Emergency Department and Inpatient Utilization Reductions and Cost Savings Associated with Trauma Center Mental Health Intervention: Results from a 5-year Longitudinal Randomized Clinical Trial Analysis

STUCTURED ABSTRACT

Objective:

The American College of Surgeons Committee on Trauma (ACS/COT) now requires that trauma centers have in place protocols to identify and refer patients at high risk for the psychological sequelae of traumatic injury. No investigations have documented reductions in utilization and associated potential cost savings associated with trauma center mental health interventions.

Summary Background Data:

The investigation was a randomized clinical trial analysis that incorporated novel 5-year emergency department/inpatient health service utilization follow-up data.

Methods:

Patients were randomized to a mental health intervention, targeting the psychological sequalae of traumatic injury (n= 85) versus enhanced usual care control (n= 86) conditions. The intervention included case management that coordinated trauma center-to-community care linkages, psychotropic medication consultation, and psychotherapy elements. Mixed model regression was used to assess intervention and control group utilization differences over time. An economic analysis was also conducted.

Results:

Over the course of the 5 years intervention patients demonstrated significant reductions in emergency department/inpatient utilization when compared to control patients (F [19,3210] =2.23, P =0.009]. Intervention utilization reductions were greatest at the 3-6 month (intervention 15.5% versus control 26.7%, Relative Risk [RR] = 0.58, 95% Confidence Interval [95% CI] 0.34, 1.00) and 12-15 month (intervention 16.5% versus control 30.6%, RR = 0.54, 95% CI 0.32, 0.91) post-injury time points. The economic analysis suggested potential intervention cost savings.

Conclusions:

Mental health intervention is associated with significant reductions in emergency department and inpatient utilization as well as potential cost savings. These findings could be productively integrated into future ACS/COT policy discussions.

Mini-Abstract:

The study team conducted a five-year follow-up of a randomized clinical trial that assessed the association between a trauma center-based mental health intervention and emergency department and inpatient service utilization. Patients randomized to the intervention demonstrated significant reductions in utilization with associated potential cost savings when compared to control patients.

INTRODUCTION

Every year approximately 30 million Americans incur traumatic physical injuries and between 1.5 and 2.5 million of these individuals require inpatient hospital admission.^1,2 Posttraumatic stress disorder (PTSD) and other related psychological sequelae occur in approximately 20-40% of hospitalized injury survivors.^3–6 PTSD and mental health comorbidity are associated with a broad spectrum of functional impairments including limitations in physical function and work-related disabilities.^7–9

The American College of Surgeons Committee on Trauma (ACS/COT) now requires that trauma centers have in place protocols to identify patients at high risk for the psychological sequelae of traumatic injury and have a referral process in place for patients who screen positive for being at high risk.^10,11 A growing body of literature now documents that stepped trauma center-based mental health screening, intervention and referral procedures can reduce the symptoms of PTSD and related comorbidities.^5,10–13 Stepped care interventions incorporate proactive case management that serves to link the injured patient to outpatient and community health services, and has the potential to reduce unnecessary emergency department health service use.¹⁴ Psychopharmacologic consultation and psychotherapeutic element delivery targeting PTSD and related comorbidity has also been included in these effective stepped care procedures.^12,13

The new mental health requirement for identification and referral is similar to the ACS/COT requirement for trauma centers alcohol screening and intervention, with the caveat that no trauma center based mental health intervention directly targeting PTSD and related comorbidity is currently required. Prior investigation informing the ACS/COT alcohol requirement demonstrated emergency department and trauma center inpatient utilization reductions as well as cost-savings associated with alcohol screening and brief intervention procedures.^15,16 Literature review, however, revealed no investigations that have documented reductions in health care utilization and associated cost savings derived from trauma center mental health screening, intervention, and referral procedures.

The current investigation is a 5-year longitudinal analysis of a trauma center-based stepped care intervention that targeted the psychological sequalae of traumatic injury with systematic screening, intervention, and referral procedures. The original investigation examined one year patient-reported (e.g., post-traumatic concerns, symptomatic distress, functional status) and utilization outcomes and documented clinically and statistically significant reductions in patients’ post-injury concerns as well as a non-significant trend toward reductions in emergency department utilization for patients randomized to the intervention when compared to patients randomized to the usual care control condition.¹⁷ The current follow-up data analysis incorporates novel 5-year longitudinal emergency department and inpatient health service utilization data. The investigation hypothesizes that the 5-year longitudinal utilization analyses will reveal significant reductions in combined emergency department/inpatient health service utilization in patients randomized to the intervention when compared to patients randomized to the enhanced usual care control condition. An economic analysis was also conducted to identify potential cost savings associated with trauma center mental health intervention and procedures.

