One-year prospective follow-up of motor vehicle accident victims

Pergamon S0005-7967(96)@1038-1

ONE-YEAR

PROSPECTIVE

Behav. Res. Ther. Vol. 34, No. 10, pp. 775-786, 1996 Copyright © 1996 ElsevierScienceLtd Printed in Great Britain. All rights reserved 0005-7967/96 $15.00 + 0.00

FOLLOW-UP

VEHICLE ACCIDENT

OF MOTOR

VICTIMS

EDWARD B. BLANCHARD *l, EDWARD J. HICKLING 1'2, KRISTINE A. BARTON 1, ANN E. TAYLOR j, WARREN R. LOOS 2 and JACQUELINE JONES-ALEXANDER ~ tCenter for Stress and Anxiety Disorders, University at Albany-SUNY, 1535 Western Avenue, Albany, NY 12203, U.S.A. and 2Capital Psychological Associates, Albany, NY, U.S.A. (Receh,ed 20 April 1996) Summary--One-hundred and thirty-two victims of motor vehicle accidents (MVAs), who sought medical attention as a result of the MVA, were assessed at three points in time: 1-4 months post-MVA, 6 months later, and 12 months later. Of the 48 who met the full criteria for Post-Traumatic Stress Disorder (PTSD) initially, half had remitted at least in part by the 6-month follow-up point and two-thirds had remitted by the 1-yr follow-up. Using logistic regression, 3 variables combined to correctly identify 79% of remitters and non-remitters at the 12-month follow-up point: initial scores on the irritability and foreshortened future symptoms of PTSD and the initial degree of vulnerability the subject felt in a motor vehicle after the MVA. Four variables combined to predict 64% of the variance in the degree of post-traumatic stress symptoms at 12 months: presence of alcohol abuse and/or an Axis-il disorder at the time of the initial assessment as well as the total scores on the hyperarousal and on avoidance symptoms of PTSD present at the initial post-MVA assessment. Copyright © 1996 Elsevier Science Ltd

INTRODUCTION

It is well recognized that varying proportions of individuals exposed to a traumatic event which threatens life or bodily integrity will develop Post-Traumatic Stress Disorder (PTSD) acutely, and likewise, that varying proportions will remit over time (Green, 1994; Davidson & Foa, 1993; American Psychiatric Association, 1994). Much of the early evidence for the remission was gathered retrospectively (e.g. NVVRS, Kulka, Schlenger, Fairbank, Hough, Jordan, Marmar & Weiss, 1988) and represents an inference process, that is, a smaller percentage of individuals exposed to an earlier trauma are currently positive for PTSD than at some earlier point after trauma exposure. For example, the NVVRS found about 15% of veterans exposed to combat in Southwest Asia met the criteria for PTSD when assessed in 1985, whereas almost 30% would have been positive for PTSD at some time in the past. In addition to some pioneering prospective studies of natural disaster survivors who were assessed early after the disaster and then some years later (e.g. Green, Lindy, Grace, Gleser, Leonard, Korol & Winget, 1990) the past few years have seen an increase in prospective follow-up evaluations of trauma victims. For example, Rothbaum and Foa (1993) assessed 94 female rape victims shortly after the assault (average of 12 days) and again at weekly intervals for 3 months. Initially, 94% of the sample met the symptomatic criteria for PTSD (except for the duration criteria). By 3 months, the percentage was 47%. Clearly remission is occurring. Likewise, McFarlane (1988) in Australia, followed up 315 fire fighters exposed to severe, out of control, brush fires with questionnaire assessments at 3, 8, 11 and 29 months post-exposure to the trauma. Thirty-two percent met the criteria for PTSD at the initial assessment, while 63 (19.7%) developed it (as a delayed onset PTSD) over the course of the follow-up. Remission occurred in 64 (20.3%) individuals. In a large American epidemiologic study, Kessler, Sonnega, Bromet, Hughes and Nelson (1995) found 459 cases of PTSD out of 5877 individuals assessed. Relying on retrospective report, they found approximately 28% remission in cases of PTSD from all causes 1 yr after the trauma. *Author for correspondence. 775

776

Edward B. Blanchard et al.

