[refining a computational PNH screening algorithm]

Detecting patients potentially suffering from paroxysmal nocturnal hemoglobinuria via screening algorithm by retrospective electronic health record analysis

---

Objective

The primary objective of this study is to develop a screening algorithm to detect patients at high risk for PNH via patient electronic record analysis in order to reduce the delay in PNH diagnosis. This algorithm needs to score a probability of having a high risk for PNH or absence of having a high risk for PNH. Secondary objective of this study is to evaluate the current diagnostic pathway for employing flow cytometry with FLAER in cases suspected of PNH.

Methodology

A computational algorithm will be developed for retrospective analysis of electronic health records (EHR) of patients aged 14 years and older, with at least one abnormal hemoglobin lab medical record coded in the AZ Delta patient database. 

The algorithm will also screen for risk factors reflected in measurements, keywords and codes  e.g.: (hemolytic) anemia, dysphagia, dyspneaoe, fatigue, headache, weakness, erectile dysfunction, hemoglobinuria, hematuria, cytopenia, chronic kidney disease, abdominal pain, bone marrow dysfunction and/or thrombosis.

A system will be set up to rank the at-risk patients and visualize their patient characteristics along with their risk factors and results from flow cytometry

Generative and language models can optimize the identification of PNH risk factors in medical texts from patient files

Impact and future directions

Despite a simple diagnostic test, many PNH (Paroxysmal Nocturnal Hemoglobinuria) cases in Belgium remain undetected due to the disease’s rare and complex presentation in various medical specialties. We are developing an algorithm to identify potential PNH patients retrospectively, with plans to expand and implement this system in other hospitals for early diagnosis and treatment, aiming to decrease complications and improve patient outcomes. Early treatment could reduce complications, morbidity and mortality.

General info and contact

Keywords (#): PNH, rare diseas, algorithm, risk scoring, OMOP

RADar project research lead: Ir. Kim Denturck

RADar project researchers: Ir. Kim Denturck, Laura Burgelman, Prof. Dr. Ir. Peter De Jaeger

Principal investigator: MD Dries Deeren

Timeline: December 2023 – December 2025

Status: Ongoing

Publications / presentations:
Submission of publication in progress, Poster- in category Non malignant hematology- on 38th GAM of the BHS conference 03-04/02 2023, Brussels, link.

Funding: Grants via Sobi, Amgen, Novartis and Roche, contract with Alexion pending