12-phase flow (12)

1. 1FRAME
   Frame as a SIGHT-THREATENING established corneal ulcer requiring pathogen-stratified work-up + intensive topical therapy — NOT undifferentiated red-eye triage (that is ophtho.acute-red-eye.core.v1, which routes here). The four-pathogen partition (bacterial / fungal / Acanthamoeba / herpetic) and the scrape-before-empiric / no-patch / no-empiric-steroid spine are set here. Perforation → endophthalmitis is recognised then routed OUT.
   advance: definitive-keratitis scope confirmed; upstream-triage and endophthalmitis concerns delineated by engine_id
2. 2ENTRY
   Recognise the presenting pattern: painful red eye + corneal infiltrate/opacity; contact-lens red eye; vegetative-trauma ulcer; prior-HSV dendrite/zoster; or pain-out-of-proportion ± ring infiltrate. Capture which routing engine sent the patient (acute-red-eye / conjunctivitis / corneal-abrasion) and carry over their findings.
   actions: workup.acute_red_eye
   advance: keratitis entry pattern recognised; upstream carryover state ingested
3. 3CONTEXT
   Build the pathogen prior: contact-lens wear + hygiene + water exposure (Pseudomonas / Acanthamoeba); vegetative/organic trauma (filamentous fungal); prior HSV / V1 zoster (herpetic); immunocompromise / ocular-surface disease (Candida, atypical); pain severity relative to signs; infiltrate morphology + margin; epithelial-defect/fluorescein pattern. This phase assigns the bacterial-vs-fungal-vs-Acanthamoeba-vs-HSV pretest distribution.
   inputs: contact_lens_use_and_hygiene, mechanism_of_trauma_vegetative_matter, water_exposure_lens_in_water, prior_hsv_or_zoster_ocular_disease, pain_severity_relative_to_signs, infiltrate_morphology_margin, epithelial_defect_and_fluorescein_pattern, immunocompromise_or_ocular_surface_disease
   actions: workup.acute_vision_loss
   advance: pathogen pretest prior assigned by risk factor + morphology
4. 4RED_FLAGS
   Recognise the ulcer that needs same-hour ophthalmology + admission: hypopyon / dense AC reaction; ≥3 mm / central / deep stromal vision-threatening ulcer; impending or actual perforation (descemetocele, Seidel-positive) — perforation → endophthalmitis arm routed OUT to ophtho.acute-vision-loss.core.v1; severe pain-out-of-proportion (Acanthamoeba); rapidly destructive contact-lens ulcer (Pseudomonas). These mandate scrape-before-empiric, hourly therapy, and disposition escalation.
   inputs: hypopyon_and_anterior_chamber_reaction, infiltrate_size_and_central_location, corneal_thinning_or_perforation_signs
   actions: calc.qsofa, workup.acute_vision_loss
   advance: perforation / endophthalmitis / vision-threat / Acanthamoeba red flags screened and routed by engine_id if positive
5. 5INITIAL_WORKUP
   CORNEAL SCRAPE + smear (Gram, Giemsa, KOH/calcofluor white) + culture (blood, chocolate, Sabouraud dextrose; non-nutrient agar overlaid with E. coli for Acanthamoeba) BEFORE starting broad empiric fortified topical antibiotics for any significant / central / large / atypical / non-resolving ulcer. Document baseline acuity each eye, infiltrate size with a measured diagram/photo, and AC reaction. Small peripheral non-vision-threatening ulcer may be treated empirically with monotherapy fluoroquinolone without scrape per AAO PPP — the size/location/atypia threshold is explicit.
   inputs: corneal_scrape_smear_and_culture, visual_acuity_each_eye
   actions: panel.cbc, panel.inflammation, panel.cmp
   advance: scrape/culture obtained for significant ulcer (or documented threshold for empiric monotherapy); baseline acuity + infiltrate map recorded
6. 6BRANCHING_WORKUP
   Pathogen decision tree: feathery margins + satellites + endothelial plaque + vegetative trauma → fungal (KOH/calcofluor, Sabouraud); pain-out-of-proportion + ring infiltrate + lens-in-water → Acanthamoeba (non-nutrient agar + in-vivo confocal microscopy / PCR — low threshold given diagnostic-delay-driven prognosis); branching terminal-bulb dendrite / prior HSV → herpetic (clinical ± PCR); dense suppurative mucopurulent contact-lens ulcer → Pseudomonas-skewed bacterial. Repeat scrape / biopsy / confocal if smear-culture negative but non-resolving.
   inputs: confocal_microscopy_or_pcr
   actions: workup.acute_vision_loss
   advance: pathogen class assigned (organism-directed or empiric-by-prior); Acanthamoeba/fungal actively excluded if any pointer
7. 7DIFFERENTIAL
   Terminal differential with pivot findings: microbial keratitis vs corneal abrasion (epithelial defect WITHOUT stromal infiltrate / no AC reaction pivot — route back to ophtho.corneal-abrasion.core.v1); vs sterile / marginal infiltrate (small, peripheral, sub-epithelial, intact epithelium, lid-margin staph hypersensitivity, minimal pain pivot); vs recurrent corneal erosion (recurrent on-waking pain, prior abrasion, no infiltrate pivot); Acanthamoeba vs HSV stromal (pain-out-of-proportion + ring + lens-in-water vs dendrite + prior HSV + reduced corneal sensation pivot); vs anterior uveitis (deep ciliary ache + cell/flare WITHOUT a corneal infiltrate or epithelial defect pivot — route ophtho.acute-red-eye.core.v1 / uveitis arm).
