Stepped Collaborative Care Targeting Posttraumatic Stress Disorder Symptoms and Comorbidity for US Trauma Care Systems

Key Points

Question

Can a brief stepped collaborative care intervention for injured patients at a trauma center delivered by front-line clinicians reduce posttraumatic stress disorder symptoms compared with usual care?

Findings

In a randomized clinical trial with 635 injured patients from 25 US trauma centers, intervention patients demonstrated significant posttraumatic stress disorder symptom reductions compared with those who received usual care at 6 months, but not 12 months, postinjury. Subgroup analyses revealed larger posttraumatic stress disorder treatment effects at trauma centers with good or excellent protocol implementation.

Meaning

In this study, a well-implemented brief intervention for injured patients reduced posttraumatic stress disorder symptoms; policy efforts should incorporate these findings into national trauma center requirements and verification criteria.

Abstract

Importance

To date, few multisite investigations have evaluated early interventions for injured patients with posttraumatic stress disorder (PTSD) symptoms.

Objective

To simultaneously assess the effectiveness and implementation of a brief stepped collaborative care intervention targeting PTSD and comorbidity.

Design, Setting, and Participants

A stepped-wedge cluster randomized clinical trial was conducted at 25 US level I trauma centers. Participants included hospitalized survivors of physical injury who underwent a 2-step evaluation for PTSD symptoms. Patients reporting high levels of distress on the PTSD Checklist (PCL-C) were randomized (N = 635) per the stepped-wedge protocol to enhanced usual care control (n = 370) or intervention (n = 265) conditions. The study was conducted from January 4, 2016, through November 2018. Data analysis was performed from November 4, 2019, to December 8, 2020.

Interventions

The Trauma Survivors Outcomes and Support collaborative care intervention included proactive injury case management that assisted patients transitioning from hospital inpatient to outpatient and community settings. The intervention also integrated evidence-based pharmacotherapy and psychotherapeutic elements targeting PTSD symptoms and comorbidity.

Main Outcomes and Measures

The primary study outcome was PTSD symptoms assessed with the PCL-C at baseline in the surgical ward and at 3, 6, and 12 months postinjury. Secondary outcomes included depressive symptoms, alcohol use, and physical function. Subgroup analyses examined the effect of baseline risk factors for enduring PTSD and quality of protocol implementation on study outcomes. Primary statistical analyses were conducted using the intent-to-treat sample.

Results

A total of 327 men (51.5%) were included in analysis; mean (SD) age was 39.0 (14.2) years. The investigation attained follow-up of 75% to 80% of the participants at 3 to 12 months. The intervention lasted a mean (SD) of 122 (132) minutes. Mixed model regression analyses revealed statistically significant changes in PCL-C scores for intervention patients compared with control patients at 6 months (difference, −2.57; 95% CI, −5.12 to −0.03; effect size, 0.18; P < .05) but not 12 months (difference, −1.27; 95% CI, −4.26 to 1.73; effect size, 0.08; P = .35). Subgroup analyses revealed larger PTSD treatment effects for patients with 3 or more baseline risk factors for enduring PTSD and for patients, including firearm injury survivors, treated at trauma centers with good or excellent protocol implementation. Intervention effects for secondary outcomes did not attain statistical significance.

Conclusions and Relevance

A brief stepped collaborative care intervention was associated with significant 6-month but not 12-month PTSD symptom reductions. Greater baseline PTSD risk and good or excellent trauma center protocol implementation were associated with larger PTSD treatment effects. Orchestrated efforts targeting policy and funding should systematically incorporate the study findings into national trauma center requirements and verification criteria.

Trial Registration

ClinicalTrials.gov Identifier: NCT02655354

This randomized clinical trial examines the use of a brief stepped collaborative care intervention in patients who have developed posttraumatic stress disorder as a result of traumatic physical injury.

Introduction

Annually, between 2.5 million and 3.0 million individuals in the US are so severely injured that they require inpatient hospital admission.¹ The symptoms of posttraumatic stress disorder (PTSD) and related comorbidity occur frequently after traumatic physical injury.^2,3,4,5,6 Traumatically injured patients commonly present with associated depressive symptoms, alcohol and drug use disorders, and other comorbidities.^3,7,8,9,10 Prospective studies suggest that comorbid conditions and other factors, such as preinjury trauma exposure and intentional injury mechanisms (eg, firearm injury), may act as risk factors that contribute to an enduring course of PTSD symptoms.^{11,12,13,14,15,16,17} After injury, PTSD and related comorbidities are associated with a broad profile of functional and health-related impairments and associated societal costs.^{7,18,19,20,21,22}

