A Risk Score for Predicting Post-traumatic Stress Disorder in Adults in a Chinese Earthquake Area

Abstract

OBJECTIVES:

This study aimed to identify risk factors for post-traumatic stress disorder (PTSD) and develop a risk score model for predicting PTSD in adults in a Chinese earthquake area.

Methods:

Questionnaires covering demographic information, earthquake experience and social support were administered to subjects experiencing a major earthquake. The PTSD Checklist-Civilian Version questionnaire was used for PTSD diagnosis. Subjects were randomly assigned to training (70%) or validation (30%) subsets. A risk score model for predicting PTSD risk was established, based on logistic regression of PTSD risk factors that were significant on univariate analysis of the training data.

Results:

In total, 9556 subjects completed questionnaires; PTSD prevalence was 4.5%. Divorced or widowed status, various adverse earthquake events and low levels of social support were identified as risk factors for PTSD. When tested in the validation dataset, the risk score model had good discriminative power and a good fit between predicted and observed values.

Conclusions:

The risk score shows an acceptable predictive value and may be useful for early prediction of PTSD, in Chinese earthquake areas.

Keywords

Earthquake Post-traumatic stress disorder Risk factors Risk score

Introduction

On 12 May 2008 at 14:28 h, an earthquake of magnitude 8.0 on the Richter scale hit Wenchuan County, Sichuan, China. It caused extensive damage to residential areas and buildings. The death toll and declared missing exceeded 90000, and over 400000 people were injured. In Sichuan and neighbouring Gansu and Shaanxi provinces an estimated 5 million buildings collapsed, 21 million buildings were damaged and about 10 million people were made homeless.

Many people who experience disasters develop mental and physical illnesses;¹ in the last decade, post-traumatic stress disorder (PTSD) has been the most commonly studied of these. PTSD is probably the most frequent and debilitating psychological disorder to occur after traumatic disasters and events.^2,3 However, not all individuals exposed to trauma will develop PTSD; the reported incidence of PTSD varies between 1.5% and 67%, depending on the nature of the event, the length of time after the event, the sample used and other factors.^{4– 6}

After a person experiences trauma, physiological, societal, economic and other factors may collectively contribute to the development of PTSD. A number of studies have been conducted to describe the risk factors for PTSD after natural disasters. The risk factors associated with this disorder vary from study to study and include sex, age, severity of injury, financial loss, social support and previous mental illness.^{7 – 9} Although several risk factors for PTSD have been identified, the cumulative risk rendered by their combination is unknown. Therefore a global assessment of the impact of these variables on the development of PTSD is needed. However, few studies have attempted to incorporate these risk factors into a risk score. Such an evaluation tool would be of great importance in enabling psychiatrists to stratify individuals after a disaster, according to their PTSD risk. By predicting the absolute risk for an individual, earthquake victims who are likely to develop PTSD could be promptly identified, enabling the implementation of timely measures to protect their health.

In the present study, stepwise logistic regression was used to identify a group of risk factors associated with PTSD among victims of the 2008 Sichuan earthquake. These factors were then incorporated into a risk score model for the assignment of risk of PTSD.

Subjects and methods

Subjects

Subjects were selected using a multistage stratified-cluster sampling method. First, counties affected by the 2008 Sichuan earthquake were classified into three zones depending on the severity of destruction. The counties in each zone were numbered and three counties were selected at random from each zone. A systematic sampling approach was then used to select randomly 63 of the temporary prefabricated shelters provided by the government after the earthquake, in the urban and rural areas of these nine counties. Adults (aged ≥ 18 years) living in the selected shelters, who had been living in the affected area at the time of the earthquake, were invited to participate in the study.

The study was conducted at West China Hospital, Sichuan University, Sichuan, China. Written informed consent was obtained from all study participants. The study protocol was approved by the Institutional Review Board of West China Hospital, Sichuan University, Sichuan, China.

