Blockchain and Machine Intelligence will transform global healthcare the most effectively…* http://www.forbes.com/sites…
Well we will see if this actually works or produces any significant gains. I see it more as “There is a pot of money. People want answers and solutions. So let’s figure out a way we can charge for something of value and get some of that …”.Note that major revisions I recall of both the air traffic system and IRS computers which appear to be way less difficult take quite a bit of time and don’t exactly go off without major problems and hitches with billions wasted. (And for that matter nobody has figured out how to stop spam and plenty of smart people behind that. Why not unleash Watson on spam!)
Watson’s more interested in solving cancer than spam:* http://www.washingtonpost.c…SigmaAlgebra and I had an exchange the other day about Machine Intelligence and his position is that we have a higher chance of finding the cure for cancer than being able to model human intelligence.In recent years, the words “pattern recognition” have been appropriated by investors from the Data Science and AI sectors (nothing wrong with this, incidentally). Calculating the probability of cancer occurrence is where the “pattern recognition” and quantum processing power of distributed AI would be really powerful.
Spend a few days on mammography data and you’ll see that’s there is more to the story. (First bone up on logistic regression and hazard rates.)
Thanks for suggesting.I read ‘Improvement of breast cancer relapse prediction in high risk intervals using artificial neural networks’ paper which compares this methodology with standard Cox proportional hazard model.I’ve been thinking for a while that there are a number of issues with the mathematical tools we’ve had for medical diagnosis to-date. Issues to do with false positives, correlation not being the same as causation, representative sampling, lack of complete knowledge about genome-protein interactions and the power (or lack of) of non-invasive machine imaging.Instead of logistic regression and its “best of straight line” fit, AI researchers are applying more topographic algorithms to try and gain more of a 3D matrix model pinpointing where the mammary gland has a higher probability of showing an abnormality.I thought of the limitations of the maths and the medical tools after my Dad died, following a coma. It became really obvious that the most sophisticated fMRI brainscans, EEG machines, breathing monitors etc are still only providing an ability to see inside and understand a % rather than 100% of what’s happening inside our bodies, brains and consciousness.There’s still a lot of research and technical innovation needed.
There are other models beside the cox proportional models. These include accelerated failure models and cox models with time dependent variables. Other approaches are stratified proportional models. Sure there may be other non-linear approaches, but each their strengths and limitations. Outside of the data, the issue is what are you trying to minimize or maximize? Early detection? False Positives? False Negatives? Radiation exposure? Lower death rates? Total Treatment Costs? Certain treatments? ….
What the end-objective is is itself a classic Operational Research problem.If the end-user of the tool and dataset is an insurance company, the OR model is optimizing variables for highest likelihood of cancer (so the medical premium charges can be higher) — sad but true.If the end-user is a doctor, they want lower death rates because they’re bound by Hippocratic Oath to try to save as many lives as possible.If the end-user is the patient, they want lowest treatment costs and lowest radiation exposure for peace of mind.
Got to give you a C- for those answers, sorry!
Why? It’s true.Which variables to include in the algorithm do depend on the motivations, biases and vested interests of the end-user of the tools and datasets.* http://www.nytimes.com/2015…
This is a good take on the dissonance that often occurs when statistics meet medicine: http://econlog.econlib.org/…. And I agree with the notion that in healthcare, we often fail to properly value the cost of treatment as it compares to risk of illness.
Thanks, lucky you working on this project!When we dig deeper, it’s to do with fundamental flaws in how economics defines and models for value, risk and trade-offs.Price (cost) is often used as a proxy for value when value is much more complex than can be fitted onto a price elasticity curve (e.g., in demand-supply; ISLM).Value, in its more complete factor variable form, carries with it implicit and explicit “comparative perceptions”.These perceptions may include Patient 1 FEELING better than Patient 2 even when the costs of treatment and risk metrics for Patient 1 and 2 are exactly the same.This is something that IBM Watson, Google and others are missing in their knowledge corpus because it’s something economics and mathematics hasn’t yet been able to model for in a coherent way.There’s a genuine need to bridge / close those dissonance (signal:noise) gaps.It’s a conversation I had with the CTO of Philips Healthcare at the Deep Learning Summit earlier this year.It’s a BIG, HARD PROBLEM for Machine Intelligence and also for Economics, Finance & Accounting and Mathematics to solve.I happen to be crazy enough to think about these problems and even madder still to actually invent tools to solve it.Haha!
most doctors also don’t fully understand the risk elements either
not necessarily.
