- Email: [email protected]
Continuous visual field test supervision may not always be necessary1
Continuous Visual Field Test Supervision May Not Always Be Necessary Rosita E. Van Coevorden, MD, Richard P. Mills, MD, Ying Ying Chen, MD, Howard S. Barnebey, MD Objective: To assess the effect of supervision on computerized visual field (VF) performance and to determine what patient characteristics predict poor unsupervised performance. Design: Randomized, crossover, cross-sectional, clinical trial. Participants: Two hundred unselected patients with definite or suspect glaucoma or neuro-ophthalmic VF indication participated. Intervention: All patients completed two 30 –2 tests of one eye on a Humphrey perimeter, one with continuous active technician supervision and one without supervision after the initial 2 minutes of the test. Main Outcome Measures: Visual field reliability and global VF indices were measured. Results: Supervision had a positive effect on overall reliability (P ⫽ 0.04) but not on individual reliability parameters or any of the global VF indices. There was no difference between Humphrey Field Analyzers I and II in the need for supervision. Predictors of need for supervision were low educational level and a prior test result with false-positive responses. Predictors of an unreliable test were advanced age and a prior test with a high proportion of fixation losses. Conclusion: Supervision is necessary for those with risk factors for unsatisfactory perimetry such as advanced age, low level of formal education, and prior test results with false-positive responses or high fixation losses; in the remainder, omission of supervision can be considered. Ophthalmology 1999;106:178 –181 It is generally acknowledged that patient performance on computerized perimetry improves with active, continuous technician involvement. Textbooks on perimetric technique contain phrases such as “Patients must be . . . given feedback throughout the test,”1 “During the test there should be occasional voice contact with the patient,”2 and “Do not leave the patient alone.”3 The manual of operations for the Collaborative Initial Glaucoma Treatment Study states, “During the test, the monitoring technician should give occasional encouragement and reminders to the patient . . .”4 For most ophthalmic practices, continuous technician supervision of visual field (VF) tests is an expensive requirement. With the cost-containment pressures of managed care, there is considerable interest in abandoning supervision for at least some patients who seem to be able to perform well unsupervised. Yet there are no data on whether patients should be left alone during testing, whether unsupervised test results would be different in predictable ways,
Originally received: July 1, 1997. Revision accepted: August 3, 1998. Manuscript no. 97359. From the Department of Ophthalmology, University of Washington, Seattle, Washington. Presented in part at the ARVO annual meeting, Fort Lauderdale, Florida, May 1997. Supported in part by NIH Grant EY01730, and in part by an unrestricted award from Research to Prevent Blindness, Inc., New York, New York. The authors have no proprietary interest in Humphrey–Zeiss, Inc. Reprint requests to Richard P. Mills, MD, University of Washington, Ophthalmology, Box 356485, Seattle, WA 98195-6485.
178
or whether there are patient characteristics that would predict when supervision is still necessary. The release of the Humphrey Field Analyzer II (HFA II), with its improved ergonomic design, has prompted speculation that patients may be left alone more easily during testing. This study was designed to assess the effect of supervision during computerized threshold perimetry with HFA models I and II in routine clinical use and to determine what patient characteristics were predictive of the need for supervision.