This improved to approximately 38% remission by 2 yr and seemed to reach an asymptote of 60% remission by about 6 yr post-trauma. Interestingly, participation in mental health treatment (although not necessarily for PTSD) was associated with significantly more rapid remission than non-participation (about 25% remission among untreated participants at l yr versus 32% remission among those who received treatment at some point in the follow-up (not necessarily in the first year post-trauma)). Within the study of victims of motor vehicle accidents (MVAs), there have been several prospective follow-up studies. In Oxford (UK), Mayou, Bryant and Duthie (1993) assessed 200 consecutive admissions to the Emergency Room resulting from MVAs. They were able to reassess 174 three months later and 171 (86%) at a 1-yr follow-up. Using the Present State Examination (Wing, Cooper & Sartorious, 1974), 14 (8%) had developed PTSD by the 3-month follow-up. Altogether 19 (11%) were positive for PTSD at some point in the 1-yr follow-up study. Five of the initial 14 had remitted by the 1-yr follow-up. No data were provided on potential predictors of remission. Feinstein and Dolan (1991), also in the UK, assessed 48 MVA victims hospitalized for leg fractures when they were initially hospitalized and at 6 weeks and 6 months post-MVA. Twelve (25%) met DSM-III-R (American Psychiatric Association, 1987) at the 6-week assessment while only 7 (14.6%) were still positive at 6 months. From the initial assessment, Impact of Event Scale (IES) scores (Horowitz, Wilmer & Alvarez, 1979), a measure of intrusion and avoidance symptoms, predicted clinical status at 6 months as did report of above average alcohol consumption at time of the initial assessment. Green, McFarlane, Hunter and Griggs (1993), in an Australian study, examined 24 hospitalized MVA victims at 4 weeks post-accident and again at 6 months and 18 months post-MVA. Eighteen (75%) were available at 18 months. Two participants (8.3%) met DSM-III-R criteria for PTSD at the initial assessment while 7 were subclinical. At 18 months, 5 of the 7 subclinical PTSD participants now met criteria for full PTSD (the other two cases were not available for reassessment). Brom, Kleber and Hofman (1993), in The Netherlands, assessed 151 MVA victims (out of 738 approached) with the Dutch version of the IES (Horowitz et al., 1979). Half received a brief treatment over the next 6 months, while half (n = 83) constituted a no-treatment control. Twenty-four percent of the latter dropped out. On average, the reduction in IES scores for the controls was from 17.4 to 7.4. No specific data on PTSD diagnoses was available but post-traumatic stress symptoms, as measured by the IES, clearly declined over time. In a preliminary report from our laboratory (Blanchard, Hickling, Taylor, Loos & Forneris, 1995) we reported on a 6-month prospective follow-up of 98 MVA victims who had sought medical attention as a result of the accident. Forty (41%) met DSM-III-R criteria for PTSD, based upon the CAPS (Clinician-Administered PTSD Scale) (Blake, Weathers, Nagy, Kaloupek, Klauminzer, Charey & Keane, 1990). At 6 months, 20 of the 40 initial PTSDs available for follow-up still met the full criteria for PTSD. No data on potential predictors of remission were reported. It is thus clear that relatively little information is available on prospective follow-up of MVA victims, and the two large-scale reports (Mayou et al., 1993; Brom et al., 1993) did not use a diagnostic instrument known to be sensitive to the diagnosis of PTSD. Moreover, almost nothing is known about factors which might be associated with remission of PTSD over the first 12 months of follow-up. Feinstein and Dolan identified an initial sub-syndromal form of PTSD as a risk factor among MVA victims with broken legs for developing delayed onset PTSD but gave no hints on factors related to remission. In these times of increasing cost consciousness in mental health, it would be of interest to know what factors (including participation in treatment) would predict the remission of PTSD over the first year and likewise, what factors are associated with symptom maintenance. With adequate knowledge on this point and limited treatment resources, one might engage in a triage effort, sending those at risk to remain ill for early treatment. In this paper, we report on the 1-yr prospective follow-up of 158 MVA victims who sought medical attention following the MVA and who were initially assessed 1 4 months post-MVA (thus allowing enough time for the patient to fully meet the temporal criterion for PTSD). The paper has three parts, one documents the clinical status of all Ss over the first year and an 18-month

One-year prospective follow-up o f M V A victims

777

follow-up on the part of the sample who had an initial diagnosis of PTSD. The second part presents data related to predictors of improvement or remission for 48 participants who initially met the criteria for full PTSD. The last part examines changes in subjective distress, as measured by standardized psychological tests, and in role impairment for those participants initially diagnosed with PTSD. METHOD

Participants We initially assessed 158 MVA victims, including 50 males and 108 females. Inclusion criteria were 17 yr of age or greater and that the individual had sought medical care within 48 hr of the MVA. The participants were initially assessed from 1-4 months after the MVA. They were reassessed 6 months and 12 months after the initial interview. Individuals who had met initial criteria for PTSD and who were still symptomatic at 12 months were also seen for an 18-month evaluation. Participants were paid $50 for participating in the initial assessment, $50 for the 6-month follow-up assessment and $75 for the 12-month follow-up assessment (and $50 for the 18-month follow-up). All gave written informed consent for all of the procedures. We were able to obtain data from 132 participants at the 1-yr follow-up representing 83.5% of the original sample. Of those for whom data are unavailable at 12 months, 12 had moved, left no forwarding address and were unreachable through family contacts, 13 refused or dropped out of the study at 6 months or at 12 months, and one had died. In Table 1 is the basic demographic information on three subsets of 12-month follow-up participants, subdivided by initial diagnostic status (see below) into full PTSD, sub-syndromal PTSD, and non-PTSD. Comparable information on the dropouts is also provided. Comparisons of completers to dropouts revealed that significantly more minority participants discontinued participation (Z2 (1, n = 158)= 16.63, P = 0.00005); there were no other significant differences. There was a nonsignificant (P = 0.09) trend for more initial full PTSDs to drop out.

Initial diagnosis Initial diagnosis for the presence of PTSD was by means of a structured interview developed by the National Center of PTSD (Boston Branch), the CAPS (Clinician Administered PTSD Scale; Blake et al., 1990). It has been shown to have adequate reliability and validity (Weathers & Litz, 1994). Four doctoral level assessors, each with over 5 yr of experience in assessing for PTSD in Vietnam veterans, administered the structured interview. One participated in the development of the CAPS: he trained the other 3 assessors. All interviews were tape recorded. Table 1. Demographic and diagnostic information on l-yr follow-up completers and drop-outs Initial diagnostic status of completers Full PTSD (n = 48)

Sub-syndroma[ PTSD (n = 42)

Non-PTSD (n = 42)

Total completer sample (n = 132)

Drop-outs (n = 26)

10/38 79.2%

14/28 66.7%

20/22 52.4%

44/88 66.7%

6/20 76.9%

34.8 (10.4) 19-60

35.9 (11.8) 17 65

38.1 (15.3) 17-71

36.2 (12.6) 17-71

31.7 (11.8) 18-73

Ethnici O' Caucasian/Minority (% Minority)

43/5 10.4%

40/2 4.8%

40/2 4.8%

123/9 6.8%

17/9 34.6%

Education Some college/H S or less (% college)

39/9 81.3%

30/12 71.4

31 / 11 73.8%

100/32 75.8%

17/9 65.4%

Marital status Married/not married (% married)

19/29 39.6%

19/23 45.2%

19/23 45.2%

57/75 43.2%

8/18 30.8%

Characteristic Sex (M/F) (%F) Age X (SD) Range

Initial diagnosis of drop-outs PTSD-% of initial sample Sub-PTSD~% of initial sample Non-PTSD--% of initial sample

14 (22.6%) 3 (6.6%) 9 (17.6%)

778

Edward B. Blanchardet al.

