The following is a guest article by Keith Schmidt, MD
Every practice that sends patients home with pre-procedure instructions is already running a patient education program. It just is not staffed, measured, or budgeted, so nobody calls it that. The instructions go out on paper or through the portal, the patient goes home, and the questions start arriving through the phone tree, the inbox, and the nurse line. The work gets done by whoever picks up.
That informal system is now the target of a growing category of software: chatbots, text assistants, voice agents, and portal add-ons that promise to answer routine patient questions automatically. Some of these will be useful. Many will not. The practices that make good decisions here will be the ones that measured the problem before they shopped for a solution. This article is about how to do that measurement, and what to demand from any tool once you have the numbers.
Why the Problem is Invisible on Your Dashboards
Pre-procedure questions do not show up as a line item anywhere. A call about whether to hold a blood thinner is logged, if at all, as a generic phone encounter. A portal message asking where to park is closed by a medical assistant and disappears. A patient who cannot reach anyone at 9 p.m. and shows up unprepared appears the next morning as a cancellation or a delay, which is coded as a scheduling event rather than a communication failure.
Because the cost is spread across staff time, physician interruptions, and lost procedure slots, no single report captures it. Vendors know this, which is why their pitches lean on plausible-sounding but unverifiable savings. The only defense is your own data.
A Five-Day Audit Any Practice Can Run
You do not need an analytics platform to measure this. You need one workflow, five business days, and a tally sheet. Pick a single procedure type or a single clinic location. Then, for one week, have the people who actually field patient contacts record four things for every procedure-related question that reaches them:
The channel: phone, portal message, text, voicemail, or in person.
The category: logistics such as arrival time, location, parking, or driver requirements; medication questions; diet and fasting; post-procedure expectations; or something that needed clinical judgment.
Whether the answer was already in the written instructions the patient received.
How long it took to resolve, including any handoff to a nurse or physician.
Keep it aggregate. No names, no identifiers, no chart numbers. This is an operations count, not a clinical record.
At the end of the week, you will have something most practices have never seen: the actual volume of questions, the share that are repeats of instructions already given, how many required a clinician, and the staff minutes consumed. In our own experience running this kind of tally in pain and anesthesia clinics, the majority of contacts are logistics and clarification, and a large share are answered verbatim on the instruction sheet the patient was holding. Your numbers will differ, and that is the point. Get your own.
What the Numbers Tell You
If most questions are repeats of written instructions and cluster in the evening and early morning, you have an access problem, not a content problem. The instructions are fine. Patients cannot get a follow-up answer when they need one. That is the case where automation of some kind may help.
If a large share of questions require clinical judgment, automation is the wrong first move. You need a clearer escalation path to a nurse, and possibly better instructions.
If the questions are mostly about things the instructions never covered, you have a content problem. Fix the document before buying anything.
Only the first pattern justifies a software purchase, and even then only if the tool meets a short list of conditions.
Five Requirements For Any Tool That Answers Patients
Whatever a vendor calls the product, hold it to these.
The practice owns the content. Every answer a patient receives should come from instructions the practice wrote or approved, specific to its sites, physicians, and protocols. A generic library is how a patient learns the wrong arrival time. Ask to see exactly where the answers come from and who can edit them.
The tool retrieves rather than invents. A large language model that composes answers from its general training is a liability in a clinical setting. The safer architecture matches a patient’s plainly worded question to an approved answer and delivers that answer. If there is no approved answer, it should say so and route to a human. Ask the vendor what happens when the question is not covered.
Escalation is designed in, not bolted on. New symptoms, medications not addressed in the instructions, chest pain, bleeding: these go to a person, immediately, every time. Ask for the list of triggers and test them yourself.
The channel is one patients already use. A phone number they can call or text without downloading an app or creating a login will reach more of your patients than any portal. Adoption is the metric that matters, and a tool nobody uses saves nothing.
Every exchange is reviewable. If a physician cannot read what patients asked and what they were told, the tool is not accountable to anyone. Reviewability also turns the tool into a feedback loop: a question that keeps recurring becomes a line the practice adds to its own instructions.
The Measure is Prepared Patients, Not Message Volume
Vendors will report how many messages the system handled. That number is easy to inflate and tells you nothing. The outcomes that matter are the ones you counted in your audit: fewer repeat questions reaching staff, fewer clinician interruptions for non-clinical questions, and fewer patients arriving unprepared. Run the same five-day tally after implementation and compare. If the numbers did not move, the tool did not work, no matter what the usage report says.
Patient education around procedures is a clinical responsibility that the modern practice has quietly outsourced to a sheet of paper and a phone tree. Software may help close that gap, but only when the practice keeps authorship of the content, keeps judgment with clinicians, and keeps score itself. Count first. Then decide.
About Keith Schmidt
Keith Schmidt, MD, is an anesthesiologist and interventional pain physician in the Chicago area. He is a co-founder of PrepQ Health, a company that builds patient communication tools, and has a financial interest in it. No product is discussed in this article. He can be reached at [email protected].
