Could a symptom-checker app shorten the long wait for an endometriosis diagnosis?

A digital symptom checker could plausibly shorten the notoriously long wait for an endometriosis diagnosis — a modelling study estimated a reduction of more than four years, while saving money — but the tools that exist have not yet been validated well enough to rely on [s1][s2]. The promise is real and the readiness is not, and the distance between the two is the whole story.

The problem the app is aimed at

Endometriosis affects about 10% of reproductive-aged women and is a leading cause of infertility, yet it carries an average diagnostic delay of up to 10 years [s2]. That delay is driven partly by the absence of a noninvasive way to identify the disease, which lets it progress while women cycle through appointments [s2]. It is one of the clearest cases in medicine where earlier recognition, not a new treatment, is the lever — which is exactly the kind of task a symptom-checker app is pitched to help with, by prompting the right questions and the right referral sooner. The corpus has covered how general symptom checkers perform on accuracy and triage; this is the same class of tool aimed at one underdiagnosed condition.

What the modelling study estimated

Researchers built a Markov decision process model — a simulation, not a trial — to compare a digital symptom checker for endometriosis against standard care, from a societal perspective over a 40-year horizon [s1]. In the model, the symptom checker reduced diagnostic delay by 4.36 years, generated 0.049 quality-adjusted life years per person, and saved $5,196.22 in costs, producing an incremental net monetary benefit of $10,089.00 at a $100,000-per-QALY threshold [s1]. Probabilistic sensitivity analysis put the net benefit at $12,398.92 (95% CI $11,893.11 to $12,904.72), and the tool remained cost-effective across a range of assumptions [s1].

Two caveats are load-bearing. First, these are outputs of a model, and a model's conclusions are only as good as its inputs and structure — the study is the first economic evaluation of its kind, not a measured result in patients [s1]. Second, the model itself specified the conditions under which the value holds: the greatest benefit required the checker to have a sensitivity and specificity of at least 0.7, compliance above 45%, and a time horizon of at least 10 years [s1]. In other words, the favourable economics depend on a tool that is accurate enough and actually used — neither of which can be assumed.

Whether the tools are good enough yet

That is where the second piece of evidence matters. A 2026 scoping review of endometriosis screening tools examined 18 studies, sorting the tools into four categories: questionnaire-based, app-based, machine-learning and AI-driven, and subtype-focused models [s2]. Of the 14 that reported psychometric performance, the area under the ROC curve ranged from 0.77 to 0.95, with the machine-learning and AI-driven models reporting the highest accuracy [s2]. An AUC of 0.7 is the floor the economic model required, so the better tools clear it on paper [s1][s2].

But the review's conclusion is a warning, not an endorsement. It found significant gaps across the tools, "particularly in external validation, comprehensive psychometric evaluation, and cultural adaptation" [s2]. A model that performs well on the data it was built on, but has not been tested on independent populations, is exactly the kind of tool that disappoints when deployed — and the cultural-adaptation gap echoes a problem seen across women's digital health, where products built and validated narrowly generalise poorly.

What it means for a reader

The case for a symptom-checker app in endometriosis is genuinely promising: the diagnostic delay is long, costly and harmful, and both a cost-effectiveness model and the accuracy figures of the better tools point in a hopeful direction [s1][s2]. But it is not yet a solved problem a reader should rely on. The economic benefit is modelled, not observed, and the tools remain under-validated on independent populations [s1][s2]. The reasonable use today is as a prompt to raise symptoms with a clinician sooner — a condition managed by symptom control, not cured — not as a substitute for the specialist assessment that still makes the diagnosis.

Sources

• [s1] Economic evaluation of a digital symptom checker for endometriosis using a Markov decision process model — npj Digital Medicine (2026)
• [s2] Bridging the Diagnostic Gap: Reviewing Current Endometriosis Screening Tools and Models — Journal of Women's Health (2026)