Pupil examination: validity and clinical utility of an automated pupillometer.
(Tests, problems and exercises)
Periodic health examinations (Equipment and supplies)
Physical diagnosis (Equipment and supplies)
Privitera, Claudio M.
Larson, Merlin D.
Holland, Martin C.
|Publication:||Name: Journal of Neuroscience Nursing Publisher: American Association of Neuroscience Nurses Audience: Professional Format: Magazine/Journal Subject: Health care industry Copyright: COPYRIGHT 2005 American Association of Neuroscience Nurses ISSN: 0888-0395|
|Issue:||Date: Feb, 2005 Source Volume: 37 Source Issue: 1|
|Topic:||Event Code: 440 Facilities & equipment|
|Geographic:||Geographic Scope: United States Geographic Code: 1USA United States|
Abstract: Pupillary size and reactivity have long been a critical
component of the clinical assessment of patients with neurological
disorders. The pupillary examination may provide critical information
related to new or worsening intracranial pathology and facilitate prompt
intervention to minimize further neuronal damage. With this in mind,
intensive care nurses caring for neurologically impaired patients
frequently must perform pupillary examinations in concert with assigning
a Glasgow Coma Scale score. The purpose of this study was to test the
accuracy and reliability of an automated pupillometer compared with the
standard manual examination as a preliminary step in assessing the
usefulness of automated pupillometry in the critical care setting.
Twenty patients in the intensive care units of a teaching hospital were
examined by two groups of three examiners using both the manual
examination with a penlight or similar light source and a portable
automated pupillometer capable of measuring pupil size and reaction.
Measurements by a static pupillometer before and after each pupillary
examination were used to determine the mean "'true" size
of the pupil. This study found that the automated pupillometer is more
accurate and reliable than the manual examination in measuring pupil
size and reactivity. For these reasons, such a device may be a
beneficial addition in the clinical assessment of neurologically
Pupil size, shape, and reactivity have long been used as indicators of neurological function in brain-injured patients, particularly in comatose patients. Pupillary response to light measures the integrity of neuronal pathways governing pupil size. The oculomotor nerve is the efferent link and the optic nerve is the afferent link. Changes in pupillary size and reactivity may indicate second or third cranial nerve involvement and in the brain-injured patient often are signs of increased intracranial pressure and herniation (Chesnut, Gautille, Blunt, Klauber, & Marshall, 1994). For those patients with severe head injury, an unfavorable outcome was inversely related to pupillary responsiveness and pupil size (Ritter et al., 1999). Because a decreased interval between injury and clinical intervention has been associated with improved outcome (Chesnut et al.), the ability to infer the presence of a significant intracranial process from the pupillary examination may help in the triage of patients. An accurate and reliable pupil examination also may be useful in determining whether to transport an unstable brain-injured patient to a computed tomography (CT) scanner, an endeavor not without risk (Hurst et al., 1992).
Many studies have attempted to correlate the pupillary examination with radiologic findings and neurological outcomes (Chesnut et al., 1994). Such studies have been largely inconclusive, however, because of the inconsistencies in the pupillary measurements and the lack of standard definitions for pupillary dilatation, pathologic anisocoria (i.e., unequal pupils), and reactivity. The reason these parameters have been so poorly defined likely relates to the fact that manual pupillary examination lacks precision, in part because of a great deal of interexaminer variability (Litvan et al., 2000). Manual pupillary assessment is highly subjective, particularly when the examiner estimates pupil size in millimeters without the use of a pupil gauge. When such a gauge is available, it is often pictured on the neurological assessment section of the flow sheet; therefore the examiner must make a mental image of a patient's pupil size until he or she is ready to begin documentation.
In this study, the interexaminer variability in pupillary measurements using the manual examination as compared with a portable automated pupillometer was critically examined and quantified. The variability of the manual and pupillometric examination in relation to pupil size and the ability of the manual examination to detect pupillary reactivity as compared with the pupillometer also were analyzed. The researchers hypothesized that measurements obtained using the automated pupillometer would provide intensive care nurses with more objective and reliable data when performing pupillary examinations.
Approval for the study was obtained from the institutional review board at the University of California at San Francisco. Twenty randomly selected patients, ages 4 to 87 years, who were in the intensive care units (ICUs) of the San Francisco General Hospital were examined. Patients with a history of ophthalmological disease were excluded.
