Predictive Value of a Cross-Cultural Asthma Case-Detection Tool in an Elementary School Population

Predictive Value of a Cross-Cultural Asthma Case-Detection Tool in an Elementary School Population Stanley P. Galant, MD*; Linda J.R. Crawford, RN, PNP‡; Tricia Morphew, MS§; Craig A. Jones, MD㛳; and Stanley Bassin, PhD¶ ABSTRACT. Objective. Bronchial asthma, which affects ⬃5 million US children, is vastly underdiagnosed and treated, particularly among minorities and those of low socioeconomic status. Because current methods of detecting those at greatest risk of asthma in a multicultural setting appear inadequate, we assessed the validity and reliability of a new asthma questionnaire across 3 dominant cultures in Orange County, California (white, Hispanic, and Vietnamese). Methods. Children in grades 1, 3, and 5 and their families, in 3 different schools representative of these major ethnic groups, were randomly selected to participate in the validation process. Two schools with low socioeconomic status and dominant Hispanic or Vietnamese minorities were designated inner-city schools, whereas the third school was a suburban school with predominately white students. Participants completed a 7-question, 11-element questionnaire in their primary language, followed by an asthma evaluation (history, physical examination, and spirometry) by an asthma specialist (who was blinded with respect to the results of the questionnaire), at their respective schools. The physician then made a determination regarding the presence and severity (according to National Institutes of Health guidelines) of asthma. Several weeks later, the entire student body was asked to complete the questionnaire at home and return it to school for analysis. Validation of each item was evaluated for sensitivity, specificity, and positive and negative predictive values, and application of univariate analyses provided an estimated probability of an asthma diagnosis by the asthma specialist. A “bestfit” algorithm was determined with all 11 elements, if possible, and an abbreviated algorithm that selected the fewest-question combination that yielded the best asthma predictability was established. Reliability was established with the percent agreement between the 2 questionnaires and the ␬ statistic. Results. Of the 401 children/families who participated in the validation analysis, 45% were Hispanic, 22% white, 19% Vietnamese, and 15% other. The overall prevalence of asthma specialist-diagnosed asthma was 28%,

From the *Children’s Hospital of Orange County, Orange, California; ‡Orange County School Nurses Organization, Costa Mesa, California; §Southern California Chapter of the Asthma and Allergy Foundation of America, Los Angeles, California; 㛳Division of Allergy and Immunology and Department of Pediatrics, Los Angeles County and University of Southern California Medical Center and Keck School of Medicine, University of Southern California, Los Angeles, California; and ¶California State University, Pomona, California. Accepted for publication Apr 5, 2004. DOI: 10.1542/peds.2003-0575-F Reprint requests to (S.P.G.) 1201 W. La Veta Ave, Suite 501, Orange, CA 92868. E-mail: [email protected] PEDIATRICS (ISSN 0031 4005). Copyright © 2004 by the American Academy of Pediatrics.

http://www.pediatrics.org/cgi/content/full/114/3/e307

with 65% of cases being graded as intermittent and 35% as persistent. Sixty-two percent of the children had not been previously diagnosed with asthma. There were no significant differences among cultures in sensitivity or specificity for any of the individual questions or the complete or abbreviated algorithms. The abbreviated algorithm with 3 questions, ie, question 1 (asthma in the past 2 years), question 4 (cough, chest tightness, trouble breathing, or wheezing with exercise), and question 6 (same symptoms in the morning or day in the past 4 weeks) yielded comparable sensitivity and specificity for the complete algorithm in all groups. The abbreviated algorithm had >86% predictability in detecting children with persistent asthma and 56% predictability in detecting children with intermittent asthma. Reliability was also excellent, with percent agreement usually > 80% and ␬ values of >.70. Conclusions. This asthma detection tool has been shown to be suitable for detecting persistent asthma in a multicultural inner-city population, as well as in a suburban setting. An abbreviated algorithm with 3 questions and >80% predictability in detecting persistent asthma seems ideal for evaluating large numbers of school-aged children. The school setting is an excellent site for identifying children with asthma. Although there is concern that subjects detected in the school setting might not have access to ongoing medical care, case detection is an important first step that could lead to earlier diagnosis and treatment. Reducing the barriers to good care in inner-city environments is the next step. Pediatrics 2004; 114:e307–e316. URL: http://www.pediatrics.org/cgi/ content/full/114/3/e307; asthma, elementary school, detection tool, validation, cross-cultural. ABBREVIATIONS. NIH, National Institutes of Health; SES, socioeconomic status; ED, emergency department; NPV, negative predictive value; PPV, positive predictive value.

