A Prospective Assessment of Resilience in Trauma Patients Using the Connor-Davidson Resilience Scale

Abstract

Background:

Resilience, or the ability to adapt to difficult or challenging life experiences, may be an important mediator in trauma recovery. The primary aim of this study was to describe resilience levels for trauma patients using the validated Connor-Davidson Resilience Scale (CD-RISC).

Methods:

Adult trauma patients admitted to a Level 1 trauma center (June 2022-August 2022) were surveyed at the time of admission and by phone between two weeks and one-month after the original survey to obtain follow-up scores. We utilized the validated CD-RISC score, a 25-question survey with 5 subfactors (Tenacity, Positive Outlook, Social Support, Problem Solving, Meaning & Purpose). Each question scored from 0 to 4 (maximum score 100, representing the highest resilience). Patient factors were collected from the electronic medical record and trauma health registry. Wilcoxon signed-rank test and multivariable linear regression were used to understand associations with CD-RISC scores.

Results:

We enrolled 98 patients. Median age was 50 y (IQR 32–67) and 74% were male. Baseline median CD-RISC score on admission was 88 (IQR 81–94). Follow-up surveys (N=64) showed a median score of 89.5 (80–90.5) (p=NS). No demographic variable was significantly associated with increasing baseline CD-RISC score. Increased length of stay (β=1.03), insurance (β=−7.50), and unknown race (β=23.69) were correlated with follow-up CD-RISC score. Subfactor “Meaning and Purpose” decreased at follow-up but was not statistically significant (p=0.05).

Conclusion:

Validated tools that can accurately distinguish variability in resilience scores are needed for the trauma patient population to understand its relationship with long-term patient health outcomes.

Background

After sustaining a major injury, 80% of severe trauma survivors score below normal for overall well-being at 18-month follow-up.¹ Two years after injury, trauma patients suffer from persisting pain, functional deficit, socioeconomic difficulties, and mental health impairment.² After more than 20 years from initial injury, nearly half of patients suffer from one symptom of PTSD.³ This lasting impact can predispose a patient for future injury as mental illness is a risk factor for traumatic revictimization.⁴ It is vital to explore the role of resilience in trauma patients’ psychological and physical recovery, yet research seeking to understand and promote resilience in trauma patients is limited.⁵

Resilience is defined by the American Psychological Association as the process and outcome of successfully adapting to difficult or challenging life experiences.† Resilience is a dynamic characteristic that can be positively or negatively modified by external events.^6,7 There is ongoing research examining the utility of resilience measures and the potential mediating effects of resilience in the context of trauma.^8–11 Exploring how a patient’s resilience will change after trauma is beneficial to healthcare workers and policy makers so that they can identify necessary interventions to support patient recovery. Several scales exist to measure resilience.¹² The Connor-Davidson Resilience Scale (CD-RISC) is a well-studied, validated, and internally consistent 25-item self-report measure developed in 2003 to assess resilience levels in clinical practice, treatment-outcome studies, and biological research.⁶

The aim of our study was to measure resilience in trauma patients using the CD-RISC scale at admission and after discharge. We hypothesized certain patient factors or injury-specific characteristics are associated with high resilience.

Methods

Patient Population

We performed this study prospectively at an urban Level 1 Trauma Center. Patients were selected from our inpatient census, which included patients admitted to the Trauma Service or the Orthopedic Trauma Service. Patients were enrolled between June 27th, 2022 and August 5th, 2022. Only patients who were over 18 years of age and had been admitted to the hospital for more than two midnights were eligible for inclusion. We included patients admitted to the trauma service for two midnights to ensure that our study would include patients with a serious traumatic injury. Prisoners, pregnant patients, patients who were not able to speak, did not speak English, physically or cognitively unable to consent, or requiring acute psychiatric hospitalization were excluded from this study. The trauma units were rounded on weekdays between 11 am and 2 pm to enroll eligible patients during index hospitalization to the study. Study enrollment and baseline screening were performed by two research personnel (SM and DRC), under supervision by the PI of the study). Eligible patients who were awake, willing to participate in the study, and signed informed consent forms were administered the Connor-Davidson Resilience Scale (CD-RISC).

