Development and Implementation of a Dynamic Framework for Assessment and Improvement of Registries’ Data Quality
Author(s)
Mai Duong, PhD1, Nahila Justo, PhD, MBA, MPhil2, Noami Berfeld, MSc1, Charlotte Pettersson, MSc3, Hai Nguyen, PhD1, Kristina Kastreva, PhD4, Vitaliy Matyushenko, MSc5, Lenka Mokrá, MSc6, Ali Farjana, BSc7, Neil Bennett, BSc8, Emma Watson, BSc7, Dino Masic, MSc7, Annie Poll, PhD7.
1Thermo Fisher Scientific, Evidera RWDSS, London, United Kingdom, 2Thermo Fisher Scientific (Evidera RWDSS) and Karolinska Institute, Stockholm, Sweden, 3Thermo Fisher Scientific, Evidera RWDSS, Stockholm, Sweden, 4University Hospital "Alexandrovska", Medical University Sofia, Sofia, Bulgaria, 5Ukranian SMA Registry, Kharkiv, Ukraine, 6Institute of Biostatistics and Analyses, Brno, Czech Republic, 7TREAT-NMD Services Ltd, Newcastle Upon Tyne, United Kingdom, 8TREAT-NMD Services Ltd, Newcastle upon Tyne, United Kingdom.
1Thermo Fisher Scientific, Evidera RWDSS, London, United Kingdom, 2Thermo Fisher Scientific (Evidera RWDSS) and Karolinska Institute, Stockholm, Sweden, 3Thermo Fisher Scientific, Evidera RWDSS, Stockholm, Sweden, 4University Hospital "Alexandrovska", Medical University Sofia, Sofia, Bulgaria, 5Ukranian SMA Registry, Kharkiv, Ukraine, 6Institute of Biostatistics and Analyses, Brno, Czech Republic, 7TREAT-NMD Services Ltd, Newcastle Upon Tyne, United Kingdom, 8TREAT-NMD Services Ltd, Newcastle upon Tyne, United Kingdom.
OBJECTIVES: To develop a dynamic framework for assessing and improving data reliability, extensiveness, and coherence across six TREAT-NMD registries to support a study in spinal muscular atrophy.
METHODS: A Data Quality Framework and Improvement Process (DQFIP) was developed based on EMA guidance and FAIR (Findable, Accessible, Interoperable, and Reusable) principles and implemented across registries taking part in the ongoing study. It comprises Metadata (variable definitions, coding, formats, quality metrics, KPIs, and validation/verification) and Process (automated assessment scripts, manual review, query logs and resolutions). The Improvement Process involves root cause investigation, data quality assurance, and improvement planning: and it is implemented iteratively in a continuous feedback loop.
RESULTS: Data quality was assessed for approximately 400 patients and 150 variables by TREAT-NMD, in collaboration with ThermoFisher Scientific (Evidera). In the first extraction, high missingness was found in some variables (e.g., comorbidities, treatment stop date). Most queries pertaining to data reliability were due to outdated 'flag-type' variables (e.g., ‘currently/ previous’ status of ventilation not being updated despite updates to respective date variables). Coherence-related queries included date format and dose unit inconsistencies. Other general queries included absence of registry enrolment and loss to follow-up dates. After receiving queries from the first extraction, registry curators made necessary corrections and completed missing data, which resulted in better quality in the second extraction after one year. Query frequency dropped to under 10% for most variables and under 5% for critical ones (treatments and outcomes). For continuous quality improvement, meetings with TREAT-NMD and registry curators were organized to investigate root causes and discuss remedial measures, including adjustments in data capture, edit checks, and processes.
CONCLUSIONS: Clinical registries' protocol-driven prospective data collection allows for adaptability and quality improvement throughout the data lifecycle. The DQFIP continues to improve the quality of participating registries, supporting research sustainability and benefiting the patient community.
METHODS: A Data Quality Framework and Improvement Process (DQFIP) was developed based on EMA guidance and FAIR (Findable, Accessible, Interoperable, and Reusable) principles and implemented across registries taking part in the ongoing study. It comprises Metadata (variable definitions, coding, formats, quality metrics, KPIs, and validation/verification) and Process (automated assessment scripts, manual review, query logs and resolutions). The Improvement Process involves root cause investigation, data quality assurance, and improvement planning: and it is implemented iteratively in a continuous feedback loop.
RESULTS: Data quality was assessed for approximately 400 patients and 150 variables by TREAT-NMD, in collaboration with ThermoFisher Scientific (Evidera). In the first extraction, high missingness was found in some variables (e.g., comorbidities, treatment stop date). Most queries pertaining to data reliability were due to outdated 'flag-type' variables (e.g., ‘currently/ previous’ status of ventilation not being updated despite updates to respective date variables). Coherence-related queries included date format and dose unit inconsistencies. Other general queries included absence of registry enrolment and loss to follow-up dates. After receiving queries from the first extraction, registry curators made necessary corrections and completed missing data, which resulted in better quality in the second extraction after one year. Query frequency dropped to under 10% for most variables and under 5% for critical ones (treatments and outcomes). For continuous quality improvement, meetings with TREAT-NMD and registry curators were organized to investigate root causes and discuss remedial measures, including adjustments in data capture, edit checks, and processes.
CONCLUSIONS: Clinical registries' protocol-driven prospective data collection allows for adaptability and quality improvement throughout the data lifecycle. The DQFIP continues to improve the quality of participating registries, supporting research sustainability and benefiting the patient community.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
RWD60
Topic
Real World Data & Information Systems
Topic Subcategory
Data Protection, Integrity, & Quality Assurance, Reproducibility & Replicability
Disease
Rare & Orphan Diseases