Can big data reduce healthcare treatment costs by determining which treatment regimes are cost effective on a customized individual basis?

I think you might be hard pressed to find a doctor that makes any treatment decisions that aren't informed by proven research trials. Doctors don't change their approach based on every trial that comes out though, he or she must determine the quality of the trial and whether or not it actually applies to their patient. For instance, there is a new "blood thinner" on the market called dabigatran that has been proven to be as safe as warfarin but does not require close monitoring (patients on warfarin can have serious consequences if they become under- or overanticoagulated) but patients with certain anticoagulant indications have not been studied and so warfarin is still the mainstay of treatment for them. The patient with back pain didn't just show up and get wheeled straight to the OR so that his neurosurgeon could make extra money but without going through necessary steps as dictated by specific guidelines. The patient had to first get an X-Ray to assess for bony abnormalities or joint space narrowing and then go on a 6 week trial of NSAID (such as Advil) therapy with physical therapy before he could even get an MRI. The only people who can get an MRI without going through above measures are people who show signs of significant neurologic damage such as muscle weakness or spasticity, increased reflexes, loss of bladder control, ect. The only people who can qualify for surgery are patients whose MRI shows herniation into the joint space compressing the spinal cord. It's possible that the patient you spoke of had severe herniation in his C-spine and milder herniation in his lumbar spine (surely the doctor would have chosen to do the more severe if the two first). The spinal fusion he had could have relieved some of the pressure on his spine in general, which could account for the lumbar herniation becoming less symptomatic. Maybe his C-spine was in pain due to spondylolisthesis, in which case the surgery could have been very necessary if it was severe. In fact, without knowing what that individual patient had, it's difficult to make an argument either way with regards to following guidelines and/or individualized medicine. I guess to me, following guidelines is really a guideline. It's what you should do for most patients. If the only thing doctors did was follow guidelines, well, we wouldn't need doctors, would we? You can find examples of doctors that negligently do the most expensive treatments but I have yet to meet one like that. Then again, I haven't had too much experience as of yet. Just keep in mind that there are a lot of factors going into any clinical decision and it's difficult enough to accurately judge something when you've got the training and access to all of the patients medical history- and nearly impossible without. Imaging is getting more precise, lab tests are becoming more specific, we are moving towards individualized medicine but the transition isn't cheap. Perhaps one day it will mean less spending, but by then there will be something even better on the market. If you want to cut costs on your medical bills, eat right and get plenty of exercise! And go for an annual physical. :)

It will be very very difficult. I have some experience with this. A few years ago Illinois Workers Compensation Act added a Utilization Review provision that allows the employer/carrier to challenge treatment which falls outside evidence based guidelines. The  treating doctor would propose X, contact the carrier for authorization,  and an escalating series of reviews  under Utilization Review Guidelines (URAC) would occur to certify or non-certify the care. Sometimes this resulted in more conservative care before proceeding with the original procedure, sometimes it resulted  the patient getting a 2nd opinion and litigating ( read, incur claim expense which is different from medical expense, but still a payout of money). If the patient had group medical coverage, sometimes he ignored the UR non-certification, got the care he wanted and fought with the employer/ WC carrier at the end about which carrier should bear the cost finally. In one instance , we had a man with a definite lumbar and cervical disc injury. Surgery was non-certified for both levels. he went through group, had his neck done first and while he was recuperating, his low back resolved... without the additional, contested surgery. But that is a single instance that I can recall, and I have handled or touched a   decent volume of claims and cases. Treating doctors offices have hired support staff to deal with the certification process. UR companies have  been created to implement the guidelines. Carriers have added staff to deal with this issue. What I have not seen is a bottom line cost comparison of how the claim expense administrative costs compare to the savings for non-certfied treatment. (I am sure it is out there. There is an organization which tracks this. I just haven't seen it. If someone can link to a report, I would be  grateful). I have seen egregious abuses of unnecessary, ineffective but profit making procedures and care (unfortunately chiropractors and unscrupulous physical therapists, and independent MRI owners are the ones I saw most). Something needed to be done to address that abuse, but there is a strong  reluctance from the legitimate patient and the legitimate treating doctors to have a disembodied non-medical  algorithm tell them what they can and cannot do to  treat the patient. So you can gather all the data, perform all the studies, identify all the biases, but it will come down to  figuring out what to do with the flesh and blood guy in front of you whose back still hurts and who, in the doctor's opinion, needs 6 more weeks of Physical therapy beyond what the Big Data says. I'd love to see some  braking mechansim happen, but it would be a   generationally slow process to change a lot of people's mindsets who don't want something Big telling them what to do. (Also, I have not heard  any  discussions about whether always sticking with data supported care makes for a closed system which   prevents finding out if some alternative course of care is better, cheaper, more effective. I'll leave that one for the medical statisticians and epidemiologists).