Assessments

Assessment of the earthquake victims took place in December 2008, ∼6 months after the earthquake. According to criteria for PTSD set out in the Diagnostic and Statistical Manual of Mental Disorders (4th edition) (DSM-IV),¹⁰ symptoms usually develop within the first 3 months after the traumatic event and will then last ≥ 3 months. Therefore, 6 months after the earthquake is an appropriate time for assessment of earthquake victims. In addition, some research has indicated that PTSD may be more accurately diagnosed at 6 months,¹¹ and the 6-month timepoint has been selected as the investigation time in other studies of PTSD.^12,13

All interviewers participated in a 2-day training programme to ensure that they had the same understanding of the questionnaires. They administered several self-report questionnaires to the study participants face to face, covering earthquake experience, social support and symptoms of PTSD.

Demographic data

Background information on age, sex, marital status, occupation, nationality, education level, income level and residency was recorded.

Earthquake experience

A survey of exposure to traumatic events (including being buried/injured, amputation, loss of family members, loss of colleagues/friends, property loss, witnessing death and participation in the rescue work) was completed.

Social support

The Social Support Rating Scale¹⁴ was used to assess the current social support available to the subject, such as support from parents, teacher, relatives and friends. This scale includes 10 items measuring objective support, subjective support and support utilization. The maximum score is 66. This questionnaire has been shown to have good validity and reliability in Chinese populations.^15,16

PTSD symptoms

The PTSD Checklist-Civilian Version (PCL-C)¹⁷ was used to diagnose PTSD. The PCL-C was constructed in accordance with DSM-IV¹⁰ and includes 17 items scored from 0 to 4 (0, none; 1, slight; 2, moderate; 3, severe; 4, extreme). The checklist assesses all the core symptoms of PTSD, including intrusion, avoidance and hyperarousal. A score of 50 on the PCL-C has been recommended as a cut-off point indicating high PTSD symptom load, and provides good diagnostic sensitivity (0.82) and specificity (0.83).¹⁷

Statistical Analyses

In order to develop risk scores and test their validity within a single dataset, all subjects were randomly assigned to either a training subset (70% of the subjects) or a validation subset (30% of the subjects) using computer-generated random numbers. An initial screening of the variables in the training dataset was performed using univariate analysis of the associations between each variable and the presence of PTSD. All factors with significance at the univariate level (P < 0.2) were entered as independent variables in a logistic regression model for PTSD. The coefficients from the logistic regression were then translated into a point-based system to give a risk score;¹⁸ the risk score was calculated by multiplying the regression coefficients by 10 and rounding off to the nearest integer.¹⁹ An overall score was calculated for each participant by adding together the score for each variable in the risk model.

The risk score was tested in the validation dataset. Model discrimination was assessed by calculating the area under the receiver operating characteristic (ROC) curve. To assess the agreement between the predicted and actual outcomes, the predicted risk based on the risk scores was divided into quintiles, and the predicted and observed PTSD rates were compared within each quintile. Differences between the predicted and observed rates were used to calculate the Hosmer–Lemeshow χ² statistic to measure the goodness of fit. All statistical analyses were performed using SPSS^® software, version 15.0 (SPSS Inc., Chicago, IL, USA).

Results

A total of 10 080 adults from nine different counties of the earthquake region were invited to participate in the study. Of these, 9 556 (95.0%) completed questionnaires. The mean age was 44.5 years (range 18 – 96 years); 4 332 (45.3%) subjects were male and 5 224 (54.7%) were female. Of the total, 9390 (98.3%) subjects were Han Chinese. Overall, 430 subjects had a PCL-C score indicative of PTSD, giving a PTSD prevalence rate of 4.5%. A total of 6 690 subjects were assigned to the training dataset; 2 866 were assigned to the validation dataset.

The variables with significance at the univariate level among the training dataset that were entered into the regression model and assigned a risk score are given in Table 1. Subjects with a higher score across these characteristics had higher predicted risks of PTSD. The predicted probability of PTSD can be estimated by adding the sum of the individual risk scores (Table 1) and reading the result from the nomogram given in Fig. 1. For example, a subject with a risk score of 65 would have a predicted PTSD probability of ∼0.70. Although higher scores are associated with a higher risk of PTSD, the relationship is not a linear one. An increase in the risk score from 0 to 40 increased the risk of PTSD to only 0.18, whereas an increase in the risk score from 40 to 60 resulted in a dramatic increase in the risk of PTSD, from 0.18 to 0.60.

