Hybrid Fuzzy Systems for Clinical Decision Support: Integrating Neural Networks and Evolutionary Algorithms at Tikur Anbessa Specialized Hospital (Preprint)

BACKGROUND 1.1. Background of the Study In recent decades, artificial intelligence (AI) has revolutionized numerous sectors, including healthcare, by enhancing the quality and speed of decision-making through data-driven technologies [1]. Clinical decision support systems (CDSS) are a notable application of AI, aiming to assist healthcare professionals in diagnosing diseases, recommending treatments, and predicting patient outcomes [2]. However, traditional CDSS often struggle with the inherent uncertainty and imprecision present in medical data. Fuzzy logic, introduced by Zadeh, offers a means to manage uncertainty and vague information through linguistic rules and membership functions, making it well-suited for clinical environments where data may be incomplete or imprecise [3]. To further improve system performance, fuzzy systems have been integrated with other AI techniques such as neural networks and evolutionary algorithms, forming what is known as hybrid fuzzy systems [4]. These systems can learn from data (via neural networks) while optimizing fuzzy rules and parameters (via evolutionary methods), making them robust for complex decision-making scenarios. In the context of healthcare, hybrid fuzzy systems have been shown to improve diagnostic accuracy, reduce human error, and enhance treatment planning [5]. Neuro-fuzzy models allow the system to learn complex nonlinear relationships in medical datasets, while evolutionary algorithms help fine-tune these models to achieve higher precision [6]. This combination allows for adaptive, intelligent systems capable of supporting clinicians in high-stakes environments such as intensive care units and emergency rooms. Tikur Anbessa Specialized Hospital (TASH), the largest referral hospital in Ethiopia, faces increasing challenges related to high patient volumes, resource constraints, and complex clinical decision-making processes [7]. Despite advances in digitization, decision support tools in TASH remain rudimentary, and much of the diagnostic and treatment planning relies on manual processes and physician intuition. There is a growing need for intelligent systems that can assist medical professionals in making timely and accurate decisions, particularly under uncertainty. This study seeks to explore the application of hybrid fuzzy systems specifically integrating fuzzy logic, neural networks, and evolutionary algorithms for clinical decision support at Tikur Anbessa Specialized Hospital. By tailoring such a system to the hospital's specific context, this research aims to bridge the gap between AI innovation and real-world healthcare delivery in Ethiopia. OBJECTIVE 1.3. Objectives of the study 1.3.1. General Objective The general objective of the study was to assess a hybrid fuzzy system integrating neural networks and evolutionary algorithms for clinical decision support at Tikur Anbessa Specialized Hospital. 1.3.2. Specific Objectives 1. To identify the key challenges faced in clinical decision-making processes at TASH. 2. To design a hybrid fuzzy-neuro-evolutionary model suitable for clinical data environments. 3. To evaluate the adaptability of the proposed model using selected patient case data. 4. To assess the perceived usefulness of the system among clinicians. METHODS 3.1. Research Design This study adopts a quantitative cross-sectional design supported by exploratory elements to assess the effectiveness of a hybrid fuzzy-neuro-evolutionary system for clinical decision support. A cross-sectional design allows the collection of data from clinicians at a single point in time, enabling the examination of relationships between model design, perceived usefulness, adaptability, and clinical decision-making effectiveness [32]. The study further incorporates design science principles by developing and evaluating an AI-driven system within a real-world hospital setting, as is common in healthcare technology research [33]. 3.2. Research Approach A quantitative research approach is employed to statistically assess the relationships among the study variables. This approach facilitates hypothesis testing, numerical measurement, and generalizability of results using structured instruments [34]. Quantitative data will be supported by limited qualitative insights gathered from open-ended survey questions and interviews, primarily for contextualizing user feedback. 3.3. Population and Sampling The study population includes healthcare professionals (physicians, residents, and senior nurses) at Tikur Anbessa Specialized Hospital. From a total population of approximately 375 clinicians, a sample was drawn using stratified random sampling to ensure representation across departments (e.g., internal medicine, surgery, pediatrics). A minimum sample size of 194 was targeted based on the Krejcie and Morgan formula for a 95% confidence level and 5% margin of error [35]. 3.4. Data Collection Instruments Data was collected using structured questionnaires and a prototype evaluation form. The questionnaire consists of closed-ended Likert-scale items designed to measure perceptions of model design, adaptability, usefulness, and decision-making effectiveness. Additionally, a system usability testing checklist and case-based evaluation form was used to assess the prototype's real-time performance. This approach ensures triangulation between self-reported perceptions and system metrics [36]. 3.5. Variables and Measurement The dependent variable is Decision-Making Effectiveness, operationalized through clinician-rated improvements in accuracy, speed, and confidence. Independent variables include: • Clinical Challenges (e.g., delay, uncertainty) • Model Design (clarity, functionality) • Model Adaptability (system learning, flexibility) • Perceived Usefulness (alignment with clinical needs) Each variable is measured using multiple Likert-scale items (1 = Strongly Disagree to 5 = Strongly Agree) adapted from validated scales in healthcare and technology acceptance research [46]. 