Intraobserver and interobserver agreement in visual inspection for xanthochromia: implications for subarachnoid hemorrhage diagnosis, computed tomography validation studies, and the Walton rule

Marshman, Laurence A.G., Duell, Ryan, Rudd, Donna, Johnston, Ross, and Faris, Casandra (2014) Intraobserver and interobserver agreement in visual inspection for xanthochromia: implications for subarachnoid hemorrhage diagnosis, computed tomography validation studies, and the Walton rule. Neurosurgery, 74 (4). pp. 395-400.

[img] PDF (Published Version) - Published Version
Restricted to Repository staff only

View at Publisher Website: http://dx.doi.org/10.1227/NEU.0000000000...
 
4
8


Abstract

Background: Visual inspection for xanthochromia is used to diagnose subarachnoid hemorrhage (SAH), to validate computed tomography subarachnoid hemorrhage diagnosis and was used to determine the Walton rule. No study has assessed the reliability of xanthochromia.

Objective: To determine intraobserver and interobserver xanthochromia agreement.

Methods: Mock cerebrospinal fluid samples contained increasing concentrations of human oxyhemoglobin, bilirubin, and albumin. Non-color-blind observers randomly assessed samples against a white background twice under significantly differing illumination. Specimens were recorded as red, orange, yellow, or clear.

Results: Results were obtained for 26 observers (11 male, 15 female observers). We found that 19.2% of samples were misclassified: red, 11.7%; orange, 28.5%; yellow, 29.6%; and clear, 22.1% (χ = 213.2; P < .001). Of the yellow misclassifications, 88% were misclassified as clear. Female observers correctly classified samples significantly more frequently than male observers (P = .03). Intraobserver agreement differed significantly from expected for both male (χ = 105.6; P < .001) and female (χ = 99.9; P < .001) observers regardless of illumination. Interobserver agreement was poor regardless of sex (χ for male observers = 176.96, P < .001; χ for female observers = 182.69, P < .001) or illumination (χ for bright = 125.64, P < .001; χ for dark = 148.48, P < .001). Overall, there was 75% agreement in 46% of the tests and 90% agreement in only 36% of the tests.

Conclusion: This simple laboratory study would be expected to maximize agreement relative to clinical practice. Although non-color-blind female observers significantly outperformed non-color-blind male observers, both intraobserver agreement and interobserver agreement for xanthochromia were prohibitively poor regardless of sex or illumination. Yellow was most frequently misclassified, 88% as clear (ie, true positives were commuted to false negatives). Xanthochromia is therefore highly unreliable for subarachnoid hemorrhage diagnosis and computed tomography validation. The Walton rule requires urgent clinical revalidation.

Item ID: 36852
Item Type: Article (Research - C1)
ISSN: 1524-4040
Keywords: observer agreement, xanthocromia
Related URLs:
Additional Information:

One comment on this article was published with reply from the original authors. The links to the comment can be found in the Related URLs field:

Sabour, Siamak (June 2014) Intraobserver and interobserver agreement in visual inspection for xanthochromia: implications for subarachnoid hemorrhage diagnosis, computed tomography validation studies and the Walton rule: methodological mistake (with reply by Marshman et al). Neurosurgery, 74(6): pp. E702-703.

Date Deposited: 27 Feb 2015 02:48
FoR Codes: 11 MEDICAL AND HEALTH SCIENCES > 1109 Neurosciences > 110999 Neurosciences not elsewhere classified @ 45%
11 MEDICAL AND HEALTH SCIENCES > 1103 Clinical Sciences > 110323 Surgery @ 45%
11 MEDICAL AND HEALTH SCIENCES > 1103 Clinical Sciences > 110316 Pathology (excl Oral Pathology) @ 10%
SEO Codes: 92 HEALTH > 9201 Clinical Health (Organs, Diseases and Abnormal Conditions) > 920118 Surgical Methods and Procedures @ 70%
92 HEALTH > 9201 Clinical Health (Organs, Diseases and Abnormal Conditions) > 920199 Clinical Health (Organs, Diseases and Abnormal Conditions) not elsewhere classified @ 30%
Downloads: Total: 8
More Statistics

Actions (Repository Staff Only)

Item Control Page Item Control Page