No consensus exists on clinical or epidemiologic

Lack of Correlation of Symptoms With Specialist-Assessed Long-term Asthma Severity* Molly L. Osborne, MD, FCCP; William M. Vollmer, PhD; Kathryn L. Pedula, MS; John Wilkins, MD; A. Sonia Buist, MD; and Mark O Hollaren, MD, FCCP Study objectives: To validate three indicators of asthma severity as defined in the National Asthma Education Program (NAEP) guidelines (ie, frequency of symptoms, degree of airflow obstruction, and frequency of use of oral glucocorticoids), alone and in combination, against severity as assessed by pulmonary specialists provided with 24-month medical chart data. Design: Cross-sectional comparison of questionnaire and clinical-based markers of asthma severity with physician-assessed severity based on chart review. The pulmonologists did not have access to the results of the baseline evaluations when making their severity assessments. Setting and participants: Study participants were 193 asthmatic members (age range, 6 to 55 years) of a large health maintenance organization who underwent a baseline evaluation as part of a separate longitudinal study. This evaluation consisted of spirometry, skin prick testing, and a survey that included questions on symptoms and medication use. The participants in the ancillary study were selected, based on their baseline evaluation, to reflect a broad range of asthma severity. Results: Based on the chart review, 86 of the study subjects (45%) had mild disease, 90 (45%) had moderate disease, and 17 (9%) had severe disease. This physician-assessed severity correlated highly (p < 0.013) with NAEP-based indices of severity based on oral glucocorticoid use (never, infrequently for attacks, frequently for attacks, and daily use) and on spirometry (FEV 1 > 80% predicted, 60 to 80% predicted, and <60% predicted). It did not, however, correlate with current asthma symptoms (< once/week, 2 to 6 times/week, daily) (p 0.87). A composite severity score based on spirometry and the glucocorticoid use data still provided an overall agreement of 63%, with a weighted kappa of 0.40. Conclusions: While current symptoms are the most important concern of patients with asthma, they reflect the current level of asthma control more than underlying disease severity. Investigators must therefore use caution when comparing groups of patients for whom severity categorization is based largely on symptomatology. This observation, that symptoms alone do not reflect disease severity, becomes even more important as health-care delivery moves closer to protocols/practice guidelines and best treatment programs that rely heavily on symptoms to guide subsequent treatment decisions. (CHEST 1999; 115:85 91) Key words: asthma; asthma severity; asthma symptoms; chart review Abbreviations: ED emergency department; NAEP National Asthma Education Program *From the Division of Pulmonary and Critical Care Medicine (Drs. Osborne and Wilkins), Portland Veterans Administration Medical Center; Kaiser Permanente Center for Health Research (Dr. Vollmer and Ms. Pedula), Division of Pulmonary and Critical Care Medicine (Dr. Buist), Oregon Health Sciences University; and the Department of Medicine (Dr. O Hollaren), Oregon Health Sciences University, Portland, Oregon. This project was supported by NIH grant HL 48237. Manuscript received June 16, 1997; revision accepted May 6, 1998. Correspondence to: Molly Osborne, MD, FCCP, P3 Pulm Pulmonary/Critical Care, VA Medical Center, 3710 SW US Veterans Hospital Rd, Portland, OR 97207 No consensus exists on clinical or epidemiologic measures that can be used to classify patients based on disease severity. 1 4 This may be related in part to confusion in the literature between the underlying severity of disease and current level of control. The term severity is sometimes used to refer to both concepts. Conceptually, however, level of control may correlate only weakly with the underlying severity of disease. 5 As an example, a patient with severe asthma taking daily steroids and exercising appropriate allergen avoidance may be well-con- CHEST / 115 / 1/ JANUARY, 1999 85

