The results from this study show that even relatively minor road traffic injuries can lead to a significantly lower HRQoL, especially among women. The findings show different patterns of HRQoL after RTI, depending on sex, age, injured body part, and levels of injury severity. It was found that all degrees of injury severity (MAIS classification) led to a significantly lower HRQoL for women, primarily in the younger age groups, compared to the reference group. However, for men, this difference in HRQoL was found only for more severe injuries (MAIS3+). These results confirm findings from other studies, where men reported significantly better HRQoL after an RTI than women. The results also confirmed that the HRQoL generally decreases with higher injury severity [
6,
9,
22] and that women do not recover their HRQoL in the same way as men [
9,
10,
22,
23]. However, this study adds additional knowledge since we had the possibility to assess the natural reduction of HRQoL by using a reference group. If this natural reduction is not considered in the comparison of HRQoL of those with injuries, the results indicate that older people with injuries are more negatively affected in relation to HRQoL, in all injury severities, compared to their younger counterparts. However, if the reference group’s reduction of HRQoL over age is considered in the analysis, then the interpretation becomes the opposite. Hence, younger individuals have a larger difference from the reference group in HRQoL, independent of the injury severity, compared to the older individuals, due to the reduction of HRQoL with increased age, since the level of HRQoL is already reduced for the older individuals. In other words, the difference between the injured group and the reference group was reduced with increased age, but women showed a lower HRQoL compared to men (Table
2), regardless of age and injury severity. Women in the older age groups had not recovered to the reference group’s level regardless of the injury severity, while men in the age group of 30–64 years with less severe injuries (MAIS1 and 2) (MAIS1) and men in the age group of 65 years with injuries classified as MAIS2 did reach the same level as the reference group. However, participants with more severe injuries (MAIS3+) did not reach the reference group’s HRQoL levels, regardless of age and sex. One conclusion is therefore that all levels of injury severity lead to a lower HRQoL for women, primarily in the younger age groups, whilst only more severe injuries (MAIS3+) lead to a lower HRQoL among men. Whether women suffer more from the psychological effects of RTI or if women tend to communicate about their injuries in different ways than men need to be further investigated. A qualitative study from our research group indicated that men and women report similarities regarding the experienced consequences of an injury, but they also report different consequences, both considering the type and the severity [
6]. There is evidence that women have worse physical outcomes after an injury due to their smaller body structure [
24]; however, more studies are needed to determine if the difference in outcome can be explained by differences in crash mechanism, from treatment variables, or from premorbid sex differences.
Not surprisingly, the results show that more severe injuries (MAIS3+) gave a substantially lower HRQoL in general, but also that it is related to gender and injured body part. Furthermore, the results show that even less severe injuries (MAIS1) can give considerably lower HRQoL, especially for women (see Fig.
2a, b for details on body parts). In accordance with previous studies, injuries to the cervical-, thoracic and lumbar spine, upper- and lower extremities, and pelvis result in significantly lower HRQoL [
6]. For the MAIS2 classified injuries, the low HRQoL was most substantial in relation to injuries to the cervical-, thoracic, and lumbar spine. For the MAIS1 classified injuries, the low HRQoL was most common in relation to injuries to the lumbar- and thoracic spine with a negative difference of 10–14% for the injury group compared to the reference group. In line with previous studies, the results show that injuries to the spine have the most impact on HRQoL, independent of the injury severity [
9,
11,
25].
Strengths and limitation
The main strengths of the study are that it includes a nationwide sample of RTIs and the comparison of injury severity and injured body parts, which gives an estimation of the impact of the different injuries and injury severities on HRQoL, both in total and for each gender, specifically in relation to a reference group without known injuries. However, some limitations also need to be mentioned in relation to this study, mainly arising from the low response rates for those with injuries and the reference group. Due to the low response rate, an extensive analysis was performed to identify possible biases. The analysis did not show a clear and consistent pattern of non-participation between gender and age groups, or between the injury group and the reference group. Both the injury group and the reference group followed similar non-response patterns, which suggests that major differential bias between the two groups is unlikely. We have also recognized the lower participation of younger individuals, especially men (who are healthier than older men); however, this pattern is common in public health surveys, and the health status of the study population is probably overestimated. On the other hand, partial non-response in any of the EQ5D dimensions assessed was very low, thus, preventing any further underestimation. Moreover, the low response rate also raises the question of power in the analysis, especially the sub-group analysis. To address this issue, a post-hoc power analysis was calculated based on a MCI of 0.082, which has previously been identified in HRQoL studies [
26,
27]. The study results show a standard deviation of 0.15 in the reference group and 0.25 for the injury group. According to the post-hoc power calculation, between 50 and 200 persons were required in each comparison group, giving a power of 80% at 5% significance. In the results where we have a larger difference between the groups than 0.082, the power is above 80%; however, sometimes is it less, e.g., in the case of differences between MAIS1 and MAIS3+ for women aged 7–17 years, the difference is 0.22, which gives a power of 65%. Ideally, the number of participants would have been larger in each sub-group, giving the opportunity to always detect statistically significant differences as small as 0.082. Thus, adding additional participants would probably generate more statistically significant results; however, it is plausible that this would not change the conclusions of the results.
Although the current study has the above-mentioned limitations, the generalizability of the results should be sound considering an adult injury population as we have not been able to identify a systematic bias regarding the non-responses in the different groups and injury severities included. However, the results might only be transferable to adult injury populations who seek emergency care.
Moreover, we also need to consider the skewness of the data in regard to the analysis. Although we did apply different variance stabilizing methods, the skewness was not corrected. With regard to the skewness of the data, it is important to remember that HRQoL data in injury populations are naturally negatively skewed due to a large number of individuals who report a high HRQoL; hence, the mean of the population is lower than the mode. However, when we compared the EQ5D mean scores with the ones published by Sun et al. [
28] (with a representative sample of 57,009 citizens from Stockholm; response rate of 61%), the EQ5D mean scores for the current study population were slightly lower (0.05–0.06 points) compared with the mean scores attained by Sun et al. [
28]. Furthermore, when comparing the current mean scores with another Swedish study conducted by Burström et al. [
15], who used the EQ5D, the mean scores of the two studies are even more similar, although Burström et al. published their data 9 years before the data collection for this study. These comparisons suggest that the reference group used in this study is likely to be representative of the target population. The response rate analysis is presented in more detail in Monárrez-Espino et al. [
13].