Ok, at this point Disqus isn’t parent-child indenting accordingly so it’s not clear which comment you’re replying to?(1.) IBM Watson.(2.) Issues with mathematical tools for medical diagnosis.(3.) Operational Research.(4.) Flaw in economic modeling.Also, can you specify WHY “not necessarily”? That will help us understand what you’re reading that gives you your views.Thanks!
Ah, yes: Definitely multi-objective optimization!Right?Maybe.So, want a point on a non-inferior set.There’s a book on that by Jared Cohon, one of my dissertation advisors, and long and until recently President at CMU. Yes, I took his course!
Honestly, if you could get patients involved in that question without using big math terms, you probably could get some more generalized models(or variations of models) for people who have to do early screening/considering surgery.The answers in practical and emotional terms are different for different people at different points in their lives
Can I have that paper for personal reasons?And yes.I should tell you that genome to genome research is a high priority right now for breast/ovarian cancer researchers because of the wide range of risk associated with BRCA1/2 and the sheer amount of people who either have BRCA1/2 but as a VUS or don’t have it but have family histories of very deadly versions of breast/ovarian cancer. (which is why there is serious hype around PALB2 http://ghr.nlm.nih.gov/gene… howeve there is a slew of genes being worked on in this category, and upenn/basser is working on individualized genetic risk assessment)
http://www.biblioteca.uma.e…
which cancer are we talking about curing is the better question…..
We like to think that, regardless of the type of cancer, there is 100% objectivity in the tools we select to model risk occurrence.Again, the reliability of the tools boil down to three factors:(1.) How we measure for bias.(2.) Correlation ≠ Causation.(3.) Genetic and lifestyle variance.
watson is always interesting
Fred,I’m happy to see you all are stepping into this arena. There is certainly an opportunity for massive impact – patients are begging for change – but I agree with your sentiment that how, when, and why is a lot less clear. The ball of yarn is so tightly wound that simply navigating and understanding the sales process takes a year or more to figure out, draining most startups initial financing.Naturally, this would lead a lot of entrepreneurs to try direct to consumer. Every other day I read about the “Uber for Healthcare”. This model is not new and we’ve seen a lot of success in private pay, elective procedures. Smaller, more specialized clinics are stealing customers away from hospitals. LASIK and surgical centers are great examples of this working, but still require high capital costs and expensive personnel to execute. However, health systems are acquiring these clinics in hopes of keeping patients within their referral network. So if you’re a health IT startup, like us, you generally find yourself back in a complex sell to a system which move at snails pace.Secondarily, as far as scaling D2C to a billion-dollar story, the people that eat up 80% of our healthcare costs do not have disposable income. They struggle to pay rent, eat quality food, etc. So the degree to which you are able to sell direct to consumer is still limited by high costs to reach these people, and the complexity of completing the transaction via medicare / medicaid reimbursements. “Ubering” this process is a bit of a misnomer. I’d love to see it happen but I’m skeptically optimistic. Our government is a huge dampener on innovation compounded by an aging baby boomer population – aka silver tsunami – causing all sorts of problems.As the co-founder of Patient IO and first-time healthcare entrepreneur, I grossly overestimated the complexity of this industry. Please continue driving this research and conversation so this remains topical. Too many young minds and smart money are not focused on some of these core social problems.Salute,Colin
@tophatalpha:disqus email me at [email protected]
.This is a great effort and good luck with it.There have to be some better delivery mechanisms developed and quickly. Why every politician is not talking about the absurdity of having health insurance policies compartmentalized by state is nuts.Trump made a good point on this.Why is this so damn difficult?The “pricing” of everything in the hospital, health care delivery pipeline is a perfect example of where a smidgen of competition would nudge things in a better direction.Good luck with this project.JLMwww.themusingsofthebigredca…
Or the absurdity of calling it “insurance”. “Insurance” is for unforeseen accidents so you can hedge your risk. I wish there were another mechanism, not called “insurance”, for managing regular care.
