Several communities in the country face different challenges when seeking health care options. Criteria for selecting a provider can range from anywhere such as distance to an individual's medical history. Simple google searches only involve hospital types and overall ratings, and they cannot grant specific and relevant information, especially in cases affecting those of low income or minority communities. For example, Native American communities have lower life expectancies, different medical conditions, and less access to healthcare than the general public and could benefit from more specific searches. African American women face a rate of death during childbirth at a rate of three times higher than the general public. Sickle cell anemia affects African American communities at a rate much higher than the general. Low income communities benefit from locations considering their budget. Targeted searches are crucial to receiving proper medical care especially taking into account complication rate. Also, the general ratings of hospital do not reflect a hospital’s performance in different medical fields; hospitals can and do perform differently. For example, a hospital may rank overall lower than another hospital but does better in a particular field, e.g., pulmonology.
What it does
The product is a website which allows users to enter in one or more fields including procedure, medical history, age, sex, race/ethnicity, location, etc. Upon clicking the search button, results of the best matching nearby health providers appear with location, match rate, complication rate, estimated cost, and top physicians.
How we built it
We built our website off the CMS Synthetic dataset (Medicare claims that are de-identified). We uploaded the Carrier, Beneficiary, Inpatient, and Outpatient datasets from CMS Synthetic Public Use files to GCP’s Cloud SQL service. We queried the database to search for patients with similar chronic conditions to the user’s as well as gender, age range, race, and procedure. We aggregated various performance metrics for each facility and physician (such as 30-day readmissions for complications, death rate, and cost). We matched each physician with their respective facility and ranked them on overall performance, which we also used to create a match score for the facility with the user. The queried results are displayed on a domain.com website, which also features the use of SnapKit’s Bitmoji feature. Users query on the compare page, and the results are displayed on the results page.
Challenges we ran into
We had initial issues with query speed, so we used GCP’s Cloud SQL service to hasten our queries. We also had issues with updating our domain, but those issues seemed to have resolved.
Accomplishments that we're proud of
We are proud of our integration of multiple technologies provided by the sponsors, and we hope we innovated something new to be used by users around the country to locate healthcare facilities and physicians targeted towards their needs.
What we learned
We learned the humanitarian aspects of healthcare equity that go into providing targeted care to individuals. To do so, we used technologies encouraged by MLH in a unique way to summarize and bring the confusing Medicare database to life.
What's next for HPL
Currently, we are using synthetic data; however, in the future, HPL intends to use real CMS patient records which would update every three months instead of using years old data. Additionally, we will implement an appointment feature which allows users to make appointments with the recommended healthcare providers directly through the website.