Expectations for Recovery Important in the Prognosis of Whiplash Injuries

Summary

Background:
Individuals' expectations on returning to work after an injury have been shown to predict the duration of time that a person with work-related low back pain will remain on benefits; individuals with lower recovery expectations received benefits for a longer time than those with higher expectations. The role of expectations in recovery from traumatic neck pain, in particular whiplash-associated disorders (WAD), has not been assessed to date to our knowledge. The aim of this study was to investigate if expectations for recovery are a prognostic factor after experiencing a WAD.

Methods and Findings:
We used a prospective cohort study composed of insurance claimants in Sweden. The participants were car occupants who filed a neck injury claim (i.e., for WAD) to one of two insurance companies between 15 January 2004 and 12 January 2005 (n = 1,032). Postal questionnaires were completed shortly (average 23 d) after the collision and then again 6 mo later. Expectations for recovery were measured with a numerical rating scale (NRS) at baseline, where 0 corresponds to “unlikely to make a full recovery” and 10 to “very likely to make a full recovery.” The scale was reverse coded and trichotomised into NRS 0, 1–4, and 5–10. The main outcome measure was self-perceived disability at 6 mo postinjury, measured with the Pain Disability Index, and categorised into no/low, moderate, and high disability. Multivariable polytomous logistic regression was used for the analysis. There was a dose response relationship between recovery expectations and disability. After controlling for severity of physical and mental symptoms, individuals who stated that they were less likely to make a full recovery (NRS 5–10), were more likely to have a high disability compared to individuals who stated that they were very likely to make a full recovery (odds ratio [OR] 4.2 [95% confidence interval (CI) 2.1 to 8.5]. For the intermediate category (NRS 1–4), the OR was 2.1 (95% CI 1.2 to 3.2). Associations between expectations and disability were also found among individuals with moderate disability.

Conclusions:
Individuals' expectations for recovery are important in prognosis, even after controlling for symptom severity. Interventions designed to increase patients' expectations may be beneficial and should be examined further in controlled studies.

Introduction

Whiplash is an acceleration-deceleration mechanism resulting from an energy transfer to the neck that may lead to a variety of clinical manifestations (whiplash-associated disorders [WAD]) [1]. Despite substantial research over many years, understanding prognostic factors for WAD recovery is sparse [1,2]. High pain intensity in the neck and head, radicular signs/symptoms, older age, female gender, and tort insurance systems are associated with delayed recovery [2]. Psychological and psychosocial factors such as general psychological distress [3], early depressive symptoms [4], passive coping strategies [5], perceived helplessness [6], and higher levels of somatisation [7] are also associated with poor prognosis in WAD. Several studies also indicate a substantial difference in recovery between countries [2,8]. Such variation might reflect differences across study populations, different inclusion criteria, or different health-care seeking behaviour, but might also reflect societal and individual differences in the perception or expectations of the injury and for recovery. Thus expectations might in turn influence prognosis. Results from studies of uninjured individuals' perceptions of the symptoms that may arise from WAD and those from mild traumatic head injuries suggest that there are differences between countries in the perceptions of presence and duration of symptoms. For example in Lithuania and Greece, symptoms are expected to be fewer and to be present for a shorter duration than is the case in Canada [9,10].

The concept of expectations for recovery is complex. Janzen et al. have suggested a conceptual model for health expectations, which can also be applied to recovery expectations [11]. Such expectations are, apart from the symptoms/signs of the disease or injury, influenced by prior health and prior knowledge of the condition, as well as by psychological characteristics such as anxiety, self efficacy, and the patient's beliefs. Systematic reviews of research on various clinical conditions and outcomes have concluded that expectations contribute to the patients' outcome, including recovery duration and return to work [12,13].

Over the past 10–15 y, considerable attention has been given to WAD in the Swedish media. WAD are often considered to have severe consequences. Such “prior knowledge” might have an impact on expectations of the prognosis for individuals experiencing WAD. To our knowledge, the role of recovery expectations in individuals with traumatic pain has not been studied. Therefore, the objective of this study was to assess the prognostic value of individuals' expectations for recovery after WAD.

Methods

Study Population

A cohort was formed from incident cases of injury claimants to two Swedish traffic insurers: Trygg-Hansa and Aktsam. The inception period was between January 15, 2004 and January 12, 2005. A weekly computer-based search of new claims was performed at the insurer's information technology department. The following criteria were used for the search: claimant age 18–74 y; car occupants who filed an injury claim within 30 d of a collision; and no fatal injury to another occupant in the same car. Identifying information on each injured person was transferred to Karolinska Institutet (KI) on a weekly basis. The following day, questionnaires were sent from our research centre at KI . The questionnaire included questions concerning the collision, socioeconomic and preinjury health status, and pain and other symptoms experienced since the collision. We included questions assessing mental health and questions that could capture information about the injured person's expectations for a complete recovery. Because of a risk of low or delayed response during the summer holiday of 2004, claims made during 16 June and 3 August 2004 were excluded. At 6 mo postinjury, a follow-up questionnaire was administered to those who responded to the baseline questionnaire. Here we asked about the disability and signs and symptoms that the participants attributed to the injury 6 mo earlier. At 6 mo after the last inception into the study, we also retrieved information from the insurers on all claimants about whether they had completed their claim, and whether the claim was settled or not.

