Headache in Traumatic Brain Injuries From Blunt Head Trauma

Population

In the main cohort study, we enrolled children younger than 18 years with GCS scores of 14 to 15 after nontrivial blunt head trauma who presented to the ED within 24 hours of the initial injury. We defined trivial trauma as that resulting from ground level falls or running into stationary objects with no evidence of head trauma other than scalp abrasions or lacerations. We excluded patients with penetrating head trauma, preexisting neurologic disease impeding clinical assessment, patients transferred to the ED with neuroimaging already obtained, and patients with bleeding disorders or ventricular shunts. We did not assess for the presence of headache in patients who were either physically unable to speak or preverbal, including but not restricted to all patients younger than 2 years.

Patient Assessment and Variable Definitions

Clinicians completed standardized history and physical examinations before cranial CT (if obtained) and documented the findings onto study case report forms. Headache was defined as any complaint of pain to the head at the time of ED evaluation (ie, a history of headache that had resolved was not considered a headache for purposes of analysis). For those patients with headaches, clinicians assessed headache severity (categorized as mild/barely noticeable, moderate, severe/intense, or unclear), timing of onset (before the head injury, within 1 hour of the injury, 1–4 hours after injury, >4 hours after injury, or unknown), and headache location (diffuse, only at site of injury, at occiput only and clearly due to restraining backboard, other, or unclear). Two clinicians independently evaluated a convenience sample of 4% of patients to assess interobserver agreement of clinical findings.⁸

Definitions of Isolated Headache

We defined isolated headache in 2 ways based on the absence of other specific clinical findings on initial ED history and physical examination (Table 1). The first definition (termed “extensive” definition of isolated headache) was based on an extensive list of variables potentially associated with TBI, and the second definition was based solely on the variables in the PECARN prediction rule for children 2 to 18 years.² We analyzed the data based on these 2 definitions of isolated headache as the literature suggests that clinicians often assess children from the vantage point of having either no signs or symptoms other than a single finding of concern (extensive definition) or having no signs or symptoms other than headache defined solely by the age-specific PECARN prediction rule variables.^9,10 One difference between the 2 definitions is that the extensive definition does not consider the mechanism of injury, because it is not in itself a symptom or sign of TBI, and we wished the extensive definition to reflect patient clinical characteristics alone. However, because severe mechanism of injury is 1 of the factors in the PECARN prediction rule, patients with headaches and severe mechanisms of injury did not meet the PECARN rule-based definition of isolated headache.

Outcomes

We had 2 outcomes: (1) ciTBI and (2) TBI on CT. We defined ciTBI as death due to TBI, neurosurgical procedure, intubation for at least 24 hours for TBI, or hospitalization for 2 or more nights due to the head trauma in association with TBI on CT. We defined TBI on CT as any acute traumatic intracranial finding or a skull fracture depressed by at least the width of the table of the skull. Cranial CT scans were obtained at the discretion of the treating providers. We completed follow-up procedures for all patients discharged from the ED to determine outcomes.²

Analysis

We summarized the data by using counts, percentages, and 95% confidence intervals (CIs) for categorical variables and medians and interquartile ranges for continuous variables. We analyzed outcomes in the following groups: (1) all patients with histories of headaches after the traumatic event who also had other signs or symptoms suggestive of TBI (ie, nonisolated headaches); (2) patients with isolated headaches–extensive definition; and (3) patients with PECARN-isolated severe headaches (ie, severe headaches with no other factors in the PECARN head trauma rule for those 2–18 years). We also assessed the risk of TBI in those with severe headaches plus 1 other PECARN prediction rule finding to determine the incremental risk of ciTBI of 1 other PECARN variable in addition to severe headache. For all analyses, we excluded patients who had missing documentation of any PECARN prediction rule factor for children 2 to 18 years. For the isolated headache–extensive definition, we also excluded patients from the analysis if they had >1 other of the extensive findings missing or marked as unknown. We calculated κ statistics and 95% CIs for the presence of headache and headache severity, and used the Fleiss-Cohen weighted κ with standard quadratic weights for headache severity.

We conducted an exact multivariable logistic regression analysis to assess whether TBI on CT was less likely in those with isolated compared with nonisolated headaches, adjusted for the severity of mechanism of injury. We also conducted 2 multivariable logistic regression analyses in all patients with headaches (combining the isolated and nonisolated groups) to assess the association between the 2 TBI outcomes and the severity (categorized as mild or moderate/severe) and location (categorized as diffuse or localized) of headache. A localized headache was defined by a headache only at the site of injury, at the occiput only and clearly due to backboard, and other. We were unable to run multivariable models to assess the effect of timing of the headache due to low risk of the 2 TBI outcomes in the “>1 hour after injury” group. For these 2 analyses, we excluded patients whose headaches started before the head injury. We adjusted for all variables included in the extensive definition of isolated headache and also included the age of the patient (categorized as 2–<5 years, 5–10 years, or >10 years). Seizure after head trauma and neurologic deficits were not included in the regression models due to low prevalence of these predictors. All variables were entered as dichotomous variables in the regression models except age, which was entered as the noted 3-level categorical variable. Additionally, we adjusted for time from injury to time of ED evaluation in hours (categorized as <1 hour, 1–4 hours, or >4 hours), because time from injury potentially influences the relationship between the severity of the headache and the outcomes.

We used sequential regression models to impute for missing data used in the multivariable logistic regression analyses. We limited imputation to those patients with headaches. We used linear regression models to impute continuous variables and generalized logistic models to impute categorical variables.¹¹

We used SAS/STAT software (version 9.3; SAS Institute, Inc, Cary, NC) for all analyses and imputed missing values for regression models by using IVEware (University of Michigan, Ann Arbor, MI).¹² Because this was a secondary analysis of the parent cohort study, we did not estimate sample size needs for this analysis.