⭐Introduction to our API

Overview

Abstractive Health offers an API solution that helps you gain deep insights of your patients: (1) summarizes clinical documentation (2) retrieve a patient's full medical records nationwide through health information exchanges, and (3) query the record for any question you may have to take action. Leveraging our advanced, clinician-grade AI, we generate comprehensive summaries that condense the essential information from medical records and give you access to the underlying structured data as JSONs, CDAs, and FHIR. Our API processes file types of (i) PDFs (including scanned and handwritten documents), (ii)CDAs, and (iii) FHIR; and summarizes it all using our advanced AI model.

Retrieve data from the Health Information Exchange (HIE)

Our API allows you to retrieve medical records nationwide through our integration with Carequality®, giving you access to data from major health systems like Athena Health, CVS Health, eClinicalWorks, Epic, NextGen, and PointClickCare. You'll have comprehensive access to available medical data, including progress notes, discharge summaries, consults, procedure notes, lab results, immunization records, radiology reports, scans, x-rays, and more. Depending on the volume of medical data, it typically takes between 2 to 4 minutes to retrieve all documents per patient search.

Summarize the full record - including scans and PDFs

Our API also enables batch processing of healthcare records, including CDAs, FHIR, and scanned documents and handwritten notes, with summaries generated in under 2 minutes. We can handle records up to 1000 pages.

Our technology for generating AI summaries

We generate our summaries using a combination of state-of-the-art open-source models, such as Clinical LLaMA and Clinical GPT, alongside proprietary technology. Through our clinical research at Weill Cornell Medicine, we have demonstrated our AI generated summaries are effectively near comparable to those written by physicians. In addition, our technology was awarded third out of 200+ companies by the U.S. Department of Veterans Affairs as a leader in trustworthy AI solutions.

The content and length of each summary depends on the original documents being summarized. Typically, summaries average 500 words and follow this structure:

AI Sample Summary

Patient Name:

Tommy Abstractive

History of Present Illness:

Tommy Abstractive initially presented to the emergency room on 12/17/2018 with left ankle pain after he twisted his ankle while playing basketball. On x-ray and CT, he was found to have a nondisplaced medial malleolus fracture, displaced posterior malleolus fracture, and medial clear space widening along with mid diaphyseal fibular fracture. He was reduced and splinted in the ED. He underwent ORIF of the left ankle four days later. A follow-up x-ray on 02/06/2019 showed a healed ankle fracture with intact hardware in acceptable alignment. Since surgery, Mr. Abstractive has been attending physical therapy, doing home exercises, and is making good progress without pain or a limp with ambulation.

Allergies:

Penicillin
Anticonvulsants
Sulfonamide antibiotics

Family History:

Tommy's clinical family history reveals a mix of medical conditions, including heart disease, hypertension, osteoarthritis, and type 2 diabetes among immediate family members.

Tommy presents as a 32-year-old male with a stable social history. He reports occasional alcohol use and identifies work-related stress as his primary challenge.

Medications:

Ibuprofen 400mg q 4-6 hours. Administered 12/17/18
Acetaminophen 650-1000 mg q 4-6 hours. Administered 12/17/18

Past Medical History:

Orthopedics consult and reduction/splinting of fracture in the ED on 12/17/2018 ORIF of the left ankle on 12/21/2018 Ice, elevation, and ace wrap for swelling Weight bearing as tolerated with a camboot Follow up x-ray on 02/06/2019 Physical therapy, including range-of-motion and other at-home exercises

Follow-Ups:

Increase activity as tolerated
Elevate/ace wrap for swelling prn
Continue PT; new script given
Continue at home ankle exercises
F/U 3 months with left ankle X-Rays.

When generating the summaries, we send you coverage scores that represent how similar the summary sentences are to the sentences in the original content. The scores range from 0 to 1. We internally consider anything below a 0.4 as a low score and scores above 0.4 as good. Coverage scores are not a measure of clinical accuracy, but can be used as a rough threshold to determine if you desire to implement an internal human-in-the-loop workflow before presenting the end results to a clinician.

Further, the coverage score output can be used to map a summary sentence back to the best sentence within the source data. We provide the 5 best potential sentences from the original source data. The output will provide source sentence and source section numbers.

Generating a comprehensive patient summary

To generate comprehensive summaries from patient data via Carequality® (i.e. what we provide in the standalone UI application of Abstractive Health), you would chain the followings APIs together once each one is successful:

Get token
Search patient
Retrieve patient documents
Request a summary (using documents you retrieved from Carequality®)
Retrieve a summary
Return treatment note to Abstractive Health

We allow this versatility so that you could combine both Carequality external documents and your own local documents of a patient for summarization. To do this, you would include other documents that you possess in Step #4 in addition to the Carequality documents returned in Step #3.

What you need to get started

You don't need an engineering team to start using our API, we designed it in a way that anyone with basic coding skills can get started by following our step by step guide. You will be able to try the API for free and see if it fits your use case. If you have any problems or additional questions you can contact us at [email protected].