OBJECTIVES: To model clinical and situational variables that may affect likelihood of physicians to order physical restraints.
DESIGN: Cross-sectional, factorial survey.
SETTING: One academic medical center.
PARTICIPANTS: One hundred eighty-nine physicians: interns in all specialty practices and resident and attending physicians in departments of surgery, general internal medicine, family practice, emergency medicine, and psychiatry.
MEASUREMENTS: Vignettes were randomly generated using different values of six situational and eight clinical variables. Each physician received five unique vignettes for which they indicated their likelihood to order restraint on a 10-point scale.
RESULTS: Nine hundred six distinct vignettes were completed. The mean likelihood that physicians would order restraint was 3.9±3.0 (range 0 (not at all) to 9 (absolutely)). Exploratory regression analysis on physician's likelihood to restrain with independent variables of secondary diagnosis, patient age, sex, time of day, familiarity and trust with requesting nurse, patient behavior, vital signs, oxygen saturation, and dehydration explained 12.5% of variance (F=5.43, P<.001). Independent factors of unsafe patient behavior (P=.001) and secondary diagnosis of dementia (P=.06) resulted in greater likelihood of ordering restraint, whereas lack of trust in the judgment of the reporting nurse (P=.008) resulted in lower likelihood of ordering restraints.
CONCLUSION: Patients' clinical status had less influence on physicians' likelihood of ordering physical restraints than the working relationship with the requesting nurse or the patient's behavior. Interdisciplinary team approaches with active physician input for nonrestraint strategies in the management of patient behavior is emphasized to minimize restraint use.