Provider data management is the work of keeping a health plan's data about doctors correct: which doctor practises at which address, which specialties each one has, which hospitals each one works at, and whether each one still accepts patients on that plan. It is done mainly in the United States. There, a doctor, clinic or hospital is called a provider, and a health insurer is called a health plan. The same software also cleans the data about the plan's members.
A health plan publishes a directory of the doctors in its network. Members use it to find a doctor whose bills the plan will pay. The same data decides how the plan pays each claim. But the data goes out of date all the time. Doctors move clinics, retire, change phone numbers and stop taking new patients. Audits by the American government have found errors in a large share of directory entries. A member who calls a listed doctor and learns that she left two years ago cannot get care in time. Worse, the member may visit a doctor outside the network by mistake and receive a large bill. American law now requires plans to check their directories every 90 days.
The software checks many sources at once. It reads the doctors' own updates, the credentialing files, the national register of doctor ID numbers, state licence records and the plan's own claims. Claims are a strong clue: if a doctor has billed only from a new address for six months, the directory is probably out of date. Machine learning models compare all the sources and give each fact a confidence score. Where the sources disagree, the software flags the record for a person to check, often by phone. Fixing and filling in records in this way is called data quality and enrichment. Keeping one correct record for each doctor across all systems is the job of master data management (MDM).
Say a health plan in Illinois lists 40,000 doctors. One of them, a children's doctor, is listed at two clinics. For the past eight months, every one of her claims has come from only one of them, and the other clinic's website no longer shows her name. The software marks the second address as probably wrong, with a confidence of 92%. An employee in the plan's provider data team phones that clinic, confirms that the doctor has left, and removes the address. The corrected record goes into the plan's health plan administration system and its online directory.
Provider data management differs from credentialing. Credentialing checks, once and then every few years, that a doctor is qualified and licensed. Provider data management keeps the everyday facts correct between those checks. Both feed the same directory. Clean doctor data is one small but necessary part of interoperability, because a record cannot be sent to a doctor whose address is wrong.