trolled, while a noncompliant patient with mild to moderate disease may have more severe, disruptive symptoms. Recognizing these difficulties, most efforts to characterize severity have incorporated multiple indicators, including poor symptom control, type and amount of medication use, and objective measurements of lung function. 1,5 For example, 1991 National Asthma Education Program (NAEP) guidelines 6 characterize mild, moderate, and severe asthma according to frequency of symptoms; pulmonary function; school or work absences; methacholine sensitivity; drug regimen and response to medications; degree of exercise tolerance; and healthcare utilization. Despite the usefulness of these guidelines from a conceptual perspective, they do not readily translate into a simple severity index that could be used in clinical and epidemiologic settings. In fact, the guidelines point out that, because of the highly variable nature of asthma, classification of patients into mild, moderate, and severe disease categories is necessarily imperfect, and that the categories may overlap. The problem is further compounded by the fact an individual s underlying disease severity may vary over time. The frequency and/or severity of self-reported symptoms of asthma would seem an intuitively appealing gauge of disease severity, and, as noted above, have been used to help define severity in a variety of settings. 1,5,7 While current symptoms are important, they may more closely reflect the current level of control and/or compliance with medical treatment than underlying disease severity. In support of this concept, the literature reflects a poor correlation between reported symptoms and objective measures of lung function. 5,8,9 This report summarizes our efforts to validate the usefulness of three specific dimensions for assessing asthma severity: frequency of symptoms, degree of airflow obstruction, and self-reported frequency of use of oral glucocorticoids. This analysis was undertaken as part of a larger longitudinal study to define predictors of hospital-based care in asthma. Physician-assessed severity assessments (based on chart review) served as our gold standard. Materials and Methods We report on baseline, cross-sectional data collected as part of a longitudinal study to characterize risk factors for hospital-based care in asthma. All subjects gave informed consent. In the case of minors, both the children and their parents consented to the study. Sample and Research Setting Participants were members of a large health maintenance organization (Kaiser Permanente, Northwest Division) who either were hospitalized for asthma during the 2 years prior to recruitment or had antiasthma medications dispensed at least twice in the year prior to recruitment. At the time of recruitment, all participants reported having physician-diagnosed asthma and indicated that they experienced ongoing symptoms consistent with asthma. By design, participants ranged in age from 6 to 55 years. From the original cohort of 914 subjects, we randomly selected 193 for chart review. In order to provide a good mix of patients with mild, moderate, and severe asthma, participant selection was stratified based on an a priori severity index computed from the available baseline data (see below). Each participant underwent a single clinical evaluation that included spirometry and an interviewer-administered questionnaire requesting information about current respiratory symptoms and medication use. Because of the spirometry protocol, participants were required to be relatively asymptomatic at the time of clinical evaluation, although each could have had an exacerbation requiring steroids up to 2 weeks before the evaluation. The clinical evaluations of 914 participants were performed systematically throughout a 13-month period (January 1993 through January 1994; about 70 participants/month) in order to collect data uniformly throughout the year. For logistical reasons, we did not attempt to assess variation in peak expiratory flow rate as part of the baseline assessment. Questionnaires We developed two questionnaires for use in the study. The first was developed for participants in the 6- to 14-year-old age group and was completed by a parent. A second questionnaire was used for participants aged 15 to 55 years. Both were administered in interview format. The questionnaires covered a wide variety of factors affecting patients with asthma, including respiratory symptoms, asthma and allergy characteristics, and medication use. They were adapted from relevant sections of the American Thoracic Society Division of Lung Disease Respiratory Symptom Questionnaire and the International Union Against Tuberculosis and Lung Disease Bronchial Symptoms Questionnaire. 