Figure 1:

Predicted probability of post-traumatic stress disorder (PTSD) using a risk score based on data from 6690 adults in a Chinese earthquake area

Table 1:

Logistic regression results and assigned risk scores for variables that were significantly predictive of post-traumatic stress disorder, based on data from 6 690 adults in a Chinese earthquake area

Variable	Coefficient	P-value	Odds ratio	95% confidence intervals	Risk score
Marital status
Married	0.801	0.004	2.23	1.29, 3.86	8
Divorced	1.307	0.001	3.70	1.65, 8.28	13
Widowed	1.132	0.005	3.10	1.40, 6.86	11
Being buried/injured	0.687	0.000	1.99	1.40, 2.83	7
Amputation	1.852	0.002	6.37	1.99, 20.39	19
Loss of family member	1.081	0.000	2.95	2.27, 3.84	11
Loss of colleague/friend	0.281	0.035	1.32	1.02, 1.72	3
Property loss	0.640	0.003	1.90	1.24, 2.90	6
Witnessing death	0.375	0.013	1.45	1.08, 1.95	4
Participation in the rescue work	1.738	0.000	5.69	4.04, 8.00	17
Social Support Rating Scale¹⁴ score
≤ 25	0.863	0.000	2.37	1.61, 3.49	9
26 – 40	0.394	0.005	1.48	1.13, 1.95	4

The risk score was tested in the validation dataset. The area under the ROC curve was 0.767 (95% confidence intervals 0.724, 0.810), indicating that the risk score model had good discriminative power in the validation population.

Using the quintiles of predicted risk, the predicted and observed risks agreed closely (Fig. 2), demonstrating that the risk score had effective calibration. For example, a subject in the highest risk quintile had a mean predicted risk of 0.302, which agreed closely with the observed risk of 0.293. The observed and predicted risks were consistently within 2.5% of each other. The goodness of fit was confirmed by the Hosmer–Lemeshow χ² statistic (χ² = 9.746, 8 degrees of freedom, P = 0.283).

Figure 2:

Mean observed and predicted risks of post-traumatic stress disorder (PTSD) in the five quintiles of risk, predicted using a risk score based on data from 6690 adults in a Chinese earthquake area

Discussion

In the present study, certain marital statuses, various adverse earthquake events and low levels of social support were identified to be associated with PTSD after the 2008 earthquake, which is consistent with the findings of previous studies.^{20 – 22} The present results confirmed the importance of adverse earthquake events in the development of PTSD. Events such as bereavement, personal injury, amputation and witnessing death are known to increase levels of distress and fear in disaster victims.^{23 – 25} Such events may also be associated with the psychological stress of reconstructing the lives of the victims and their families. It is therefore important that victims are helped to undergo the mourning process and readjust to normal life after an earthquake.

In times of crisis, individuals may seek tangible or emotional support from others.^26,27 In the present study, most of the significant PTSD risk factors (except property loss) were concerned with emotional loss or suffering. Other variables such as income level were not significantly associated with PTSD; property loss was significantly related to PTSD and was entered into the risk model, but its risk score was much lower than that of most of the other variables. This suggests that tangible support is less important than emotional support.

A marital status of divorced or widowed and loss of a family member were assigned to a higher risk score (> 10), indicating that emotional support from family should be a special concern. In the family-centred Chinese culture, family is likely to be one of the most important sources of social support. Therefore, optimal emotional support, especially from family (including empathy, concern, affection, acceptance, intimacy, encouragement and caring) is important for the prevention of PTSD after an earthquake. The prompt reunion of family members (in addition to the provision of psychological intervention, particularly to bereaved and divorced or widowed victims) should be a priority in postearthquake management.

In the present study, the overall PTSD prevalence was 4.5% in a sample of 9 556 subjects. This rate is lower than that reported in other earthquake victims.^28,29 Differences in the severity of the disasters, the characteristics of the study subjects and the study methodology may account for the observed differences in PTSD prevalence. In addition, the Chinese government responded quickly and effectively to the damage and problems caused by the 2008 Sichuan earthquake, which may have played an important role in reducing the prevalence of PTSD.

