Defining “therapeutically inconsequential” head computed tomographic findings in patients with blunt head trauma☆☆☆

1. Introduction 

Each year, well over a million patients present to US and Canadian emergency departments (EDs) for evaluation after blunt head trauma.1., 2., 3., 4. Most patients have minor injuries and require no immediate investigation, but approximately 20% undergo computed tomographic (CT) head scanning,2., 5. which reveals traumatic abnormalities in less than 20% of imaged patients.6., 7., 8., 9., 10. The small proportion of abnormal scans has stimulated research into reducing unnecessary imaging.7., 10., 11., 12., 13., 14., 15., 16., 17., 18. Although many detected lesions require specific directed care, some do not.7., 10., 11. Many patients with abnormal scans are admitted for serial examination, but recent work suggests that observation may not be necessary for all injuries.19 If it proves possible to distinguish which patients with abnormal findings do not require any specific intervention or proceed to poor outcomes, the following benefits could be realized: (1) there could be further reductions in CT imaging of patients with head injury, with an associated reduction in radiation exposure; (2) health care costs associated with imaging, admission, and observation could be lessened; (3) more efficient decision instruments for CT imaging could be created because the instruments would not have to identify all patients with “abnormal” imaging but instead would focus on identifying only patients with “therapeutically consequential” injuries.

Previous work on adult blunt head trauma suggests that many abnormalities identified on CT scanning are “therapeutically inconsequential” and rarely, if ever, require directed interventions.20., 21., 22., 23., 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., 36., 37., 38., 39. This work proposes that the CT findings listed in Figure 1 are important in terms of intervention, but speculation centers on CT scans that lack these findings. Confirmation that these lesions, in concert with the clinical examination, are therapeutically inconsequential would obviate the need to identify them when they occur in isolation (in the absence of other more serious injuries).

View Large Image Download to PowerPoint

Figure 1. CT abnormalities considered to be important from a neurosurgical therapeutic standpoint.

This is a substudy of the National Emergency X-radiography Utilization Study (NEXUS) II, a study that seeks to develop a clinical decision instrument to guide CT imaging of patients with minor head injury. This substudy has 3 goals: (1) to describe the prevalence of therapeutically inconsequential CT findings among patients with blunt head injury who undergo CT scanning; (2) to determine the frequency of “important neurosurgical outcomes” among these patients (Figure 2); and (3) to estimate whether altered mentation and coagulopathy are important codeterminants of outcome in such patients. When CT scans are considered separately from the clinical findings, we expect that CT scans with therapeutically inconsequential findings will be associated with some poor outcomes because we know that even normal CT scan results can be associated with a poor outcome (eg, diffuse axonal injury). Subsequently, the incidence of important neurosurgical outcomes in the group with therapeutically inconsequential CT findings should be at least as high as the incidence within the group of head-injured patients with normal CT scan results, if taken out of context from clinical information such as a low Glasgow Coma Scale score. Therefore, we chose to evaluate clinical findings to see whether we could identify the anticipated important neurosurgical outcomes among the patients with “therapeutically significant” CT findings. We ultimately seek to determine whether patients with these CT abnormalities require any specific clinical response and, therefore, whether it is important to obtain a CT scan for these patients solely to avoid overlooking one of these lesions.

View Large Image Download to PowerPoint

Figure 2. Study definitions.

2. Materials and methods 

We conducted a multicenter, prospective, observational study of blunt head trauma patients who underwent CT imaging. The protocols and methodology were reviewed and approved by the human subjects committees of the participating institutions.

2.3. Data collection and processing 

On enrollment and before CT imaging, managing clinicians collected data on age, sex, and the presence or absence of 18 specific signs and symptoms associated with head injury, including 2 variables, “altered level of alertness” and “coagulopathy,” identified a priori as being important predictors of outcome in patients with minimal or no injury evident on CT (Figure 3). Patient information was entered into a dedicated computer, which generated an imaging requisition or voucher. To enforce data collection before head CT scanning, radiology technicians performed head CT imaging only when they received such a voucher. Incomplete enrollment data could be entered into the computer for hemodynamically unstable patients and were completed at the earliest appropriate opportunity.