MVA victims were initially diagnosed using DSM-III-R (American Psychiatric Association, 1987) criteria. They were termed full PTSD (n = 62, 39.2% of initial sample) if they met the full criteria. If they met the criteria for Cluster B (one reexperiencing symptom) and either Cluster C (at least 3 avoidance and psychic numbing symptoms) or Cluster D (at least 2 hyperarousal symptoms), (but not both) they were termed sub-syndromal PTSD (n = 45, 28.5%). Finally, if they met the criteria for only one symptom Cluster or no Clusters, they were classified as non-PTSD (n = 51, 32.3%). As a local reliability check, an advanced doctoral student in clinical psychology listened to 15 randomly selected tapes and scored the CAPS without knowledge of the initial diagnosis. Kappa for diagnostic agreement was 0.810 (P < 0.0005). The mean correlation for the CAPS scores on the 17 individual symptoms used to diagnose PTSD was 0.975 (P < 0.0001). Follow-up assessment

All participants initially agreed to follow-up assessments 6 and 12 months after the initial assessment and were given explicit appointments. Approximately one week before the follow-up appointment, a packet of questionnaires was mailed to the participant with a reminder of his/her appointment. Follow-up telephone calls were placed to confirm or re-schedule appointments. Altogether, we were able to gather 6-month data on 145 (91.8%) MVA victims and 12-month data on 132 (83.5%). For the 18-month follow-up, we attempted to follow-up those initially diagnosed as PTSD, with special emphasis on those initial PTSDs who still met the full criteria at 12 months. We obtained 18-month interviews on 35 (73% of those available at 12 months) and on 13 of 16 (81.3%) of those who were still full PTSD at 12 months. We also followed up the 5 initial sub-syndromal cases who developed delayed onset PTSD during the first year. For the most part, follow-up assessments were completed by the same initial assessor and were conducted face-to-face. In two instances follow-up interviews were completed by telephone. At the follow-up interviews we adapted the LIFE (Longitudinal Interval Follow-up Evaluation) methodology to assess week-by-week changes in physical symptoms, effects of the MVA on driving, and individual symptoms of PTSD. This methodology utilizes personal events (e.g. birthdays, holidays, etc.) as anchors to help the respondent recall when noticeable changes occurred in symptoms. We also utilized the LIFE to assess psychosocial status (as a measure of role impairment) and status of Axis-I co-morbidity as well as any treatment, psychological or pharmacological, received during the interval. We also noted changes in litigation status and any new personal or family trauma. Finally, current status of PTSD symptoms was assessed at each visit using the CAPS-II (Blake et al., 1990). From the LIFE follow-up of individual symptoms of PTSD, it was possible to make a week-by-week diagnosis of full PTSD, sub-syndromal PTSD and non-PTSD. These data were transformed into month-by-month diagnoses on the basis that a participant was called positive for the month if he/she was positive for at least one week of that month. Each physical injury was assessed on a 0 to 3 scale as to whether, by the participant's report, it remained the same (3), was improved but still limited activity (2), was markedly improved but still noticeable (1) or completely remitted (0). A monthly index of physical injury was calculated by summing the scores for each injury described by the participant [range 0 to 3 x n (number of injuries)] and dividing by 3 × n. Thus, a participant who was completely healed would receive a score of 0 while someone who was not at all improved would receive a score of 1.0 (3xn + 3×n). Other initial information

The assessors gathered information on the MVA and the participant's reaction to it, including information on extent of physical injury to the participant and others involved in the MVA, on subjective reactions to the MVA, the effects on the participant's driving behavior, and status of legal and insurance issues, using a locally designed structured interview. Participants were asked about previous MVAs and reactions to them. If anyone was injured, an assessment of possible PTSD from that earlier accident was made. We also assessed for other previous trauma using the format of Breslau, Davis, Andreski and Peterson (1991).

One-year prospective follow-up of MVA victims

779

We assessed for pre-MVA and current Axis-I psychopathology using the SCID-NP (Spitzer, Williams, Gibbon & First, 1990a); we also assessed for possible Axis-II psychopathology using the SCID-II (Spitzer, Williams, Gibbon & First, 1990b). Two of the assessors had been trained in use of these instruments by personnel from NYS Psychiatric Institute. They, in turn, trained the other two assessors. Finally, we assessed for pre-MVA and current psychosocial functioning using the LIFE-Base, a structured interview developed by Keller, Lavori, Friedman, Nielsen, Endicott, McDonald-Scott and Andreasen (1987) for this purpose. The four assessors were trained in its use and the use of the LIFE by personnel at Brown University under Keller's direction. Using the LIFE-Base participants were assessed for performance in major role functions (work if employed 30 hr/week or more, school if a full-time student, or homemaking if neither of the other two applied) on a 1 (no impairment and high level of performance) to 5 (severe impairment, not able to meet demands of the role) scale. Relationships with all first degree relatives and spouse or partner on a 1 (very good, very close emotional relationship) to 5 (very poor, no emotional relationship, avoids family members) scale were assessed. Similar ratings were made for relationships with friends and participation in recreation. These ratings were repeated prospectively on a month-by-month basis as part of the LIFE follow-up interviews. For the purpose of this paper, we used the initial post-MVA assessment ratings and those for month 12. Furthermore, we used only one role rating (as described above); finally, we averaged the ratings of family relationships across all first degree relatives plus spouse or partner into a single family relationship variable. Psychological tests. We administered the Beck Depression Inventory (Beck, Ward, Mendelson, Mock & Erbaugh, 1961), the State-Trait Anxiety Inventory, (Spielberger, Gorsuch & Lushene, 1970) and the Impact of Events Scale (Horowitz et al., 1979) at each assessment occasion as measures of subjective distress. We also made Global Assessment Scale (GAS) ratings on each occasion. Degree of initial physical injury was scaled using the Abbreviated Injury Scale (American Association for Automotive Medicine, 1985) by a physician blind to diagnostic status who had all of the participant's reports of injury and treatment as well as basic demographic information. Any patient who had a diagnosable Axis-I disorder, including PTSD, was given an explicit referral for possible treatment, after the initial assessment. RESULTS The first question to be addressed by these results is what happens over time to the MVA victims in terms of their diagnostic status. Shown in Table 2 is a month-by-month frequency distribution of diagnoses (full PTSD, sub-syndromal PTSD, and non-PTSD) within each of the initial diagnostic sub-groups. Table 2. Month-by-month diagnostic status of M V A victims as a function o f initial diagnosis Initial Dx.