B

ronchial asthma, the most common chronic disease in childhood, has a reported prevalence of 4.3% to 6.7%.1 However, recent studies indicate a much higher prevalence, reaching 30% in inner-city populations, with a large percentage of cases not having been previously diagnosed.2,3 Furthermore, bronchial asthma is the most frequent cause of hospitalization and school absenteeism.4,5 To the extent that asthma morbidity, which is so prevalent today, is attributable to underdiagnosis and lack of appropriate treatment, schools and the medical community are faced with the issue of defining their roles in the care of children with asthma. This problem is particularly relevant for schools serving low-income and minority populations, PEDIATRICS Vol. 114 No. 3 September 2004

e307

TABLE 1.

Characteristics of Selected Pediatric Validation Studies

Study

No. of Questions

School-Based

Inner City

Random Selection Attempted in Community Served

Multilingual

Asthma Specialist Validation

Hall et al29 Bauer et al27 Wolf et al20 Jenkins and Clarke24 Glasgow et al7 Frank et al31 Sole´ et al25 Joseph et al30 Redline et al16 Jones et al28 Current study

4 5 (3)† 3‡ 2 5 12 8 5 7§ 7§ 7§

No Yes Yes Yes Yes No No Yes Yes Yes Yes

No Yes Yes No No No No Yes No Yes Yes

No Yes No No No No No No No No Yes

Yes No No No No No No No No Yes Yes

Yes/no* No Yes Yes Yes No Yes/no* No Yes Yes Yes

* Not all patients were evaluated by a specialist. † Second year. ‡ Eight elements. § Eleven elements.

which have been shown to have disproportionately large asthma burdens, with increased hospitalization and mortality rates.6 Even in a population with a reported 27.1% prevalence of “self-reported asthma” in the past 12 months among children 6 to 7 years of age, the school health program does not routinely screen children for asthma.7 There are now adequate data suggesting that delays in diagnosis may influence asthma outcomes.8 For these reasons, it is imperative that there be a simple, standardized, screening tool that can identify children at high risk for asthma in the school setting. Two basic methods have been used historically to assess the prevalence of asthma.9 Most commonly, a written or videotaped questionnaire has been used; less frequently, physiologic measurements of bronchial lability have been performed.10–12 Although the latter method may be more objective, it is more invasive and is difficult to administer to children in large epidemiologic settings. In addition, it exhibits less sensitivity and in some cases less specificity than the questionnaire.13,14 Since the 1950s, written questionnaires have been used to assess the prevalence of asthma.15 Although surveys have improved during the years, problems related to the lack of a universal definition of asthma remain, resulting in previously established physician diagnoses of asthma often underestimating true asthma prevalence.3,6 In addition, specific questions, particularly related to wheezing, may be poorly understood by the parent,16,17 may be difficult to translate into different languages,18 and may address a symptom not present among all children with asthma (ie, cough-variant asthma).19 The ideal detection tool should be noninvasive, simple, easily understood even by less-educated populations, and consistent across cultures.9,12,13 To have good predictive values, it must have good repeatability, sensitivity, and specificity. Validation must be established with an accepted standard method, which is thought to be the approach involving history, physical examination, and spirometry used by asthma specialists.20,21 The International Study of Asthma and Allergy in Childhood questionnaire (Appendix A) (which is e308

primarily a survey tool, rather than a case-detection tool), with both written and videotaped questionnaire formats, has been the most frequently used model using self-reported techniques to determine the prevalence of asthma and allergic disease worldwide and is currently being used in 56 countries.18,22 Derived from the International Union against Lung and Tuberculosis Disease,23 it relies heavily on the symptoms of wheezing for the diagnosis of asthma under a variety of conditions. Furthermore, although several studies of the International Study of Asthma and Allergy in Childhood questionnaire showed adequate sensitivity and specificity, the questionnaire has been evaluated only among children 6 to 7 years and 13 to 14 years of age. Validation methods using physician evaluations, bronchial hyperreactivity assessments, and comparisons with established questionnaires have been variable and less than robust.24–26 A growing number of studies with different validation methods have been performed with pediatric populations.7,14,20,24,25,27–31 Although potentially useful, several have methodologic problems (Table 1). Few randomly selected subjects from the community for validation,27 and only some used validation by an asthma specialist.7,16,20,24,28 Although several were school-based, few questionnaires examined innercity populations.20,27,28,30 Only 2 studies were crosscultural (ie, simultaneous validation in ⬎1 language),28,29 and none, to our knowledge, was validated in the Vietnamese asthma population in the United States. In addition, sensitivities (range: 61-100%) and specificities (range: 55-100%) were variable (Table 2). Only 2 questionnaires14,28 identified those at highest risk according to asthma severity (intermittent or persistent), as established by the new National Institutes of Health (NIH) guidelines.32 These factors limit the applicability of these observations for the general multiracial populations of children in many communities in the United States. Because of these potential limitations, we evaluated the predictive value of a novel asthma questionnaire, developed by Jones et al,28 at school sites for 3 diverse ethnic groups in Orange County, 2 in the inner

CROSS-CULTURAL ASTHMA CASE-DETECTION TOOL

Pediatric Asthma Validation Statistics Across Studies

TABLE 2.