The 25-question version of the CD-RISC was utilized. The CD-RISC can be subdivided into five lifestyle factors: 1) Personal competence, high standards, and tenacity (Tenacity), 2) trust in one’s instincts, tolerance of negative affect, strengthening effects of stress (Positive Outlook), 3) positive acceptance of change and secure relationships (Social Support), 4) control (Problem-Solving), and 5) spiritual influences (Meaning & Purpose).⁶ These factors were not shared with patients. Patients were read out the questions and asked to give a score on a 5-point scale (0–4), with 0 representing “Not at all confident” and 4 representing “Completely confident”. The patient’s score on each question was recorded and the scores for the 25 total questions were summed together. The score range is from 0 to 100, with higher scores reflecting greater resilience.

Study participants were contacted by phone beginning two weeks and up to one-month after the initial survey; follow-up was performed by three of the coauthors (SM, DRC, PNH). During the second contact, patients were readministered the CD-RISC. Each statement was read, and the answer was recorded. To minimize bias, the survey administrator provided only the directions written on the survey as would have been available had the respondent completed the questionnaire individually.

Demographics and pre-hospital data were collected from the electronic medical record and the trauma registry. These variables included age, gender, race (including categories of White, Black, Other, and Unknown), ethnicity (including categories of Hispanic, Non-Hispanic, Declined to answer, and Unavailable), marital status, employment status, trauma activation level (level 1, or full team activation; level 2, or limited trauma activation; level 3, or emergency department team activation only; or trauma consult without activation), and patient origin (scene or transfer from another facility). Injury mechanism, hospital length of stay, number of days in the intensive care unit, emergency department and discharge disposition, comorbidities, insurance status, and primary payor were also retrieved from the trauma registry.

Analysis

The data analysis for this study was generated using Stata/SE v17.0 (StataCorp, College Station, TX). Descriptive analysis was used to illustrate patient demographics. Categorical variables are presented as counts and proportions. Normally distributed continuous variables are presented as the mean, standard deviation, and range. Skewed variables are presented with the median and interquartile range. Wilcoxon rank sum test, and chi-square tests were performed, as appropriate, to test for differences in patient characteristics between those who followed up and were lost to follow-up. Wilcoxon signed-rank test was used to describe the relationship between admission and follow-up CD-RISC scores. Multivariable linear regression was performed to describe the relationship between baseline and follow-up CD-RISC scores with various patient demographics and injury characteristics.

Based on a power calculation assuming the baseline CD-RISC score would be 80±5, a sample of 25 patients at baseline and 25 patients at follow-up would be powered to detect a 5% decrease in score. Per our protocol, we enrolled patients for a specified time period and were overpowered to detect this difference. Findings were considered statistically significant if p ≤ 0.05.

This prospective observational study was approved by the Institutional Review Board.

Results

A total of 98 patients were included in this study, of which 64 (65.3%) were reached for follow-up (Figure 1, Table 1). The median age was 50 years (interquartile range [IQR] 32–67) and over half (53.1%) were unemployed. Nearly 74% of patients were male. Most patients (65.3%) were of white race. Black patients accounted for 27.6% of the total cohort. Most patients (52.2%) presented as Category 2 activation and more than half (56.8%) were brought directly from the scene of injury. Less than one-third (27.8%) of patients were a victim of violence. The injury severity score ranged from 1 to 50 with similar distribution between mild (1–9), moderate (10–15), severe (16–24), and profound (25+). The median hospital length of stay was 9 days (IQR 4–14), during which over one-third (35.7%) of patients experienced an in-hospital complication. Five patients did not answer at least one question in the baseline survey, with a range of 1 to 3 questions omitted; three patients did not answer at least one question in the follow-up survey, with a range of 1 to 11 questions omitted.

Figure 1.

Flow diagram of prospective cohort enrollment.

Table 1.