3.6. Data Analysis Techniques Data was analyzed using SPSS v26. Descriptive statistics (mean, standard deviation) was summarized respondent demographics and item responses. Multiple linear regression was used to assess the predictive influence of the independent variables on decision-making effectiveness. Correlation analysis was explored inter-variable relationships. Where applicable, ANOVA may be applied to test differences across clinical departments. These statistical techniques are standard in evaluating health information systems [37]. 3.7. Reliability and Validity Measures Instrument reliability will be evaluated using Cronbach’s Alpha, with thresholds of ≥ 0.70 considered acceptable for internal consistency [38]. Content validity will be ensured through expert review by medical informatics and AI specialists. A pilot test with 20 clinicians will refine question clarity and scale performance. Construct validity will be supported via exploratory factor analysis (EFA) to confirm that items align with their intended constructs. Table 1 Cronbach’s Alpha a test Reliability Statistics Cronbach's Alpha N of Items .925 34 Source: Survey Result, (2025) The reliability analysis presented in Table 1 shows that the measurement instrument achieved a Cronbach’s Alpha coefficient of 0.925 across 34 items, indicating excellent internal consistency. This result exceeds the commonly accepted threshold of 0.70 and falls within the range classified as outstanding reliability for multi-item scales [38]. The high alpha value confirms that the items used to measure clinical challenges; model design, model adaptability, perceived usefulness, and decision-making effectiveness are consistently capturing their intended constructs. This finding is consistent with prior studies on clinical decision support and intelligent systems, which emphasize the need for reliable and stable measurement instruments when evaluating complex, technology-driven constructs in healthcare environments [8], [11], [13]. In studies involving hybrid intelligent systems and clinician perceptions, high internal consistency is particularly critical due to the multidimensional nature of constructs such as usability, adaptability, and decision effectiveness [4]. Similar levels of reliability have been reported in empirical evaluations of fuzzy-based and neuro-fuzzy clinical decision support systems, where well-structured instruments were necessary to ensure dependable statistical inference [11], [14], [17]. Moreover, the strong reliability outcome supports the suitability of the instrument for advanced multivariate analyses, including correlation and multiple regressions, as recommended [37]. In line with design science research principles [33], the robustness of the measurement tool strengthens the credibility of subsequent findings and ensures that observed relationships among variables reflect true associations rather than measurement error. Overall, the result confirms that the survey instrument is methodologically sound and appropriate for assessing clinician perceptions of hybrid fuzzy decision support systems in resource-constrained healthcare settings such as Tikur Anbessa Specialized Hospital. 3.8. Ethical Considerations Ethical clearance was secured from the Institutional Review Board (IRB) of Addis Ababa University and Tikur Anbessa Specialized Hospital. All participants were received informed consent forms explaining the purpose, voluntary nature, and confidentiality of the study. No identifiable personal or medical data will be collected. Prototype testing was used anonymized synthetic or historical patient cases to avoid any ethical breaches in real-time diagnosis or treatment [39]. RESULTS 4. Results and Discussion Out of the 194 distributed surveys, 185 responses were collected, representing a diverse group of clinicians at Tikur Anbessa Specialized Hospital. Of the collected responses, 95 were from male clinicians, while 90 were from female clinicians. This distribution ensures a balanced representation of both genders, allowing for a comprehensive analysis of perceptions and experiences with the hybrid fuzzy decision support system. The gender distribution reflects a fair representation of the hospital's clinical staff, providing valuable insights into how different demographic groups engage with and evaluate the system’s effectiveness in improving clinical decision-making. Table 2: Descriptive Statistics Clinical Challenges Statements N Mean Std. Deviation I often face difficulty making decisions due to incomplete or uncertain patient data. 185 3.2703 1.39191 Lack of access to real-time clinical information delays my decisions. 185 3.6865 1.25069 Decision-making is heavily dependent on individual clinical experience. 185 4.0324 1.03684 Existing support tools are insufficient for complex case decisions. 185 3.6000 1.02257 I experience high cognitive load during diagnosis due to ambiguous symptoms. 185 3.4865 1.06887 Valid N (listwise) 185 Source: Survey Result, (2025) The descriptive statistics in Table 2 indicate that clinicians at Tikur Anbessa Specialized Hospital experience substantial challenges in their clinical decision-making processes. The highest mean score was recorded for the statement “Decision-making is heavily dependent on individual clinical experience” (M = 4.03, SD = 1.04), suggesting that in the absence of robust clinical decision support systems, clinicians rely heavily on personal expertise and judgment. This finding is consistent with previous studies reporting strong dependence on individual clinical experience in resource-constrained healthcare settings where formal decision support tools are limited or insufficiently developed [10], [12], [27]. Similarly, the relatively high mean score for “Lack of access to real-time clinical information delays my decisions” (M = 3.69, SD = 1.25) highlights systemic information gaps that adversely affect timely and effective decision-making. Prior studies have shown that delays in accessing up-to-date patient