Blog / Healthcare Lead Generation / Healthcare data

Healthcare Contact Data Accuracy in 2026: What the Evidence Shows

What published research actually says about provider data accuracy, what levels are realistic, and the benchmarks worth holding a vendor to.

HK

Harshith Kumar

Healthcare data writer
Updated 6 Nov 2026
10 min read
Why Healthcare Contact Data Is So Hard to Keep Up to Date

TABLE OF CONTENTS

Reading progress

QUICK ANSWER

Published evidence suggests healthcare provider data is far less accurate than vendor claims imply. A CMS audit of Medicare Advantage directories found 48.74% of 10,504 reviewed locations carried at least one deficiency and 50.14% of providers had at least one error, in files regulated health plans must legally maintain. Realistic benchmarks are per-field and per-segment rather than a single percentage.

Healthcare data vendors routinely advertise accuracy in the mid to high nineties. The best available independent evidence points somewhere very different, and the gap is worth understanding before you set expectations for a campaign.

The most rigorous public measurement comes from the Centers for Medicare & Medicaid Services, which audited Medicare Advantage provider directories directly rather than relying on self-report. Those directories are maintained by regulated health plans under a legal obligation to keep them current.

Even so, CMS found that 48.74% of the 10,504 locations reviewed had at least one deficiency, and 50.14% of the 5,602 providers reviewed had at least one error. This guide sets out what that implies for commercial data and what benchmarks are realistic.

What the CMS Audit Actually Found

CMS reviewed provider directories from 52 Medicare Advantage Organizations between November 2017 and July 2018, covering 5,602 providers at 10,504 locations. It remains the most substantial independent measurement of provider data accuracy available publicly.
Error type Share of deficiencies What it means
Provider not at that location
39.61%
The single largest error category
Provider not at any listed location
26.43%
Record wholly disconnected from reality
Phone number needs updating
13.09%
Contact point wrong
Address needs updating
6.91%
Location wrong
Suite number needs updating
4.53%
Partial address error
Wrongly shown as accepting patients
4.19%
Status wrong
Wrongly shown as not accepting
4.46%
Status wrong, other direction
Specialty or name errors
Under 1%
Identity rarely wrong
Two things stand out. First, 66% of all deficiencies were location errors: the provider was not where the directory said. Second, identity errors were rare. Who someone is tends to be right; where they are and what they are doing tends to be wrong.

Key Takeaways

  • Roughly half of audited directory locations carried an error, in legally mandated files.
  • Location is the weak point, not identity. 66% of deficiencies were location errors.
  • Accuracy varied from 4.63% to 93.02% between organisations, so averages conceal enormous variation.
  • Vendor claims measure deliverability, which is a narrower question than accuracy.
  • Benchmarks should be per-field and per-segment, never a single number.
  • Measure it yourself — a 20-record manual check settles more than any claim.

Why Vendor Claims and Evidence Diverge

A vendor claiming 95% accuracy and an audit finding 50% error are usually not contradicting each other. They are measuring different things.
Vendor figures almost always describe email deliverability: the proportion of addresses that will not hard bounce. That is mechanically testable at scale and genuinely achievable at high rates. The CMS audit measured whether the provider was actually at the listed location, practising the listed specialty, accepting patients as stated. Those are harder questions and they fail far more often.
So both numbers can be true simultaneously. A file can be 95% deliverable and still have half its location or affiliation data wrong, because an email address survives a move that invalidates everything around it.

The Forces Driving Decay

Three documented forces explain why location and affiliation data fails at these rates.
Workforce movement. Research in Annals of Internal Medicine from a Weill Cornell Medicine team found annual physician turnover rose from 5.3% to 7.6% between 2010 and 2018, a 43% relative increase. Among registered nurses, BLS projects about 180,800 openings per year against a base of roughly 3,465,400.
Structural consolidation. AHA data shows 3,567 of 5,121 US community hospitals now belong to a health system. Each acquisition changes affiliations, domains and purchasing authority across many provider records at once.
Source latency. HHS has stated in the Federal Register that provider practice addresses change frequently and are therefore unlikely to be maintained accurately, since updates occur only when providers choose to submit them. Commercial files built on that foundation inherit the lag.

24 TO 48 HOUR DELIVERY

Want Accuracy Reported for Your Segment?

Send your target criteria and we will confirm counts and share sample records with verification dates attached.

Realistic Accuracy Benchmarks by Field

Because fields decay at different rates, benchmarks only make sense per field. Use these as the basis for what to ask a provider.
Field Expectation Why
Name and credential
Very high
Identity rarely changes; CMS found under 1% error
Specialty
High if NPI-derived
Structured source; weak if title-inferred
Email deliverability
High on recent verification
Mechanically testable at scale
Practice location
Moderate
The largest error category in the CMS audit
Employer or affiliation
Moderate
Consolidation changes it in bulk
Role or seniority
Lower
Often inferred once and never revisited
Phone number
Moderate
13.09% of CMS deficiencies
Practice setting
Lower
Frequently inferred rather than verified
The pattern is consistent: fields describing who someone is hold up well, and fields describing where they are and what authority they hold decay quickly. Any single accuracy percentage averages across both and therefore describes neither.

Why Segment Matters More Than Average

The most under-appreciated finding in the CMS data is not the average but the spread. Location inaccuracy rates by organisation ranged from 4.63% to 93.02%, against an average of 44.97%.
That range tells you something important about averages in this field: they are nearly uninformative about any particular slice. A vendor quoting a database-wide figure is giving you the midpoint of a distribution that may be extremely wide, and your segment could sit anywhere in it.
Commercial files vary the same way, for the same reasons. Providers refresh high-demand specialties frequently and leave smaller segments untouched for much longer. This is why every accuracy question should be framed around your specific segment, and why a database average should be treated as a non-answer.