Great to see this topic under the USV lens. I will share the content with healthcare consultants and architects.
The state-to-state healthcare walls are bigger and more secure than anything Trump wants to build at the Mexican border. My insurance policy fundamentally provides me only w/ in-state coverage. Sure, if I’m traveling and need to go to a hospital emergency room I’m covered, but I can’t shop for costly surgical procedures state-to-state. My insurance carrier won’t cover. I can buy a loaf of bread anywhere I want, but I can’t similarly shop for a hernia operation outside of NYS w/out foregoing coverage. Of course, cost is only one variable when seeking med care, but cost transparency and a system that allows interstate medical coverage immediately changes the competitive landscape. Let’s start by unshackling the regulatory handcuffs.
You could certainly buy national coverage? How does a regional health insurance company begin to negotiate on costs with hospitals and providers across the country and keep its costs with the 15% margin ceilings?
MVP, Technology should start where the signal to noise is the lowest, spend a day at LA county hospital, operated by the residents, fellows and attending physicians at USC. It provides free healthcare to hundreds of thousands of Californians. If your product can’t help this group, go back to step one.
Don’t you mean highest?
No! You are just measuring what you already know.
Oh, so you mean finding out new facts and requirements (that you didn’t know of earlier), which may lead you to PMF? or even just P?
You’re right and the tip about starting where signal to noise is lowest applies to other tech too, not just medical tech.I like problems where currently signal <<< NOISE and try to solve it so SIGNAL >>> noise. This means the signal to noise is low at the start and I need to figure out how to make the tools to either tune into the signal and/or amplify it or to dial down the noise.
Separate signal from noise?Yes, definitely use Wiener filtering!Maybe!Look up work by J. Tukey, D. Brillinger with some connections with R. Hamming and C. Shannon.For more? Sure, Kalman filtering. There’s a nice presentation in D. Luenberger (long at Stanford).Does all this classic work apply to the problem you are considering? Likely not!
Thanks, that’s helpful.A few years ago, I did look at Kalman filter — as it applied to Energy Derivatives and entropic modeling.There’s usefulness in the classic tools. I’m not rejecting them off-hand; simply reasoning through WHY and HOW useful each is and also how they are in composite.I’m aware there are missing data inputs. It isn’t just that the mathematical tools aren’t quite a fit for solving the problem, it is the fact that there’s a HUGE dataset which hasn’t previously been collected in a systematic way that lends to the data then being analyzed by mathematic tools, including the classic ones.Someone pointed me to towards Amplitudihedron as an example where scientists have an intuition about the existence of something but it either doesn’t seem to make mathematical sense OR the mathematical tools needed to resolve it don’t exist yet.* http://www.wired.com/2013/1…* http://www.wired.com/2014/0…The problem I’m solving is less obscure and much more commercial than high-end Quantum Physics; although it is connected because in 2014 QP postulated for the existence of a Perceptronium particle “the most general substance that feels subjectively self-aware” (Professor Max Tegmark, MIT) and high-end QP gets applied in areas of Machine Intelligence and Neuroscience modeling.The implications of this would be that our “stardust” of the Universe and of us contains Perceptronium. It would be somewhere within our bodies and beings, maybe our natural neural nets (synapses and nodes).Meanwhile, Sir Roger Penrose and Dr Stuart Hameroff have proposed their “Orchestrated objective reduction Orch-OR” theory for quantum and medical consciousness. Furthermore, their theory argues that consciousness could exist in the microtubules.[I looked into their work because my Dad was in a coma (later died) so I wanted to understand more about where medical research was on consciousness.]Anyway, years before Tegmark published his paper, I set out to make a system to measure and model consumer perceptions.Now, the classic approach in market research would involve one of these methods:(1.) Put out a survey containing:* Yes (agree) / No (disagree) boxes* Ask customers to rate something (1-5 or 1-10 or 1-100% scale)* Ask customers open-ended questions and let them answer freely in textboxes.