View Large Image Download to PowerPoint

Figure 3. Blunt head trauma questionnaire.

We collected initial and follow-up CT reports for each enrolled patient at least 1 month after the date of injury, according to NEXUS II protocol. From these reports, we identified all patients who had initial CT findings that met predefined criteria for therapeutically inconsequential. Figure 2 contains an explicit definition of a therapeutically inconsequential CT scan, along with other study definitions. Two trained reviewers (CA and AHS) independently evaluated a subset of 100 CT report summaries (consisting of 50 therapeutically inconsequential reports and 50 reports of more severe injury) to assess interrater agreement in identifying cases with therapeutically inconsequential findings.

Discharge summaries, notes from follow-up appointments, and CT reports were collected from the 6 sites participating in the follow-up portion of the study. From these records, we extracted a presenting Glasgow Coma Scale score,40 mechanism of injury, coagulation status (which was also collected prospectively, but some patients had a level of consciousness that precluded informing the managing physician), and whether the patient received any “directed intervention.” The same records were used to calculate the Glasgow Outcome Scale score41 at each patient's final, charted evaluation. A second reviewer completed independent abstractions about use of directed intervention and Glasgow Outcome Scale score; subsequently, interobserver agreement was assessed using the weighted κ statistic.42 Outcomes involving disagreement between abstractors were resolved by third-party review by a coinvestigator (WRM).

3. Results 

From April 1999 to December 2000, the 18 sites enrolled 8,374 blunt head trauma victims. Abnormal traumatic findings were identified on 863 CT scans (10.3%; 95% confidence interval [CI] 9.7% to 11.0%), including 155 scans exhibiting isolated therapeutically inconsequential lesions (18.0% of abnormal scans; 95% CI 15.5% to 20.7%). The prevalence of therapeutically inconsequential findings among all imaged patients was 1.9% (95% CI 1.6% to 2.2%).

The population with therapeutically inconsequential injuries exhibited a unimodal age distribution (median 39 years; interquartile range 25 to 60 years) in which 20- to 31-year-old patients were the largest group. Men constituted 66% (n=103) of the group (Figure 4). Figure 5 details the various mechanisms of injury.

View Large Image Download to PowerPoint

Figure 4. Age and sex of 155 head trauma patients with therapeutically inconsequential CT findings.

View Large Image Download to PowerPoint

Figure 5. Age and mechanism of injury in 155 head trauma patients with therapeutically inconsequential CT findings. MVC, Motor vehicle crash.

Of the 155 patients with therapeutically inconsequential injuries, 81 presented to 1 of the 6 sites obtaining detailed outcome information. Median follow-up was 1 week (interquartile range 4 days to 2.5 weeks), with a mean of 6 weeks. Five of the 81 patients had less than 24 hours of follow-up information: 4 were discharged to home from the ED in stable condition, and the fifth patient was admitted to the hospital for overnight observation for possible cervical spine injury. All 5 patients remained unchanged from baseline neurologic status at admission and were assigned a Glasgow Outcome Scale score of 5 according to their functional capacity at their final evaluation.

Eight of 81 patients (10%) with a therapeutically inconsequential CT finding received some form of directed intervention (Table 1). According to the Glasgow Outcome Scale score, 73 (90%) of the 81 patients had a “functional outcome” (Glasgow Outcome Scale score 4 or 5), whereas 8 (10%) patients had a “poor outcome” (Glasgow Outcome Scale score 1, 2, or 3; Table 2). Combining both outcome measures, 71 (88%) patients did well, whereas 10 (12%) patients had an important neurosurgical outcome.

Table 1. Ten patients requiring neurosurgical intervention, with a poor outcome, or both.