Efficacy and effectiveness spectrum investigations have documented that early psychotherapeutic and psychopharmacologic interventions targeting the symptoms of PTSD can be effective.^{6,17,23,24,25,26,27,28,29} Some acute care medical investigations that have tested broad reach interventions with more generalizable samples have yielded mixed results.^{30,31,32,33,34,35} Comprehensive disease management strategies, such as collaborative care, that combine proactive case management, pharmacotherapy targeting PTSD, and behavioral intervention elements embedded within injury care management have demonstrated effectiveness in reducing PTSD symptoms while also broadening intervention reach.^25,36,37,38

Pragmatic clinical trials aim to recruit generalizable samples of patients in real-world settings to influence policy decisions regarding the adoption of health care practices.^39,40,41 Over the past decade, the Trauma Survivors Outcomes and Support team has established a stakeholder partnership with the American College of Surgeons Committee on Trauma (ACS/COT), allowing the results of pragmatic comparative effectiveness trials to be directly translated into policy for US trauma care systems.^{42,43,44,45,46} Previous investigations suggest that less than 10% of US level I and II trauma centers routinely provide postinjury screening or integrated treatment targeting PTSD.⁴⁷ The 2014 ACS/COT guidelines suggest PTSD screening, intervention, and referral as a best practice, but unlike mandates for alcohol, do not require that trauma centers implement these practices as part of verification criteria.⁴³

By necessity, pragmatic clinical trials may incorporate methods derived from implementation science approaches to both advance the sustainable delivery of screening and intervention procedures and document shifting health care system contexts.^{48,49,50,51,52} The events of 2020, including the coronavirus disease 2019 pandemic^53,54,55 and increases in firearm sales and firearm injuries,^{56,57,58,59,60,61} provide examples of the shifting context of life-threatening exposures for US patient populations. The effect of these developments on patients reinforces the need to improve screening and intervention procedures targeting PTSD and related conditions for patients presenting to acute care medical settings.

This pragmatic comparative effectiveness trial hypothesized that injured patients receiving a brief stepped collaborative care intervention would demonstrate significant reductions in PTSD symptoms compared with control patients receiving enhanced usual care. A secondary hypothesis was that the intervention could effectively reduce depressive symptoms, alcohol use, and postinjury functional impairments. An additional issue was whether baseline surgical ward risk factors for enduring PTSD symptoms affected intervention treatment effects. Individual sites’ ability to initiate the intervention protocol was evaluated via an implementation process assessment, and the investigation explored whether the quality of protocol implementation affected any observed PTSD treatment effects.

Methods

Design

The Trauma Survivors Outcomes and Support pragmatic trial orchestration was carried out at University of Washington’s Harborview Medical Center, in close collaboration with the National Institutes of Health Care Systems Research Collaboratory.⁶² The Western Institutional Review Board approved the protocol before study initiation.⁶³ All patients provided written informed consent prior to protocol participation and received financial compensation.⁵⁰ This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials.

Investigative procedures are detailed in the study protocol (Supplement 1) and are briefly described here. Sites recruited into the study constituted a representative subsample of all US level I trauma centers.⁵⁰ The investigation used a stepped-wedge cluster randomized design, and sample size estimates that include adjustments for the clustering of individual observations within sites have been previously described (Supplement 1).⁵⁰ All sites began the protocol recruiting control patients and were randomized sequentially to initiate the intervention. Patients were assessed at baseline in the surgical ward as trauma inpatients and again 3, 6, and 12 months after the injury. Recruitment for the trial began January 4, 2016, and 12-month patient study follow-up ended November 2018. Data analysis was performed from November 4, 2019, to December 8, 2020.

Injured patients aged 18 years or older were included in the trial. To ensure adequate follow-up rates, patients were required to provide 2 sets of contact information.

The study team had previously developed a 10-domain electronic health record screen to detect patients with a high likelihood of PTSD distress as hospitalized inpatients (Supplement 1).⁶⁴ Patients identified by the screen as likely to have high levels of distress, indicated by a score of 3 or more positive electronic health record domains, were then formally screened for study entry with the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV)⁶⁵ PTSD Checklist (PCL-C), anchored to the acute injury event.^37,38,66 Patients with high levels of distress, as indicated by a score 35 or higher on the PCL-C, were randomized into the trial (Figure 1).

Figure 1. Flow of Trauma Center Patients Through the Trial.

In the stepped-wedge cluster randomized design, all sites initiate recruitment of control patients and then sequentially, per the randomization scheme, begin intervention patient recruitment; thus, all sites recruit both control and intervention patients. PTSD indicates posttraumatic stress disorder.

Before patient recruitment, each of the 25 sites was randomized electronically by a study biostatistician to 1 of 4 waves in the stepped-wedge design. Recruiters at the 25 sites were aware of each patient’s intervention or control group status at the time of the baseline surgical ward interview but were instructed not to inform patients of their status until after completion of the baseline interview. All follow-up interviewers were blinded to patient intervention or control group status.

