A study on investigation of the usage of decision support systems and evidence-based medicine relations via machine learning algorithms

Information systems have been popularized in all areas of our lives, and today information systems literacy has also been formed as a concept. With the pandemic, this concept has become more important, especially for physicians. At the same time, information systems, decision support systems in medicine, and evidence-based medicine approaches have become supported and work is continuing rapidly. Therefore, it is inevitable that physicians will be aware of current developments in this field. The aim of this study is to determine the literacy of information systems of physicians working in the academic field but to demonstrate the sensitivity of DSS and evidence-based medicine with a machine learning approach. The dataset was obtained from the physicians working in the academic field in Turkey. By using web-based survey, all the information including some demographic attributes, the usage level of the DDS and EBM systems, and the score of ISL belonging to 810 physicians has been recruited from this survey. A comprehensive analysis process including both statistical significance tests and machine learning algorithms was conducted to determine the group differences about the ISL scale and classify the experienced and inexperienced people on the usage of DDS and EBM systems. The range F1-score for each machine learning algorithm was satisfying $$\left( {F1_{{{\text{DDS}}}} = .818\;{\text{to}}\;.992,\;F1_{{{\text{EBM}}}} = .797\;{\text{to}}\;.955} \right)$$ F 1 DDS = . 818 to . 992 , F 1 EBM = . 797 to . 955 . From the view of both training and testing performance for DDS, random forest algorithm provided the best results than its competitors. In the examination of EBM usage, while C5.0 was found better in terms of training results, extreme gradient boosting machine with linear kernel has the highest F1-score on testing performance. In the study, the ISL scale was applied to physicians working in medical schools and its effectiveness was showed to a great extent. Statistical significance test results were obtained for the title, graduation institution, specialty institution and specialty type characteristics of the physicians. However, we associated DSS, EBM and system and application approaches with the scope of both aforementioned attributes and the ISL scale. We conducted a comprehensive comparative study by using various machine learning algorithms on the subject to the study and obtained useful insight about the factors on the usage level of DDS and EBM systems. It is expected that the results will be a guide for all physicians and administrators working in the academic field.