10 12 The two questions used to define our symptom index were: (1) During the last 4 weeks, have you been wheezing or had coughing or shortness of breath? (2) Have you had nighttime asthma symptoms (awoke with wheezing, shortness of breath, or cough)? Spirometry Subjects performed spirometry using standardized methods and equipment that met or exceeded American Thoracic Society requirements 13,14 (Spirotech S 550 dry rolling seal spirometer [Graseby Andersen, Spirotech Div; Smyrna, GA] using a digital shaft encoder with 10-mL volume resolution). The best FEV 1 was chosen for analysis and expressed as a percent of predicted FEV 1 using the prediction equations of Knudson et al. 15 NAEP-based Asthma Severity Scale The study questionnaire incorporated several facets of severity assessment as defined in the NAEP Expert Panel Report. 6 These included the previously noted questions on frequency of both daytime and nocturnal symptoms and the frequency of oral glucocorticoid use. We combined this information with baseline spirometry to construct three scaled severity indices (Table 1). The symptoms index combined the information on daytime and nocturnal symptom frequency and divided patients into three 86 Clinical Investigations

Table 1 Components of Asthma Severity based on NAEP Guidelines Score Value Full Sample (N 914), % Chart Review Sample (N 193), % Symptoms 1/wk 0 38* 24 2 to 6/wk 1 33 30 Daily 2 29 46 Oral Steroids Not used 0 58 40 50% of time 1 31 39 50% of time 2 9 18 Daily 3 1 3 Spirometry FEV 1 80% predicted 0 71 62 FEV 1 60 to 80% predicted 1 17 18 FEV 1 60% predicted 2 12 20 *Percentages do not add up to 100% due to rounding. Spirometry available for only 831 subjects in full sample. had also intended to abstract data on symptoms, we found that they were seldom mentioned in the chart. Similarly, there was little objective information on pulmonary function; 79% of patients had no information on peak flow, 12% had a measurement recorded during an acute exacerbation, and 9% had measurements recorded on an ongoing basis. Spirometry was rarely performed. Because we felt comfortable that the NAEP-based severity score performed well in the extreme cases, we purposely oversampled subjects with severity in the range less likely to be appropriately classified (eg, severity scores of 2 to 5). Sixty percent of our chart review sample came from this group, and an additional 20% was selected from above and below this range. Within each of these three ranges, we further stratified our selection to be balanced by sex and age (6 to 18 years vs 19 to 55 years). We also required all participants to have performed spirometry at the baseline evaluation. A total of 200 charts were identified in this manner, of which we were able to review 193. We hoped that the chart review would not only validate the severity score, but would help us to identify the component(s) of the score, if any, that warranted revision. groups based on overall frequency of asthma symptoms: less than once a week; 2 to 6 times a week; or daily. The individual categories for each index correspond approximately to the cut points for mild, moderate, and severe asthma as defined in the NAEP guidelines. We summed these three indices to provide an overall severity score, which was then used to categorize subjects as mild (0 to 2), moderate (3 to 4), or severe (5 to 7). We shall refer to this as our NAEP-based severity scale. Because hospitalization and emergency room use were primary outcome variables for our larger longitudinal study, we did not include them in constructing our severity scale. In order to evaluate this scoring system, we conducted a chart review for a subset of the participants and determined a physician-assessed severity rating that served as our gold standard. The pulmonologists who conducted the medical record review used their best clinical judgment, based on information available in the record, to assign their severity ratings. Details of the chart review process are outlined below. Chart Review We performed the chart review using previously established methodology. 