Patient Cohort Characteristics

Variable	Total Cohort n (%)* n=98	Follow-Up n (%)* n=64	Lost to Follow-Up n (%)* n=34	p-value
Age in years, median (IQR)	50 (32–67)	44 (29–65)	56 (40–71)	0.09
Male sex	72 (73.5%)	47 (65.3%)	25 (34.7%)	0.99
Race
Black	27 (27.6%)	20 (74.1%)	7 (25.9%)	0.26
White	64 (65.3%)	38 (59.4%)	26 (40.6%)	0.09
Employment				0.76
Employed	29 (29.6%)	20 (31.3%)	9 (26.5%)
Unemployed	52 (53.1%)	33 (51.6%)	19 (55.9%)
Retired	11 (11.2%)	8 (12.5%)	3 (8.8%)
Unknown	6 (6.1%)	3 (4.7%)	3 (8.8%)
Comorbidities
Hypertension	33 (33.7%)	21 (32.8%)	12 (35.3%)	0.80
Smoking	20 (20.4%)	13 (20.3%)	7 (20.6%)	0.97
Alcohol use	11 (11.2%)	4 (6.3%)	7 (20.6%)	0.03
Mental or Personality Disorders	10 (10.2%)	5 (7.8%)	5 (14.7%)	0.28
Trauma activation levels				0.47
Category 1	31 (32.3%)	19 (30.2%)	12 (36.4%)
Category 2	52 (52.2%)	37 (58.7%)	15 (45.45%)
Category 3	3 (3.1%)	1 (1.6%)	2 (6.1%%)
Consult	10 (10.4%)	6 (9.5%)	4 (12.12%)
Patient origin				0.43
Direct from scene	54 (56.8%)	35 (56.5%)	19 (57.6%)
Transfer	36 (37.9%)	25 (40.3%)	11 (33.3%)
Injury Severity Score				0.865
1–9 (mild)	30 (30.6%)	21 (32.8%)	9 (26.5%)
10–15 (moderate)	19 (19.4%)	13 (20.3%)	6 (17.7%)
16–24 (severe)	23 (23.5%)	14 (21.9%)	9 (26.5%)
25+ (profound)	26 (26.53%)	16 (25.0%)	10 (29.4%)
Hospital LOS in days, median (IQR)	9 (4–14)	9 (4–14)	9.5 (6–14)	0.62
Victim of Violence	27 (27.8%)	19 (29.7%)	8 (24.2%)	0.57
Penetrating injury mechanism	22 (22.5%)	17 (26.6%)	5 (14.7%)	0.18
In-hospital complications	35 (35.7%)	19 (29.7%)	16 (47.1%)	0.08
30-day Readmission	8 (8.2%)	6 (9.4%)	2 (5.9%)	0.54
Discharge disposition				0.26
Inpatient rehabilitation	18 (19.0%)	9 (14.5%)	9 (27.3%)
Home or self care (routine discharge)	41 (43.2%)	31 (50.0%)	10 (30.3%)
Skilled nursing facility	31 (32.6%)	18 (29.0%)	13 (39.4%)
Insurance				0.068
Yes	87 (88.8%)	57 (89.1%)	30 (88.2%)
No	6 (6.1%)	2 (3.1%)	4 (11.8%)
Unknown	5 (5.1%)	5 (7.8%)	0 (0.0%)

^*Categories are shown if samples are small. Data presented as n (%) unless otherwise specified.

Follow-up CD-RISC scores were obtained a median 17 days (14–38) after baseline score. There was no significant difference in patient characteristics between the patients who were able follow-up and those who did not (Table 1).

On admission, patients had a median (IQR) baseline CD-RISC score of 88 (81–94) (Table 2). At follow-up, the median CD-RISC score increased to 89.5 (80–90.5), p=NS. On Wilcoxon signed-rank test, baseline CD-RISC subgroup and follow-up CD-RISC scores were not significantly different, although Factor 5 (Meaning and Purpose) approached significance (p=0.054). The range and distribution of each Factor is depicted in Figure 2. On multivariable linear regression, no variable was significantly correlated with increasing baseline CD-RISC score (Table 3).

Table 2.

