Implementing WHO Cervical Cancer Screening Algorithm in FHIR + CQL for OpenMRS

Development
1

Hi everyone,

I’m Dan Heslinga, a primary care physician and founder of Hopena Health. I’ve been building a FHIR R4 clinical decision support implementation of the WHO cervical cancer screening and treatment guidelines, targeting OpenMRS deployments in low- and middle-income countries.

What it does

The project implements WHO Algorithm 5 (HPV DNA primary screening + VIA triage) as computable CDS covering the full screening cascade:

  • Screening: Eligibility checks (age, sex, HIV status), scheduling, due-for-screening alerts
  • Triage: VIA assessment routing for HPV-positive women
  • Treatment: Ablation eligibility checklist, modality selection (thermal ablation vs excision)
  • Follow-up: Post-treatment and post-triage-negative retesting, including WLHIV double follow-up per WHO Rec 33

Both general population (age 30-49, 5-10 year interval) and WLHIV (age 25-49, 3-5 year interval) pathways are supported.

Technical approach

  • 5 CQL libraries (~2,000 lines of clinical logic) compiled to ELM JSON
  • 4 PlanDefinition ECA rules with patient-view triggers
  • CIEL terminology — all clinical concepts use OpenMRS concept dictionary codes mapped to SNOMED/LOINC/ICD-10
  • Client-side CQL evaluation using cql-execution, following the architecture described in Bacher et al., “FHIRing up OpenMRS” (AMIA 2024)
  • 14 synthetic test patients with 82 test assertions passing
  • Licensed under Apache 2.0

The CQL evaluates against FHIR resources retrieved from the OpenMRS FHIR2 Module. I worked around the current limitation where OpenMRS stores treatment procedures as Observations rather than Procedure resources.

Published IG and source code

The IG follows WHO SMART Guidelines L3 conventions (one CQL library per decision table, PlanDefinition ECA rules).

Looking for feedback

I’d love to hear from:

  1. OpenMRS implementers working on cervical cancer screening programs — does this match your clinical workflow? What’s missing?
  2. FHIR/CQL developers — feedback on the CQL patterns, PlanDefinition structure, or terminology bindings
  3. Anyone interested in piloting this at an OpenMRS site

This is a v0.1.0 release. Known gaps include: colposcopy pathway (not in Algorithm 5), pregnancy deferral, hysterectomy exclusion, and age-based exit from screening.

Thanks for any feedback!

Dan