Initial

1

2

3

4

5

6

7

8

9

10

11

12

12-month CAPS

48 0 0

NS 47 0 l

NS 45 2 1

NS 40 4 4

** 33 7 8

*** 27 10 11

*** 21 12 15

*** 24 9 15

*** 22 l0 16

*** 22 8 18

*** 20 5 23

*** 20 6 22

*** 20 8 20

*** 16 7 25

0 42 0

* 1 37 4

*** l 28 13

*** 1 21 20

*** 0 16 26

*** 2 14 26

*** 2 12 28

*** 3 9 30

*** 5 7 30

*** 5 8 29

*** 4 6 32

*** 4 8 30

*** 4 7 31

*** 3 9 30

0 0 42

0 2 40

0 2 40

0 2 40

0 2 40

0 1 41

0 2 40

0 3 39

* 0 4 38

* 0 4 38

0 3 39

0 1 41

0 1 41

0 l 41

PTSD (n = 48) PTSD Sub Non

Sub-syndromal (n = 42) PTSD Sub Non

Non-PTSD (n = 42) PTSD Sub Non n

=

(132)

X2 * ** ***

~t 0.05 0.01 0.001

Critical value 3.841 6.635 10.827

E d w a r d B. B l a n c h a r d et al.

780

We compared the distribution in each month to the initial distribution using Guilford's Z-' Test for Difference between Correlated Proportions (Guilford, 1965, p. 242). For those initially diagnosed with PTSD, by month 3 of the follow-up (approximately 4--7 months post-MVA) a significant proportion (8/48, 16.7%) had remitted partially or completely. At 7 months the proportion showing full or partial remission was 50% and by the 1-yr CAPS interview, two-thirds had remitted partially or in full with slightly over half now meeting the criteria for non-PTSD. Examining those initially diagnosed with sub-syndromal PTSD, one sees two phenomena: first, a generally more rapid rate of remission with significance reached in month 1, and 30 out of 42 (71.4%) fully remitted by month 7. Interestingly, after month 7 there is only limited improvement on a group basis, since 12 of 42 (28.6%) have not remitted to the non-PTSD level by the l-yr CAPS interview. The second phenomenon is the deterioration of some initial sub-syndromal PTSD Ss over time. Altogether, 7 different participants showed enough deterioration (increased symptoms) to move into the full PTSD category at some point over the 1-yr follow-up. (A detailed analysis of these participants with delayed onset PTSD is available separately (Buckley, Blanchard & Hickling, 1996).) At months 8 and 9 the proportion of initial sub-syndromal PTSDs who had deteriorated is significant. This may coincide with the anniversary of the MVA. Finally, one sees among those with an initial diagnosis of non-PTSD occasional deterioration to sub-syndromal PTSD. This became significant in months 8 and 9, mirroring the deterioration of the sub-syndromal PTSDs, and again perhaps coinciding with an anniversary of the MVA. No initial non-PTSD deteriorated to the criteria for full PTSD.

18-month jollow-up data In Table 3 are the month-by-month diagnoses for the 35 initial PTSDs followed from month 12 to month 18 and the 5 initial sub-syndromal PTSD Ss who developed delayed onset PTSD and who were thus followed for the additional 6 months. Examining the data on the 35 initial PTSDs, including the 13 of 16 (81.3%) who still met the full criteria for PTSD based on the CAPS interview at 12 months, the main finding is of stability of diagnosis. There is very little additional improvement over the next 6 months among those with full PTSD at the 1-yr follow-up point and some slight deterioration at some points. In no instances were the change in frequencies significant. For the 5 sub-syndromal PTSDs who were followed, 2 of the 3 with delayed onset PTSD had remitted at the 18-month point and one other had become essentially symptom free. Because of the reduced sample sizes, no analysis of predictors was made on these 12 to 18-month data. Prediction of rem&sion There are two prediction problems for this portion of the results. For those participants with an initial diagnosis of PTSD, we would like to be able to predict two things: (a) their clinical status (full PTSD or remission) (we have collapsed the sub-syndromal PTSDs (partial remission) and the non-PTSD (full remission)) and (b) their degree of post-traumatic stress symptoms (PTSS) as indicated by the 1-yr CAPS score. For the first part, logistic regression is the appropriate statistic since the criterion is dichotomous. For the second part we have used stepwise multiple regression.

Table 3. Month-by-month diagnoses for initial PTSDs for months 13 through 18 Month Dx. initial