	Patient Age, y/Sex	Presentation	Initial CT Summary	Directed Intervention	GOS Score	Abnormal Alertness	Abnormal Behavior	Coagulopathy
	25/M	GCS 8	Normal/tiny focus anterior limb of the right internal capsule likely a nonspecific focus of calcification; CT 2: small left temporal hemorrhagic contusion is not larger but is now associated with minimal edema	ICP monitor	4	Yes	Yes	INR 1.2
	27/F	GCS 8	Hemorrhagic contusion (3 mm in the right medial frontal lobe)	None	3	Yes	Yes	Normal
	37/M	GCS 9	Few tiny areas of contusion in high left cortical area	None	3	Yes	Yes	Normal
	39/F	GCS 3, full cardiac arrest, precipitating DIC	SAH (occipital/parietal junction); possible hemorrhage interpeduncular cistern	ICP monitor	3	Yes	Yes	DIC
	40/M	GCS 14: oriented, eyes were open, able to obey verbal commands but level of alertness and behavior were markedly abnormal	Intraparenchymal hemorrhage (small right inferior frontal area); SAH (right anterior inferior frontal lobe)	Craniotomy	4	Yes	No	INR 3.4
	46/F	Insufficient data available for follow-up (marked “unknown” for most presenting signs and symptoms)	Normal; CT 2: (8 h later) small SAH in intrapeduncular fossa and over the convexity	ICP monitor	1	Unknown	Unknown	Normal
	59/F	GCS 15: oriented, eyes open, and able to obey verbal commands but altered level of alertness; alcohol was involved; DNR order	Right interhemispheric bleed (probably SDH) anteriorly	Craniotomy	1	Yes	No	INR 1.4
	73/F	GCS 14: able to obey verbal commands but not oriented, altered level of alertness, complained of headache; severely coagulopathic postcraniotomy	SAH (cortical sulci and Sylvian fissure)	Craniotomy	1	Yes	No	INR 3.1
	82/M	GCS 14: eyes shut, oriented and able to obey verbal commands, altered level of alertness, alcohol was involved	SDH (small, left temporal/parietal area); SAH (cortical sulci, right parietal area)	ICP monitor	2	Yes	Yes	Normal
	85/M	GCS 14: eyes open, able to obey verbal commands but alcohol was involved, vomited repeatedly	SDH (left temporal lobe, small); SAH (small, in left temporal and frontal/parietal junction sulci)	Craniotomy	3	No	Unknown	Unknown

GOS, Glasgow Outcome Scale; M, male; F, female; GCS, Glasgow Coma Scale; ICP, intracranial pressure; DIC, disseminated intravascular coagulation; SAH, subarachnoid hemorrhage; SDH, subdural hematoma; DNR, do not resuscitate.

Table 2. Final outcomes among 81 patients with therapeutically inconsequential injuries.

	Outcome (GOS Score)	No. of Patients (%)	No. of Directed Interventions	Final Outcome Classification
	5	65 (80)	0	Functional
	4	8 (10)	2	Functional
	3	4 (5)	2	Poor
	2	1 (1)	1	Poor
	1	3 (4)	3	Poor

Presenting Glasgow Coma Scale scores were recorded for 80 patients and were less than 15 for 8 of the 9 patients who had a known Glasgow Coma Scale score and an important neurosurgical outcome. The ninth patient with a Glasgow Coma Scale score of 15 had an altered level of alertness as a result of alcohol consumption, as well as an elevated international normalized ratio (INR). The initial Glasgow Coma Scale score and level of alertness were not recorded in a tenth patient with an important neurosurgical outcome. By comparison, the initial Glasgow Coma Scale score was abnormal (<15) in 23 (32%) of 71 patients who had neither a directed intervention nor a poor Glasgow Outcome Scale score.

Coagulopathy was documented for 7 (9%) patients and excluded in 67 (91%) others; coagulation status was unknown in 7 patients. Two coagulopathic patients, aged 13 and 84 years, were discharged home without any disability. The other 5 patients were among the group of 8 patients who received a neurosurgical intervention. Among 58 patients with normal mentation and coagulation and minimal CT findings, we found poor outcomes or need for intervention in no cases (95% CI 0.0% to 6.2%).

The UCLA site enrolled 753 blunt head injury patients whose initial CT scans were interpreted as normal. We obtained 6-month telephone follow-up information for 568 (75.4%) patients, of whom 564 (99.3%; 95% CI 98.2% to 99.8%) were alive or had died more than a week after their injury and from a cause unrelated to head injury. Four patients died within the first week after injury: all were multiple trauma patients for whom a neurologic cause of death could not be definitively excluded (Table 3).

