INTRODUCTION: Pulse oximetry is a widely used tool, unfortunately there is a paucity of data investigating its accuracy in Intensive Care Units (ICU) and if they are able to meet mandated FDA criteria as claimed by them in critically ill patients. AIM: To assess bias, precision and accuracy of pulse oximeters used in ICU and factors affecting them. MATERIALS AND METHODS: A prospective cohort study, including 129 patients admitted to the ICU of a tertiary referral centre. Pulse oximetry and blood gas were done simultaneously. Pulse oximetry was done using two pulse oximetres: Nonin and Philips. All physiological variables like haemoglobin, lactate, use of vasopressors and blood pressure were recorded. Bland Altman curves were constructed to determine bias and limits of agreement. Effect of physiological variables on bias and difference between performance characteristics of bias was determined using SPSS. RESULTS: Pulse oximetry overestimated arterial oxygen saturation (SaO) by 1.44%. There was negative correlation between bias and SaO (r=-0.32) and positive correlation with lactate (r=0.16). The Philips pulse oximeter had significant higher bias and variability than Nonin pulse oximeter. (2.49±2.99 versus 0.46±1.68, mean difference = 1.98, 95% C.I. = 1.53 - 2.43, p-value <0.001). CONCLUSION: Pulse oximetry overestimates SaO. Bias tends to increase with rising lactate and hypoxia. There is heterogeneity in performance of various pulse oximetry devices in ICU.