Dx. 12-month

13

14

15

16

17

18

18-month CAPS

13 5 17

14 7 14

14 6 15

14 5 16

13 4 18

12 5 18

12 6 17

12 5 18

3 2 0

3 2 0

3 2 0

3 2 0

2 3 0

2 3 0

2 3 0

I 3 I

Initially diagnosed with PTSD PTSD Sub Non

35 0 0

Initially sub-syndromal PTSD PTSD Sub Non

0 5 0

O n e - y e a r p r o s p e c t i v e f o l l o w - u p o f M V A victims

781

Restriction of predictor battery. As is obvious from the description of the initial and follow-up data gathering efforts, we have a very large potential battery of predictors. (Altogether, we examined 101 potential predictors since there is almost no data available to guide the selection.) To help reduce this battery we initially calculated the simple bivariate correlation coefficient between each predictor and the 1-yr CAPS score and the point biserial correlation between each predictor and l-yr clinical status (full PTSD or not). Only those predictors which individually accounted for 4% of variance in one of the two criteria were retained. In addition, we have specified for the multiple regression analysis that each new variable added to the equation must account for at least 3% of new variance (AR 2 = 0.03 or higher). Prediction of l-yr clinical status of initial PTSDs In Table 4 is the result of logistic regression analysis to identify variables which would predict who continued to meet the full criteria for PTSD and who had shown full or partial remission. As indicated in Table 4, using base rates, one would be correct 66.7% of the time if everyone was predicted to be a remitter. Examining the results in Table 3, one can see that prediction is improved to 79.2% correctly classified using three variables. Moreover, prediction of who continues to have PTSD improves from 0 to 62.5%. To illustrate how these three variables work, one needs to remember that logistic regression yields coefficients for an equation to predict the natural logarithm (In) of an odds ratio (OR), that is the probability of a S being classified PTSD or less than PTSD. Using base rates from our data on initial PTSDs, the O R of PTSD to less than PTSD is 0.333/0.667 or 0.50 to 1; conversely the odds of being less than PTSD is 2.0 to 1. As an example, consider the situation in which a S has values for the three predictors of the mean score for the sample plus one standard deviation (rounded to a whole number for the two CAPS items). This yields an equation: In (OR) = -6.6571 + 0.5251(3.85 + 2.43 = 6.28, rounded to 6) + 0.4723(1.92 + 2.43 = 4.35, rounded to 4) + 0.0379(70.8 + 25 = 95.8, rounded to 96) or In (OR) = 2.0205 O R = 7.5421. This means that the likelihood of such an MVA victim, who initially met the criteria for PTSD, still meeting the criteria 1 yr later is about 7.5 to 1. As a second example in the opposite direction, consider the case in which the variables are at the mean score for the sample minus one standard deviation (again rounded to a whole number for the two CAPS items). This yields a second equation: In ( O R ) = -6.6571 + 0.5251(3.85 - 2 . 4 3 = 1.42, rounded to 2) + 0.4723(1.92 - 2.43 = - 0 . 5 1 , rounded to 0) + 0.0379(70.8 - 25 = 45.8, rounded to 46) or In (OR) = - 1 . 9 7 2 9 O R = 0.1391. This means that the likelihood of such an MVA victim, who initially met the criteria for PTSD, still meeting criteria 1 yr later is 0.1391 or about 1 to 7. Turned the other way, such an individual is about 7 to 1 to have remitted in full or in part. Table 4. Logistic regression to predict I-yr clinical status among initial PTSDs Percent correctly identified PTSD

Less than PTSD

Overall

0 50.0 56.3 62.5

100 78.1 90.6 87.5

66.7 68.8 79.2 79.2

BRT 34, tO--B

B 0.5251 0.4723 0.0379 -6.6571

S.E. 0.1944 0.1889 0.0186 2.0234

Wald

df

. . . 7.294 6.254 4.141 10.824

.

Sig.

Predictor

0.0069 0.0124 0.0419 0.0010

Base rate Initial CAPS-13, Irritability Initial CAPS-I1, Foreshortened Future Vulnerability in Auto. at Initial Assessment Constant

. I I I I

E d w a r d B. B l a n c h a r d et al.

782

Table 5. Summary of final multiple regression to predict post-traumatic stress symptoms at 1 yr in initial PTSDs

Multiple Variable Alcohol abuse at time of initial assessment Sum of initial CAPS hyperarousal symptoms Pre-MVA Axis-I! disorder Sum of initial CAPS avoidance symptoms Constant

Change

Sig. of F

B

13

t

P

R

R2

in R 2

for change

85.53 1.06 18.07 1.12 - 11.64

0.476 0.347 0.275 0.223

4.97 3.62 2.91 2.29 - 1.51

0.000 0.0008 0.0056 0.0268 0.1394

0.563 0.712 0.770 0.798

0.317 0.506 0.593 0.637

-0.189 0.087 0.044

-0.0001 0.0039 0.0268

One-year post-traumatic stress symptoms of initial PTSDs Shown in Table 5 is the final multiple regression equation for predicting l-yr CAPS scores as indication of PTSS for those participants initially diagnosed with PTSD. One can determine from Table 5 that with four variables we are able to account for 64% of variance in 1-yr CAPS scores, representing a multiple R of 0.798. Cross-validation. As is well-known, any multiple regression equation capitalizes to some degree on chance associations present in the sample. One way to correct for these chance associations is through cross-validation. We have approached this issue in the following way: we randomly selected half (n = 24) of our original sample of initial PTSDs, leaving a second randomly selected sub-sample of 24. We then calculated multiple regression equations, using the steps described above, with the criterion of total CAPS score. This yielded two sets of significant predictors and two multiple Rs (one for each half). This whole process was repeated three times to yield a total of six sets of predictors and six multiple Rs. Those predictors which regularly appear in the various regression equations are probably the ones in which one should have the most faith. A summary of these six cross-validations, and of the variables selected in the initial analysis (Table 5), is presented in Table 6. Examining Table 6, one can see that three of the variables identified in the stepwise multiple regression analysis of the entire sample (Table 5), namely presence of alcohol abuse at the time of the initial assessment, presence of any Axis-II Personality Disorder at the initial assessment, and sum of CAPS scores on the Cluster D (Hyperarousal) symptoms, all load regularly in the cross validation. Four other potential predictors also emerge two or more times in the cross validation. Among the four are the CAPS score on Symptom 11, sense of foreshortened future, the CAPS score on Symptom 13, irritability, and the perceived vulnerability when driving or a passenger in a vehicle, all three of which load in the logistic regression, and the initial AIS score as an indication of degree of physical injury. The one variable from the regression equation in Table 5 which does not show up in the cross-validation is initial CAPS score on Avoidance symptoms. Thus, it may be a less important indicator of 1-yr status. Finally, all of the cross-validation regressions yield multiple R values in the same range as that in Table 5, or higher. The role of treatment in remission of PTSD. Given the results from Kessler et al.'s (1995) retrospective survey on the apparent early importance of treatment in remission of PTSD, we entered treatment participation (pharmacological, psychological, both pharmacological and psychological) as a potential predictor. Of the 48 participants with an initial diagnosis of PTSD, 23 entered some form of treatment. None of the treatment participation predictors was a

Table 6. Summary of predictors identified in cross-validation analysis Cross-validation sample Predictor variable

1

Alcohol abuse at time of initial assessment Pre-MVA Axis-II disorder Sum of initial CAPS hyperarousal Sxs. Sum of initial CAPS avoidance Sxs. Relations with family, post-MVA (social support) Initial CAPS, Symptom 11, Foreshortened Future Initial CAPS, Symptom 13, Irritability Abbreviated Injury Scale score Vulnerability in auto at initial assessment Subjective estimate of functioning at initial assessment Final multiple R

