# The questionnaire arrived after the surgery

> An audit of 1,532 patients at Royal North Shore Hospital who did not complete their digital pre-operative questionnaire in 2025 found that much of the failure sat in the pathway, not with patients. 236 were invited two days or less before surgery, on the day or afterwards, 9 were invited after they had died, and more than a third never proceeded to surgery. The root cause was weak integration between the questionnaire platform and the electronic medical record.

*Source: https://getpathways.ai/articles/the-questionnaire-arrived-after-the-surgery*

Sometime in 2025, an invitation went out from Royal North Shore Hospital asking
a patient to complete their pre-operative health questionnaire. It was a
routine message, one of thousands the hospital's digital platform sends.

The patient had already died.

It happened nine times. Twenty-three other patients were asked to complete
their pre-op assessment after their operation was over. Twenty-one were invited
while they were acutely unwell, or already lying in a hospital bed.

None of these patients failed to engage. The system failed to notice.

These cases come from an audit presented at the ASA CSC by a team from RNSH's
Department of Anaesthesia, Perioperative Medicine and Pain Management and
Northern Sydney Local Health District. They are the starkest findings. They
aren't the most important ones. What matters more is what they reveal about
where digital perioperative care actually breaks, and who it breaks for.

## The front door of surgery

For most patients, the first real step towards an operation isn't a hospital
visit. It's a set of questions about their heart and lungs, their medications,
their previous anaesthetics, how far they can walk before they need to stop.

Answered weeks ahead, those questions buy the anaesthetic team time. Time to
notice the heart murmur nobody has investigated. The snoring that turns out to
be untreated sleep apnoea. The blood thinner that has to be stopped in advance.
The frailty that a few weeks of prehabilitation could meaningfully improve.

The answers also decide what kind of review a patient needs, from a short
phone call to a full pre-admission clinic. Early assessment and prehabilitation
are associated with better outcomes and lower healthcare use. RNSH has run its
pre-operative health questionnaire online since 2024.

The logic was sound. But a significant share of patients never completed it.
And when a questionnaire goes unanswered, the work doesn't disappear. It falls
to nurses phoning patients one by one, or it lands on the day of surgery, when
the options are fewest.

## Opening 1,532 files

The easy explanation for a missed digital questionnaire is the patient. They're
older. They're not comfortable with technology. English isn't their first
language. It's an explanation that asks nothing of the hospital.

The RNSH team decided to test it. Three junior medical officers took every
patient invited to complete the questionnaire in 2025 who didn't finish it,
1,532 people in all, and went through their medical records one at a time.
They looked at demographics, surgery type, clinic appointments, delays and
outcomes. The goal was to separate what the system could fix from what
genuinely belonged to the patient.

What they found reads less like a story about patients and more like a story
about three things: a clock, a broken connection, and the wrong channel.

## The clock

Start with when the invitations went out. Of the 787 non-completers with a
recorded interval between invitation and surgery, most were invited with weeks
or months to spare. But a large tail was not.

![Bar chart titled "30% were invited too late to answer": of 787 non-completers, 186 were invited more than 90 days before surgery, 365 three to 90 days before, 53 two days before, 160 on the day and 23 after surgery. The last three, 236 patients, are highlighted.](/assets/articles/rnsh-fig1-invitation-timing.svg)

*Figure 1. Timing of questionnaire invitations relative to surgery among 787
non-completers.*

Fifty-three patients were invited two days before their operation. One hundred
and sixty were invited on the day itself. Twenty-three were invited afterwards.
Together that is 236 people, 30% of the group, who were asked to complete a
pre-operative assessment when there was no longer any "pre" left to speak of.

No patient completes a meaningful health assessment in that window. More to the
point, no clinician has time to act on one.

The pattern gets worse when you sort patients by how urgently they need
surgery. NSW triages elective surgery into urgency categories, with clinically
recommended timeframes of 30, 90 and 365 days. The audit labels them A, B and C.

![Bar chart titled "The most urgent patients had the least time": median days from invitation to surgery were 4 for Category A (most urgent), 49 for Category B and 149 for Category C (least urgent).](/assets/articles/rnsh-fig2-median-days-by-category.svg)

*Figure 2. Median days from questionnaire invitation to surgery among
non-completers, by urgency category.*

Category C patients, the least urgent, had a median of 149 days between
invitation and operation. Category B had 49. Category A, the most urgent, had
four.

Four days. Within that window, the questionnaire has to reach the patient, be
opened, be completed, be reviewed by a nurse or anaesthetist, and be acted on:
a blood test ordered, a medication adjusted, a cardiology letter chased, a
clinic booked. Prehabilitation, which can genuinely change outcomes, needs
weeks.

The patients with the most to gain from early assessment were given the least
time to complete it.