   advance: single best diagnosis selected; abrasion/sterile-infiltrate/uveitis look-alikes excluded or routed; co-infection (bacterial-on-fungal, polymicrobial) flagged
8. 8RISK_STRATIFICATION
   Severity → disposition: vision-threatening ulcer = central, ≥3 mm, deep stromal, hypopyon, perforation risk, monocular patient, immunocompromised, contact-lens Pseudomonas, fungal/Acanthamoeba, or non-resolving on empiric — each lowers the admission / same-hour-ophthalmology threshold. Baseline counting-fingers-or-worse acuity and completely central ulcer are the SCUT steroid-benefit subgroup markers (Srinivasan SCUT 2012 PMID 21987582).
   inputs: infiltrate_size_and_central_location, visual_acuity_each_eye
   actions: calc.news2
   advance: vision-threat severity tier + SCUT steroid-eligibility subgroup assigned
9. 9TREATMENT
   Pathogen-stratified intensive topical ladder (regimen_axes): (1) empiric broad fortified / fluoroquinolone topical HOURLY around-the-clock after scrape; (2) bacterial-targeted by smear/culture; (3) fungal → topical natamycin 5% (filamentous, esp. Fusarium — MUTT I) ± voriconazole, oral antifungal NOT routinely (MUTT II); (4) Acanthamoeba → biguanide (PHMB / chlorhexidine) + diamidine, prolonged; (5) herpetic → topical ganciclovir / oral aciclovir-valaciclovir, debridement for dendrite; (6) adjunct topical steroid ONLY after infection controlled (≥48 h appropriate antibiotic) AND HSV/fungal/Acanthamoeba excluded (SCUT — no overall benefit, subgroup only). NEVER patch; NO topical anaesthetic; cycloplegic for comfort; stop contact-lens wear. Therapeutic/tectonic keratoplasty for perforation (non_pharm).
   inputs: corneal_scrape_smear_and_culture, pregnancy, creatinine
   advance: pathogen-appropriate hourly topical therapy started; steroid gate enforced; no-patch/no-anaesthetic guardrails documented; surgery arranged if perforating
10. 10DISPOSITION
    Vision-threatening ulcer (central / ≥3 mm / hypopyon / fungal / Acanthamoeba / contact-lens Pseudomonas / monocular / non-resolving) → admit or same-hour ophthalmology with ability to deliver hourly topical therapy + daily review. Perforation / suspected endophthalmitis → emergent ophthalmology and route OUT to ophtho.acute-vision-loss.core.v1. Small peripheral non-vision-threatening ulcer on reliable empiric monotherapy → close outpatient ophthalmology review within 24-48 h. Benign look-alike confirmed (abrasion / sterile infiltrate) → route back to ophtho.corneal-abrasion.core.v1 / manage as sterile.
    inputs: visual_acuity_each_eye
    advance: disposition + route-out/route-back engine_id documented; hourly-therapy capability confirmed for admitted ulcers
11. 11MONITORING
    Daily slit-lamp review while severe: infiltrate size, epithelial-defect area, AC reaction, thinning — measured against the baseline map. Expect slow change; clinical worsening at 48-72 h on appropriate therapy mandates re-scrape / culture review / consider fungal-Acanthamoeba-polymicrobial / adjust regimen — NOT silent escalation of the same agent. Re-image (confocal) if non-resolving. Watch corneal thinning → perforation route-out.
    inputs: infiltrate_size_and_central_location, corneal_thinning_or_perforation_signs
    actions: panel.inflammation
    advance: objective improvement (shrinking infiltrate, re-epithelialisation, quieter AC) OR non-response re-evaluation triggered
12. 12FOLLOWUP
    Taper topical therapy with re-epithelialisation; manage residual scar / irregular astigmatism affecting acuity (corneal-scar density + irregular astigmatism are the dominant residual-VA drivers — Menda JAMA Ophthalmol 2020 PMID 31804657) with rigid-lens trial / referral for optical or therapeutic keratoplasty. HERPETIC: long-term oral aciclovir 400 mg BID ×12 mo prophylaxis halves recurrence, esp. stromal (HEDS NEJM 1998 PMID 9696640). CONTACT-LENS: hygiene re-education (no overnight wear, no water exposure, replace case, switch to daily disposables) — the dominant preventable driver (Stapleton Ophthalmology 2008 PMID 18538404; Carnt Ophthalmology 2022 PMID 35952937). Counsel contralateral-eye risk.
    inputs: prior_hsv_or_zoster_ocular_disease, contact_lens_use_and_hygiene
    advance: scar/astigmatism plan + HSV-prophylaxis decision + contact-lens prevention counselling documented