ASSESSMENT OF THE WALKING ability of patients

190 Assessing Walking Ability in Subjects With Spinal Cord Injury: Validity and Reliability of 3 Walking Tests Hubertus J. van Hedel, PT, MS, Markus Wirz, PT, Volker Dietz, MD, FRCP ABSTRACT. van Hedel HJ, Wirz M, Dietz V. Assessing walking ability in subjects with spinal cord injury: validity and reliability of 3 walking tests. Arch Phys Med Rehabil 2005;86: 190-6. Objective: To assess the validity and reliability of 3 timed walking tests (Timed Up & Go [TUG], 10-meter walk test [10MWT], 6-minute walk test) in subjects with spinal cord injury (SCI). Design: Cross-sectional study and repeated assessments. Setting: The SCI center of a university hospital in Switzerland. Participants: Validity was assessed by using the data of 75 patients with SCI, and reliability was determined with 22 patients with SCI. Intervention: Patients performed the timed tests and the Walking Index for Spinal Cord Injury II (WISCI II) on the same day. Three measurements within 7 days were taken to assess reliability. Main Outcome Measures: The measures were scatterplots, correlation coefficients (r), and the Bland-Altman plot. Validity was determined in patients with different walking abilities. Results: Overall, correlation of the 3 timed walking tests was excellent with each other ( r.88) and moderate with the WISCI II ( r.60). The correlation between the timed tests for patients with poor walking ability remained high ( r.70) but decreased in WISCI II ( r.35). High correlation coefficients (r.97) were found for intra- and interrater reliability. However, TUG and 10MWT reliability were negatively influenced by a poor walking function. Conclusions: The 3 timed tests are valid and reliable measures for assessing walking function in patients with SCI. Key Words: Gait disorders, neurologic; Rehabilitation; Walking. 2005 by American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation ASSESSMENT OF THE WALKING ability of patients with movement disorders is an important outcome measure in rehabilitation. Valid and reliable walking tests that quantify gait performance allow therapists to evaluate the development of the patient s walking ability throughout his/her rehabilitation. Valid refers to the ability of an instrument in From the Spinal Cord Injury Center, Balgrist University Hospital, Zürich, Switzerland (van Hedel, Wirz, Dietz); and Department of Medical Physics & Biophysics, Catholic University Nijmegen, Nijmegen, The Netherlands (van Hedel). Supported by the Swiss National Science Foundation (NCCR on Neuronal Plasticity and Repair). No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors(s) or upon any organization with which the author(s) is/are associated. Reprint requests to H.J. van Hedel, PT, MS, Spinal Cord Injury Center, Balgrist University Hospital, Forchstr 340, CH-8008 Zurich, Switzerland, e-mail: hvanhede@balgrist.unizh.ch. 0003-9993/05/8602-8801$30.00/0 doi:10.1016/j.apmr.2004.02.010 assessing what it is intended to measure; reliable refers to the reproducibility of measurements, that is, to what extent replicated measurements agree. 1 Several timed walking tests have been developed for this purpose. The validity and reliability of these tests have been investigated in several populations, varying from healthy elderly people 2,3 to patients with stroke, 4,5 Parkinson s disease, 6,7 or lower-limb amputation. 8 In patients with a spinal cord injury (SCI), 2 walking tests have been validated: the Spinal Cord Injury Functional Ambulation Inventory 9 and the Walking Index for Spinal Cord Injury (WISCI), 10 which was revised later into the WISCI II. 11 The WISCI II has an ordinal scale that quantifies a patient s walking ability; a score of 0 indicates that a patient cannot stand and walk and the highest score of 20 is assigned if a patient can walk more than 10m without walking aids or assistance. We introduced this test, together with 3 timed walking tests, in our SCI center. The timed walking tests are the Timed Up & Go (TUG) test, the 10-meter walk test (10MWT), and the 6-minute walk test (6MWT). The TUG measures the time (in seconds) it takes a patient to stand up from an armchair, walk 3m, return to the chair, and sit down. 12 This test was originally developed as a clinical measure of balance in elderly people. 13 The 10MWT measures the time (in seconds) that it takes a patients to walk 10m; it assesses the short-duration walking speed. This test has been used in gait studies of patients with neurologic movement disorders in general, 4 as well as of patients with stroke 5 and Parkinson s disease. 7 The 6MWT measures the distance (in meters) walked within 6 minutes. This test is useful in assessing cardiovascular exercise capacity in elderly patients with congestive heart failure or chronic lung disease 14-17 and walking ability in patients with acquired brain injury. 