ambulatory contact directory data quality research
An ambulatory contact directory is useful only when a staff member can reach the intended organization, confirm the required service, and record what changed. A row can look complete while its telephone number reaches a closed location, its fax belongs to another department, or its payer status has expired. This review asks how a small practice can measure those failures without presenting external directory studies as a local prediction.
The recommended unit is an attempted use of one directory record for one defined purpose. Record creation, attempted contact, successful connection, service confirmation, discrepancy discovery, correction submission, verified correction, and scheduled recheck should remain separate events. That sequence shows where work stops and which records remain unresolved.
This is an operational evidence review, not legal, privacy, contracting, billing, or clinical advice. The published percentages below describe particular insurer directories, specialties, methods, and years. They help define audit questions but do not set a target for an independent office.
Define the measure
Start by identifying the population of records that the practice actually relies on: referral destinations, laboratories, pharmacies, payers, imaging centers, vendors, and community resources. Do not mix archived contacts with active destinations. Give each active record a stable identifier so that a correction does not erase its history.
Define successful use before sampling. A connected telephone call may prove reachability but not specialty, network status, secure-fax ownership, or willingness to accept a referral. Use separate fields for each assertion. If the receiving office gives an ambiguous answer, record unknown rather than forcing a pass.
A useful discrepancy taxonomy includes disconnected line, reassigned line, wrong department, closed location, wrong specialty, outdated clinician affiliation, rejected fax, inactive portal link, and unverified network participation. The taxonomy should also distinguish a suspected error from one confirmed against an authoritative second source.
Keep four clocks: age since last verification, time from failed use to correction request, time from request to updated record, and time from update to independent confirmation. These clocks answer different questions. A fast edit is not equivalent to a correct destination.
Evidence from provider-directory studies
The selected literature measures directory accuracy through secret-shopper calls, cross-directory comparison, stakeholder interviews, and record-matching methods. A California mammography study reported directory inaccuracies ranging from 23% to 38% across more than 33,000 observations in 2018 and 2019 (PubMed). That range belongs to the plans, specialties, years, and verification method studied; it is not a benchmark for a solo practice contact list.
A Maryland Medicaid analysis found that insurance participation could be verified for 46% of listed providers in 2018 and 56% in 2019 (PubMed). A separate cross-insurer study found much lower consistency for addresses and telephone numbers than for specialty labels (PubMed). Together, these studies support separating reachable contact, correct network status, specialty, location, and appointment availability rather than collapsing them into one pass/fail field.
Persistence studies add a time dimension. Researchers who recontacted inaccurate listings found that many remained inaccurate months later (PubMed; PubMed). This supports recording both the last verification date and the outcome of an attempted use. It does not establish a universal reverification interval.
| Evidence field | Local numerator | Local denominator | Important exclusion |
|---|---|---|---|
| Reachable destination | successful connections | attempted contacts | calls abandoned before the destination could be tested |
| Correct service | destinations confirming the intended service | reached destinations | unresolved or ambiguous answers |
| Verified correction | corrected entries confirmed through a second source | entries submitted for correction | edits not independently rechecked |
| Known staleness | entries beyond the practice review rule | active directory entries | inactive records retained only for audit history |
Machine-readable directories and natural-language processing may help identify candidate discrepancies (PubMed; PubMed). Automation does not prove that a telephone number reaches the intended office or that a payer relationship is current. A local workflow should preserve the source used, the verifier, the date, the result, and unresolved uncertainty.
Build a local protocol
Select records by a reproducible rule. A small office might audit every destination used during a defined month plus a random sample of other active records. Preserve the complete denominator, attempted records, exclusions, nonresponses, and unresolved cases. Replacing unreachable records during the audit must not remove them from the original sample.
For each test, document the intended service, channel, date, staff role, verification source, outcome, and next action. Avoid including patient information in the directory audit. If a call or portal test would disclose unnecessary patient details, use an approved alternative verification method.
Require a second check for material edits such as telephone, fax, location, specialty, or network status. The confirmer should see the evidence and the prior value. A second person is preferable where staffing permits; otherwise separate the initial change and later confirmation in time and retain both timestamps.
Track correction outcomes. Submitted, accepted, reflected in source, and verified through use are not interchangeable. If an insurer or outside organization controls the published listing, keep the practice's request receipt and the latest observed public value. The local record can warn staff while an external correction is pending.
Review the specialist office contact directory maintenance workflow for operational ownership and the referral appointment outcome capture workflow for closure evidence after a referral leaves the practice.
Interpret results carefully
Report counts beside rates and show the observation window. A 10% error rate means something different when it represents one of ten records versus one hundred of one thousand. Break results out by field and destination type only when the subgroup denominators remain visible.
Do not claim that an edit caused better access unless the study design supports that conclusion. Seasonal closures, payer updates, mergers, office moves, and changes in referral mix can alter the observed rate. Repeat the same protocol and explain definition changes before comparing periods.
The most useful output is a worklist with verified corrections, unresolved records, responsible owners, and next review dates. A single accuracy score cannot show which destination may misroute sensitive information or delay a referral.
Frequently Asked Questions
Does a connected call prove the record is accurate?
No. It proves only that the tested number connected. Specialty, location, secure-fax ownership, payer participation, and appointment availability require their own verification steps.
How should nonresponses be counted?
Keep them visible as unresolved attempts. Excluding them from the denominator can make the directory appear more accurate than the tested workflow.
Is there one correct reverification interval?
The selected studies do not establish one interval for every practice. Choose a written risk-based rule, record the last verified date, and also recheck after a failed use or credible change notice.
Can software replace confirmation calls?
Software can identify conflicts and stale fields, but it does not prove that a destination currently serves the intended purpose. Use automation to prioritize review, then preserve human verification evidence.