Research Review By Dr. Michael Haneline©


Download MP3

Date Posted:

October 2020

Study Title:

Development and validation of a model predicting post-traumatic headache six months after a motor vehicle collision in adults


Cancelliere C, Boyle E, Côté P, Holm LW, Salmi L, Cassidy JD

Author's Affiliations:

Faculty of Health Sciences & Centre for Disability Prevention and Rehabilitation, Ontario Tech University and Canadian Memorial Chiropractic College, Ontario, Canada; Department of Sport Science and Clinical Biomechanics, Faculty of Health Sciences, University of Southern Denmark, Denmark; Division of Epidemiology and Institute of Health Policy, Management and Evaluation, Dalla Lana School of Public Health, University of Toronto, Ontario, Canada.

Publication Information:

Accident Analysis & Prevention 2020; 142: 105580. doi: 10.1016/j.aap.2020.105580

Background Information:

Post-traumatic headache (PTH) has been defined as a new headache, or worsening of a pre-existing headache, that begins within seven days after head injury, whiplash, craniotomy, or regaining of consciousness following trauma (1). PTH is associated with mild traumatic brain injury (MTBI) and whiplash-associated disorders (WAD), which affect approximately 600 per 100,000 persons annually.

MTBI and WAD may lead to post-concussion symptoms (PCS), of which PTH is the most common symptom. Studies have shown that 30% of persons with PTH have persistent symptoms for up to one-year post-injury (2, 3).

Even though PTH is fairly common, very little is known about its prognostic factors. Thus, the objective of this study was to develop and validate a prediction model for reporting PTH six months following a motor vehicle collision (MVC) in adults who reported PTH during their initial interview.

Pertinent Results:

The baseline characteristics of the cohorts were similar in most categories; however, the percentage of participants that reported severe headache pain at baseline was greater in the Saskatchewan cohort (36.5%) than in the Swedish cohort (13.6%). On the other hand, the percentage of participants that reported PTH at six months was greater in the Swedish cohort (46.4%) as compared to Saskatchewan cohort (31.8%).

The multivariable analysis produced a prediction model that included 12 predictor variables, with eight of them found to be useful in predicting PTH, including:
  1. Age
  2. Work status
  3. Headache pain intensity
  4. Symptoms in arms or hands
  5. Dizziness or unsteadiness
  6. Stiffness in neck
  7. Pre-existing headache
  8. Lower recovery expectations
Patients were categorized by the model as "positive", meaning they will have PTH 6 months after the collision, at a probability level of 0.75, which is considered a high threshold that maximizes specificity. Thus, anyone who scores a probability level of 0.75 or higher on the model would be categorized as "positive" for the likely presence of PTH at six months following a traffic collision.

The specificity of the development (Saskatchewan-relevant) model was 99.8% and the likelihood ratio (LR) was +8.0, while the validation (Sweden-relevant) model’s specificity was 95.5%, with a LR +6.0.

Outcome information was missing for 20.3% of the Saskatchewan participants and 25.6% of the Swedish participants. Data on predictors was missing for 3.5% of the Saskatchewan participants and 2.4% of the Swedish participants. These participants’ data were excluded from the analysis.

Clinical Application & Conclusions:

Based on the performance of the model that was developed in this study, clinicians can collect information from patients on the eight predictors mentioned above to identify patients who will be likely to report ongoing PTH 6 months after a traffic collision.

Clinicians can also provide PTH patients advice on healthful lifestyles and symptom management (e.g. proper sleep, nutrition, exercise, coping, and stress management), and for those patients who are at risk of developing ongoing PTH, clinicians should monitor them more closely and deliver self-management strategies more intensely.

MVCs often result in injuries and symptoms beyond the head and neck that may impact PTH outcomes, so offering evidence-based care for conditions like back pain, for instance, may help to improve outcomes for patients with PTH (4).

Study Methods:

This was a secondary analysis of a population-based, incidence cohort study that included all adults treated for motor vehicle injuries in the province of Saskatchewan, Canada who were followed for one-year after injury. In addition, a validation cohort from Sweden was included that involved a secondary analysis of data from an incidence cohort study of personal injury claimants to two Swedish traffic insurers that insured about 20% of cars driven in Sweden during 2004. Sweden was selected as the validation cohort because of the similarities between Saskatchewan and Sweden (e.g. similar climates, similar car safety laws and speed limits, universal access to health care, and comparable insurance claims processes).