METHODS

Design Overview

The investigation was a follow-up analysis of data collected as part of a randomized comparative effectiveness trial that evaluated a stepped mental health intervention procedure versus enhanced usual trauma center care. Previous publication provides a detailed description of the study design.¹⁷ Study recruitment occurred over an 18-month period from March 2014 through September 2015. The University of Washington Institutional Review Board approved all study procedures prior to protocol initiation and written informed consent was obtained from each patient participant. For adolescent participants, parental consent was obtained prior to adolescent assent.

Participants

Patients included in the study identified as either female and male survivors of intentional and unintentional injuries ages ≥ 14 who were admitted to the University of Washington’s Harborview Level I Trauma Center. Patients included in the study were screened for high levels of psychological distress at the time of their injury admission as manifested by a score of ≥ 35 on the PTSD Checklist Civilian Version (PCL-C)¹⁸, a score of ≥ 10 on the 9-item Patient Health Questionnaire (PHQ-9)¹⁹, or a score of ≥ 1 on the PHQ-9 item 9 suicide assessment (“thoughts that you would be better off dead or of hurting yourself”) answered “at least several days”, “more than half the days” or “nearly every day”. Patients were also required to have ≥ 3 posttraumatic concerns at the time of their baseline interview. Patients were only excluded if they required immediate psychiatric intervention (e.g., self-inflicted injury after a suicide attempt), were not Washington State residents, or were currently incarcerated; non-English speaking patients were also excluded from the protocol.

Measures

Emergency Department and Inpatient Health Service Utilization

Emergency department and inpatient health service utilization was assessed using the Emergency Department Information Exchange System (EDIE) developed by Collective Medical Technologies.^20–22 EDIE is a novel clinical informatics tool that aggregates emergency department visits for the population of patients presenting to any Emergency Department in Washington, Oregon and other US states in real-time. EDIE is currently integrated into the medical record at the University of Washington and Harborview Medical Center. For the purposes of the current trial, EDIE allowed blinded population-based 5-year follow-up of all emergency department and inpatient visits across Washington State for the intent-to-treat sample. Of note, the current investigation extended observations from an initial report of EDIE utilization that only contained 1-year post-injury emergency department visits for the intent-to-treat sample of 171 patients.¹⁷

Covariates

For all hospitalized inpatients, PTSD symptoms were assessed with the PCL-C which demonstrates excellent reliability and validity across trauma-exposed patient populations.¹⁸ The PHQ-9 was used as a continuous measure to assess depressive symptoms¹⁹ and has established reliability/ validity in acute care and primary care patients.²³ The Medical Outcomes Study 12/36-item Short Form (SF-12) is a reliable and valid tool to assess physical and mental health function among traumatically injured populations.²⁴ The Mental Component Summary Score (MCS) was used to assess baseline mental health function. The investigation determined injury severity at baseline during the index admission from the medical record International Classification of Disease Codes using the Abbreviated Injury Scale and Injury Severity Score.²⁵ Counts of medical comorbidities were also assessed through a review of trauma registry ICD codes.

Randomization

Randomization occurred in a 1:1 ratio according to a computer-generated random assignment sequence. Randomization occurred in blocks of four to six patients.

Intervention

A detailed description of the intervention procedures is included in study team prior publications.¹⁷ Briefly, intervention patients received a patient-centered mental health screening, intervention, and referral procedure targeting the psychological sequalae of traumatic injury that occurred over the course of the six months after the index injury admission. Patients randomized to the intervention condition were visited by a masters in social work (MSW) intervention team member at bedside. The social worker asked about each patient’s unique concerns and treatment preferences and scheduled ongoing times to meet/call the patient during the initial days and weeks post-injury. The social worker developed a unique treatment plan specific to the preferences of each participant, and with patient consent incorporated family members/other supportive caregivers into post-injury treatment planning. The social worker managed and coordinated care across acute, specialty and outpatient care delivery sectors, with a goal of reducing unnecessary emergency department utilization/inpatient service use. They also reviewed care plans with appropriate outpatient and primary care providers. The social worker gave the study team 24/7 cell phone number to all intervention patients and encouraged spontaneous calls or texts to answer any post-injury concerns that arose, including patient and family member questions regarding the necessity of returning to the emergency department for post-injury concerns.

