Utility of clinical decision rule for intensive care unit admission in patients with traumatic intracranial hemorrhage

Accepted Manuscript Utility of clinical decision rule for intensive care unit admission in patients with traumatic intracranial hemorrhage Brandt D. Whitehurst, M.D., Jared Reyes, M.Ed., Stephen D. Helmer, Ph.D., James M. Haan, M.D., FACS PII:

S0002-9610(16)30676-6

DOI:

10.1016/j.amjsurg.2016.09.057

Reference:

AJS 12131

To appear in:

The American Journal of Surgery

Received Date: 19 May 2016 Accepted Date: 30 September 2016

Please cite this article as: Whitehurst BD, Reyes J, Helmer SD, Haan JM, Utility of clinical decision rule for intensive care unit admission in patients with traumatic intracranial hemorrhage, The American Journal of Surgery (2016), doi: 10.1016/j.amjsurg.2016.09.057. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Utility of clinical decision rule for intensive care unit admission in patients with traumatic intracranial hemorrhage

RI PT

Short Running Title: Clinical Decision Rule for ICU Admission in tICH

Brandt D. Whitehurst, M.D.a, Jared Reyes, M.Ed. a, Stephen D. Helmer, Ph.D.a,b, James M.

a

SC

Haan, M.D.a,c,*

Department of Surgery, The University of Kansas School of Medicine – Wichita, 929 N. Saint

c

M AN U

Francis St., Room 3082, Wichita, KS 67214, USA; Departments of bMedical Education and Trauma Services, Via Christi Hospital Saint Francis, 929 N. Saint Francis St., Room 3082,

Wichita, KS 67214, USA

AC C

EP

April 3 – 6, 2016

TE D

Presented at the 68th Annual Meeting of the Southwestern Surgical Congress, Coronado, CA,

Corresponding Author: James M. Haan, M.D., FACS, Department of Trauma Services, Room 2514, Via Christi Hospital St. Francis, 929 N. Saint Francis St., Wichita, KS 67214 Phone: 316-268-5538, FAX: 316-291-7892, E-mail: [email protected] 1

ACCEPTED MANUSCRIPT

Abstract BACKGROUND: Recent literature suggests the majority of traumatic intracranial hemorrhage does not require intervention. One recently described clinical decision rule was

RI PT

sensitive in identifying patients requiring critical care interventions in an urban setting. We sought to validate its effectiveness in our predominately rural setting.

METHODS: A retrospective study was conducted of adult patients with traumatic intracranial

SC

hemorrhage. The rule, based on age, initial Glasgow coma scale score, and presence of a nonisolated head injury, was applied to externally validate the previously reported findings.

M AN U

RESULTS: In our population, the rule displayed a sensitivity of 0.923, specificity of 0.251, positive predictive value of 0.393, and negative predictive value of 0.862. The area under curve was 0.587. While our population has a similar adjusted head injury severity score as that from which the rule was developed, significant differences in age and intracranial hemorrhage pattern

TE D

were noted.

CONCLUSIONS: The rule displayed decreased performance in our population, most likely secondary to differences in age and intracranial hemorrhage patterns. Prospective evaluation and

AC C

EP

cost-savings analysis are appropriate subsequent steps for the rule.

2

ACCEPTED MANUSCRIPT

KEYWORDS: Trauma; Traumatic intracranial hemorrhage;

RI PT

Clinical decision rule; Triage;

AC C

EP

TE D

M AN U

SC

Intensive care unit admission

3

ACCEPTED MANUSCRIPT

Short Summary Intracranial hemorrhage is a frequent finding on computed tomography of the head in trauma populations. Routine intensive care unit (ICU) admission has been the standard-of-care

RI PT

for such patients. A clinical decision rule for determining appropriateness of ICU admission for intracranial hemorrhage patients has recently been described. We examined a retrospective cohort of traumatic intracranial hemorrhage patients from our trauma center and found it had

SC

high sensitivity, but degraded specificity in appropriately selecting patients for ICU admission. Specific contributing factors to this loss of performance were advanced age, mechanism of

AC C

EP

TE D

M AN U

injury, prevalence of subdural hematomas, and delayed presentation.