18 The validity and reproducibility of these timed tests, however, have not been assessed in patients with SCI. Our purpose in this study was to investigate the concurrent validity and reliability of these 3 timed walking tests. Because walking ability varies considerably among patients with SCI, we also investigated the validity of these tests in subgroups of patients with good and poor walking function. METHODS Assessment of Validity and Reliability To establish the validity of the gait assessment, the data of patients with a traumatic or ischemic SCI were included in this cross-sectional study. Subjects were selected from among all patients who were admitted for rehabilitation or ambulant check-up and were seen by a physical therapist in our hospital between May 2001 and August 2003. We selected those patients who were able to perform most of the walking tests (WISCI II score, 0) and who had no additional gait-impairing deficits. Patients who entered the study between May and December 2001 were scored with the original WISCI; those scores were recalculated after publication of the WISCI II. The study conformed with the Declaration of Helsinki, and the administration of the tests conformed to our hospitals

WALKING ABILITY IN SPINAL CORD INJURY, van Hedel 191 Table 1: Characteristics of the Patient Populations in Both Assessments Characteristics Assessment of Validity (75 patients 100%) Assessment of Repeatability (22 patients 100%) Age (y) Mean SD 54 20 52 20 Range 17 84 21 77 Gender, n (%) Female 30 (40) 8 (36) Male 45 (60) 14 (64) ASIA class, n (%) A 4 (5) 1 (5) B 3 (4) 0 (0) C 7 (9) 3 (14) D 61 (81) 18 (82) Level of lesion, n (%) Cervical 25 (33) 7 (32) Thoracic 21 (28) 7 (32) Lumbar 21 (28) 7 (32) Sacral 8 (11) 1 (5) NOTE. Values may not equal 100% due to rounding. Abbreviations: ASIA, American Spinal Injury Association Impairment Scale; SD, standard deviation. clinical standards. The characteristics of the 75 patients in the study of the validity of the tests are reported in table 1. Whenever possible, patients performed all 4 tests. However, at the onset of rehabilitation, patients could not always perform the 6MWT. Only measurements from tests performed on the same day were included in our analysis. To establish the reliability of the gait assessment, patients were tested on 3 occasions within 7 days. They were not permitted to wear dress shoes for the tests. 19 The patients were all inpatients at some point between October 2001 to August 2003, and met the following criteria. They could perform all the walking tests, had no additional gait-impairing deficits, and were willing to participate in the study. The tests were always performed in the morning, before other therapies were given. Intrarater reliability was determined by comparing the 2 measurements performed by 1 therapist (test 1, test 2), whereas interrater reliability was determined between 2 measurements taken by 2 different therapists (test 1, test 3). The characteristics of the 22 patients included in this reliability study are shown in table 1. All patients were informed about the study and gave written consent. Subgroups Walking ability varies considerably among patients with SCI. Therefore, we investigated the concurrent validity for subgroups of patients with good and poor walking function. The grouping of the patients was based on criteria of the WISCI II. A first comparison was made between patients with a WISCI II score of less than or equal to 10 (group 0 10) and those with a score between 11 and 20 (group 11 20). A second comparison was made between patients who were unable to walk freely or who needed assistance (WISCI II categories 0 8, 10, 11, 14, 17) and those who could walk without assistance (WISCI II categories 9, 12, 13, 15, 16, 18 20). These groups were designated dependent and independent walkers, respectively. Statistical Analysis Concurrent validity was determined by assessing the relationship between the various tests. Because the WISCI II has an ordinal scale, correlations including the WISCI II were quantified using the Spearman rank correlation ( ). The correlations among the TUG, the 10MWT, and the 6MWT were determined with the Pearson correlation coefficient (r) or the Spearman rank correlation, as appropriate. Intra- and interrater reliability were determined by calculating the Pearson correlation coefficient between tests 1 and 2 and tests 1 and 3, respectively. Furthermore, the Bland-Altman plot 20 was used to assess the degree of agreement between tests performed by the same therapist and between 2 different therapists. This method plots the average of 2 measurements versus the difference between these 2 measurements. The differences between the 2 measurements were assessed by using a paired t test or Wilcoxon signed-rank test, as appropriate. With the Bland-Altman plot, the relationship between the measurement error and the true value can be evaluated. 