As in prior study team trials, stepped up care was available for patients with specific post-injury symptomatic concerns. Patients with alcohol use problems or other potential injury risk behaviors, could receive motivational interviewing delivered by the social worker and embedded within care management.^12,26 Patients with high levels of PTSD and/or depressive symptoms, could receive cognitive behavioral therapy elements that were also delivered by the social worker and embedded within care management.^13,27 Psychiatric consultation was also available for patients with a preference for PTSD medication treatment and related conditions such as insomnia. The social worker participated in weekly, computerized decision support tool-facilitated MD/PhD caseload supervision to ensure adherence to the protocol and intervention processes.

Enhanced usual care control condition

Patients in the control condition underwent informed consent, both EHR and in-person screenings, baseline surgical ward evaluation, and blinded follow-up interviews. At the termination of the surgical ward interview, the nurses of all patients in the usual care arm of the investigation were contacted by the study research staff. The research staff reviewed the nature and severity of the individual patient’s posttraumatic concerns and level of emotional distress (e.g., PTSD and depression symptom status). At the time of the investigation, nurse notification of patient concerns and emotional distress constituted an enhancement to usual trauma center care.

Data Analyses

The study team first examined the intervention and control groups for differences in baseline demographic, and clinical characteristics. The primary analysis examined emergency department and inpatient health service utilization over the course of the 5-years after the index injury admission for patients randomized to intervention and control conditions; this longitudinal assessment was designed to build upon and extend prior analyses that only examined EDIE documented emergency department visits over the course of the first year after injury.¹⁷ Mixed model regression analyses were used for the intent-to-treat (N = 171) sample to determine if patients in the intervention and control groups manifested different patterns of emergency department/inpatient utilization over the 5-year follow-up time period.^28–30

First, unadjusted models were fit containing time (quarter), intervention, and intervention by time interactions. Building upon prior study team analyses of EDIE data, the dependent variable for these models was a combined, dichotomous variable indicating one or more emergency department or inpatient visits assessed quarterly over the course of the 5-years after the index injury admission. The study team also performed sensitivity analyses including additional adjustments for relevant clinical and demographic characteristics that were not evenly distributed at baseline (i.e., with P values < 0.10), including SF-12 MCS scores and categories of medical comorbidity. Sensitivity analyses also included adjustment for emergency department and inpatient utilization occurring 3 months prior to the index injury admission. Finally, sensitivity analyses utilized mixed model regression to further delineate discrete time periods over the course of the 5-year follow-up that were associated with intervention treatment effects. The study team used SAS Software Version 9.4³¹ and STATA SE³² version 17.0 for all analyses.

Economic Analyses

Prior trauma care system economic analyses have adopted a variety of approaches relevant to the societal, payor, and trauma center institutional perspectives.^{15,16,33–37} The current investigation was informed both by recent ACS/COT mental health guidelines encouraging payor frameworks that reimburse for trauma center-based service delivery and the limitations imposed by the novel use of EDIE health care information exchange 5-year longitudinal follow-up data.^{10,15,16,20–22,34–37} EDIE provided longitudinal follow-up over 5 years for all emergency department and inpatient visits incurred by the intent-to-treat sample. However, EDIE data does not routinely include cost, charges and reimbursements for emergency department and trauma center inpatient visits. Therefore, mean expenditures per emergency department and inpatient encounter were extrapolated from the Medical Expenditure Panel Survey (MEPS)³⁸. Expenditures documented in MEPS include direct payments for care including payor (i.e. Medicare, Medicaid, private) and out-of-pocket costs. The number of per patient emergency department and inpatient visits were multiplied by mean expenditures, respectively.