Study

N

Sensitivity, %

Specificity, %

PPV, %

NPV, %

Age Range

29

197 1103, year 1 1179, year 2 129 168 179 157 52 66 262 109 636 401

100 94 96 74 85 98 61 92 89 77 80 91/77/87* 83/56/66*

55 87 96 73 81 69 95 100 100 93 75 88 88

48 55 82 NR 61 74 75 NR NR 63 50 84 68

97 99 100 NR 94 97 NR NR NR 96 92 94 87

6 mo to 18 y 5–12 y 5–12 y 6–12 y 13–14 y 5–6 y 5–15 y 6–7 y 13–14 y 7–11 y 4–13 y 3–18 y 5–11 y

Hall et al Bauer et al27 Wolf et al20 Jenkins and Clarke24 Glasgow et al7 Frank et al31 Sole´ et al25 Joseph et al30 Redline et al16 Jones et al28 Current study

NR indicates not reported. * Persistent asthma/intermittent asthma/overall, with the abbreviated algorithm (yes to question 1 asthma or positive responses to both question 4 and question 6).

city, with low socioeconomic status (SES) (Hispanic and Vietnamese), and 1 in a suburban (predominantly white) setting, for comparison. This questionnaire also attempts to identify children at highest risk for asthma, ie, those with persistent asthma, who have the greatest need for additional evaluation and treatment, as established by the new NIH guidelines.32 METHODS Study Population Three elementary schools reflecting a dominant ethnic population, ie, white (Oak Grove School, 64%), Vietnamese (Morningside School, 54%), and Hispanic (Pio Pico, 100%), were asked to participate after permission was obtained from the school district. Inner-city schools (Morningside and Pio Pico) were so designated on the basis of low SES, as evidenced by reduced-price or free lunches for ⱖ85% of the children and the presence of a large minority population. At each school, classrooms from grades 1, 3, and 5 were selected with random numbers, without previous knowledge of asthma history. An attempt was made to enroll ⬃15% of the student body for the validation studies.

Asthma Detection Survey The asthma detection survey includes 7 questions that address 11 elements of a child’s respiratory health. Each 1-page survey form is printed in English on 1 side and Spanish or Vietnamese on the other (Appendix B). Selected demographic information is requested, and each survey has an individual preprinted identifier number. The first question, which contains 5 elements with yes/no answer options, inquires whether the child has experienced repeated episodes of asthma, cough, chest tightness, trouble breathing, or bronchitis in the past 2 years. This is the only place in the survey that the word asthma is used. Questions 2 through 7 use graded answer options. Wheeze is not used as an isolated symptom, being combined with cough, chest tightness, and trouble breathing under a variety of conditions in questions 2 to 7, for reasons noted above. In questions 2 to 4, information about the frequency of emergency department (ED) visits, school absences, and exercise-related difficulties is requested. The focus of questions 5 through 7 shifts to medication use and symptom (cough, chest tightness, trouble breathing, or wheezing) frequency during the most recent 4-week recall period; the graded answer options correspond to the severity criteria described in the 1997 NIH Guidelines for the Evaluation and Treatment of Asthma.32 The protocol presented was approved by the institutional review board of Children’s Hospital of Orange County.

Study Design Parents of children in the randomly selected classrooms received an introductory letter stating that they would be invited to participate in a study at a mobile clinic at their school. If they

consented, then they would complete a respiratory-asthma questionnaire and their child would be evaluated by an asthma specialist at the school site. To reduce selection bias, no promise of health care was indicated, although there would be an opportunity to enroll in a low-cost health plan, if so desired. For their time and effort, the families were offered a $50 gift certificate for groceries if they participated. Additional, randomly selected classrooms were invited until an optimal number of children at each school were found to participate. Upon arrival at the school, parents were greeted by school personnel and then given the questionnaire in their primary language, with minimal prompting by bilingual speakers (to simulate the home or “real-world” situation). After completing the questionnaire, the family proceeded to a series of stations where vital signs were recorded, spirometry was performed by trained pediatric respiratory therapists, and an asthma-related history was recorded by bilingual nurses. Procedures at the last station for the asthma evaluation, ie, the history recording and physical examination, were performed by a boardcertified allergist with the help of bilingual nurses. The final diagnosis regarding the presence and severity of asthma (either intermittent or persistent asthma) was made by using the NIH guidelines.32 Participating physicians were unaware of the responses to the asthma screening questionnaire, which had been completed previously. The study was conducted between August 2002 and March 2003. All evaluation forms were identified with preprinted numerical indicators, to ensure confidentiality. These numerical indicators were later correlated with the subjects’ names by school personnel, to provide the parents with the conclusions of the survey. To evaluate the reliability of the questionnaire, the survey tool was distributed to the entire student body several weeks later, taken home, and returned for comparison with the initial survey responses determined at the school site.

