Objective The present study was designed to compare safety and efficiency of deep sedation and moderate sedation during upper gastrointestinal endoscopy in inpatients having comorbidities and classified ASA class Ⅱ-Ⅲ. Methods A prospective, randomized and doubleblinded study was performed. Onehundred and twenty inpatients were enrolled, with 40 patients in each group. Inpatients with comorbidities and ASA class Ⅱ-Ⅲ were randomized into moderate sedation (Group M, midazolam+sufentanil),deepsedation with two anesthetics (GroupD2, propofol+sufentanil) and deep sedation with three anesthetics (Group D3, propofol+midazolam+sufentanil). The vital signs, sedation induction time, procedure time, recovery time, patients’ comfortability and satisfaction, endoscopists’ satisfaction, sedationrelated complications and procedurerelated complications were recorded. Results There was no statistical difference in the three groups regarding age, gender, body mass index, ASA classification, or procedure time. Moderate sedation could significantly shorten induction time (P=0.001) and recovery time (P=0.002). Compared with group M, patients in group D2 and group D3 had higher risks of sedatiorelated adverse events. Endoscopists were satisfied with the different levels of sedation in all three groups. Patients in group D2 and group D3 had more comfortability (P＜0.001). Compared with group D2, patients in group D3 required significantly less propofol to reach deep sedation (P＜0.001), and it resulted in a lower incidence of circulation and respiration adverse events. Conclusion For most inpatients with comorbidities and classified ASA class Ⅱ-Ⅲ, routine gastrointestinal endoscopy can be performed safety and efficiency with either moderate or deep sedation. For the patient with unstable cardiopulmonary status, moderate sedation might be a better choice instead of deep sedation.