How to Measure Accuracy Yourself

Given how little vendor percentages tell you, measuring directly is the only reliable route. A sample of 200 records supports all of this.
  • Validate all emails to establish deliverability, then set that question aside.
  • Manually verify 20 records against practice websites, hospital directories and board certification tools.
  • Check location specifically, since that is where the CMS evidence says errors concentrate.
  • Count fill rates on the two or three fields you filter by.
  • Deduplicate on NPI to test whether the count was inflated.
  • Calculate the median verification date rather than reading the best few.
Scored against the CMS baseline, two or three location errors in twenty is roughly consistent with professionally maintained data. Eight or more suggests a file that has not been independently verified at all.

What Good Accuracy Reporting Looks Like

A provider reporting honestly on accuracy looks different from one marketing it, and the difference is visible in four specific behaviours.
  • Per-field figures rather than one headline number, because fields decay at different rates.
  • Segment-level figures for your specific filters rather than the database.
  • A non-zero unresolved category. Every genuine process produces records it cannot confirm.
  • Per-record verification dates, so freshness is auditable rather than asserted.
A provider volunteering that location data is less reliable than identity data is telling you something true and verifiable against the published evidence. One claiming uniform high accuracy across all fields is making a claim the independent research does not support for anyone.

Want Per-Field Accuracy for Your Filters?

Share your targeting criteria and we will confirm what is available and how recently it was verified.

Planning Campaigns Around Real Accuracy

Accepting realistic accuracy changes how you plan rather than whether you proceed.
  • Budget for a usable subset, not the full record count, and confirm counts before purchase.
  • Weight targeting toward stable fields. Specialty and credential hold up better than role and setting.
  • Validate immediately before sending, not at purchase.
  • Report engagement on delivered messages so deliverability gaps do not masquerade as audience preference.
  • Treat location-dependent tactics cautiously, since that is the weakest field in the evidence.
The last point has practical consequences for territory assignment and direct mail, both of which depend on exactly the field the CMS audit found least reliable. Plan for verification there rather than assuming it.

How MedicoLeads Reports Accuracy

MedicoLeads reports accuracy per segment rather than as a single database-wide figure, because the published evidence shows database averages conceal very wide variation between segments.
Counts are confirmed against your exact filter combination before purchase, each record carries its own verification date and method, and records that could not be resolved are removed rather than absorbed into the delivered count.
Verification runs in layers because deliverability and record accuracy require different evidence, which is precisely the distinction the CMS findings illustrate. The method is described in our explanation of how layered verification works.

Conclusion

The best independent evidence on healthcare provider data accuracy is sobering: roughly half of audited directory locations carried at least one error, in files that regulated organisations were legally required to maintain. Identity data held up well; location and affiliation data did not.
Treat single accuracy percentages as describing deliverability only, ask for per-field and per-segment figures, and measure a sample yourself. A provider who reports their weak fields honestly is more useful than one claiming uniform accuracy that no published research supports.

Frequently Asked Questions About Healthcare Data Accuracy

Less accurate than vendor claims suggest. A CMS audit of Medicare Advantage provider directories found 48.74% of 10,504 reviewed locations carried at least one deficiency and 50.14% of 5,602 providers had at least one error, in directories regulated health plans are legally obliged to maintain.

They are measuring different things. Vendor figures almost always describe email deliverability, which is mechanically testable and genuinely achievable at high rates. The CMS audit measured whether providers were at the listed location practising as stated. A file can be highly deliverable and still have much of its location data wrong.

Identity fields. CMS found specialty and name errors accounted for under 1% of deficiencies, while 66% were location errors. Name, credential and NPI-derived specialty hold up well. Practice location, affiliation, role and practice setting decay considerably faster and should be verified rather than assumed.

Because the variation is enormous. In the CMS data, location inaccuracy by organisation ranged from 4.63% to 93.02% against an average of 44.97%. Commercial files vary similarly, since providers refresh high-demand specialties often and leave smaller segments untouched. Always ask about your specific segment.

Take a 200-record sample: validate all emails, manually verify 20 records against practice websites and hospital directories with particular attention to location, count fill rates on your filter fields, deduplicate on NPI, and calculate the median verification date. Two or three location errors in twenty is roughly normal.

Per-field rather than headline figures, segment-level rather than database-wide, a non-zero unresolved category, and per-record verification dates. A provider who volunteers that location data is weaker than identity data is describing something the published evidence supports.

Yes. Budget for a usable subset rather than the headline count, weight targeting toward stable fields such as specialty and credential, validate immediately before sending rather than at purchase, and treat location-dependent tactics like territory assignment and direct mail with extra verification.

The pressures are increasing. Physician turnover rose 43% between 2010 and 2018, and 3,567 of 5,121 US community hospitals now sit inside health systems, with each acquisition changing affiliations in bulk. Verification methods have improved, but the underlying rate of change has risen alongside them.

Ready to see accuracy reported honestly?

Tell us the specialties, roles and regions you need. We will confirm counts and share sample records with verification dates.

ABOUT THE AUTHOR

Picture of William Coogan

William Coogan

William Coogan is a healthcare marketing professional with 15 years of experience in communication strategy. At MedicoLeads, he writes about data-driven outreach that helps healthcare providers reach and engage their audiences.

FREE SAMPLE

See records with verification dates

Get sample healthcare records showing when each was last confirmed and by what method.
No commitment. Reply within one business day.

TALK TO US

Ready to Work From Verified Healthcare Data?

Tell us who you want to reach. Our data experts will confirm available counts and share a sample within one business day.
Scroll to Top