Recovery from mild traumatic brain injury (mTBI) is primarily based on the resolution of post-concussive symptoms back to a premorbid level. However, the “good old days” bias means fewer premorbid symptoms are retrospectively recalled, thus skewing the determination of recovery relative to pre-injury. The objectives of this study were to investigate the “good old days” bias in pediatric mTBI and demonstrate the implications of this bias on perceived recovery. Children and adolescents 2–18 years old (mean = 10.9, SD = 4.4, N = 412) were recruited after sustaining an mTBI. Ratings of premorbid symptoms were provided: (a) in the Emergency Department (ED; by parents), (b) retrospectively at a 1-month follow-up (by parents and adolescents), and (c) retrospectively at a 3-month follow-up (by parents and adolescents). Parent ratings of premorbid symptoms decreased by 80% from the ED to 1-month post-injury (p < .001) but were stable from 1 to 3 months post-injury (p < .05). Adolescents premorbid ratings declined from 1 to 3 months post-injury. Slow recovery did not have a differential impact on premorbid reporting. Using premorbid ratings obtained in the ED, instead of retrospective symptom reporting at the time of follow-up, suggests that a significant minority of patients believed to be “not recovered” actually have the “same or lower” symptom ratings at 1 (29%) and 3 months (41%) post-injury compared with before the injury. The “good old days” bias is present in pediatric mTBI by 1-month post-injury, influences retrospective symptom reporting, and has measureable implications for determining recovery in research and clinical practice.
Mild traumatic brain injuries (mTBIs) are a complex pathophysiological process affecting the brain through a neurometabolic cascade, induced by traumatic biomechanical forces sufficient to temporarily disrupt mental status, and most often associated with symptomatology that includes headaches, dizziness, nausea, poor balance, and problems with cognitive functioning (Giza & Hovda, 2001; McCrory et al., 2009). mTBIs represent a major health concern for children and adolescents. The annual incidence of mTBI is estimated by the World Health Organization to be ∼0.6 per 100 people (Cassidy et al., 2004), with children and adolescents accounting for 45% of injuries (Canadian Institute for Health Information, 2007). In the youngest children with developing and maturing brains, approximately one of six children will sustain an mTBI before the age of 10 (Langlois, Rutland-Brown, & Thomas, 2006).
The majority of symptoms associated with mTBIs resolve within days to weeks, but a minority of people have symptoms that may persist beyond this timeframe (i.e., 11% of pediatric patients remain symptomatic at 3 months post-injury; Barlow et al., 2010). Fortunately, it is rare to have problems persist to the 12-month post-injury mark (e.g., 2.3%; Barlow et al., 2010; McCrory et al., 2009), but even 1–3 months of disruption to normal functioning is marked, especially in a child who is developing and expected to be acquiring new skills (Halstead & Walter, 2010). Tracking recovery from an mTBI is primarily based on the resolution of subjectively reported symptoms in comparison with baseline (premorbid) levels of functioning. Because there are no baseline data for the vast majority of children and adolescents, clinicians often rely on a comparison of current symptoms (obtained at some point post-injury, possibly weeks or months) to retrospectively reported premorbid symptoms as part of their determination of recovery from the mTBI.
Subjective reporting of premorbid functioning is vulnerable to normal human biases. Studies by Mittenberg and colleagues (1992) and Hilsabeck and colleagues (1998) demonstrated that persons with head injuries report fewer premorbid symptoms than would be found in healthy controls. This suggests that symptoms are attributed to the head injury and not to normal functioning. Although this was referred to as “expectation as etiology,” Gunstad and Suhr (2001) suggested that the paucity of reported premorbid symptoms is a “good old days” bias. The “good old days” bias was described as “... common for all individuals to selectively remember being healthier in the past and to fail to remember having common maladies” (p. 324, Gunstad & Suhr, 2001). The “good old days” bias infers a role in the perceived persistence of post-concussive symptomatology by minimizing the likelihood that these symptoms existed prior to the injury and attributing current symptoms directly to an injury (rather than being common experiences). The “good old days” bias has been demonstrated in adult mTBI populations when compared in a cross-sectional design with healthy community controls (Iverson, Lange, Brooks, & Rennison, 2010; Lange, Iverson, & Rose, 2010), with the bias being related to failure on performance validity testing (more bias in those who do not pass validity tests), but not litigation status.