Patients and Methods Two hundred consecutive patients undergoing VF testing for glaucoma, glaucoma suspect, and neuro-ophthalmic indications were examined twice using the 30 –2 full-threshold strategy on a Humphrey Field Analyzer, once with continuous technician supervision and once without such supervision. The eye to be tested (right or left), the perimeter model to be used (HFA I or HFA II), and the order of testing (supervised or unsupervised first) were randomly selected for each patient using a table of random numbers. Patients were excluded only for mean defect worse than ⫺15.00 dB and best-corrected vision worse than 20/60 because such patients usually have trouble with fixation or attentiveness and clearly require technician supervision. Patients were not excluded if they had no prior perimetric experience or if they had comorbidities precluding optimal performance. Of 200 patients tested, there were 34 with no prior perimetric experience. Before all tests, patients were oriented to testing, including how to pause the test by holding down the response button, and given the demonstration test if they were perimetric novices. The foveal threshold test and the first 30 to 60 seconds of the full test were always performed with the technician present. The physiologic blind spot was located again, if fixation
Van Coevorden et al 䡠 VF Test Supervision Not Always Necessary Table 1. Perimeter Model and Supervision Benefits Supervised (S)
Unsupervised (U)
HFA I (n ⴝ 99)
HFA II (n ⴝ 101)
Reliable Reliable Not reliable Not reliable
Reliable Not reliable Reliable Not reliable
80 (80.8%) 10 (10.1%) 2 (2.0%) 7 (7.1%)
80 (79.2%) 11 (10.9%) 7 (6.9%) 3 (3.0%)
losses were being recorded despite stable fixation behavior of the patient as observed on the video monitor. The testing then proceeded with no technician present during an unsupervised test or continuous active technician involvement, including periodic encouragement and recentering, during a supervised test. A minimum of 15 minutes of rest was given between the first and second VF test. For each patient, the following information was recorded: age, highest educational level attained, diagnosis, and number of prior VF tests. For those patients with prior VFs, the reliability parameters (fixation losses, false-positives, and false-negatives) from the immediate previous test on the study eye were transcribed. For each VF test, the following data were recorded: test time, mean defect, short-term fluctuation, pattern standard deviation, and corrected pattern standard deviation, percentage fixation losses, percentage false-positives, percentage false-negatives, and the presence or absence of any reliability flags (XX) on the printout indicating low patient reliability. A VF score was calculated according to a method similar to that used by the Collaborative Initial Glaucoma Treatment Study.4,5 In brief, each test location on the total deviation probability plot is assigned a value from 0 to 4 depending on the highest significance level on the printout that is shared by that point and at least two neighboring points. The VF score is the sum of the values from the 74 test locations not within the physiologic blind spot on the 30 –2 test. These variables were analyzed to determine the effect of supervision by subtracting the unsupervised result from the supervised result and testing for significance using the Mann–Whitney nonparametric test. Tests for differences between HFA I and HFA II in patients requiring supervision to perform a reliable test used a chi-square statistic. The set of possible predictors of need for supervision or reliability was investigated using Mann–Whitney and proportion tests.
Results All patients completed the test protocol, 99 on the HFA I and 101 on the HFA II. With unreliability as the criterion (manufacturer’s “XX” flag) on any of the reliability parameters, we tested for a difference between the HFA I and HFA II in the proportion of patients who benefited from supervision (Table 1). A benefit from supervision was defined as a reliable result with supervision and an unreliable result without supervision. The proportion of patients benefiting from supervision was 10.1% for HFA I and 10.9% for HFA II, while the proportion of cases in which the result was unreliable with supervision and reliable without supervision was 2% for the HFA I and 6.9% for HFA II. We found that supervision was significantly associated with overall reliability (P ⫽ 0.04), but there was no significant difference between the HFA I and HFA II in the benefit from supervision (McNemar tests). The distribution of VF scores for supervised and unsupervised VF and for the score differences between them (S–U) is listed in Table 2. Although the order of testing was randomized, we conducted an analysis of the effect
of test order on the effect of supervision and reliability to detect any pronounced fatigue or learning influences. No significant relationship was found to these or to overall VF score. There was no significant effect of supervision on mean defect, short-term fluctuation, pattern standard deviation, corrected pattern standard deviation, VF score, or on the individual components of reliability (fixation losses, false-positives, and false-negatives) (Table 3). Interestingly, the only significant factor was the impact of supervision on test time between machines. It slightly increased test time on HFA II and decreased it on HFA I. For the remaining analyses, data from the patients tested on the HFA I and HFA II were combined. Careful subjective comparison of the 200 pairs of fields showed four instances of glaucomatous defects that were present on the supervised field but absent on the unsupervised field, whereas in one instance the defect was present on the unsupervised field but absent on the supervised field. In none of these five cases was the glaucomatous field loss missed entirely, potentially leading to a misdiagnosis. The relation of covariates age, highest educational level attained, prior VF experience, and previous VF reliability was examined (Table 4). Patients were categorized as benefiting from supervision if they had a reliable result with supervision and an unreliable result without supervision. The analysis showed that lower education level and higher previous visit false-positives had a significant association with benefit from supervision. A logistic regression model using the 127 patients with prior VF results and the same predictor variables as examined individually in Table 4 was not significant, and model selection through backward elimination chose the null model with no predictors. The same covariates were analyzed as possible predictors of unreliability (Table 5). Older age and higher previous visit fixation losses were significantly associated with unreliability. A regression model using the 127 patients with prior VF results and the same predictor variables as in Table 5 was significant (P ⬍ 0.001), with the strongest predictors being age and number of fixation losses on prior fields.