CD-RISC Score

Variable	Baseline Score (median, IQR) n=98	Follow-Up Score (median, IQR) n=64	p-value
CD-RISC 25	88 (81–94)	89.5 (80–90.5)	0.57
Factor 1 (Tenacity)	3.68 (3.50–4.00)	3.75 (3.25–4.00)	0.49
Factor 2 (Positive Outlook)	3.28 (2.85–3.71)	3.42 (3.00–3.71)	0.66
Factor 3 (Social Support)	3.60 (3.20–4.00)	3.70 (3.30–4.00)	0.69
Factor 4 (Problem-Solving)	3.66 (3.00–4.00)	3.66 (3.00–4.00)	0.95
Factor 5 (Meaning and Purpose)	4 (3.50–4.00)	3.50 (3.00–4.00)	0.05

Figure 2.

CD-RISC Factor Groups at Baseline and Follow-Up

Table 3:

Multivariate linear regression for baseline CD-RISC score

Variable	ß-coefficient (SE)	95% CI	p-value
Age	−0.001 (0.108)	−0.22 – 0.22	0.992
Gender
Male	REF	REF	REF
Female	2.4 (3.66)	−4.87 – 9.78	0.505
Race
Black	REF	REF	REF
White	−3.65 (3.76)	−11.17 – 3.87	0.335
Other	5.61 (15.10)	−24.59 – 35.81	0.712
Unknown	7.31 (10.04)	−12.76 – 27.38	0.469
Employment
Employed	REF	REF	REF
Unemployed	−4.05 (3.49)	−11.02–2.92	0.250
Retired	−1.02 (5.75)	−12.52 – 10.48	0.860
Unknown	−2.40 (6.84)	−16.07 – 11.27	0.727
Victim of Violence	−7.84 (6.42)	−20.68 – 5.01	0.227
Injury Severity Score
1–9 (mild)	REF	REF	REF
10–15 (moderate)	3.08 (4.40)	−5.73 – 11.88	0.488
16–24 (severe)	3.14 (4.45)	−5.77 – 12.05	0.484
25+ (profound)	2.16 (4.30)	−6.43 – 10.75	0.617
Hospital LOS in days	0.10 (0.23)	−0.37 – 0.56	0.680
Penetrating or Blunt Mechanism	8.42 (6.42)	−4.42 – 21.26	0.195
Insured	−0.57 (2.72)	−6.84 – 5.70	0.857
Discharge Disposition
Rehab (Inpatient)	REF	REF	REF
Home or Self-Care (Routine Discharge)	−1.93 (5.00)	−11.93 – 8.07	0.701
Skilled Nursing Facility	−0.54 (4.26)	−9.05 – 7.97	0.899
Long-Term Care	−1.15 (19.36)	−39.87 – 37.58	0.953
Home with Services	−1.83 (7.74)	−17.31 – 13.66	0.814
Comorbidities
Hypertension	1.35 (4.10)	−6.84 – 9.54	0.743
Smoking	−3.14 (4.10)	−11.34 – 5.06	0.447
Diabetes	6.00 (4.39)	−2.78 – 14.78	0.177
Alcohol Use Disorder	1.80 (5.67)	−9.54 – 13.14	0.752

When examining follow-up scores, increased length of stay and unknown race identity were associated with an increased CD-RISC score at follow-up. Insurance was significantly correlated with a decreased CD-RISC score at follow-up. (Table 4)

Table 4:

