# How Dmand AI builds its healthcare data, from CMS records to verified contacts

Dmand AI starts from public registries and Medicare files, adds its own executive, org hierarchy and technology data, keys each record to one identifier and links it to the sites and systems around it. This page covers each step, the sources behind each data type, how often they update and what you can see for free.

30 days free. Self-serve. No credit card. [Start free trial](https://app.dmand.ai/signup)

Last updated: 2026-10-25

## Dmand AI combines CMS, Medicare and HRSA records with its own executive, hierarchy and technology data

Public registries are self-reported, and they often disagree. A [2013 review by the HHS Office of Inspector General](https://oig.hhs.gov/oei/reports/oei-07-09-00440.asp) found inaccurate data in 48% of the NPI registry records and 58% of the Medicare enrollment records it sampled, most often addresses. So no Dmand AI record rests on one file: each source is matched against the others before it reaches a profile.

### HCP records start from the NPI registry and Medicare enrollment

Identity, specialty and practice address come from NPPES; enrollment, PAC ID and group billing from PECOS; procedures, prescribing and industry payments from Medicare files and Open Payments.

### Executive records come from career data and organization sources

Names, titles and role dates come from licensed business data and public sources such as leadership pages, conference agendas and press. Dmand AI sorts each title into a department and a level.

### Hospital and site records come from CMS facility files and cost reports

Type, CCN and services come from the Provider of Services file, beds and finances from hospital cost reports, owners from Medicare enrollment, and clinics and other sites from NPI organization records and HRSA health center data.

### Medical group records come from Medicare billing and NPI records

A group’s PAC ID, clinicians and locations come from PECOS and Care Compare, its organization NPI from NPPES, and its owner from Dmand AI’s hierarchy research.

### Technology records come from install data and CMS reporting

The EHR at a practice or hospital comes from the clinician’s own CMS reporting on certified health IT, the site where they work or its parent; installed hospital systems come from Dmand AI’s technology data.

### See every source by data type

| Data type | Public sources | Dmand AI data |
| --- | --- | --- |
| Physicians and HCPs | NPPES NPI Registry (CMS); Medicare enrollment records (PECOS); Care Compare, Doctors and Clinicians (CMS), including facility affiliations; Medicare Part B and Part D files; CMS Open Payments; state license boards; ClinicalTrials.gov; PubMed | Dmand AI Affiliation Data |
| Healthcare executives and practice staff | Owners and officers on file in PECOS; organization websites and leadership pages; conference agendas; press | Dmand AI Executive Data, with titles sorted into 38 departments and 700+ sub-departments |
| Hospitals and health systems | CMS Provider of Services file; CMS hospital cost reports; PECOS ownership; NPPES; Medicare inpatient and outpatient hospital files | Dmand AI Organization Profile Data; Dmand AI Org Hierarchy Data |
| Sites of care | NPPES organization records; CMS Provider of Services files; HRSA health center data; state licensing filings | Dmand AI Organization Profile Data; Dmand AI Org Hierarchy Data |
| Medical groups | PECOS; Care Compare, Doctors and Clinicians; NPPES | Dmand AI Org Hierarchy Data |
| EHR and technology | CMS certified health IT reporting; HRSA health center data | Dmand AI Technology Data |
| Contacts | NPPES practice phone | Waterfall enrichment across 20+ healthcare-specific sources, including licensed business data providers |

## Each record is keyed to a permanent identifier: the NPI for clinicians, the CCN for certified facilities and the PAC ID for Medicare enrollment

Names change and addresses get spelled many ways; identifiers stay the same. Dmand AI joins sources on these numbers first and standardizes names and addresses second, so a clinician who changes jobs or a hospital that rebrands keeps one record.

### NPI for clinicians and organizations

The 10-digit National Provider Identifier CMS issues to individual providers (Type 1) and organizations (Type 2), kept apart in Dmand AI. A deactivated NPI is flagged and hidden from search unless you ask for it.

### CCN for certified facilities

The six-character CMS certification number of each Medicare-certified hospital and facility. One hospital is one CCN, so a merger or a new name does not split a hospital into two records.

### PAC ID for Medicare enrollment

The PECOS Associate Control ID that Medicare enrollment gives clinicians and groups. It shows which group a clinician assigns billing to, which is how group links are made.

### Dmand ID for organizations in the hierarchy

Sites of care, medical groups and parent organizations in the org hierarchy carry a Dmand ID, starting SOC for a site and NG for a group or parent, so a site without a CCN still keeps a stable record.

## Dmand AI places a person at a site only from a record that ties them to it, and each link keeps its type

6.4M+ affiliations tie clinicians to 675K+ sites of care. Practicing at a site, being employed there, billing through a group and running a hospital are different facts, so each link records which one it is and where it came from.