20 This method also allows for testing whether the measurements are normally distributed by using a 95% confidence interval (CI). RESULTS As expected, the results of the walking tests evaluated for their validity varied profoundly among the 75 patients. The time required for the TUG varied between 8 and 156 seconds (mean standard deviation, 36 27s), and that for the 10MWT varied between 6 and 190 seconds (mean, 30 28s). The distance walked in 6 minutes varied between 23 and 475m (mean, 205 120m). The WISCI II score varied between 3 and 20 (median, 13). Correlations Between the WISCI II and the Timed Walking Tests for All Patients Figure 1 shows the relation between the WISCI II and the TUG, 10MWT, and 6 MWT. There was a very good significant correlation 1 between the WISCI II and the TUG (.76; number of observations [n] 67) for the patients and a moderate to good significant correlation between the WISCI II and 10MWT (.68, n 67). These negative correlation coefficients indicate that an increase in walking function as indicated by the WISCI II score was associated with a decrease in the time needed to perform the TUG or 10MWT. There was also a moderate to good significant correlation between the WISCI II score and the 6MWT (.60, n 60). The positive correlation coefficient indicates that an improved walking function, as indicated by the WISCI II score, was associated with larger distances walked within the 6 minutes. Correlations Among the Timed Walking Tests for All Patients Figure 2 shows the scatterplots of the data obtained in the 3 timed walking tests for all 75 patients. Although the relationship between the TUG and the 10MWT appeared to be linear, the correlation between the 6MWT and the 2 other timed tests was exponential. There were excellent significant associations between the TUG and the 10MWT (r.89, n 70), the 6MWT and the TUG (.88, n 62), and the 6MWT and the 10MWT (.95, n 62). Correlations Between WISCI II and the Timed Walking Tests: Subgroups For patients with a WISCI II score of 0 to 10, the relationship between the WISCI II and the 3 other walking tests was little to none and did not differ from zero (WISCI II vs TUG:.16, n 20; WISCI II vs 10MWT:.24, n 20; WISCI II vs 6MWT:.22, n 13). For the group with a WISCI II score of 11 to 20 (n 47), the relationship was fair and signif-

192 WALKING ABILITY IN SPINAL CORD INJURY, van Hedel (n 23),.70 (n 15), and.96 (n 15) between the TUG and the 10MWT, the 6MWT and the TUG, and the 6MWT and the 10MWT, respectively. These values were.79,.78, and.93 for the patient group with WISCI II scores of 11 to 20 (n 47). Similar results were found for the dependent and independent groups. For the dependent group, the correlation coefficients were also calculated between the TUG and the 10MWT (r.88, n 27), the 6MWT and the TUG (.74, n 18), and between the 6MWT and the 10MWT tests (.92, n 19). For the independent group, these values were.86 (n 43),.88 (n 44), and.94 (n 43), respectively. Intra- and Interrater Reliability Figure 3 shows the scatterplots of the repeated measurements of walking function in patients by the same therapist (fig 3A) and by 2 different therapists (fig 3B). When expressed in correlation coefficients, the intrarater reliability assessed for 22 patients was excellent for the TUG (r.979, P.001), the 10MWT (r.983, P.001), and the 6MWT (r.981, P.001). One therapist measured 2 patients only. Therefore, the interra- Fig 1. Relationships between the timed walking tests and WISCI II. Relationships between (A) the WISCI II and TUG, (B) the WISCI II and 10MWT, and (C) the WISCI II and 6MWT. Subjects were divided into 2 groups according to the WISCI II independent walkers (without use of parallel bars or assistance) and dependent walkers (dependent on parallel bars or the support of another person). icant (WISCI II vs 10MWT:.49), or moderate to good and significant (WISCI II vs TUG:.65; WISCI II vs 6MWT:.64). A similar relationship was found when dependent and independent walkers were grouped. Figure 1 shows the relationships between the WISCI II and the 3 timed walking tests for both the dependent and independent subgroups. For the dependent group, the relationships between the WISCI II and TUG or 6MWT were little to none (WISCI II vs TUG:.22, n 23; WISCI II vs 6MWT:.21, n 15). A fair but not significant relationship existed between the WISCI II and the 10MWT (.35, n 24). The relationships were higher for the independent group and therefore similar to the 11 20 WISCI II score group. The significant correlation coefficients were.66 (n 44),.48 (n 43), and.65 (n 45) for the WISCI II versus the TUG, 10MWT, and 6MWT, respectively. Correlations Among Timed Walking Tests: Subgroups The correlations between the 3 timed walking tests were all significant and between good and excellent. Only small differences in correlation coefficients existed between the poor walking patients and the good walking patients. For patients with a WISCI II score of 0 to 10, the correlation coefficients were.92 Fig 2. Relationships between the timed walking tests. Relationship between (A) the 10MWT and TUG tests, (B) the TUG and 6MWT, and (C) the 10MWT and 6MWT.

