The concept of postconcussional syndrome (PCS), which refers to a symptom cluster following traumatic brain injury (TBI),1 has gained significance in light of recent data showing that incidence of mild TBI is much higher than previously thought. According to a 2003 report by the Centers for Disease Control (CDC),2 mild traumatic brain injury (TBI) results in approximately 500,000 emergency department (ED) visits without hospitalization and almost 200,000 hospitalizations per year in the U.S. Given the high incidence of mild TBI and publication of diagnostic criteria for PCS,3—5 it is surprising that no prospective studies on the criteria-based diagnosis of PCS are available. To our knowledge, this study is the first to prospectively measure the prevalence of the criteria-based diagnosis of PCS and to evaluate the specificity of PCS diagnostic criteria to brain injury. Results of this study may be important in refining diagnostic criteria and to guiding clinical decision making.
Diagnostic criteria for PCS were first proposed in 1992 in the International Classification of Diseases, Tenth Revision (ICD-10), which included clinical and research criteria sets for PCS (Code F07.2).3,4 ICD-10 clinical criteria3 require a history of TBI and the presence of three or more of the following eight symptoms: 1) headache, 2) dizziness, 3) fatigue, 4) irritability, 5) insomnia, 6) concentration or 7) memory difficulty, and 8) intolerance of stress, emotion, or alcohol. During preparation of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), the diagnosis of postconcussional disorder was proposed.6 Criteria for this diagnosis were published in modified form in a DSM-IV appendix of provisional criteria sets designated as needing further research.5 DSM-IV criteria are: A) history of TBI causing "significant cerebral concussion;" B) cognitive deficit in attention and/or memory; C) presence of at least three of eight symptoms (e.g., fatigue, sleep disturbance, headache, dizziness, irritability, affective disturbance, personality change, apathy) that appear after injury and persist for ≥3 months; D) symptoms that begin or worsen after injury; E) interference with social role functioning; and F) exclusion of dementia due to head trauma and other disorders that better account for the symptoms. Criteria C and D set a symptom threshold that requires symptom onset or worsening to be contiguous to the injury, distinguishable from pre-existing symptoms, and have a minimum duration.
We hypothesized that ICD-10 criteria would be more inclusive than DSM-IV criteria, that both ICD-10 and DSM-IV criteria would be fulfilled more often by patients with brain injury than by those with extracranial trauma, and that DSM-IV criteria would have greater specificity to TBI.
Patients with TBI or trauma not involving the brain were prospectively recruited from January 1999 to February 2001 at Ben Taub General Hospital, a Level I trauma center in Houston, Texas. Study criteria for all patients were: arrival to the trauma center within 24 hours of injury; blood alcohol level of <200 mg/dL; 16 years of age or older; fluent English or Spanish speaker; resident in the hospital catchment area (Harris County, Texas); not an undocumented alien, incarcerated, homeless, or on active duty military service; no spinal cord injury; no previous hospitalization for TBI; no substance dependence,5 mental retardation, psychosis, or previous central nervous system disturbance; and no other condition preventing standard administration and interpretation of the outcome measures.
Additional TBI study criteria were: diagnosis of nonpenetrating TBI; lowest postresuscitation Glasgow Coma Scale7 (GCS) score >8; computed tomography (CT) scan of the brain performed within 24 hours of arrival; and if mild TBI, no surgery under general anesthesia during the admission. Additional study criteria for extracranial trauma were: diagnosis of trauma not involving the brain and no evidence of brain injury.
Of 448 TBI and 256 extracranial trauma patients who met the above criteria and consented , 178 TBI and 104 extracranial trauma patients were followed up 3 months after injury. Follow-up rates of TBI (39.7%) and extracranial trauma patients (40.6%) did not differ significantly. t1 and t2 present sample demographic and injury features. Of the TBI patients, 161 (90%) were classified as mild TBI, defined as a closed head injury producing a GCS score of 13—15 upon arrival with no later deterioration of GCS <13. Seventeen patients (10%) were classified as moderate TBI, defined as a GCS score of 9—12 upon arrival with no later deterioration of GCS <9. The difference in Injury Severity Score between the TBI and extracranial trauma groups was not statistically significant (t = 1.34, 280 df, p = 0.18).