### Ownership links come from CMS filings and hierarchy research

Owners on file with Medicare, cost reports and Dmand AI’s own research place each hospital, group and site under the organization that owns or runs it.

### Practice links come from the locations clinicians file with CMS

Practice addresses in NPPES and Care Compare, and the hospitals CMS lists for each clinician, become links to sites, labeled primary practice, employer or affiliated.

### Group links come from Medicare billing reassignment

When a clinician assigns Medicare billing to a group in PECOS, Dmand AI links them to that group, based on enrollment rather than an email domain.

### Leader and staff links come from roles at the organization

Executives, managers and staff are placed at the organization and site they work for, so the people who run a site sit beside the clinicians who practice there.

## Dmand AI draws each health system from the IDN down to its divisions, hospitals, outpatient departments, medical groups and practice sites

Dmand AI’s data team builds the tree and checks it against CMS ownership and enrollment filings, CMS facility files, the NPI registry and state licensing filings. Outpatient departments sit under the hospital whose CCN they bill under, and 99% of hospitals are placed under their parent organization.

## Dmand AI finds work emails, personal emails and mobile numbers through waterfall enrichment across 20+ healthcare-specific sources

The NPI registry lists a practice phone and no email. When you ask for a contact, Dmand AI asks its sources one after another, including licensed business data providers, matches each result to the person and their site, and checks it before it reaches your list.

### One waterfall for each contact type

Email and mobile searches run separately, so a source with no mobile number never holds up the email search.

### Matched to the person, not the domain

Results are matched on a clinician’s NPI and site, or a leader’s role and organization, which keeps common names at large systems apart.

### Labeled by type and status

Each email is marked professional or personal and deliverable or catch-all, and each mobile number valid, with the date of the last check. Many hospital mail servers accept any address, so a catch-all email is never shown as deliverable.

### Charged only when found

A lookup that finds nothing uses no credits, and you can ask for a fresh check on any contact.

## Dmand AI refreshes its records every month, and any user can flag a wrong field from the profile it is on

Each public source is picked up as it publishes. A change then takes the same path as the first build: it is matched to one record, the links around it are rechecked, people review what software cannot settle, and the update reaches the profile and your saved lists.

### New source files arrive on their own schedules

Registry and Medicare files publish weekly, monthly, quarterly or yearly, and each is loaded as it lands.

### Each change is matched to one record

Names and places are standardized, so “Saint Louis” and “St. Louis”, or two spellings of one suite, count as one place.

### Links to sites and systems are rechecked

A new practice address or billing group moves a clinician’s links, and a new owner moves a site in the tree.

### The data team reviews what software cannot settle

Possible duplicates, mergers and changes of ownership go to people before they change a record.

### Updates reach the profile and your saved lists

The profile shows the new value with its source, and job changes are flagged in the lists that hold the person.

### See how often each source publishes

| Source | Publisher | Publishes | What it updates in Dmand AI |
| --- | --- | --- | --- |
| NPI registry (NPPES) | CMS | Weekly changes, a full file each month | Names, credentials, taxonomy, practice addresses, deactivations |
| Medicare enrollment (PECOS) | CMS | Quarterly | Enrollment, PAC IDs, group billing, owners on file |
| Care Compare, Doctors and Clinicians | CMS | Monthly | Practice locations, group and hospital affiliations |
| Provider of Services files | CMS | Quarterly | Facility type, CCN, services, certification |
| Hospital cost reports | CMS | As new filings post | Beds, discharges, revenue and margin |
| Medicare Part B, Part D and hospital service files | CMS | Yearly | Procedures, prescribing, hospital services |
| Open Payments | CMS | Yearly | Industry payments to clinicians |
| HRSA health center data | HRSA | As HRSA publishes | Health center sites and their EHR |
| ClinicalTrials.gov and PubMed | National Library of Medicine | As records are added | Trials and publications |
| Dmand AI Executive, Org Hierarchy, Affiliation and Technology Data | Dmand AI | Every month | Leaders, hierarchy, links, installed systems |
| Contacts | 20+ healthcare-specific sources | When you ask | Emails and mobile numbers, with the date checked |

To fix a record, use Suggest update on its profile in the app, or request data that is not there yet; both go to the Dmand AI data team. To remove your own profile, use the [removal request page](https://dmand.ai/legal/remove-my-information).

## Dmand AI publishes answers on free profiles and keeps affiliations, org trees and contacts in the app

Free pages answer what a search asks: who a hospital is, how big it is and who owns it. The links between people and places, the part the data team builds, open in the app.

