Rothbaum 2012.

Methods	Design: RCT
Participants	Setting: participants were recruited from a hospital trauma centre. Type of trauma exposure: patients admitted to the emergency department experiencing a DSM‐IV criterion A trauma. Inclusion criteria: adults aged 18–65 years who had experienced a traumatic event within 72 hours of presentation in the emergency department; afraid that they might be killed or seriously injured during the event; able to be contacted following discharge and to return for follow‐up appointments; and alert and oriented and able to provide informed consent. Exclusion criteria: non‐English speaking; lost consciousness > 5 min during the event; history of a serious mental illness; currently suicidal or reported current substance dependence. Sample size: 5608 individuals assessed for eligibility; 4219 did not meet inclusion criteria; 1249 refused to take part or were excluded for other reasons; 137 randomised. Mean age: 31.5 (SD = 11.6) years Gender: 48 (35.0%) men; 89 (65.0%) women Ethnicity: 13.1% black; 78.8% white; 1.5% Native American; 6.6% other Country: USA
Interventions	Group 1: modified PE: n = 69 Patients received 3 × 1‐hour sessions of a modified PE intervention, distributed 1 week apart. PE consisted of imaginal exposure, processing time, breathing retraining, psychoeducation and homework tasks (including addressing in vivo exposure goals). Group 2: assessment only: n = 68
Outcomes	PTSD: PTSD PSS‐I; PDS Other: BDI‐II, Additional Treatment Inventory; Standardized Trauma Interview Follow‐up: 4 and 12 weeks
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Envelopes containing computer‐generated patient random assignments (either to immediate intervention or assessment only) were given to the patient and their nurse after the initial evaluation to ensure that assessors remained blind.
Allocation concealment (selection bias)	Low risk	Envelopes containing computer‐generated patient random assignments (either to immediate intervention or assessment only) were given to the patient and their nurse after the initial evaluation to ensure that assessors remained blind.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessors were blinded.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Missing data handled with multiple imputation with mixed‐effect models being used to obtain predicted mean values for outcomes at each assessment point. There was a disproportionately higher level of missing data in the intervention group at initial outcome (31.9% compared to 19.1%) and 12 weeks' follow‐up (39.1% compared to 27.9%).
Selective reporting (reporting bias)	Unclear risk	Outcomes were not reported in study protocol.
Other bias	High risk	Authors affiliated with experimental intervention.