* Ask customers to do some type of psychometric test(2.) Track their social media activity via cookies, e.g.:* which sites they’re visiting* duration of site interaction* which products they’re clicking the “buy” button on(3.) Run Probability & Statistic modeling on those data inputs with some human interpretation or text mining on what they wrote freely in the textbox.The text mining would be similar to what IBM Watson does in its Personality Insights algorithm or what Natural Processing does with skip grams and Google-Stanford’s Word2Vec and Sentence2Vec.Voilà, supposedly a comprehensive profile of the user.But I believe we can do better than this. We can do better in initial data input classification as well as in decision-modeling tools.I’m taking a leaf from Da Vinci who wrote: “All our knowledge has its origins in our perceptions.”By this, he didn’t just mean optics, machine vision (and their Kalman filters) and the pattern recognition generated by Probability tools. Their inputs could be defined as the “objective” variables.He also means the superposition entanglement of the “subjective” variables. These would be emotions, natural language, culture (discrete: gender, age, education etc; non-discrete: social grouping because often people belong to a mix of groups simultaneously, tastes in movies, music, media, activities etc).This loops back to my previous question about whether the language of mathematics, including Probability+Logic+Statistics+EuclideanSpaces, is sufficient enough to enable us to measure and model emotions, natural language, culture (discrete: gender, age, education etc.; non-discrete: social grouping because people often belong to a mix of groups simultaneously, tastes in movies, music, media, activities etc).If classical maths can’t yet measure and model those factors in a sensible and systematic way, then new tools would need to be invented…From point of data capture (so surveying tools and cookies) to data analytics (filters, database structures, visualizations) to results interpretation and actionable insights (decision-making).Anyway, I always appreciate more pointers on tools from you!
I started my career in applied math and computing around Washington, DC. It was a hot job market: Once I sent a few resume copies and, then, in two weeks went on seven interviews and got five offers. My annual salary quickly went to six times what a new, high end Camaro cost. My wife was working on her Ph.D.; we were saving money rapidly; we had a great time going to high end restaurants, the fancy wine and cheese shops on Wisconsin Avenue in Georgetown, concerts, plays, taking French pastries, baking potatoes wrapped in foil, French bread, some decent grape juice from between Macon and Dijon, porterhouse steaks, and charcoal to Shenandoah, etc. I got a nicer violin and bought her a nice piano. We had two late model cars and were saving money rapidly.So, some of that work was in digital signal processing where I encountered Fourier series, the Fourier transform, auto-correlation, covariance, power spectra, the fast Fourier transform, digital filtering, the Kalman filter, etc.At one point I quickly readR. B. Blackman and J. W. Tukey, The Measurement of Power Spectra: From the Point of View of Communications Engineering.and, then, in a big rush wrote some software to illustrate how to measure power spectra of ocean waves.Just this week happened to see:http://projecteuclid.org/do…which is a PDF of The Annals of Statistics 2002, Vol. 30, No. 6, 1595–1618 JOHN W. TUKEY’S WORK ON TIME SERIES AND SPECTRUM ANALYSIS1 BY DAVID R. BRILLINGER University of California, Berkeley. which is a review of just such things, with lots of emphasis on the fast Fourier transform, etc.IIRC, Brillinger, e.g., as inDavid R. Brillinger, Time Series Analysis: Data Analysis and Theory, Expanded Edition.was a Tukey student at Princeton.There is much more to that literature, and that PDF has a lot.When I was pursuing such things in DC, I had yet to learn probability and stochastic processes at the level of, say,Jacques Neveu, Mathematical Foundations of the Calculus of Probability.Bummer. So, I had a lot of experience and intuition; then when I got the real math, via Neveu, etc., it was terrific.In that field, much of the best literature is from Russians! Otherwise the French!One good, succinct, but mathematically solid reference on Kalman filtering isDavid G. Luenberger, Optimization by Vector Space Methods.So, it’s fun and profit in Banach space, i.e., with the Hahn-Banach result. There’s usefulness in the classic tools. I’m not rejecting them off-hand; simply reasoning through WHY and HOW useful each is and also how they are in composite. Again, some math such as Kalman filtering is like an airline flight in the South Pacific from island B to island C. If you are at island A, want to get to island D, then you can take a boat to B, the airline to C, and another boat to your destination D.The flight from B to C is based on starting with some assumptions, using those as hypotheses in some theorems with some proofs, and then having the conclusions of those theorems get you to C. It’s much the same as just the Pythagorean theorem in geometry: The main assumption is just a right triangle and knowledge of the lengths of the two legs. Then the theorem gives you the square of the length of the hypotenuse. It’s essentially the same for Wiener filtering, Kalman filtering, regression analysis, principle components analysis, mathematical programming, power spectral estimation, optimal control theory, ….The time I was doing power spectral estimation? Some guys wanted to study the missile launch hover control system of the US SSBN fleet. That’s the system that automatically floods/blows the buoyancy tanks in an attempt to keep the submarine at the right depth, in possibly rough seas, for firing missiles.So, they had a lot of data on ocean waves. So, they wanted the power spectra of that data and, then, to use that to generate sample paths of stochastic processes with that power spectra and, then, use those sample paths as input to candidate control systems.Gee, now that I type