4

ACCEPTED MANUSCRIPT

Background Traumatic brain injury (TBI) is a frequent occurrence in the adult trauma population with an estimated incidence of 1,700,000 cases per year with 275,000 requiring hospitalization.1

RI PT

Traumatic intracranial hemorrhage (tICH) is a frequent finding on imaging for TBI patients that require hospitalization. Clinical observation, often in an intensive care unit (ICU), is the

standard-of-care in many trauma centers for patients with tICH. Intensive care unit observation

SC

affords timely identification of symptoms associated with hemorrhage progression and secondary brain injury.2-5 Due to rising health care expenditures, appropriate utilization of high cost

M AN U

resources, such as ICU monitoring, has received great interest.6 Because published data suggest that rates of clinical progression and neurosurgical intervention for tICH is low,7,8 the costeffectiveness of routine ICU admission is now questioned.9-12

A previously published clinical decision rule, derived from a retrospective review of 432

TE D

tICH patients in an urban Level 1 trauma center, was designed to identify patients at high risk for requiring a critical care or neurosurgical intervention, and thereby needing ICU admission. The decision rule utilizes three clinical parameters: Glasgow coma scale (GCS) score <15, presence

EP

of non-isolated head injury, and age greater than 65.4 While simple and rapid to utilize, the performance characteristics of this clinical decision rule have not yet been externally validated.

AC C

While interested in this rule as an aid for assigning appropriate disposition, we suspected population characteristics, such as our significant rural catchment area, would impact its performance. Therefore, we sought to examine the performance characteristics of this clinical decision rule in our population.

5

ACCEPTED MANUSCRIPT

Methods Utilizing the International Classification of Diseases, Ninth Revision (ICD-9) codes for traumatic intracranial hemorrhage (851-854) to query our trauma registry, we identified a cohort

Level I trauma center between January and December 2013.

RI PT

of all patients, 18 years of age or greater, treated at our American College of Surgeons verified

Database capture from our trauma registry included patient demographics (age and

SC

gender) and markers of injury and intervention such as: mechanism of injury, injury severity score (ISS), GCS score, ethanol use and level, initial vital signs, ICU length of stay, need for

M AN U

mechanical ventilation and ventilator days, hospital length of stay, disposition, and mortality. Retrospective chart reviews were subsequently conducted to identify clinical, physical, and objective parameters for analysis as well as the occurrence of critical care interventions. Clinical parameters recorded included loss of consciousness, nausea, emesis, headaches,

TE D

focal neurologic deficits, amnesia and seizures. Physical parameters recorded included the presence of focal neurologic deficits, non-frontal scalp injury including hematoma and lacerations, basilar skull fracture signs such as raccoon eyes, Battle’s sign, and hemotympanium.

EP

Objective parameters recorded included time from injury to presentation to the hospital, use of anticoagulants, presence of isolated head injury, and initial GCS score. Because hemorrhage size

AC C

and volume are not utilized in applying the clinical decision rule, imaging reports were utilized to determine hemorrhage type in our cohort. Critical care interventions included arterial catheterization, central line placement,

intracranial pressure monitoring, mechanical ventilation, use of vasopressor, antiarrhythmic or anti-hypertensive drips, transfusion of blood products, or performance of advanced cardiac life support protocols. Therapeutic and surgical procedures such as interventional radiology

6

ACCEPTED MANUSCRIPT

procedures, craniotomy, skull fracture elevation, or Burr hole placement were considered critical care interventions for purposes of this study. Data were abstracted, summarized, and reported as mean with 95% confidence interval or

RI PT

median with interquartile range as appropriate for the respective data type. The data were

examined for statistical relationships using SPSS Software, Version 19 (IBM Corp., Somers, New York). Univariate analyses were conducted using chi-square tests. Comparable to the

SC

methods employed by Nishijima et al.,4 a binary decision tree was used to validate the clinical decision rule. A receiver operating characteristic curve was plotted with the area under the curve

M AN U

(AUC) used to evaluate the performance of the decision rule for distinguishing between true positives (ICU admissions requiring critical intervention) and false positives (ICU admissions not requiring critical intervention). This study was approved for implementation by the

Results

TE D

institutional review board of Via Christi Hospitals Wichita, Inc.

