The MRI Safety Screening Call in Haitian Creole: Live Translation Ends the 11-Minute Interpreter Hold

How Many of Last Month's No-Shows Spoke Spanish at Home?

Most imaging center owners cannot answer that question, because the RIS field for preferred language is either blank or holds whatever the front desk guessed at registration. But every scheduler in the building can tell you the pattern from memory: the patients who need an interpreter are the ones who arrive without having fasted, arrive wearing deodorant for a screening mammogram, arrive on the wrong day, or do not arrive at all.

That is not a patient problem. It is a communication problem with a dollar figure on it, and until this year the only fix was a per-minute interpreter line and a lot of hold music.

Here is what a bad version looks like. A Haitian Creole-speaking patient is scheduled for a lumbar MRI at 3:20pm. The pre-screen call the day before never connects properly, so the safety questionnaire is done in the hallway with the patient already gowned, the technologist holding a phone on speaker, waiting eleven minutes for a Creole interpreter to come on the line. Meanwhile the 3:40 patient is in the waiting room and the magnet is idle.

The Form That Cannot Be Guessed At: MRI Safety Screening

Of all the paperwork in a diagnostic building, the magnetic resonance safety questionnaire is the one where a language gap is genuinely dangerous rather than merely expensive. It asks about pacemakers and implanted defibrillators, aneurysm clips, cochlear implants, neurostimulators, insulin pumps, surgical hardware, metal fragments in the eye from grinding or welding, tattoos, pregnancy, and whether the patient can lie flat for forty minutes. A patient who nods along without understanding is a safety event waiting to happen.

The rest of the pre-visit conversation is less dangerous but more costly. Nothing to eat or drink after midnight for a fasting glucose or lipid panel. Hold biotin supplements before thyroid testing. Drink thirty-two ounces of water and do not empty your bladder before a pelvic ultrasound. No deodorant, powder or lotion before a screening mammogram. Bring your implant card. Bring your prior films or tell us where they were done. Hold metformin after iodinated contrast if instructed. Every one of those, delivered badly, produces either a wasted slot or a repeat visit.

And the obligation is not optional. Section 1557 of the Affordable Care Act requires meaningful language access at facilities receiving federal funds, including qualified interpreters and a notice that language assistance is available. Most centers meet it with a contracted phone line, a laminated language identification card at the front desk, and hope.

What Gemini Changed About the Middle of a Phone Call

Live translation used to be a scheduled service. You booked an interpreter, or you dialed a line and waited in a queue whose length depended on how common the language was — short for Spanish, long for Haitian Creole, Karen, Dari or Amharic.

In 2026 it became something that happens inside a conversation already in progress. Gemini 3.5 Live Translate detects more than seventy languages and translates speech to speech close to real time. Gemini 3.1 Flash Live, released in March, handles audio to audio across more than ninety languages and is good at doing something else mid-conversation — looking a record up, checking a slot, writing an answer down. The practical change for an imaging center is this: the interpreter is no longer a person you wait for, it is a capability the call already has from the first ring, in whatever language the patient opens with.

flowchart TD
  A["Pre-visit call to patient, day before scan"] --> B["Spoken language detected on first reply"]
  B --> C["Safety and prep questions asked in that language"]
  C --> D{"Any implant, metal or pregnancy answer flagged?"}
  D -->|No| E["Prep confirmed, arrival time repeated back"]
  D -->|Yes| F["Warm transfer to qualified interpreter and MR safety officer"]
  F --> C
  E --> G["Preferred language written to RIS for every future visit"]
  G --> H["Reminder texts sent in the same language"]

Tuesday at 5:10pm, Running the Next Day's List

The scheduler starts the pre-visit calls for Wednesday's MRI and CT list at ten past five, because that is when working patients answer. Twenty-two calls. Four of the patients speak Spanish, one speaks Vietnamese, one speaks Haitian Creole.

