The document describes a project to develop a guideline-based clinical decision support system (CDSS) for COVID-19 using openEHR archetypes and the Guideline Definition Language (GDL). A team from Zhejiang University extracted diagnostic and treatment guidelines from Chinese sources and represented them in openEHR templates and GDL rules. Two CDS applications were implemented and are being deployed in hospitals to aid diagnosis and screening of COVID-19 patients. The project aims to demonstrate how open standards like openEHR can facilitate rapid development and sharing of computerized clinical guidelines.

1. Dr. Shan Nan
(nanshan@zju.edu.cn)
On behalf of all the team members
Zhejiang University
2020.07.24
Guideline-based CDSS for COVID-19

2. Me and my group
Shan Nan Ph.D
• PhD of Biomedical Engineering
• Post Doc in Zhejiang University.
• Experienced on CIG and rule-based CDSS
Medical Informatics Group, the earliest medical
informatics research group in China

3. The openEHR team
Prof. Huilong Duan
OpenEHR Ambassador
Prof. Xudong Lu
OpenEHR Ambassador
Tianhua Tang
PhD Student
Mengyang Li
PhD Student
Hongshuo Feng
MSc Student
OpenEHR team with Dr. Rong Chen,
the inventor of GDL
Yijie Wang
Informaticist
Dr. Heather Leslie
External consultant
Kuai Yu
Software Engineer

4. Motivations

5. COVID-19 global crisis
Informaticists are obligatory to take actions!

6. Efforts in informatics communities
Evidence-based CDSS and knowledge base have not been reported

7. Timeline of the Chinese guideline
7th Edition
Mar. 14
6th Edition
Feb. 19
5th Edition
Feb. 3
4th Edition
Jan. 27
3rd Edition
Jan. 23
2nd Edition
Jan. 22
1st Edition
Jan. 16
First Case
Dec. 26
Under Control
Mar. 23

8. Sharing the COVID-19 knowledge
• Different Language
• Narrative guidelines needs
proper interpretation
• Difficult to integrate with
heterogeneous
information systems
https://gmcc.alibabadoctor.com/prevention-manual

9. OpenEHR approaches have the potential
1500 people
87 countries

10. Info model and a formal guideline language
OpenEHR archetypes make the
information model easy to
understand, reusable and sharable
The Guideline Definition Language
provides a formal way to
computerize narrative guidelines

11. Our mission
• Specify a guideline acquisition approach
using openEHR
• Develop and disseminate a computer
interpretable guideline for COVID-19
• Implement CDS applications for COVID-19

12. COVID-19 CIG R&D

13. A rapid guideline acquisition approach
We used openEHR archetypes and Guideline Definition Language to capture the knowledge
in guidelines and rapidly developed CDSS applications.

14. A collaboration platform to manage the KA

15. Extract IF-THEN relations
We extracted the Diagnostic Criteria, Clinical Classification, Clinical Warning Sign,
Differential Diagnosis, Treatment, and Discharge sections in the guideline.

16. Extract entities
Entities in each sentence were marked up and extracted. Then they are mapped to
openEHR concepts.

17. Elicit COVID-19 related archetypes
Li M, Leslie H, Qi B, Nan S, Feng H, Cai H, Lu X, Duan H, Development of an openEHR Template for COVID-
19 Based on Clinical Guidelines, JMIR, Vol22, No 6 (2020): June. https://www.jmir.org/2020/6/e20239/
Collect data elements from guideline
https://ckm.openehr.org/ckm/templates/1013.26.291

18. Author rules in GDL2 Editor
RESOURCE_DESCRIPTION
original_author:Hash<string,,string>[1]
other_contributors:string[0..*]
lifecycle_state:terminology_code[1]
details:DETAIL
other_details:Hash<string,string>[0..1]
DETAIL
language: terminology_code[1]
purpose:string[1]
keywords:string[0..*]
use:string[0..1]
misuse:string[0..1]
original_resource_url:Hash<string,,string>[0..1]
other_details:Hash<string,,string>[0..1]
copyright:string[0..1]
original_language:terminology_code[1]
descrption:RESOURCE_DESCRIPTION
data_bindings:DATA_BINDING
pre_conditions:ASSERTION
rules:RULE
templates:OUTPUT_TEMPLATE
default_actions:ASSIGNMENT[*]
model_id:string[1]
template_id:string[0..1]
predicates:ASSERTION[*]
type:string[1]
elements:ELEMENT
DATA_BINDING
id:string[1]
when:ASSERTION[*]
then:ASSIGNMENT[*]
priority:integer[1]
RULE id:string[1]
name:string[1]
model_id:string[1]
template_id:string[0..1]
object:OBJECT
OUTPUT_TEMPLATE
id:string[1]
path:string[1]
ELEMENT
term_definitions:Hash<string,hash<string,TERM>>[1]
term_bindings:Hash<string,TREM_BINDING>[0..1]
TERMINOLOGY
code:string[1]
text:string[1]
description:string[1]
other_items:Hash<string,string>[0..1]
TERM
code:string[1]
TERM_BINDING
gdl_version:string[0..1]
id:string[1]
concept:string[1]
definition:GUIDELINE_DEFINIITON
terminology:TERMINOLOGY
GUIDELINE_DEFINIITON
GUIDELINE
AUTHORED_RESOURCE
TOP
LEVEL