Discussion Supervision was significantly associated with increased test reliability (P ⫽ 0.04). However, if a Bonferroni or other correction for multiple statistical tests had been applied, the significance of supervision would have disappeared entirely, despite the large sample size of 200 patients. The effect size of supervision was much smaller than had been anticipated during the study design. No difference in the beneficial effect of supervision on reliability was observed between the older HFA I and the newer HFA II perimeter models, despite the better ergonomic design of the HFA II. None of the individual reliability parameters or global VF indices showed any significant effect of supervision. Conventional wisdom Table 2. Summary of Visual Field Score Data Quantiles
Supervised (S) Unsupervised (U) S⫺U
Minimum
25%
50%
75%
Maximum
0 0 ⫺113
4 3 ⫺7
24.5 17.5 0.5
57 70 10
252 230 72
179
Ophthalmology Volume 106, Number 1, January 1999 Table 3. Effect of Supervision and Perimeter Model on Reliability Parameters and Global Indices HFA I
HFA II
Both I and II
Mean (S ⫺ U)
SD
Mean (S ⫺ U)
SD
Mean (S ⫺ U)
SD
SDP for Machine Effect
P for Supervision Effect
⫺0.02 0.00 ⫺0.01 0.03 ⫺0.12 ⫺0.09 ⫺0.04 ⫺0.39 ⫺3.0
0.12 0.11 0.07 1.97 1.68 1.29 0.71 1.90 25.5
0.00 0.01 0.01 ⫺0.20 0.27 0.34 0.16 0.40 2.2
0.13 0.05 0.15 1.44 1.72 1.44 0.88 1.77 23.2
⫺0.01 0.00 0.00 ⫺0.09 0.08 0.13 0.06 0.01 ⫺0.38
0.13 0.08 0.12 1.72 1.71 1.38 0.81 1.87 24.5
0.657 0.847 0.136 0.162 0.685 0.204 0.134 0.013 0.281
0.139 0.407 0.084 0.317 0.517 0.145 0.429 0.700 0.233
Fixation losses False-positive False-negative MD CPSD PSD SF Test time VF score
MD ⫽ mean deviation; CPSD ⫽ corrected pattern standard deviation; PSD ⫽ pattern standard deviation; SF ⫽ short-term fluctuation; VF ⫽ visual field.