Table 3. Head trauma patients with normal initial CT scan results in whom an important neurosurgical outcome could not be ruled out.

	Patient No.	Age, y	Outcome
	Patients who died within 1 week of injury
	1	92	Died of cardiac arrest during surgery for open reduction and internal fixation of orthopedic injuries
	2	59	Cardiac arrest; died of a respiratory infection
	3	64	Fall; no change from neurologic baseline 3 days later; history of alcoholism, Korsakoff disease, Wernicke disease
	4	80	Died of respiratory distress as a result of cervical spine injury and paralysis, DNR
	Patients with a poor outcome >1 wk after injury
	5	93	Died months later, relative noted the fall in the nursing home was the beginning of downhill trend
	6	90	Had 24-h care before injury, GOS 3, able to do even less after the injury (no change in GOS score)
	7	86	Had 24-h care before injury, GOS 3, able to do even less after the injury (no change in GOS score)
	Patients with documented neurosurgical injuries
	8	84	Developed chronic SDH, remained at neurologic baseline (was severely demented); however, he bled into his injuries months later and died; neurosurgical intervention withheld as per DNR orders
	9	68	Developed chronic SDH, remained at neurologic baseline of GOS 5, no neurosurgery

Sufficient data to determine the Glasgow Outcome Scale score were available for 513 (68.1%) of the follow-up cases. Telephone interview confirmed functional outcomes in 505 cases (98.4%; 95% CI 97.0% to 99.3%), whereas 9 cases, including the 4 cases just discussed, either died or had a poor outcome (including 2 patients who were functioning at the same Glasgow Outcome Scale score before the date of injury; Table 3).

Telephone contact was made for 835 (59.2%) of 1,410 blunt head injury victims who did not undergo CT imaging at initial presentation. Of this group, 765 (91.6%) agreed to participate in the survey, whereas 70 (8.4%) patients refused. No patient reported a directed intervention, and family or surrogates denied any deaths as a result of undiagnosed head injury.

Raw interrater agreement for a designated therapeutically inconsequential CT finding was 98%, with a κ of 0.92. Interrater agreement of neurosurgical intervention was 100%. Interrater agreement of Glasgow Outcome Scale scores was 52 (95%) of 55 cases, with 3 cases requiring third-party review. Using only functional outcome as Glasgow Outcome Scale scores 5 and 4 and poor outcome as Glasgow Outcome Scale scores 1 to 3, there was no disagreement. The weighted κ statistic agreement for 52 of 55 was 0.82.

4. Limitations 

This study has 2 primary limitations: the small number of observed adverse outcomes among patients with inconsequential injuries and variable lengths of follow-up used to assess long-term outcomes. It is unlikely, given the relative rarity of therapeutically inconsequential CT findings, that the former problem can be overcome or that a larger study will ever be completed. As for the latter, although it is possible that a few patients deteriorated after discharge and did not return to the same hospital, most patients who deteriorate do so within the first few days of initial presentation.41 In addition, our use of early Glasgow Outcome Scale scores may well have overestimated long-term disability of patients with therapeutically inconsequential head CT findings (and relative to the patients with normal head CT results, who were evaluated at least 6 months after their injury) because Glasgow Outcome Scale scores frequently improve with extended observation.41 In contrast, the study is strengthened by the inclusion of a large and diverse population, prospective data collection, and rigorous adherence to medical record review rules, with high levels of interrater agreement.

5. Discussion 

Our results suggest that minor CT findings are rare among blunt head injury patients. Less than 2% of patients who were selected for CT scanning had therapeutically inconsequential findings. Among this group, however, a significant minority had an important neurosurgical outcome, with 10% requiring directed interventions related to the head trauma and an additional 2% having a poor outcome, as reflected by a low Glasgow Outcome Scale score.