WALKING ABILITY IN SPINAL CORD INJURY, van Hedel 193 Fig 3. Intra- and interrater reliability of 3 timed walking tests. Scatterplots showing (A) intrarater and (B) interrater results of (top) the TUG, (middle) the 10MWT, and (bottom) the 6MWT.

194 WALKING ABILITY IN SPINAL CORD INJURY, van Hedel ter reliability could be determined for only 20 subjects. The interrater reliability was excellent for the TUG (r.973, P.001), the 10MWT (r.974, P.001), and the 6MWT (r.970, P.001). Figure 4 shows the plots for the reliability of the walking tests by using the Bland-Altman plot. The differences between the 2 measurements were plotted versus the averages of the 2 measurements. The closer the points are located to the horizontal axis, the better the reliability. For the TUG, the mean difference between 2 measurements made by the same therapists was 3.3 7.0 seconds. This differed significantly from zero (Wilcoxon signed-rank test, P.001). Similarly, the intrarater measurements for the 6MWT ( 20.5 27m) also differed significantly from zero (paired t test, P.002). Because these differences were significantly greater than zero, intrarater reliability could not be assessed further and the 95% CI is not indicated in figure 4. There was no significant difference between the intrarater measurements for the 10MWT (0.5 6.0s). The same was true for the interrater assessments for all 3 timed walk tests (TUG, 0.3 7.5s; 10MWT, 0.1 7.0s; 6MWT, 14.8 33.6m). A normal distribution of the differences between the measurements was found only for the interrater assessments of the TUG, because 95% or more of the differences between the measurements were within the CI (fig 4B). A normal distribution was not present in the other tests. For the TUG and the 10MWT, it appears that the differences increased when the average time needed to perform the test was higher (figs 4A, 4B). For patients who performed the tests within 40 seconds, the repeatability was very good (figs 4A, 4B). DISCUSSION Our purpose in this study was to investigate the concurrent validity of 3 timed walking tests in patients with SCI. The main findings were as follows. One, the WISCI II correlated well with the 3 timed walking tests when all patients were included. Two, excellent correlation coefficients existed when the 3 walking tests were compared with each other. Three, in patients with severe walking disabilities, there was little correlation between the WISCI II and the 3 timed tests. The correlation coefficients between the 3 tests remained excellent within this subgroup of patients. Four, the correlation coefficients calculated for the intra- and interrater reliability were excellent (r.97). However, the Bland-Altman plot unexpectedly indicated that interrater reliability was better than intrarater agreement and that repeatability of the TUG and 10MWT depended on the patient s walking performance. The validity of the timed walking tests appeared to be good in patients with SCI. Compared with the results of other studies, the moderate to good correlations of the 3 tests with the WISCI II indicate that the tests can be used in both clinical practice and research to assess walking function in patients with SCI. This agrees with observations made in other subject groups. In healthy elderly people, good correlations were found between the TUG and the Berg Balance Scale (r.76) and between the TUG and the Tinetti Balance Scale (r.74). 21 Furthermore, the TUG had a good correlations with the physical subscales of the Sickness Impact Profile (mobility control,.46; mobility range,.36) in patients with an unilateral lower-limb amputation. 8 The correlation between the TUG and the WISCI II in our study was in a similar range. The 10MWT has been validated for patients with neurologic diseases, 4 but patients with SCI were not included in that study. A pronounced walking disability was described in patients using aids and in those with a sensory impairment. The 6MWT correlated closely with the peak oxygen consumption in patients with heart failure (r.64) and pulmonary disease (r.73). 