information significantly reduce decision quality and increase the likelihood of diagnostic errors [10], [31]. The perception that “Existing support tools are insufficient for complex case decisions” (M = 3.60, SD = 1.02) further reflects the limitations of conventional rule-based or manual systems in addressing complex and uncertain clinical scenarios, as documented in earlier research on clinical decision support systems [9], [27]. Moderate mean values for “I experience high cognitive load during diagnosis due to ambiguous symptoms” (M = 3.49, SD = 1.07) and “I often face difficulty making decisions due to incomplete or uncertain patient data” (M = 3.27, SD = 1.39) indicate persistent challenges associated with data uncertainty and diagnostic ambiguity. These findings align with studies conducted in developing-country healthcare environments, where incomplete medical records and variable data quality increase clinicians’ cognitive burden and complicate clinical reasoning processes [7], [12]. Previous research suggests that such challenges can be effectively mitigated through intelligent decision support systems incorporating fuzzy logic, which are specifically designed to handle imprecision, uncertainty, and ambiguity in medical data [3], [13], [14]. Overall, the results presented in Table 2 underscore the critical need for advanced and adaptive clinical decision support systems capable of reducing reliance on individual experience, alleviating cognitive load, and enhancing decision-making in environments characterized by uncertainty and limited real-time information. These findings provide strong empirical justification for the adoption of hybrid fuzzy-based decision support models in tertiary healthcare institutions such as Tikur Anbessa Specialized Hospital. Table 3: Descriptive Statistics Model Design Statements N Mean Std. Deviation The hybrid system is logically structured and easy to understand. 185 3.9027 1.00609 The system integrates fuzzy logic, neural networks, and evolutionary algorithms effectively. 185 3.9405 .98452 I can easily interpret the system’s diagnostic suggestions. 185 3.9459 1.01473 The design supports decision-making under uncertain or imprecise clinical conditions. 185 3.8432 1.13361 The interface of the system is clear and user-friendly. 185 4.0649 .93606 The model aligns well with existing clinical workflows. 185 4.0757 1.08583 The technical components are well-coordinated to deliver accurate outcomes. 185 3.7946 1.07894 Valid N (listwise) 185 Source: Survey Result, (2025) The descriptive statistics in Table 3 indicate that clinicians hold generally positive perceptions regarding the design of the proposed hybrid clinical decision support system. All items recorded mean values above the midpoint of the scale, suggesting broad agreement that the system is well designed, understandable, and supportive of clinical decision-making. The highest mean scores were observed for “The model aligns well with existing clinical workflows” (M = 4.08, SD = 1.09) and “The interface of the system is clear and user-friendly” (M = 4.06, SD = 0.94), highlighting the importance of workflow compatibility and usability in promoting clinician acceptance of decision support technologies. These findings are consistent with prior studies emphasizing that system usability and workflow integration are critical determinants of effective clinical decision support adoption [11], [28]. High mean values for “I can easily interpret the system’s diagnostic suggestions” (M = 3.95, SD = 1.01) and “The hybrid system is logically structured and easy to understand” (M = 3.90, SD = 1.01) indicate that the system’s architecture supports transparency and interpretability, which are essential for clinician trust and sustained use. Previous research on clinical decision support systems highlights that interpretability and logical system design significantly enhance clinician confidence and reduce resistance to technology-assisted decision-making [9], [10], [30]. Furthermore, the strong agreement with “The system integrates fuzzy logic, neural networks, and evolutionary algorithms effectively” (M = 3.94, SD = 0.98) and “The design supports decision-making under uncertain or imprecise clinical conditions” (M = 3.84, SD = 1.13) demonstrates the perceived effectiveness of the hybrid intelligent approach. These results align with earlier studies showing that hybrid models combining fuzzy logic with learning and optimization techniques are well suited for handling uncertainty, complexity, and nonlinear relationships inherent in medical decision-making [3], [4], [17], [23]. The relatively high mean score for “The technical components are well-coordinated to deliver accurate outcomes” (M = 3.79, SD = 1.08) further suggests that clinicians perceive the system as technically coherent and reliable. Prior research has emphasized that seamless coordination among intelligent components is essential for achieving accurate and clinically meaningful outputs in hybrid decision support systems [4], [16]. Overall, the findings presented in Table 3 demonstrate that the proposed hybrid model is perceived as well designed, interpretable, and compatible with clinical practice. This supports the argument that carefully structured hybrid intelligent systems can enhance clinician trust, facilitate adoption, and improve decision-making performance in complex and uncertain healthcare environments such as tertiary hospitals. Table 4: Descriptive Statistics Model Adaptability Statements N Mean Std. Deviation The system adapts to new clinical scenarios effectively. 185 3.8432 .95113 The system improves its performance over time. 185 3.4811 1.04305 The model updates its recommendations when given additional patient data. 185 2.9892 1.22470 The system handles diverse patient conditions without performance drops. 185 2.8919 1.26370 The model adjusts well to diagnostic rule changes. 185 3.4486 1.10769 The system is capable of learning from clinician feedback. 