19. Create a guideline

20. Import archetypes
27 archetypes were imported here.

21. Data binding

22. Edit a list of rules
E.g. if CT image report contains ground glass opacities, patchy shadows, interstitial
changes, consolidation, or infiltration, the patient has a COVID-19 risk factor.

23. Logical structure of the GDL rule set
COVID-19
Diagnosis &
Treatment
Discharge
Yes
Yes
Critical case classification
Clinical_Warning.v0
Classfication.v0
Suspect cases diagnosis
White_cell_count.v0
Lymphocyte_count.v0
Suspected_Diagnosis.v0
Yes
No
Exclude
No
No
General treatment
General_Treatment.v0
Diagnosis
IgG_Antibody_Detection.v0
IgM_Antibody_Detection.v0
Confirmed_Diagnosis.v0
Discharge criteria
Out_Hosipital.v0
Body_Temperature_Monitor.v0
Nucleic_acid_test_result.v0
Supportive treatment
Respiratory_support_Treatment.v0
Circulation_support_Treatment.v0
Convalescent_plasma_Treatment.v0
Blood_Purification_Treatment.v0
Continuous_Renal_Replacement_Therapy.v0
Other_Treatment.v0
No
Yes
Totally 18 GDL rules have been authored.

24. Rule validation
The rules were validated with patient data adopted from a published patient case report[1].
[1] First case of 2019 novel coronavirus in the United States. N Engl J Med 2020;382(10):929–936

25. Disseminate and share the rules
Templates and rules were shared in GitHub and announced in openEHR disclosure
https://discourse.openehr.org/t/chinese-covid-19-diagnosis-and-
treatment-decision-support-openehr-templates-and-rules/516
https://github.com/ZJU-BME-VICO/openEHR-COVID-19
S Nan, T Tang, H Feng, Y Wang, M Li, X Lu, H Duan, Rapid Development and Dissemination of a Computer
Interpretable Guideline for COVID-19, JMIR, under revision

26. Implementing a CDSS

27. Clinical requirements
Although COVID-19 has been largely under control in China since March 2020, small scale
breakouts have still been reported in Jilin, Guangdong, Beijing and most recently, Xinjiang.
COVID-19 screening in the hospital is still required for a certain period of time.
Scenario 2: Screening susceptive cases in
management dept. in hospital
Scenario 1: COVID-19 diagnosis at
fever/Respiratory/Infection clinics
GDL rules

28. Implementation framework
We use a CDS platform (named Tracebook) to build CDS applications by
configuring rules and UI components rather than coding from scratch.
GDL can be applied here
UI can be designed here

29. Mapping between GDL and Drools
A GDL Rule
A Drools Rule
Archetypes => Object Data Model GDL Expression => MVEL
Unfortunately, there is currently a lack of open source/free GDL execution engine…

