Health System Phone Access Benchmarks: Hold Times, Abandonment and Booking Leakage
· Alex Dalyac, Founder & CEO
Health system scheduling lines fail twice: patients wait to reach a person, then hear the next slot is weeks away. Public targets are a 30-second answer and under 5% abandoned; most VA call centers miss them.
An AI receptionist in front of the call center answers every call, books in the EHR, and hands clinical calls to staff. Below: benchmarks, our data, and pilot thresholds.
Why are health system scheduling phone lines so hard to get through?
Centralized patient access centers are staffed for average volume, so Monday mornings, lunch hours and evenings overflow into holds, phone trees and callbacks. The best public benchmark comes from the Veterans Health Administration, which sets a target of answering in 30 seconds or less with 5% or fewer calls abandoned.
Across 285 VHA facilities, average speed of answer was 87 seconds in October 2014 and 69 seconds in September 2016, and nearly 80% of call centers still missed one or both targets (Griffith et al., AJMC 2019).
| Measure | Public target | Observed |
|---|---|---|
| Average speed of answer | 30 seconds or less (VHA) | 87 s in 2014, 69 s in 2016 across 285 VHA facilities |
| Abandonment rate | 5% or less (VHA) | 12% in 2014, 8.3% in 2016 |
| Hold time after the IVR | 2 minutes or less (CMS, Medicare plan lines) | Measured quarterly by CMS secret shoppers |
| Calls that reach no one | Not tracked by most centers | 14-20% of primary care calls arrive after hours or with all lines busy (our data) |
Answer speed mattered: patients at the slowest-answering VHA sites were less likely to say they got urgent appointments when needed.
The CMS figures come from its Part C and Part D call center monitoring, which measures hold time from the end of the IVR or greeting until a live person answers. It applies to Medicare plans, not hospitals, but is a fair bar for a scheduling line.
What happens once a patient reaches a scheduler?
Reaching a person does not guarantee a timely visit. In one large US health system's online booking in October 2026, the average wait to a third open new-patient slot was 50 days, and 77% of doctors couldn't see a new patient within two weeks. For 46% of them, a similar doctor under 30 minutes away could.
In our October 2026 mystery-shopper calls to large US health systems, schedulers rarely bridged that gap on their own.
| What the caller asked for | What the scheduler did | Share of calls |
|---|---|---|
| A booked-out doctor, no hint | Offered a sooner similar doctor unprompted | 8% |
| A booked-out doctor, then asked "anyone sooner?" | Found one after searching a median 88 seconds | 67% |
So 92% of sooner options were missed unless the patient knew to ask, and finding one took a minute and a half on hold.
Nationally, the average wait for a physician visit is now 31 days (AMN Healthcare 2025 survey). Patients who hear a long wait and hang up to look elsewhere are booking leakage the call center never sees.
Where should automation stop when symptoms come up?
Automation should stop the moment a caller describes symptoms the practice has defined as urgent: the call goes to a nurse, the on-call provider or a staff queue with a summary, never to a booking flow. The practice writes those rules; the AI applies them on every call.
In our data, urgent clinical calls are a small but steady share, and they cluster after hours.
| Setting | Urgent clinical share | What the AI does |
|---|---|---|
| Primary care, two New York practices | 6% of business calls | Flags and transfers to staff |
| Gastroenterology, after-hours calls | 19% of after-hours calls | Flags post-procedure symptoms to the on-call team |
| Behavioral health, small practices | 3.5% of business calls | Flags urgent calls to the clinician |
Most after-hours calls are routine, so a single on-call queue for all of them buries the urgent ones.
The other 81% of after-hours GI calls were booking, records, billing, rescheduling and prep questions that can be handled without waking anyone. See how warm transfer works.
How does an AI receptionist fit in front of a call center?
