Baseline Dissociation and Prospective Success in Special Forces Assessment and Selection
Abstract
Introduction
Although dissociation at the time of trauma (peritraumatic dissociation) has been shown to predict the development of posttraumatic stress disorder (PTSD), it is not yet known whether the tendency to dissociate under nonstressful circumstances (i.e., at baseline) can also serve as a predictor of vulnerability to stress in healthy individuals.
Method
Baseline symptoms of dissociation (CADSS) were assessed in 774 active duty male soldiers enrolled in Special Forces Assessment and Selection (SFAS).
Results
Soldiers who endorsed experiencing any symptoms of dissociation at baseline were significantly less likely to be successful in SFAS. The greater the number of symptoms of dissociation endorsed at baseline, the greater the likelihood of failure.
Discussion
These data explain our earlier findings of fewer symptoms of dissociation in elite troops and may have relevance for the selection and hiring of personnel for nonmilitary, at-risk professions. Better screening may status (positively or negatively) in SFAS. Of the 794 SFAS candidates who were given the study recruitment speech, 774 candidates enrolled in the study. Thus, the refusal rate was three percent. Information on the 20 individuals who refused to participate in the study was not available to the research team.
Participants
Of the 794 candidates approached, 774 (97%) active duty male soldiers (mean age 26, SD=4) agreed to participate in the study. All participants were enrolled in a US Army SFAS program. The participants’ mean years of service in the Army was 4.9 (SD=3.2). Two-hundred eighty soldiers (36%) were married, 403 (52%) soldiers were single, and 86 (11%) soldiers were divorced. Eighty-six percent or 677 candidates were enlisted and 110 (14%) were officers.
Procedure
After providing informed consent, participants completed the self-report portion of the Clinician Administered Dissociative Symptom Scale (CADSS). The CADSS is a reliable, valid, self-report instrument designed to assess state symptoms of dissociation in response to a specified stressor.14 Subjects were instructed to complete the CADSS using the week previous to enrollment in the course as their reference point. Subjects were instructed to inform the research team (orally and in writing) if during the previous week they had experienced any traumatic or highly stressful events. We did not include the clinician-observer component of the CADSS given the low intercorrelation coefficients for this component. After completing the CADSS, participants commenced participation in SFAS.
Data analysis
In order to test the hypothesis that symptoms of dissociation would be significantly related to success or failure in SFAS, the following variables were created: total CADSS scores (the sum of individual CADSS items); classification scores indicating whether or not subjects reported symptoms of dissociation at baseline (1=yes; 0=no); and two additional classification scores designed to classify subjects in a binary fashion based on whether or not their CADSS total score was greater than, at, or below a specified value (less than 5; equal to, or greater than, 5; less than 11; equal to or greater than 11). These classification cut-off points were selected based on the distribution of responses from subjects in this study and on the mean pre-stress CADSS dissociation scores noted in our previous studies.15 Chi-squared analyses were performed to test whether subjects endorsing baseline symptoms of dissociation (any, greater than 5 points, greater than 11 points) were more likely to fail SFAS compared to peers who did not report such symptoms.
Receiver operator characteristics (ROC) curves ROC curves were created by using the CADSS baseline total score (the test variable) in order to predict outcome in the SFAS program (the state variable, where 1= failed SFAS). ROC graphs were created for the group as a whole and for the sub-group of subjects whose CADSS score was 1 or greater. For both ROC graphs, the area under the curve as well as coordinate points for the curve were calculated (SPSS 11.5). The null hypothesis assumption was that the true area under the curve equals 0.5. With regard to the parameters for the standard distribution of error, the distribution assumption was nonparametric and the confidence interval 95 percent.16 Although this method is also a regression model, it offers an advantage over the logistic regression format in that a classification table corresponding to specific scores on the CADSS and to the likelihood of success or failure in the course is possible.
