I attended a seminar hosted by the Department of Epidemiology and Biostatistics and Division of Cardiology at UC-San Francisco. It explored the possibilities and challenges that Big Data presents in the area of Healthcare. I attended to learn more about what types of use cases are being considered in this area.
One speaker talked about testing for treatment effectiveness, and noted that traditional methods involve experiments on predetermined sample sizes with a given “condition” (e.g. high blood pressure) – classic hypothesis testing to see if the treatment is effective or not. However, testing in this manner may not necessarily be efficient.
It is possible for one person who has a given condition to NOT require treatment. It is also possible that someone who requires treatment does not get it, simply because that person was inadvertently excluded from the sample size taken. The results are mixed…
– you offer a number of people an effective treatment, and the condition dissipates…
– you may incur unnecessary healthcare cost prescribing it to someone that ultimately does not require it, and…
– you miss that one person that does need it.
The promise of Big Data in Healthcare, as I see it, goes beyond two frequently cited use cases: 1. improving population health, and 2. decreasing healthcare costs. I suggest that a complementary “use case” is kept in mind – ensuring that an individual who needs treatment is accurately identified.
The Healthcare industry today has more datatypes available at their disposal – from sport-related wearables, social networks, Web-based surveillance, etc. The data can enable medical professionals to derive a “holistic” patient profile to improve identifying individuals requiring some “treatment”.
An improved profile is key to achieving the two use cases noted previously while ensuring that medical professionals continue to focus on individuals as best as they can. That is not without its challenges, and I will address that in my next blog.