Statistical Analyses The validity of each survey item as a case-detection tool for asthma was evaluated through calculation of standard parameters (ie, sensitivity, specificity, positive predictive value [PPV], and negative predictive value [NPV]) and application of univariate discriminate analysis procedures. The latter provided the estimated probability of an asthma diagnosis by an allergy specialist on the basis of parent responses to each survey item (negative or positive). A large-scale school screening effort underway in Orange County uses scanning software and case-detection criteria developed and validated in Los Angeles County by Jones et al.28 Validation of this comprehensive (complete) algorithm, which considers previously diagnosed cases and potential undiagnosed cases with symptom experience indicative of asthma, was examined in our study population. Briefly, the scanning software generates 1 of 3 letters for each child, on the basis of prior diagnosis and probability of asthma. Letters reflect predicted asthma status, ie, not likely, possible, or probable asthma. Because children who have 50% to 79% probability receive a “possible” letter, we included this group in our positive category for calculation of validation statistics (negative screen: ⬍50% probability and no prior diagno-

http://www.pediatrics.org/cgi/content/full/114/3/e307

e309

TABLE 3. Distribution of Demographic Characteristics for Each Study Population (Reliability, Validity, and Survey) No. of Patients

Overall Age, y (mean ⫾ SD) (range) Female Male Vietnamese Hispanic Caucasian Other/unknown* Language survey completed among Hispanic participants Spanish English Asthma prevalence Total prevalence Prior diagnosis New diagnosis

Reliability Sample

Validity Sample

193 7.8 ⫾ 1.5 (5–11) 108 (57.1%) 81 (42.9%) 48 (24.9%) 110 (57.0%) 17 (8.8%) 18 (9.3%)

401 7.8 ⫾ 1.5 (5–11) 162 (51.7%) 151 (48.2%) 75 (18.7%) 179 (44.6%) 88 (21.9%) 59 (14.7%)

89 (80.9%) 21 (19.1%)

137 (76.5%) 42 (23.4%) 113/401 (28.2%) 43 (38%)/25 (58%)† 70 (62%)/15 (21%)†

* Number for which ethnicity is unknown within each sample: reliability: N ⫽ 2; validity: N ⫽ 14; survey: N ⫽ 106. † Persistent asthma.

sis; positive screen: ⱖ50% probability and/or prior diagnosis). Figures based on an 80% cutoff point (probable asthma) were also evaluated and are available upon request. Participating school nurses indicated that a subset of 3 questions would be more practical for internal review (case detection by school personnel), which prompted development of an abbreviated scoring algorithm. Results based on the item analysis described above contributed to the development of this abbreviated version. Responses to the 3 most predicative items were considered in combination, and validity estimates were calculated. ␹2 tests were used to assess potential cross-cultural differences in validity estimates for each item, as well as the complete and abbreviated scoring algorithms. This analysis was extended to include examination of potential differences in nonresponse rates for each survey item. Stability reliability of survey items and the complete and abbreviated scoring algorithms was measured through calculation of percent agreement and the ␬ statistic. ␹2 tests were used to evaluate cross-cultural variations in percent agreement. Confidence intervals (95%) surrounding ␬ values for the overall group were computed. Results based on examination of potential cultural differences in ␬ values were assessed but are not reported, because of limited variation and sample size for particular ethnic groups. All analyses were conducted with SPSS version 10.0 (SPSS Inc, Chicago, IL).

RESULTS Demographic Features

Demographic data that describe children included in the reliability and validity analyses are shown in Table 3. Reliability, which was determined by analyzing the survey questionnaire responses obtained several weeks apart, was assessed for 193 children, including 57% Hispanic, 25% Vietnamese, 9% white, and 9% other children. The average age was 7.8 years (range: 5–11 years), and 57% of the children were female. Eighty-one percent of the Hispanic families responded in Spanish, whereas 19% of the Hispanic families responded in English. Children represented in the reliability sample were not excluded from participation in the validity study. The demographic composition of the group of 401 validity study participants was similar to that of the reliability sample (Table 3). A slightly smaller percentage of participants were female and Hispanic, compared with the e310

reliability sample, with greater representation of Vietnamese and white children. Approximately 23% of Hispanic participants responded to the English version of the survey. A similar language determination for the reliability and validity analyses was not possible for the Vietnamese population. The prevalence of asthma in the validation sample of 401 children was 28% (stratified according to ethnicity: Vietnamese: 22.7%; Hispanic: 24.0%; white: 33.0%; other: 40.7%; P ⬍ .05). Among the 113 children with asthma, 65% were classified as having intermittent asthma and 35% as having persistent asthma; 62% were newly diagnosed and 38% were previously diagnosed. Persistent asthma was found in 21% of newly diagnosed cases and 58% of previously diagnosed cases. Abbreviated Algorithm

