Intelligent triage method for early diagnosis autism spectrum disorder (ASD) based on integrated fuzzy multi-criteria decision-making methods

Abstract

BackgroundAutism spectrum disorder (ASD) symptoms and severity levels vary from patient to patient, so treatment and healthcare will vary. However, little attention has been given to developing an autistic triage method for ASD patients concerning four issues: hybrid triage criteria, multi-selection criteria problems, criteria importance, and trade-off based on the inverse relationship between autistic criteria. Therefore, this study aims to develop a new method for triaging ASD patients and classifying them according to their severity of disorder using Fuzzy Multi-Criteria Decision Making (fMCDM) methods. MethodsTwo methodology phases have been conducted: the first phase is to identify and preprocess the ASD dataset, including 988 autistic patients with 42 medical and Sociodemographic criteria. In the second phase, two fMCDM methods were used to develop the triage method. The fuzzy Delphi Method (FDM) is used to select the most influential criteria among the 42 based on thirteen psychologists in the psychological field. Then Fuzzy-Weighted Zero-Inconsistency (FWZIC) is used to assign weights to the important criteria according to four psychologists' opinions. Accordingly, the Processes for Triaging Autism Patients (PTAP) method has been developed for the first time for triaging and classifying patients into three severity levels: minor, moderate, and urgent. ResultsFor the preprocessed phase, 538 out of 988 patients were obtained as a new ASD dataset underwent data cleaning to capture only autism patients. For the second phase, the FDM results have selected 19 out of 42 criteria and can control the bias of psychologists' opinions, FWZIC has assigned the appropriate weights for the 19 criteria, and the PTAP method triages the 538 patients into three severity levels: 36 minor injuries, 432 moderate injuries, and 70 urgent injuries. More complex statistical analyses have been presented using MedCalc statistical software. Three physicians in the psychological field gave their subjective judgements for the diagnosis of 46 random samples of patients. The sensitivity results were 86.67%, 80%, and 90.91%, while the specificity results were 93.55%, 88.46%, and 94.29% for urgent, moderate, and minor levels, respectively. In addition, the accuracy was 91.30% for urgent, 84.78% for moderate, and 93.48% for minor. This assessment led to a deduction that a proposed ASD triage method can be applied with high performance. ConclusionsThe developed triage method can be used for early autism diagnosis application and support clinical treatment utilizing the advantages of fMCDM techniques of multidimensional criteria. Four medical criteria were selected from the psychologists, while Sociodemographic acquired high attention with 15 selected criteria. For the correlation analysis of the 19 used criteria, the ‘Wave’ criterion has the highest correlation with the triage level and obtained 0.4523. On the contrary, the “Pointing with the index finger” criterion has the lowest correlation and obtained −0.0542. Limitations and future works have also been reported in this study. The study confirms the efficacy of the proposed triage method compared with previous studies in five comparative points with 100%.