Also available to the study team were estimates of intervention costs. Intervention costs accrued over the 6-month post-index trauma center time period. Intervention costs were estimated from time allocation data collected by the study team and were subdivided into costs related to medication consultation versus all other costs. Medication consultation costs were assigned an MD psychiatrist average national hourly wage taken from the Bureau of Labor Statistics (BLS), while all other costs were assigned social work average national hourly wage from the BLS.³⁹ Fixed intervention costs accounted for the time spent by the social worker and MD on administrative activities, an average of 1 hour per week for each, valued at the BLS standard wages and divided by the number of patients in the intervention group. All costs were converted to 2021 dollars using the consumer price index for medical care expenditures.⁴⁰

For the final economic analysis, the study team calculated total costs of the intervention and total cost offset from reduced utilization, which was considered the benefit. ^35,36 We used a 3% discount rate in year 2, which is the standard defined by the Panel on Cost Effectiveness for Health and Medicine.³⁵ The study team calculated a Benefit Cost Ratio (BCR) which establishes a measure of return for each dollar spent on the intervention.^35,36

RESULTS

Seven thousand three hundred and twelve injured patients were evaluated for study participation, and 939 patients were approached by study research staff. Consent was attempted in 308 patients with 56 patients (18%) refusing consent. Eighty patients did not meet mental health distress inclusion criteria, and one patient withdrew, leaving 171 patients for randomization into the longitudinal portion of the investigation.

Patients randomized to the intervention (n = 85) and usual care control (n = 86) conditions did not demonstrate significant differences for the majority of injury, demographic and clinical characteristics including age, sex, marital status, race, ethnicity and insurance status.¹⁷ Intervention and control group patients did differ at baseline with baseline SF-12 MCS (Intervention mean = 48.7, standard deviation = 9.3, versus Control mean 45.9, standard deviation = 45.9, P = 0.07) and number of chronic co-morbid conditions (Intervention percentage one or more = 34.% versus Control percentage one or more = 45.3%, P < 0.01).

Patients in the intervention condition demonstrated a pattern of reduced emergency department and inpatient utilization over the course of the 5 years after injury when compared to patients in the control condition (Figure 1). Reductions attained statistical significance at the 3-6 month (Intervention 15.5% versus Control 26.7% any emergency/inpatient utilization, Relative Risk (RR) = 0.58, 95% Confidence Interval (CC) = 0.34, 1.00, P < 0.05) and 12-15 month intervals (Intervention 16.5%, versus Control 30.6% any emergency department/inpatient utilization, RR = 0.54, 95% CI = 0.32, 0.91, P < 0.05) periods after the injury (Table 1).

Figure 1.

Intervention and control group, proportion with one or more emergency department or inpatient visits at each quarter, over the 5 years after index injury. N=171 at all time points; ED = Emergency Department; m = months.

Table 1.

Patients with Any Emergency Department and/or Inpatient Utilization by Control and Intervention Group Randomization Status

Months After Index Injury Admission	Total Sample, % (N = 171)	Control, % (n = 86)	Intervention, % (n=85)	RR (95% CI)
Baseline to 3	46.0	45.4	46.6	1.03(0.77,1.36)
>3-6	20.3	26.7	15.5	0.58(0.34,1.00)*
>6-9	24.2	28.6	20.4	0.71(0.44,1.15)
>9-12	20.3	23.7	17.5	0.74(0.43,1.26)
>12-15	22.5	30.6	16.5	0.54(0.32,0.91)*
>15-18	17.5	19.7	15.5	0.79(0.43,1.43)
>18-21	17.2	21.7	13.6	0.63(0.34,1.14)
>21-24	20.0	25.7	15.5	0.61(0.35,1.04)
>24-27	19.2	23.7	15.5	0.66(0.37,1.15)
>27-30	16.5	14.8	18.4	1.25(0.68,2.28)
>30-33	17.9	15.8	20.4	1.29(0.73,2.28)
>33-36	16.2	20.7	12.6	0.61(0.33,1.13)
>45-48	20.3	23.7	17.5	0.74(0.43,1.25)
>57-60	14.6	16.8	12.6	0.75(0.39,1.45)

Point estimates and relative risk are derived from unadjusted mixed model regression

^*P≤0.05;

Relative Risk (RR); Confidence Interval (CI)

Unadjusted and adjusted mixed model regression demonstrated a significant group by time interaction over the course of the 5 years after the index admission (Table 2) A consistent pattern of visit reduction for intervention patients when compared to control patients was observed over the course of the first 9 quarters after the index admission with regression analyses revealing a significant main effect for this time period (t[1366] = −2.32, P < 0.05). In contrast, regression analyses demonstrated no significant main or group by time interaction effect between quarters 10 through 20.