Univariate discriminant analysis procedures revealed the following 5 items to be the most unique predictive classifiers for asthma status (from most to least predicative): question 1 asthma, question 4, question 6, question 1 trouble breathing, and question 1 chest tightness. These results contributed to development of the abbreviated scoring algorithm. Other considerations included the percentage of responders for each item, the balance of sensitivity and false-positive findings, and the ease of application in a school or pediatric care setting. The abbreviated version classifies a child as a potential asthmatic (requiring additional evaluation for diagnosis of asthma) when responses are positive to question 1 asthma and/or both question 4 and question 6. Therefore, the responses to those questions formed the basis for the abbreviated algorithm. Reliability

The reliability of each survey item scaled according to application as a case-detection tool and the complete and abbreviated scoring algorithms was

CROSS-CULTURAL ASTHMA CASE-DETECTION TOOL

assessed (Table 4). ␬ values indicated acceptable reliability for the majority of items, with substantial agreement being observed for question 1 asthma, question 1 chest tightness, question 3 (missed school), and the complete algorithm (␬ ⬎ .70). Although total percent agreement was slightly lower for question 4 (exercise-induced symptoms) (74.8%), this reliability indicator was ⬎80% for all other items. Reliability findings were consistent across ethnic groups, with the exception of question 1 trouble breathing and question 5 (medicine use). Stability in reporting experience of trouble breathing was slightly lower among Vietnamese participants; however, 80% answered in accordance with their earlier responses. Reported medicine use in the past 4 weeks (question 5) was less reliable among children in the other/unknown category (61.2%), compared with Hispanic (83.1%), Vietnamese (87.5%), and white (94.1%) participants. Validation of Individual Questions

Response rates for particular questions did not show ethnic variation (Table 5). More than 97% of all participants answered questions 2 to 7. Response rates for the health condition items represented in question 1 ranged from 80% to 90%, with a higher rate being observed for question 1 cough (95%). It was interesting to note that, among Hispanic participants, the percentage of nonresponders for specific items was higher among the Spanish-speaking Hispanic subjects than among the English-speaking participants; 21.2% of Spanish-speaking subjects did not respond to question 1 asthma, compared with 4.8% (P ⬍ .05), and 20.4% did not respond to question 1 chest tightness, compared with 7.1% (P ⬍ 05). Item analysis showed consistent patterns of sensitivity and specificity across cultures, although repeated episodes of cough were significantly less specific for Vietnamese and Hispanic participants (45.6% and 39.5%, respectively, compared with white, 69.5%, and other, 67.7%; P ⬍ .001). Predictive values are directly influenced by prevalence. Lower TABLE 4.

specificity for repeated episodes of cough and lower prevalence of asthma among Vietnamese and Hispanic participants likely contributed to the decreased positive predictive ability of this item for these participants. For all other items, significant differences in PPVs and NPVs were not detected at the Bonferroni adjusted level of .001. Algorithm Validity

The validity of the complete and abbreviated algorithms with all ethnic groups is shown in Table 5. Overall, sensitivity was 63% for the complete algorithm and 66% for the abbreviated algorithm. The overall specificity was 87% for the complete algorithm and 88% for the abbreviated version. Algorithm Validity Across Cultures

Two important comparisons are shown in Table 6. There were no statistically significant differences in validity (sensitivity, specificity, PPV, and NPV) among the ethnic groups with either the complete or abbreviated algorithm. In addition, the abbreviated questionnaire compared well with the complete algorithm for all groups. Validity Stratified According to Asthma Severity

In Table 7, it can be seen that asthma severity greatly affected the sensitivity of each survey item and both algorithm scoring methods to detect asthma. Case identification of persistent asthma was consistently greater than that of intermittent asthma. This observed increased sensitivity of detection for persistent asthma was statistically significant with either the complete (P ⫽ .02) or abbreviated (P ⫽ .008) algorithm. Item analysis revealed similar trends, but differences in validity estimates according to severity were less pronounced for the following items: question 1 cough, question 1 bronchitis, question 2 (ED visits), question 3 (missed school), and question 4 (exercise-induced symptom experience).

