UveitisDDx user manualدليل استخدام UveitisDDx

Level 0. Quick start 1 minute Free

1. Choose the region in the top bar (on a phone: ☰ Menu → Region). If unsure, leave Worldwide. On the Free plan Worldwide is the only region.
2. Enter the profile. On the Case tab, open Profile and set age at onset, duration, course and anatomical class (SUN). Set Laterality here (unilateral or bilateral): your answer is always the one used. If you leave it blank, it is worked out from the eyes in which you record findings as present (see segments).
3. Tap findings in the other tabs. One tap means present (green). The × button, or a second tap on an OD/OS eye button, means looked for and absent (red). A third tap on an eye button clears it.
4. Open Results. The ranking updates after every entry.
5. Read Must not miss first, then Test before steroids, then the ranked list. The bars are a relative weight of evidence, not a probability.

The differential is computed on your device. Nothing is sent anywhere unless you save a case, make a PDF or opt a case into pooled learning. Do not type names, dates of birth or record numbers.

The nine segments, and recording each eye Free

The Case tab holds nine segment tabs, in the order you examine: Profile, History, Symptoms, Anterior, Vitreous, Fundus, Imaging, Labs, Systemic. A number on a tab counts the entries already made in it. The search box above the tabs searches every segment and the synonyms (type "snowballs" or "placoid"); the ⓘ button beside a finding shows its definition. The app remembers the segment you were on.

• Profile: demographics, pattern (onset, laterality, duration, course, SUN class) and host.
• History: surgery and injections, injury, drugs and vaccines, exposures, host factors and the response to treatment (prompt response, relapse on tapering, worse on corticosteroid, little or no response).
• Anterior, Vitreous, Fundus, Imaging: the examination, front to back. Imaging covers FA, ICG, OCT, autofluorescence, OCT angiography, ultrasound, electrophysiology and systemic imaging.
• Labs and Systemic: HLA typing, serology, ocular fluids, then musculoskeletal, skin, gastrointestinal, mucocutaneous and neurological findings.

Per eye. Most examination findings carry two buttons, OD (right) and OS (left); graded findings (for example vitreous cells) show a row of levels for each eye. The engine reasons at patient level: a finding present in either eye counts as present, and one counts as absent only if you recorded it absent in at least one eye and present in neither. Laterality is the answer you set in Profile; if you leave it blank, it is worked out from the eyes in which you record findings as present (one eye: unilateral; both: bilateral). Either way, record each sign in the eye where you see it.

Present, absent, not examined Free

These are three different statements, and the engine treats them differently.

For laboratory and imaging tests the same buttons mean positive (present) and negative (×). A test not done stays untouched and is never read as negative.

Why absent findings carry weight. A quiet, white eye argues against anterior-segment disease; no hypopyon, no synechiae and no keratic precipitates are information. Record a negative only when you looked.

“Fully examined — nothing else found.” The Anterior, Vitreous, Fundus and Imaging tabs each have this switch. It records that you completed that part of the examination. It does not change the live ranking, and unticked segments are not penalised. It matters for pooled learning (see Pooled learning): only in a ticked segment does a finding you left unrecorded count as absent for the research pool.

The chip bar Free

The strip under the top tabs, on the Case and Results screens, lists the whole case as chips: categorical answers (“Age at onset: Over 60 years”), present findings, and absent findings prefixed with ×. Per-eye entries show the eye: “Floaters (OD+OS)”, “× Red eye (OD+OS)”. Tap a chip to remove it. Reset clears the case after a confirmation; Restart in the header and ☰ → New case do the same. When you are editing a visit of a saved case, a Save visit chip appears at the end.

☰ → Share link copies the whole case as a URL (variable and level codes only, never free text); opening it restores the case.

Reading Results Free + PRO

• The ranked list. Each diagnosis shows its urgency tag (Emergent, Urgent, Routine), a one-line description, a bar and explains x/y: how many of your present findings the diagnosis accounts for. Tap a row to open it: what it explains, what it does not explain, what is against it (findings you marked absent), typical findings you have not recorded (tap one to go and ask), its investigations, teaching pearls and pitfalls, and a link to EyeWiki.
• Groups. Closely related entities are grouped, for example “Intermediate uveitis, 2 members”. The group takes the rank of its best member; tap it to see the members.
• The score is a relative weight of evidence, not a probability. The underlying data are published frequencies of varying quality, not a validated epidemiological model. Read the list as “which diagnoses this picture favours, and roughly by how much”. Never read 85 as an 85% chance of disease.
• Strips and banners. A red or amber Safety alert strip counts the Must not miss entries. An Unexplained alarming finding banner means the leading diagnosis leaves an alarming finding unexplained; that finding needs its own differential.
• Free plan: the three leading diagnoses, then a line “n more diagnoses in Pro”. Pro: the full list.