The inclusion criteria were:
  • All adults who made a bodily injury claim or were treated by a registered health professional for a motor-vehicle related injury; and
  • were at least 18 years old at the time of the collision.
Participants were excluded if they:
  • Did not occupy a motor vehicle at the time of the collision,
  • completed their baseline questionnaire > 30 days after their collision,
  • reported being hospitalized > 2 days as a result of the collision,
  • reported lost consciousness > 30 min after the collision (Saskatchewan cohort only), or
  • reported that the “accident caused” headache had an intensity < 3/10 on the 11-point numerical rating scale (NRS) at the time of completing their baseline questionnaire.
Participants in both cohorts completed questionnaires on baseline data which included sociodemographic factors, preinjury health status, injuries and other symptoms experienced since the collision, depressive symptomology and recovery expectations. In addition, validated measures were used where possible, including the numeric rating scale (NRS) for pain intensity, overall general health, and the Centre for epidemiological Studies – Depression Scale (CES-D). Follow-up data was collected by computer-assisted telephone interview in Saskatchewan and self-reported questionnaire in Sweden.

Participants were considered to have PTH 6 months after the collision if they self-reported a headache that was ≥ 3 on the NRS as a result of the accident.

Thirty-six baseline characteristics were selected to be candidate predictors for the target outcome (PTH). Candidate predictors were grouped according to five domains (5):
  1. Personal factors (e.g. age, gender, lifestyle);
  2. impairments (physical and psychological);
  3. activity limitations;
  4. participation restrictions; and
  5. environmental factors (barriers or facilitators in the physical, social and attitudinal environment).
Predictors were selected for multivariate analysis following a preliminary analysis of all predictors and removal of those predictors that did not contribute usefully in the multivariable model.

Since some participants did not complete the study, sensitivity analyses were conducted to assess the impact of missing outcome data on the study’s overall conclusions. The following 3 different scenarios were tested to determine whether they would lead to different conclusions: 1) assumed “no PTH” for those with a missing outcome value; 2) assumed “PTH” (worst case scenario); and 3) “carry forward the last observation” (whether at six weeks or three months). The multivariable analysis was repeated for the 3 scenarios to determine each scenario’s effect on the study’s outcome.

Study Strengths / Weaknesses

  • This study involved a large population-based cohort of patients in Saskatchewan who were injured in MVCs; thus, ensuring the stability and accuracy of the prediction model that was developed. To augment the study’s validity, a geographic validation study was conducted on a similar population in Sweden.
  • The prediction model was tested in an independent population, which resulted in similar predictive ability, specificity and positive likelihood ratios.
  • When available, valid psychometric measures were utilized to develop predictors.
  • The subject matter knowledge and terms used to develop the model were clinically relevant, resulting in predictors that can easily be collected and utilized by clinicians.
  • Pain intensity questions were not the same between the cohorts, wherein participants in the Saskatchewan cohort were asked about “average” pain over the past week, while participants in the Swedish cohort were asked about “current” pain.
  • The definition of headache that was used is somewhat equivocal, given that there is no gold standard diagnosis of PTH. Consequently, the authors relied on published studies to select a 30-day cut-off for the onset of PTH and a ≥ 3 NRS cut-off for pain intensity.
  • An a priori decision was made to reduce false positives (those who are classified as having a high probability of the target outcome that do not develop the outcome), although this results in a high number of false negatives (those who are classified as having a low probability of the target outcome that do develop the outcome). The authors reasoned that due to the lack of evidence for effective PTH treatments, it would be better to identify true positives in order to inform a patient of their prognosis and encourage self-management strategies.
  • Some potentially important predictors may not have been included, such as psychological factors, lifestyle factors, factors related to compensation, etc.

Additional References:

  1. Headache Classification Committee of the International Headache Society, 2013. The international classification of headache disorders, 3rd edition (beta version). Cephalalgia 2013; 33(9): 629–808.
  2. Ferrari R, Russell AS, Carroll LJ, Cassidy JD. A re-examination of the whiplash associated disorders (WAD) as a systemic illness. Ann Rheum Dis 2005; 64(9): 1337–1342.
  3. Cassidy JD, Cancelliere C, Carroll LJ et al. Systematic review of self-reported prognosis in adults after mild traumatic brain injury: results of the international collaboration on mild traumatic brain injury prognosis. Arch Phys Med Rehabil 2014; 95(3 Suppl): S132–151.
  4. Cassidy JD, Boyle E, Carroll LJ. Population-based, inception cohort study of the incidence, course, and prognosis of mild traumatic brain injury after motor vehicle collisions. Arch Phys Med Rehabil 2014; 95(3 Suppl): S278–285.
  5. Leonardi M, Steiner TJ, Scher AT, Lipton RB. The global burden of migraine: measuring disability in headache disorders with WHO’s Classification of Functioning, Disability and Health (ICF). J Headache Pain 2005; 6(6): 429–440.