I won't see big-data driven diagnostic in my life time.  Here are three reasons: 1. Using population statistic on the diagnostic of one person is fundamentally illogical. It's like saying to me: "The average life span in the U.S. is 80 years.  I predict you will die on your 80th birthday." Well, I'm quite likely to die earlier, or I may die later, but the chance of me dying on my 80th birthday is actually really small! Now let's apply this algorithm to an individual patient who came in with a fever.  Your data set says: "80% of the fever reported is caused by a cold, no treatment needed.  10% is caused by an infection, you need to prescribe antibiotics. 5% is caused by...." So what do you do with the next patient coming in?  You'll say "you have a cold. Go home, rest, and drink a lot of water." And this machine doctor will diagnose 100% of his fever patient as having a cold, because that's what the data says if you apply population statistics on an individual.  There are so many of these type of pseudo-scientific, but fundamentally illogical claims out there, sometime I wonder if people just need to go back to high school and take a math class.  You can NOT use population statistics on an individual! 2. In order to make a right diagnosis, you need to know at least 2 things: a), the disease and its pathology, and b), the individual patient. Each person reacts differently to the same pathogen.  Some people get high fever, some people has low fever accompanied by a skin rash. A family doctor will often know, if he's been treating the individual for a long time, that "my patient X has had flu, chickenpox, etc., and he never had a high fever with them. But he has these symptoms", so then even a 101 degree fever may be alarming to the family doctor and he will insist the patient coming in for a thorough examination.  In a word, you need to know the patient.  If you are missing patient information, you can't make a right diagnosis. For "big data", that means you need to collect a vast amount of data about each individual, because you don't actually know which data is going to be important. You'll first need to set a baseline about the individual: what did he eat, what did he drink, how many times he pee'd, what is the composition of his pee, what is his blood analysis during this time, what is his brain activity, etc... and then you need all these data sets again for the rest of his life because people change.  I'm not making things up - all of these data points actually are useful for specific diagnosis, but I'll be shocked if anybody let you collect these data from him. It's 1000 times worse than the NSA, and you have less ability than the NSA to keep their data secure. This is basically treating people like lab rats. Maybe in an Orwellian novel you'll have all newborn implanted a chip to collect their biological data and beam it to your database for future diagnostic, but the potential for abuse is so great I don't expect this to ever happen.  3. Most of the data sets available are garbage.  The way most of these data sets were created was as follows: a researcher goes out to collect data, he doesn't really know what were the drivers for his topic.  He had have some hypothesis. So he went out collecting the usual demographics data, plus the key variables to test his hypothesis.  He won't collect a lot of data not directly related to his topic because it was expensive to collect these data.  Now he got something he can publish.  That's pretty much the "big data" those guys are talking about. Same with Pharma companies. They only collect data to prove a point. Anything else is unnecessary. Even worse, what if you collected some unnecessary data and someone ended up with an "unrelated adverse event"?  The FDA will ask you to do a thorough investigation, recruit more patient to your clinical trial to prove this adverse event is indeed unrelated to your product.  You just shoot yourself on the foot.  That's why pharma companies never, ever, collect clinical data that is not absolutely necessary to get a product approved! But for a diagnosis, you often need hundreds of variable to account for hundreds of possibilities.  Nobody has collected any data set with all these variables yet.  If you want to start collecting all these data now, you can but it's a very expensive endeavor. So what about the incomplete data sets you already have? Honestly, they are garbage.  It's like someone trying to paint a garden when he himself can see only 2 color, and even worse, he's trying to predict what his neighbor's garden will look like based on what he sees in his garden. Personally, I think enabling individuals to take better care of themselves will do more to cut healthcare cost than anything else. http://www.cosmesys.com

It can, with one condition: you get the correct data for the treatment regimens you have in mind. The issue is that the data are hidden in the Big Pharrna's, the instrument makers' , the clinics' and the hospitals' accounting departments which are not required by the laws to be transparent in their reports. Garbage in, garbage out.

Certainly, although there is a lot of lower hanging fruit when it comes to reducing healthcare cost. Big Data is typically a Round 2 effort for cost-effectiveness. There are several well designed trials on ways to reduce the cost of healthcare without using patient-level information, say in shifting patients from more expensive settings of care to less expensive ones. Our customers at CliniCast often have VERY good ideas on how to manage the cost of patient care, say in standardizing an episode of cancer treatment, and look to us to help them standardize that operation across the organization. As someone who builds fancy Big Data systems for a living, my hypothesis is there is more money to be saved in consistently applying relatively simple algorithms to a broader population than looking for additional opportunities on a per patient basis. The reason for this, as stated in the other excellent answers, is the amount of missing data and the operational frictions in applying different protocols to similar patients.