16 An initial pilot review of 12 charts was performed by two pulmonologists (JW and MO), who also conducted the subsequent chart reviews. Their independent ratings of asthma severity agreed in 11 of the 12 charts, which were not used in the final analysis. Although we did not perform a formal interrater reliability analysis, the process used to develop our assessment of severity was designed to help ensure a high level of reliability. In addition, we were careful to resolve any uncertainty in the assessment of severity throughout the chart review by discussing the categorization in detail, using both inpatient and outpatient charts. In these cases, classifications were made by consensus of both physicians. In reaching their assessments, the reviewers only used chart information for the 2 years prior to baseline evaluation, and they were blinded to the study data. As part of their chart review, the two pulmonologists collected information on frequency of exacerbations of asthma over the 2-year period (requiring a therapeutic intervention such as increased medical treatment for asthma or addition of an antibiotic), visits to the emergency department (ED) or urgent-care clinic, hospitalizations, and medication prescribing. Although we Statistical Methods All analyses were performed using the SAS software package (SAS Institute; Cary, NC). Standard methods were used to analyze contingency tables. For 2 2 tables, p values are based on the continuity-adjusted 2 statistic. Where appropriate, we used the Mantel-Haenszel 2 statistic to test for trends in contingency tables. The weighted kappa statistic 17 was used to evaluate the agreement between the NAEP-based and chartbased severity scales. An intermediate weight of 1 2 was assigned to mild vs moderate and moderate vs severe disagreements. vs severe disagreements had weights of 0. Unless otherwise stated, all p values are two-sided and a p value 0.05 indicates significance. Results Selected demographic characteristics of the chart review sample are presented in Table 2. The chart review sample was deliberately stratified by age and sex. Overall, 107 of the patients (55%) in the chart review sample had a physician-assessed severity rating of moderate or severe. Table 3 contrasts these 107 subjects and the 86 mild subjects with respect to several factors noted during the chart review. As expected, patients classified as having moderate to severe disease based on the chart review were more likely than those with mild disease to have exacerbations and ED care noted in their charts and were more likely to have been prescribed inhaled antiinflammatory agents and burst oral steroids. Table 3 also shows, however, that no single factor provided perfect separation of the two groups. The use of bursts of oral glucocorticoids provided the best separation between the groups, and would appear to correlate well with factors that influenced physicianassessed severity rankings. Nonetheless, bursts of oral glucocorticoids were prescribed for 19% of CHEST / 115 / 1/ JANUARY, 1999 87

Table 2 Demographic Information patients classified as having mild asthma and were not prescribed for 24% of patients in the moderate to severe groups. Although the type and frequency of asthma symptoms were not routinely noted for most office visits, asthma exacerbations, defined as a change in symptoms requiring intervention by a health-care provider, did appear to be reliably documented. Exacerbations reflected in Table 3 include telephone contacts for acute asthma symptoms as well as urgent care and ED visits. Table 3 Summary of Chart-based Information for Participants Stratified by Physician-assessed Severity Physician-assessed Severity (N 86), % Moderate/ (N 107), % Exacerbations in past 2 years None 48 9 0.001 1 38 13 2 14 25 3 or more 0 52 ED visit in past 2 years No 73 42 0.001 Yes 27 58 Taking cromolyn or inhaled corticosteroids No 66 31 0.001 Yes 34 69 Burst oral steroids No 81 24 0.001 Yes 19 76 Chart Review Sample (N 193), % Age Group, yrs 3to14 36 15 to 34 29 35 to 55 36 Sex Male 48 Female 52 Smoking Status Never 75 Former 19 Current 6 Race Caucasian 92 Other 8 Socioeconomic status* Upper middle/upper class 22 Middle class 57 Working/lower class 21 *Categories by self-report. Concordance of the physician-assessed and NAEPbased severity scores is presented in Table 4. Though highly correlated, substantial disagreement between them still exists. For example, the two scales agreed exactly in only 53% of the cases. This discordance is reflected in the weighted kappa score of 0.31. Of particular concern were the few cases in which mild subjects were classified as severe and vice