Prospective evaluation of criteria for obtaining thoracolumbar radiographs in trauma patients☆

Introduction 

Thoracolumbar spine injury is a potentially serious cause of significant morbidity and mortality in blunt trauma patients 1, 2, 3. Early identification of these injuries can likely decrease long-term morbidity. Although extensive study has been conducted to derive indications for imaging the cervical spine after blunt trauma, indications for obtaining thoracolumbar radiographs in blunt trauma patients have not been well studied and remain controversial 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15. This is despite evidence that thoracolumbar spine injuries are more common than cervical spine injuries after blunt trauma (16).

The purpose of this study was to prospectively evaluate the ability of defined clinical criteria to identify patients with thoracolumbar spine injuries. We hypothesized that all clinically important thoracolumbar spine injuries would be identified by certain high-risk criteria. Patients were considered at high-risk for thoracolumbar spine injury if any of the following criteria were present: complaints of thoracolumbar spine pain or thoracolumbar spine tenderness on examination, a decreased level of consciousness, a distracting painful injury, an abnormal neurologic examination, or evidence of intoxication with ethanol or drugs. These criteria were largely derived from the screening criteria advocated for evaluation of cervical spine injury as well as criteria previously suggested as important for the identification of thoracolumbar spine injuries 8, 9, 10, 11, 12, 13, 14, 15.

Materials and methods 

Protocol 

Each subject underwent a complete physical examination by the physician caring for the patient during the initial ED evaluation. Selected examination findings were documented on a structured data collection form and were completed or confirmed by the faculty emergency physician prior to obtaining radiographs. The presence or absence of each individual study criteria was recorded. If a criterion was not able to be determined, it was considered “unknown.”

All patients undergoing thoracolumbar radiography had both antero-posterior and lateral radiographs obtained. Oblique or cone downed radiographs were obtained at the discretion of the radiologist. Computed tomography (CT) scan or magnetic resonance imaging (MRI) of the spine was also obtained at the discretion of the treating physician.

Radiology Department logs were reviewed to identify any blunt trauma patients who underwent ED thoracolumbar imaging but were not enrolled in the study. Also, both ED continuous quality improvement community and trauma registry records were reviewed to determine if any patient was later found to have a thoracolumbar spine injury but was not enrolled in the study. In such cases, the chart was reviewed to determine the type of injury and the ED presentation of that patient.

Patients were assessed for all the clinical findings in Table 1. Patients with any one of these “high-risk” clinical criteria were considered in the “high-risk” group for thoracolumbar spine injury. In addition, patients were assessed for a severe mechanism of injury (with description of that injury), as a possible additional predictive criterion.

Table 1. High-risk Criteria

	Complaints of thoracolumbar spine pain
	Thoracolumbar spine tenderness on midline palpation
	Decreased level of consciousness
	Abnormal peripheral neurologic examination
	Distracting painful injury
	Evidence of intoxication with ethanol or drugs

Each of the screening criteria was prospectively defined to varying degrees and examples provided. Patients were considered to have a decreased level of consciousness if their Glasgow coma score (GCS) was less than 15, they were disoriented to person, place or events, or had an inappropriate or delayed response to external stimuli. An altered peripheral neurologic examination was considered present if any of the following were found on neurologic questioning or testing: weakness, numbness, or paresthesias. A distracting painful injury was considered present if any of the following were found: long bone injuries, crush injuries, burns, or any other injury resulting in pain that altered the patient’s perception. Patients with a GCS < 13 or undergoing emergent resuscitative procedures (intubation, thoracotomy, immediate transport to the operating room from the ED) were considered “unstable” and there was not a requirement to prospectively complete the entire study form. For these cases, the physicians completed the data sheet with as much information as possible. These unstable patients were considered high-risk for the purposes of data analysis. Generally, these patients would have at least one high-risk criterion (e.g., a decreased level of consciousness) present. A severe mechanism of injury was defined as any one of the following: a high speed (> 45 mph) or rollover motor vehicle collision, a motor vehicle collision requiring extrication, ejection from a motor vehicle, motorcycle accident greater than 20 miles per hour (mph), automobile versus pedestrian at > 5 mph, or a fall greater than 10 feet.

All radiographic studies were interpreted by a faculty radiologist (board certified or eligible). Diagnoses of thoracolumbar spine injury were based on the final interpretation of all radiographic studies. Only those patients with an acute injury were included in the group with injuries. Injuries that were chronic in nature were not included in the acute injury population. When injuries were of unclear age, prior radiographs were reviewed for comparison, when available.