Reliability of Survey Items and the Complete and Abbreviated Scoring Algorithms Overall and According to Ethnicity

Survey Items*

Q1: asthma Q1: cough Q1: trouble breathing Q1: chest tightness Q1: bronchitis Q2: ED visits Q3: missed school Q4: exercise-induced Q5: medicine use Q6: day symptoms Q7: night symptoms Complete algorithm† Abbreviated version‡

Overall (N ⫽ 193)

Agreement, %

␬ (95% CI)

Agreement, %

Vietnamese (N ⫽ 48)

Hispanic (N ⫽ 110)

Caucasian (N ⫽ 17)

Other (N ⫽ 18)

0.86 (.73–.98) 0.64 (.52–.76) 0.69 (.53–.85) 0.71 (.51–.92) 0.60 (.29–.91) 0.44 (.22–.66) 0.73 (.63–.83) 0.47 (.34–.60) 0.60 (.47–.72) 0.57 (.44–.70) 0.57 (.44–.69) 0.63 0.63 (.50–.76)

96.4 83.1 91.0 94.9 95.5 88.3 87.3 74.8 83.1 81.5 81.0 88.6 86.2

97.4 84.1 80.0 91.2 100 87.2 85.1 62.5 87.5 77.1 77.1 87.5 83.4

97.1 84.3 96.0 96.0 94.3 88.6 87.8 78.9 83.1 83.2 80.2 89.1 87.9

100 76.5 93.8 92.8 85.7 94.2 88.3 70.6 94.1 93.8 94.1 88.2 93.8

86.7 80.0 92.9 100 100 83.3 88.8 88.3 61.2 72.2 83.3 88.9 76.5

Cross-cultural Comparison ␹2

NC 0.8 (P ⫽ .855) 8.4 (P ⫽ .038) NC NC 1.1 (P ⫽ .788) 0.3 (P ⫽ .962) 6.6 (P ⫽ .085) 8.3 (P ⫽ .040) 3.4 (P ⫽ .329) 2.5 (P ⫽ .478) 0.1 (P ⫽ .993) 2.7 (P ⫽ .442)

CI indicates confidence interval; NC; not calculated because of limited variation. * Questions 2 to 7 were coded as ever versus never (relevant to specified time). † Yes to question 1 asthma or symptom experience indicative of asthma (probability of asthma: ⱖ50% or ⱖ80% with more stringent criteria). ‡ Specifically, yes to question 1 asthma or positive responses to both question 4 and question 6.

http://www.pediatrics.org/cgi/content/full/114/3/e307

e311

TABLE 5. Validity of Each Survey Item as a Case-Identification Tool for Distinguishing Asthmatic From Nonasthmatic Children, Compared With the Standard Method of Physician Assessment of Asthma (No/Yes) Survey Items*

Response Rate for Each Item, %

Presence of Condition Among Responders, %†

Sensitivity, %

Specificity, %

PPV, %

NPV, %

Q1: asthma Q1: cough Q1: trouble breathing Q1: chest tightness Q1: bronchitis Q2: ED visits Q3: missed school Q4: exercise-induced Q5: medicine use Q6: day symptoms Q7: night symptoms Complete algorithm¶ Abbreviated version#

85.5 95.0 86.8 83.8 83.3 98.8 99.3 99.3 99.8 99.8 99.8 100 99.0

14.0 59.5 21.8 13.9 9.9 13.1 43.0 42.3 24.1 31.1 31.3 26.9 26.7

50.0 (43/86) 85.7 (90/105) 51.1 (47/92) 36.1 (30/83) 27.7 (23/83) 26.8 (30/112) 69.1 (76/110) 82.0 (91/111) 47.8 (54/113) 62.5 (70/112) 56.6 (64/113) 62.8 (71/113) 65.5 (72/110)

98.1 (252/257) 50.4 (139/276)‡ 88.7 (227/256) 93.3 (236/253) 96.0 (241/251) 92.3 (262/284) 67.0 (193/288) 73.2 (210/287) 85.0 (244/287)㛳 80.9 (233/288) 78.4 (225/287) 87.2 (251/288) 88.2 (253/287)

89.6 (43/48) 39.6 (90/227)‡ 618 (47/76) 638 (30/47)§ 697 (23/33) 57.7 (30/52) 44.4 (76/171) 54.2 (91/168)§ 55.7 (54/97)㛳 560 (70/125) 508 (64/126)㛳 65.7 (71/108)§ 67.9 (72/106)

85.4 (252/295) 90.3 (139/154) 83.5 (227/272) 81.7 (236/289) 80.1 (241/301) 76.2 (262/344)㛳 85.0 (193/227) 91.3 (210/230) 80.5 (244/303) 84.7 (233/275) 82.1 (225/274)㛳 85.7 (251/293) 86.9 (253/291)

Overall, N ⫽ 401; prevalence: 28.2%. * Questions 2 to 7 were coded as ever versus never (relevant to specified time). † Percentage based on those who responded to each question, reflecting a positive answer/classification. ‡ Cross-cultural comparison indicates significant difference with respect to validity parameter (␹2 test, Bonferroni adjusted level), P ⱕ .001. § P ⱕ .01. 㛳 P ⱕ .05. ¶ Yes to question 1 asthma or symptom experience indicative of asthma (probability of asthma ⱖ50% or ⱖ80% with more stringent criteria). # Specifically, yes to question 1 asthma or positive responses to both question 4 and question 6.