Attrition bias was evaluated by t-tests for continuous variables and chi-square tests for categorical variables, at the 95% confidence level. Compared to patients who were successfully followed up, those lost to follow-up had less formal education and higher rates of males, assault, primary Spanish speakers and Hispanic ethnicity. Follow-up was not significantly associated with age, GCS and Injury Severity Score.
Injury Severity Measures.
The GCS score7 is the sum of ratings of eye-opening, motor responsiveness, and verbalization. The Injury Severity Score (ISS),8 as modified for this study, is the sum of squares of the highest score on the Abbreviated Injury Scale (AIS)9 in each of the three most severely injured body regions, excluding the head. Higher ISS and lower GCS scores indicate greater injury severity.
Structured Interview for PCS.
A structured interview (available from the authors) was created to evaluate DSM-IV and ICD-10 criteria for PCS. The interview queried the DSM-IV and ICD-10 postconcussional symptoms and the DSM-IV clinical significance criterion. For each symptom, the interview determined if the DSM-IV and/or ICD-10 thresholds were met.
The DSM-IV cognitive deficit criterion was evaluated by six scores derived from three neuropsychological tests. The Rey Complex Figure10 is a measure of visual memory in which a design is copied, redrawn from memory immediately afterward, and redrawn again after 30 minutes. In this study, the single test score used was the number of design elements correctly redrawn at delay (possible range 0—36). The Selective Reminding Test10 is a measure of verbal memory in which a word list is recalled after 6 repetitions and again 30 minutes later. The three scores used were sum of long-term storage (possible range 0—72), sum of consistent long-term retrieval (0—72), and delayed recall (0—12). The measure of attention was the Symbol Digit Modalities Test11 in which visual symbols are converted to digits using written and spoken responses. The two scores used were the number of correct written and spoken substitutions (possible range 0—110) in 90 seconds. For all three tests, higher scores indicate better performance. These tests have been recommended by two National Institutes of Health panels12,13 for use in TBI research as measures of attention and memory. It is understood that these tests are not cognitively pure and that effects of other impairments may be present, such as visuoconstructive and executive-function deficits in the Rey Complex Figure.
The threshold for classifying a test score as abnormal was statistically adjusted for demographic background according to an accepted psychometric procedure.14 First, a linear model was fitted to predict each of the six test scores from age (years), gender, education (years), racial-ethnic group (African-American, Hispanic, non-Hispanic Caucasian, other), and language of test administration (English, Spanish). Second, a residual score was obtained for each patient by subtracting the demographically predicted score from the observed test score. The distribution of residual scores defined the normal range of test performance, independent of demographic background. Finally, each patient’s residual score was compared to this distribution and was classified as abnormal if >1 standard deviation from average, in the direction of impairment. This method of demographic adjustment was used because our sample was much more racially/ethnically diverse than available norms for these tests. A potential disadvantage of this procedure is that the prevalence of cognitive deficit in the extracranial trauma patients was statistically defined rather than empirically estimated. However, the procedure did not bias the difference in prevalence of cognitive deficit between TBI and extracranial trauma patients, which was the variable of interest.
Diagnosis of Postconcussional Disorder (DSM-IV).
The DSM-IV diagnosis of postconcussional disorder5 was given if criteria C, D, and E were satisfied by the structured interview and if at least one of the six neuropsychological test scores was statistically abnormal as defined above. Criterion A (history of TBI) was omitted from analysis because the study design required this criterion to be met by all patients with TBI and by none with extracranial trauma. For patients evaluated before 3 months after injury, a symptom was considered to meet the duration threshold of criterion C if the symptom persisted through the entire postinjury interval and fulfilled all other requirements. Criterion F (exclusion) was presumed for all patients because the study selection criteria excluded cases with preexisting neurologic disorders and because Dementia Due to Head Trauma was unexpected in the largely mild TBI sample.
Diagnosis of PCS (ICD-10).
The ICD-10 diagnosis of PCS3 was given if at the time of evaluation, three or more of the symptoms in the ICD-10 definition were reported to have been present for at least one week. The duration threshold was set ad hoc because it is unspecified in ICD-10. The TBI history criterion was omitted from analysis.
