Overview of Recent Applications of Artificial Intelligence for Real World Evidence Development
Author(s)
Blein C1, Ludwikowska M2, Tournier C3, Leboucher C3, Toumi M4, Bakhutashvili A5, Aballea S6
1Creativ-Ceutical, lyon, 69, France, 2Creativ-Ceutical, Cracow, Poland, 3Creativ-Ceutical, Lyon, France, 4Creativ-Ceutical, Paris, France, 5MARCO POLO, Luxembourg, Luxembourg, 6Creativ-Ceutical, Paris, 75, France
Presentation Documents
OBJECTIVES
: To provide an overview about trends in current research and status of development and use of AI methods in RWE studies conducted using electronic health records (EHR) or claims databases.METHODS
: A literature review of recent (2020-2021) applications of AI methods in RWE studies using electronic medical records or claims database was implemented in EMBASE and MEDLINE via Ovid. Different types of applications of AI were categorized based on titles and abstracts and selected examples were reviewed within each category.RESULTS
: Of the 1,122 records screened, 408 articles were selected into 3 categories. Category 1 (n=74/18%): AI is used in study methods to identify population, prevalence of a condition, patients’ characteristics – risk factors, treatment patterns, health outcomes and adverse reactions. Category 2 (n=314/77%): Research that aim to develop and validate AI tools to assess prognosis, predict survival, identify drug interactions, or identify health outcomes Category 3 (n=20/5%): Research that aim to develop and validate AI tools that address challenges of RWE studies, such as imputing missing data, measuring patients’ similarity for heterogeneous EMR data, predicting clinical variables not completely reported in EHR and leveraging temporal sequential patterns from EHRs The main 4 types of AI methods used were machine learning, deep learning, NLP and process mining. Key challenges encountered are related to data source, data bias, data aggregation, lack of receptivity and black box effect. Lack of receptivity by external stakeholders is a major barrier. The black box problem is serious enough that specific regulations are beginning to emerge. The EU's GDPR now includes a right to explanation to deal with algorithmic opacity.CONCLUSIONS
: The pace of adoption of AI is accelerating. AI methods provide potential for significant advances in the field of RWE, but key challenges including lack of receptivity and black box effect remain to be addressed.Conference/Value in Health Info
2022-05, ISPOR 2022, Washington, DC, USA
Value in Health, Volume 25, Issue 6, S1 (June 2022)
Code
RWD65
Topic
Methodological & Statistical Research, Real World Data & Information Systems, Study Approaches
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics, Electronic Medical & Health Records, Health & Insurance Records Systems, Literature Review & Synthesis
Disease
No Additional Disease & Conditions/Specialized Treatment Areas