The other end of the curve has its own problem. A questionnaire answered five
months before surgery describes the patient as they were in autumn, not the one
who arrives in spring with new medications or a fall in between. Early is good.
Early and never refreshed is not.

## The broken connection

The second thread runs through the patients who should never have been invited
at all.

Beyond the 21 who were invited while acutely unwell or admitted, and the nine
invited after death, the auditors found that 4.9% of non-completers had been
transferred to another hospital and 5.6% had asked to be removed from the list
or had declined surgery.

Follow the whole cohort through and the scale becomes clear.

![Flow chart titled "Where 1,532 missed questionnaires led": 385 (25%) were ophthalmology patients and 1,147 (75%) had other planned surgery. Of the 1,147, 814 (71%) proceeded to surgery and 333 (29%) did not. Of the 814, 724 (89%) had surgery at RNSH and 90 (11%) elsewhere. Across the full cohort, 541 patients (35.3%) never proceeded to surgery.](/assets/articles/rnsh-fig3-cohort-flow.svg)

*Figure 3. Outcomes for the 1,532 patients who did not complete their digital
pre-operative questionnaire.*

Of the 1,147 patients outside ophthalmology, 333 never proceeded to surgery and
90 had their operation somewhere else. Across the full cohort, 541
non-completers, more than a third, never proceeded to surgery.

Every one of those facts was recorded in the hospital's electronic medical
record. The admission. The transfer. The cancellation. The death. None of it
reached the questionnaire platform, so none of it stopped the invitation going
out. The authors put the root cause plainly: weak integration between the
digital platform and the EMR.

That gap has costs that rarely show up in a dashboard. There is the human cost
to a family who receives a reminder addressed to someone they have just lost.
There is the labour cost to the staff who upload patients, troubleshoot the
system and follow up on invitations that should never have been sent. And there
is a measurement cost. When more than a third of "non-completers" never went on
to surgery, a raw completion rate mixes patient engagement with waiting-list
churn, and makes patients look less willing than they are.

Then there is the silence. In 24.4% of cases, the record held no documented
reason for non-completion at all. Without a loop between platform and record, a
disengaged patient and an invitation that should never have existed look
exactly the same.

## The wrong channel

The patient factors are real, and the audit doesn't pretend otherwise. Of the
1,532 non-completers, 64% were over 65, and 14.5% had a preferred language
other than English.

The clearest signal comes from comparing surgical specialties.

![Scatter plot titled "Older patients, more missed questionnaires": incompletion rate against average patient age for 14 surgical specialties. ENT and gynaecology sit near age 50 and 11%; vascular, urology and plastics near age 61 to 67 and 19 to 20%; ophthalmology is highlighted alone at the top right, near age 72 and 23%.](/assets/articles/rnsh-fig4-age-vs-incompletion.svg)

*Figure 4. Incompletion rate against average patient age by surgical
specialty. Values read from the poster's Figure 2 and approximate.*

The older a specialty's patients, the more often they missed the questionnaire.
ENT and gynaecology, with average ages around 50, sat near 11%. Ophthalmology
sat alone in the top right corner: an average age in the low 70s and an
incompletion rate close to one in four. With 385 patients, it was also the
largest single group of non-completers.

There is a cruel logic to that. Many ophthalmology patients are having surgery
precisely because they can't see well. They are then asked to read and complete
a health questionnaire on a phone screen.

It is common to call this a digital literacy problem. That framing makes
completion the patient's job. A more useful framing is channel fit. A web form
asks for a smartphone or computer, good eyesight, confident English and some
familiarity with medical terms. A phone conversation asks for almost none of
that. Older patients, visually impaired patients and patients from culturally
and linguistically diverse backgrounds aren't necessarily unwilling. They are
being offered the wrong door.

## A different question

Put the three threads together and the original question starts to look like
the wrong one.

"Why don't patients fill in the form?" assumes the form arrived at a sensible
time, to someone who still needed it, in a format they could use. For a large
share of the 1,532, it didn't. The invitation came too late to matter, or went
to someone in a hospital bed, at another hospital, or no longer alive. Or it
arrived on a screen the patient couldn't comfortably read.

The better question is whether the pathway gave patients a fair chance to
answer. Framed that way, the audit points to three fixable problems: timing,
validity and access. The first two are integration problems, squarely within
reach of the hospital and its vendors. Even the third is less about the
patient's ability than about the channel the hospital chooses to offer.

## PAINITE: a pathway built around where patients fall through

The RNSH team's response is a program called PAINITE, short for Perioperative
Assessment & Intelligent Navigation, Integration & Triage Efficiency. It shares
its name with one of the rarest gemstones on earth, which suits a program built
to find the patients a pathway usually loses.

PAINITE doesn't tear up the existing pathway. It adds a tool at each point
where the audit found it stalling.