The goals of this study were to: (a) longitudinally study the presence of the “good old days” bias in symptom reporting following an mTBI in children and adolescents, as well as (b) elucidate how the bias can impact clinical practice and research. The primary hypothesis was that the perceived level of baseline functioning will increase (i.e., retrospective ratings of premorbid symptoms will decrease) as participants get further in time from their injury; evidence of the “good old days” bias. A second hypothesis was that those with poor or slow recovery would show a stronger “good old days” bias compared with those who recover. A third hypothesis was that the “good old days” bias would lead to increased numbers of patients being considered to have elevated symptomatology by clinicians when relying on a comparison of current functioning to premorbid symptoms that are retrospectively reported at 1 and 3 months after the injury.
Participants included 412 children/adolescents (and their parents) between the ages of 2 and 18 who presented to the Emergency Department (ED). Inclusion criteria were sustained an injury to the head, a Glasgow Coma Scale score of 13–15 of 15, and at least one reported symptom (e.g., dizziness, confusion, headache, balance issues, nausea). Exclusion criteria included alcohol or drug use at the time of the injury/assessment, injury due to suspected child abuse, loss of consciousness for >30 min, an inability to complete the questionnaires, or sustaining a second injury during the study period (i.e., only one child presented to the ED with a second injury during his/her enrolment). Litigation status was not documented for this sample (in our experience [Barlow et al., 2010], this has been a rare occurrence in this population).
Parents completed the post-concussion symptom inventory (PCSI; Glass, Natale, Janusz, Gioia, & Anderson, 2005) to provide ratings of symptoms following their child's injury. Adolescents (13–18 years) also completed self-report PCSI ratings. The PCSI is a 26-item scale with somatic, cognitive, and emotional symptoms that are often reported following an mTBI. For the parent version, each symptom is rated on a 5-point scale of severity, ranging from “never” (0) to “almost always” (4) (total severity score range = 0–104). For the adolescent self-report version, each symptom is rated on a 7-point scale of severity, ranging from “never” (0) to “almost always” (6) (total severity score range = 0–156). This study used both a pre-injury (premorbid) version and a post-injury version, with ratings being based on functioning over a few days. The post-injury version also contains a single question for rating how different someone is compared with before their injury (a 5-point Likert rating scale, ranging from 0 [“no difference”] to 4 [“major difference”]). This single question of how different someone is compared with pre-injury is not included in the PCSI total score.
Proxy ratings of children's symptoms were obtained by parents: (a) while in the ED at the time of injury (n = 412), (b) over the phone at 1 month following the injury (n = 402), and (c) over the phone at 3 months following the injury (n = 141). One parent completed the current and premorbid symptom ratings in the ED, with the same parent completing the ratings at 1 and 3 months. Retrospective ratings completed by adolescents were obtained at (1) 1 month (n = 162) or (3) 3 months post-injury (n = 62). Not all parents and adolescents completed retrospective ratings at 3-months post-injury because this was added to the study design after data collection had already started. The collection of these data was approved by the University of Calgary Conjoint Health Research Ethics Board.
The total score (severity of symptoms) on the PCSI was the primary dependent variable for analyses. Dichotomizing of the sample into two groups (group 1: “recovered” or group 2: “not recovered”) was based on the single question on the PCSI asking how different a person is currently compared with before the injury (i.e., this is the 5-point Likert rating scale at the end of the PCSI mentioned previously, with rating of “no difference” being categorized as “recovered,” whereas rating any level of difference between 1 and 4 was categorized as “not recovered”). Recovery was based on either parent endorsement (for analyses of parent data) or adolescent self-report (for analyses of adolescent data). Repeated-measures analyses of variance (ANOVAs) and paired-samples t-tests were used to examine differences in symptom reporting over time. Comparisons of current symptoms to premorbid symptoms (reported in the ED, at 1 month, or at 3 months post-injury), to determine whether someone has the same or fewer symptoms versus more symptoms, were examined using chi-square analyses. Alpha was set a priori at p < .05 for all analyses.
The mean age of the 412 children was 10.9 years (SD = 4.4, range = 2–18 years), with 62.7% of the sample being boys and the majority being Caucasian (85.8%). Considering severity indicators for the injury, 94.4% had a Glasgow Coma Scale score of 15/15, 19.9% experienced some loss of consciousness (<30 min), and post-traumatic amnesia was reported in 32.2% of the sample. Of the 13.3% of patients who were sent for neuroimaging by the emergency physician, only seven had positive findings (24.1% of those sent for imaging, 1.7% of total sample) that suggested a complicated mTBI. For the 162 adolescents who also completed the self-report version of the PCSI, the mean age was 15.0 years (SD = 1.3, range = 13–18), with 61.7% of the sample being boys (and the majority being Caucasian).