Must not miss Free

The panel on the right of Results (below the list on a phone). It answers a different question from the ranking: what can I not afford to be wrong about? It is built from rules, not from rank, so a diagnosis can appear here while ranked far down the list. An entry gets in for one of these reasons, shown under its name:

• A safety rule fired. For example, chronic vitritis in an adult over 40: exclude vitreoretinal lymphoma.
• It is plausible on the whole picture, or fits part of the picture (an alarming finding the leading diagnosis does not explain).
• It is a look-alike to exclude: the dangerous mimic or association of a leading diagnosis (for intermediate uveitis: sarcoid, TB and MS).

Entries are ordered by urgency times plausibility, so an emergency stays visible even when it is not the top explanation. This panel is complete on the Free plan, always.

Test before steroids Free

The panel under Must not miss. It lists the tests to have in hand before starting corticosteroids or immunosuppression, because those drugs can worsen or mask an infection or a lymphoma. Each line names the test and what it is for (“Syphilis serology”, “IGRA and chest imaging”, “Diagnostic vitrectomy and MRI brain”, “HIV test and CD4 count”, and others such as aqueous PCR for herpesviruses or Toxoplasma serology). A line appears when a matching diagnosis is plausible or a rule fires on your findings, not because of its rank. The tick box is read-only: it fills in when you record the corresponding test in the Labs or Imaging tab. Like Must not miss, it is complete on the Free plan.

Ask next Free + PRO

The strip at the top of the ranked list. Rather than asking you to fill two hundred fields, the engine names the single question, sign and test whose answer would most reduce uncertainty among the leading diagnoses per unit of cost. Chips read Ask:, Examine: and Test: followed by the finding. Tap a chip: the app jumps to that row in the Case tab and highlights it. Record present or absent and return to Results; the suggestions are recomputed.

Free shows one suggestion; Pro shows all three. The worked example below shows one answer flipping a case.

Why this rank PRO

Open a diagnosis and press Why this rank. The table shows every term of its score, and the rows sum exactly to the total:

Read it as an audit: a leading diagnosis resting on grade C cells or on a single heavy term deserves less trust than one supported by several cited rows.

Classification criteria PRO

A panel on the right of Results checks the case against the published criteria that are relevant to the leading diagnoses: SUN Working Group 2021 sets (for 25 uveitic entities) and the International Criteria for Behçet Disease 2014. Each set shows its items with ✓ met, ✗ not met or ? not recorded, and an overall status: Met, Met (exclusion not tested), Possible, Not met or Excluded. Recorded and Record next chips jump to the rows that confirm or complete the criterion. Items read as absent because you did not record them are flagged assumed absent; items answered by a close substitute are flagged proxy criterion.

Classification is not diagnosis. The criteria were written to define homogeneous groups for research, with high specificity; a case can be real disease and fail them. They are shown beside the ranking and never change a score. A set that is Met after the exclusion work-up supports the diagnosis; one that is Not met does not exclude it.

Targeted work-up PRO

A collapsed panel under Test before steroids; tap its title to open. For the case in hand it lists the investigations that matter, with Why (a Test before steroids item, or its role: confirm, exclude, baseline), Moves and Done. Moves reads like Diagnosis X% → Present: Y% / Absent: Z%: the weight of the leading diagnosis before the test and after each possible result. Use it to pick the test that most changes your management, and to know in advance what a negative result buys you. Done ticks when you record the result.

Regions and fallback PRO

Disease frequencies and the mix of diseases differ by place: Behçet in Riyadh is not Behçet in Boston. The Region menu in the top bar offers six priors: Worldwide, Middle East (pooled), Gulf (Saudi Arabia, UAE, Kuwait, Qatar, Oman, Bahrain), Levant / Egypt (Jordan, Syria, Lebanon, Palestine, Iraq, Egypt), Europe / UK and North America.

Fallback. For each disease and finding the engine uses the most specific regional value that exists and otherwise goes up one level: Gulf → Middle East → Worldwide, Levant / Egypt → Middle East → Worldwide, Europe and North America → Worldwide. Where a region has no data of its own you are therefore looking at the wider figure, not at nothing. Set your default region under ☰ → Account (Default region); a saved case also stores its own region.

The Free plan is fixed to Worldwide: the other regions are listed marked PRO.

Worked example: the “intermediate uveitis” that is not Free + PRO

A woman of 72 is referred as intermediate uveitis, not responding. Bilateral, longer than three months, chronic, a white eye without pain, complaining of floaters. Region: Worldwide. Press ☰ → New case first.