Patients were examined by two groups of examiners, each consisting of one neurosurgical attending physician, two neurosurgical interns, and four advanced practice nurses with an average of 15 years of experience assessing pupils of neurologically impaired patients in both the emergency department and ICU. Participants were trained to use the automated pupillometer by an intensive care nurse proficient in using the device. Examiners became proficient in using the device within 20 minutes.
Each examiner in each group examined each patient in succession during a single session. A maximum of 5 minutes elapsed between individual examiners' evaluations. Each examiner measured one pupil manually and with a dynamic pupillometer and then repeated this process for the other pupil. Pupil size and reactivity were measured under dim light conditions with a penlight or other similar light source.
In keeping with standard nursing practice in the hospital's ICU, a pupil gauge was not placed next to the eye during the manual examination. The order of examiners, the method of measuring the pupils, and the pupil examined first were randomized for each patient. Using a static pupillometer, one of the authors measured pupil sizes before and after each examiner performed the pupillary measurements to determine "true" pupil size. The author also monitored the environment to ensure light conditions were dim and consistent for all participants. The examiners' manual size measurements were recorded in 0.5 mm units. Manual pupillary reactions were categorized as nonreactive, sluggish, or brisk. Examiners did not see the results of readings obtained with the static and dynamic pupillometer devices or other examiners' results.
The NeurOptics ForSite dynamic quantitative pupillometer is a portable, battery-operated, automated device with an LCD screen and a digital video camera that measures pupil size and reactivity. The measurement occurs in three automated steps: targeting, video acquisition, and analysis. By pressing the "scan" button, the targeting phase is initiated, and a live video image is displayed in the LCD window. The operator centers the image of the patient's pupil within a large area bounded by red corner marks. At the end of the targeting phase, the pupillometer verifies whether an acceptable pupil image is within the camera's field of vision. If it is, a 3-second capture of pupil images commences. A white light stimulus is flashed at the patient's targeted eye. The pupil light response is calculated in the analysis phase, and the results are summarized, saved, and displayed in the LCD window.
Two different types of pupillometers were used in this study: static and dynamic. The static pupillometer measures only the pupil size and does not stimulate pupillary reactivity. The device was standardized using artificial pupils of known sizes and has a maximum error of 0.1 mm. The dynamic pupillometer stimulates the pupil with a light flash of a fixed intensity and duration (0.8 seconds). A video of the pupillary response is recorded by the digital video camera for both types of pupillometers at 40 frames per second (fps) for 3.2 seconds and saved in the device memory. Each frame then is processed automatically to evaluate the diameter of the pupil as a function of time.
The static pupillometer averages pupillary size over a period of 3.2 seconds. The dynamic pupillometer averages pupillary size before pupillary constriction, that is, during an initial latency period of approximately 200 milliseconds. To evaluate pupillary reaction using the dynamic pupillometer, a normative model was created by the manufacturer; measurements of noninjured participants were taken in different ambient light and eye accommodation conditions.
The amplitude of light reflex is defined as the ratio of the change in pupil size to the initial pupil size. The amplitudes of the responses ranged from less than 10% for small pupils to more than 40% for medium and large pupils. This is mainly due to the nonlinear length-tension property of the pupillary sphincter and dilatory muscles (Usui & Clark, 1978.).
The device first analyzes the velocity profile of the pupillary reaction. The velocity profile of a reactive pupil consists of an initial zero velocity, followed by a large negative velocity, then a small positive velocity, corresponding to the latency, constriction, and recovery phases of pupillary reaction, respectively (Fig 1). Only velocity profiles containing all three distinct phases are considered to be reactive. In addition to analyses of velocity profiles by the pupillometer, all data were downloaded via an infrared transceiver to an external computer where all pupillary profiles were inspected visually to confirm the output from the pupillometer. The pupillometer then categorizes the response of a reactive pupil as brisk or sluggish, according to the manufacturer's normative data. In defining the two categories, the manufacturer defined two confidence limits for each range of pupil size that determine the boundaries of normal response. A pupillary reaction is categorized as brisk if the response falls above the lower confidence limit; otherwise it is categorized as sluggish.
[FIGURE 1 OMITTED
For each reading of each eye of each patient, the error was calculated as the absolute value of the divergence from the interpolated true size. To test whether pupillometer measurements were more accurate than manual measurements, the difference between the examiners' errors for each reading of each eye was calculated, summarized by medians and 95% confidence intervals, and tested by the nonparametric Wilcoxon signed-rank test. These appeared to be the most appropriate methods because the differences in errors appeared to be abnormally distributed, and random effects models found little within-patient or within-eye correlation in the error differences.
