Self-Removal of Medical Devices in the ICU: A Retrospective Study.

Introduction Self-removal of medical devices (SRMD) is common in the intensive care unit (ICU). Most studies of this issue concentrate on self-extubation, leaving self-removal of other devices less well studied. Methods This is a retrospective chart review utilizing the MIMIC III database. Free-text notes were examined for reports of patients removing medical devices. Predictive factors and the outcome of mortality were analyzed. Univariate analysis of categorical variables was performed using chi-square and continuous variables using the t-test. Multivariate analysis was performed with logistic regression. Covariates were gender, age, tobacco abuse, delirium, alcohol abuse, psychiatric history, drug abuse, dementia, and brain trauma. Results Overall, 5.3% of ICU patients pulled at least one device. The number of devices pulled per 1,000 ICU days was 21.5. The devices most pulled were nasogastric tubes (11.87), peripheral intravenous catheters (3.1), and Foley catheters (2.0). Death was more likely for patients who removed devices: 56% versus 40% (p<0.001). SRMD was an independent risk factor for death, with an adjusted odds ratio (OR) of 1.9 (p<0.001). On univariate analysis, a history of delirium was the most substantial predictive factor at 16.3%, followed by a history of alcohol withdrawal (16.2%) and alcohol abuse (9.9%), p<0.001 for all comparisons. On multivariate analysis, history of delirium remained the most decisive independent risk factor with an OR of 3.15, followed by a history of alcohol withdrawal (OR 2.0) and a history of drug abuse (OR 1.7), p<0.001 for all comparisons. Conclusions SRMD occurs with clinically important frequency in the ICU. Nasogastric tubes are poorly tolerated. SRMD is an independent predictor of ICU mortality. Delirium plays a central role in SRMD, with alcohol use and drug use as additional essential factors. This suggests a focus on delirium control as the best way to prevent SRMD and improve patient safety.