Parent ratings of their children's premorbid functioning obtained in the ED, at 1 month, and at 3 months post-injury are presented in Figs 1 and 2. When considering those participants with or without parent proxy ratings at 3 months post-injury, there were not statistical differences in the children regarding age, F(1,412) = 0.68, p = .411, gender, χ2(1) = 0.11, p = .742, proportion with positive loss of consciousness, χ2(1) = 0.73, p = .394, proportion with positive post-traumatic amnesia, χ2(1) = 0.06, p = .810, or the proportion with positive post-traumatic confusion, χ2(1) = 0.04, p = .837.
Parent proxy ratings completed in the ED suggested a mean premorbid severity score of 7.1 (SD = 10.1; Fig. 1), with only 34.5% reporting no premorbid symptoms. There was a significant decline over time in premorbid symptom ratings, F(2,135) = 20.43, p < .001, suggesting that a “good old days” bias was present by 1-month post-injury—mean = 1.4, SD = 4.1, t(400) = 12.76, p < .001; 75.6% retrospectively report no premorbid symptoms—and did not change statistically from 1 to 3 months post-injury—mean = 0.8, SD = 2.5, t(139) = 0.93, p = .36; 83.0% retrospectively report no premorbid symptoms.
In those who did or did not recover (Fig. 2), again significantly fewer premorbid symptoms were reported over time in both samples—recovered, F(2,87) = 14.10, p < .001; not recovered, F(2,43) = 6.60, p = .003. The decline in premorbid symptoms was significant from the ED to 1-month post-injury in both groups—recovered, mean in ED = 6.0, SD = 8.8, mean at 1 month = 0.9, SD = 2.7, t(255) = 9.73, p < .001, Cohen's d = 0.88; not recovered, mean in ED = 9.3, SD = 12.0, mean at 1 month = 2.3, SD = 5.7, t(139) = 8.18, p < .001, Cohen's d = 0.79—but not significant from 1 to 3 months post-injury in both groups—recovered, mean at 3 months = 0.3, SD = 1.3, t(107) = 1.30, p = .20, Cohen's d = 0.24; not recovered, mean = 2.3, SD = 4.4, t(31) = 0.46, p = .65, Cohen's d = 0.01.
Adolescent ratings of premorbid symptoms obtained at 1 and 3 months post-injury are presented in Fig. 3. When considering those adolescents with or without self-reported premorbid ratings at 3 months, there were not statistical differences regarding age, F(1,160) = 0.09, p = .769, gender, χ2(1) = 0.15, p = .695, proportion with positive loss of consciousness, χ2(1) = 0.16, p = .693, proportion with positive post-traumatic amnesia, χ2(1) = 3.104, p = .078, or the proportion with positive post-traumatic confusion, χ2(1) = 1.08, p = .298. Adolescent ratings indicated a significant decline in premorbid symptoms from 1 month (mean = 6.8, SD = 11.9; 40.7% rated self as symptom-free before injury) to 3 months post-injury—mean = 5.0, SD = 14.6; t(55) = 2.91, p = .005; 54.8% rated self as symptom-free before injury. The decline in premorbid symptoms from 1 to 3 months was significant in those who reported being recovered (premorbid at 1 month, mean = 5.3, SD = 8.2; premorbid at 3 months, mean = 2.6, SD = 6.7), t(30) = 2.19, p = .037; Cohen's d = 0.37, but only approached significance in those who did not recover (premorbid at 1 month, mean = 9.4, SD = 15.7; premorbid at 3 months, mean = 4.7, SD = 7.5), t(24) = 2.03, p = .054, with a small-to-medium effect size (Cohen's d = 0.41).
When considering only those patients with possible protracted recovery (i.e., parent reports that their child is different), which is most representative of those patients who are referred to see a specialist or recruited into a research study, important differences emerged based on how a return to their premorbid level of symptoms was determined. The determination of a return to premorbid functioning for this study is based on a child having the same or fewer current symptoms compared with before they were injured, which was done by comparing current symptoms to either: (a) retrospective premorbid symptoms obtained at 1 or 3 months post-injury or (b) actual premorbid symptoms obtained in the ED. At 1-month post-injury and using “retrospective” premorbid symptoms (Fig. 4), only 7.9% of the children were determined to have returned to their premorbid level. However, at 1-month post-injury and using premorbid symptoms obtained in the ED (Fig. 4), 29.2% of the children were determined to have actually returned to their premorbid level, χ2(1) = 22.80, p < .001, odds ratio = 5.4 (95% confidence interval = 2.6–11.2). Similarly, at 3 months post-injury and using retrospective premorbid symptoms (Fig. 4), only 9.4% of the children were determined to have returned to their premorbid level. However, at 3 months post-injury and using premorbid symptoms obtained in the ED (Fig. 4), 41.1% of the children were determined to have actually returned to their premorbid level, χ2(1) = 10.40, p < .001, odds ratio = 6.7 (95% confidence interval = 1.9–24.2).