Because poor outcomes can occur in the presence of a normal CT scan result, such as in diffuse axonal injury, a small group of poor outcomes among patients with therapeutically inconsequential CT scans is not unexpected. The key question is: Could these patients with an important neurosurgical outcome have been identified clinically? In our series, all 9 patients with important neurosurgical outcomes, for whom presenting information was available, had at least an abnormal mental status documented on presentation. Eight patients had a presenting Glasgow Coma Scale score of less than 15, whereas the ninth presented with a Glasgow Coma Scale score of 15 but an altered level of alertness as a result of alcohol consumption, and a coagulopathy. In addition, abnormal coagulation was documented for 5 of the 7 patients, for whom coagulation status was known, who underwent directed interventions, which is consistent with previous work43 and suggests that the patient's coagulation status may be more important than presence of therapeutically inconsequential CT findings in predicting the potential for adverse outcome. It is important to obtain a head CT scan for patients with these CT abnormalities but with the qualifier that it is necessary only in the presence of identifying abnormal clinical characteristics.

Just as a normal head CT scan result is compatible with severe brain injury, so also it is possible that there were other brain abnormalities, not visible on CT, causing the important neurosurgical outcomes. Although patients with a therapeutically inconsequential head CT finding were more likely to have an important neurosurgical outcome than were patients with normal imaging results, CT findings themselves do not seem to provide independent prognostic information. These injuries seem to indicate a higher risk for adverse outcomes, but this risk appears to apply only to patients who exhibit other clinical impairments such as baseline neurologic dysfunction, which identifies them as higher risk and therefore in need of a head CT scan.

A Glasgow Outcome Scale score of 4 may be a significant injury, and these patients may need to be identified for a head CT scan even if they receive no directed intervention. Although there is no doubt that a Glasgow Outcome Scale score of 4 is a significant injury, unfortunately current therapy for traumatic brain injury is limited, particularly in the acute setting. We chose to focus on patients for whom there would be a change in acute management. We thought that patients should be referred for neurologic rehabilitation according to their clinical disabilities, rather than the magnitude of CT findings, if they are not to be treated with a directed intervention.

The number of patients with minimal CT injuries and adverse outcomes is small, and we cannot completely exclude the possibility of an important neurosurgical outcome in a patient with an inconsequential CT abnormality, even in the presence of a normal level of alertness. However, given the size of our study, combined with the infrequency of therapeutically inconsequential head CT findings, it is unlikely that a larger study will ever be conducted to further clarify this risk. Thus, we may be forced to accept some degree of uncertainty about these injuries.

Our confidence limits indicate that therapeutically inconsequential CT findings are rare, appearing in less than 2.2% of patients undergoing CT imaging, or 0.55% of blunt head injury patients (assuming that approximately 25% of acute head injury patients undergo imaging),1., 2., 5. which corresponds to 1 therapeutically inconsequential injury for every 180 head injury victims. Although we found no unexpected poor outcomes among these patients, our upper confidence limit of 6.2% suggests that such outcomes could occur once in every 16 patients with therapeutically inconsequential injuries, or approximately once in every 2,900 head injury patients. To place this number in a clinical perspective, approximately 25,000 full-time equivalent emergency physicians in the United States evaluate about 1.4 million patients with blunt head injuries each year.1., 44. Thus, on average each emergency physician can expect to encounter at most one unexpected adverse outcome in a patient with a therapeutically inconsequential brain injury in every 50 years of practice. The true risk of injury is probably substantially lower than this figure, and the use of head injury precautions for discharged patients should limit the potential for adverse events even further.

Our study also suggests that poor outcomes are unusual among blunt trauma patients who have normal initial tomographic imaging results, as well as among patients who receive no imaging. None of the patients we contacted required a directed intervention within a week of injury. One patient, who remained unchanged from his neurologic baseline, could have had neurosurgical intervention for chronic subdural hematomas (his do-not-resuscitate protocol was followed instead), but no changes in the initial evaluation and treatment of this patient would have clearly averted this outcome. The few remaining patients who ultimately had a poor outcome were either unchanged from their pre-injury status or had other traumatic injuries that were considered more likely to explain their deterioration.

Therapeutically inconsequential CT abnormalities are uncommon among head injury patients selected for CT scanning. Although a significant number of patients with these findings go on to have an important neurosurgical outcome, this seems too uncommon in the absence of some independent clinical predictors of poor outcome, including baseline alteration in mental status and possibly altered coagulation status.