185 3.3730 1.15454 The model is flexible in processing incomplete or noisy data. 185 3.5135 1.09400 Valid N (listwise) 185 Source: Survey Result, (2025) The descriptive statistics in Table 4 reveal moderate perceptions of adaptability regarding the proposed hybrid clinical decision support system. While several items indicate positive adaptive capabilities, others suggest areas where adaptability remains limited, particularly in dynamic and highly variable clinical contexts. The highest mean score was recorded for “The system adapts to new clinical scenarios effectively” (M = 3.84, SD = 0.95), indicating that clinicians generally perceive the system as capable of responding to novel or unfamiliar cases. This finding is consistent with prior research showing that hybrid intelligent systems integrating fuzzy logic and learning mechanisms are better suited to accommodate variability and uncertainty in clinical environments [3], [4], [17]. Moderate mean values for “The model is flexible in processing incomplete or noisy data” (M = 3.51, SD = 1.09), “The system improves its performance over time” (M = 3.48, SD = 1.04), and “The model adjusts well to diagnostic rule changes” (M = 3.45, SD = 1.11) suggest that clinicians recognize the system’s learning and adaptation potential, although these capabilities may not yet be fully realized. Previous studies emphasize that while adaptive learning is core strength of neuro-fuzzy and hybrid systems, its effectiveness often depends on the availability of high-quality data and continuous model tuning [5], [21], [23]. In contrast, relatively lower mean scores for “The model updates its recommendations when given additional patient data” (M = 2.99, SD = 1.22) and “The system handles diverse patient conditions without performance drops” (M = 2.89, SD = 1.26) indicate perceived limitations in real-time learning and generalization across heterogeneous patient profiles. These findings align with existing literature that highlights challenges in implementing fully adaptive clinical decision support systems, particularly in resource-constrained settings where data streams may be fragmented, incomplete, or inconsistently updated [7], [12], [26]. The moderate agreement with “The system is capable of learning from clinician feedback” (M = 3.37, SD = 1.15) further suggests that feedback-driven learning mechanisms are present but may require stronger integration into routine clinical workflows. Prior research indicates that effective incorporation of clinician feedback is essential for improving system adaptability, acceptance, and long-term performance [11]. Overall, the findings in Table 4 suggest that while the proposed hybrid model demonstrates promising adaptive features particularly in handling uncertainty and evolving clinical scenarios its real-time updating and generalization capabilities remain areas for further enhancement. These results underscore the need for continuous model refinement, improved data integration, and stronger feedback mechanisms to fully realize the adaptive potential of hybrid fuzzy-based clinical decision support systems in tertiary healthcare environments. Table 5: Descriptive Statistics Perceived usefulness Statements N Mean Std. Deviation The system helps me make better clinical decisions. 185 3.4054 1.05448 Using the system enhances the speed of my decision-making. 185 3.4919 1.04319 The system improves the accuracy of diagnoses. 185 3.5243 1.17061 The system is relevant to my day-to-day clinical needs. 185 4.0541 .89522 Overall, the system is a valuable tool in clinical practice. 185 4.0973 .91559 Valid N (listwise) 185 Source: Survey Result, (2025) The descriptive statistics in Table 5 indicate that clinicians generally perceive the proposed hybrid clinical decision support system as a useful tool in their daily practice. The highest mean scores were recorded for “Overall, the system is a valuable tool in clinical practice” (M = 4.10, SD = 0.92) and “The system is relevant to my day-to-day clinical needs” (M = 4.05, SD = 0.90), suggesting strong perceived relevance and practical value. These findings are consistent with prior research emphasizing that perceived usefulness is a key determinant of clinician acceptance and sustained use of clinical decision support systems [11], [29], [30]. Moderate mean values for “The system improves the accuracy of diagnoses” (M = 3.52, SD = 1.17), “Using the system enhances the speed of my decision-making” (M = 3.49, SD = 1.04), and “The system helps me make better clinical decisions” (M = 3.41, SD = 1.05) indicate that clinicians recognize the system’s potential to enhance decision quality and efficiency, although the perceived impact on diagnostic performance is not uniformly strong. Similar patterns have been reported in earlier studies, where clinical decision support systems were valued for their supportive role but were not viewed as complete replacements for clinician judgment, particularly in complex or uncertain cases [7], [10], [31]. The relatively higher variability in responses related to diagnostic accuracy and decision speed suggests differences in individual clinician experience, familiarity with intelligent systems, and clinical context. Previous research indicates that perceived usefulness tends to increase as users gain greater exposure to the system and as the system becomes more tightly integrated into routine workflows [11], [28]. Overall, the results presented in Table 5 demonstrate that the proposed hybrid model is perceived as relevant and valuable by clinicians, supporting its potential for adoption in clinical settings. However, the moderate ratings related to decision speed and accuracy suggest that continued system refinement, user training, and workflow integration may be necessary to fully realize its perceived usefulness and maximize its impact on clinical decision-making. Table 6: Descriptive Statistics Decision-Making Effectiveness Statements N Mean Std. Deviation I am more confident in my decisions when supported by the system. 185 4.2270 .87373 The system helps reduce diagnostic errors. 185 3.5189 1.06879 My decisions are made more quickly with system support. 