A total of 355 adult patients with traumatic intracranial hemorrhage were identified as

EP

meeting the initial study criteria. During the review process, 14 patients were excluded from further analysis. Nine patients did not have imaging or imaging reports demonstrating a tICH

AC C

despite the sometimes-associated presence of findings such as skull fractures. Two registry entries did not correlate with an electronic medical record. One patient was not evaluated by the trauma team. One patient without a trauma mechanism was transferred to another service after a characteristic hypertensive bleed was identified. The final patient presented with pulseless electrical activity secondary to asphyxiation from hanging and did not undergo intracranial imaging. The remaining 341 patients were included in the final analyses.

7

ACCEPTED MANUSCRIPT

Examination of the patient population meeting inclusion criteria revealed that 63.3% were male and the majority were over the age of 65 with a mean age of 61.6 years (Table 1). Patients were on average moderately brain-injured as evidenced by a median GCS score at

RI PT

presentation of 15 and median ISS of 10. Alcohol was detectable in 16.4% of patients. Few patients presented in shock (3.5%). Anticoagulant therapy was common and noted in 39.2% of patients. A significant proportion (18.6%) of patients presented more than twenty-four hours

SC

after the inciting event.

Falls were the most common injury mechanism (59.2%), with motor vehicle collisions

M AN U

representing the second most common mechanism (21.4%; Table 2). Isolated head injury was present in the majority of patients (71.8%). Multiple hemorrhages were present in 41.1% of patients with subdural hematoma being the most frequent type of hemorrhage, whether isolated or not (Table 3). Mechanical ventilation was the most frequently performed critical care

TE D

intervention (22.9%) followed by central or arterial line placement (13.5%; Table 4). Neurosurgical intervention was necessary in 15.9% of patients. Average hospital length of stay was 5.1 ± 5.9 days (median = 3 days) and we observed a 10.3% mortality rate in our population.

EP

Validation of the clinical decision rule used the occurrence of a critical care intervention as the gold standard for indicating requisite ICU admission (Table 5). In our population, the rule

AC C

displayed a sensitivity of .92 (95% CI .89 - .95), specificity of .25 (95% CI .21- .30), positive predictive value of .39, and negative predictive value of .86. The AUC was .587 (95% CI .53 .65; P = .008). In evaluating those patients with delayed presentation (>24 hours), the rule suffered from diminished sensitivity, and loss of statistical significance. Numerous presenting parameters were associated with an increased likelihood of undergoing a critical care intervention such as: GCS less than 9 (P < .001), increasing ISS (P<.001), presence of multiple tICH (P = .011), systolic blood pressure less than 90 mmHg (P < 8

ACCEPTED MANUSCRIPT

.001), and presence of alcohol (P < .001; Table 6). In the delayed presentation sub-population, there was no statically significant difference in presenting GCS, ISS, presence of multiple tICH, SBP<90, or mortality.

RI PT

Using the decision rule, eight patients were inappropriately identified as low-risk for requiring a critical care intervention. Of the eight, six underwent neurosurgical interventions consisting of two craniotomies, two Burr holes, and two skull fracture elevations. Of this group,

SC

4 of the 6 presented more than 48 hours after the inciting event; the remaining two patients had focal depressed skull fractures requiring elevation. Among the non-neurosurgical group, one

M AN U

patient required central venous line placement following a myocardial infarction, the other patient was taken to the operating room for facial fracture repair and returned to the ICU intubated until a 12-hour follow-up head CT was obtained.

