Health data interoperability is the ability of different health systems to exchange a patient's records in a shared format, so that each doctor can see the whole history. Without it, a patient's data stays scattered. Consider a 64-year-old woman in the United States. Over ten years, she has seen a skin doctor, a family doctor and a heart specialist, and she has had blood tests at two labs. Each of them keeps its own records in its own system. Her insurer holds a record of every claim. Some reports were sent by fax. Nobody has the whole picture.
Healthcare produces more data than almost any other industry, and uses less of it. The companies in this scene gather the data, clean it and make it usable, for three purposes: care, payment and research. Most of their customers are American hospital groups and insurers, and much of the software is built by engineers in India. The products fall into three groups.
The first group moves records between systems. FHIR APIs let one application read from and write to a hospital's records in a standard format, called FHIR. American government rules now require record systems to offer such APIs. But much of American healthcare still sends documents by secure cloud fax, because email is not trusted with patient records. Medical record retrieval fetches records from other hospitals through whichever channel each one uses.
The second group brings the data together. A healthcare data platform collects a hospital group's or an insurer's records, claims and lab results in one place, and matches them to each patient. Provider data management keeps clean data about the doctors themselves: who practises where, and which insurance plans each accepts.
The third group reads the records. Most of a record is free text: notes and reports written in sentences, which older software could not use. Clinical data abstraction uses language models to read these notes and turn them into data. Reading clinical text with software is called clinical NLP.
Here is what happens for the woman. Her new heart specialist in Michigan asks for her history. Provider data management supplies the current address and fax number of each of her past doctors. Her family doctor's records arrive through a FHIR API. A small skin clinic that has no API sends its pages by cloud fax. The data platform matches everything to her, although one lab spelt her name differently. A language model reads eight years of notes and finds that she stopped one heart medicine in 2019 because it gave her a cough. The specialist sees the whole history on one screen.
All of this must protect the patient's privacy. Records are encrypted when they are sent and stored. Researchers usually receive de-identified data, with names and other identifying details removed.
India is building these connections nationally. Under the Ayushman Bharat Digital Mission, each person can get an ABHA number. With the person's consent, records from hospitals, labs and apps can be linked to it and shared in the same FHIR format.