Multivariate linear regression for Follow-Up CD-RISC score

Variable	ß-coefficient (SE)	95% CI	p-value
Age	0.16 (0.11)	−0.06 – 0.38	0.15
Gender
Male	REF	REF	REF
Female	−0.05 (3.73)	−7.61 – 7.51	1.0
Race
Black	REF	REF	REF
White	0.18 (4.51)	−9.34 – 8.97	0.97
Unknown	23.69 (9.91)	3.56 – 43.82	0.02*
Employment
Employed	REF	REF	REF
Unemployed	−2.57 (3.45)	−9.58 – 4.44	0.46
Retired	−11.36 (6.08)	−23.70 – 0.98	0.07
Unknown	2.98 (6.47)	−10.14 – 16.11	0.65
Victim of Violence	0.79 (7.39)	−14.21 – 15.79	0.92
Injury Severity Score
1–9 (mild)	REF	REF	REF
10–15 (moderate)	1.08 (4.99)	−9.04 – 11.20	0.83
16–24 (severe)	−2.65 (4.36)	−11.49 – 6.20	0.55
25+ (profound)	−5.12 (4.43)	−14.12 – 3.88	0.26
Hospital LOS in days	1.03 (0.24)	0.55 – 1.52	<0.001*
Penetrating or Blunt Mechanism	−8.68 (6.77)	−22.41 – 5.06	0.21
Insured	−7.50 (2.69)	−12.95 - −2.05	0.008*
Discharge Disposition
Rehab (Inpatient)	REF	REF	REF
Home or Self-Care (Routine Discharge)	4.59 (5.59)	−6.76 – 15.93	0.42
Skilled Nursing Facility	−7.74 (5.51)	−18.93 – 3.44	0.17
Long-Term Care	−32.80 (17.47)	−68.26 – 2.66	0.07
Home with Services	1.29 (7.95)	− 17.43 – 14.86	0.87
Comorbidities
Hypertension	4.05 (4.09)	−4.25 – 12.35	0.33
Smoking	−4.88 (4.64)	−14.30 – 4.55	0.30
Diabetes	2.73 (4.56)	−6.52 – 11.99	0.55
Alcohol Use Disorder	4.53 (6.33)	−8.31 – 17.38	0.48

Asterisk (*) indicates statistical significance where p < 0.05

Discussion

Increasing evidence suggests resilience plays an important role in trauma recovery. This study examined the psychometric properties of CD-RISC in a sample of 98 trauma patients. The patients median baseline CD-RISC score was 88 (81–94). Approximately 65% of patients were able to follow-up, and the median follow-up CD-RISC score was 89.5 (80–90.5). On multivariable linear regression, no patient or injury characteristic was significantly correlated with increasing baseline CD-RISC score. On follow-up, increased length of stay and unknown race were significantly positively correlated with CD-RISC score whereas insurance was significantly negatively correlated with CD-RISC score.

Overall, our patients had high CD-RISC scores at baseline and follow-up. In Connor and Davidson’s original study, the CD-RISC score in the general population was reported as 82 (IQR 73–90).⁶ The CD-RISC is responsive to change in many groups including clinical populations, students, and healthcare workers.^{8–10,13–19} Other studies of trauma subpopulations have showed similar or lower CD-RISC scores (patients with traumatic brain injury had a mean 76.8±17.3, and geriatric patients with orthopedic fractures had a mean 73.03)).^8,19 A shorter version, the CD-RISC 10, has been used in patients with spinal cord injury and showed high reliability, validity, and practicality.²⁰ Future study is needed to demonstrate adequate reliability and validity of the CD-RISC within the larger trauma patient population. An alternative would be to develop trauma-specific scales. In a study of three trauma centers, the Functional Outcomes and Recovery after Trauma Emergencies (FORTE) project, resilience was measured using an unvalidated subscale of the trauma-specific QoL (T-QoL) questionnaire. In FORTE, patients with low resilience were less likely to have returned to work or school, and more likely to report chronic pain, functional limitations, and PTSD symptoms.²¹ More recently, a revised T-QoL (RT-QoL) has been developed for the trauma population that measures recovery and resilience as part of a combined domain with physical well-being.²²

In our study, subgroup and total baseline CD-RISC and follow-up CD-RISC scores were not significantly different. Rainey et al. reported a similar finding, whereby resilience scores remained stable from time of injury and 12-month follow-up, regardless of injury type, etiology, or severity.²³ In our study, no patient or injury characteristic was significantly associated with baseline CD-RISC score, which contrasts with Rainey et al. who found the baseline CD-RISC 10 score was positively correlated with education level and employment at baseline.²³ There may be other factors which we did not capture in the CD-RISC or the electronic medical record that more accurately predict resilience. In studies of patients after traumatic brain injury, higher preinjury productivity, life satisfaction, and social connection as well as lower preinjury substance misuse were associated with higher resilience levels.^8,9 Patient frailty has also been suggested as a correlate of patient resilience.²⁴

Interestingly, having insurance was correlated with decreasing CD-RISC follow-up scores (β=−7.50). Of the patients able to follow-up, 57 patients (89.1%) were insured, of whom 18 patients (31.6%) were on Medicaid and 24 (42.1%) were on Medicare. There are documented disparities in rates of surgery, time to surgery, and complications based on insurance type across the general trauma and orthopedic trauma literature.^25–27 In our patient population, many patients are uninsured up until receiving care and application for insurance occurs in the hospital, which may have skewed our results. Increased length of stay correlated with increasing CD-RISC follow-up scores (B=1.03). Our prior research has shown that patients with greater indicators of major trauma, such as longer hospitalization, are likely to engage with a trauma recovery services (TRS). TRS is the psychosocial support program at our institution that provides various services such as peer mentors, counseling, referrals to wraparound service programs, and more.²⁸ It is possible that access to hospital-based support programming improved patient resilience at follow-up, or enabled providers to identify support services.

