Objective Blood pressure is one of vital signs of human, and necessary for clinical judgment. We conduct a retrospective study to verify the consistency between invasive and non-invasive blood pressure measurements based on MIMIC-Ⅲ database. Methods A total of 11 671 patients with 226 093 pairwise measurements of noninvasive and invasive blood pressure measurements are extracted from the MIMIC-Ⅲ database. On the one hand, exploratory data analysis is used to compare the differencebetween invasive and noninvasive measurements from the perspective of age and blood pressure stratification, and certain results are presented in Bland-Altman plot and box plot. On the other hand, we study the association between blood pressure and ICU mortality. Results Our results indicate that at different ages, discrepancies between invasive and non-invasive systolic blood pressure are small in the range of 80-120 mmHg; noninvasive systolic blood pressure is higher than invasive systolic blood pressure as pressure <80 mmHg, and discrepancies increase with the decreasing of the blood pressure; while as pressure is higher than 120 mmHg, noninvasive blood pressure is lower than invasive blood pressure, and discrepancies increase with the increasing of the blood pressure. And Bland-Altman plot shows that results of the two types of measurements exist proportional bias which is difficult to be eliminated. Systolic non-invasive blood pressure is associated with a higher ICU mortality than systolic invasive blood pressure in the same range (< 80 mmHg) , non-invasive mean blood pressure and invasive mean blood pressure show better agreement. Conclusions Our study shows the discrepancies exist between invasive and noninvasive blood pressure measurements during abnormal blood pressure at different ages. In assessing patients' prognosis, non-invasive and invasive mean blood pressure can be interpreted in a consistent manner.