An evidence-fused neutrosophic framework for uncertainty-aware treatment selection

Abstract

This paper presents an integrated decision-support framework for selection of healthcare treatment based on the use of Neutrosophic Logic, Dempster-Shafer Theory (DST), and Interval-Valued Fuzzy Sets (IVFS) in an extended Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) structure. Healthcare decision-making often involves incomplete clinical data, conflicting expert opinions, and context-dependent variability, which are not adequately addressed by conventional MCDM approaches. Existing methods exhibit key limitations: classical models assume precise inputs, fuzzy models capture vagueness but not indeterminacy, and existing neutrosophic approaches lack a mechanism for resolving inter-expert conflict prior to ranking. The proposed framework addresses these gaps through a sequential uncertainty-handling process in which neutrosophic logic models truth, indeterminacy, and falsity, IVFS captures variability via interval-valued representations, and DST performs evidence-theoretic fusion to reconcile conflicting expert inputs before ranking. To overcome these problems, the proposed framework is used to transform neutrosophic evaluations into fuzzy representations by using interval-valued representations, which liberalizes the treatment of uncertainty. The DST systematically integrates expert judgment to support structured evidence fusion while avoiding premature consensus. The extended TOPSIS method is subsequently applied to generate treatment rankings using belief-weighted neutrosophic scores. The framework is evaluated using sensitivity analysis and Monte Carlo simulation, where variations in criterion weights and stochastic perturbations representing expert variability are introduced to assess ranking stability under uncertainty. The framework is applied to an illustrative numerical evaluation in a healthcare setting, where treatment options are assessed in terms of efficacy, adverse effects, cost, recovery period, and patient satisfaction. Sensitivity analysis and Monte Carlo simulations are employed to validate the approach, demonstrating its robustness and stability under varying weighting schemes and expert opinion perturbations. Results from a synthetic illustrative scenario indicate stable and interpretable rankings under weight perturbations and stochastic noise, suggesting robustness while requiring further validation with real clinical data. The proposed approach provides an uncertainty-aware decision-support tool for clinicians, administrators, and policymakers, offering interpretable treatment prioritization while remaining scalable for complex healthcare environments.