versa. Even with modification of the questionnaire-based cut points to maximize agreement between the two rating systems, the weighted kappa statistic only increased to 0.34. Similar results were observed when stratifying for age or sex (data not shown). In an attempt to determine possible reasons for the lack of agreement between the NAEP-based and physician-assessed severity scores, we cross-tabulated the component indices of the NAEP-based score with the physician-assessed rating (Table 5). Although the level of oral glucocorticoid use and baseline spirometry values both significantly correlated with the level of physician-assessed severity, we found no correlation between the level of current asthma symptoms and physician-assessed severity. This lack of correlation between physician-assessed asthma severity and patients symptoms also persisted after further analysis in which symptoms were separated into daytime and nocturnal asthma symptoms. As a result of these findings, we computed a revised severity score based only on lung function data and glucocorticoid treatment information. Cut points were determined empirically to provide maximum agreement with the physician-assessed measure of severity. This improved the overall agreement with the physician-assessed severity scale to 63%, with a weighted kappa of 0.40 (Table 6). Discussion This analysis grew out of our efforts to develop a simple index of asthma severity for use in a longitudinal study of 914 individuals, aged 3 to 55 years, Table 4 Concordance of Original NAEP-based Severity and Physician-assessed Severity Original NAEP-based Severity Score (Cutpoints) Physician-assessed Severity, % (N 86) Moderate (N 90) (N 17) Total (0 to 2) 72 51 12 51 0.001 Moderate 24 32 23 28 (3 to 4) (5 to 7) 3 17 65 15 88 Clinical Investigations

Table 5 NAEP-based Severity Components Stratified by Physician-assessed Severity Physician-assessed Severity, % (N 86) Moderate/ (N 107) Symptoms 1/wk 23 25 0.870 2 to 6/wk 31 28 Daily 45 47 Oral steroids Don t use 59 25 0.001 50% of time 35 42 50% of time 6 28 Daily 0 5 Spirometry FEV 1 80% predicted 72 53 0.013 FEV 1 60 to 80% predicted 16 20 FEV 1 60% predicted 12 27 with asthma. Our initial scale incorporated three dimensions of severity suggested by the National Asthma Education Program (symptoms, corticosteroid use, and lung function. 6 Attempts to validate this scale against expert physician opinion based on medical record review did not show significant correlation between current symptom frequency and the physician-assessed severity score. A revised score omitting self-reported symptoms served to marginally improve agreement with the physician-assessed severity score. Our findings are consistent with several recent reports. Severity of disease has been shown in one study to correlate with airway obstruction, 3 while several other investigators have found weak correlations with asthma symptoms and severity. 5,8,9 A recent review highlights some of the difficulties in this process, in which the authors emphasize the various approaches that have been tried, without arrival at an agreed-upon best approach. 5 Although several methods, including questionnaire-based assessment, were discussed in the review, no single method integrated symptoms (eg, wheeze, dyspnea, cough, sputum production) with duration or intensity. Several studies have looked at the correlation between asthma symptoms and objective measures of lung function, including both FEV 1 and peak expiratory flow rate, 1,7,8,18 and have failed to note a correlation. Some have proposed that patients develop a perception tolerance to their asthma symptoms, noticing them less over time, and thus symptoms become a less accurate yardstick of disease severity. 19 Table 6 Concordance of Final NAEP-based Severity Score (Excluding Symptoms Component) and Physician-assessed Severity Final NAEP-based Severity Score (Cutpoints) Physician-assessed Severity, % (N 86) Moderate (N 90) (N 17) (0 to 1) 77 46 6 0.001 Moderate (2 to 3) 22 52 47 (4 to 5) 1 2 47 Studies demonstrate both the heterogeneity of dyspnea perception in patients and the disparity in dyspnea perception between physicians and their patients. For example, Kikuchi et al 20 asked whether dyspnea and chemosensitivity to hypoxia and hypercapnia are factors in fatal asthma attacks. They studied 11 patients with asthma who had near-fatal attacks, 11 patients with