Results
Of the 774 who participated in the study and in SFAS, 318 successfully passed the course; 456 candidates did not. The mean CADSS score at baseline was 2.39 (SD=4.5; range=44). The distribution was not normal (skewness 3.30). Of the 774 subjects, 425 subjects (55.0%) did not endorse any symptoms of dissociation at baseline; 349 subjects (45.0%) endorsed such symptoms. As shown in Tables 1, 2, and 3, the pass rates in SFAS were significantly different between the group of candidates who reported dissociation and those who did not (Score of 1 or greater on the CADSS versus CADSS score of zero: Chi-Square=4.5; df=1; asymptotic significance [asymp. sig.] (2-sided) p<0.035; Fisher’s Exact Test, Exact Significance: p<0.04 (2-sided); p<0.021 (1-sided); Score of five or more on the CADSS versus a score of less than 5 on the CADSS: Chi-Square=9.3; df=1; asymp. sig. (2-sided) p<0.002; Fisher’s Exact Test, Exact Significance: p<0.002 (2-sided); p<0.001 (1-sided); Score of 11 or more on the CADSS versus a score of less than 11: Chi-Square=10.7; df=1; asymp. sig. (2-sided) p<0.001; Fisher’s Exact Test, Exact Significance: p<0.001 (2-sided); p<0.001 (1-sided).
ROC curve data
When an ROC curve is created, the area may take values between 1 and zero. A value of 1 or zero would indicate that the test is always right or always wrong, respectively. If the test performs no better than chance at detecting the state variable (for example, failure in SFAS), the area under the curve would be 0.5. Using the variable “total dissociation score” as the “test variable” and “status” as the “state variable” (value of the state variable=failure in SFAS), ROC analyses indicated that the area under the curve was 0.6 (nonparametric standard of error=0.021); p<0.01. Thus, the total dissociation score performed better than chance at predicting likelihood of failure at SFAS.
Table 4 lists the coordinate points of the ROC analysis and indicates, for a given score on the CADSS, the probability of being status (positively or negatively) in SFAS. Of the 794 SFAS candidates who were given the study recruitment speech, 774 candidates enrolled in the study. Thus, the refusal rate was three percent. Information on the 20 individuals who refused to participate in the study was not available to the research team.
TABLE 4.
Predicting failure at SFAS
| ROC coordinates of the curve: area under the curve=0.55; std error=0.02; asymp. sig: p<0.02 | ||
|---|---|---|
| Positive if greater than or equal to the following: | Sensitivity | 1-Specificity |
| -1.00 | 1.000 | 1.000 |
| .50 | .482 | .406 |
| 1.50 | .386 | .308 |
| 2.50 | .307 | .233 |
| 3.50 | .246 | .189 |
| 4.50 | .206 | .138 |
| 5.50 | .175 | .097 |
| 6.50 | .151 | .075 |
| 7.50 | .129 | .057 |
| 8.50 | .112 | .047 |
| 9.50 | .094 | .035 |
| 10.50 | .081 | .025 |
| 11.50 | .070 | .022 |
| 12.50 | .059 | .019 |
| 13.50 | .053 | .016 |
| 14.50 | .042 | .016 |
| 16.00 | .033 | .006 |
| 17.50 | .031 | .006 |
| 18.50 | .024 | .006 |
| 19.50 | .018 | .006 |
| 20.50 | .015 | .006 |
| 23.50 | .009 | .003 |
| 25.50 | .007 | .003 |
| 27.00 | .007 | .000 |
| 36.00 | .002 | .000 |
| 45.00 | .000 | .000 |
Participants
Of the 794 candidates approached, 774 (97%) active duty male soldiers (mean age 26, SD=4) agreed to participate in the study. All participants were enrolled in a US Army SFAS program. The participants’ mean years of service in the Army was 4.9 (SD=3.2). Two-hundred eighty soldiers (36%) were married, 403 (52%) soldiers were single, and 86 (11%) soldiers were divorced. Eighty-six percent or 677 candidates were enlisted and 110 (14%) were officers.
Procedure
After providing informed consent, participants completed the self-report portion of the Clinician Administered Dissociative Symptom Scale (CADSS). The CADSS is a reliable, valid, self-report instrument designed to assess state symptoms of dissociation in response to a specified stressor.14 Subjects were instructed to complete the CADSS using the week previous to enrollment in the course as their reference point. Subjects were instructed to inform the research team (orally and in writing) if during the previous week they had experienced any traumatic or highly stressful events. We did not include the clinician-observer component of the CADSS given the low intercorrelation coefficients for this component. After completing the CADSS, participants commenced participation in SFAS.