Patients in the control group underwent informed consent, both PTSD screenings, baseline surgical ward evaluation, and follow-up interviews. The enhanced aspect of usual care consisted of nurse notification of each patient whose score was 35 or higher on the PCL-C. Previous investigation suggests that, after hospital discharge, usual posttraumatic care includes routine surgical, primary care, and emergency department visits, as well as some mental health specialty appointments.^36,37,38

Stepped Collaborative Care Intervention

Acute care medical stepped collaborative care protocols targeting PTSD and related comorbidity have been described previously (Supplement 1).^36,37,38 Briefly, stepped collaborative care treatments bring together evidence-based medication and psychotherapeutic intervention elements with proactive case management strategies that aim to reduce care fragmentation for injured patients. Patients who demonstrate enduring PTSD symptoms after initial treatment receive stepped-up care that can include medication adjustments or the addition of new psychotherapeutic elements.

After completion of the usual care control phase of recruitment, the principal investigator (D.Z.) visited each trauma center in order to train front-line social workers, nurses, physicians, and other health care professionals via a 1-day, onsite intervention workshop.^50,67 The workshop provided an overview of the core case management, psychopharmacology, and motivational interviewing and cognitive behavioral therapy elements of the stepped collaborative care intervention. After the 1-day workshop, the study team initiated regular site supervisory calls in which the site interventionists presented cases to supervising psychiatric, research coordinator, and other study team members. The supervisory calls also addressed protocol issues related to recruitment, regulatory compliance, and staff turnover. The 25 sites’ intervention and staffing activities were documented in the REDCap database.⁶⁸

Patient-Reported Outcome Assessments

The PCL-C was used to assess the symptoms of PTSD. At baseline in the surgical ward, patients were asked to rate their symptoms since the injury event; the 3-, 6- and 12-month interviews queried patients about their symptoms over the month prior to the interview. Prior investigations, including studies by the research team with injured patients, have established the psychometric equivalence of the DSM-IV and DSM-5 versions of the PCL.^{65,69,70,71,72,73} Because the study team had validated the 10-domain EHR screen with the PCL-C for DSM-IV and DSM-IV CAPS,⁶⁴ all primary surgical ward and follow-up assessments were performed with the DSM-IV version of the PCL (PCL-C).^69,74

With regard to secondary outcomes, the 9-item Patient Health Questionnaire (PHQ-9) brief depression severity measure was used to assess depressive symptoms.⁷⁵ The Alcohol Use Disorder Identification Test 3-item version (AUDIT-C) was used to assess alcohol use problems before and after the injury hospitalization.^37,76,77 The investigation used the Medical Outcomes Study Short Form Physical Components Summary Score (MOS SF PCS) SF-12 at baseline in the surgical ward to assess physical function in the month before the injury admission, and the SF-36 at 3-, 6-, and 12-month follow-up to assess postinjury physical function.^37,78,79

Medical record data from the 25 sites’ trauma registries were used to derive injury severity scores and injury mechanisms.^80,81,82,83 Laboratory toxicologic test results, insurance status, length of hospital and intensive care unit stays, and other clinical characteristics were obtained from trauma registries.

Implementation Process Assessment

Over the course of the multiyear study, the Rapid Assessment Procedure Informed Clinical Ethnography (RAPICE) approach was used to assess the processes of protocol implementation.^44,50 RAPICE used data from multiple sources, including study team intervention documentation, and recruitment, regulatory, and other clinical trial logs. The RAPICE approach also embedded participant observation within clinician investigators immersed in the pragmatic trial implementation coupled with regular mixed method consultation.⁴⁴

Over the course of the trial, 4 core domains of the implementation process were identified: site attainment of recruitment milestones, quality of intervention implementation, leadership stability and turnover, and regulatory adherence. A coding framework was developed, and sites were grouped by fair or poor vs good or excellent implementation status based on their scores across the four domains (eTable 1 in Supplement 2).

Statistical Analysis

All primary statistical analyses were conducted using the intent-to-treat sample. The first goal of the statistical analysis was to examine and compare patterns over time in the symptoms of PTSD. To determine whether patients in the intervention and control groups manifested different patterns of change in PTSD symptoms over the year after injury, the study team used mixed-effects regression models.^84,85,86 Mixed-effects regression was used to account for both repeated measures for individuals over time (ie, baseline, 3-month, 6-month, and 12-month time points) and site-level cluster randomization. This analytic approach was replicated for all secondary outcomes, including the PHQ-9, AUDIT, and SF MOS PCS scores. For all dependent variables, models were fit containing time categories, intervention, and intervention by time interactions, as well as stepped-wedge period and site. Baseline characteristics identified to be significantly different across the 2 groups were included as design variables in regression analyses. In addition, to better understand clinically relevant PTSD symptom reductions over time, the study team compared intervention and control group 10-point PCL-C symptom reductions.³⁸ The study team conducted sensitivity analyses to assess the impact of removing baseline covariates, as well as including the DSM-5 version of the PCL as the dependent variable.

