Aristotle University Of Thessaloniki (AUTH)
Reporting completeness and quality of systematic reviews of prognostic prediction models for covid-19: a living mapping and qualitative assessment of the research landscape
Abstract
dc:descriptionA prognostic prediction model (PPM) combines various prognostic factors to estimate a patient's risk of specific outcomes, supporting medical decision-making, particularly when healthcare resources are limited. Systematic reviews (SRs) of PPMs serve as a key research design for critically appraising, validating, and evaluating these models, thereby drawing conclusions about their quality and usefulness. The growing volume of SRs evaluating PPMs has led to the emergence of overviews of reviews (OoRs), which collect and synthesize evidence from multiple SRs to provide actionable insights for policymakers and clinicians. In this dissertation, we first conducted a methodological overview of systematic reviews of prognostic prediction models for COVID-19, published between 2020 and 2022, to assess their reporting completeness and transparency. Ten SRs were identified, including between 6 and 606 models developed using conventional statistical or machine learning (ML) methods. Most lacked a pre-defined protocol or adequate details on study selection, data collection, and reporting of included primary studies and models, with only one providing publicly available data. Additionally, the overall risk of bias was judged as high for the majority of SRs using the ROBIS tool. Furthermore, this thesis critically assessed the reporting completeness of COVID-19 prognostic models developed using either conventional statistical or ML methods, according to the TRIPOD and TRIPOD+AI guidelines. A total of 53 studies describing 71 prognostic models were identified; adherence to both guidelines was low overall, with significantly poorer compliance among ML-based studies compared to conventional model studies. More specifically, the reporting weaknesses included incomplete predictor and outcome definitions, lack of clarity around missing data handling, inadequate performance evaluation and validation procedures, and missing sample size justification in all studies. Finally, an evaluation of the reporting completeness of OoRs of prognostic prediction models using the PRIOR checklist was conducted to identify items for improvement and potential adaptation of the tool for the specific context of prognostic models. Seven OoRs were retrieved, showing moderate overall PRIOR adherence. Gaps occurred in reporting methodological components as none of the included overviews fully described the data collection process, management of discrepant data or synthesis methods used. In addition, only one addressed primary study overlap across the included SRs, and none provided sufficient detail on the assessment of reporting bias or certainty of evidence. Guidance for conducting SRs of PPMs is urgently needed, alongside greater awareness and education in minimum reporting standards and quality criteria. Moreover, tailored reporting guidelines are essential to address specific methodological challenges of PPMs overviews, thereby enhancing transparency, reproducibility, and utility in this growing field of evidence synthesis.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Talimtzi, Persefoni
- Ταλιμτζή, Περσεφόνη
Subjects
dc:subject × 13- Προγνωστικά μοντέλα
- Ανασκόπηση συστηματικών ανασκοπήσεων
- Πληρότητα καταγραφής
- Prognostic prediction models
- COVID-19
- Overview of systematic reviews
- Reporting completeness
- Άλλες Ιατρικές Επιστήμες
- Ιατρική και Επιστήμες Υγείας
- Ιατρική, άλλοι τομείς
- Other Medical Sciences
- Medical and Health Sciences
- Medicine, miscellaneous
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/60901