The lesson: a quiet white eye with chronic vitritis in a patient over 60 is not a pars planitis until lymphoma has been excluded, and a poor steroid response is information. On the Free plan steps 1 to 4 work exactly as shown (Ask next gives that first suggestion); steps 5 and 6 are Pro.

My cases Free + PRO

A private, de-identified case log for signed-in users. New case asks for a title, age (0–90; 90 means 90 or older), sex, region and tags. No names, dates of birth, record numbers or dates: notes and tags are scrubbed of anything that looks like one, and you are warned before saving. Visits are dated only as a day offset from the first visit.

• Visits. New visit copies the previous visit into the Case tab; update it and press the Save visit chip. Each visit stores the findings, the knowledge-base version that produced the ranking, and the engine's top-ranked diagnosis at that time, so you can see the picture evolve. Load reopens a visit for editing.
• Final diagnosis. Search the diagnosis (or type free text) and grade the certainty: 1 Proven (organism or cells identified), 2 Meets classification criteria, 3 Clinical, stable 3+ months, 4 Presumed. Tick the basis (PCR, culture, biopsy, serology, criteria, imaging, course). Cases wait in the Awaiting final diagnosis filter, and are flagged after 90 days.
• Outcomes. Visual acuity OD/OS and complications (CME, glaucoma, cataract, band keratopathy, hypotony, CNV, ERM, RD, phthisis).
• Private reference. A field for your hospital number so you can find the patient again. It is stored only in this browser, never sent to the server and never in an export.
• PDF PRO: a case summary for the notes or for teaching.
• Offline. Edits made offline are kept on the device, shown as pending sync, and sent when you reconnect.

Free keeps up to 5 cases. Pro is unlimited and adds follow-up visits with outcomes and the PDF.

Pooled learning (opt-in) Free

The engine starts from the literature. With your permission, confirmed cases can later refine the literature priors. In a case's detail, the switch Share de-identified case for research is off by default, is set per case, and can be switched off again at any time.

What is shared (only for opted-in cases that have a final diagnosis): region, an age band (not the raw age), sex, the final diagnosis, its certainty and basis, whether the engine was consulted, the number of days from first visit to the final diagnosis, a coarse creation month, and the structured findings of the first visit (answers, examination findings, and which segments you ticked as fully examined). Each row gets a fresh random identifier that cannot be traced to a patient or a clinician.

What is never shared: your identity or account, case or user identifiers, titles, notes, tags or any free text, the private reference, and calendar dates.

How it is used. A case counts only if its final diagnosis has stood for 30 days; certainty 4 (presumed) never teaches; certainty 3 teaches findings but not tests; and a disease is touched only once at least 20 such cases exist. Examined findings count, and a finding counts as absent only when you ticked Fully examined for its segment; a test counts only if it was done. Pooled data update the literature value, they never overrule it, and a learned value never reaches 0% or 100%.

Nothing reaches users automatically. A candidate knowledge base has to beat the current one on blind vignettes and on held-out real cases, an editor must sign off every large change, and it is then published as a normal, versioned release. Institutional use of real patient data also needs that institution's ethics approval.

Offline use and installing the app PRO

• Install. Use your browser's own install option: in Chrome or Edge the install icon in the address bar, or the browser menu → Install UveitisDDx; on iPhone or iPad, Share → Add to Home Screen. The app then opens full-screen like any other.
• How long it works offline. The app and the knowledge base are cached on the device. The offline licence runs for up to 30 days from the last time you opened the app online (or until your Pro period ends, if earlier), and renews every time you start it online. After that, reconnect once and it renews. An Offline pill shows in the header when you have no connection.
• What refreshes on reconnecting. A newer knowledge-base release is downloaded (“Knowledge base updated to version …”), and cases and visits saved offline are synced to your account.
• Signing out removes the offline copy from that device.
• The Free plan is online-only (the pill reads Offline (Pro works offline)).

Teaching mode Free + PRO

Open the Teaching tab (or ☰ → Teaching mode). It holds a bank of more than a hundred clinical vignettes across the diseases in the knowledge base.

1. Read the vignette. The Case n / N button opens the Teaching Case Bank (search and category filters); Previous and Next step through the vignettes.
2. Commit your differential first: up to five diagnoses in the “Add a diagnosis…” boxes.
3. Press Reveal the engine ranking. You see your score, the expected diagnosis, the must-not-miss diagnoses you did not list, the teaching point and the engine's differential with an EyeWiki link for each. Hide and try again resets the attempt.
4. Open in Case Engine loads the vignette into the Case tab so you can add findings, follow Ask next and open Why this rank.

Scoring: 3 points if any expected diagnosis is on your list, 1 more if your first choice is an expected diagnosis, and 1 for each must-not-miss diagnosis in the vignette's rubric that you listed. With residents: have them commit and discuss before revealing, then spend the time on the must-not-miss they omitted. Free: the first 5 vignettes. Pro: the full bank.