TE D

Comments

In the modern healthcare setting, efficient resource utilization has become an important consideration due ever rising costs and limited resources. Traditionally, close ICU observation

EP

was endorsed to aid in early identification and intervention of secondary brain injury.2 As data have accumulated demonstrating that a significant proportion of patients with tICH do not

AC C

require additional interventions,8 the traditional mantra of ICU observation, neurosurgical consultation,12 and repeat head imaging has been challenged.10 Given this situation, interest has grown in predictive models that identify presenting patient characteristics associated with need for ICU admission.6 One such clinical decision rule is that of Nishijima et al.4 We were interested in the utility of this rule both for its ease of application, requiring only three readily obtained clinical parameters, and for its design to minimize the number of patients incorrectly identified as low-risk for critical care interventions. We found these compatible with our 9

ACCEPTED MANUSCRIPT

primary goals of simplicity and safety in utilizing any clinical decision rule. However, we were concerned that the rule was developed retrospectively in a patient population likely distinct from our own; therefore, we sought to retrospectively validate its performance characteristics in our

RI PT

own patient population prior to implementation.

In its original population, the clinical decision rule had an excellent reported sensitivity of 98% (95% CI 94 – 99), with a specificity of 50% (95% CI 44 – 56), and an AUC of 0.74 (95%

SC

CI .70 – .77). While the sensitivity remained relatively high in our population , both the

specificity and AUC were much lower than expected. In addition to the slight decrease in

M AN U

performance that is expected when applying such a model to a sample different from the training sample, the diminished performance of the clinical decision rule is likely the result of clear distinctions between the original population and our population.

Between the two populations, the frequency of isolated head injury and abbreviated head

TE D

injury score was similar at approximately 74 vs 72% and 4 vs 3, respectively. However, our population had a decreased overall severity of injury, which may have impacted rule performance. This suggests differences in rule performance may be secondary to increased high-

EP

energy mechanisms and subsequent poly trauma in the original population and not the actual severity of tICH. In our population, subdural hematomas were far more prevalent (50.4 vs 37%);

AC C

while both intraparenchymal hemorrhages and subarachnoid hemorrhages were comparatively lower in our population. Because invasive monitoring and intracranial pressure monitoring has a more significant role in the treatment of tICH causing a mass effect, such as with subdural hematomas as opposed to intraparenchymal hemorrhage or subarachnoid hemorrhage, this may account for our increased rate of neurosurgical intervention. With regard to mechanism of injury, falls were present in approximately 60% of our patients as compared to approximately 30% in the original population. Conversely, motor 10

ACCEPTED MANUSCRIPT

vehicle collisions represented a much larger proportion of the original population as compared to ours (approximately 40% vs. 20%, respectively). Our trauma catchment includes substantial

predominance of other injury mechanisms in our population.

RI PT

rural geographic areas, which have fewer vehicles and less dense traffic, likely accounting for the

Because age is an integral parameter of the rule, the significant age disparity between the two populations (original 48 [IQR 30 – 63] vs ours 67 [IQR 45 – 82]) is perhaps the most

SC

significant factor in accounting for the apparent degraded performance of the rule. Despite the similarity in head injury severity, there was a disproportionately larger number of elderly patients

M AN U

in our population. We suspect low-energy mechanisms, such as a fall from standing, are more prevalent in elderly patients with tICH, and may explain the finding of similar head injury severity with decreased overall injury severity in our population.

Due to the risk of harm in delayed recognition of a need for critical care interventions, the

TE D

presence of false negatives and examining factors related to their occurrence is critical for safe implementation. In each instance, an elderly patient with a remote fall, isolated head bleed, and normal GCS presented due to mild symptoms such as headache, nausea, or mild gait disturbance.

EP

Such patients may delay presentation because they are frequently cognitively appropriate (i.e. GCS = 15) and may not have resources or access to seek immediate intervention. We did not

AC C

find any false negatives wherein a patient suffered a significant cognitive decline leading to neurosurgical intervention.

Our study has the limitations incumbent with a retrospective chart review: occasional

inaccuracy or omission of recorded data and unrecognized errors or oversights of the chart reviewer. For example, loss of consciousness was almost universally noted on admission, but neurological symptoms such as amnesia, nausea, or emesis were only noted when in the

11

ACCEPTED MANUSCRIPT

affirmative. A prospective validation of the Nishijima et al.4 decision rule in a hospital with an urban catchment is an appropriate next step. Despite its decreased performance in our population, the decision rule may still have

RI PT

value in reducing unnecessary ICU admissions because of its simplicity. Care must be taken

AC C

EP

TE D

M AN U

SC

when applying this rule to patients with delayed presentations and skull fractures.