Results 

A total of 2404 blunt trauma patients were enrolled. The mean age of patients enrolled was 36.9 ± 18.2 years. A total of 152 (6.3%, 95% CI 5.4–7.4%) patients were diagnosed with an acute thoracolumbar spine injury. Of the 152 patients with injury, all (100%, 95% CI 98–100%) had at least one of the high-risk criteria. The prevalence of each of the individual criteria in patients with and without injuries is shown in Table 2. No single high-risk criterion identified all patients with thoracolumbar injuries.

Table 2. High-risk Criteria for Thoracolumbar Injuries

		Injury Present (n = 152)	Injury Absent (n = 2,252)	Relative Risk	Sole§ Criterion	PPV	NPV
	Complaints of spinal pain	110 (72%, 65–79%)	1,463 (65%, 63–67%)	1.1 (1.0, 1.3)	5	7.0% (5.8–8.4%)	96% (94–97%)
	Spinal tenderness on palpation	108 (71%, 63–78%)	1,624 (72%, 70–74%)	1.0 (0.9, 1.1)	2	6.2% (5.1–7.5%)	94% (92–96%)
	Decreased level of consciousness	43 (28%, 21–36%)	502 (22%, 21–24%)	1.3 (1.0, 1.7)	9	7.9% (5.8–10.5%)	94% (93–95%)
	Abnormal neurologic examination	14 (9%, 5–15%)	213 (9%, 8–11%)	1.0 (0.6, 1.6)	2	6.2% (3.4–10%)	94% (93–95%)
	Distracting painful injury	63 (41%, 34–50%)	560 (25%, 23–27%)	1.7 (1.4, 2.0)	7	10% (7.9–13%)	95% (94–96%)
	Ethanol or drug intoxication	23 (15%, 10–22%)	372 (17%, 15–18%)	1.0 (0.6, 1.4)	0	5.8% (3.7–8.6%)	94% (93–95%)
	Cervical spine injury	7 (4.6%, 1.9–9.3%)	31 (1.4%, 0.9–1.9%)	3.3 (1.5, 7.5)	0	18% (7.7–34%)	93% (92–94%)
	Severe mechanism of injury	92 (61%, 52–68%)	821 (36%, 34–38%)	1.6 (1.4, 1.9)	0	10% (8.2–12.2%)	96% (95–97%)

Results are shown with 95% confidence intervals in parenthesis. Calculations of relative risk ratios do not include those patients in whom the finding was unknown. PPV = Positive Predictive Value; NPV = Negative Predictive Value.

§ Sole criterion indicates this finding was the single criterion present in a patient with thoracolumbar spine injury.

Of the 2252 patients without radiographic injuries, 2158 patients had at least one of the high-risk criteria. Only 88 patients in the study were without any of the high-risk criteria. The six high-risk criteria had a sensitivity of 100% (95% CI 98–100%), specificity of 3.9% (95% CI 3.1–4.8%), positive predictive value of 6.6% (95% CI 5.6–7.6%), and negative predictive value of 100% (95% CI 97–100%). The sensitivity and specificity for the six high-risk criteria as well as other combinations of the high-risk criteria are shown in Table 3.

Table 3. Sensitivity and Specificity for the High-risk Criteria* and Various Combinations of Criteria

		Sensitivity	Specificity
	High-risk criteria (any of the six)	100% (97.6–100%)	3.9% (3.1–4.8%)
	High-risk criteria and severe mechanism of injury	100% (97.6–100%)	2.0% (1.5–2.7%)
	High-risk criteria excluding intoxication	100% (97.6–100%)	4.6% (3.7–5.5%)
	High-risk criteria excluding intoxication and abnormal neurologic examination	98.7% (95.3–99.8%)	5.7% (4.8–6.8%)
	High-risk criteria excluding intoxication, abnormal neurologic examination, and distracting injury	92.8% (89.1–97.3%)	11.1% (9.8–12.4%)

* High-risk criteria include any of the following: complaints of thoracolumbar spine pain, thoracolumbar spine tenderness, decreased level of consciousness, abnormal neurologic examination, distracting painful injury, or evidence of intoxication with ethanol or drugs. %95% confidence intervals are provided in parentheses.