Persistence of post-concussion symptoms suggests ongoing and/or poor recovery from an mTBI, which often necessitates increased healthcare resources. However, the determination of symptom persistence following an mTBI is typically based on a retrospective account of functioning prior to the injury and clinicians trust that this account is accurate. In the present sample of 412 parent–child pairs, retrospectively reported premorbid symptoms decreased by 80% from the ED to 1-month post-injury. Moreover, the number of children rated as having zero premorbid symptoms goes from only 35% in the ED to 76% at 1 month and 83% at 3 months. Although there did not appear to be an increased effect of the “good old days” bias on parent report at 3 months post-injury (perhaps due to the symptom scores already being quite low, representing a basal effect), adolescent self-report suggests ongoing decreases in retrospectively reported premorbid symptoms up to 3 months post-injury. It is postulated that these decreases over time in premorbid symptoms reflect the “good old days” bias.
The presence of a “good old days” bias impacting retrospective symptom reporting following an mTBI in this pediatric sample is consistent with the previous literature in adult samples (Gunstad & Suhr, 2001; Iverson et al., 2010; Lange et al., 2010). Adult mTBI patients from a hospital-based clinic retrospectively reported fewer premorbid symptoms at 1.8 months post-injury than healthy control volunteers from the community, although this was only a small effect size (Lange et al., 2010). Iverson and colleagues (2010) reported a medium-to-large effect size when comparing retrospective premorbid symptom reporting by mTBI patients attending a worker's compensation clinic to the same sample of healthy volunteers from the community. Both of these adult studies employed a cross-sectional methodology, rather than longitudinal, and the comparison group was younger (Iverson et al., 2010), better educated (Iverson et al., 2010; Lange et al., 2010), and had a higher proportion of women (Iverson et al., 2010). Despite the methodological differences between the current study and Lange and colleagues (2010) and Iverson and colleagues (2010), it is likely that retrospective reporting of symptoms is subject to a “good old days” bias following an mTBI, but perhaps not in non-injured healthy adults (Sullivan & Edmed, 2012).
One aspect of the “good old days” bias to consider is whether a slower recovery (i.e., increased duration since a negative event) results in a larger bias. In our longitudinal study with pediatric mTBI patients, the findings did not support our second hypothesis. Parent-ratings for both those who recovered and those who did not recover demonstrated a “good old days” bias that impacted the perception of premorbid symptoms. Although adolescent data may suggest that those who recover are more likely to exhibit the bias, the small-to-medium effect size (d = 0.41) coupled with an alpha that approached significance (p = .054) suggest a lack of power and suggest that both those who do and those who do not recover exhibit the “good old days” bias. The exact reason(s) why a protracted recovery did not result in a stronger bias on premorbid ratings is not clear, but it is possible that the bias is established very soon after the injury (sometime between injury and the 1-month follow-up) and therefore is not altered by the effects of recovery or non-recovery at later time points. Obviously, the effect of recovery or non-recovery on the extent of the bias warrants further exploration.
Disentangling the effects of a concussion, accounting for pre-existing problems, and monitoring recovery over time is truly a tour de force for any clinical setting. Typically, a clinician relies on a retrospective account of premorbid functioning as a marker of recovery for current symptoms. However, as this study has clearly demonstrated, the retrospective recall of said symptoms is subject to bias as early as 1-month post-injury. Perhaps one of the more important findings of this study is that the clinical implications of a “good old days” bias on symptom reporting are tangible, demonstrable, and quantifiable. When making a determination of whether someone has returned to their premorbid level of symptoms, considerable differences are found depending on which premorbid symptoms are used for the comparison. Using the premorbid symptoms reported in the ED, rather than retrospectively reported at the follow-ups, results in up to one-third fewer patients being deemed to have symptomatology that exceeds their baseline. Considering the prevalence and healthcare costs associated with an mTBI, this 5–7-fold improvement in a clinician's identification of a patient's return to premorbid functioning would be substantial.
