Healthcare language interpreter services need an operational review after procurement. A signed contract shows that a service has been arranged; it does not show whether staff can access it or whether patients' communication needs were met. A useful audit combines encounter records, service logs and staff feedback, while keeping the limits of each data source visible.
What a useful audit does
- Separates survey findings from proof of cause.
- Counts comparable encounters within a defined period.
- Records language needs and how they were addressed, not just a preferred-language label.
- Connects access problems to an owner, a tested response and follow-up measurement.
What the research actually shows
A 2026 JAMA Network Open study analysed 230 primary care physicians who reported patients with limited English proficiency. Among them, 50.4% reported regularly using at least one professional interpreting modality and 53.0% reported regularly using at least one ad hoc modality. These are physician self-reports, not percentages of patient visits, and the categories are not mutually exclusive.
The survey was conducted from October 2024 to April 2025. Of 753 eligible physicians, 267 responded, with 230 included in the analysis. The authors identify a modest response rate, self-reporting, limited generalisability beyond primary care and a cross-sectional design that cannot establish causality. The study supports investigating access barriers; it does not prove that every institution already has enough interpreting capacity or that a different contract cannot help.
A 2024 study by Slade and colleagues distributed a questionnaire to 750 clinicians at one safety-net hospital and analysed 221 responses, a 29.5% response rate. Efficiency pressures and cumbersome access led the reported barriers; cost ranked lowest. Lowest-ranked is not absent, and findings from that hospital should not be presented as the explanation for every practice.
An earlier qualitative study of 20 residents at two hospitals found underuse despite readily available interpreters. Its interviews describe decisions shaped by time, convenience and perceived communication needs. They do not test whether changing suppliers or contract terms would improve use. Treat the studies as prompts for local investigation, not as a universal procurement verdict.
Map the written process to the real access path
For covered entities, 45 CFR 92.8(d) specifies written language-access procedures. These include how staff recognise limited English proficiency, obtain qualified interpreters and translators, identify qualified bilingual staff and locate translated materials. The inventory includes languages, issuance dates and electronic access instructions, plus coordinator contact details where applicable.
Rehearse the procedure with the people who must use it, including overnight and temporary staff. Can they find access details, identify the right language, connect the equipment and escalate a failed request? A written process is only one part of implementation. A one-minute target may be a local operational goal, but it is not a universal compliance threshold.
Section 92.7 sets the coordinator requirement for covered entities with at least 15 employees. Section 92.9 addresses training for relevant employees, including new employees and people affected by material policy changes. It requires completion records to be made at the time and kept for at least three calendar years. An audit should examine that minimum together with other applicable retention obligations, not flag every deletion of an older record as a violation.
Section 92.11 addresses notices of available assistance, including annual provision, requests, websites, physical locations and specified communications. It also contains individual-level compliance options, including documented opt-outs with conditions. This summary is not a complete legal checklist: have the responsible compliance team confirm coverage, applicable exceptions and current requirements before changing policy.
Define the denominator before calculating a rate
A patient's preferred language is useful information, but does not by itself establish limited English proficiency or the need for an interpreter at every visit. Similarly, a blank interpreter field may reflect missing documentation rather than proof that no assistance occurred. Joining a count of patients to a count of visits creates a misleading rate.
For a local reporting measure, define a population of encounters during a stated period and record assessed language-assistance needs. Then classify how those needs were addressed: professional interpreting, assessed language-concordant communication, refusal of an offered service, other documented arrangements, or unknown outcome. Keep unknown need separate too. Review classifications with the language-access and clinical teams; do not silently remove refusals or unknowns to improve the headline number.
One proposed descriptive metric is: encounters with documented professional interpreting divided by encounters with documented language-assistance need. Both counts must use the same population, period and encounter identifier. Report the other outcome categories alongside it. This is a service-use metric, not a standalone legal-compliance score or a target that must always reach 100%.