has indicated that fixation losses and false-negative responses can be minimized by intermittent reminders to the patient during testing. Apparently, the active technician involvement during the first few minutes of testing, even in the unsupervised group, was sufficient to eliminate an overall significant difference in the proportion of fixation losses and false-negative responses. The effect of supervision on test time, increasing test duration on the HFA II but decreasing it on the HFA I, is difficult to explain since there are no differences between the two models that might create such a result. Patient characteristics that could be easily observed by technicians before testing commenced were tested for their predictive value of a benefit from supervision. Educational level below grade 12 and a prior test with false-positive responses were significant predictors of need for supervision. Potential predictors of a tendency to perform an unreliable test were advancing age, especially older than 70 years of age, and a prior test with a high proportion of fixation losses. A correction for multiple
statistical tests was not used because the number of variables tested against each dependent variable was small. In addition, the effect of supervision would have been rendered even less beneficial with such a correction in place. More than 85% of the patients in this study showed no beneficial effect of supervision; that is, they performed the test equally well regardless of supervision as evaluated by the measures of reliability, VF indices, VF score, and subjective comparisons of defect density. Taking our analysis of significant predictors of supervision benefit and reliability together, it would seem advisable to supervise patients older than 70 years of age, those who had a prior VF test with many fixation losses or falsepositive responses, and/or who had not finished high school. If the patient lacks these or other obvious risk factors for unsatisfactory perimetry such as poor visual acuity or advanced VF loss, and the initial 1 to 2 minutes of the test proceed uneventfully, consideration can be Table 5. Predictors of Reliability
Table 4. Predictors of Benefit from Supervision
Potential Predictor
n
Age Education No. of prior VF Previous fixation losses† Previous false-positives† Previous false-negatives†
200 200 199 127 127 127
Age ⬎ 70 yrs Prior VF
200 199
No Benefit Benefit from from Supervision Supervision P for (mean) (mean) Difference* 59.04 14.64 4.30 0.10 0.03 0.04
61.81 13.19 4.00 0.10 0.07 0.01
Proportion (%)
Proportion (%)
30.7 55.1
47.6 52.4
0.234 0.026 0.404 0.819 0.023 0.225
n
Age Education No. of prior VF Previous fixation losses† Previous false-positives† Previous false-negatives†
200 200 199 127 127 127
58.10 14.63 4.30 0.08 0.03 0.04
One or Both Fields Unreliable (mean)
P for Difference*
64.30 13.90 4.15 0.22 0.04 0.02
0.013 0.067 0.700 0.023 0.037 0.069
Proportion Proportion (%) (%) 0.188 0.999
VF ⫽ visual field. * Mann-Whitney tests for continuous variables; chi-square proportion tests. † For those patients with prior visual field tests.
180
Potential Predictor
Both Fields Reliable (mean)
Age ⬎ 70 yrs Prior VF
200 199
28.8 54.4
47.5 55.0
0.038 0.943
VF ⫽ visual field. * Mann-Whitney tests for continuous variables; chi-square proportion tests. † For those patients with prior visual field tests.
Van Coevorden et al 䡠 VF Test Supervision Not Always Necessary given to leaving the patient alone for the duration of the test. Much technician time would be saved at the cost of only a few VF tests that must be repeated because of unreliability.
References 1. Walsh TJ, ed. Visual Fields: Examination and Interpretation, 2nd ed. San Francisco: American Academy of Ophthalmology, 1996;121 (ophthalmology monographs; 3).
2. Anderson DR. Perimetry With and Without Automation, 2nd ed. St. Louis: Mosby, 1987;207. 3. Stein HA, Slatt BJ, Stein RM. The Ophthalmic Assistant, Fundamentals and Clinical Practice, 5th ed. St. Louis: Mosby, 1988;424. 4. Collaborative Initial Glaucoma Treatment Study. Manual of Operations, revised March, 1997;7–22. 5. Lichter PR, Mills RP, the CIGTS Study Group. Quality of life study— determination of progression. In: Anderson DR, Drance SM, eds. Encounters in Glaucoma Research 3: How to Ascertain Progression and Outcome. Amsterdam: Kugler Publications, 1996;149 – 63.
181
Originally received: July 1, 1997. Revision accepted: August 3, 1998. Manuscript no. 97359. From the Department of Ophthalmology, University of Washington, Seattle, Washington. Presented in part at the ARVO annual meeting, Fort Lauderdale, Florida, May 1997. Supported in part by NIH Grant EY01730, and in part by an unrestricted award from Research to Prevent Blindness, Inc., New York, New York. The authors have no proprietary interest in Humphrey–Zeiss, Inc. Reprint requests to Richard P. Mills, MD, University of Washington, Ophthalmology, Box 356485, Seattle, WA 98195-6485.
178
or whether there are patient characteristics that would predict when supervision is still necessary. The release of the Humphrey Field Analyzer II (HFA II), with its improved ergonomic design, has prompted speculation that patients may be left alone more easily during testing. This study was designed to assess the effect of supervision during computerized threshold perimetry with HFA models I and II in routine clinical use and to determine what patient characteristics were predictive of the need for supervision.
