To assess the differences between ultrasound cardiac output monitor (USCOM) and thermodilution (TD) systematically in cardiac function monitoring of critically ill patients. The Chinese and English literatures about the clinical trials which using USCOM and TD to monitor cardiac function published in CNKI, Wanfang database, China biomedical literature database, VIP database, China Clinical Trial Registration Center, PubMed, Embase and Cochrane Library were searched by computer from the establishment to December 2018. Some indicators, like cardiac output (CO), cardiac index (CI), stroke volume (SV) and other parameters were used to evaluate cardiac function. Literature search, quality evaluation and data extraction were conducted independently by two authors. The tailored Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) was used for literature quality evaluation. EndNote X6 was used for literature screening and management. RevMan 5.3 was used for Meta-analysis. Funnel chart analysis was used for publication bias. A total of 26 studies involving 772 patients were included. Among them, there were 5 literatures found that the agreements of cardiac function between the USCOM and TD methods were poor. Meta-analysis showed that there was no significant difference between the two methods in CO and CI monitoring [CO: mean difference (MD) = -0.06, 95% confidence interval (95%CI) was -0.17 to 0.05, P = 0.31; CI: MD = -0.04, 95%CI was -0.13 to 0.05, P = 0.38]. Subgroup analysis of different TD methods [pulmonary artery catheter (PAC), pulse indicator continuous cardiac output (PiCCO)] and different windows of USCOM ultrasonic probe [aorta (AA), pulmonary artery (PA)] in CO monitoring was not shown significant difference yet (PAC: MD = -0.07, 95%CI was -0.18 to 0.04, P = 0.23; PiCCO: MD = 0.09, 95%CI was -0.31 to 0.50, P = 0.65; AA windows: MD = -0.14, 95%CI was -0.31 to 0.02, P = 0.09; PA windows: MD = -0.00, 95%CI was -0.15 to 0.14, P = 0.95; AA/PA windows: MD = 0.23, 95%CI was -0.40 to 0.86, P = 0.47). However, the difference in SV was statistically significant between the USCOM and TD method (MD = 1.48, 95%CI was 0.04 to 2.92, P = 0.04). Funnel chart showed that the literature distribution of CO and CI monitoring were basically symmetrical, indicating that the bias of literature publication is small. USCOM has good consistency with TD method in monitoring the cardiac function parameters of CO and CI, and different windows of ultrasonic probe of USCOM have no significant influence on the monitoring results, but there is significant difference in the consistency of the two methods in SV monitoring.