A patient-follow-up study, which was previously published 23, was conducted among a population-based sample of injury patients of all severity levels. This study followed injury patients aged 15 years and older who attended the ED of the Dutch Injury Surveillance System (a representative continuous registry of intentional and unintentional injuries of 17 hospitals in the Netherlands). Surveys were conducted at 2 months, 5 months, 9 months and two years after initial treatment. This study was conducted with the approval of the Ethics Committee Erasmus MC University Hospital.

Between 8 October 2001 and 31 December 2002 a sample was selected of 8,564 patients aged 15 years and older who attended the ED of the Dutch Injury Surveillance System 23. The patients were treated at the ED, followed by either hospital admission or direct discharge to the home environment. The sample of patients consisted of victims of traffic, home and leisure, occupational and sport accidents. The sustained injuries varied from minor to severe injury, single and multiple injury and hospitalized and non-hospitalized patients. The sample of patients was stratified, over sampling patients who were hospitalized. Each injury patient of the selected sample received a postal questionnaire 21/2 months after the injury and 3,167 (37%) responded. The first questionnaire was made anonymous for privacy reasons. At 5, 9 and 24 months a follow-up questionnaire was sent to patients that responded to the preceding questionnaire. For these questionnaires the patients needed to give permissions by an informed consent form. The 5, 9 and 24 months follow-up questionnaire were completed by respectively 2,384, 2,295 and 1,781 patients. The present study used a sample of 1,781 respondents (i.e 21% of the original sample) on the two year post-trauma survey, which assessed both posttraumatic stress symptoms and HRQoL 23. To adjust the data for non-response, a non-response analysis was conducted 23. Multivariate logistic regression analysis was used to examine if variables age, sex, type of injury, external cause of the injury, hospitalization and length of stay, health status and ambulance transport were possible determinants of non-response. The significant variables were used to adjust for response bias by inverse probability weighting 24. Additionally, the data were adjusted for stratification of the sample of ED patients 23

The follow-up questionnaire included items regarding demographics of the patient, accident category, type of injury, health care use and the Impact of Event Scale (IES), which was used to assess symptoms of posttraumatic stress indicative of PSTD 25. The IES consists of 15 items, which measure intrusive re-experiences of the trauma and avoidance of trauma-related stimuli. By combining the 15 items the total IES-score, ranging from 0 through 75, can be calculated. Wohlfarth et al. showed that a cut-off score of 35 on the total IES-score produced a sensitivity of .89, and a specificity of .94 against the DSM-IV diagnostic criteria for PTSD as the gold standard 26. Therefore, we assumed that an IES-score higher than 35 (IES ≥ 35) represents symptoms of posttraumatic stress indicative of PTSD. The Dutch translation of the IES has been found to be valid and reliable 27.

Additionally, the questionnaire included items to measure functional outcome and HRQoL. HRQoL is an index of perceived functional outcome of an illness and disability that is anchored between 0 (worst imaginable health state or death) and 1 (full health), thus allowing comparison between the health status of patients with distinct diseases. To measure HRQoL, multi-attribute utility instruments (MAUIs) such as the Health Utility Index (HUI) or the EQ-5D may be used 2829. These instruments require the patient to report his or her health state with a standardised generic health state classification system, which is then converted into a health utility score using utility weights derived from the general population. Despite the similarities in obtaining the health utility score, there are important variations between the instruments regarding the health domains included in the health classification system and the methods applied to derive the utility weights 30. As a result of these variations, the distinct instruments yield different utilities for similar health states. To overcome omissions in measuring HRQoL it is important to use several instruments that have complementary health domains 31.

Therefore, to measure functional outcome and HRQoL, the questionnaire included the EQ-5D and the HUI mark 3 (HUI3). With the EQ-5D classification system, respondents describe their health in three levels of severity on the health domains mobility, self-care, usual activities, pain/discomfort and anxiety/depression 32. Subsequently, the weight of that health state is computed by a formula that firstly yields a partial weight score for each domain depending on the reported level and secondly adds the utility weights (also referred to as the 'tariff'), which are based on preference data of the general population of the UK 33.

For instance, a patient reports some problems with walking and performing usual activities, as well as moderate pain or discomfort (EQ-5D profile 21221). Full health has a utility value of 1. Because the health state of the patient deviates from the best possible health state (EQ-5D profile 11111), a fixed reduction of -.081 is applied. For the problems with walking, performing usual activities and moderate pain or discomfort reductions are applied of -0.069, -0.036 and -0.123 respectively. This is results in a utility of 0.691. The complete algorithm to calculate EQ-5D utilities is published by Dolan et al. 34

The questionnaire included 19 items regarding the presence of one or more chronic disease(s) prior to the injury to assess comorbidity 35. Comorbidity is defined as the presence of any coexisting medical conditions or disease processes additional to the injury that the injury patients sustained 36.

For analysis of the data the Statistical Package for the Social Sciences version 14.0 was used (SPSS Inc, Chigaco, Ill). The IES-score can be calculated if all IES items are completed. In 8% of the cases data of one of the 15 IES items was missing. For these cases, the missing IES item was estimated by calculating the median value of 5 nearby points. The missing data was then imputed by the estimated values 37. If more than one of the 15 IES items was missing, data were not imputed. Chi-square statistics (dichotomous variables) and Student t tests (continuous variables) were used to test for differences between injury patients with IES scores higher or lower than 35.