Table 2.

Mixed Model Regression Results Comparing Intervention and Control Group Emergency Department/Inpatient Utilization Over the 5-Years After the Index Injury Admission

	F Value (df)	P-value
Model/Variables
5 Year Unadjusted
Intervention	1.62 (1, 3211)	0.20
Quarter	8.36 (19, 3211)	<0.0001
Intervention*Quarter Interaction	1.93 (19, 3211)	<0.0001
5 Year Adjusted
Intervention	1.17 (1, 3210)	0.28
Quarter	8.36 (19, 3210)	<0.0001
Intervention*Quarter Interaction	1.93 (19, 3210)	0.01
Any ED/Inpatient utilization 3 months prior to index injury admission	20.26 (1, 3210)	<0.0001
Comorbid medical conditions	6.73 (3, 3210)	<0.01
Baseline SF-12 MCS*	0.08 (1, 3210)	0.78

^*SF-12 MCS = Medical Outcomes Study 12 item Short Form Mental Health Composite Summary Score

On average intervention providers spent 6.40 hours (SD=4.62 hours) with each intervention patient over the course of the 6-months after the injury. Conservative estimates suggested that approximately one-third (mean= 2.20 hours; Standard Deviation = 2.10 hours) of the total time was spent providing medication consults with an MD psychiatrist. Comparative cost analyses that incorporated intervention cost relative to cost offsets from health care utilization revealed a positive cost offset of $6,496.60 and a benefit-cost ratio of 12.33 over the two years following intervention implementation (Table 3).

Table 3.

Intervention and Control Patient Costs and Benefit Cost Ratio Analysis

	A			B			C±
	Intervention (n = 85)			Control (n = 86)			Difference
	Year 1	Year 2†	Total	Year 1	Year 2†	Total
Discounted Costs
Fixed costs per patient*	$91.73	$45.25	$136.98	-	-	-	($136.98)
Variable costs per patient**	$389.91	-	$389.91	-	-	-	($389.91)
Total							($526.89)
Discounted Offset Costs
ED costs per patient (average)***	$2,870.60	$1,536.24	$4,406.84	$3,390.83	$2,190.23	$5,581.6	$1,174.22
Inpatient stay costs per patient (average)****	$3,449.60	$3,558.45	$7,008.5	$4,261.86	$8,068.57	$12,330.43	$5,322.38
Total							$6,496.60
Benefit Cost Ratio							12.33

^±Column C shows the difference in discounted costs and discounted offset costs between the intervention and control patients. The benefit-cost ratio represents the average potential per patient dollars offset by a decrease in ED and inpatient stays ($6,496.60) relative to the per patient intervention costs ($526.60).

^†Year 2 costs and offset discounted by 3%

^*Fixed costs are costs of maintaining the intervention, divided by the number of patients and would decrease as enrolled patients increase; fixed costs include one hour per week of social worker and psychiatrist time, using BLS standard wages over the first 18 months.

^**Variable costs are the time spent managing medication and counseling for each patient, on average, using BLS standard wages

^***Emergency department costs are the average expenditures on ED visits, per patient using MEPS

^***Inpatient visit costs are the average expenditures on IP stay, per patient using MEPS

DISCUSSION

This is the first investigation to report that a trauma center-based mental health intervention is associated with significant reductions in emergency department and inpatient health service utilization in the months and years after an index injury hospitalization. The current investigation corroborates and extends findings from previous trials of trauma center-based mental health screening, intervention and referral procedures.^{5,12,13,26,41,42} Prior randomized clinical trial investigations have established the ability of stepped trauma center-based mental health interventions to significantly reduce the symptoms of PTSD and related comorbid conditions.^{5,12,13,26,41,42} One prior investigation has demonstrated significant improvements in physical function in patients randomized to a stepped care intervention that included proactive case management, psychopharmacology and psychotherapy components when compared to a usual trauma center care control condition.¹²