Data analysis
In order to test the hypothesis that symptoms of dissociation would be significantly related to success or failure in SFAS, the following variables were created: total CADSS scores (the sum of individual CADSS items); classification scores indicating whether or not subjects reported symptoms of dissociation at baseline (1=yes; 0=no); and two additional classification scores designed to classify subjects in a binary fashion based on whether or not their CADSS total score was greater than, at, or below a specified value (less than 5; equal to, or greater than, 5; less than 11; equal to or greater than 11). These classification cut-off points were selected based on the distribution of responses from subjects in this study and on the mean pre-stress CADSS dissociation scores noted in our previous studies.15 Chi-squared analyses were performed to test whether subjects endorsing baseline symptoms of dissociation (any, greater than 5 points, greater than 11 points) were more likely to fail SFAS compared to peers who did not report such symptoms.
Receiver operator characteristics (ROC) curves ROC curves were created by using the CADSS baseline total score (the test variable) in order to predict outcome in the SFAS program (the state variable, where 1= failed SFAS). ROC graphs were created for the group as a whole and for the sub-group of subjects whose CADSS score was 1 or greater. For both ROC graphs, the area under the curve as well as coordinate points for the curve were calculated (SPSS 11.5). The null hypothesis assumption was that the true area under the curve equals 0.5. With regard to the parameters for the standard distribution of error, the distribution assumption was nonparametric and the confidence interval 95 percent.16 Although this method is also a regression model, it offers an advantage over the logistic regression format in that a classification table corresponding to specific scores on the CADSS and to the likelihood of success or failure in the course is possible.
Results
Of the 774 who participated in the study and in SFAS, 318 successfully passed the course; 456 candidates did not. The mean CADSS score at baseline was 2.39 (SD=4.5; range=44). The distribution was not normal (skewness 3.30). Of the 774 subjects, 425 subjects (55.0%) did not endorse any symptoms of dissociation at baseline; 349 subjects (45.0%) endorsed such symptoms. As shown in Tables 1, 2, and 3, the pass rates in SFAS were significantly different between the group of candidates who reported dissociation and those who did not (Score of 1 or greater on the CADSS versus CADSS score of zero: Chi-Square=4.5; df=1; asymptotic significance [asymp. sig.] (2-sided) p<0.035; Fisher’s Exact Test, Exact Significance: p<0.04 (2-sided); p<0.021 (1-sided); Score of five or more on the CADSS versus a score of less than 5 on the CADSS: Chi-Square=9.3; df=1; asymp. sig. (2-sided) p<0.002; Fisher’s Exact Test, Exact Significance: p<0.002 (2-sided); p<0.001 (1-sided); Score of 11 or more on the CADSS versus a score of less than 11: Chi-Square=10.7; df=1; asymp. sig. (2-sided) p<0.001; Fisher’s Exact Test, Exact Significance: p<0.001 (2-sided); p<0.001 (1-sided).
ROC curve cata
When an ROC curve is created, the area may take values between 1 and zero. A value of 1 or zero would indicate that the test is always right or always wrong, respectively. If the test performs no better than chance at detecting the state variable (for example, failure in SFAS), the area under the curve would be 0.5. Using the variable “total dissociation score” as the “test variable” and “status” as the “state variable” (value of the state variable=failure in SFAS), ROC analyses indicated that the area under the curve was 0.6 (nonparametric standard of error=0.021); p<0.01. Thus, the total dissociation score performed better than chance at predicting likelihood of failure at SFAS.
Table 4 lists the coordinate points of the ROC analysis and indicates, for a given score on the CADSS, the probability of being right (sensitivity) or of being wrong (1 minus the specificity) in predicting whether a subject who endorsed any particular score on the CADSS would fail the SFAS course. As noted in Table 4, the probability of being wrong in predicting that a candidate who obtains a particular dissociation score on the CADSS can be calculated. For example, using these Signal Detection Methods, the likelihood of being wrong in predicting that a candidate obtaining a score of 5 or more on the CADSS will fail SFAS is approximately 9.0 percent; similarly, the probability of error in predicting that a candidate who obtains a score of 11 or greater on the CADSS will fail SFAS is less than 2.0.
TABLE 1.
