JMIR Publications

ABSTRACT

Background:

Fuzzy logic has increasingly been explored as a flexible alternative to traditional statistical approaches in healthcare modelling, particularly in contexts characterized by ambiguity, complexity, and incomplete information. While widely applied for prediction, classification, and risk stratification, its role in causal inference remains insufficiently examined and often methodologically fragmented.

Objective:

This systematic review aimed to study the practical use of fuzzy logic frameworks to solve causal questions in healthcare, emphasizing their methodological background, comparative performance, and integration with formal causal inference tools.

Methods:

A systematic search was conducted in six major databases (PubMed, Web of Science, ScienceDirect, SpringerLink, Scopus, IEEE Xplore) to identify peer-reviewed studies published between 2014 and 2025 that applied fuzzy modelling to healthcare settings with causal objectives. The review adhered to PRISMA 2020 guidelines and used a modified PICO framework to structure eligibility. Data was extracted on modelling strategies, comparator methods, healthcare domains, and alignment with causal frameworks. The risk of bias was independently assessed using adapted versions of the Joanna Briggs Institute (JBI) Checklist and the PROBAST-AI tool.

Results:

Thirty-seven studies met the inclusion criteria. The most common fuzzy methods were Fuzzy Inference Systems (FIS), Fuzzy Cognitive Maps (FCM), and Neuro-Fuzzy models. Applications spanned infectious diseases, oncology, cardiovascular health, mental health, and occupational health. Of the fourteen studies that included comparator methods, five demonstrated superior predictive performance of fuzzy models compared with conventional statistical or machine learning approaches, three reported broadly comparable results, and nine provided insufficient details for robust comparison. Only two studies (5%) explicitly applied formal causal frameworks, such as fuzzy-set Qualitative Comparative Analysis or DEMATEL+ANP, while six others adopted heuristic causal reasoning without counterfactual structures. Most studies showed moderate to high risk of bias, most frequently due to limited validation, unclear sampling strategies, or inadequate definition of outcomes.

Conclusions:

Fuzzy logic demonstrates potential for improving causal thinking in healthcare, particularly in scenarios involving high-dimensional data, non-linear interactions, and epistemic uncertainty. Its ability to incorporate expert knowledge and provide interpretable, rule-based outputs align well with the needs of clinical and policy decision-making. However, methodological transparency remains inconsistent, and integration with established causal inference frameworks is rare. Future research should prioritize the development of hybrid models that explicitly link fuzzy logic with counterfactual and graphical causal approaches, adopt standardized reporting practices, and evaluate these models through external validation in real-world healthcare environments. Such efforts are essential to move beyond fragmented applications and establish fuzzy logic as a credible paradigm for causal inference in complex health systems. Clinical Trial: This systematic review was prospectively registered in the PROSPERO database (CRD420251044493)