Patients and Methods Two hundred consecutive patients undergoing VF testing for glaucoma, glaucoma suspect, and neuro-ophthalmic indications were examined twice using the 30 –2 full-threshold strategy on a Humphrey Field Analyzer, once with continuous technician supervision and once without such supervision. The eye to be tested (right or left), the perimeter model to be used (HFA I or HFA II), and the order of testing (supervised or unsupervised first) were randomly selected for each patient using a table of random numbers. Patients were excluded only for mean defect worse than ⫺15.00 dB and best-corrected vision worse than 20/60 because such patients usually have trouble with fixation or attentiveness and clearly require technician supervision. Patients were not excluded if they had no prior perimetric experience or if they had comorbidities precluding optimal performance. Of 200 patients tested, there were 34 with no prior perimetric experience. Before all tests, patients were oriented to testing, including how to pause the test by holding down the response button, and given the demonstration test if they were perimetric novices. The foveal threshold test and the first 30 to 60 seconds of the full test were always performed with the technician present. The physiologic blind spot was located again, if fixation
Van Coevorden et al 䡠 VF Test Supervision Not Always Necessary Table 1. Perimeter Model and Supervision Benefits Supervised (S)
Unsupervised (U)
HFA I (n ⴝ 99)
HFA II (n ⴝ 101)
Reliable Reliable Not reliable Not reliable
Reliable Not reliable Reliable Not reliable
80 (80.8%) 10 (10.1%) 2 (2.0%) 7 (7.1%)
80 (79.2%) 11 (10.9%) 7 (6.9%) 3 (3.0%)
losses were being recorded despite stable fixation behavior of the patient as observed on the video monitor. The testing then proceeded with no technician present during an unsupervised test or continuous active technician involvement, including periodic encouragement and recentering, during a supervised test. A minimum of 15 minutes of rest was given between the first and second VF test. For each patient, the following information was recorded: age, highest educational level attained, diagnosis, and number of prior VF tests. For those patients with prior VFs, the reliability parameters (fixation losses, false-positives, and false-negatives) from the immediate previous test on the study eye were transcribed. For each VF test, the following data were recorded: test time, mean defect, short-term fluctuation, pattern standard deviation, and corrected pattern standard deviation, percentage fixation losses, percentage false-positives, percentage false-negatives, and the presence or absence of any reliability flags (XX) on the printout indicating low patient reliability. A VF score was calculated according to a method similar to that used by the Collaborative Initial Glaucoma Treatment Study.4,5 In brief, each test location on the total deviation probability plot is assigned a value from 0 to 4 depending on the highest significance level on the printout that is shared by that point and at least two neighboring points. The VF score is the sum of the values from the 74 test locations not within the physiologic blind spot on the 30 –2 test. These variables were analyzed to determine the effect of supervision by subtracting the unsupervised result from the supervised result and testing for significance using the Mann–Whitney nonparametric test. Tests for differences between HFA I and HFA II in patients requiring supervision to perform a reliable test used a chi-square statistic. The set of possible predictors of need for supervision or reliability was investigated using Mann–Whitney and proportion tests.