asthma who had not had near-fatal attacks, and 16 normal subjects. Their results suggested that reduced chemosensitivity to hypoxia and blunted perception of dyspnea may predispose patients to fatal asthma attacks. Another study suggested that at least some patients are more accurate in guessing peak flow rate than physicians. Specifically, Shim et al 21 demonstrated that physicians were quite inaccurate in estimating the peak expiratory flow rate by examining patients, whereas the patients themselves were far more accurate in guessing the measurement. In addition, patients were able to tell whether the peak expiratory flow rate was better, the same, or worse from day to day. One interpretation of these data is that patients symptoms reflect changes in airflow obstruction rather than absolute lung function. Despite the evidence from these studies, several recently developed, multidimensional asthma severity scales have incorporated symptoms. Bailey et al 1 developed a severity scale after analyzing data from 199 adult asthmatic patients. The authors used standard multivariate and psychometric techniques with data from asthma patients to develop and evaluate a series of scaled questions, in order to address the broad issue of underlying asthma severity. Specifically, they used measures of asthma duration and pulmonary function combined with simple questionnaire scales (assessment by physicians, incidence of various respiratory symptoms and disease, subjective measures of bother with respect to symptoms, medications used and their side effects, and healthcare utilization). In that study, three factors were shown to correlate with physician-assessed severity: symptom intensity, airflow impairment, and management intensity. In that study symptom intensity CHEST / 115 / 1/ JANUARY, 1999 89

Figure 1. Interplay of severity, management, and control in asthma. Case A; moderate disease, poor control: easily controlled with inhaled corticosteroids and -agonist as needed; winds up in ED 2 to 3 times yearly because of noncompliance. Case B; severe disease, good control: well-controlled with low-dose oral corticosteroid, leukotriene modifier, long-acting -agonist and short-acting bronchodilator as needed; no unscheduled health-care utilization. reflected more than symptom frequency, which may contribute to the differences between their observations and ours. Blanc et al 7 developed another severity scale that used parameters of respiratory symptom frequency, asthma history, and past and present use of asthma medications. Rosier et al 22 developed a pediatric functional asthma severity scale based on responses to six survey questions quantifying symptoms and activity restriction due to asthma. These investigators correlated this score with school absence due to asthma, functional impairment, medical visits for asthma, and use of asthma medications. Our data and others suggest that asthma symptoms are better suited to quantifying the level of current control than for assessment of underlying disease severity. This is a very useful concept for the clinician. Level of control incorporates a short-term assessment (ie, days to weeks) of a patient s asthma. This is typically measured in terms of lung function and/or symptoms. We know that the level of asthma control can vary markedly over time. If care is protocol-based, then a knowledge of the regimen, plus symptom frequency, would be a good indicator of severity (barring issues of patient noncompliance). Indeed, the new NAEP guidelines, 23 which divide pharmacologic therapy into quick-relief medications and long-term control medications, allow protocol-based therapy to identify the level of severity for a given patient. Level of control is different from severity, although it is certainly correlated with severity. Figure 1 illustrates this. Global severity is a determinant of level of control, but its impact is modulated by medical management, self-management, and environmental exposures. Similarly, level of control, in conjunction with a wide array of person factors, determines various health outcomes, including health-care utilization, quality of life, and functional status. In conclusion, it is not yet clear what role quantification of asthma symptoms may have in assessing asthma severity. Symptoms are the most important concern of patients with asthma, and they are also an important focus for physicians treating