A complaint of thoracolumbar spine pain was the criterion most often present in injured patients. In addition, there were 52 (34%, 95% CI 27–42%) patients with thoracolumbar spine injury who had complaints of either thoracolumbar spine pain or had tenderness on palpation of the spine as the only positive high-risk criteria. Of these, five patients had only complaints of thoracolumbar spine pain and two patients had only thoracolumbar spine tenderness. The remaining 45 had both findings. The presence of spinal tenderness was unable to be evaluated in 146 (6%) patients and complaints of thoracolumbar spinal pain were unable to be evaluated in 127 (5%) patients. Only 12 (8%) patients with thoracolumbar spine injury were unable to be assessed for complaints of thoracolumbar spine pain. Only 13 (9%) patients with thoracolumbar spine injury were unable to be assessed for thoracolumbar spine tenderness.

In those patients with thoracolumbar spine injury, a distracting painful injury was the sole high-risk criteria in seven (5%, 95% CI 2–9%) patients, although two of these patients had multiple other criteria unable to be assessed due to the degree of pain from the distracting injury. The presence of a distracting painful injury was unable to be evaluated in 143 (6%) of patients. Only 10 (7%) patients with thoracolumbar spine injury were unable to be assessed for this criterion.

An abnormal peripheral neurologic examination was the sole high-risk criteria in only two (1%, 95% CI 0–4%) patients with thoracolumbar spine injury. An abnormal peripheral neurologic examination was unable to be evaluated in 129 (5%) patients. Only 12 (8%) patients with thoracolumbar spine injury were unable to be assessed for this criterion.

A decreased level of consciousness was the sole high-risk criterion in nine (6%, 95% CI 3–11%) patients with thoracolumbar spine injury. In these nine patients, the GCS was 13 or lower. The mental status was unable to be completely evaluated in 23 (1%) patients. Only 4 (3%) patients with thoracolumbar spine injury were unable to be assessed for this criterion.

There were no (0%, 95% CI 0–2%) patients with thoracolumbar spine injury in whom ethanol or drug intoxication was the sole high-risk criteria. There were, however, three patients who had a decreased level of consciousness from ethanol intoxication as the only clinical criterion of their thoracolumbar spine injury. Only 18 of the 2252 patients without thoracolumbar spine injury had intoxication as the single high-risk criterion for thoracolumbar imaging. Ethanol or drug intoxication was unable to be evaluated in 109 (5%) patients. Only 12 (8%) patients with thoracolumbar spine injury were unable to be assessed for this criterion.

A severe mechanism of injury was found present in 913 (38%) patients who underwent thoracolumbar spine imaging. Ninety-two (61%, 95% CI 52–68%) of the 152 patients with thoracolumbar spine injury had a severe mechanism of injury. Forty-one patients underwent thoracolumbar imaging solely for a severe mechanism of injury and none of these patients had spine injury. For that reason, the addition of a severe mechanism of injury to the high-risk criteria did not improve the sensitivity of the criteria but somewhat worsened the specificity (Table 3).

Three patients were found to have thoracolumbar spine injuries during their hospital evaluation but were not enrolled in the study. One patient was a 60-year-old woman involved in a motor vehicle accident. The patient had a GCS of 14 and multiple rib fractures on arrival in the ED, and was admitted to the intensive care unit for pain control. On hospital day #3, the patient complained of back pain with ambulation. A compression fracture of L2 was identified. This patient apparently was not fully evaluated for thoracolumbar spine injuries while in the ED. It would appear, however, that she would have met at least two of the high-risk criteria. The second patient was a 30-year-old man involved in a motorcycle accident. This patient had a GCS of 14 on arrival and was found to have abrasions across the back, as well as fracture or dislocations of the 2nd through 5th metatarsals. On hospital day #2, the patient underwent thoracolumbar spine imaging for “continued complaints of back pain” and was found to have an L3 superior end plate injury. This patient had at least one and probably two of the high-risk criteria if they were applied in the ED. The final patient identified with a thoracolumbar spine injury but not enrolled in the study was a 22-year-old man found to have a transverse process fracture of L4 on abdominal CT scan. This patient may not have had any high-risk criteria on ED presentation and the patient did not undergo thoracolumbar radiographs for evaluation of the lumbar spine. In addition, the patient received no treatment for this injury.

The NPV was high (94–96%) for each of the individual screening criteria. However, this primarily reflects the low prevalence of injury (6.3%) in the study population.

