Patients are different regarding their health conditions, illnesses and ailments, and mobility. These and other factors could affect their specific feelings about the thermal environment. Current methods of predicting thermal sensations were developed based on research on a group of healthy people. Therefore, the use of these methods with patients should be verified. This paper aimed to verify selected thermal comfort models in terms of their reliability to predict the average assessment of thermal conditions in the patient rooms and the percentage of dissatisfied patients. Thermal condition analyses were based on microclimate parameter measurements and extensive questionnaires among patients, done in parallel in the hospital’s patient rooms. The research was carried out in 23 hospitals located in the following provinces in Poland: (Warmińsko-Mazurskie, Kujawsko-Pomorskie, Mazowieckie, Wielkopolskie, Łódzkie, Świętokrzyskie, Lubelskie and Dolnośląskie). Overall, 890 patients from 14 different specialities participated in the research. Actual thermal sensations of patients differed from predictions obtained from selected thermal comfort models. A comparison of the actual thermal sensations with the PMV model indicates the shift of the comfort curve towards cooler rather than neutral conditions. A comparison of the same to the adaptive model predictions indicates that patients had a higher dissatisfaction rate and preferred cooler thermal conditions than predicted. Research findings show that normative models are not fully reliable in predicting patients’ thermal sensations.