22 The correlation coefficient with the WISCI II in our study was good. In summary, all 3 timed walking tests showed good correlations with the WISCI II and, therefore, are valid for use with in patients with SCI. Furthermore, good validity can also be expected, because the correlations between the 3 tests were excellent and did not depend on the patients walking abilities. When we investigated the influence of walking ability on the test validity by dividing the patients into groups with good and poor walking abilities, the correlations between WISCI II and timed walk tests changed. For patients with severely impaired walking function, the correlations were low. Patients with a WISCI II score less than or equal to 10 showed a positive correlation between the WISCI II and the TUG test and a negative correlation between the WISCI II and the 6MWT. This means that an improved walking function, as indicated by the WISCI II score, was associated with the increased time needed to perform the TUG or a decreased distance covered in 6 minutes. When the patients were divided according to other criteria of the WISCI II (dependent and independent walkers), the expected correlation returned. It might therefore be questioned whether the ranking of the categories in the WISCI II was done in an optimal way. Furthermore, in SCI, the patient s goal is to achieve independent walking, even if orthoses or other walking aids must be used. Therefore, it might be better to score independent gait as being superior to gait requiring external assistance. The intra- and interrater reliabilities were excellent, when using correlation coefficients. Unexpectedly, the interrater reliability assessed by the Bland-Altman plot 20 was superior to the intrarater reliability. The finding that subjects performed the TUG and 6MWT better during the second assessment by the same therapists may indicate that the first measurement affected the second one. Both tests assess a more difficult task, that is, balance for the TUG 13 and cardiovascular exercise for the 6MWT, 14-17 when compared with the 10MWT. The slightly better performance on the second test may indicate that the patients became rapidly familiarized with these tests. The differences between the measurements of 2 therapists for the 10MWT and 6MWT lacked normal distribution, which may be because of the relatively small sample size (n 20). Furthermore, the Bland-Altman plots indicate that the reliability was excellent when the TUG and the 10MWT were completed within 40 seconds but became worse for SCI patients with poor walking ability. In these patients, the 2 measurements differed considerably. We assume that this is because of a high day-to-day variability in performance. Our subjects were not randomly recruited from the patient population. Most of their characteristics, however, were comparable to those of a representative sample of 414 Swiss patients, who received their first rehabilitation between 1990 and 1994. 23 The mean age of our subjects was about 10 years higher and could be explained by 2 factors: (1) we included patients who had already finished their first rehabilitation and (2) the percentage of patients with ischemic etiologies, which occur later in life, is slowly increasing. 23 Age affects walking speed negatively, 3,24 therefore, the measured test durations may be longer and the distances may be shorter than expected for the patient population. Although this may have affected the reliability negatively, we do not believe that the validity has been influenced. CONCLUSIONS The 3 timed walking tests are valid measures for assessing functional gait in patients with SCI. These tests appear to have

WALKING ABILITY IN SPINAL CORD INJURY, van Hedel 195 Fig 4. Bland-Altman plot to assess reliability of walking tests. (A) Intrarater and (B) interrater reliability plots using the Bland-Altman plot, that is, the differences between scores are plotted against the averaged scores for (top) the TUG, (middle) the 10MWT, and (bottom) the 6MWT. The 95% Cls of the TUG and 6MWT are not shown for the intrarater reliability, because the differences between the measurements differed significantly from zero and, therefore, reliability could not be assessed.