![Timeline titled "The proposed PAINITE pathway": request for admission, questionnaire, correspondence chasing, pre-admission clinic, day-prior phone call and surgery. Four arms attach to it: a voicebot conversation that takes the health history, a snifferbot that reads referrals and requests what's missing, a perioperative harness for rule-based triage, and simultaneous day-prior calls with issues flagged to staff. Ophthalmology is piloted first.](/assets/articles/rnsh-fig5-painite-pathway.svg)

*Figure 5. The proposed PAINITE pathway: four arms added to the existing
perioperative journey.*

The easiest way to see how it is meant to work is to follow a patient through
it.

Take an illustrative patient: a 79-year-old woman booked for cataract surgery.
Her failing eyesight is the reason she needs the operation. It is also why a
link on her phone is so hard to read.

Her pathway begins the moment the request for admission is lodged. Instead of a
link, she gets a phone call. A voicebot works through her health history in
conversation, asking follow-up questions where her answers need them. She can
ask what a question means. She doesn't need an app, a login or good eyesight.

Her GP's referral mentions a heart condition but not the details. Rather than
waiting for a nurse to notice the gap and start a round of phone calls, a
second tool, the snifferbot, reads the incoming correspondence, pulls out the
key facts, and automatically requests what's missing.

Once her information is complete, the perioperative harness triages her using
rule-based logic. It decides whether she needs a frailty coordinator, an
anaesthetic nurse or an anaesthetist, and it moves complex cases to the front
of the queue. Every routing decision can be traced back to the rule that
produced it.

The day before surgery, the voicebot calls every patient on the list at once to
go over fasting and medication instructions. For most patients, the call ends
there. If she mentions she hasn't stopped a medication she was told to stop,
her call is flagged to staff, who decide what happens next.

She arrives on the day assessed, prepared, and on the right list.

## Three design choices worth noticing

The first is that PAINITE uses language AI where the work is language. Talking
with patients and reading referral letters are exactly the tasks conversational
AI and document extraction do well, and exactly the tasks that consume clinic
staff time today.

The second is that it uses rules where a decision is made. Triage runs on
explicit, rule-based logic rather than a model's judgement. That makes every
decision explainable to a clinician, auditable by a governance committee, and
predictable from one week to the next.

The third is that it keeps people where judgement is needed. The day-prior
calls don't resolve a fasting breach or a missed medication stop. They find it
and hand it to someone qualified to decide. Automation handles the volume.
Clinicians handle the exceptions.

## Why start with eyes

PAINITE isn't designed to switch on all at once. The plan begins with a
voicebot pilot for ophthalmology patients, followed by a process evaluation for
quality assurance. Only then do the voicebot, snifferbot and perioperative
harness extend to other surgical specialties. AI projects in NSW public
hospitals are also assessed under the NSW AI Assurance Framework before they go
live.

Ophthalmology is the right first test because it is where the web form fails
hardest. It had the largest group of non-completers, the oldest patients, and
the barrier no amount of reminder messages can fix: patients who struggle to
see a screen. If a voice channel can lift completion there, the case for the
rest of the pathway largely makes itself.

## How we'll know if it works

The audit gives PAINITE a clear baseline, and the measures of success follow
directly from what went wrong. Invitations should go out with enough lead time
for each patient's urgency category, and Category A patients should no longer
be working with four days. Invitations to patients who are admitted,
transferred, removed from the list or deceased should be stopped before they
are sent. Completion in ophthalmology, and among patients who prefer a language
other than English, should rise.

The less visible measures matter just as much: the staff hours spent chasing
patients and correspondence, and the number of day-of-surgery problems such as
fasting breaches and medications not stopped. And the 24.4% of cases with no
documented reason should shrink. A pathway that knows why a patient didn't
respond can do something about it.

## The lesson for every hospital

Go back to that first invitation, the one sent to a patient who had already
died. It would be easy to dismiss it as a glitch. It is more useful to read it
as a symptom of a pathway that acts on a list while the information that should
change that list sits in another system.

Any hospital running a digital pre-operative pathway can ask the same questions
the RNSH team asked. When are our invitations going out, relative to surgery and
urgency? Who are they going to? What happens when the record changes? And who
is our channel leaving out?

The answers may be uncomfortable. They are also fixable. Before concluding that
patients won't engage with digital care, it is worth checking whether the
pathway ever gave them a fair chance to.

The questionnaire that arrives after the surgery isn't a fact about the
patient. It's a fact about the system. And systems can be redesigned.

*Source: Zhang D, Marks A, Doane M, Tinker B, Elhage T, de Solom T. Barriers to
current perioperative digital platforms – how can innovation solve these
issues? Poster, ASA CSC. Department of Anaesthesia, Perioperative Medicine and
Pain Management, Royal North Shore Hospital; Medical Workforce and Education
Unit, Northern Sydney LHD. Figures redrawn from the poster's data.*