this, hmm, it occurs to me that they were throwing out the baby and keeping the bathwater! What they really wanted as input to their work was the power spectrum, just the power spectrum, and, then, to heck with the generated sample paths.Or, at sea we have the power spectrum of the process, say, from the sea state, and a real sample path to the present. Now, how do we use those two to extrapolate to the next, say, 30 seconds and, then, say how to flood/blow the tanks based on that extrapolation? Look, I just dreamed this up just typing now; so, it can’t be classified yet; so go chase after Hillary instead! :-)Use that to make money in the stock market? Ask J. Simons since I have to believe that he thought of all such things a long time ago.So, in some classic terms from Tukey at Bell Labs, they wanted to measure noise color, right, much as telephone engineers at Bell Labs wanted to do for Western Electric and AT&T.So, the software I wrote, quickly, less than a week, took data and measured the power spectra and, then, ran white noise through a digital filter to generate sample paths with the measured power spectrum.One of the lessons, explicit in Blackman and Tukey and illustrated clearly by my software, was just how much data it took to get an accurate estimate of a power spectrum. It turns out, at low frequency, really, at high resolution at whatever frequency but at low frequency typically want a lot of resolution, it takes a lot of data.My fast work impressed the customer, and my company won a nice software development contract.Later when I was considering a Ph.D. program, I noticed that so far in my career I’d studied Tukey’s lemma, a version of the axiom of choice, and his convergence and uniformity in topology, stepwise regression, power spectral estimation, and exploratory data analysis so wondered if there was a career in such things and a suitable Ph.D. program so wrote Tukey and asked him. I got a nice letter back from G. Watson claiming that there was a good career and that Princeton had the best program. Net, I got accepted to grad school at Princeton.From some other work I’d done, I got accepted similarly to grad school at the Division of Applied Math at Brown, the Operations Research program at Cornell, and Mathematical Sciences at Johns Hopkins. Since my wife was already in grad school at Hopkins, I went there.Q. So, whether to use some math?A. Do you have the assumptions for island B? Then, would getting to island C help?For Now, the classic approach in market research would involve one of these methods: and * Ask customers open-ended questions and let them answer freely in textboxes. Okay, I brushed up against such things, if only from my wife who got her Ph.D. in essentially mathematical sociology, e.g., with a lot of emphasis on survey data and its analysis with statistics. One of her profs was J. Coleman, and another was P. Rossi. Each was at one time President of the American Sociological Association.At a party in her department, by then I’d encountered enough in uses of applied statistics to remark to her and Rossi that what they were doing should be terrific for careers in marketing research.I was correct: Today she could be a leading person in Internet ad targeting.From listening to my wife talk about surveys, usually the “open ended” questions were to identify what additional issues might be addressed with additional multiple choice questions. So, they didn’t try very hard to analyze the responses to the open ended questions for direct application of the analysis. Sure, today might look for key words in the text boxes and try to do something with those.Another big point from her work was the challenge of the steps:(1) Guess some multiple choice questions that may be relevant. Sure, the usual suspects are age, education, and income.(2) Ask the questions and hope that people answer honestly.(3) Use the answers to build a model of their interests. “consumer perceptions,” categories, e.g., young, urban professional, Chablis and Brie set, dual income, no kids (DINKS), drives a Toyota Prius, gluten-free, 100% organic, etc.(4) Use that model to predict their buying behavior for various products and ads.So, it’s several steps with errors at each one. Bummer.In the ad targeting for my project, my approach is simpler, more direct, with fewer steps and no logically intermediate theories about or categories of human behavior.The goal is essentially just to get users to click on ads (e.g., without click fraud). So, (A) see what users click on what ads. Also, (B) as a byproduct of the main work of my project, have some data on the users. Then use (A) and (B) and some math I’ve derived to determine what ads to show to what users to maximize my revenue from clicks. So far my Web site is not using cookies, and generally I’m working hard to protect user privacy.For such work in applied statistics, keep in mind that, given enough data, essentially the most powerful statistical technique is just simple, old cross tabulation. We use something else when we don’t have enough data to use cross tabulation (it’s easy to find cases when the amount of data that would be needed for cross tabulation would be absurdly large, even for big data) or where we want a simple, intuitive explanation. Still, cross tabulation should be used more often.Uh, yes, my ad targeting math is not cross tabulation!