185 3.9351 .99242 I can manage uncertain cases better with the system. 185 3.2703 1.39191 The system helps me arrive at evidence-based conclusions. 185 3.6865 1.25069 The quality of my clinical decisions has improved with the system. 185 4.0324 1.03684 The system supports better risk assessment in patient care. 185 3.6000 1.02257 I rely on the system’s recommendations in complex cases. 185 3.4865 1.06887 The system improves coordination between departments in shared decision-making. 185 3.9027 1.00609 My overall decision-making effectiveness has improved with the use of the system. 185 3.9405 .98452 Valid N (listwise) 185 Source: Survey Result, (2025) The results indicate that clinicians perceive high confidence in their decisions when supported by the system, with a mean score of 4.227. This suggests that the clinical decision support system (CDSS) provides reliable guidance, enhancing decision-making confidence, which aligns with previous studies on CDSS effectiveness [8], [10], [11], [29]. High confidence in decision-making can reduce hesitation and improve patient outcomes, particularly in complex clinical scenarios. The system also contributes to reducing diagnostic errors, though the mean score of 3.519 indicates a moderate effect. This demonstrates that while CDSS supports clinicians in avoiding mistakes through evidence-based recommendations, its impact may vary depending on user experience and case complexity. This finding is consistent with prior research reporting variability in system adoption and reliance among clinicians [11], [28], [30]. Clinicians reported that decision-making becomes faster with system support, with a mean of 3.935. The system streamlines workflow by providing relevant information and recommendations quickly, which supports previous studies highlighting the efficiency benefits of CDSS in high-pressure environments [8], [9]. Faster decision-making can improve patient care and reduce delays in treatment delivery. Management of uncertain cases received a lower mean score of 3.270 and the highest standard deviation of 1.392, indicating significant variability in perceptions. This suggests that while the system provides some support, clinicians still rely heavily on their judgment when dealing with ambiguous or complex cases. This limitation is consistent with earlier research emphasizing the need for human oversight in atypical or rare clinical scenarios [1], [12]. The system helps clinicians arrive at evidence-based conclusions, with a mean of 3.687, demonstrating moderate effectiveness in supporting clinical reasoning. This finding aligns with prior studies showing that hybrid intelligent systems, including fuzzy logic and neural networks, improve the alignment of clinical decisions with best practice guidelines [14], [17]. Similarly, the quality of clinical decisions is perceived to have improved with the system (mean = 4.032), which supports previous evidence that CDSS can enhance accuracy and reliability in patient care [10], [11], [13]. Risk assessment in patient care is moderately enhanced by the system, with a mean score of 3.600. CDSS assists clinicians in evaluating patient risks through structured analysis and rule-based guidance, although variability in responses suggests differences in user familiarity or system application. This finding is consistent with research highlighting the role of fuzzy logic and hybrid systems in structured risk evaluation [6], [20], [25]. Reliance on system recommendations in complex cases shows a moderate mean of 3.487, reflecting that clinicians often combine system guidance with personal expertise [27]. The system also improves coordination between departments in shared decision-making, with a mean of 3.903. This aligns with studies demonstrating that CDSS facilitates collaboration by integrating patient information and supporting consensus across multidisciplinary teams [8], [31]. Overall, respondents perceived that their decision-making effectiveness has improved with system use, reflected in a mean of 3.941. This supports prior findings that CDSS adoption enhances clinical workflow efficiency and overall decision quality [11], [12], [30]. In summary, the descriptive statistics suggest that CDSS is perceived as a valuable tool for improving confidence, decision quality, decision speed, and inter-department coordination. However, its effectiveness in managing uncertain cases, reducing diagnostic errors, and guiding complex decisions is more variable. These findings are consistent with previous studies indicating that while CDSS enhances clinical decision-making, human judgment remains critical, especially in complex or atypical cases [1], [8], [10], [11], [28]. Table 7 Correlation matrixes between variables CC MD MA PU DME Clinical Challenges Pearson Correlation 1 Sig. (2-tailed) N 185 Model Design Pearson Correlation .194** 1 Sig. (2-tailed) .008 N 185 185 Model adaptability Pearson Correlation .168* .228** 1 Sig. (2-tailed) .023 .002 N 185 185 185 Perceived usefulness Pearson Correlation .280** .171* .313** 1 Sig. (2-tailed) .000 .020 .000 N 185 185 185 185 Decision-Making Effectiveness Pearson Correlation .110 .378** .403** .206** 1 Sig. (2-tailed) .136 .000 .000 .005 N 185 185 185 185 185 **. