Background: Early identification of multidrug-resistant (MDR) infections in critically ill patients remains a major clinical challenge. Delayed recognition often leads to inappropriate empirical therapy, increased mortality, and escalation of antimicrobial resistance. Clinical prediction tools may facilitate early risk stratification and optimize antimicrobial decision-making. Objectives: To develop and internally validate a composite clinical risk score for predicting MDR infections in critically ill patients. Methods: A prospective observational study was conducted among 100 critically ill patients receiving antimicrobial therapy. Independent predictors identified through multivariable logistic regression were incorporated into a weighted risk score. Model performance was evaluated using ROC analysis, calibration testing, and diagnostic accuracy metrics. Results: MDR prevalence was 42%. Independent predictors included ≥5 antibiotics (aOR 3.41; p=0.005), ICU stay >8 days (aOR 2.89; p=0.019), Gram-negative infection (aOR 3.27; p=0.018), and diabetes mellitus (aOR 2.21; p=0.046). A composite 8-point risk score demonstrated increasing MDR probability across low (12%), moderate (41%), and high-risk groups (78%) (p<0.001). Model discrimination was good (AUROC 0.78), with stable bootstrap validation (0.77). Calibration was acceptable (Hosmer–Lemeshow p=0.62). Conclusion: The ICU-MDR Risk Score provides a practical bedside tool with good predictive performance for early identification of high-risk patients.