Patients were recruited at the trauma center ED or during hospitalization. Diagnoses of TBI and extracranial trauma were made by trauma center physicians, GCS ratings by trauma center staff, and ISS ratings by an AIS-certified research nurse. The ED protocol for management of patients with a history of impaired consciousness included a CT scan of the head. Scans were reinterpreted by the project neuroradiologist from the original films. The structured interview and neuropsychological tests were administered by bachelors- or masters-level personnel at an average of 85.6 days (standard deviation = 20.1) after injury. The study protocol was approved by the institutional review boards of participating institutions.
Prevalence of PCS was compared between DSM-IV and ICD-10 criteria with the McNemar test for paired observations. The odds ratio (OR) and 95% confidence interval (CI) were used to evaluate the extent to which the PCS diagnosis and criteria were specific to TBI. A higher odds ratio reflects greater specificity by indicating that PCS was more prevalent in patients with TBI than those with extracranial trauma. Specificity of the DSM-IV and ICD-10 diagnoses to TBI was compared by the CI of the difference between the two odds ratios. The 95% confidence level was used throughout.
t3 shows the numbers of brain-injured and extracranial trauma patients who did and did not meet the DSM-IV and ICD-10 diagnostic criteria for PCS (other than the history of TBI criterion). It is seen that the prevalence of PCS was about six times higher with ICD-10 than with DSM-IV criteria. The same discrepancy was seen whether or not brain injury was present. The difference in prevalence between ICD-10 and DSM-IV criteria was highly significant in both TBI (χ2 = 93.01, 1 df, p <0.001) and extracranial trauma patients (χ2 = 33.03, 1 df, p < 0.001).
Specificity of PCS Criteria to Brain Injury
The odds ratios in t3, representing the specificity of the PCS criteria to brain injury, indicate that the TBI patients fulfilled both DSM-IV and ICD-10 criteria more often than did the extracranial trauma patients. However, the odds ratio reached statistical significance for the ICD-10 but not for the DSM-IV criteria. Therefore, support for the specificity of PCS criteria to brain injury was limited.
The difference between the odds ratios in t3, each representing the specificity of DSM-IV or ICD-10 criteria, fell short of conventional statistical significance (95% CI: 0.96—1.71). The lower odds ratio for the DSM-IV criteria is in the direction opposite to the hypothesis that DSM-IV criteria would be more specific to TBI.
Brain-injured patients with CT intracranial abnormalities had no greater prevalence of PCS, using either DSM-IV (OR = 1.62, 95% CI: 0.58—4.55) or ICD-10 criteria (OR = 0.82, 95% CI: 0.32—2.13).
Specificity of DSM-IV criteria
t4 compares the numbers of TBI and extracranial trauma patients who did and did not meet the DSM-IV criteria requiring cognitive deficit, postconcussional symptoms, and clinical significance. The two DSM-IV symptom criteria (C and D) were combined for analysis. The odds ratios in t4 represent the specificity to TBI of each DSM-IV criterion. Although all criteria were fulfilled at a higher rate after TBI than after extracranial trauma, the odds ratio was statistically significant only for the symptom criterion. The odds ratios for the cognitive deficit and clinical significance criteria did not reach statistical significance. These results indicate that while the DSM-IV symptom criteria have some specificity to TBI, specificity of the DSM-IV cognitive deficit and clinical significance criteria was limited.
This is the first prospective study to describe the prevalence and specificity of the criteria-based diagnosis of PCS and to compare between DSM-IV postconcussional disorder and ICD-10 PCS. The results address the recommendations by CDC for research on symptoms following mild TBI2 and by the World Health Organization (WHO) Task Force on Mild TBI15 for studies to prospectively compare outcomes after mild TBI with appropriate control groups. The major findings are that there was a large difference between the prevalence of PCS using DSM-IV and ICD-10 criteria, that ICD-10 criteria were the more inclusive of the two criteria sets, and that the DSM-IV and ICD-10 criteria sets both had limited specificity to brain injury because PCS criteria can be met after trauma whether or not the brain is injured.
The higher prevalence of ICD-10 PCS, which was predicted based on the smaller ICD-10 criteria set, was observed in patients with and without brain injury. The lower prevalence of DSM-IV postconcussional disorder can probably be explained by the exclusion of the majority of patients by the DSM-IV cognitive deficit and clinical significance criteria. Although the DSM-IV criteria define a higher symptom threshold, requiring a 3-month duration and discriminability from preexisting symptoms, inequality of symptom thresholds was not a major discriminating factor.