Akin to the disease-specific quality of life tools that have been developed for different clinical populations, including trauma patients, a validated and specific resilience tool may be better suited to measure resilience in the trauma patient population. Ideally, a resilience screening tool specific for the trauma patient population would help with early identification of individuals at high risk of a poor long-term recovery. We obtained follow-up CD-RISC scores a median 17 days (14–38) after baseline because our research team is interested in the ability to distinguish a patient’s resilience trajectory soon after injury. A disease-specific resilience tool would also provide sufficient granularity and variability in resilience scores by asking questions about outcomes and resources most pertinent to the trauma patient population. This is contrast to our results using the CD-RISC, which revealed relatively high scores for our entire population at baseline and follow-up. Our concern about the CD-RISC scale for the trauma population is that, in our study, it did not reveal sufficient variation in our study population to identify high-risk individuals. Accurately measuring resilience is especially needed as the literature suggests resilience is a mediator of long-term patient outcomes, such as medication adherence, life satisfaction, anxiety, depression, return to work, substance misuse, and long-term chronic pain.^10,11,13,29 Large decreases or persistent low scores could indicate the need for early and targeted delivery of heightened support and care, such as case management, to aid in recovery. Conceptualizations of resilience posit it can be taught and strengthened.³⁰ Future study is needed to evaluate whether intensive interventions, such as evidence-based trauma psychosocial support programs, can modify resilience. TRS participation has already been shown to be associated with higher patient satisfaction and better treatment adherence to follow-up plans.^28,31

Our study should be interpreted in light of several limitations. Resilience may be a baseline characteristic, but we were unable to ascertain resilience scores prior to injury. Another limitation is that participants may have been reluctant to report lower scores because of attitudes, beliefs, and perceptions on sensitive questions. Some patients opted to skip certain questions, expressing dissatisfaction with the wording of or lack of comfortability with certain CD-RISC statements, which may have introduced reporting bias and inflated CD-RISC scores. Future studies could consider methods that provide greater privacy, such as audio computer-assisted self-interviewing technology. Multiple attempts were made to contact patients, and this resulted in variation in the amount of time which passed between baseline and follow-up assessments. The loss of 35% of the sample to follow-up may have introduced non-response bias that inflated follow-up CD-RISC scores. However, patients who did and did not follow-up were similar in demographic and injury characteristics. It is possible that the groups differed in unmeasured characteristics that can affect resilience, such as in food security, housing stability, and financial stability. These are among the specific social needs of our patients that our institution measures in a social determinants of health screening tool.³² This screening tool has yet to be routinely administered to our trauma patient population, and future study will investigate its expansion to this patient group. The only resilience assessment used in this study was CD-RISC. The CD-RISC score assesses characteristics of resilience, and does not assess the resiliency process.⁶ Alternative resilience surveys, like the Brief Resilience Scale or the revised T-QoL survey, or other assessments of mental health, like depression or PTSD, could be used in the future to assess validity. Lastly, this is a single center study with a small sample size.

In conclusion, trauma patients had relatively high baseline and follow-up resilience scores as per the CD-RISC self-report measure. Possession of insurance and unknown race were correlated with a decreased CD-RISC score on follow-up, whereas increased length of stay was correlated with an increased CD-RISC score on follow-up. While we did find some variation in CD-RISC scores in our study, there was not strong evidence that this score would sufficiently identify patients with high risk for poor outcomes. Overall, these results highlight the need for a validated, effective, and specific tool to assess resilience in the trauma patient population and to further investigate the mechanisms to modify resilience to improve trauma patient outcomes.