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed). Source: Survey Result, (2025) The correlation matrix in Table 7 examines the relationships among Clinical Challenges (CC), Model Design (MD), Model Adaptability (MA), Perceived Usefulness (PU), and Decision-Making Effectiveness (DME). The analysis reveals several significant associations that provide insights into factors influencing clinicians’ decision-making performance. Clinical Challenges (CC) shows a weak but significant positive correlation with Model Design (MD) (r = 0.194, p < 0.01) and Model Adaptability (MA) (r = 0.168, p < 0.05). This indicates that as clinical challenges increase, the emphasis on well-structured model design and adaptable systems becomes more important. The positive relationship suggests that addressing clinical challenges through careful design and adaptability can enhance system usability. These findings are consistent with previous studies indicating that clinical decision support systems need to accommodate diverse clinical contexts and complex scenarios to be effective [8], [11], [28]. Clinical Challenges (CC) is also moderately correlated with Perceived Usefulness (PU) (r = 0.280, p < 0.01), suggesting that clinicians facing more challenges perceive the system as more beneficial. This aligns with prior research showing that perceived usefulness increases when clinicians recognize that the system can reduce cognitive load and support problem-solving in challenging clinical situations [38], [39]. The correlation with Decision-Making Effectiveness (DME) (r = 0.110) is not statistically significant (p = 0.136), implying that clinical challenges alone may not directly improve decision-making outcomes unless the system design and adaptability adequately address these challenges [7], [31]. Model Design (MD) has significant positive correlations with Model Adaptability (MA) (r = 0.228, p < 0.01), Perceived Usefulness (PU) (r = 0.171, p < 0.05), and Decision-Making Effectiveness (DME) (r = 0.378, p < 0.01). This indicates that well-designed models contribute to both the adaptability of the system and the perceived usefulness, ultimately enhancing decision-making effectiveness. These findings are consistent with prior studies emphasizing that system architecture and design quality directly impact user satisfaction, perceived usefulness, and clinical performance [9], [28]. A robust model design ensures that the system aligns with clinical workflows and user needs, supporting more accurate and timely decisions [8], [10]. Model Adaptability (MA) is positively correlated with Perceived Usefulness (PU) (r = 0.313, p < 0.01) and Decision-Making Effectiveness (DME) (r = 0.403, p < 0.01). This strong correlation highlights that adaptable systems, which can adjust to different clinical contexts, significantly enhance both how useful clinicians perceive the system to be and their overall decision-making performance. This supports previous findings that flexible, context-sensitive systems are critical for effective clinical decision support, particularly in complex hospital environments [14], [17], [23]. Adaptable systems can accommodate patient variability, unexpected cases, and workflow differences, improving clinician confidence and decision quality. Perceived Usefulness (PU) shows a significant positive relationship with Decision-Making Effectiveness (DME) (r = 0.206, p < 0.01). This suggests that the more clinicians perceive the system as useful, the more effective their decisions become. This aligns with the Technology Acceptance Model (TAM), which posits that perceived usefulness is a key determinant of system adoption and performance outcomes [38], [46]. Systems that are regarded as beneficial and reliable are more likely to be integrated into daily practice, leading to improvements in decision quality, speed, and confidence [11], [30]. Finally, the matrix indicates that while Clinical Challenges alone have limited direct impact on decision-making effectiveness, Model Design, Adaptability, and Perceived Usefulness act as critical mediators. Their positive and significant correlations with Decision-Making Effectiveness demonstrate that the structural and functional qualities of the system are essential for translating clinical challenges into improved decision outcomes. This finding is consistent with prior studies emphasizing that CDSS effectiveness depends not only on the complexity of clinical problems but also on system quality, adaptability, and clinician perception [8], [10], [24], [31]. Table 8 Model Summary Model Summarya Model R R Square Adjusted R Square Std. Error of the Estimate 1 .867a .751 .746 .76427 a. Predictors: (Constant), Perceived usefulness , Model Design , Clinical Challenges , Model adaptability b. Dependent Variable: Decision-Making Effectiveness Source: Survey Result, (2025) The multiple regression analysis in Table 8 indicates a strong relationship between the set of predictors Clinical Challenges, Model Design, Model Adaptability, and Perceived Usefulness and Decision-Making Effectiveness. The multiple correlation coefficient, R = 0.867, demonstrates a strong positive linear association, suggesting that these variables collectively have a substantial impact on clinicians’ decision-making performance. This finding aligns with previous studies reporting that clinical decision support systems (CDSS) designed with high usability and adaptability significantly improve decision outcomes in hospital settings [8], [10], [11]. The model explains a substantial portion of the variance in Decision-Making Effectiveness, with R² = 0.751, indicating that approximately 75.1% of the variability in decision-making outcomes can be accounted for by the independent variables. This high explanatory power is consistent with prior research, which found that factors such as system design quality, adaptability, and perceived usefulness are critical determinants of clinical performance and workflow efficiency [9], [28], [30]. A model explaining more than 70% of the variance is generally considered robust in healthcare technology research, reflecting strong predictive capability [11]. The Adjusted R² = 0.746 further confirms the reliability of the model after adjusting for the number of predictors and sample size. The minimal difference between R² and Adjusted R² suggests that all four predictors contribute meaningfully to explaining decision-making effectiveness, rather than being redundant or irrelevant. This observation supports earlier studies emphasizing that the combination of well-designed system architecture, adaptive functionality, and clinician-perceived usefulness is essential for translating system features into practical improvements in decision-making [14], [17], [23]. The standard error of the estimate (0.76427) indicates the average deviation between observed and predicted values. For a 5-point Likert scale used in