A number of studies have identified associations between PTSD in earthquake areas and a variety of risk factors. However, the present study was the first to derive a risk score for individuals. It was able to identify individuals at risk and indicate the magnitude of that risk. Using several characteristics such as marital status, traumatic events due to the earthquake and the level of social support, the risk score effectively identified the PTSD risk of individuals in the earthquake area. The predicted risks showed successful discrimination, separating the highest and the lowest risk individuals, and successful calibration, with the predicted risks agreeing closely with the observed risks. The mathematically formulated predictive equation in the present study used the best combination of predictors rather than considering just one factor at a time.

The risk of PTSD is known to become higher as the number of risk factors present for an individual increases.^30,31 However, few studies have quantitatively examined the magnitude of the PTSD risk with different combinations of risk factors. The present study showed that if no or only one risk factor was present, the incidence of PTSD was very low, varying between ∼0.5% and 3%. However, with the presence of two or more risk factors, the incidence of PTSD rose sharply from 3% to 80%.

The model presented here has a potential role in predicting the possible occurrence of PTSD immediately after an earthquake. More significantly, the use of the scoring system developed in the present study allowed some refinement to the prediction of risk, as the model differentiated those at very low risk from those at intermediate or high risk. This process was more precise than assuming that victims above one cut-off point had an equally high risk. The risk score model for PTSD presented here will allow mental care providers to identify individuals at the highest risk of PTSD who may benefit most from management by psychiatrists, without referring the entire population. One way to use the risk score is to select a clinically important level of predicted risk (for example a ≥ 70% risk of progression to PTSD) and to refer only those subjects who score accordingly (in this case, those who score ≥ 65 points). We believe that our risk score can facilitate the early identification of individuals most at risk of developing PTSD, especially as the information included in the risk score is readily obtained. This is particularly important as early intervention is an effective method in the treatment and prevention of PTSD, and has been shown to be successful in reducing PTSD severity.³²

In conclusion, the risk score for PTSD presented here, based on a logistic regression model, shows an acceptable predictive value with favourable applicability, and may provide a tool for the early identification of adults at risk of PTSD in Chinese earthquake areas.

Footnotes

Acknowledgement

This work was partly funded by the National Basic Research Program of China (863 Program 2008AA022601 and 2008AA022603).

Conflicts of interest: The authors had no conflicts of interest to declare in relation to this article.

References

Norris

, Friedman

, Watson

: 60,000 disaster victims speak: part I. An empirical review of the empirical literature, 1981-2001. Psychiatry 2002; 65: 207–239.

Bromet

, Dew

: Review of psychiatric epidemiologic research on disasters. Epidemiol Rev 1995; 17: 113–119.

Schnurr

, Friedman

, Bernardy

: Research on posttraumatic stress disorder: epidemiology, pathophysiology, and assessment. J Clin Psychol 2002; 58: 877–889.

Garrison

, Bryant

, Addy

: Posttraumatic stress disorder in adolescents after Hurricane Andrew. J Am Acad Child Adolesc Psychiatry 1995; 34: 1193–1201.

Goenjian

, Najarian

, Pynoos

: Posttraumatic stress disorder in elderly and younger adults after the 1988 earthquake in Armenia. Am J Psychiatry 1994; 151: 895–901.

Davis

, Siegel

: Posttraumatic stress disorder in children and adolescents: a review and analysis. Clin Child Fam Psychol Rev 2000; 3: 135–154.

Armenian

, Morikawa

, Melkonian

: Loss as a determinant of PTSD in a cohort of adult survivors of the 1988 earthquake in Armenia: implications for policy. Acta Psychiatr Scand 2000; 102: 58–64.

Kwon

, Maruyama

, Morimoto

: Life events and posttraumatic stress in Hanshin-Awaji earthquake victims. Environ Health Prev Med 2001; 6: 97–103.

Naeem

, Ayub

, Masood

: Prevalence and psychosocial risk factors of PTSD: 18 months after Kashmir earthquake in Pakistan. J Affect Disord 2011; 130: 268–274.

10.

American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders, 4th edn. Washington, DC: American Psychiatric Association, 1994.

11.