Univariate logistic regression and multivariate logistic regression analyses (enter method) were used to determine the predictive value of patient demographics, accident category and severity level of the sustained injuries with regard to posttraumatic stress symptoms indicative of PTSD (IES ≥ 35) at two-year post-trauma. To dichotomize severity level, the injury diagnoses were categorized into two severity classes (mild versus moderate to severe) as previously tested by an international expert group 38. The injury severity class moderate to severe comprises injuries such a skull-brain injury, fracture/dislocation of the vertebral column, fracture of pelvis and hip fracture. The injury severity class mild comprises injuries such as superficial injury, concussion and wrist fracture.

For the analysis of the association between IES ≥ 35 and HRQoL, we selected participants that filled in both EQ-5D and HUI3. To test differences between participants with and without PTSD regarding their responses on each of the EQ-5D and HUI3 health domains, the non-parametric Wilcoxon-Mann-Whitney test was conducted. Differences regarding the mean EQ-5D and HUI3 summary scores were tested with a one-way ANOVA. P-values < 0.05 were considered to indicate statistical significance.

Stepwise multiple regression analyses (enter method) was applied to investigate the association between demographics (block 1), hospitalization and comorbidity (block 2) posttraumatic stress symptoms indicative of PTSD (IES ≥ 35) (block 3) and HRQoL measured with the EQ-5D and HUI3.

Regarding the respondents on the 24-month follow-up questionnaire, the average age was 44.5 years old and 46% were female. Over one half (54%) was injured due to home and leisure accidents. The sustained injuries of all the respondents consisted mostly of superficial injury/open wounds (51%) and upper extremity fractures (13%). After treatment at the ED, 9% of the respondents were admitted to hospital. Approximately one third (31%) had one or more pre-existing comorbid conditions. Table 1 shows the characteristics of the injury patients, accident category and hospitalization status.

With reference to the 1,781 respondents that completed the 24-month follow-up questionnaire, 1,585 (89%) filled in the EQ-5D and the HUI3 and 1,380 (77.5%) filled in the IES.

EQ-5D - Table 2 shows the responses on the EQ-5D of injury patients with IES scores higher or lower than 35. The calculated mean EQ-5D summary score for injury patients with IES scores ≥ 35 was 0.56, whereas for injury patients with lower IES scores the mean EQ-5D summary score was 0.87 (t = 112.0; p < 0.001). Respondents with posttraumatic stress symptoms indicative of PTSD reported significantly more problems on all five EQ-5D health domains (p < 0.001). Differences in reported problems between patients with IES scores higher or lower than 35 were largest for EQ-5D health domains pain/discomfort (82% versus 28%) and anxiety/depression (53% versus 11%).

When the responses of hospitalized and non-hospitalized injury patients with IES ≥ 35 and IES < 35 are presented separately, again patients with symptoms indicative of PTSD (IES ≥ 35) report significantly more problems on each of the EQ-5D health domains, resulting in a mean EQ-5D utility loss of 0.32 for non-hospitalized patients (t = 112.2; p < 0.001) and 0.23 for hospitalized patients (t = 22.1; p < 0.001). Compared to the injury patients without PTSD indications, injury patients with symptoms indicative of PTSD (IES ≥ 35) at 24-months post-trauma also had significantly lower mean EQ-5D utility scores at the 21/2-month (t = 105.0, p < 0.001), 5-month (t = 100.1, p < 0.001) and 9-month (t = 38.1, p < 0.001) follow-up.

Figure 1 shows the mean EQ-5D utility score of non-hospitalized and hospitalized patients with and without symptoms indicative of PTSD at 21/2, 5, 12 and 24 month follow-up

HUI3 - Table 2 also shows the responses on the HUI3 domains reported by injury patients with IES-scores higher or lower than 35. For patients with IES ≥ 35 the calculated mean HUI3 summary score was 0.51 and for patients with lower IES-scores 0.83 (t = 81.1; p < 0.001). Respondents with posttraumatic stress symptoms indicative of PTSD (IES ≥ 35) reported significantly more problems on all HUI3 health domains, except hearing where a reverse association was found (p < 0.001). Differences in reported problems between patients with IES ≥ 35 and IES < 35 were largest for the HUI3 health domains emotion (92% versus 33%) and pain (84% versus 38%). Analysing the responses of non-hospitalized and hospitalized patients with and without PTSD indications (IES ≥ 35) separately shows that hospitalized patients with symptoms indicative of PTSD (IES ≥ 35) reported most problems. Non-hospitalized patients with lower IES-scores reported least problems on the HUI3 health domains. Symptoms indicative of PTSD (IES ≥ 35) were associated with a mean utility loss of 0.33 in non-hospitalized patients (t = 80.8; t < 0.001) and 0.24 in hospitalized patients (t = 15.9; t = 0.001)

The models tested to predict HRQoL measured with EQ-5D and HUI3 were both statistically significant (EQ-5D: F = 80.27, p < 0.001; HUI3: F = 118.55, p < 0.001). Table 3 shows that posttraumatic stress symptoms indicative of PTSD (IES ≥ 35) are associated with decreased HRQoL, even after controlling for possible confounders.