Both a priori and exploratory secondary analyses were performed. Baseline surgical ward risk factors were identified that were associated with higher levels of enduring PTSD symptoms; a priori secondary analyses stratified the intent-to-treat sample to identify risk groups that manifested differential treatment effects. Post hoc exploratory analyses examined the influence of quality of implementation and firearm injury on observed treatment effects. Statistical significance was set at 2-sided P < .05. The study team used SAS, version 9.4 (SAS Institute Inc) and SPSS version 25 (SPSS Software IBM) for all analyses.

Results

The 635 injured participants recruited and randomized into the investigation were diverse patients with substantial histories of preinjury trauma (Table 1). The population included 308 women (48.5%) and 327 men (51.5%). Mean (SD) age was 39.0 (14.2) years. Three patient characteristics across the usual care (n = 370) vs intervention (n = 265) groups were observed to be significantly different at baseline across the 2 groups: age (mean [SD], 39.9 [14.8] vs 37.6 [13.4] years; P = .047), sex (female, 161 [43.5%] vs 147 [55.5%]; P = .003), and electronic health record PTSD diagnosis (55 [14.9%] vs 60 [22.6%]; P = .01) .

Table 1. Baseline Patient Characteristics.

Characteristic	Patients, No. (%)
	All (N = 635)	Usual care control (n = 370)	Intervention (n = 265)
Electronic heath record variables
Sexa
Female	308 (48.5)	161 (43.5)	147 (55.5)
Male	327 (51.5)	209 (56.5)	118 (44.5)
ICU admission	377 (59.4)	230 (62.2)	147 (55.5)
Prior inpatient hospitalization	249 (39.2)	157 (42.4)	92 (34.7)
Tobacco use	356 (56.1)	211 (57.0)	145 (54.7)
Psychiatric diagnosis	246 (38.7)	141 (38.1)	105 (39.6)
PTSD diagnosisb	115 (18.1)	55 (14.9)	60 (22.6)
Positive blood alcohol concentration	157 (28.4)	86 (29.1)	71 (27.7)
Demographic
Age, mean (SD), yc	39.0 (14.2)	39.9 (14.8)	37.6 (13.4)
Raced
White	315 (49.6)	188 (50.8)	127 (47.9)
Black	218 (34.3)	129 (34.9)	89 (33.6)
American Indian	15 (2.4)	9 (2.4)	6 (2.3)
Asian	4 (0.6)	2 (0.5)	2 (0.8)
Pacific Islander	4 (0.6)	1 (0.3)	3 (1.1)
Mixed	20 (3.1)	9 (2.4)	11 (4.2)
Other	59 (9.3)	32 (8.7)	27 (10.2)
Hispanicd
Yes	102 (16.1)	52 (14.1)	50 (18.9)
No	530 (83.9)	316 (85.9)	214 (81.1)
Educational level
<High school	123 (19.5)	74 (20.1)	49 (18.6)
High school/GED	314 (49.7)	181 (49.2)	133 (50.4)
Associate degree	118 (18.7)	66 (17.9)	52 (19.7)
Bachelor’s or graduate degree	77 (12.2)	47 (12.8)	30 (11.4)
Marital status, married/living with partner	178 (28.0)	99 (26.8)	79 (29.8)
Employed	376 (59.6)	208 (56.4)	168 (64.1)
Insurance
Private	224 (35.4)	131 (35.5)	93 (35.4)
Public	329 (52.1)	199 (53.9)	130 (49.4)
None	79 (12.5)	39 (10.6)	40 (15.2)
Acute care injury and medical factors
Intentional injury
Assault	53 (8.3)	26 (7.0)	27 (10.2)
Stabbing	53 (8.3)	31 (8.4)	22 (8.3)
Firearm	128 (20.2)	79 (21.4)	49 (18.5)
Unintentional injury	401 (63.2)	234 (63.2)	167 (63.0)
Injury severity categorye
0-8	137 (24.1)	74 (23.3)	63 (25.1)
9-15	197 (34.7)	109 (34.4)	88 (35.1)
≥16	234 (41.2)	134 (42.3)	100 (39.8)
Traumatic brain injuryf
None	388 (68.3)	208 (65.6)	180 (71.7)
Mild	101 (17.8)	58 (18.3)	43 (17.1)
Moderate	56 (9.9)	33 (10.4)	23 (9.2)
Severe	23 (4.0)	18 (5.7)	5 (2.0)
No. of comorbid medical conditions
0	192 (31.3)	110 (31.3)	82 (31.4)
1	117 (19.1)	68 (19.3)	49 (18.8)
2	91 (14.8)	52 (14.8)	39 (14.9)
≥3	213 (34.7)	122 (34.7)	91 (34.9)
Length of stay for index visit, mean (SD), days	12.9 (12.7)	13.1 (12.3)	12.6 (13.2)
Clinical assessments
No. of previous serious traumas before injury admission, mean (SD)g	4.5 (3.1)	4.4 (3.1)	4.5 (3.2)
Preinjury PTSD symptoms lasting ≥30 dg	219 (34.5)	125 (33.8)	94 (35.5)
Baseline PCL-C total score, mean (SD)h,i	52.1 (11.9)	50.7 (11.2)	54.0 (12.6)
Baseline PHQ-9 depression total score, mean (SD)h	14.1 (5.8)	13.9 (5.8)	14.3 (5.8)
PHQ-9 Item 9 suicide positivej	157 (24.8)	90 (24.4)	67 (25.3)
Preinjury AUDIT-C score, mean (SD)	3.7 (3.4)	3.7 (3.3)	3.7 (3.5)
Preinjury self-report drug usek
Stimulantsl	135 (21.4)	77 (20.9)	58 (22.1)
Opioids	62 (9.8)	44 (11.9)	18 (6.8)
Marijuana	302 (47.7)	177 (47.8)	125 (47.5)
Preinjury SF-12 PCS score, mean (SD)	49.4 (9.7)	49.1 (9.7)	49.8 (9.8)
Surgical admission risk factors for enduring PTSDm
<3n	424 (66.8)	259 (70.0)	165 (62.3)
≥3n	211 (33.2)	111 (30.0)	100 (37.7)

Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test–Consumption Items⁷⁷; GED, general educational development; ICU, intensive care unit; PCL-C, PTSD Checklist Civilian Version⁶⁶; PHQ, Patient Health Questionnaire⁷⁵; PTSD, posttraumatic stress disorder; SF-12 PCS, Short-Form Health Survey–Physical Component Summary.⁷⁹

^aP = .003.

^bP = .01.

^cP = .047.

^dPatients were asked to self-identify racial/ethnic group classifications as provided by the investigators; 1 study participant was missing race data and was included in the other category and 3 study participants were missing ethnicity data.

^eFor injury severity, the categories are mild (0-8), moderate (9-15), and severe (>16).^80,81

^fTraumatic brain injury severity was coded on the basis of a previously validated algorithm for hospitalized inpatients that assigned Maximum Abbreviated Injury Scale head injury scores of 1 to 2 for mild, 3 for moderate, and 4 or higher for severe.⁸³

^gDerived from the 3-month interview Trauma History Screen; this value may underestimate the rates of previous PTSD symptoms because patients who were missing data were included in the no PTSD symptom group.

^hFor PCL-C and PHQ-9 baseline assessments, inpatients were asked to report symptoms since the injury event. A score on the PCL-C of greater than or equal to 35 indicates high levels of posttraumatic distress. A score on the PHQ-9 of greater than or equal to 10 indicates high levels of postinjury depressive symptoms.

ⁱP = .001.

^jPHQ-9 Item 9 suicide-positive is a score of 1 or greater.

^kSingle item self-report dichotomized as none vs at least monthly use.

^lStimulants include cocaine and amphetamines.

^mRisk factors for enduring PTSD symptoms include (1) 5 or more preinjury traumas, (2) PTSD diagnosis from inpatient electronic health record screen, (3) non-White race from inpatient electronic health record, (4) 3 or more medical comorbidities, (5) use of stimulants, and (6) intentional injury.

ⁿP = .04.

The investigation attained follow-up rates of 80.2% at 3 months, 77.3% at 6 months, and 75.1% at 12 months (Figure 1). Follow-up rates did not substantially differ for patients assigned to the control and intervention groups over the course of 12 months.

Intervention Time, Activities, and Processes of Care

Site intervention teams spent a mean (SD) of 122 (132) minutes (median, 81 minutes; range, 3-890 minutes; interquartile range, 45-147.5 minutes) with each patient over the course of the year after injury. Approximately 70% of intervention activity occurred during the first 3 months postinjury, 20% occurred between months 3 and 6, and 10% occurred between months 6 and 12.