We included claimants who met the criteria outlined above, and who (i) reported WAD (defined below), (ii) were not hospitalized for more than 2 d, and (iii) did not have other injury claims reported during the inception period. We excluded those who were not Swedish residents, those with complete recovery at baseline, those who had missing data on expectations for recovery, and those who, at follow up, reported that they had experienced any new injury during the past 6 mo. The study was approved by the Regional Committee on Ethics at Karolinska Institutet, Stockholm.

Definition of WAD

WAD was defined as having answered “yes” to any of the following questions in the questionnaire: “Do you have or have you had pain/ache in the neck due to the accident?” or “Do you have or have you had reduced neck movement that you relate to the accident?” Cervical fractures were excluded.

Outcome

Disability was assessed with the seven-item Pain Disability Index (PDI) questionnaire, which ranges from 0 to 70, where 0 is no disability due to pain [14]. PDI was trichotomised with cut-off scores at the median and at the 75th percentile.

Exposure

Expectations for recovery were measured with a numerical rating scale (NRS 0–10) where the respondents were asked to rate how likely it was that he/she would have a complete recovery. The anchors were labeled “not likely” (0) and “very likely” (10). For the analysis, we reversed the scale and arbitrarily trichotomised it into those who stated that they were less likely to make a full recovery (NRS 5–10), an intermediate category (NRS 1–4), and those who stated that they were very likely to make a full recovery (NRS 0).

Potential Confounders

We used the baseline information when choosing potential confounders. The choice was guided by knowledge of prognostic factors for neck pain and WAD (Box 1). Postcollision symptoms were (i) pain referring to three specified anatomic regions and one option for “other” region, all assessed with numerical rating scales (NRS 0–10); (ii) number of pain areas from pain drawings [15]; and (iii) mental health, using the Hospital Anxiety and Depression Scale (HADS) [16,17]. We assessed psychological stress after the collision with the Impact of Event Scale (IES) [18,19]. In the analysis of potential confounders, HADS subscales and IES were all treated both as categorical variables according to suggested cut-off scores [16,20] and as continuous variables.

Moreover, we used a subscale of the Pain Management Inventory to assess passive coping [21,22]. The presence and severity of eight pain-associated symptoms were assessed with a checklist, with a five-grade verbal scale ranging from “no symptoms” to “unbearable symptoms.”

Statistical Analysis

We used multivariable polytomous logistic regression [23]. We first assessed the crude associations between the exposure and pain disability. The exposure was treated both as a continuous and categorical variable, thus we built two different regression models. We determined the role of confounders by considering each potential confounder in turn in the models [24]. If any single variable changed any of the crude estimates by more than 10%, they were entered in the final model [24]. We also evaluated the final full model by removing one confounder at a time. We tested for the presence of interaction by stratifying for gender, educational level, and level of pain intensity. Results are presented as crude and adjusted odds ratios (ORs) with 95% confidence intervals (CIs).

Responders were compared with nonresponders, to determine potential bias as a result of selective participation. We had information on all claimants regarding age, gender, and whether they had completed the injury claim or not. We also compared those with WAD who responded to the baseline questionnaire with those who were successfully followed for 6 mo.

All statistical analyses were performed with the use of SPSS software package version 14.0 [25].

Results

Between January 15, 2004 and January 12, 2005, there were 4,603 eligible claimants fulfilling the search criteria at the insurance companies. Of these, 676 were filed during the summer of 2004, and were excluded. Another 52 had an unknown address, were not car occupants, or were not Swedish residents. Of the remaining claimants 3,875 were successfully approached with questionnaires. Figure 1 illustrates the inclusion and exclusions of study participants, as well as the attrition from the study.

Attrition Analysis

Compared to the responders, the nonresponders were younger (mean age 37.2 y versus 39.3 y), had a lower proportion of females (43.9% versus 56.1%), and a higher proportion of uncompleted claims (40.7% versus 17.7%). There were no essential differences between the WAD claimants who completed both questionnaires compared to those who completed only the baseline questionnaire with respect to sociodemographic characteristics, recovery expectations, physical and mental symptoms at baseline, and proportion of settled claims.