+ + +

2

3 + +

4

5

6 +

+ +

+ + + +

0.949

0.826

+ + +

0.927

+

+ + 0.838

+ + 0.785

0.800

Number of times appears

Variables found in Table 5

3 3 2 0 1 3 2 3 2 1

* * * *

0.798

O n e - y e a r prospective follow-up o f M V A victims

783

Table 7. Ratings of role performance variables for participants with initial PTSD based on 12-month clinical status Initial assessment value

12-month

Variables Major role Function ~ (work, school homemaking)

Relations with family2 (average of all 1st degree relatives plus spouse/partner) Relations with friends Recreation participation

12-month assessment value

diagnosis

~"

(SD)

~"

(SD)

PTSD (n = 16) Sub-PTSD (n = 7) Non-PTSD (n = 25) PTSD Sub-PTSD Non-PTSD PTSD Sub-PTSD Non-PTSD PTSD Sub-PTSD Non-PTSD

3.4

(1.2)

2.8

(1.7)

2.7

(1.6)

1.4

(0.8)

3.0

(1.4)

2.2

(1.6)

2.2 2.3 2.3 2.0 2.4 2.3 3.9 3.1 3.0

(0.8) (1.0) (1.0) (1.3) (1.6) (1.2) (0.8) (1.9) (1.4)

2.5 2.1 2.2 2.6 2.1 2.0 3.3 2.1 2.3

(0.7) (0.6) (I.0) (1.4) (1.1) (I.2) (1.2) (1.3) (1.3)

JRating l - - n o impairment, high level of performance; 3--mild impairment; 5--severe impairment; 2Rating l--very good, very close emotional relationship; 3--fair, believes relationship needs to be closer; 5--very poor. no emotional closeness, avoids family member.

significant correlate of 12-month clinical status or 12-month CAPS score. For example, the correlations of any treatment with 12-month clinical status was 0.030 and with 12-month CAPS score was 0.149 (P = 0.313). These results seem to justify combining treated and untreated participants in all of the analyses.

Collateral changes in participants initially diagnosed with PTSD Shown in Table 7 are the mean ratings on role performance from the initial and 12-month follow-up assessments for those participants initially diagnosed with PTSD. They are subdivided by 12-month clinical status into full PTSD (no remission), sub-syndromal PTSD (partial remission) and non-PTSD (full remission). The mean scores from the psychological tests, as measures of subjective distress, from the initial and 12-month assessments are shown in Table 8. Each set of scores was initially subjected to a 2-way (Sub-group x Time) M A N O V A which was followed as appropriate by univariate repeated measures ANOVAs. Since no comparable data are available, in the interest of fully exploring the data set, we calculated correlated t-tests for each diagnostic sub-group for each measure. These results are shown in Tables 7 and 8. Examining the role performance ratings, we found an overall main effect of Time [F(4,41) = 4.90, P = 0.003] but no main effect of Sub-group or interaction of Sub-group x Time. Univariate Table 8. Psychological test scores for participants with initial PTSD based on 12-month clinical status Initial assessment Measure

12-month assessment

12-month diagnosis

X

(SD)

k

(SD)

PTSD Sub Non

20.6 13.4 13.0

(11.4) (2.2) (8.1)

20.3 10.9 9.6

00.5) (7.2) (10.3)

PTSD Sub Non

41.7 36.4 30.2

(14.2) (16.2) (17.4)

39.5 29.0 10.5

(17.4) (9.7) (9.5)

PTSD Sub Non

75.0 64.4 58.3

(18.3) (18.6) (16.2)

77.1 59.7 50.4

(16.0) (18.8) (16.4)

PTSD Sub Non

64.4 58.1 57.4

(18.6) (9.1) (10.9)

76.3 60.7 53.4

(16.3) (12.5) (16.1)

PTSD Sub Non

57.0 55.0 57.0

(8.9) (13.2) (12.7)

56.8 63.0 70.4

(11.6) (7.1) (14.8)

Beck Depression Inventory

Impact of Events

State-Anxiety

Trait-Anxiety

Global Assessment Scale

784

Edward B. Blanchardet al.