12

ACCEPTED MANUSCRIPT

References 1. Faul M, Xu L, Wald MM, Coronado VG. Traumatic Brain Injury in the United States: Emergency Department Visits, Hospitalizations and Deaths 2002–2006. Atlanta (GA):

RI PT

Centers for Disease Control and Prevention, national Center for Injury Prevention and

Control; 2010. Available at: http://www.cdc.gov/traumaticbraininjury/pdf/blue_book.pdf. Accessed April 15, 2016.

SC

2. Graham DI, Ford I, Adams JH, et al. Ischaemic brain damage is still common in fatal nonmissile head injury. J Neurol Neurosurg Psychiatry, 1989;52,346-350.

M AN U

3. Vos PE, Battistin L, Birbamer G, et al. EFNS guideline on mild traumatic brain injury: report of an EFNS task force. Eur J Neurol, 2002;9,207-219.

4. Nishijima DK, Shahlaie K, Echeverri A, Holmes JF. A clinical decision rule to predict adult patients with traumatic intracranial haemorrhage who do not require intensive care unit

TE D

admission. Injury, 2012;43,1827-1832.

5. Nishijima DK, Haukoos S, Newgard CD, et al. Variability of ICU use in adult patients with minor traumatic intracranial hemorrhage. Ann Emerg Med, 2013;61,509-517.

EP

6. Hukkelhoven CW, Steyerberg EW, Habbema JD, Maas AI. Admission of patients with severe and moderate traumatic brain injury to specialized ICU facilities: a search for triage

AC C

criteria. Intensive Care Med, 2005;31,799-806. 7. Homnick A, Sifri Z, Yonclas P, et al. The temporal course of intracranial haemorrhage progression: how long is observation necessary? Injury, 2012;43,2122-2125. 8. Carlson AP, Ramirez P, Kennedy G, et al. Low rate of delayed deterioration requiring surgical treatment in patients transferred to a tertiary care center for mild traumatic brain injury. Neurosurg Focus, 2010;29,E3.

13

ACCEPTED MANUSCRIPT

9. Sifri ZC, Livingston DH, Lavery RF, et al. Value of repeat cranial computed axial tomography scanning in patients with minimal head injury. Am J Surg, 2004;187,338-342. 10. Bee TK, Magnotti L, Croce MA, et al. Necessity of repeat head CT and ICU monitoring in

RI PT

patients with minimal brain injury. J Trauma, 2009;66,1015-1018.

11. Nishijima DK, Sena M, Holmes JF. Identification of low-risk patients with traumatic brain

2011;70,E101-107.

SC

injury and intracranial hemorrhage who do not need intensive care unit admission. J Trauma,

12. Huynh T, Jacobs D, Dix S, et al. Utility of neurosurgical consultation for mild traumatic

AC C

EP

TE D

M AN U

brain injury. Am Surg, 2006;72,1162-1165.

14

ACCEPTED MANUSCRIPT

Table 1 Patient population and clinical characteristics N

Value

Number of observations

341

100%

Age (years)

341

61.6 ± 22.6

Age greater than 65

178 216

Initial GCS score

340

SC

Male gender

RI PT

Parameter

63.3%

15 (13, 15)

259

76.2%

21

6.2%

60

17.6%

341

10 (9, 18)

341

3 (3, 4)

56

16.4%

56

199.7 ± 110.9

341

144.3 ± 30.9

Diastolic blood pressure (mmHg)

340

88.6 ± 19.4

Heart rate

341

86.4 ± 19.3

Respiratory rate

341

17.0 ± 7.5

Temperature (ºF)

331

97.8 ± 1.0

Oxygen saturation (%)

331

97.3 ± 3.0

12

3.5%

132/337

39.2%%

338

1.4 ± 5.2

GCS 9 - 12 GCS 3 – 8 Injury severity score AIS head

TE D

Positive blood alcohol level

M AN U

GCS 13 - 15

52.2%

Blood alcohol concentration Initial vital signs

AC C

EP

Systolic blood pressure (mmHg)

Systolic blood pressure <90 mmHg Anticoagulant therapy Interval from injury to presentation (days)

15

ACCEPTED MANUSCRIPT

Delayed presentation (≥ 1 day)

63

18.6%

AC C

EP

TE D

M AN U

SC

RI PT

Data are presented as percent, mean ± SD, or median (interquartile range).