In the current investigation the observed utilization reductions were greatest over the first 27 months after the index injury admission, with statistically significant peaks at the 3-6 months and 12-15 months post-hospitalization. Of note, no significant utilization reductions were observed for the first 3 months immediately after the index injury admission or between the 27 months post-injury and the 5-year post-index admission time points. This pattern of intervention utilization reductions suggests that the mental health intervention was less effective at reducing recurrent emergency department and inpatient visits directly related to medical/surgical complications in the immediate weeks and months following the index admission.^14,43 Rather, the mechanism of observed intervention utilization reductions between 3- and 27-months post-injury was more likely to have been attributable to reductions in anxiety related emergency department visits and/or reductions in habitual use of the emergency department for routine care visits. Although these hypothesized mechanisms were not directly assessed in the current study, future investigation could productively evaluate these potential mechanisms of intervention associated utilization reductions.

The investigation’s economic analysis suggests that the direct and fixed costs of the intervention were substantially offset by the increased costs of control group emergency department and inpatient visits. The benefit-cost ratio and positive cost offset suggested that cost savings are likely to accrue through reductions in emergency department visits and inpatient stays. Although no prior investigations of trauma center-based mental health interventions have examined utilization reductions and associated potential cost savings, trauma center studies of alcohol screening and brief intervention have included economic analysis.^15,16 Gentilello and colleagues conducted a cost-benefit analysis with targeted relevance to trauma center provider and institutional perspectives, as well as payor and federal agencies responsible for health care costs and insurance legislation.¹⁵ Gentilello and colleagues reported a net cost saving associated with alcohol screening and brief intervention, yet acknowledged as a limitation of their analytic approach, the exclusion of a series of costs relevant to a societal perspective including reduced individual productivity and quality of life. Hinde and colleagues evaluated the ACS/COT alcohol screening and brief intervention requirement on hospital readmissions and costs in the state of Arizona.¹⁶ Hinde and colleagues reported that the ACS/COT requirement was associated with a reduction in the probability of hospital readmission but no impact on readmission and total health care costs. Limitations of the Hinde et al approach to economic analyses included the inability to capture recurrent hospital admission to different trauma centers beyond the trauma center affiliated with the index injury admission and the inability to determine if alcohol screening and intervention was actually delivered to an individual patient.¹⁶

The approach to the economic analysis in the current investigation has limitations. Guidelines supporting the ACS/COT psychological sequelae screening and referral requirement have attempted to provide billing information to providers working to reimburse trauma center-based mental health service delivery.^10,44,45 In concert with ACS/COT efforts to reimburse trauma center mental health service delivery, the economic analyses in the current investigation adopted a payor perspective that focused primarily on potential emergency department and inpatient utilization cost offsets of the intervention. It is acknowledged that from a trauma center institutional perspective, reducing recurrent emergency department/inpatient utilization could potentially reduce hospital revenues.³⁴ A key issue in the current study is that over 95% of patients were either covered by public insurance such as Medicaid/Medicare or were uninsured.¹⁷ Of particular note, Medicaid and Medicare relative to private payors generally contribute negative contribution margins for trauma centers, and overall lower reimbursement to health systems.^46–48

The investigation is novel in its use of EDIE data which provides 5-year population level follow-up emergency department and inpatient utilization data for the intent-to-treat sample. The use of EDIE allows for consolidation of emergency and inpatient utilization from all regional health care facilities in Washington State. However, EDIE currently yields little information that can be used to derive costs for an index and recurrent emergency department and inpatient visits. The study economic analysis is therefore limited by the use of national estimates for utilization costs, which were averages of emergency department and inpatient costs per visit.⁴⁹ Another limitation of the investigation is that it was conducted at a single trauma center site. Future investigation could attempt replications of the findings linking mental health intervention with utilization reductions and associated costs savings, across multiple trauma center sites.

Beyond these considerations, this investigation contributes to a growing body of literature supporting mental health screening, intervention, and referral procedures at US trauma centers.¹¹ A series of investigations have demonstrated the effectiveness of trauma center-based interventions in reducing the symptoms of PTSD and related comorbidities after injury.^{5,12,13,26,41,42} The current investigation expands this evidence-base by documenting reductions in emergency department and inpatient service utilization associated with trauma-center based mental health intervention. These findings could be productively integrated into future ACS/COT policy discussions regarding mental health screening, intervention, and referral procedures for US trauma care systems.

Data Availability:

The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.