this study, this error is relatively small, demonstrating that the model predicts decision-making effectiveness with reasonable accuracy. This precision is consistent with previous findings showing that robustly designed CDSS can reliably support clinicians’ decisions, reducing cognitive load and enhancing workflow efficiency [11], [29], [30]. Overall, the model summary suggests that Decision-Making Effectiveness can be strongly predicted by Clinical Challenges, Model Design, Model Adaptability, and Perceived Usefulness, highlighting the importance of system design and user perception in clinical settings. These results are consistent with prior research demonstrating that the effectiveness of CDSS depends not only on the complexity of clinical cases but also on how well the system adapts to user needs and supports evidence-based decision-making [8], [10], [28], [31]. Table 9 ANOVA Analysis ANOVAa Model Sum of Squares df Mean Square F Sig. 1 Regression 35.326 4 8.831 15.120 .000b Residual 105.139 180 .584 Total 140.465 184 a. Dependent Variable: Decision-Making Effectiveness b. Predictors: (Constant), Perceived usefulness , Model Design , Clinical Challenges , Model adaptability Source: Survey Result, (2025) The ANOVA results in Table 9 assess the overall significance of the regression model in predicting Decision-Making Effectiveness based on the independent variables Clinical Challenges, Model Design, Model Adaptability, and Perceived Usefulness. The regression model produced an F-value of 15.120 with a corresponding p-value of 0.000, indicating that the model is statistically significant at the 1% level. This demonstrates that the set of predictors collectively explains a significant proportion of variance in decision-making effectiveness, which aligns with prior studies emphasizing that the combined effect of system design, adaptability, and perceived usefulness strongly influences clinical outcomes [8], [10], [28]. The sum of squares for regression (35.326) and the mean square for regression (8.831) reflect the variability in decision-making effectiveness accounted for by the independent variables. The residual sum of squares (105.139) and mean square (0.584) represent the unexplained variability in the dependent variable. The comparison of these values through the F-test confirms that the model’s predictive power is significantly greater than chance, consistent with previous research reporting that well-structured clinical decision support systems (CDSS) significantly improve decision quality, speed, and confidence when integrated into hospital workflows [11], [29], [30]. The ANOVA findings complement the R² value of 0.751 reported in the model summary, supporting the conclusion that approximately 75.1% of the variance in decision-making effectiveness can be explained by Clinical Challenges, Model Design, Model Adaptability, and Perceived Usefulness. This high explanatory power is consistent with earlier studies demonstrating that CDSS features such as adaptability, user-centric design, and perceived usefulness are crucial for enhancing clinician decision performance and reducing cognitive workload [9], [14], [17], [23]. Overall, the ANOVA results confirm that the regression model is both statistically significant and practically meaningful. The significance of the F-test indicates that the independent variables jointly contribute to improving clinical decision-making effectiveness, reinforcing prior evidence that CDSS implementation requires a combination of robust model design, adaptive functionality, and perceived usefulness to optimize healthcare outcomes [8], [10], [28], [31]. These results provide a strong basis for further analysis of individual predictors using the coefficients table to identify which factors exert the greatest influence on decision-making effectiveness. Table 10 Multiple Regression Coefficients Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 1.957 .310 6.308 .000 Clinical Challenges .055 .026 -.018 2.115 .036 Model Design .259 .058 .298 4.466 .001 Model adaptability .293 .064 .319 4.578 .001 Perceived usefulness .092 .045 .061 2.044 .043 a. Dependent Variable: Decision-Making Effectiveness Source: Survey Result, (2025) 1. Clinical Challenges The regression analysis shows that Clinical Challenges have a positive but small effect on Decision-Making Effectiveness (B = 0.055, Beta = -0.018), with a t-value of 2.115 and Sig. = 0.036. This indicates that clinical challenges are statistically significant predictors, meaning that as clinical challenges increase, there is a slight but measurable influence on decision-making effectiveness. The negative Beta reflects that, after accounting for other variables, the challenges may slightly reduce effectiveness, highlighting the complex interplay between clinical difficulties and decision-making outcomes. Previous studies have reported similar findings, noting that clinical complexity and challenges can affect clinicians’ reliance on decision support systems. While CDSS improves overall performance, higher clinical complexity sometimes slows decision-making or introduces variability in outcomes, indicating the importance of supporting clinicians in challenging cases [11], [28]. Clinical challenges can negatively moderate the effect of system quality on decision-making, aligning with the small negative Beta observed in this study [28]. Furthermore, the result suggests that addressing clinical challenges directly through system design or workflow optimization is essential. Integrating intelligent alerting, case prioritization, and decision support tailored to complex scenarios helps mitigate the negative impact of clinical challenges [8]. Therefore, even though Clinical Challenges show a small effect, this study confirms that CDSS implementation must account for case complexity to maximize decision-making effectiveness. 