The limited specificity of both PCS criteria sets to brain injury can probably be explained by the fact that PCS criteria (other than the history of TBI criterion) were often fulfilled by trauma patients who did not have injury to the brain. Because the DSM-IV cognitive deficit and clinical significance criteria were not specific to TBI, they did not improve the specificity of the DSM-IV criteria set. Future research could compare symptom reports to neuropsychological test results in order to determine if the TBI and extracranial trauma patients differ in over- or underestimating their true impairments. Our findings do not necessarily imply that the ICD-10 criteria have greater specificity to TBI, because a direct comparison between the DSM-IV and ICD-10 odds ratios did not reach conventional statistical significance. In addition, the relatively low prevalence of the DSM-IV postconcussional disorder would have curtailed the power of statistical tests with this criteria set.
To implement the DSM-IV cognitive deficit and exclusion criteria, some ad hoc decisions were necessary. First, in order to maximize detection of cognitively impaired patients, we set the threshold for cognitive deficit to be very low. The alternative strategy would have been to set the cognitive deficit threshold higher, for example by requiring a greater number of abnormal test scores. This alternative strategy would have reduced the prevalence of cognitive deficit in all patients, without changing the difference in prevalence of cognitive deficit between the TBI and extracranial trauma groups. Additional problems of the DSM-IV cognitive deficit criterion are that cognitive impairment after TBI tends to resolve by 3 months postinjury, as concluded by the recent WHO task force,16 and that the criterion is incompatible with ICD-10 research criteria.4 There is a structural incompatibility between the DSM-IV cognitive impairment criterion and the ICD-10 criteria for research (different from the ICD-10 clinical criteria used in this study) which exclude the diagnosis of PCS in the presence of cognitive impairment.4
The diagnosis of DSM-IV postconcussional disorder is to be excluded when a patient meets criteria for Dementia Due to Head Trauma, which is probably rare after mild TBI, and when another mental disorder better accounts for the symptoms.5 In this study we decided to allow the diagnosis of postconcussional disorder in patients who met criteria for other anxiety and mood disorders, since DSM-IV does not exclude this possibility. Exclusion of postconcussional disorder in these cases would have further widened the discrepancy between the prevalence of DSM-IV and ICD-10 criteria.
Strengths of this study include the prospective cohort design, defined source population and sampling frame, appropriate comparison group, sample size, inclusion of nonhospitalized patients, and evaluation of attrition bias.15 Limitations of the study are, first, that attrition was substantial and was associated with demographic background. However, attrition bias for injury severity was not detected and there was no evidence that attrition influenced the prevalence or specificity of PCS criteria. Second, prevalence of PCS could have been overestimated in brain-injured patients because the purpose of the study was disclosed to patients and because injury history was not concealed from the interviewers. Third, the single follow-up evaluation at 3 months postinjury did not allow for change in PCS prevalence to be measured. Requiring a shorter symptom duration in patients evaluated before 3 months after injury might have caused the prevalence of DSM-IV postconcussional disorder to be overestimated. Fourth, recruitment was confined to a single trauma center, excluding patients who were untreated or who were treated in nonhospital settings. Finally, by excluding patients with GCS scores <8, the results do not address PCS after severe brain injury.
Our finding of a large difference in prevalence between the DSM-IV and ICD-10 diagnoses, if confirmed in future research, would imply that the two PCS criteria sets will often result in diagnostic disagreement. The observed frequency of PCS symptoms in patients without brain injury supports the recommendation of the CDC report2 that complaints of postconcussional symptoms are not sufficient to make the diagnosis of mild TBI. The findings also support the WHO task force16 criticism that in diagnosing PCS, linking residual symptoms to TBI is the major problem. These difficulties suggest that refinement of the PCS diagnostic criteria of DSM-IV and ICD-10 is needed before the criteria can be recommended for routine use.
This study was supported by grants R49/CCR612707-01 (Depression after Mild to Moderate Traumatic Brain Injury, Principal Investigator: Harvey S. Levin) from the Centers for Disease Control and H133A70015 (Traumatic Brain Injury Model System of TIRR; Principal Investigator: Walter M. High, Jr.) from the National Institute on Disability and Rehabilitation Research.
This study was also presented in part at the meeting of the International Neuropsychological Society, Toronto, Ontario, Canada, February 13—16, 2002.