Discussion 

In this series, all the enrolled patients with thoracolumbar spine injury after blunt trauma were identified by the high-risk clinical criteria. Of the six high-risk criteria, all except intoxication were sometimes present as sole indicators for identifying the presence of thoracolumbar spine injury. These study findings also suggest that physicians who evaluate blunt trauma patients can assess whether each of the individual criteria is absent or present. The inability to assess an individual criterion was present in < 10% of patients with thoracolumbar spine injury and was due primarily to a decreased level of consciousness.

The majority of patients (75%) with thoracolumbar spine injury had complaints of thoracolumbar spine pain or tenderness to the thoracolumbar spine on palpation. Prior studies have suggested that 60–81% of patients with thoracolumbar spine injuries will have thoracolumbar spine pain or tenderness 12, 14. In this study, nearly a third of the patients with injury had one of these findings as the only clinical indicator of thoracolumbar spine injury. This emphasizes the importance of assessing these factors in blunt trauma patients.

A distracting painful injury was also very common in patients with thoracolumbar spine injury and the sole indicator in seven patients. Patients with major traumatic injuries have been previously shown to be at risk for thoracolumbar injury 11, 14. Several authors have advocated thoracolumbar spine imaging in patients with major traumatic injuries despite the absence of other specific indications of thoracolumbar injury 11, 14. Our results would support that approach of imaging the thoracolumbar spine if major distracting injuries are present.

One-fourth of the patients with a thoracolumbar spine injury had a decreased level of consciousness, and this finding was the sole indicator in nine patients. These nine patients all had a GCS < 13. This should not be surprising because patients with decreased levels of consciousness often have an unreliable physical examination due to their impaired mentation 8, 18, 19, 20. Any alteration in the mental status is believed to be an important indicator for cervical spine imaging 8, 9. The degree of depression that results in an unreliable physical examination of the thoracolumbar spine and warrants thoracolumbar imaging, however, is controversial. Several studies suggest trauma patients with a GCS < 8 or GCS < 10 should undergo thoracolumbar spine imaging, but patients with GCS scores higher should not 12, 15. One study of blunt trauma patients with a GCS from 13 to 15 demonstrated those with a GCS of 13 or 14 and thoracolumbar spine fractures less often had associated thoracolumbar spine pain or tenderness than those with thoracolumbar spine fractures and a GCS of 15 (11). The authors of that study recommended imaging the thoracolumbar spine of blunt trauma patients with a GCS below 15.

Only 9% of patients with thoracolumbar spine injury in our study had an abnormal neurologic examination, and it was the sole indicator in only two patients with injury. This is generally consistent with prior studies in which 6–26% of patients with thoracolumbar spine injury had an abnormal neurologic examination 12, 14. Dropping this element from the screening criteria would decrease the sensitivity of the screening criteria from 100% to 98.7%, but only raise the specificity by 1.1%.

No patient with thoracolumbar spine injury had ethanol or drug intoxication as the sole positive clinical screening criteria. This finding questions the utility of this specific criterion as a sole indication for obtaining thoracolumbar spine radiographs in blunt trauma patients. Although intoxication alone has been shown to be clinically important in detecting cervical spine injuries, it has not been specifically studied for the thoracolumbar spine 8, 9. For thoracolumbar evaluation, intoxication may be clinically important only when it results in a depressed level of consciousness (14). If intoxication were excluded from the study as a high-risk criterion (see Table 3), the sensitivity of the criteria would not change, but the specificity would improve slightly.

Sixty-one percent of patients with thoracolumbar spine injury had a severe mechanism of injury. No patient with thoracolumbar spine injury, however, had a severe mechanism of injury as the sole positive predictive criteria. If we added severe mechanism of injury (see Table 3) to the high-risk screening criteria, the sensitivity of the criteria would not improve but the specificity would decrease by half. A future study attempting to derive criteria that maintained sensitivity but increased specificity might consider including mechanisms of injury with a high-risk association of thoracolumbar spine injury. Falls ≥ 10 feet, ejection from a motor vehicle, and motor vehicle crashes ≥ 50 mph are mechanisms of injury previously shown to increase the risk of thoracolumbar spine injury (12). One recent study has suggested that mechanism criteria could aid in the detection of cervical spine injury (21).

The presence of a cervical spine injury had a good positive predictive value in this study. Non-contiguous, multi-level vertebral fractures have been documented to occur in 4.5–16.7% of patients with vertebral fractures 16, 22, 23. Although not included in the specific criteria evaluated in this study, we would include the presence of a cervical spine injury as an indication for thoracolumbar spine imaging.