Thanks! I’ll get reading and applying!
I hope people use better passwords than usual for their healthcare information… https://www.youtube.com/wat…
Being in Nashville (the self-proclaimed US healthcare capital due to the large hospital management companies here) and having spent 20 years in Silicon Valley, I am glad to see this type of analysis and focus. I’ve recently watched local incumbents grumble about ACA and then see the light. But as others have said in the comments, most of the problems in healthcare are because of fiduciary responsibilities, not patient care. So you have a choice – work within the system for scale yet be strangled by the resistance to change, or work outside the system and be strangled by lack of scale. I am actively watching companies in both spaces and learning from them all. I look at companies like Change Healthcare and Castlight and see some progress on transparency and engagement, but in the end most progress will require changes in patient attitudes towards price/quality/convenience/trust and the payers’ support of new ways of solving health problems.I hope they also look at the role that CMS (Medicare, Medicaid) plays in all of this. It is extremely hard to evaluate risk/return opportunities when the government has such a long lever.
Can you fill me in on lifestyle in nashville? My gf is considering an onc fellowship at Vanderbilt.
Would love to offer my west coast-biased opinions about the Nashville lifestyle. Connect on LI and we can talk further. https://www.linkedin.com/in…
I’m headed to Nashville tomorrow to attend the Health Further event. Would love to learn more about the healthcare scene there.
I’ll be at the conference. Connect on LI and we can figure out how to meet in person. https://www.linkedin.com/in…
Definitely very astute insights. Particularly on CMS. We spend a huge amount on seniors (obviously we will spend a ton as they are generally sickest) and it is pretty inefficient in many cases, and while private insurance pays more generally, CMS is the baseload of pretty much every hospital system so they have massive power. Unfortunately often times they wield it ineffectively, one reason being politics obviously. One decent thing they have started is making much more of their data open / available to the public. They can serve as a replacement to CPT and proprietary coding that is so messed up, and i would fully support the federal govt. funding a new coding system.
Yup. Also the FDA – anything that gets attached o a phone for medical testing in the long term will need FDA clearance as a medical device. Which means the FDA and the Apple store will have to look through code and firmware upgrades every time you need o make a change (the FDA even more often if you do anything on your hosts)
I’m so excited to follow this series with all of the learnings. By far one of the fields that I am most interested in.
Ah, yes: Definitely multi-objective optimization!Right?Maybe.So, want a point on a non-inferior set.There’s a book on that by Jared Cohon, one ofmy dissertation advisors, and long and until recently President at CMU. Yes, I took his course!
The key is transparency in healthcare. Patients need more control over medical data, prices, treatments.. More data synchronized between doctors, hospitals and patients means more data available and more data transparent. Major issues are: privacy and sharing, identity and ownership. Here blockchain tech might be beneficial.
Agree on transparency. I’m not sure the blockchain provides any more value than modern, open API’s would.
depends for what. While it is totally cheaper for me to have my glasses die an honorable death in India, or suddenly decide I hate my nose and go to south korea, if I needed sustained care, it isn’t reasonable to go to india or south korea