TABLE 6.

Validity of Algorithm Across Ethnic Groups

Ethnic Group

No. (Asthma Prevalence)

Algorithm*

Sensitivity, %

Specificity, %

PPV, %

NPV, %

75 (23%)

C A C A C A C A

77 71 54 63 66 71 67 58

88 91 82 84 92 87 97 94

65 71 49 55 79 77 94 88

93 91 85 88 84 87 81 77

Vietnamese Hispanic

179 (24%)

Caucasian

88 (33%)

Other

59 (41%)

* C indicates complete algorithm; A, abbreviated algorithm.

TABLE 7.

Sensitivity According to Asthma Severity Classification

Survey Items*

Intermittent (N ⫽ 73)

Persistent (N ⫽ 40)

Q1: asthma Q1: cough Q1: trouble breathing Q1: chest tightness Q1: bronchitis Q2: ED visits Q3: missed school Q4: exercise-induced Q5: medicine use Q6: day symptoms Q7: night symptoms Complete algorithm† Abbreviated version‡

36.0% (18/50) 82.9% (58/70) 36.8% (21/57) 27.3% (15/55) 22.6% (12/53) 23.6% (17/72) 63.9% (46/72) 77.5% (55/71) 32.9% (24/73) 50.0% (36/72) 42.5% (31/73) 52.1% (38/73) 55.7% (39/70)

69.4% (25/36) 91.4% (32/35) 74.3% (26/35) 53.6% (15/28) 36.7% (11/30) 32.5% (13/40) 78.9% (30/38) 90.0% (36/40) 75.0% (30/40) 85.0% (34/40) 82.5% (33/40) 82.5% (33/40) 82.5% (33/40)

␹2 9.4 1.4 12.1 5.6 1.9 1.0 2.6 2.7 18.4 13.4 16.9 9.0 6.9

(P (P (P (P (P (P (P (P (P (P (P (P (P

⫽ ⫽ ⬍ ⫽ ⫽ ⫽ ⫽ ⫽ ⬍ ⬍ ⬍ ⫽ ⫽

.002) .237) .001) .018) .170) .309) .104) .099) .001) .001) .001) .003) .008)

Overall, N ⫽ 401; prevalence: 28.2%. * Questions 2 to 7 were coded as ever versus never (relevant to specified time). † Yes to question 1 asthma or symptom experience indicative of asthma (probability of asthma: ⱖ50% or ⱖ80% with more stringent criteria). ‡ Specifically, yes to question 1 asthma or positive responses to both question 4 and question 6.

DISCUSSION

This study carefully evaluated the question of whether a novel asthma case-detection tool could detect children at high risk for asthma in several e312

cultures. Our data suggest that this is the case. The tool, which was described recently by Jones et al,28 has worked well in identifying at-risk children 5 to 11 years of age (mean: 7.8 years) in both inner-city

CROSS-CULTURAL ASTHMA CASE-DETECTION TOOL

and suburban school sites, regardless of Hispanic, white, or Vietnamese ethnicity. This survey tool had an overall sensitivity of 83% and specificity of 88% (with the abbreviated algorithm) for children at greatest risk, with persistent asthma diagnosed by the asthma specialist during the validation procedure. Sensitivity was lower (56%) for those with less risk, with intermittent asthma. In terms of ethnicity, the prevalence across cultures of physician-diagnosed persistent asthma was similar for the Hispanic group (9%) and the white group (10%) and was slightly higher (12%) for the Vietnamese and other groups (data not shown). Validation fared well in all groups, in terms of sensitivity, specificity, PPV, and NPV. Although only trends could be analyzed, because the number in each group was small, there did not appear to be large differences, although the sensitivity of the algorithm was somewhat lower for the Hispanic group (Table 6). The patterns of response rates for question 1 through question 7, as well as the repeatability of each question, were not significantly different among the 3 ethnic groups. However, in the Hispanic population, nonresponses to certain questions, such as question 1 asthma and question 1 chest tightness, were fewer among subjects who answered in English (Table 3). This could reflect better SES and educational background for the English-speaking Hispanic families. Because physician diagnosis of asthma is known to vastly underestimate its presence, particularly among minorities in inner-city environments, it was of great interest to evaluate the usefulness of an abbreviated algorithm, particularly for those with negative responses to question 1 asthma (because they stated no or did not answer the question). In comparison with question 1 asthma alone, we found that using a combination of questions in the algorithm analysis, particularly combinations of question 1 asthma and question 4 (exercise-induced asthma) and question 6 (daytime asthmatic symptoms), resulted in excellent sensitivity of 83% and specificity of 88% for those with the highest risk, with persistent asthma (Table 7). This abbreviated algorithm usually yielded results as good as or better than those of the complete 7-element algorithm (Table 6). Unique to this questionnaire was the incorporation of wheezing with other symptoms, including cough, trouble breathing, and chest tightness, instead of using wheezing as a “stand-alone” diagnostic tool, for reasons given earlier.16–19 In addition, this approach is consistent with the observation that the most frequent presenting symptoms of asthma among children are cough (92%), wheeze (90%), and shortness of breath (83%).33 Redline et al16 recently reported that cough and breathing problems were more useful criteria than was wheezing. However, because wheeze is such a common stand-alone criterion in many surveys, we analyzed the value of wheeze as a stand-alone criterion and evaluated its effect in adding it to the abbreviated algorithm. Because the question was used only by the physician evaluator and could include prompting, the result is only an estimate of its value. For wheeze as a stand-alone crite-