The Creole call now goes like this. The patient answers, says allo, and the conversation continues in Creole from the second sentence. The safety questions get asked in full — all of them, not the three the technologist can manage in broken Spanish. The patient mentions a plate in his wrist from a 2019 fracture. That answer is flagged, and the call is handed to a qualified interpreter and the technologist so the implant can be identified properly against the manufacturer's card before Wednesday, rather than at 3:20pm with the patient on the table.

The Vietnamese-speaking patient confirms she understood the fasting instruction, and repeats her arrival time back. The preferred language gets written into the RIS, which means her reminder text tomorrow morning goes out in Vietnamese and every appointment she books for the next five years starts in the right language. That last part is the compounding bit and it is the part centers forget to build.

What Three Recovered Slots a Week Are Worth

Assume a center where about nine percent of the patient panel is more comfortable in a language other than English, that the panel's no-show and improper-prep rate runs around twenty-seven percent against fourteen percent for everyone else, and that the center runs 340 appointments a week. All illustrative — put your own numbers in.

That is roughly $44,000 a year on recovered capacity alone, before you count the interpreter minutes. If your center currently buys phone interpretation at, say, $1.85 a minute and burns 240 minutes a month on routine scheduling and prep calls, that is another $5,300 a year of spend that mostly moves to the routine, low-risk conversations that no longer need a live human on the line. Note what that sentence does not say: it does not say you cancel the interpreter contract.

The Conversations That Must Stay With a Qualified Interpreter

Do not use this for anything clinically significant, and be strict about where that line sits. Consent for iodinated or gadolinium contrast is a consent conversation. Delivering a result, especially an abnormal one, is a clinical conversation. Any positive answer on the safety questionnaire — an implanted device, a possible metal fragment in the eye, a possible pregnancy — stops the automated path and goes to a qualified human interpreter and your magnetic resonance safety officer, every time, no exceptions, and that rule belongs in your written procedure rather than in someone's head.

There are practical limits too. Regional dialects and low health literacy defeat a straight translation — a patient may hear every word correctly and still not know what "nothing by mouth after midnight" means for the coffee she drinks at 6am. Elderly patients on a poor cell connection are a hard case in any language. And a deaf or hard-of-hearing patient needs sign language interpretation, which is a different service entirely and is not solved by any of this.

One more, aimed squarely at the owner: document what you did. If you change how language assistance is delivered, write the new procedure down, record which conversations are handled which way, and keep the record. If a complaint ever lands, "we used a translation tool" is a weak answer and "here is our written language access procedure, here is the escalation rule, and here is the log" is a strong one.

Frequently asked questions

Can we use this for the safety screening itself, or only for scheduling?

Use it to ask the questions and to capture the answers, and treat any flagged answer as an automatic stop. The screening is a safety document, so the technologist still reviews and signs it, and anything positive gets re-verified with a qualified interpreter before the patient goes near Zone IV. Used that way it makes the screening better, because the full questionnaire actually gets asked instead of an abbreviated version.

Which languages should we set up first?

Pull your last twelve months of registrations and count. Most centers find that three languages cover the large majority of their non-English panel, and it is usually not the three they expected — one metropolitan center's third-largest language is often something like Amharic, Karen or Dari rather than the one everyone assumes. Set up those three properly, then add the rest.

What does it do to call length?

Expect the routine calls to get shorter, because the wait for an interpreter disappears, and expect the complicated ones to get longer, because you are now having a real conversation you previously could not have. That trade is a good one — the long call happens the day before, on the phone, instead of at 3:20pm with a magnet sitting idle.

Do we still need our interpreter contract?

Yes. Keep it, and expect the minutes to shift rather than vanish. You will use fewer minutes on prep and scheduling and roughly the same or more on consent, results and flagged safety answers, which is exactly the reallocation you want.

Since almost all of this happens on the phone: CallSphere builds AI voice and chat agents that answer an imaging center's line and web chat around the clock, book and reschedule appointments, and capture the caller's details — including the language they actually called in — so the pre-visit conversation starts right and the record follows the patient to the next visit.