So it’s ‘hump day’ of National Health IT Week and we’ve already talked about how EHR’s are capturing a treasure-trove of rich data, and the burgeoning enthusiasm for analytics and predictive modeling that is nature consequence of having this data.
But what about the whole “big data” thing? I’ve been on the fence as less than enthusiastic about all the hand waving and bell ringing surrounding the big data movement in healthcare, simply based upon our inability as an industry to really do anything particularly useful with the “little data” we already have. We need look no further than all of the angst and heartfelt bickering over the very modest data requirements of Meaningful Use Stage I requirements to validate this assessment of our industry readiness for “big data”. This isn’t saying that there have not been pockets of incredibly talented individuals, applying extraordinary effort, to do cool things with data, and yielding some glimmers of hope for “big data”. But as an industry, we have not really done much with the data we already have to understand what works, and doesn’t work, and use that insight to change our behavior. And when it comes to data, I don’t think you can run before you can walk, and in my mind big data is Olympic-caliber running. It’s also important to consider my opinion only in the context of strongly structured discrete data like lab results and coded clinical data entered into EHRs. This distinction is relevant since when it comes to things like digital imaging, PACS, and advanced visualization algorithms, the broader healthcare market has arguably been dealing with “big data” for more than a decade.
But I’m beginning to change my mind on big data in healthcare. And rather than walking before you run, big data may present on opportunity to leap ahead in deriving value from data in very focused areas, even before full competence in ‘small data’ analytics has been achieved. The reason for my change in thinking is really related to (a) an evolving understanding of how big data technologies can be applied to existing data problems, and (b) compelling new sources of data, and potential solutions, that have never before been possible.
Big data isn’t just about doing analysis of twitter feeds and facebook posts (which encompass all three V’s of big data – volume, variety and velocity) it can also be about having the cost-effective processing horsepower to do much more sophisticated analytics on the clinical or billing data we already have. Or the data that we’re going to have from our EHRs. Rather than testing clinical or financial models against a month’s worth of data, or a quarter’s worth of data, now we can test those same models against a year, or five, or a decade’s worth of data. This same processing horsepower means analysis that might have taken days or weeks can now be done in minutes or hours, which makes the results that much more valuable in impacting clinical care and changing frontline staff behaviors. And to the extent these big data technologies can be adopted and applied to today’s data analytics needs by mere mortals, then all the better. For example, Hadoop has been a centerpiece in the whole big data hype circus, and historically has been a complex beast that can only be mastered by the most advanced and sophisticated IT shops. And who in their right mind with all the other challenges facing healthcare IT (ICD-10 conversion, EHR implementation, HIPPA privacy audits, flat budgets, health information exchanges, etc.) wants to try something like that? But what we’re seeing is a maturing of the Hadoop technology stack and a vendor ecosystem developing around the platform with solutions that make it much more practical that healthcare organizations will be able to try out some of these big data solutions. For example, Informatica has recently announced support for Hadoop such that any transformations or data quality rules created in Informatica can be run on Hadoop unchanged – taking advantage of the lower cost and higher performance of the Hadoop platform while avoiding much of the complexity and specialized skills that have traditionally been a huge barrier to adoption. One potential area this sort of approach could be applied is crowdsourcing medical decisions – a topic I have previously written about that I think has very real implications for the providers aspiring to become “learning healthcare organizations” which you can read here.
There was also a good report from InformationWeek titled 2013 Healthcare IT Priorities that observed a growing proliferation of data coming from mobile devices and personal medical monitors, which in my mind will inevitably hit all three V’s (volume, velocity and variety) that traditionally define big data. They further state that we are rapidly heading towards a future where acquiring data ceases to be the problem, but figuring out what to do with it becomes the real challenge.
In this same vein, I also believe that some of some of the more consumer-oriented “big data things” have potential promise in healthcare. The creation of an individual insurance market is going to drive a tectonic shift in the perspective of payors as they reorient their sales and marketing from selling coverage in large chunks to employers, and instead need to understand and target the far more finicky individual consumer with a very different perspective on value and customer service than their traditional employer buyer. In this situation, all those social media feeds and the sentiment analysis they can reveal become very compelling with a demonstrable ROI, as does old-fashioned analytics on big data such as web click-stream analytics to optimize the consumer experience on their websites.
There is also clear potential in combining consumer data with clinical data to provide a more complete 360 degree view of the patient for providers – bringing key information such as what over-the-counter drugs a patient may have bought over the last 30-60-90 days that may have a marked adverse reaction with a prescription they are taking, or just be clinically undesirable (such as someone with hypertension taking an OTC antihistamine for example). Providing this insight to their provider at the time of the patient’s next visit – or better still, apply rules in near-real-time as soon as the data becomes available and alerting the patient’s care team even when no visit is scheduled – can really change the role of physicians in orchestrating a patient’s health and wellness rather than simply treating symptoms and disease during an office visit or hospital admission.
Tomorrow we will move on to what is motivating healthcare providers to finally take a genuine interest in analytics and business intelligence, with that thing being align financial incentives.