### Free on dmand.ai

Profiles and directories for US hospitals, medical groups and HCPs: identifiers, type, size and address, each hospital’s parent organization, a clinician’s primary practice, leaders’ names and titles, technology categories, and ranked lists of the 10 largest.

### In the Dmand AI app

All of a record’s affiliations and colleagues, full org trees, EHR vendors and products, 100+ filters and contacts, which use credits only when found. Exports, CRM sync, signals and the Claude connector depend on the plan, and every plan starts with 30 days free.

## Healthcare data sources and methodology FAQs

### Where does Dmand AI’s healthcare data come from?

From public records and Dmand AI’s own data. Public sources include the CMS NPI registry (NPPES), Medicare enrollment records (PECOS), Care Compare, the CMS Provider of Services files, hospital cost reports, Medicare claims files, Open Payments, HRSA health center data, state license boards, ClinicalTrials.gov and PubMed. Dmand AI adds its executive, org hierarchy, affiliation and technology data, and finds contacts through waterfall enrichment across 20+ healthcare-specific sources.

### Is Dmand AI’s data built on NPPES and PECOS?

Yes, for clinicians and medical groups. NPPES supplies each NPI, name, taxonomy and practice address; PECOS supplies Medicare enrollment, PAC IDs, the groups clinicians bill through and the owners on file. Dmand AI matches the two against each other and against Care Compare, because the two registries often disagree.

### How accurate is the NPI registry?

The NPI registry is self-reported, and CMS does not verify most of what providers enter. A 2013 HHS Office of Inspector General review found inaccurate data in 48% of the NPPES records and 58% of the PECOS records it sampled, mostly in addresses, and the two systems disagreed for 97% of the records. So Dmand AI never relies on one file: it matches sources against each other and sends unclear cases to its data team.

### What is a CCN, and why does Dmand AI key hospitals to it?

A CCN, or CMS certification number, is the six-character number CMS gives each Medicare-certified hospital and facility. Dmand AI counts one hospital per CCN, so a rebrand or a merger does not create a second record, and it places outpatient departments under the hospital whose CCN they bill under.

### What is a Medicare PAC ID?

The PAC ID, or PECOS Associate Control ID, is the number Medicare enrollment assigns to a clinician or group. When a clinician assigns billing to a group, the two PAC IDs connect, and Dmand AI uses that connection to link clinicians to their medical groups.

### How does Dmand AI know which hospital or group a clinician works at?

From records that tie the two together: the practice locations a clinician files with CMS, the hospitals CMS lists them with and the group they bill Medicare through. Each link is labeled primary practice, employer, affiliated or billing group, and 6.4M+ such links tie clinicians to 675K+ sites of care.

### How does Dmand AI build a health system’s org hierarchy?

Its data team places each hospital, medical group and site under the organization that owns or runs it, checked against CMS ownership and enrollment filings, CMS facility files, the NPI registry and state licensing filings. Outpatient departments sit under their hospital, and 99% of hospitals are placed under their parent organization.

### How does waterfall enrichment find emails and mobile numbers?

When you ask for a contact, Dmand AI checks 20+ healthcare-specific sources one after another until one finds it, runs a separate waterfall for each contact type, matches the result to the person and site, and labels it with its status and the date checked. A lookup that finds nothing uses no credits.

### How often is Dmand AI’s data updated?

Records for HCPs, executives and sites of care refresh every month. Each public source is loaded as it publishes: the NPI registry weekly with a full file each month, Care Compare monthly, Medicare enrollment and the Provider of Services files quarterly, and Medicare utilization and Open Payments yearly. Contacts are checked when you ask.

### How do I report a wrong record in Dmand AI?

Use Suggest update on the record’s profile in the app. The Dmand AI data team checks it against the sources and fixes the record. To remove your own profile instead, use the removal request page on dmand.ai.

### Does Dmand AI use Medicare claims data?

Yes. Procedure volumes and prescribing come from Medicare Part B and Part D files, and hospital services from Medicare inpatient and outpatient files, each with its year. They cover Medicare claims only, not commercial or all-payer claims.

### What can I see for free, and what needs an account?

Free profiles and directories show identifiers, type, size, address, each hospital’s parent organization, a clinician’s primary practice, leaders’ names and titles and technology categories. Full affiliations, org trees, EHR vendors and contacts are in the app, and every plan starts with 30 days free, self-serve, with no credit card.

---

**Check the method on the records you know best.** 30 days free, self-serve, no credit card. [Start free trial](https://app.dmand.ai/signup) or [book a demo](https://dmand.ai/book-a-demo).

Canonical: https://dmand.ai/how-our-data-is-built