patients with asthma. Inevitably, however, symptoms reflect both level of current control and disease severity. Clinical investigators must therefore use caution when comparing groups of patients in whom severity categorization is largely based on symptomatology. This observation that symptoms alone do not reflect disease severity becomes even more important as health-care delivery moves closer to utilizing protocols, practice guidelines, and best-treatment programs that rely heavily on symptoms to guide subsequent treatment decisions. Simple formulas are not likely to replace physician judgment in assessing the disease severity level in patients with asthma. References 1 Bailey WC, Higgins DM, Richards BM, et al. Asthma severity: a factor analytic investigation. Am J Med 1992; 93:263 269 2 Busse W, Maisiak R, Young K. Treatment regimen and side effects of treatment measures. Am J Respir Crit Care Med 1994; 149:544 550 90 Clinical Investigations

3 Enright PL, Lebowitz MD, Cockcroft DW. Physiologic measures: pulmonary function tests. Asthma outcome. Am J Respir Crit Care Med 1994; 149(2 pt 2):S9 S18 4 Bailey W, Wilson S, Weiss K, et al. Measures for use in asthma clinical research: overview of NIH Workshop. Am J Respir Crit Care Med 1994; 149(2 pt 2): S1 S8 5 O Connor GT, Weiss ST. Clinical and symptom measures. Am J Respir Crit Care Med 1994; 149:S21 S28 6 National Asthma Education Program. Guidelines for the diagnosis and management of asthma. Publication no. 91-3042. Bethesda, MD: National Institutes of Health, 1991 7 Blanc PD, Jones M, Besson C, et al. Work disability among adults with asthma. Chest 1993; 104:1371 1377 8 Apter AJ, ZuWallack RL, Clive J. Common measures of asthma severity lack association for describing its clinical course. J Allergy Clin Immunol 1994; 94:732 737 9 Juniper EF, Kline PA, Vanzieleghem MA, et al. Effect of long-term treatment with an inhaled corticosteroid (budesonide) on airway hyperresponsiveness and clinical asthma in nonsteroid-dependent asthmatics. Am Rev Respir Dis 1990; 142:832 836 10 Burney PG, Chinn S, Britton JR, et al. What symptoms predict the bronchial response to histamine? Evaluation in a community survey of the bronchial symptoms questionnaire (1984) of the International Union Against Tuberculosis and Lung Disease. Int J Epidemiol 1989; 18:165 173 11 Burney PG, Laitinen LA, Perdrizet S, et al. Validity and repeatability of the IUATLD (1984) Bronchial Symptoms Questionnaire: an international comparison. Eur Respir J 1989; 2:940 945 12 Samet JM. Epidemiologic approaches for the identification of asthma. Chest 1987; 91:74S 78S 13 Becklake M, Crapo RO. Lung function testing: selection of reference values and interpretive strategies. Am Rev Respir Dis 1991; 144:1202 1218 14 Enright PI, Johnson LJ, Connett JOE, et al. Spirometry in the Lung Health Study: I. Methods and quality control. Am Rev Respir Dis 1991; 143:1215 1223 15 Knudson RJ, Lebowitz MD, Holberg CJ, et al. Changes in the normal maximal expiratory flow-volume curve with growth and aging. Am Rev Respir Dis 1983; 127:725 734 16 Osborne ML, Vollmer WM, Buist AS. Diagnostic accuracy of asthma within a health maintenance organization. J Clin Epidemiol 1992; 45:403 411 17 Fleiss JL, Cohen J, Everitt BS. Large-sample standard errors of kappa and weighted kappa. Psychol Bull 1969; 72:323 327 18 Li D, German D, Lulla S, et al. Prospective study of hospitalization for asthma: a preliminary risk factor model. Am J Respir Crit Care Med 1995; 151:647 655 19 van Schayck CP, Dompeling P, Rutten MPM, et al. The influence of an inhaled steroid on quality of life in patients with asthma or COPD. Chest 1995; 107:1199 1205 20 Kikuchi Y, Okabe S, Tamura G, et al. Chemosensitivity and perception of dyspnea in patients with a history of near-fatal asthma. N Engl J Med 1994; 330:1329 1334 21 Shim CS, Williams MH Jr. Evaluation of the severity of asthma: patients versus physicians. Am J Med 1980; 68:11 13 22 Rosier MJ, Bishop J, Nolan T, et al. Measurement of functional severity of asthma in children. Am J Respir Crit Care Med 1994; 149:1434 1441 23 National Asthma Education and Prevention Program Expert Panel. Clinical Practice Guidelines. Expert Panel report 2: Guidelines for the diagnosis and management of asthma. Publication no. 97-4051. Bethesda, MD: National Institutes of Health/National Heart, Lung, and Blood Institute, 1997 CHEST / 115 / 1/ JANUARY, 1999 91