It answers first, on every line, and resolves what it can: booking, rescheduling, cancelling, insurance acceptance, directions, forms. Everything else is transferred to the right team with a summary, so schedulers start each transferred call already knowing the reason.
| Option | Answers at peak | Books in the EHR | Typical cost |
|---|---|---|---|
| IVR phone tree | Yes, then queues to a person | No | Low per call, but adds hold time |
| Outsourced or offshore call center | Depends on staffing | Yes, with EHR access | Per agent hour |
| Answering service | Yes | Usually takes a message | About $1.50 a minute or more |
| AI receptionist | Yes, every call at once | Yes, directly | From $1 a call |
At a 99-second median call, per-minute billing costs more than twice a per-call price.
The AI books into the same EHR your digital front door uses (Epic, athenahealth, eClinicalWorks and others), and can text callers your patient portal or intake link. It also searches similar doctors' schedules in about 1 second, versus 88 seconds for a human, and offers a sooner option when the first choice is booked out (load rebalancing).
Results: 75% of calls resolved without a human
At a 12-location gastroenterology group in New York where the AI answers all calls, 75% of calls were resolved without a human in September and October 2026, and 25% were transferred. 33% of calls arrived outside 8 am to 6 pm on weekdays.
| Measure | Gastroenterology, 12 locations | Primary care, two practices |
|---|---|---|
| Resolved without a transfer | 75% | 52% (staff take transfers by design during open hours) |
| Slot searches ending in a booking or reschedule | 54% | 69% |
| Median call length | 126 s | 99 s |
| Calls outside 8 am-6 pm weekdays | 33% | 14-20% after hours or all lines busy |
32% of GI cancellations and reschedules came after hours, freeing slots that can be refilled the next morning.
When offered a sooner similar doctor, 42% of GI patients took it and were seen 3x sooner (median). The primary care figures cover over 8,000 business calls from April to October 2026.
What would need to be true before a health system pilots AI phone agents?
Four things: a measured baseline, written escalation rules, direct EHR booking, and a narrow first scope. Pilot one service line or one overflow window (after hours or busy-line overflow), not the whole center.
| Pilot item | What to measure or decide |
|---|---|
| Baseline | Answer speed, abandonment, hold after IVR, calls arriving after hours |
| Escalation rules | Which symptoms transfer, to whom, by hour; fallback when no one answers |
| Booking | Visit types and providers the AI may book; which providers are interchangeable |
| Success | Share resolved without a human, booking rate on slot searches, abandonment change |
Start with overflow: those calls reach no one today, so any resolution is new capacity.
How to set it up
We connect to your EHR, load your providers, visit types and escalation rules, and route overflow or after-hours calls to the AI first. Staff keep their lines; the AI takes what they can't answer.
See how YouSquared works for primary care, gastroenterology, cardiology, orthopedics, OB/GYN, and other healthcare practices. Start free on the web or book a demo to walk through your call flows with our team.
Key terms
- Average speed of answer (ASA). Average time from a call entering the queue to a person answering. The VHA target is 30 seconds or less.
- Abandonment rate. Share of callers who hang up before reaching anyone. The VHA target is 5% or less.
- Booking leakage. Patients who call, hear a long wait or never get through, and book with another provider.
- Third next available appointment. Days to a provider's third open slot, the standard access measure.
FAQ
Why are hospital clinic phone lines so slow? Calls are routed to a central center staffed for average volume, behind an IVR. At peaks, holds build; in the VHA, nearly 80% of centers missed answer-speed or abandonment targets.
Can health systems use offshore call centers for Medicaid patients? CMS guidance SMD #10-026 (2010) bars Medicaid payments to entities outside the US for State plan services, but says outsourcing call centers for enrollment or claims is not prohibited. Check your state and payer contracts.
Does open access scheduling work in primary care? Partly. A systematic review of 24 studies found every study measuring it cut third-next-available waits, but only 25% got under 48 hours, and no-shows improved only where they started above 15%.
Should we put an AI receptionist in front of our digital front door? Yes, on the phone line. It books into the same EHR and texts patients your portal link, so phone and online booking stay in one schedule.
Is it secure for a health system? Yes. YouSquared is HIPAA compliant, SOC 2 audited, and signs a business associate agreement (BAA). See Security.