Spearman rank correlations were used to quantify the dependence of errors for each method on the interpolated true pupil size. The between-examiner standard deviation (SD) provides a measure of repeatability that does not depend upon the interpolated true sizes. The between-examiner SD were calculated for each method among the three or four examiners who consecutively measured a patient data. These data then were summarized by medians and the difference between methods tested by the Wilcoxon signed-rank test.
For pupil reactivity, each method was classified as producing agreement if all of the three or four examiners consecutively evaluating the eye found the same result. Agreement rates were compared using McNemar's test for matched dichotomous data. Analyses done separately for nurse examiners and physician examiners produced qualitatively similar results to the pooled analyses presented here.
Ten patients had acute intracranial processes, including hemorrhagic stroke or traumatic brain injury. The remaining 10 patients had other illnesses, including acute pancreatitis, necrotizing fasciitis, and pneumonia. Five complete sets of measurements were deleted from the analysis because the pupillometer was unable to register a reading due to factors such as scleral edema. Seventeen patients were heavily sedated, receiving continuous infusions of propofol, fentanyl, lorazepam, ketamine and/or midazolam (Table 1). Three patients had Glasgow Coma Scale (GCS) scores of 3-5 and did not receive any sedatives.
The error in pupillary size measurement was greater for the manual examination than the pupillometer (Table 2). The median absolute error of the manual measurement compared with the interpolated true size was 0.50 mm, 95% confidence interval (CI) 0.47-0.60 mm. The median absolute error of the pupillometer measurement compared with true size was 0.23 mm (CI 0.20-0.27 mm). The median difference in error of the manual examination was greater than the error of the pupillometer by 0.27 mm, with p < 0.0001; that is, the median improvement in accuracy of the pupillometer over the manual examination was 0.27 mm (CI 0.20-0.30 mm).
In addition, the degree of error in pupillary measurement increased with the size of the pupil for both modes of measurement (Table 2). The variability of the pupillary size measurements among examiners as described by the standard deviation in pupillary size measurements was greater for the manual examination than for the pupillometer (Table 2). The median standard deviation in manual size measurements among examiners for each patient, group, and side of measurement was 0.58 mm (CI 0.50-0.58 mm), whereas that of the pupillometer was 0.15 mm (CI 0.12-0.25 mm). The between-examiner SD also increased with pupil size. (Fig 2).
[FIGURE 2 OMITTED]
The pupillometer was consistent for all but one data set (the same patient, examiner group, and side of examination) in determining pupillary reactivity. Sixty out of 71 sets (85%) were briskly reactive for all pupillometer measurements, while 10 sets (14%) were nonreactive for all pupillometer measurements. Thirty-three sets (46%) were briskly reactive for all manual measurements in the group and 9 sets (13%) were nonreactive for all manual measurements in the group. Interexaminer disagreement was smaller for the pupillometer (1.4%, CI 0%-7.6%) than for the manual examination (39%, C128%-52%). The disagreement among the manual measurements was greater by 38% (CI 25-51%) than the disagreement among the pupillometer measurements (McNemar p < .0001).
Dynamic pupillometry consistently found no reflexes in three patients for whom some manual examinations detected small light reflexes, while manual measurements sometimes found no reflexes in 27 patients found to be consistently briskly reactive by the pupillometer (Fig 3). All reflexes that were sometimes absent by manual examination but consistently briskly reactive by pupillometer were in pupils with a mean size of 2.6 mm (SD = 0.7 ram), while reflexes that were sometimes present by manual examination but consistently absent by pupillometer were in pupils with a mean size of 3.2 mm (SD = 1.2 mm).
Assessment of pupillary size, symmetry, and reactivity are an integral part of the critical care nurse's examination of the neurologically impaired patient because changes in pupillary size, asymmetry, and poor or absent reactivity often indicate neurological deterioration. According to Wilson, Amling, Floyd, and McNair (1988), nurses experienced in performing neurological examinations may be able to assess pupil size accurately without a pupil gauge, but with inconsistent interexaminer reliability of pupillary reactivity. In addition, discriminating the sluggish reaction from either a brisk or nonreactive pupil is problematic (Wilson et al.). A sluggish or nonreactive pupil may be a sign of brainstem herniation and increased intracranial pressure (ICP).