Increased dissociation is associated with increased likelihood of failure: any symptom endorsement
| Fail SFAS | Pass SFAS | Total | |
|---|---|---|---|
| Chi-square=4.5; df=1 asymp. sig (2-sided): p<0.035 | |||
| No Dissociation (CADSS = 0) | 236 | 189 | 425 |
| CADSS > 1 | 220 | 129 | 349 |
| Total | 456 | 318 | 774 |
TABLE 2.
Increased dissociation is associated with increased likelihood of failure: CADSS score of 5 or greater
| Fail SFAS | Pass SFAS | Total | |
|---|---|---|---|
| Chi-square=9.3; df=1 asymp. sig (2-sided): p<0.002 | |||
| CADSS <5 | 376 | 287 | 663 |
| CADSS > 5 | 80 | 31 | 111 |
| Total | 456 | 318 | 774 |
TABLE 3.
Increased dissociation is associated with increased likelihood of failure: CADSS score of 11 or greater
| Fail SFAS | Pass SFAS | Total | |
|---|---|---|---|
| Chi-square=10.8; df=1 asymp. sig (2-sided): p<0.001 | |||
| CADSS < 11 | 419 | 310 | 729 |
| CADSS > 11 | 37 | 8 | 45 |
| Total | 456 | 318 | 774 |
Discussion
Soldiers who endorsed experiencing any symptoms of dissociation on the CADSS at baseline were significantly less likely to be successful in Special Forces Assessment and Selection (SFAS). Indeed, the greater the number of symptoms of dissociation endorsed at baseline, the greater the likelihood of failure in the course. Fewer than nine in 100 candidates who had a baseline CADSS score of greater than 5 and fewer than two in 100 with a score greater than 11 (maximum score of 79 possible) passed the course. Thus, tendency to dissociate at baseline served as a significant predictor of military training performance under highly demanding and stressful conditions. These data may help to explain our earlier finding that elite Special Forces troops exhibited fewer stress-induced symptoms of dissociation than general troops during high-intensity training. It appears likely that the SFAS process “weeds out” soldiers who tend to dissociate and selects those who do not tend to dissociate. That said, it is important to underscore the fact that we were not able to assess other life experiences in this particular sample (for example, history of traumatic stress exposure). It is possible that this or other variables contributed to the present findings. In the future, we anticipate having permission from the military to assess other variables of interest.
With the increased operational tempo of special operations units in support of the Global War on Terror, US military leaders continue to look for ways to maximize throughput of SFAS programs. With student attrition a primary concern, directors of SFAS programs are faced with the choice of lessening requirements, introducing remedial interventions for poor performers, or improving candidate selection protocols. Many believe that lowering the criteria for successful completion of such programs would be imprudent. However, assessment of candidates for baseline dissociation could help to identify those who may benefit from remedial efforts or be better suited for other military occupations. Screening out candidates who are unlikely to succeed would better focus training resources on candidates more likely to successfully complete the course. Signal Detection Methods may significantly assist in decision making. Although they do not replace the decision-making capacity of professionals, Signal Detection Methods provide professions information about the probability that they will err in the predictions they make. Clearly the accuracy of such methods is directly related to the degree to which valid, normative databases for populations of interest have been established.
The present study has several limitations. First, the CADSS data are self-report data and we currently do not know whether they reflect a unique factor or a more generic “first factor.” Although the subjective data was predictive of future success, if these data were actually used in selection programs, potential candidates knowledgeable about this relationship might cease to report such information. Future studies are underway to clarify this issue and to assess whether the propensity to dissociation may be more objectively assessed prior to stress exposure.
A second limitation is related to the study population. The present study was limited to US military personnel. To our knowledge, the relationship between propensity to dissociation and success in selection programs for civilian professions exposed to high stress (such as search and rescue, law enforcement, or firefighting) have not yet been conducted. Thus, at present we do not know whether the findings have relevance to non-military professions. However, within the context of the current war, improved screening for military occupations may result in the identification of candidates who are at greater risk for stress-related difficulties; this capability might one day lead to improved primary intervention strategies or to better military job placement decisions.17,18
Acknowledgments
This study was made possible through a grant from the Department of Defence (author CAM) and by support from the National Center for PTSD. The views expressed in this manuscript are those solely of the authors and do not represent official views of the US Army, US Navy, or Department of Veterans Affairs.