Export and delete your data Free

• Export a case. In My cases, open a case: Export JSON (the whole case with visits, final diagnosis and outcomes), Export CSV (day offset and notes per visit) and PDF (Pro).
• Delete a case. Delete case removes the case and all its visits after a confirmation. It cannot be undone.
• Delete your account and all your cases. Open ☰ → Account → Your data: Export all my data downloads everything as JSON or CSV, and Delete my account and all my cases removes the account after you enter your password and confirm. It cannot be undone. If you cannot sign in, write to support@idoc.site from the e-mail address of your account. Financial records held by the payment processor are kept as the law requires. See the privacy notice.
• On the device. Signing out clears the offline copy of the knowledge base from that browser. The private reference of each case lives only in the browser and disappears with its site data.

Pitfalls and when not to trust the ranking Free

• Blank is not zero. An unrecorded finding counts for nothing. But a × you did not earn counts against every diagnosis that usually shows that finding. Tap × only for what you actually looked for, and for tests only for results in hand.
• Masquerades. Lymphoma, infection and other non-inflammatory mimics do not announce themselves. Read Must not miss and the Unexplained alarming finding banner every time, whatever the ranking says.
• Prior steroids and immunosuppression. Treated eyes are quieter and have fewer signs, so a sparse, flat or unexpectedly bland differential after steroids reflects suppressed disease more than an absence of it. Record the response to treatment (History), because worse on corticosteroid and little or no response are strong evidence in their own right. Review Test before steroids before starting or escalating.
• Few findings, weak ranking. With a handful of entries the order is largely the prior of the profile and region. Wait for Ask next to do its work.
• Unspecified types. A finding marked type not recorded counts toward every subtype; choose the subtype when you can.
• Weak evidence. Rows with a grade C or Consensus source are expert opinion, not a series. Check Why this rank before leaning on a diagnosis supported by them.
• The score is not a probability, the knowledge base holds a fixed list of diseases (a disease not in the list cannot appear), and the criteria panel classifies, it does not diagnose.
• Do not use the ranking to exclude a sight- or life-threatening diagnosis. Examination, imaging and the laboratory decide.

Clinical status Free

Free or Pro?

Free, for good: full case entry (nine segments, per eye, present or absent), the top three diagnoses, Must not miss, Test before steroids, one Ask next suggestion, Worldwide priors, up to 5 saved cases, the first 5 teaching cases, the Arabic interface and opt-in pooled learning. Anything that protects the patient is never behind a paywall.

Pro (US$ 7.99 per month or US$ 59 per year): the full differential, all Ask next suggestions, Why this rank with PubMed references, classification criteria, targeted work-up, all six regional priors, unlimited cases with follow-up visits and outcomes, PDF case summary, offline use and the full teaching bank. Side by side: Plans and pricing.

Is there a trial?

Once plans are open, a new account starts with a 14-day Pro trial, no card needed. At the end it continues on the Free plan unless you subscribe. Nothing you saved is deleted. My plan in the ☰ menu shows the date your trial or period ends.

How do I pay?

Sign in, open Plans and pricing and choose monthly or annual. Payments are processed by Paddle, our Merchant of Record: Paddle is the seller, collects the applicable VAT or sales tax, and issues your invoice and receipt. Your card statement shows PADDLE.NET. The final amount, tax included, is shown at checkout before you pay. Subscriptions renew automatically until cancelled. If a renewal fails, Pro stays on for 7 days while the payment is retried. Change the card or download invoices under Manage billing (on the pricing page, or in My plan).

Trainees can get Pro for about US$ 35 per year by coupon: write to support@idoc.site from your institutional address.

Can I cancel, and can I get a refund?

Cancel at any time under Manage billing. You keep Pro until the end of the period you paid for, then the account moves to Free. A full refund is available on request within 14 days of any payment, with no reason needed: see the refund policy. Trials and complimentary access have nothing to refund.

What happens to my cases if I stop paying?

Nothing is deleted. All your cases stay readable and editable. The Free plan holds 5 cases: beyond that you cannot add more to your account until you return to Pro. A new case created while over the limit is kept on that device only (marked this device only) and syncs by itself if you go back to Pro.

Institutions and complimentary access

Institutions: five or more seats on one invoice. Complimentary access: for colleagues in countries where online payment is not possible, and for reviewers and teaching faculty. Write to support@idoc.site and we will arrange it.

Support

Write to support@idoc.site. The most useful report is a case the engine gets wrong: describe the findings you entered (no identifiers), what it ranked and what you think is right.

Plans and pricing · About & privacy · Terms · Privacy · Refunds