Study Population

The baseline questionnaire was completed on average 23 d (median 20 d) postinjury, with 4% responding within 7 d and 81% within 30 d of the collision. Table 1 presents a description of the 1,032 persons in the study population. Table 2 shows the proportion of the three categories of expectations stratified by baseline neck pain intensity and disability. In all 278 persons (27%) in both pain strata rated themselves to be very likely to make a full recovery. As expected, those with mild pain intensity were more likely to expect complete recovery compared to persons with intense pain, but 123 (17%) of those in the mild pain strata were pessimistic about their recovery. On the contrary, only 22 (7%) in the moderate/severe pain group had rated it very likely to have complete recovery, and 160 (50%) fell into the lower group of expectations.

Recovery Expectations and Disability

The crude ORs of the associations between expectations for recovery and pain disability are presented in Table 3. It also shows the crude OR for demographic and prior health factors, and for the confounding factors. After adjustment for confounders, the odds of high disability in individuals who stated they were less likely to make a full recovery were 4.2 times the odds in individuals who stated they were very likely to make a full recovery (Table 3). There was a dose response relationship, in that the intermediate category (NRS 1–4) was also associated with severity of disability, but the ORs were smaller. These estimates were independent of whether we modelled the confounders; anxiety, depression, and posttraumatic stress as category variables or as continuous variables. When stratifying the data by mild (NRS 0–4) versus moderate/severe (NRS 5–10) baseline neck pain, we found a stronger effect in the mild neck pain strata. However, there were only 22 individuals in the joint category of “moderate/high neck pain intensity” and stating “very likely to recover,” which precluded a formal interaction analysis (Table 2).

Discussion

Expectations for recovery were an important factor in the prognosis of WAD, for both moderate and high disability, with a clear dose response relationship. To our knowledge this is the first study to assess these associations in WAD. The significance of the findings is that in addition to an understanding of the injury and its clinical symptoms, the patient's perception of recovery is necessary information for a proper understanding of WAD. Only 27% of the study participants reported that they were very likely to make a complete recovery. Interventions with the goal of changing expectations and beliefs may therefore benefit a large proportion of individuals with WAD.

The study's strengths include the prospective design with a well-defined study population and assessment of recovery expectations early after the injury. Moreover the study is based on a well-defined population of injured persons, in which a variety of potential confounders were evaluated. We tested the robustness of the regression models both in single and combined ways, and the findings were broadly similar regardless of which statistical modelling strategy was used.

Our study has some limitations. Nonparticipants were more likely to be younger and male. However, this would have an impact on the results only if these factors were confounders in the relation between expectations of recovery and disability. The nonparticipants were also less likely to have completed their claim. Since all persons, irrespective of whether or not they are responsible for the collision, are entitled to insurance compensation in Sweden, we assume that nonparticipants were less likely to have sustained an injury or only a transient injury. Although residual confounding is possible because of misclassified or unmeasured confounders, the strength of the associations observed for high disability make it unlikely that the findings can be fully explained by residual confounding. However, it is possible that passive coping strategies, anxiety, depressive symptomatology, and other similar psychological characteristics influence the perception of disability in other ways than by being confounders. Some factors might act as mediators, or for some factors the causal change might be reversed. For instance recovery expectations might effect how individuals report intensity of pain, instead of pain intensity effecting expectations for recovery. Thus, we cannot be confident that there is a causal relationship between recovery expectations and actual recovery. Nevertheless, other studies have found that early positive information about the injury, which is one way of modifying recovery expectations, has a favourable impact on the prognosis [26,27]. In a recent intervention study in which the authors investigated the efficacy of informational group sessions for high risk patients with mild injuries after traffic collisions (including WAD), the intervention group had a 20% greater recovery rate compared to the group who received “general care” [28]. In that study expectations for recovery were not measured, but one possible explanation for the positive effect in the intervention group, may be that the expectations for recovery were modified by the group sessions. Previous findings from studies on conditions similar to WAD also suggest that individuals' expectations are important factors in recovery and return to work [12,29,30]. The fact that we found a dose response relationship between level of expectations and the level of reported disability is a further indication of a causal relationship. If future interventional studies show that modifying expectations of recovery reduced or removed this association, it would be a further step towards demonstrating causality.

The substantial influence of initial expectations on subsequent disability, even after considering the impact of baseline pain intensity and other symptoms, might be a partial explanation for anecdotal reports of long term dysfunction in persons with initially mild WAD symptoms.

In conclusion, we suggest early assessment of expectations for recovery to be made, in order to identify people at risk for poor prognosis after WAD. Furthermore, controlled studies on interventions aimed at modifying expectations are warranted. Such studies could be conducted on the population level, similar to the successful media campaign on back pain beliefs, which decreased disability claims, both in terms of incidence and time on benefits [31,32]. Alternatively interventions targeting persons in the acute phase of an injury should be evaluated. Finally, it is not inconceivable that our findings can be extended to persons with pain conditions other than WAD.

Supporting Information

Attachment 1