Pupillary assessment is especially relevant in clinical management of brain-injured patients in the ICU as an adjunct to other assessments, particularly in situations where brainstem herniation occurs in the absence of increased ICP, such as with posterior fossa lesions. In patients suffering from TBI, loss of pupillary activity or development of pupillary asymmetry greater than 2 mm were shown to be correlated with higher morbidity and mortality (Morris et al., 1998). Patients with unilateral pupillary dilatation do better than those with bilateral pupillary abnormalities (Chesnut et al., 1994).
In a study of 21 patients with acute traumatic epidural hematomas and GCS scores lower than 8, anisocoria (pupillary size difference of > 2 mm) was present in 67% of patients. Reducing the surgery interval for those with anisocoria to less than 90 minutes was associated with a better outcome (Cohen, Montero, & Israel, 1996).
In addition to abnormalities in pupil sizes, abnormalities in the pupillary response also have been used in the clinical management and prediction of outcome in a patient with severe head injury (Choi et al., 1998). A recent study demonstrated that trauma patients with a GCS of 3 and fixed and dilated pupils have no reasonable chance for survival; however, a substantial number of those with a GCS of 3 whose pupils were not fixed and dilated survived their injury (Lieberman et al., 2003).
There are two main components to pupillary measurements: size and reactivity. When considering the accuracy of pupil size measurements, not only the absolute size but also the interexaminer variability must be taken into account. Based upon the data from this study, the manual examination has a median absolute error in pupillary size measurement (0.50 mm) that is twice that of the pupillometer (0.23 mm). This error (0.50 vs. 0.23) refers to the dynamic pupillometer only, as specified in the "Pupil Size" section. The static pupillometer is used to determine the "true pupil size" and is not considered in Table 2. Furthermore, this error increases for both modalities in pupils greater than 3 mm in diameter. This result is explained by the fact that pupillary size fluctuation is maximal (SD ~ 0.5 mm) for mid-sized pupils of approximately 5 mm in diameter and smaller for larger and smaller pupils (Usui & Stark, 1978). Finding an average pupil size using a static pupillometer over an extended period of time therefore is important in yielding a fluctuation-independent "true size". Of even greater importance than the larger absolute error size in the clinical setting is the larger interexaminer disagreement for the manual examination (median SD of pupil size = 0.58 mm) compared with the dynamic pupillometer (median SD of pupil size = 0.15 mm), because it is often a change in examination that prompts a change in clinical management. This issue is further aggravated by the increase in interexaminer variability (as measured by the standard deviation) as a function of size.
These findings are particularly important because most decisions regarding clinical management of patients with abnormal pupils are made when the pupil is large and therefore most susceptible to errors and differences in measurements among physicians and nurses taking care of the patient. It is likely that the true value of pupillometry in the ICU will be in serial measurements of size and reactivity. The data presented here suggest that a trend in a patient's condition as detected by pupillary signs could be detected earlier with a portable infrared pupillometer when compared with a manual pen light examination because the smaller absolute error and variability would allow smaller changes in examination to be considered "real," instead of being attributed to examiner error and variability.
A total of 452 measurements were taken on 20 patients. Five sets of measurements were deleted because the pupillometer was unable to make a measurement, due mainly to periorbital edema or chemosis. This problem presumably has been corrected in a newer model of the pupillometer.
The interexaminer disagreement regarding pupillary reaction was much larger for the manual examination (39%) than the pupillometer examination (1%). This large disagreement among manual examiners has been found in other studies. In a study of 28 patients, researchers found that the interexaminer agreement over the reaction of pupils to light was 68% for the right pupil and 71% for the left pupil (Van Den Berge, Schouten, Boomstra, Van Drunen Little, & Braakman, 1979). In another study, examiner agreement regarding pupillary reaction was found to be poor to good among 68 pairs of nurses employed in adult and pediatric neurosurgical care units (Wilson et al., 1988). The Kappa statistic was used to determine intraclass correlation for the nominal level data of pupillary reaction. For the right pupil, the overall Kappa was 0.3842, and for the left pupil it was 0.4317, with a perfect agreement being 1.00 (Wilson et al.). The degree of disagreement among manual examiners reveals the lack of reliability of the manual examination in the clinical setting where a patient is examined by multiple examiners and the change in a patient's clinical status is based upon prior exams.