196 WALKING ABILITY IN SPINAL CORD INJURY, van Hedel good interrater reliability. In patients with poor walking ability, however, the results of the TUG and 10MWT must be interpreted with caution. Acknowledgments: We thank all the physical therapists of the Spinal Cord Injury Center of Balgrist University Hospital for performing the tests. We also thank Rachel Jurd for her help with the English language. References 1. Dawson-Saunders B, Trapp RG. Basic and clinical biostatistics 2nd ed. Norwalk: Appleton & Lange; 1994. 2. Harada ND, Chiu V, Stewart AL. Mobility-related function in older adults: assessment with a 6-minute walk test. Arch Phys Med Rehabil 1999;80:837-41. 3. Steffen TM, Hacker TA, Mollinger L. Age- and gender-related test performance in community-dwelling elderly people: Six- Minute Walk Test, Berg Balance Scale, Timed Up & Go Test, and gait speeds. Phys Ther 2002;82:128-37. 4. Rossier P, Wade DT. Validity and reliability comparison of 4 mobility measures in patients presenting with neurologic impairment. Arch Phys Med Rehabil 2001;82:9-13. 5. Smith MT, Baer GD. Achievement of simple mobility milestones after stroke. Arch Phys Med Rehabil 1999;80:442-7. 6. Morris S, Morris ME, Iansek R. Reliability of measurements obtained with the Timed Up & Go test in people with Parkinson disease. Phys Ther 2001;81:810-8. 7. Schenkman M, Cutson TM, Kuchibhatla M, Chandler J, Pieper C. Reliability of impairment and physical performance measures for persons with Parkinson s disease. Phys Ther 1997;77:19-27. 8. Schoppen T, Boonstra A, Groothoff JW, de Vries J, Goeken LN, Eisma WH. The Timed up and go test: reliability and validity in persons with unilateral lower limb amputation. Arch Phys Med Rehabil 1999;80:825-8. 9. Field-Fote EC, Fluet GG, Schafer SD, et al. The spinal cord injury functional ambulation inventory (SCI-FAI). J Rehabil Med 2001; 33:177-81. 10. Ditunno JF Jr, Ditunno PL, Graziani V, et al. Walking index for spinal cord injury (WISCI): an international multicenter validity and reliability study. Spinal Cord 2000;38:234-43. 11. Dittuno PL, Dittuno JF Jr. Walking index for spinal cord injury (WISCI II): scale revision. Spinal Cord 2001;39:654-6. 12. Podsiadlo D, Richardson S. The timed Up & Go : a test of basic functional mobility for frail elderly persons. J Am Geriatr Soc 1991;39:142-8. 13. Mathias S, Nayak US, Isaacs B. Balance in elderly patients: the get-up and go test. Arch Phys Med Rehabil 1986;67:387-9. 14. Butland RJ, Pang J, Gross ER, Woodcock AA, Geddes DM. Two-, six-, and 12-minute walking tests in respiratory disease. Br Med J (Clin Res Ed) 1982;284:1607-8. 15. Guyatt GH, Sullivan MJ, Taylor DW, Berman LB. The 6-minute walk: a new measure of exercise capacity in patients with chronic heart failure. Can Med Assoc J 1985;132:919-23. 16. Roomi J, Johnson MM, Waters K, Yohannes A, Helm A, Connolly MJ. Respiratory rehabilitation, exercise capacity and quality of life in chronic airways disease in old age. Age Ageing 1996; 25:12-6. 17. Cahalin LP, Mathier MA, Semigran MJ, Dec GW, DiSalvo TG. The six-minute walk test predicts peak oxygen uptake and survival in patients with advanced heart failure. Chest 1996;110:325-32. 18. Mossberg KA. Reliability of a timed walk test in persons with acquired brain injury.am J Phys Med Rehabil 2003;82:385-90. 19. Arnadottir SA, Mercer VS. Effects of footwear on measurements of balance and gait in women between the ages of 65 and 93 years. Phys Ther 2000;80:17-27. 20. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307-10. 21. Berg KO, Maki BE, Williams JI, Holliday PJ, Wood-Dauphinee SL. Clinical and laboratory measures of postural balance in an elderly population. Arch Phys Med Rehabil 1992;73:1073-80. 22. Cahalin L, Pappagianopoulos P, Prevost S, Wain J, Ginns L. The relationship of the 6-min walk test to maximal oxygen consumption in transplant candidates with end-stage lung disease. Chest 1995;108:452-9. 23. Zäch GA, Koch HG, Wolfensberger M, Michel D. [Demography and statistics of paraplegia] [German]. In: Zäch GA, editor. Querschnittlähmung ganzheitliche Rehabilitation. Küsnacht (Switzerland): Verlag Dr. Felix Wüst AG; 1995. p 16-9. 24. Enright PL, Sherill DL. Reference equations for the six-minute walk in healthy adults. Am J Respir Crit Care Med 1998;158: 1384-7.