220 likes | 228 Views
This paper explores the implementation of a clinical registry system, PheDRS, which aims to detect and manage chronic obstructive pulmonary disease (COPD) and other targeted patient populations. The study assesses the usefulness of existing guidelines and provides lessons learned from the implementation process.
E N D
Phenotype Detection Registry System (PheDRS) Implementation of aGeneralizable Single Institution Clinical Registry Architecture John D. Osborne, PhD, Adarsh Khare, MS, Donald M. Dempsey MS, J. Michael Wells, MD, Matt Wyatt, MSHI, Geoff Gordon, MSHI, Wayne H. Liang, MD MS, James Cimino, MD
MOTIVATION: Copd • Chronic Obstructive Pulmonary Disease (COPD) • 3rd most common cause of death in US • COPD exacerbations can cause further damage • Need to find COPD inpatients and intervene to prevent readmission
Population Health Management Patient Management Population Management Patient registry Focused around a particular condition or target population Used outside the clinical setting for epidemiological studies, outcomes research, and cohort discovery A subset of patient registries has been termed clinical registries • EMR • Orchestrate information at the patient level • Used by physicians, nurses and other healthcare providers
Clinical Registry definition • Patient registries with a clinical focus have been termed “clinical registries” or “clinical quality registries” by recent systematic reviews • Clinical registry characteristics per Hoque et al (2017) • clinical intervention component • feedback mechanism to providers
Clinical registry publications • Hoque et al (2017) found only 17 such registries from January 1980 to December 2016 ?! • Single institution was an exclusion criteria • Delayed feedback was not • In reality “feedback of data to participating clinical centers often lags well behind actual care, making data obsolete and less useful (Nelson et al, 2016)
Clinical registry resources • Existing Software Platforms • i2b2 • Rare Disease Registry • ImproveCareNow • Documentation and Architecture Guidelines • Lindoerfer and Mansmann published a checklist of design principles in 2014, updated in 2017 • Their updated checklist included 12 topics with 72 descriptive items
Approach Questions Method Try to answer from the perspective of a clinical registry implementation at UAB - PheDRS • How useful are these guidelines for implementing a clinical registry at a single institution? • Do we need them at all? • Are there additional considerations for single institution clinical registries to consider?
WHO needs guidelines - just build it! • Started in 2013 with a Cancer Registry Control Panel (CRCP) • Assist cancer registrars with case identification • Doesn’t replace registrars – just presents them with fewer cases to look at
Unfortunately only works for reportable cancer and is not generalizable to other diseases • Registry guidelines would have helped!
Phenotype Detection Registry System • What is PheDRS? • N-tier web application (Javascript, Java, Oracle SQL) • Generalize CRCP to handle other patient populations • Chronic Obstructive Pulmonary Disease (COPD) • Specific cancer (multiple myeloma) • Clinical trial candidates (precision oncology) • Allow new registries to be spun up with no (or minimal) coding • Clinical focus to allow intervention
Checklist: Applicable topics Overall Applicable Topic Guidelines Development “framework for adding a new registry” especially useful Interfaces and Interoperability Mobile devices Data Analysis General Features Training Security Encrypted data and storage • Overall, we found the CIPROS guidelines helpful for clinical registry design • Some guidelines are just good general software design principles • Half (6/12) of the articulated principles are all or almost all applicable for clinical registries
Checklist: less applicable topics • Software Architecture • N-Tier, Platform Independence • Standardization • Ontology based vocabularies and metadata, XSD for structured data exchange • Internationality (multilingual support) • Data Management & Quality • No pre-defined selection for data encoding • Organizational (informed consent) • Privacy (pseudonymization)
Lessons Learned • The ability to easily create a new registry is of huge importance • Difficult to sustain otherwise • Plan for both clinical and non-clinical registries • ETL performance matters • Co-ordination with other users • Record query execution times • Dedicated database server • Map data to standards on demand • If the data never leaves the institution – you may never need to map it
Acknowledgements • HeenaChitkara, AbakashSamal, Dale Johnson • Jessica Nichols, deNay Porter • Shri Merryweather • Yolanda Graham, Chammie Richie • Elizabeth Brown • Eddy Yang