Results All patients completed the test protocol, 99 on the HFA I and 101 on the HFA II. With unreliability as the criterion (manufacturer’s “XX” flag) on any of the reliability parameters, we tested for a difference between the HFA I and HFA II in the proportion of patients who benefited from supervision (Table 1). A benefit from supervision was defined as a reliable result with supervision and an unreliable result without supervision. The proportion of patients benefiting from supervision was 10.1% for HFA I and 10.9% for HFA II, while the proportion of cases in which the result was unreliable with supervision and reliable without supervision was 2% for the HFA I and 6.9% for HFA II. We found that supervision was significantly associated with overall reliability (P ⫽ 0.04), but there was no significant difference between the HFA I and HFA II in the benefit from supervision (McNemar tests). The distribution of VF scores for supervised and unsupervised VF and for the score differences between them (S–U) is listed in Table 2. Although the order of testing was randomized, we conducted an analysis of the effect
of test order on the effect of supervision and reliability to detect any pronounced fatigue or learning influences. No significant relationship was found to these or to overall VF score. There was no significant effect of supervision on mean defect, short-term fluctuation, pattern standard deviation, corrected pattern standard deviation, VF score, or on the individual components of reliability (fixation losses, false-positives, and false-negatives) (Table 3). Interestingly, the only significant factor was the impact of supervision on test time between machines. It slightly increased test time on HFA II and decreased it on HFA I. For the remaining analyses, data from the patients tested on the HFA I and HFA II were combined. Careful subjective comparison of the 200 pairs of fields showed four instances of glaucomatous defects that were present on the supervised field but absent on the unsupervised field, whereas in one instance the defect was present on the unsupervised field but absent on the supervised field. In none of these five cases was the glaucomatous field loss missed entirely, potentially leading to a misdiagnosis. The relation of covariates age, highest educational level attained, prior VF experience, and previous VF reliability was examined (Table 4). Patients were categorized as benefiting from supervision if they had a reliable result with supervision and an unreliable result without supervision. The analysis showed that lower education level and higher previous visit false-positives had a significant association with benefit from supervision. A logistic regression model using the 127 patients with prior VF results and the same predictor variables as examined individually in Table 4 was not significant, and model selection through backward elimination chose the null model with no predictors. The same covariates were analyzed as possible predictors of unreliability (Table 5). Older age and higher previous visit fixation losses were significantly associated with unreliability. A regression model using the 127 patients with prior VF results and the same predictor variables as in Table 5 was significant (P ⬍ 0.001), with the strongest predictors being age and number of fixation losses on prior fields.
Discussion Supervision was significantly associated with increased test reliability (P ⫽ 0.04). However, if a Bonferroni or other correction for multiple statistical tests had been applied, the significance of supervision would have disappeared entirely, despite the large sample size of 200 patients. The effect size of supervision was much smaller than had been anticipated during the study design. No difference in the beneficial effect of supervision on reliability was observed between the older HFA I and the newer HFA II perimeter models, despite the better ergonomic design of the HFA II. None of the individual reliability parameters or global VF indices showed any significant effect of supervision. Conventional wisdom Table 2. Summary of Visual Field Score Data Quantiles
Supervised (S) Unsupervised (U) S⫺U
Minimum
25%
50%
75%
Maximum
0 0 ⫺113
4 3 ⫺7
24.5 17.5 0.5
57 70 10
252 230 72
179
Ophthalmology Volume 106, Number 1, January 1999 Table 3. Effect of Supervision and Perimeter Model on Reliability Parameters and Global Indices HFA I
HFA II
Both I and II
Mean (S ⫺ U)
SD
Mean (S ⫺ U)
SD
Mean (S ⫺ U)
SD
SDP for Machine Effect
P for Supervision Effect
⫺0.02 0.00 ⫺0.01 0.03 ⫺0.12 ⫺0.09 ⫺0.04 ⫺0.39 ⫺3.0
0.12 0.11 0.07 1.97 1.68 1.29 0.71 1.90 25.5
0.00 0.01 0.01 ⫺0.20 0.27 0.34 0.16 0.40 2.2
0.13 0.05 0.15 1.44 1.72 1.44 0.88 1.77 23.2
⫺0.01 0.00 0.00 ⫺0.09 0.08 0.13 0.06 0.01 ⫺0.38
0.13 0.08 0.12 1.72 1.71 1.38 0.81 1.87 24.5
0.657 0.847 0.136 0.162 0.685 0.204 0.134 0.013 0.281
0.139 0.407 0.084 0.317 0.517 0.145 0.429 0.700 0.233
Fixation losses False-positive False-negative MD CPSD PSD SF Test time VF score
MD ⫽ mean deviation; CPSD ⫽ corrected pattern standard deviation; PSD ⫽ pattern standard deviation; SF ⫽ short-term fluctuation; VF ⫽ visual field.
