semantic-health-main-cropped.png
MIDS Capstone Project Summer 2016

Semantic Health

Finding and purchasing healthcare insurance on the health insurance exchanges brought about by the ACA 2014 laws is a difficult and frustrating experience for millions of Americans. Many have never had insurance, aren't familiar with comparing health plans, or understand what "in-network" means. They often have existing chronic conditions for which they've been undergoing periodic treatment from their doctor or prescribed a treatment for which they now need to find coverage. These can be challenging tasks for people that have been fortunate enough to have employer-purchased health insurance plans, so it easy to understand how it might appear to someone who has never dealt with the labyrinth of fine print of insurance coverage.

Healthcare exchanges exist at the Federal and State levels, and there are many commercial sites that provide online assistance to finding healthcare insurance coverage. All these sites search for healthcare insurance coverage by cost, and while that is an important dimension, there is no way to find and compare coverage that includes an existing doctor, or certain drugs, or a set of specialists. Once someone enrolls in the healthcare plan they are locked in for a year, regardless if they find out the coverage doesn't work for them, or doesn't cover what they actually need.

We can do better. What would it be like to shop for insurance the same way you shop for luggage on Amazon? You might be brand-conscious, or you could be interested in certain sizes, whether it qualifies as "carry on", different colors. Ratings by other purchasers are important and the "people like you purchased these..." suggestions can be very useful. There is a lot of data science going on behind the scenes of the Amazon product search pages and we can use similar ideas to make the purchase of health insurance much more transparent. Type in a doctor's name and you'll filter the plans by providers with that name that are in-network. Enter a chronic condition or disease, and the auto-complete suggestions that make it easy to enter the often cryptic names filter to show plans that cover those conditions or have specialists in those areas. Narrow your price window or other selection criteria. Select a specialist type and see a map of specialists near you... See what others have to say about certain doctors in the insurance plan or how quickly an insurer helps to resolve issues. These are familiar ideas to anyone that has shopped for consumer items on the internet marketplaces like Amazon. They provide a level of transparency that is not seen in today's health care and they make it much easier to understand what is available and what is the best purchase for an individual.

SemanticHealth uses Elasticsearch and machine learning algorithms to make it easy to surface insurance plans in a meaningful way. Click through data is captured so that queries can be evaluated for similarity and plans associated with different sets of needs. Over time the machine learning models improve the recommendations and make choosing health exchange insurance programs much easier. There are possibilities for sponsored search and data that currently doesn't exist on how consumers evaluate health care plans. As a technology platform the sky is the limit: chatbots to assist with details, mobile voice interfaces and more.

artboard_1.png

A better solution that uses Semantic Search and Machine Learning

Last updated:

March 30, 2017