Background/aim: To analyze the potency of a modified early warning score (EWS) to help predict hospital mortality when used for surveillance in nonacute medical wards. Materials and methods: Patients in internal medicine wards were prospectively recruited. First, highest, and last scores; and mean daily score recordings and values were recorded. Nurses calculated scores for each patient upon admission and every 4 h. The last score was the score before death, discharge, or transfer to another ward. The highest scores in total and for each single parameter were used for analysis. Results: Fifty-nine percent of 182 recruited patients had recordings eligible for data analysis. Patients admitted from the emergency room had higher mortality rates than patients admitted from outpatient clinics (15% vs. 1.5%; P = 0.01) as well as patients whose first (40% vs. 4.9%; P = 0.033) and highest scores (18.8% vs. 1.3%; P = 0.003) were equal to or more than 3. The first recorded EWS was not predictive for mortality while the maximum score during the admission period was. Conclusion: This study underlines the fact that each physiological variable of EWS may not have the same weight in determining the outcome.