A healthcare data platform is software that brings a hospital group's or an insurer's data together in one place, cleans it and matches it to each patient, so that it can be used for care, payment and research. Its buyers are mainly hospital groups and health plans in the United States. A health plan is what Americans call a health insurer. A large share of these platforms is built by engineers in India.
A hospital group's data is spread across many systems. Each hospital and clinic may run a different electronic health record. Labs, pharmacies and imaging centres keep their own systems. The insurer's health plan administration system holds the claims, the bills that the insurer has paid. A platform collects copies of all of this every day, through FHIR APIs, file transfers and even faxed documents.
Collecting the data is the easy part. Cleaning it is harder. First, the platform must match records to people. One lab writes "Smith, Mary A", a clinic writes "Mary Smith", and a claim carries only a member number. The platform compares names, dates of birth, addresses and phone numbers, and decides which records belong to the same person. This is called patient matching, a form of identity resolution. Second, it converts every code into one standard. One lab calls a test "HbA1c" and another calls it "glycated haemoglobin", but both must be stored as the same test. The platform keeps the cleaned data in a data lakehouse, one store that can hold both raw files and clean tables.
Take, for example, a health system in North Carolina with eight hospitals and 120 clinics. Insurers pay it a bonus if its patients with diabetes keep their blood sugar under control. This arrangement is called value-based care: the provider is paid partly for its patients' results, not only for each visit. Every night, the platform matches 18,000 patients with diabetes across all its systems. It finds 2,300 of them who have had no HbA1c test, the test of average blood sugar, in the past year. The list goes to the care managers, nurses who follow up with patients by phone, through care management software. Without matching, about 300 of those patients would have appeared twice under two spellings, and others would have been missed.
The same matched data also serves research. With names and identifying details removed, it becomes de-identified data. It can then be used to find patients for a clinical trial or to study how well treatments work. HIPAA, the American health privacy law, sets the rules for both uses.
A healthcare data platform differs from a general-purpose data warehouse. It already understands health data: the medical codes, the claim formats and the privacy rules. So a hospital does not have to build that knowledge itself. Newer platforms add AI assistants for care managers, and read the doctors' notes through clinical data abstraction.