ANOVAs revealed significant improvement over time (Time main effects) on Major Role Function [F(1,41)=9.27, P = 0 . 0 0 4 ] and participation in Recreation [F(I,41)= 11.83, P = 0 . 0 0 1 ] but no effects on Relationships with Family or Friends. There were no significant differential improvements among the 3 sub-groups on any of the role performance variables. Turning to Table 8, we find a significant overall main effect of Sub-group [F(10,80) = 2.87, P = 0.004, Pillais], of Time, [F(5,39) = 22.6, P < 0,001] and of the interaction of Sub-group x Time [F(10,80)= 3.09, P = 0.002, Pillais]. Follow-up univariate tests revealed significant main effects of Sub-group for the BDI, State-Anxiety, Trait-Anxiety and IES (all Ps < 0.006), significant main effect of Time for only the IES [F(1,43) = 18.7, P < 0.001] and GAS [F(1,43) = 10.72, P = 0.002] and significant interactions for the IES [F(2,43)- 8.27, P = 0.001], Trait-Anxiety [F(2,43) = 7.02, P = 0.002] and GAS [F(2,43) = 5.24, P = 0.009]. Using orthogonal contrasts, we find the group who remained PTSD at 12 months was higher than the other 2 groups combined for the BDI ( P = 0.002), State-Anxiety (P<0.001), Trait-Anxiety (P = 0.005), and IES (P = 0.002). The correlated t-test reveal that those who remained with PTSD increased significantly on Trait-Anxiety [t(14)= 3.89, P=0.002]. Those initial PTSDs who had fully remitted had significantly lower IES scores [t(23) = 10.1, P < 0.001] at 12 months their GAS score increased significantly [t(23) = 4.53, P < 0.001]. The overall picture on the measures of psychological distress is that those with initial PTSD who remitted fully or in part were less distressed initially than those who continue with PTSD and that those with continuing PTSD continue to be as distressed 12 months later as they were at the initial assessment (and on Trait-Anxiety score significantly higher at the 12-month follow-up). DISCUSSION As with the PTSD from other trauma, the PTSD resulting from MVAs shows a sizable degree of remission in the year plus (13-17 months) following the accident. Of the 48 individuals with PTSD at the initial assessment, 1-4 months post-trauma, about half have shown full or partial remission (to sub-syndromal PTSD) by 6 months and two-thirds by the 12-month follow-up point. Among those who have not remitted, at least in part, by 12 months (n = 13), there was very little additional improvement over the next 6 months (out to 18 months post-initial assessment). Rothbaum and Foa (1993) have shown a similar (and possibly steeper) remission curve in their prospective follow-up study of sexual assault victims, whereas McFarlane's (1988) report on Australian fire fighters seemed to show a more gradual remission curve, and considerable delayed onset PTSD. Other prospective follow-up studies of MVA victims have shown some degree of remission: Mayou et al. (1993) found 5 of 13 (38.5%) MVA victims, who initially met criteria for PTSD at a point 3 months post-MVA, had remitted by a 12-month follow-up. By contrast, in Feinstein and Dolan's (1991) 6-month follow-up of MVA victims hospitalized for leg fractures, 5 of 10 (50%) with PTSD had remitted by the 6-month follow-up point, a remission rate comparable with ours. These prospectively determined rates of remission, approximately 40-70% by 1 yr, are much higher than the average for all PTSD cases in the study by Kessler et al. (1995). This latter study has a much larger sample (459 cases from all causes) and thus could represent a more stable estimate. However, the difference could also arise as a difference between retrospective follow-up (Kessler et al., 1995) and prospective follow-up (our study and all of the others reviewed). Also, like other prospective studies of MVA victims, we find some delayed onset PTSD. Mayou et al. (1993) found 6/171 (3.5%) cases of delayed onset, whereas Green et al. (1993) found 5/24 (20.8%) cases developed delayed onset PTSD over their 18-month follow-up. Our rate was 7 cases in 132 (5.3%) at 12 months. Like Green et al. (1993) all of our delayed onset cases emerged from MVA victims who initially met the criteria for sub-syndromai ('sub-clinical' in Green et al.'s terminology) PTSD. One cannot tell from Mayou et al.'s (1993) report what the initial status of their delayed onset cases would be. Examining the existing literature of MVA victims, one finds almost no attention paid to factors associated with remission of PTSD. The logistic regression (Table 4) identified 3 variables which predict remission: relatively lower scores on 2 of the 17 symptoms which make up the syndrome

One-year prospective follow-up of MVA victims

785

(irritability and sense of foreshortened future) and relatively lower score on the degree of vulnerability the participant felt when either a passenger or driver in a vehicle at the time of the initial assessment. Different predictors emerge if one is interested in how symptomatic (degree of PTSD as measured by the 12-months CAPS) an individual is at 12 months. Among the stable predictors of higher PTSD scores (less remission) are two variables which could be considered predisposing factors (in that they were present prior to the MVA), a history of having met the criteria for alcohol abuse at the time of the initial assessment and a history of meeting the criteria for any Axis-II disorder. Other relatively stable predictors (of less remission) are higher scores on the hyperarousal (Cluster D) symptoms (especially including irritability), higher scores on the foreshortened future symptom, initial degree of physical injury as measured by the AIS, and perceived vulnerability in a vehicle after the MVA. (Three of these entered the logistic regression equation as described above.) Interestingly, the initial reexperiencing symptoms and avoidance symptoms (tapped by the IES scale) do not predict nor does the initial degree of fear, terror or helplessness. These latter variables have been shown to predict who initially develops PTSD in other studies of MVA victims (Mayou et al., 1993; Feinstein & Dolan, 1991; Blanchard, Hickling, Taylor, Loos & Forneris, 1996) but do not predict remission. Although there is scant literature on the topic, we do not find that the presence of personality disorder and/or previous alcohol abuse are associated with poor outcome, surprisingly. Certainly, diverse treatment studies have shown that the presence of either of these two can be associated with poorer responses to treatment. In terms of the physical injury variable, we would have expected, based on the simple bivariate correlations, one of the physical injury quotient variables to have predicted remission (or lack thereof). Instead, the sheer extent of initial physical injury is the predictor. This stands in contrast to reports by Feinstein and Dolan (1991) who did not find such an association nor did Mayou et al. (1993). We did not find that participating in any form of mental health treatment in the first year post-trauma was differentially associated with remission or with a reduction in PTSD symptoms. This finding would be consistent with the results from MVA victims reported by Brom et al. (1993). These results are contrary to the highly significant effect of treatment reported in Kessler et al.'s (1995) retrospective follow-up. Our findings may be trauma-specific. Alternatively, the retrospective (Kessler) versus prospective (present study) difference may account for the difference. We do not wish to emphasize the lack of a treatment response because: (i) there was not random assignment to treatment versus no treatment, participants self-selected; and (ii) we cannot evaluate the adequacy of either the drug or psychological treatment our participants received. Our results, which show an apparent asymptote in recovery after 1 yr, and a less rapid rate of recovery from PTSD after about 7 months, suggest that these individuals might benefit from a very focused, PTSD specific treatment. We are currently trying to develop such a treatment program. The collateral psychosocial results (impairment in role functioning and subjective distress) suggest two things: those MVA victims who develop PTSD initially but who are relatively less impaired and less distressed initially are more likely to improve on clinical status over the year. Likewise, those who initially meet the criteria for PTSD after the MVA, and who continue to meet the full PTSD criteria a year later are noticeably impaired in major role function and participation in recreational activities initially and remain impaired. Their relationships with friends even deteriorate (albeit not significantly) over the year follow-up. This suggests that some friends may have difficulty remaining supportive of a symptomatic individual over time. Their level of subjective distress, as measured by standardized psychological tests, remains fairly stable over the year (and they actually increase significantly on trait-anxiety). There are, of course, certain limitations to this research. Our initial MVA sample is not a random sample of all MVA victims or even of all injured MVA victims. While such a sample would be desirable in terms of generalizing the results, human Ss constraints restrict us to studying those MVA victims who were willing to be studied, namely volunteers. We lost 16% of our initial sample; roughly half refused and half disappeared despite our efforts to maintain contact. The retention rate is comparable with other reports on MVA victims. The one significant difference between dropouts and our sample was in terms of ethnicity. We did a poorer job of retaining minorities