Overall, at the 3-, 6-, and 12-month outcome assessment points, a greater percentage of intervention patients compared with control patients reported receiving antidepressant medication, stepped adjustments to medication dosages, and/or counseling visits (Table 2). These comparisons, however, did not achieve statistical significance. Equivalent numbers of intervention and control patients were treated at fair or poor vs good or excellent implementation sites. Additional care processes for the intervention group are described in eTable 2 in Supplement 2.

Table 2. Processes of Care for Posttraumatic Stress Disorder and Comorbidity.

Variable	No. (%)		Control vs intervention, relative risk (95% CI)
	Usual care control	Intervention
Any antidepressant medicationa
At baseline	67 (18.1)	56 (21.1)	1 [Reference]
At 3 mo	71 (24.4)	70 (32.0)	1.12 (0.82-1.52)
At 6 mo	79 (27.3)	71 (35.0)	1.06 (0.78-1.43)
At 12 mo	76 (26.2)	64 (34.0)	1.07 (0.80-1.43)
Antidepressant medication stepped up or downb
At 3 mo	42 (64.6)	46 (71.9)	1 [Reference]
At 6 mo	26 (36.1)	28 (44.4)	1.14 (0.81-1.60)
At 12 mo	22 (31.0)	26 (42.6)	1.28 (0.84-1.95)
Any counselinga
At baseline	98 (31.1)	84 (37.5)	1 [Reference]
At 3 mo	101 (35.3)	101 (47.4)	1.10 (0.84-1.44)
At 6 mo	109 (38.0)	102 (51.0)	1.11 (0.84-1.47)
At 12 mo	125 (43.4)	90 (49.2)	0.93 (0.74-1.16)
Counseling visits with discrete elementsa,c
At 3 mo	105 (36.6)	93 (43.7)	1 [Reference]
At 6 mo	118 (41.1)	97 (49.0)	1.00 (0.83-1.20)
At 12 mo	123 (42.4)	85 (45.5)	0.90 (0.69-1.17)
Implementation process assessmentd
Patients at fair or poor implementation sites	156 (42.2)	108 (40.8)	1 [Reference]
Patients at good or excellent implementation sites	214 (57.8)	157 (59.2)	1.00 (0.97-1.04)

^aThe sample size and follow-up rate for each study time point is as follows: baseline: N = 635 (100%), 3 months: n = 509 (80.2%), 6 months: n = 492 (77.5%), and 12 months: n = 477 (75.1%).

^bCalculated as a percentage of those who were taking any antidepressant at a time point.

^cThis variable is dichotomized as yes or no across all 9 counseling domains, including (1) anxiety, (2) depression, (3) traumatic brain injury, (4) suicidal ideation, (5) sleep disturbance/insomnia, (6) pain, (7) irritability, (8) alcohol use, and (9) drug use.

^deTable 1 in Supplement 2 provides the site implementation rating criteria.

The study sites received a mean (SD) of 27 (11) supervisory sessions during the course of the intervention implementation. Sites randomized to earlier stepped-wedge intervention waves were significantly more likely to receive more supervisory sessions (F_3,543 = 8.0; P < .001). There were no significant differences in the number of supervisory sessions received by sites with fair or poor vs good or excellent implementation (F_1,17 = 0.9; P < .36).

PTSD and Other Outcomes

Mixed model regression analyses revealed statistically significant changes in PCL-C scores for intervention patients compared with controls at 6 months, but not 3 or 12 months, after the index injury admission (Table 3, Figure 2A). Compared with control patients, intervention patients demonstrated approximately 5% greater 10-point PCL-C reductions at each follow-up time point (eTable 3 in Supplement 2). Patients with 3 or more baseline PTSD risk factors for higher levels of enduring PTSD symptoms manifested larger treatment effects (Table 3, Figure 2B). Patients treated at sites with good or excellent implementation also demonstrated larger treatment effects (Table 3, Figure 2C). There were no significant baseline PCL-C differences for intervention and control group patients treated at sites with good or excellent protocol implementation (net difference mean, 3.34; 95% CI, −1.39 to 8.07; P = .15) (Figure 2C).

Table 3. Change Over Time in PTSD Checklist Symptom Levels for Patients in Intervention vs Usual Care Control Conditions^a.