There are some limitations of this study to identify. First, premorbid symptoms were not rated by the adolescents at the time of injury. As a result, it is unknown how the adolescents would have rated their premorbid functioning in the absence of the “good old days” bias and only parent–proxy ratings were available. Of course, it was not plausible from our perspective to have an acutely injured patient provide ratings prior to their injury while in the ED. In fact, these ratings could have been impacted by the acute effects of the concussion and likely would have been already subjected to the effects of the “good old days” bias. Second, although the PCSI can be administered to children as young as 5 years, ratings by children 5–12 years old were not obtained. As a result, only parent–proxy ratings were available for those 12 years or younger. Third, the determination of “recovered” or “not recovered” was subjectively determined by parents and adolescents based on a single question. Due to the subjective nature of all ratings, it is entirely reasonable to believe that this determination of recovery was also influenced by the “good old days” bias. It is also possible that the subjective determination of recovery is partially (or entirely) responsible for an absence of differences in the bias between those who do and those who do not recover. An objective measure of recovery (e.g., biomarker) may produce different results. Fourth, ratings of symptoms at the 1- and 3-month follow-ups were completed over the phone, which could potentially change premorbid symptom reporting. However, the questionnaires were read over the phone question by question with the Likert rating scale being repeated each time, in order to ensure that each question would be answered in a similar manner to having the questionnaire in front of them. Fifth, not every patient provided retrospective ratings at 3 months post-injury because the study initially only included these ratings in the ED and at 1 month. Regardless, the sample size is large enough to study retrospective symptom rating longitudinally at several time points. Sixth, this study did not employ a control group. Although it would be interesting to examine whether the “good old days” bias also holds up in non-head injured patients, the goal of the present study was to employ a longitudinal within-subject design to study how this bias impacted the mTBI group specifically over time. As well, the premorbid ratings obtained in the ED actually served as the patients' own control data for subsequent ratings, thus eliminating the need for a control group. And finally, this study did not include measures of performance validity (demonstrated to increase the “good old days” bias by Iverson et al., 2010) or documentation of whether any families became involved in litigation during their enrolment in this study (although Lange et al., 2010 reported that litigation was not deemed to be related to premorbid symptom ratings in adults, this has not been investigated in children). Further investigations into the effects of failure on performance validity tests or ongoing litigation status on the “good old days” bias are warranted in children.
At this time, it is recommended that clinicians who assess patients following an mTBI should: (a) obtain baseline (premorbid) ratings of functioning as soon as possible after any injury (i.e., the bias is present by 1 month in this sample, but may actually emerge even sooner after the injury) and (b) use the earliest baseline symptom ratings as a point of comparison for determining recovery from the mTBI. As well, the development of normative data and reliable change indices for symptom scales could be an appropriate method for overcoming this bias (i.e., PCS scales remain as one of the very few scales without good normative comparisons). Because post-concussion symptom scales remain as the gold standard for determining outcome, adjusting the methodology for how premorbid functioning is quantified and used to determine recovery following an mTBI will be important for clinical and research settings.
This study contributes to the expanding literature on how our current methodology for collecting symptom ratings following an mTBI can be influenced through biases (Gunstad & Suhr, 2001, 2004; Hilsabeck et al., 1998; Iverson, Brooks, & Holdnack, 2008; Iverson et al., 2010; Lange et al., 2010; Mittenberg et al., 1992; Panayiotou, Crowe, & Jackson, 2011; Sullivan & Edmed, 2012) or even through the modality used to collect the data (Krol, Mrazik, Naidu, Brooks, & Iverson, 2011; Lange et al., 2010). The “good old days” bias is present in pediatric mTBI and influences premorbid symptom reporting. To our knowledge, this is the first study to demonstrate this bias longitudinally and in a pediatric sample, but perhaps more importantly, the first to tangibly demonstrate the implications of the bias on determining recovery in clinical and research settings. The impact from the “good old days” bias is not trivial, suggesting that accounting for this bias could potentially reduce the number of pediatric patients being deemed to have elevated symptomatology following an injury.
Funding for this study provided by an Alberta Children's Hospital Foundation project grant (grant #10000779) awarded to AM and KMB, and a summer studentship from the Canadian Institutes of Health Research training program to BK. None of the funding sources were involved in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Conflict of Interest
None of the authors have a financial interest in any of the measures in this study. BLB receives funding from a test publisher (Psychological Assessment Resources, Inc.), book royalties from Oxford University Press, and in-kind test credits for research from a computerized test publisher (CNS Vital Signs).
The authors thank Janie Williamson RN (manager, Pediatric Emergency Research Team, Alberta Children's Hospital) and the Pediatric Emergency Medicine Research Assistant Program (PEMRAP) students for recruitment and testing of patients in the emergency department, Samna Khan (McGill University, funded by a summer studentship from the Canadian Institutes of Health Research training program) and Aneesh Khetani (University of Calgary) for additional help with data collection, and Helen Carlson PhD (Neurosciences program, Alberta Children's Hospital) for assistance with formatting this manuscript.