786

Edward B. Blanchard et al.

a n d h a v e n o r e a d y e x p l a n a t i o n for this. T h e s e losses w e r e m o r e o w i n g to i n a b i l i t y to t r a c k p a r t i c i p a n t s t h a n to o u t r i g h t refusals. D e s p i t e these l i m i t a t i o n s we believe we p r o v i d e a useful p i c t u r e o f the s h o r t t e r m n a t u r a l h i s t o r y o f A m e r i c a n M V A v i c t i m s w i t h initial P T S D . Acknowledgements--This research was supported in part by a grant from NIMH, MH-48476. We acknowledge the statistical assistance of Dr James Jaccard. REFERENCES American Association for Automotive Medicine (1985). The Abbreviated Injury Scale (revision). Des Plaines, IL: American Association for Automotive Medicine. American Psychiatric Association (1987). Diagnostic and Statistical Manual of Mental Disorders (3rd edn, revised). Washington, DC: American Psychiatric Press. American Psychiatric Association (1994). Diagnostic and Statistical Manual of Mental Disorders (4th edn). Washington, DC: American Psychiatric Press. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J. & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 5, 561-571. Blake, D., Weathers, F., Nagy, L., Kaloupek, D., Klauminzer, G., Charey, D., & Keane, T. (1990). Clinician-Administered PTSD Scale (CAPS). Boston, MA: National Center for Post-Traumatic Stress Disorder, Behavioral Science Division-Boston VA. Blanchard, E. B., Hickling, E. J., Vollmer, A. J., Loos, W. R., Buckley, T. C., & Jaccard, J. (1995). Short-term follow-up of post-traumatic stress symptoms in motor vehicle accident victims. Behaviour Research and Therapy, 33, 369-377. Blanchard, E. B., Hickling, E. J., Taylor, A. E., Loos, W. R., & Forneris, C. A. (1996). Who develops PTSD from motor vehicle accidents? Behaviour Research and Therapy, 34, 1-10. Breslau, N., Davis, G. C., Andreski, P. & Peterson, E. (1991). Traumatic events and post-traumatic stress disorder in an urban population of young adults. Archives of General Psychiatry, 48, 216-222. Brom, D., Kleber, R. J. & Hofman, M. C. (1993). Victims of traffic accidents: Incidence and prevention of post-traumatic stress disorder. Journal of Clinical Psychology, 49, 131-140. Buckley, T. C., Blanchard, E. B. & Hickling, E. J. (1996). A prospective examination of delayed onset PTSD secondary to motor vehicle accidents. Journal of Abnormal Psychology, in press. Davidson, J. R. T., & Foa, E. B. (Eds.) (1993). Post-Traumatic Stress Disorder: DSM-IV and Beyond. Washington, DC: American Psychiatric Press. Feinstein, A. & Dolan, R. (1991). Predictors of post-traumatic stress disorder following physical trauma: An examination of the stressor criterion. Psychological Medicine, 21, 85-91. Green, B. L. (1994). Psychosocial research in traumatic stress: An update. Journal of Traumatic Stress, 7, 341-362. Green, M. M., McFarlane, A. C., Hunter, C. E. & Griggs, W. M. (1993). Undiagnosed post-traumatic stress disorder following motor vehicle accidents. The Medical Journal of Australia, 159, 529-534. Green, B. L., Lindy, J. D., Grace, M. C., Gleser, G. C., Leonard, A. C., Korol, M. & Winget, C. (1990). Buffalo Creek survivors in the second decade: Stability of stress symptoms. American Journal of Orthopsychiatry, 60, 43-54. Guilford, J. P. (1965). Fundamental Statistics in Psychology and Education (p. 242). New York: McGraw Hill. Horowitz, M. J., Wilmer, N. & Alvarez, N. (1979). Impact of Events Scale: A measure of subjective stress. Psychosomatic Medicine, 41, 209-218. Keller, M. B., Lavori, P. W., Friedman, B., Nielsen, E., Endicott, J., McDonald-Scott, P. & Andreasen, N. C. (1987). A longitudinal interval follow-up evaluation: A comprehensive method for assessing outcome and prospective longitudinal studies. Archives of General Psychiatry, 44, 540-548. Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M. & Nelson, C. B. (1995). Post-traumatic stress disorder in the National Comorbidity Survey. Archives of General Psychiatry, 52, 1048-1060. Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan, B. K., Marmar, C. R., & Weiss, D. S. (1988). National Vietnam veterans readjustment study advance data report: Preliminary findings from the National Survey of the Vietnam generation. Executive Summary. Washington, DC: Veterans Administration. Mayou, R., Bryant, B. & Duthie, R. (1993). Psychiatric consequences of road traffic accidents. British Medical Journal, 307, 647-651. McFarlane, A. C. (1988). The longitudinal course of posttraumatic morbidity: The range of outcomes and their predictors. The Journal of Nervous and Mental Disease, 176, 30-39. Rothbaum, B. O. & Foa, E. B. (1993). Subtypes of post-traumatic stress disorder and duration of symptons. In Davidson, J. R. T. & Foa, E. B. (Eds.) Post-Traumatic Stress Disorder: DSM-IV and Beyond (pp. 23-35). Washington, DC: American Psychiatric Press. Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). STill Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press. Spitzer, R. L., Williams, J. B. W., Gibbon, M. & First, M. B. (1990a). Structured Clinical Interview for DSM-III-R, Non-patient Edition (SCID-NP, Version 1.0). Washington, DC: American Psychiatric Press. Spitzer, R. L., Williams, J. B. W., Gibbon, M. & First, M. B. (1990b). Structured Clinical Interview for DSM-III-R Personality Disorders (SCID-II, Version 1.0). Washington, DC: American Psychiatric Press. Weathers, F. W. & Litz, B. T. (1994). Psychometric Properties of the Clinician-Administered PTSD Scale. CAPS-I. PTSD Research quarterly, 5, 2-6. Wing, J. K., Cooper, J. E., & Sartorious, N. (1974). Measurement and Classification of Psychiatric Symptoms. Cambridge, U.K.: Cambridge University Press.