Another important property of the pupillometer is its ability to detect pupillary reaction when the manual examination cannot. The current study confirms earlier reports of this phenomenon (Larson & Muhiudeen, 1995). Such cases (Fig 3) are typified by small pupil sizes (M = 2.6 mm, SD = 0.7 mm), which result in small constriction amplitudes that are difficult to detect by manual examination. The pupillary reflex can be seen clearly from the pupillary size profile, however. Because the maximum absolute error of each point in the pupillary size profile is 0.1 mm and the pattern of the profile is consistent with that of pupillary reaction, the presence of pupillary reaction in such cases is reliable.
The ability of the pupillometer to detect changes in pupillary reaction even in small pupils is important. Chesnut et al. (1994) found that pupillary dilatation precedes a noticeable decrease in reactivity as a sign of herniation. However, this was determined through manual examination. With the increased sensitivity of the pupillometer in detecting the presence of pupillary reactivity, one may be able to detect herniation in its early stages, before pupillary dilatation becomes apparent.
In addition to detecting small pupillary reactions, the use of the velocity profile by the pupillometer allows for differentiation between a true pupillary reaction and hippus (i.e., rhythmic fluctuations in pupillary diameter). There were three patients in whom the pupillometer consistently found no reflexes, whereas the manual examination sometimes detected small light reflexes. These cases are most likely due to the observer's error hi denoting hippus as a pupillary reaction.
There are two different components to the inaccuracies in manual pupillary measurements: lack of precise measurement and interexaminer inconsistency. Interexaminer inconsistency impedes the caregiver's ability to interpret the data or changes in the pupil size over time. The large error in the measurement makes it difficult to detect early changes, a finding particularly relevant for ICU nurses who regularly perform hourly pupil examinations to anticipate neurological deterioration.
The pupillometer serves as a quantitative standard that solves both problems. Because it is a mechanical measurement, there is minimal interexaminer inconsistency, as demonstrated by its smaller standard deviation in the pupillary size measurements as compared with manual examination. More precise quantitative measurements allow healthcare professionals to define parameters such as pupillary reaction (e.g., brisk vs. sluggish). The measurement precision not only allows early detection of changes in pupillary response, but also can be used in future studies to determine the definition of a "blown pupil" and other critical values as well as their relationship to CT findings and neurological outcome.
Although the current model of the device does not provide pupil reactivity measurements dichotomized into the more familiar categories of nonreactive, sluggish, or brisk, it does make available critical information with regard to amplitude and percentage of constriction velocity. New software algorithms being developed for this device should correct this limitation. As with manual pupillometry, pupillary assessment using the device on agitated or confused patients is challenging, as several seconds are needed to acquire an accurate reading. Also, as with manual pupillometry, automated pupillometry can be difficult to perform on patients with scleral or periorbital edema.
Despite its limitations, the portable dynamic pupillometer is an important device, that provides a reliable solution to the interexaminer variability often found with manual pupillometry. With its ability to detect decreased constriction velocity, an early indicator of third nerve compression secondary to a mass lesion, the handheld dynamic pupillometer is expected to become part of routine care in the ICU, the emergency department, and possibly in the prehospital setting.
The pupil examination is an integral part of the neurological examination of brain-injured patients in a neurological ICU. Performing frequent pupil assessments may provide critical and time-sensitive information regarding new or worsening intracranial pathology; therefore, an accurate examination is essential. Automated pupillometry may be useful in providing ICU nurses, as well as nurses working in other settings, with a precise and reliable measurement of pupil size and reactivity, particularly for those patients with small or sluggish pupils.
The authors wish to thank the nurses and physicians at the San Francisco General Hospital for their assistance with this project. This work was supported by the University of California, San Francisco Brain and Spinal Injury Center.
Questions or comments about this article may be directed to Michele Meeker, BSN RN, by phone at 415/206-8300 or by e-mail at firstname.lastname@example.org. She is a clinical research nurse for the department of neurological surgery at UCSF San Francisco General Hospital.
Chesnut, R.M., Gautille, T., Blunt, B.A., Klauber, M.R., & Marshall, L.E. (1994). The localizing value of asymmetry in pupillary size in severe head injury: Relation to lesion type and location. Neurosurgery, 34, 840-846.
Choi, S.C., Narayan, R.K., Anderson, R.L., & Ward, J.D. (1988). Enhanced specificity of prognosis in severe head injury. Journal of Neurosurgery, 69, 381-385.
Cohen, J.E., Montero, A., & Israel, Z.H., (1996). Prognosis and clinical relevance of anisocoria-craniotomy latency for epidural hematoma in comatose patients. Journal of Trauma, 41, 120-122.