16

ACCEPTED MANUSCRIPT

Table 2 Mechanisms of injury and head injury characteristics N

Percent

Falls

202

59.2%

Motor vehicle collision: traffic

73

21.4%

Assault, fight, legal intervention by other

31

Accident (explosion, cutting, machinery, 22

SC

animal, striking)

Suicide/self-inflicted injury Isolated head injury Loss of consciousness

6.5%

3.2%

1

.3%

1

.3%

245

71.8%

169/264

64.0%

97/337

28.8%

12/268

4.5%

AC C

EP

Focal neurological deficit

TE D

Non-frontal scalp hematoma

9.1%

11

M AN U

Motor vehicle collision: non-traffic Child abuse

RI PT

Mechanism of injury

17

ACCEPTED MANUSCRIPT

Table 3 Radiographic parameters of observed traumatic intracranial hemorrhage N

Percent

Number of observations

341

100.0%

Multiple tICH

140

41.1%

Type of bleed 237

Subarachnoid hemorrhage

123

SC

Subdural hematoma

Intraparenchymal hemorrhage

Diffuse axonal injury Skull fracture Midline shift

Herniation

TE D

Midline shift size (mm)

69.5%

36.1%

101

29.6%

8

2.3%

3

0.9%

51

15.0%

72/340

21.2%

72

7.7 ± 5.6*

19/340

5.6%

M AN U

Epidural hematoma

RI PT

Parameter

AC C

EP

tICH = traumatic intracranial hemorrhage; *Mean ± SD

18

ACCEPTED MANUSCRIPT

Table 4 Frequency of critical care and neurosurgical interventions N

Percent

Mechanical ventilation

78

22.9%

78

4.4 ± 5.4*

Mechanical ventilator days

46

Vasoactive drip

25

Anti-arrhythmic drip

17

SC

Central venous or arterial line placed

RI PT

Parameter

Blood products administered

13.5% 7.3%

5.0%

16

4.7%

3

0.9%

2

0.6%

54

15.8%

29

8.5%

Intracranial pressure monitoring

25

7.3%

Burr hole

10

2.9%

4

1.2%

M AN U

Cardiac arrest/ACLS Interventional radiology

Neurosurgical Intervention

AC C

*Mean ± SD

EP

Skull fracture elevation

TE D

Craniotomy

19

ACCEPTED MANUSCRIPT

Table 5 Statistical performance of clinical decision rule Specificity

PPV

NPV

Accuracy

AUC

P-value

Overall

.92 (.89-.95)

.25 (.21-.30)

.39

.86

.48

.59

.008

>24 hours

.81 (.77-.86)

.42 (.36-.47)

.51

.75

.59

.62

.118

AC C

EP

TE D

M AN U

SC

RI PT

Sensitivity

20

ACCEPTED MANUSCRIPT

Table 6 Comparative frequency of clinical parameters in patients with critical care interventions Critical Care

(% Positive)

P value

Glasgow Coma Scale score <15

62.1%

<.001

Glasgow Coma Scale score =15

16.8%

Glasgow Coma Scale score <9

95.0%

Glasgow Coma Scale score ≥9

21.4%

M AN U

Injury severity score 1-15

SC

Criteria

RI PT

Intervention

21.5%

Injury severity score 16-24

78.0%

42.1%

TE D

Mutliple tICH

Systolic blood pressure ≥90

32.2%

Alive

AC C

Died

EP

91.7%

No alcohol present

.011

28.9%

Systolic blood pressure <90

Alcohol present

<.001

35.6%

Injury severity score 25+

Single tICH

<.001

55.4%

<.001

<.001

30.2% 83.8%

<.001

28.3%

Delayed presentation

42.9%

Not delayed

32.4%

0114

tICH = traumatic intracranial hemorrhage

21