30. Data model mapping
Category openEHR Archetype Object Data Model
Demographic openEHR-EHR-OBSERVATION.age.v0 PatientInfo
Epidemic History openEHR-EHR-OBSERVATION.exposure_assessment.v0 EpidemicHistory
Medical Record openEHR-EHR-OBSERVATION.pf_ratio.v0
MedicalRecord
openEHR-EHR-OBSERVATION.story.v1
Exam
openEHR-EHR-CLUSTER.imaging_finding.v0
ImgExamResultopenEHR-EHR-CLUSTER.imaging_result-COVID_19.v0
openEHR-EHR-OBSERVATION.imaging_exam_result.v0
Vital Sign
openEHR-EHR-CLUSTER.inspired_oxygen.v1
PhysicalSign
openEHR-EHR-CLUSTER.level_of_exertion.v0
openEHR-EHR-CLUSTER.problem_qualifier.v1
openEHR-EHR-OBSERVATION.body_temperature.v2
openEHR-EHR-OBSERVATION.pulse_oximetry.v1
openEHR-EHR-OBSERVATION.respiration.v2
Laboratory Test
openEHR-EHR-CLUSTER.specimen.v0
LabTestResultopenEHR-EHR-CLUSTER.laboratory_test_analyte.v1
openEHR-EHR-OBSERVATION.laboratory_test_result.v1
Symptom
openEHR-EHR-CLUSTER.symptom_sign.v1
Symptom
openEHR-EHR-COMPOSITION.encounter.v1
openEHR-EHR-OBSERVATION.symptom_sign_screening.v0
openEHR-EHR-OBSERVATION.condition_screening.v0
Diagnosis
openEHR-EHR-EVALUATION.differential_diagnoses.v0
DiagnosisopenEHR-EHR-EVALUATION.health_risk.v1
openEHR-EHR-EVALUATION.problem_diagnosis.v1
Order
openEHR-EHR-EVALUATION.recommendation.v1
Order
openEHR-EHR-INSTRUCTION.medication_order.v2
openEHR-EHR-INSTRUCTION.therapeutic_order.v0
openEHR-EHR-OBSERVATION.management_screening.v0

31. Rule mapping
Guideline Section GDL Rule Drools Rule
(5) Diagnostic Criteria
COVID_Confirmed_Diagnosis.v0.gdl2
Diagnosis_Confirmed
COVID_Lymphocyte_count.v0.gdl2
COVID_Nucleic_acid_test_result.v0.gdl2
COVID_White_blood_cell_count.v0.gdl2
COVID_White_cell_count.v0.gdl2
(6) Clinical Classification COVID_Classfication.v0.gdl2 Classification
(7) Clinical Warning Sign COVID_Clinical_Warning.v0.gdl2 Clinical_Warning
(8) Differential Diagnosis COVID_Suspected_Diagnosis.v0.gdl2 Diagnosis_Suspected
(10) Treatment
COVID_Blood_Purification_Treatment.v0.gdl2
Treatment_Modern
COVID_Circulation_support_Treatment.v0.gdl2
COVID_Continuous_Renal_Replacement_Therapy.v0.gdl2
COVID_Convalescent_plasma_Treatment.v0.gdl2
COVID_General_Treatment.v0.gdl2
COVID_Immunotherapy.v0.gdl2
COVID_Other_Treatment.v0.gdl2
COVID_Respiratory_support_Treatment.v0.gdl2
(11) Discharge
COVID_Body_Temperature_Monitor.v0.gdl2
Discharge
COVID_Out_Hosipital.v0.gdl2
*For section 5, 10, and 11, there are multiple GDL rules for one section. This is because
GDL2 Editor now only allows one rule in a file, whereas these sections contain several rules.

32. Deployment and trial plan
Diagnosis and treatment CDSS for clinics CDSS for screening susceptive cases
• These applications are currently being deployed in PLA Central Theater General Hospital.
• An observational before-after study is prepared to validate the effectiveness of the CDS.
• It is expected that the miss diagnose rate of susceptive cases and confirmed cases can be
decreased.

33. Discussions

34. What do we benefit from openEHR
• Using archetypes saved our time and lifted our burden on
understanding data items while developing the CIG.
• Most archetypes in CKM can be used directly.
• Totally 27 archetypes have been used for the COVID-19 CIG, among which
26 were directly acquired from CKM and 1 (openEHR-EHR-
CLUSTER.imaging_result-COVID_19.v0) was acquired from CKM and
modified for the COVID-19 CDS purpose.
• GDL and its editor helped us to develop and test CIGs without
coding.

35. What are yet not perfect in openEHR
• Lack of an open source or free GDL execution engine.
• GDL2 Editor currently has 2 limitations.
• Data types such as DV_PROPORTION and DV_DURATION are not fully
supported in the editor.
• Difficult to support collection data type (e.g. array, set).
Not yet supported
while executing

36. Proposal to develop an open source tool
Using the experience gained from COVID-19 CIG and CDSS development, we are currently
working on developing an openEHR EL based rule editor and execution engine. We plan to
make it open source in the future.

37. ➢ Evidence-based CDSS and knowledge are required while
fighting with COVID-19.
➢ We proposed an approach to use openEHR archetypes
and GDL, and developed a CIG.
➢ Two CDS applications are under deployment in real
clinical settings.
Wrap up

38. Thanks!