2. Model Design The regression coefficient for Model Design (B = 0.259, Beta = 0.298, t = 4.466, Sig. = 0.001) shows a strong, statistically significant positive effect on Decision-Making Effectiveness. This indicates that improvements in system design such as interface usability, intuitive workflow integration, and clarity of output substantially enhance clinicians’ decision-making. Model Design has the second-highest standardized Beta in this regression, highlighting its critical role in supporting effective clinical decisions. Well-designed CDSS interfaces reduce cognitive load and support faster, evidence-based decision-making, particularly in environments with high patient volume [12], [8]. System design quality is one of the strongest predictors of decision-making performance in clinical contexts, emphasizing that design affects both accuracy and confidence in complex cases [11]. The current results reinforce the idea that investments in system design directly influence the practical effectiveness of CDSS. A well-structured model promotes efficiency and clinician satisfaction, leading to better adoption and adherence [36]. Therefore, organizations aiming to improve clinical decision-making should prioritize design improvements, aligning with broader evidence in healthcare informatics. 3. Model Adaptability Model Adaptability has the strongest effect in this study (B = 0.293, Beta = 0.319, t = 4.578, Sig. = 0.001), indicating that systems capable of adjusting to varying patient scenarios, clinician preferences, and workflow conditions significantly improve Decision-Making Effectiveness. Adaptive features, such as context-aware recommendations and customizable alerts, are essential for clinical decision support systems. Adaptability in hybrid intelligent systems allows dynamic response to uncertain or complex cases, which is critical for improving outcomes [4], [6]. Adaptive CDSS improves not only speed but also accuracy in patient care by presenting relevant information tailored to the clinician’s immediate context [11]. The findings suggest that hospitals and health institutions should prioritize adaptive system features to enhance decision-making effectiveness, particularly in environments with variable patient complexity. Integrating fuzzy logic and adaptive algorithms can help systems adjust to uncertain or evolving clinical cases, thereby maximizing support for clinicians [3], [13]. The result here confirms that adaptability is a key determinant of clinical decision support success. 4. Perceived Usefulness Perceived Usefulness shows a statistically significant positive effect (B = 0.092, Beta = 0.061, t = 2.044, Sig. = 0.043), though smaller than Model Design and Model Adaptability. This suggests that when clinicians perceive the system as useful, their overall decision-making effectiveness improves. Even a modest increase in perceived usefulness translates into measurable improvements, supporting technology acceptance theory in clinical contexts. Perceived usefulness is a major determinant of user acceptance of technology, influencing actual use and performance outcomes [29]. Clinicians are more likely to rely on and benefit from CDSS when they perceive it as valuable for their decision-making [11], [28]. The result highlights the importance of user perception in CDSS implementation. Institutions should focus on demonstrating the system’s practical value, training users effectively, and ensuring feedback loops that reinforce the system’s utility [8]. Perception of usefulness is as critical as technical functionality, influencing adoption, engagement, and ultimately, decision-making effectiveness in clinical practice. Table 11: Summary of Regression Results and Interpretation Variable Unstandardized Coefficient (B) Sig. (p-value) Interpretation Alignment with Previous Studies Clinical Challenges 0.055 0.036 Small but significant positive effect; higher clinical challenges slightly affect decision-making effectiveness Aligns with previous studies showing clinical complexity impacts reliance on CDSS and decision quality [8], [11], [28] Model Design 0.259 0.001 Strong positive effect; well-designed system improves decision-making effectiveness Consistent with research emphasizing interface usability and workflow integration as key factors in CDSS performance [8], [11], [12], [28] Model Adaptability 0.293 0.001 Strongest positive effect; adaptive systems enhance decision-making effectiveness in variable scenarios Supports studies highlighting adaptability and context-aware features as critical for effective clinical decision support [3], [4], [6], [11], [13] Perceived Usefulness 0.092 0.043 Moderate positive effect; perception of usefulness enhances decision-making effectiveness Matches findings that perceived usefulness influences acceptance, reliance, and performance outcomes in CDSS [8], [11], [28], [29] Source: Survey result,( 2024) CONCLUSIONS 5.1. Conclusion The study assessed a hybrid fuzzy system integrating neural networks and evolutionary algorithms for clinical decision support at Tikur Anbessa Specialized Hospital. The findings indicate that Decision-Making Effectiveness among clinicians is significantly influenced by the design and adaptability of the system, as well as by clinical challenges and perceived usefulness. Model Adaptability and Model Design were identified as the strongest factors, demonstrating that systems that are flexible, context-aware, and well-structured substantially improve clinicians’ ability to make effective, evidence-based decisions. Clinical Challenges and Perceived Usefulness also contributed positively, highlighting that case complexity and user perception play important but smaller roles in supporting decision-making. Overall, the study confirms that implementing a hybrid fuzzy system with adaptive and user-friendly features can enhance clinical decision-making, reduce errors, and improve patient care at Tikur Anbessa Specialized Hospital. Ensuring that the system is well-designed, adaptable to various clinical scenarios, and perceived as useful by clinicians is crucial for maximizing its effectiveness. These findings provide practical guidance for further development, adoption, and optimization of clinical decision support systems within the hospital.