Group	No. (%)b	Change baseline to 3 mo				Change baseline to 6 mo				Change baseline to 12 mo
		Mean (95% CI)		Effect size	Mean (95% CI)			Effect size	Mean (95% CI)			Effect size
		Change	Net difference		Change		Net difference		Change		Net difference
Full sample
Usual care	370 (58.3)	0.08 (−1.38 to 1.53)	1 [Reference]	1 [Reference]	−1.44 (−3.08 to 0.19)		1 [Reference]	1 [Reference]	−4.25 (−6.14 to −2.35) (P < .001)		1 [Reference]	1 [Reference]
Collaborative care	265 (41.7)	−1.65 (−3.34 to 0.04)	−1.73 (−3.96 to 0.50)	0.14	−4.02 (−5.97 to −2.07) (P < .001)		−2.57 (−5.12 to −0.03) (P = .047)	0.18	−5.51 (−7.83 to −3.2) (P < .001)		−1.27 (−4.26 to 1.73)	0.08
Baseline risk for enduring PTSD stratification
<3 Risk factors
Usual care	259 (40.8)	−2.93 (−4.68 to −1.19) (P = .001)	1 [Reference]	1 [Reference]	−3.84 (−5.77 to −1.91) (P < .001)		1 [Reference]	1 [Reference]	−6.32 (−8.51 to −4.13) (P < .001)		1 [Reference]	1 [Reference]
Collaborative care	165 (26.0)	−3.45 (−5.61 to −1.28) (P = .002)	−0.51 (−3.29 to 2.27)	0.04	−4.10 (−6.56 to −1.65) (P = .001)		−0.27 (−3.39 to 2.86)	0.02	−5.56 (−8.40 to −2.71) (P < .001)		0.76 (−2.83 to 4.36)	0.05
≥3 Risk factors
Usual care	111 (17.5)	5.93 (3.42 to 8.44) (P < .001)	1 [Reference]	1 [Reference]	3.19 (0.26 to 6.12) (P = .03)		1 [Reference]	1 [Reference]	−0.18 (−3.71 to 3.34)		1 [Reference]	1 [Reference]
Collaborative care	100 (15.7)	0.59 (−2.03 to 3.21)	−5.34 (−8.96 to −1.71) (P = .004)	0.42	−4.18 (−7.31 to −1.06) (P = .009)		−7.38 (−11.66 to −3.09) (P = .001)	0.51	−5.70 (−9.56 to −1.85) (P = .004)		−5.52 (−10.75 to −0.29) (P = .04)	0.32
Quality of implementation stratification
Fair or poor implementation
Usual care	156 (24.6)	−0.01 (−2.27 to 2.26)	1 [Reference]	1 [Reference]	−0.59 (−3.13 to 1.96)		1 [Reference]	1 [Reference]	−4.77 (−7.72 to −1.82) (P = .002)		1 [Reference]	1 [Reference]
Collaborative care	108 (17.0)	1.30 (−1.40 to 4.01)	1.31 (−2.21 to 4.83)	0.10	−0.55 (−3.62 to 2.53)		0.04 (−3.95 to 4.03)	0.00	−1.85 (−5.46 to 1.76)		2.93 (−1.73 to 7.59)	0.18
Good or excellent implementation
Usual care	214 (33.7)	0.16 (−1.73 to 2.05)	1 [Reference]	1 [Reference]	−2.03 (−4.15 to 0.09)		1 [Reference]	1 [Reference]	−3.86 (−6.33 to −1.39) (P = .002)		1 [Reference]	1 [Reference]
Collaborative care	157 (24.7)	−3.68 (−5.84 to −1.52) (P = .001)	−3.84 (−6.71 to −0.98) (P = .009)	0.31	−6.44 (−8.95 to −3.93) (P < .001)		−4.41 (−7.70 to −1.12) (P = .009)	0.31	−8.09 (−11.10 to −5.09) (P < .001)		−4.23 (−8.12 to −0.34) (P = .03)	0.26

Abbreviation: PTSD, posttraumatic stress disorder.

^aP values shown for significant findings; all other comparisons were not statistically significant. All regressions were adjusted for baseline covariate imbalances including age, sex, and index admission PTSD diagnosis.

^bAll percentages are relative to 635.

Figure 2. Posttraumatic Stress Disorder (PTSD) Symptom Levels Over Time.

PTSD symptom levels in all participants (A), stratified by surgical ward risk factors for enduring PTSD (B), and stratified by quality of trauma center implementation (C). PCL-C IV indicates PTSD Checklist-Civilian Version DSM IV (Diagnostic and Statistical Manual of Mental Disorders [Fourth Edition]).

Patients with firearm injuries treated at good or excellent implementation sites had among the largest 6-month (net difference mean, −7.81; 95% CI, −15.61 to 0.00; effect size, 0.52) and 12-month (net difference mean, −10.37; 95% CI, −19.16 to −1.59; effect size, 0.61) treatment effects (eFigure in Supplement 2). The PHQ-9, AUDIT-C, and SF-36 PCS outcome comparisons did not attain statistical significance. Sensitivity analyses that included the DSM-5 PCL and removed baseline covariates did not substantially alter the magnitude, pattern, or directionality of the observed treatment effects.