Three patients with thoracolumbar spine injuries were not enrolled in the study as they did not have imaging studies ordered in the ED. This 1.9% miss rate of stable patients with thoracolumbar spine injury is similar to a previous report of 1.7% (24). In the present study, addition of these “missed” or delayed diagnosis patients, however, does not appear to change the high sensitivity of the screening criteria. Two of these patients appear to have had high-risk criteria present on ED evaluation and should have undergone thoracolumbar spine imaging during ED evaluation. The third patient with thoracolumbar spine injury not identified in the study likely did not have any high-risk criteria present and was found to have a transverse process fracture on abdominal CT scan. This patient’s injury was not clinically important and did not require therapy. If this patient had been enrolled into the study, the sensitivity of the criteria for identifying injury would only drop to 99.3% (95% CI 96.4–100%).

Surprisingly, only 88 (4%) patients undergoing thoracolumbar spine radiography in this study were without any of the high-risk criteria. Such a low percentage was not expected when the study was begun based on general clinical experience, as it was common practice to liberally obtain thoracolumbar spine radiographs in patients arriving in spinal precautions in the ED. It is possible that either the simple process of instructing the physician staff or the process of requiring completion of the study data form before obtaining the radiographs had an effect on physician ordering behavior. Perhaps patients who had none of the criteria on the data collection sheet were less likely to undergo radiographic studies because of the information on that form. The investigators did not study that possible effect. Regardless, it appears that at the study institution, these criteria have become generally accepted and applied.

If patients without the study criteria receive definitive diagnostic studies (thoracolumbar radiography) less often than those with the criteria, there is an increased likelihood that injuries will be missed in the group without the criteria because of the less extensive evaluation (verification bias) (25). The investigators took multiple steps to detect problems with verification bias. Both CQI and trauma center records were reviewed to identify any thoracolumbar spine injuries that were missed by the study. This would detect patients with injury who were missed during the ED evaluation and received the diagnosis during a later evaluation. Only three such patients were identified. Because not every trauma patient underwent thoracolumbar radiography, it is possible that some patients with fractures were never diagnosed. It is likely that such patients would not have had any of the clinical screening criteria, because those findings would likely have led to the ordering of radiographs. Therefore, the existence of such patients would decrease the sensitivity of the criteria. The clinical practice at the study site is for all patients discharged from the ED or inpatient trauma service to be able to ambulate and function without strong analgesics. Inability to do so would have triggered additional evaluation and diagnosis. Therefore, it is most likely that all injuries of clinical significance would have been identified prior to discharge. Nonetheless, it is possible that patients with truly occult fractures were missed. The importance of truly occult injuries that do not require therapy is unclear.

Another concern when there is enrollment of so few patients without the screening criteria is the ability to draw inferences about the sensitivity and the specificity of the criteria. Unless an appropriately broad spectrum is chosen for both the “diseased” and “non-diseased” study patients, the calculated performance of the screening test could be falsely elevated (spectrum bias) (26). This study was conducted at a single institution, which serves as a regional trauma center, and may not be applicable for all hospitals. Although Level 1 trauma centers routinely provide care for severely injured patients, the prevalence of thoracolumbar spine injury was low. Therefore, it appears this study population was a fairly low-risk group overall, and possibly clinically representative.

Both the positive predictive value and specificity of these criteria were disappointingly low. Although these criteria identified all patients with thoracolumbar injury, future study should elucidate criteria that maintain this sensitivity but have improved specificity. Development of clinical guidelines would potentially allow a more meaningful reduction in thoracolumbar spine imaging. Perhaps criteria should be developed such that they identify “clinically important” thoracolumbar spine injuries.

In conclusion, the following high-risk screening criteria identified all patients with acute thoracolumbar spine injury after blunt trauma: complaints of thoracolumbar spine pain, tenderness on thoracolumbar spine palpation, a decreased level of consciousness, a distracting painful injury, or an abnormal neurologic examination. The positive predictive value and specificity of these screening criteria, however, are low. Intoxication with ethanol or drugs was not useful as an independent criterion for identifying injured patients unless it resulted in a decreased level of consciousness. Both a severe mechanism of injury and the presence of a cervical spine fracture are significantly associated with an increased risk of thoracolumbar spine injury, although neither added to the sensitivity of the clinical screening criteria.