Institutes outside Greece
Beyond the average treatment effect: risk-based approaches to medical decision making
Abstract
dc:descriptionDoctors typically align clinical treatments with average effects observed in clinical trials, assuming all patients respond similarly. However, these average effects often conceal substantial variability (heterogeneity) in patient responses, limiting their applicability for individual patient decisions. Baseline risk, or a patient's likelihood of experiencing a particular outcome, critically influences treatment effectiveness. Patients with higher baseline risk generally gain more substantial benefits from treatment compared to low-risk patients. Thus, baseline risk assessment through predictive modeling is essential for personalized medical decision-making. Healthcare data collected routinely from hospitals, insurers, and primary care providers presents significant potential for developing risk-based predictive models. Nevertheless, observational data analyses face challenges, notably confounding, because treatment assignment in routine care is non-random. Furthermore, inconsistencies in data formats complicate the integration of multiple data sources. Standardized data frameworks like the OMOP Common Data Model (OMOP-CDM) facilitate large-scale observational studies, mitigating these challenges. This thesis systematically reviews methodologies addressing treatment effect heterogeneity, categorizing them into risk-based methods, treatment effect modeling, and optimal treatment regimes. A structured, risk-based analytical framework is developed and applied across clinical trial data and observational datasets, including comparisons between antihypertensive therapies and osteoporosis treatments. Extensive simulations indicate that regression models incorporating linear interactions between baseline risk and treatment yield reliable personalized predictions even with modest sample sizes. Practical predictive models for melanoma recurrence and mortality, as well as COVID-19 outcomes, are also presented. Overall, the thesis highlights the importance of incorporating baseline risk into clinical decisions, advocating for risk-based approaches to better tailor treatments to individual patient profiles.
Degree
thesis:*- Grantor dc:publisher
- Institutes outside Greece
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Rekkas, Alexandros
- Ρέκκας, Αλέξανδρος
Subjects
dc:subject × 25- Εξατομικευμένη ιατρική
- Προγνωστικά μοντέλα
- Δεδομένα πραγματικού κόσμου
- Στρωματοποίηση κινδύνου
- Υποστήριξη ιατρικής απόφασης
- Medical decision making
- Personalized medicine
- Electronic health record
- Comparative effectiveness
- Treatment effect modeling
- Prognostic model
- Risk stratification
- Model calibration
- Μαθηματικά
- Φυσικές Επιστήμες
- Στατιστική και Πιθανότητες
- Επιστήμες Υγείας
- Ιατρική και Επιστήμες Υγείας
- Ιατρική Πληροφορική
- Mathematics
- Natural Sciences
- Statistics and Probability
- Health Sciences
- Medical and Health Sciences
- Medical Informatics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59485