Discussion

The pragmatic trial described herein documents the effectiveness of a brief stepped collaborative care intervention, delivered by front-line trauma center clinicians, in reducing PTSD symptoms at 6 months, but not 12 months, postinjury. This pattern of treatment response can be explained in part by the observation that approximately 90% of intervention activity occurred in the first 6 months after the injury. Larger PTSD treatment effects were observed for the subgroups of injured patients with 3 or more baseline risk factors for higher levels of enduring PTSD symptoms and for the subgroup of patients who were treated at trauma center sites with good or excellent protocol implementation. Firearm injury survivors treated at trauma centers with good or excellent protocol implementation also demonstrated significant PTSD treatment effects at the 6- and 12-month postinjury time points.

The results of this 25-site pragmatic trial corroborate and extend previous reports demonstrating that stepped collaborative care may be an optimal intervention approach for injury survivors treated at trauma centers.^{6,25,26,36,37,38} Collaborative care broadens the reach of mental health treatment for injury survivors by combining evidence-based PTSD pharmacotherapy and behavioral intervention elements with proactive injury case management. A recent meta-analysis²⁵ identified stepped care collaborative care interventions as the optimal treatment approach for injured trauma survivors compared with cognitive behavioral therapy alone,⁸⁷ because stepped care approaches combine effective evidence-based interventions with expanded breadth of applicability, resulting in greater overall population impact.^88,89,90 The present trial extends these earlier observations by demonstrating that a brief collaborative care intervention can be feasibly delivered by front-line trauma center clinicians and is effective when well implemented. Two recent investigations that used brief session–based cognitive behavioral therapy delivered by mental health specialty³⁵ and front-line acute care³⁰ clinicians found no PTSD treatment effects. The PTSD treatment effects observed in the present investigation were toward the middle to lower end of the continuum reported for more time-intensive multisession early cognitive behavioral therapy interventions delivered to select, disproportionately noninjured patient samples by mental health specialty clinicians.²⁴ A meta-analysis of 37 primary care intervention trials reported comparable collaborative care treatment effects for depression (standardized mean difference, 0.25; 95% CI, 0.18-0.31)⁹¹ as were observed for PTSD in the present trial.

Limitations

This investigation has limitations. An important consideration is that intervention patients had higher baseline PTSD symptom levels compared with controls, and therefore regression to the mean could explain observed treatment effects. However, regression to the mean is unlikely to comprehensively explain observed treatment effects in the present trial, because, when the protocol was well implemented, significant baseline differences between the intervention and control groups were minimized and treatment effects were enhanced. In addition, subgroup analyses stratified by baseline risk for enduring PTSD symptoms demonstrated that control patients had significantly higher symptoms from baseline to 3 months relative to intervention patients. The data on study exclusions, presented in Figure 1, are subject to error derived from site variability in reporting practices and lapses in hospital internet connectivity/software malfunction. In addition, more supervision was associated with randomization to an earlier stepped-wedge wave; however, the amount of supervision was not associated with the quality of protocol implementation.⁹² As another limitation, we acknowledge that in this injured patient population with multiple recurrent traumatic life events, many participants experienced enduring PTSD symptoms even after receiving the brief collaborative care intervention.

Conclusions

Results of this randomized clinical trial contribute to an evolving literature on early interventions for patients who experience life-threatening injuries and are at risk for the development of PTSD. A series of recent US events, including the coronavirus disease 2019 pandemic and increasing firearm violence, highlight the need for feasibly delivered hospital-based interventions that can effectively target PTSD symptoms.^{53,54,55,56,57,58,59,60} The observation that, when properly implemented, a brief collaborative care intervention can reduce PTSD symptoms suggests the possibility that broad-reach preventive interventions for life-threatening exposures can be developed and implemented across acute care medical settings.

The ACS/COT has demonstrated the ability to mandate screening and intervention procedures to address alcohol use problems at US trauma centers based on the results of pragmatic comparative effectiveness trials.^42,43,46 A key observation is that the previous ACS/COT alcohol policy mandate has resulted in variable quality of implementation across US trauma center sites.⁴⁴ Orchestrated efforts targeting policy and funding should systematically incorporate the findings of the present study into national trauma center requirements and verification criteria to ensure the sustainable implementation of high-quality screening and intervention procedures for PTSD and related comorbid conditions.

Supplement 1.

Trial Protocol

Supplement 2.

eTable 1. Trauma Center Site Implementation Quality Domains

eFigure. PTSD Symptom Levels Over Time: Firearm Injury Stratification by Quality of Trauma Center Implementation

eTable 2. Details of Intervention Care Processes (N = 265)

eTable 3. Percent of Participants With a 10-Point Reduction in PTSD Checklist (PCL-C) Symptom Levels Over Time (N = 635)

Supplement 3.

Data Sharing Statement