has indicated that fixation losses and false-negative responses can be minimized by intermittent reminders to the patient during testing. Apparently, the active technician involvement during the first few minutes of testing, even in the unsupervised group, was sufficient to eliminate an overall significant difference in the proportion of fixation losses and false-negative responses. The effect of supervision on test time, increasing test duration on the HFA II but decreasing it on the HFA I, is difficult to explain since there are no differences between the two models that might create such a result. Patient characteristics that could be easily observed by technicians before testing commenced were tested for their predictive value of a benefit from supervision. Educational level below grade 12 and a prior test with false-positive responses were significant predictors of need for supervision. Potential predictors of a tendency to perform an unreliable test were advancing age, especially older than 70 years of age, and a prior test with a high proportion of fixation losses. A correction for multiple
statistical tests was not used because the number of variables tested against each dependent variable was small. In addition, the effect of supervision would have been rendered even less beneficial with such a correction in place. More than 85% of the patients in this study showed no beneficial effect of supervision; that is, they performed the test equally well regardless of supervision as evaluated by the measures of reliability, VF indices, VF score, and subjective comparisons of defect density. Taking our analysis of significant predictors of supervision benefit and reliability together, it would seem advisable to supervise patients older than 70 years of age, those who had a prior VF test with many fixation losses or falsepositive responses, and/or who had not finished high school. If the patient lacks these or other obvious risk factors for unsatisfactory perimetry such as poor visual acuity or advanced VF loss, and the initial 1 to 2 minutes of the test proceed uneventfully, consideration can be Table 5. Predictors of Reliability
Table 4. Predictors of Benefit from Supervision
Potential Predictor
n
Age Education No. of prior VF Previous fixation losses† Previous false-positives† Previous false-negatives†
200 200 199 127 127 127
Age ⬎ 70 yrs Prior VF
200 199
No Benefit Benefit from from Supervision Supervision P for (mean) (mean) Difference* 59.04 14.64 4.30 0.10 0.03 0.04
61.81 13.19 4.00 0.10 0.07 0.01
Proportion (%)
Proportion (%)
30.7 55.1
47.6 52.4
0.234 0.026 0.404 0.819 0.023 0.225
n
Age Education No. of prior VF Previous fixation losses† Previous false-positives† Previous false-negatives†
200 200 199 127 127 127
58.10 14.63 4.30 0.08 0.03 0.04
One or Both Fields Unreliable (mean)
P for Difference*
64.30 13.90 4.15 0.22 0.04 0.02
0.013 0.067 0.700 0.023 0.037 0.069
Proportion Proportion (%) (%) 0.188 0.999
VF ⫽ visual field. * Mann-Whitney tests for continuous variables; chi-square proportion tests. † For those patients with prior visual field tests.
180
Potential Predictor
Both Fields Reliable (mean)
Age ⬎ 70 yrs Prior VF
200 199
28.8 54.4
47.5 55.0
0.038 0.943
VF ⫽ visual field. * Mann-Whitney tests for continuous variables; chi-square proportion tests. † For those patients with prior visual field tests.
Van Coevorden et al 䡠 VF Test Supervision Not Always Necessary given to leaving the patient alone for the duration of the test. Much technician time would be saved at the cost of only a few VF tests that must be repeated because of unreliability.
References 1. Walsh TJ, ed. Visual Fields: Examination and Interpretation, 2nd ed. San Francisco: American Academy of Ophthalmology, 1996;121 (ophthalmology monographs; 3).
2. Anderson DR. Perimetry With and Without Automation, 2nd ed. St. Louis: Mosby, 1987;207. 3. Stein HA, Slatt BJ, Stein RM. The Ophthalmic Assistant, Fundamentals and Clinical Practice, 5th ed. St. Louis: Mosby, 1988;424. 4. Collaborative Initial Glaucoma Treatment Study. Manual of Operations, revised March, 1997;7–22. 5. Lichter PR, Mills RP, the CIGTS Study Group. Quality of life study— determination of progression. In: Anderson DR, Drance SM, eds. Encounters in Glaucoma Research 3: How to Ascertain Progression and Outcome. Amsterdam: Kugler Publications, 1996;149 – 63.
181