A "simple" web form: the hidden work behind a surgical eligibility simulator

One evening, someone types “eye surgery am I eligible” into a search engine. The first results offer simulators that disappoint in almost identical ways. Most of them show no result on screen and only send it by email, once the address has been collected. Others ask questions that are too complex for a patient who does not know their dioptres, or on the contrary too vague to distinguish anything. And almost all of them apply generic rules, with no link to the actual practice of a surgeon. Yet an eligibility simulator can mislead in both directions: it can falsely reassure a person who will be refused at the work-up, and it can falsely discourage a person who would have been an excellent candidate.

A clinic specialised in refractive surgery wanted the opposite. The surgeon wished to publish, inside an article explaining the eligibility criteria, a small interactive tool that gives the patient an orientation faithful to the real criteria of a preoperative work-up. A reliable opinion carries responsibility, even when a form produces it. This article tells the story of the work that this “simple form” required, and why that work is precisely what separates a useful tool from a gimmick.

Encoding the surgeon’s judgement, not generic rules

The value of this simulator does not sit in its code. It sits in the thresholds it applies, and those thresholds come from the surgeon himself. The ideal age for surgery lies between 25 and 35, because refraction is stable there and presbyopia still far away. Between 45 and 55 opens a grey zone, where one hesitates to replace the still functional lens of a young presbyope. After 55, lens surgery becomes the reference option. Astigmatism beyond 3.5 dioptres corrects poorly with a laser. Monovision is a discussion about lifestyle, because the trade-off is not the same for a person who works on a screen and for a driver who drives at night.

None of these rules comes out of a generic handbook. They reflect the way this surgeon actually decides in consultation. Development therefore took the shape of a loop: the AI proposes a first version of the decision engine, the expert corrects it, and the correction gets encoded.

A concrete example: contact sports. The first version of the engine ruled out laser surgery for a boxer. The surgeon corrected it: you do not exclude LASIK for contact sports, you exclude the ICL implant, because the risk is that an impact knocks the implant against the crystalline lens. And PRK may be preferred depending on the correction. This rule was refined twice before it was right. It is the conviction that runs through our automation method: the AI speeds up the shaping, the expert remains the owner of the rules.

Designing a cautious opinion

A medical tool for the general public must never produce a diagnosis. The simulator produces an orientation, and its whole design follows from that nuance.

Each technique receives a score on three levels: probably suitable, to discuss case by case, not recommended at first sight. The engine evaluates each eye separately, then combines the two by keeping the worse of the two eyes. This simple rule avoids a dangerous trap: a very good eye must never mask a problematic one.

Every edge case receives its own dedicated answer. If the refraction has not been stable for two years, all verdicts become provisional and the simulator says so. If the right eye and the left eye lead to different techniques, the tool shows the two orientations side by side and adds a warning: very different eyes can present a limited best corrected acuity or a latent strabismus, which can contraindicate surgery. If the person does not know their correction, the tool does not guess: it explains that it cannot orient without that information, and it says where to find it.

We also chose what the form does not ask. Dry eye matters in the decision, but a patient cannot self-assess it reliably. It therefore stays out of the questionnaire, and the disclaimer mentions it explicitly among the parameters that only the preoperative examination will evaluate. An honest form only evaluates what it can evaluate.

The disclaimer, precisely, is not a legal notice endured in small print. It is an argument of transparency, displayed in plain sight under every result: the simulator orients, only the preoperative work-up decides. That sentence sums up the ethical position of the whole tool.

An experience for the novice and the connoisseur alike

The same form must serve a person who does not know their correction and a person who reads their prescription fluently. This double audience dictated almost every interface choice.

The form never asks the user to type dioptres into a numeric field. It offers descriptive ranges, with the values in brackets for the informed reader: a moderate myopia is described in words, then specified in dioptres. Every sensitive question offers an exit door, “I do not know my correction”, which leads to a useful message rather than to a distorted result. This granularity avoids the two pitfalls of common simulators: the overly technical question that blocks the novice, and the overly vague question that distinguishes nothing.

The flow advances on its own where it can: on a single-choice question, the selection triggers the next step, with no button to hunt for. At the end, the result first recalls the answers given, then shows the techniques ranked with badges readable at a glance. The novice reads the colours, the connoisseur reads the thresholds. When the two eyes diverge, two columns align the evaluations eye by eye.

Sunday evening, on a sofa. A person who has been short-sighted since high school answers the five questions in two minutes, without creating an account. The result tells her that her left eye, far more myopic than the right, points rather towards an implant, and that this difference between the eyes deserves a discussion in itself. She arrives at the consultation with the right questions instead of a certainty found online.

Privacy by construction

The simulator runs entirely in the browser. It sends no request to any server, it records nothing, and it shows no email wall before the result. A banner announces it from the introduction: no data is recorded or transmitted, everything is calculated in your browser.

This choice is not a regulatory constraint to endure. It is a feature. Answers about eyesight and age constitute health data, and the safest way to protect health data remains never to collect it. In a landscape where most simulators only deliver their result by email, a tool that asks for nothing creates precisely the trust that makes people want to book an appointment.

Under the hood: what “simple” demanded

The final component fits in a single self-contained file: the HTML, the styling and the JavaScript logic, with no framework and no external dependency. The three languages of the site, French, English and Dutch, live in the same file, so the tool drops into any page with a single line. It now serves the original article and the site’s FAQ, with no code duplication.

The invisible part demanded more care than the visible one. The decision engine was validated by a test harness that replays fourteen patient profiles, from the mild myope to the complex case, and checks that each profile receives the expected orientation. The mobile rendering was verified in a headless-driven browser, at widths of 390 and 320 pixels: that verification revealed a horizontal overflow in the two-column display, which nobody would have seen from a desktop screen. Even the scrolling has its detail: the height of the quiz changes at every step, so the scroll anchor attaches to a stable heading rather than to the form itself, to avoid screen jumps.

And in your sector?

This project is medical, but its pattern is not. It carries over to any place where an expert applies rules carried in their head: the pre-qualification of a credit file, the screening of applications before interviews, the triage of a legal request, an insurance quote, eligibility for a public benefit.

Four invariants sum up what this form confirmed for us. Your business rules are the main asset, and the AI serves to shape them fast, not to replace them. A cautious opinion, which knows its limits and says so, is worth more than a false certainty. A two-level interface, readable by the novice and credible for the connoisseur, widens the audience without sacrificing rigour. And privacy is designed from the start, because it cannot be bolted on afterwards.

This is the approach we apply to every project, described in our method: start from the real process, encode the expert’s rules, keep the human in charge of the decision. If your work rests on rules that only you know how to apply, transtorm.ai can turn them into a tool your clients use with confidence. Book a conversation below.