Ursano

, Fullerton

, Epstein

: Acute and chronic posttraumatic stress disorder in motor vehicle accident victims. Am J Psychiatry 1999; 156: 589–595.

12.

Priebe

, Grappasonni

, Mari

: Posttraumatic stress disorder six months after an earthquake: findings from a community sample in a rural region in Italy. Soc Psychiatry Psychiatr Epidemiol 2009; 44: 393–397.

13.

Chou

, Su

, Chou

: Survey of psychiatric disorders in a Taiwanese village population six months after a major earthquake. J Formos Med Assoc 2005; 104: 308–317.

14.

Xiao

: Psychological Health Scale: Social Support Questionnaire. Beijing: The Journal of Chinese Psychological Press. 1998; pp 127-131.

15.

Gao

, Chan

, Mao

: Depression, perceived stress, and social support among firsttime Chinese mothers and fathers in the postpartum period. Res Nurs Health 2009; 32: 50

16.

Tan

, Luo

, Wen

: The effect of a disastrous flood on the quality of life in Dongting lake area in China. Asia Pac J Public Health 2004; 16: 126–132.

17.

Weathers

, Litz

, Herman

: The PTSD Checklist (PCL): Reliability, Validity, and Diagnostic Utility. San Antonio: Annual Convention of the International Society for Traumatic Stress Studies, 1993.

18.

Sullivan

, Massaro

, D'Agostino

Sr ,: Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med 2004; 23: 1631–1660.

19.

Glümer

, Carstensen

, Sandbaek

: A Danish diabetes risk score for targeted screening: the Inter99 study. Diabetes Care 2004; 27: 727–733.

20.

Lee

, Chang

, Liu

: Acculturation, psychiatric comorbidity and posttraumatic stress disorder in a Taiwanese aboriginal population. Soc Psychiatry Psychiatr Epidemiol 2009; 44: 55–62.

21.

Wang

, Ni

, Chen

: Posttraumatic stress disorder 1 month after 2008 earthquake in China: Wenchuan earthquake survey. Psychiatry Res 2011; 187: 392–396.

22.

Chan

, Wang

, Qu

: Posttraumatic stress disorder symptoms among adult survivors of the 2008 Sichuan earthquake in China. J Trauma Stress 2011; 24: 295–302.

23.

Başoĝlu

, Kiliç

, Salcioĝlu

: Prevalence of posttraumatic stress disorder and comorbid depression in earthquake survivors in Turkey: an epidemiological study. J Trauma Stress 2004; 17: 133–141.

24.

Copuroglu

, Ozcan

, Yilmaz

: Acute stress disorder and post-traumatic stress disorder following traumatic amputation. Acta Orthop Belg 2010; 76: 90–93.

25.

Wang

, Zhang

, Shi

: Symptoms of posttraumatic stress disorder among adult survivors two months after the Wenchuan earthquake. Psychol Rep 2009; 105: 879–885.

26.

Shumaker

, Brownell

: Toward a theory of social support: closing conceptual gaps. J Soc Issues 1984; 40: 11–36.

27.

Stimpson

, Wilson

, Jeffries

: Seeking help for disaster services after a flood. Disaster Med Public Health Prep 2008; 2: 139–141.

28.

Lai

, Chang

, Connor

: Full and partial PTSD among earthquake survivors in rural Taiwan. J Psychiatr Res 2004; 38: 313–322.

29.

Cairo

, Dutta

, Nawaz

: The prevalence of posttraumatic stress disorder among adult earthquake survivors in Peru. Disaster Med Public Health Prep 2010; 4: 39–46.

30.

Bromet

, Sonnega

, Kessler

: Risk factors for DSM-III-R posttraumatic stress disorder: findings from the National Comorbidity Survey. Am J Epidemiol 1998; 147: 353–361.

31.

Dietrich

: Risk factors in PTSD and related disorders: theoretical, treatment, and research implications. Traumatology 2001; 7: 23–50.

32.

Peris

, Bonizzoli

, Iozzelli

: Early intraintensive care unit psychological intervention promotes recovery from post traumatic stress disorders, anxiety and depression symptoms in critically ill patients. Crit Care 2011; 15: R41.