Six measures to review together
| Measure | Definition to agree | How to interpret it |
|---|---|---|
| Language-need assessment completeness | Encounters with an assessment divided by all in-scope encounters. | Unknown need signals a data-quality gap, not a confirmed lack of need. |
| Documented response to need | Counts for each response category among encounters with documented need. | Investigate unknown outcomes and access failures; distinguish appropriate direct communication from missing interpretation. |
| Professional interpreting use | Documented professional interpreting encounters divided by encounters with documented need. | Read alongside refusals and other arrangements, not as a compliance score. |
| Time to connect | Elapsed time between a defined request timestamp and successful connection. | Show median and tail values by language, department and shift; state how retries are handled. |
| Abandoned requests | Abandoned requests divided by all in-scope requests, using an agreed retry rule. | Investigate causes; a duplicate or resolved request is not automatically an unmet need. |
| Required training evidence | Relevant employees due for training, completion dates and required retained records. | Review new roles and material procedure changes as well as initial completion. |
These are proposed operational measures, not ratios mandated by the cited studies or regulation. Vendor call logs may not capture in-person interpreting or direct language-concordant care. Reconcile sources before comparing departments, protect patient information and limit report access to authorised people.
Use the findings to choose a specific improvement
As an illustrative example, repeated overnight connection failures might lead to testing devices, access instructions, staff training and vendor availability for that shift. The appropriate response could be an internal process change, additional capacity or a revised supplier commitment. Do not assume the cause from the aggregate rate.
The Slade study describes initial improvement efforts involving video equipment, a language-access committee and a leadership role. Those are options to evaluate, not proof that any one intervention will work in a different setting. Assign an owner, define a follow-up period and check whether the identified failure becomes less frequent.
For procurement and payment questions, read medical interpretation services: costs and patient access. See also remote interpreting services and video interpreting services. Keep this audit focused on what happened in the encounter, while using contract evidence where it helps explain the result.
Keep automated tools outside the professional-interpreting count
MirrorCaption can transcribe and translate supported speech from a microphone or explicitly shared meeting-tab audio in a supported browser. It does not supply human interpreters, signed-language interpretation or an assurance of Section 1557 or HIPAA compliance. Do not count automated captions as delivery of a required professional interpreter.
An approved non-sensitive administrative or training meeting may be a suitable use to test. Internal calls and research discussions can still contain protected patient data or consequential clinical decisions; the meeting's label is not a privacy exemption. Check content, consent, organisational approval and data handling before sharing audio, and do not use the tool as a substitute for required patient-facing language assistance.
Try a non-sensitive internal conversation
Evaluate automated captions separately from your patient language-access service.
Try MirrorCaptionFrequently asked questions
Why might clinicians underuse professional interpreters?
Studies identify time pressure, access difficulties and other factors, but the findings depend on the setting and study design. A 2024 single-hospital survey ranked cost lowest among the listed barriers; it did not establish that cost never matters.
How should we calculate interpreter utilisation?
Use a defined reporting period and a consistent encounter-based denominator. Document language-assistance need and the response to it, including professional interpreting, assessed language-concordant care, refusal and missing information. Preferred language alone does not establish interpreter need.
How long must section 92.9 training records be retained?
Section 92.9 requires contemporaneous documentation of the specified training and retention for at least three calendar years. That is a minimum, not proof that deleting any record older than three years violates the section; other retention obligations may apply.
What belongs in written language-access procedures?
Section 92.8(d) calls for the applicable coordinator's current contact details, how staff identify limited English proficiency and obtain qualified interpreters and translators, qualified bilingual staff names, and an inventory of translated materials with languages, issuance dates and electronic access instructions.
Can MirrorCaption count as the required interpreter service?
No. MirrorCaption provides automated transcription and translation, not a qualified human interpreter or signed-language interpretation. Do not count its use as delivery of required professional interpreting. Even internal meetings need an appropriate privacy and content assessment.