Background: The aim of this study was to investigate trends in the practice of selective non-operative management (SNOM) for penetrating abdominal injury (PAI) and to determine factors associated with its failure. Methods: The National Trauma Data Bank for 2002-2008 was reviewed. Patients with PAI were categorized as those who underwent successful SNOM (operative management not required) and those who failed SNOM (surgery required more than 4 h after admission). Yearly rates of SNOM versus non-therapeutic laparotomy (NTL) were plotted. Multivariable regression analysis was performed to identify factors associated with failed SNOM and mortality. Results: A total of 12 707 patients with abdominal gunshot and 13 030 with stab wounds were identified. Rates of SNOM were 22·2 per cent for gunshot and 33·9 per cent for stab wounds, and increased with time (P < 0·001). There was a strong correlation between the rise in SNOM and the decline in NTL (r = - 0·70). SNOM failed in 20·8 and 15·2 per cent of patients with gunshot and stab wounds respectively. Factors predicting failure included the need for blood transfusion (odds ratio (OR) 1·96, 95 per cent confidence interval 1·11 to 3·46) and a higher injury score. Failed SNOM was independently associated with mortality in both the gunshot (OR 4·48, 2·07 to 9·70) and stab (OR 9·83, 3·44 to 28·00) wound groups. Conclusion: The practice of SNOM is increasing, with an associated decrease in the rate of NTL for PAI. In most instances SNOM is successful; however, its failure is associated with increased mortality. Careful patient selection and adherence to protocols designed to decrease the failure rate of SNOM are recommended.