Hurst, J.M., Davis, K., Jr., Johnson, D.J., Branson, R.D., Campbell, R.S., & Branson, P.S. (1992). Cost and complications during in-hospital transport of critically ill patients: A prospective cohort study. Journal of Trauma, 33, 582-585.
Larson, M.D., & Muhiudeen, I. (1995). Pupillometric analysis of the "absent light reflex." Archives of Neurology, 52, 369-372.
Lieberman, J.D., Pasquale, M.D., Garcia, R., Cipole, M.D., Mark Li, P., & Wasser, T.E. (2003). Use of admission Glasgow Coma Score, pupil size, and pupil reactivity to determine outcome for trauma patients. Journal of Trauma, 55, 437-442.
Litvan, I., Saposnik, G., Maurino, J., Gonzales, L., Saizar, R., Sica, R.E.R et al. (2000). Pupillary diameter assessment: Need for a graded scale. Neurology, 54, 530-531.
Morris, G.F., Juul, N., Marshall, S.B., Benedict, B., & Marshall, L.F. (1998). Neurological deterioration as a potential alternative endpoint in human clinical trials of experimental pharmacological agents for treatment of severe traumatic brain injuries. Neurosurgery, 43, 1369-1374.
Ritter, A.M., Muizelaar, J.P., Barnes, T., Choi, S., Fatouros, P., Ward, J., et al. (1999). Brainstem bloodflow, pupillary response, and outcome in patients with severe head injuries. Neurosurgery, 44, 941-948.
Usui, S., & Stark, L.W. (1978). Sensory and motor mechanisms interact to control amplitude of pupil noise. Vision Research, 18, 505-507.
Van Den Berge, J.H., Schouten, H.J.A., Boomstra, S., Van Drunen Little, S., & Braakman, R. (1979). Interobserver agreement in assessment of ocular signs in coma. Journal of Neurology, Neurosurgery, and Psychiatry, 42, 1163-1168.
Wilson, S.F., Amling, J.K., Floyd, S.D., & McNair, N.D. (1988). Determining interrater reliability of nurses' assessments of pupillary size and reaction. Journal of Neuroscience Nursing, 20, 189-192.
Rose Du, MD PhD, is a resident in neurosurgery for the department of neurological surgery at the University of California, San Francisco.
Peter Bacchetti, PhD, is an adjunct professor for the department of epidemiology and biostatistics at the University of California, San Francisco.
Claudio M. Privitera, PhD, is the senior research specialist at the NeurOptics, Inc., laboratory.
Merlin D. Larson, MD, is a professor of anesthesia at the University of California, San Francisco.
Martin C. Holland, MD, is an associate clinical professor and the chief of neurosurgery at the UCSF San Francisco General Hospital.
Geoffrey Manley, MD PhD, is an associate professor and the chief of neurotrauma at the UCSF San Francisco General Hospital.
Table 1. Analgesics and Sedatives Used During Examinations Drug Dose range (mean) Propofol 0-146 mcg/kg/min (34 mcg/kg/min) Fentanyl 0-1400 mcg/hr (158 mcg/hr) Lorazepam 0-8 mg/hr (1.2 mg/hr) Midazolam 0-3 mg/hr (0.2 mg/hr) Ketamine 0-10 mcg/kg/min (0.5 mg/kg/min) Table 2. Comparison of Pupillary Size Measurements Between Manual Exam and Pupillometer Manual Examination Pupillometer Absolute error in pupil size 0.54 mm 0.23 mm (Median 95% CI) (0.51-0.62) (0.20-0.31) Spearman rank correlation 0.27 0.33 coefficient between error in p = .026 p = .0044 pupil size and pupil size Standard deviation of pupil 0.58 mm 0.15 mm size (Median, 95% CI) (0.50-0.58) (0.12-0.23) Spearman rank correlation 0.35 0.34 coefficient between standard p = .0032 p = .0036 deviation in pupil size and pupil size Difference " Absolute error in pupil size 0.27 (0.20-0.40) (Median 95% CI) P<.0001 Spearman rank correlation coefficient between error in pupil size and pupil size Standard deviation of pupil 0.29 (0.14-0.46) size (Median, 95% CI) p < .0001 Spearman rank correlation coefficient between standard deviation in pupil size and pupil size * Median difference may not equal the difference in medians, because median is a nonlinear function.
|Gale Copyright:||Copyright 2005 Gale, Cengage Learning. All rights reserved.|