rion, the overall sensitivity was only 53% (63% among persistent asthmatics), whereas the specificity was high (88%). This is very similar to the results for question 1 asthma as a stand-alone criterion, with a sensitivity of 50% and a specificity of 98%. Addition of wheezing to the abbreviated algorithm resulted in better overall sensitivity of 74%, compared with 66%, and specificity of 67%, compared with 56%, for the intermittent asthma group. Addition had very little effect on the sensitivity or specificity for the persistent group. We plan to examine the value of wheezing as a stand-alone criterion more directly in future studies. Validation of the asthma detection tool has been a major problem in previously reported asthma studies (Tables 1 and 2). For adequate validation, the population should be sampled in the environment in which the tool will be used, with random selection from the community and not sampling in a clinic or hospital, where patient bias might be introduced. The questions selected for validation should not be judged (a priori) to indicate a positive or negative responder and should not be used as selection criteria for the validation process, and the questionnaire should be compared with the best standard available.9 The standard method has varied considerably in different studies and has included previously validated questionnaires,26 bronchial lability testing including exercise and bronchial constriction challenge,23,24,26 and evaluation by nonphysicians27 and by a panel of experts who did not directly examine the subjects.31 Assessment of bronchial lability for diagnosing clinical asthma has not been found to be a predictable tool14,24,26 and is therefore inadequate for validating an asthma detection tool, such as the questionnaire. The consensus is that the best standard is the asthma specialist who, in a blinded manner, examines the study population.20,21 Our study model did attempt to address these critical issues. We validated an asthma detection tool in 3 different ethnic cultures, in both inner-city and suburban settings, and made an effort to select patients randomly from the community, using schools as our major sites. There was no a priori assumption of a positive questionnaire, and we used an asthma specialist to establish a diagnosis for all children at the time of the study. Selection bias could not be totally avoided, because 41% of all invited patients participated. The small financial incentive could have resulted in recruitment of more financially needy patients, among whom asthma is thought to be more severe. Another potential limitation of this study was the assignment of severity to each case by the asthma specialist. This difficulty was illustrated by our finding that the asthma severity did not directly affect the sensitivity of detection based on ED visits, school absenteeism, and exercise-induced symptoms, although severity did correlate with most of the other parameters (Table 7). This calls into question what should be the standard method. Braganza et al34 found that pediatricians often underdiagnose the severity of asthma, according to NIH guidelines. Furthermore, Baker et

http://www.pediatrics.org/cgi/content/full/114/3/e307

e313

al35 reported major discrepancies regarding asthma severity even among asthma specialists. Some variability among the 9 asthma specialists participating in our study was thus expected. Nevertheless, asthma specialists have been found to be more likely to follow NIH guidelines than generalists and to be more effective in the validation process.36 Even with these limitations, there does not appear to be any substantial difference in this prediction model, in terms of sensitivity or specificity, among the ethnic groups, suggesting that our detection tool could be used universally for the cultures evaluated. Although our inner-city population may differ from the typical urban inner-city population, Jones et al28 found similar results for African American and Hispanic children evaluated in Los Angeles. Although we have validated a survey tool that detects children at high risk for persistent asthma, there is concern that this may not be beneficial without adequate medical follow-up services.14 A recent publication by the Department of Health and Human Services suggested identifying only children with a previous diagnosis of asthma and avoiding mass screening or case detection, “since these methods have not been shown to meet the World Health Organization or the American Academy of Pediatrics’ criteria for population or school screening programs.”37(p2) This conclusion was based mainly on several recent reports. Gerald et al,38 using a multistep approach (questionnaire, spirometry, and exercise provocation) among children with a previous diagnosis of asthma or suspected asthma, concluded that additional physiologic tests are necessary to reduce the overdiagnosis made with the questionnaire. However, spirometry and provocation testing are not practical or safe in the school setting, and neither is very specific or sensitive.14,24,26 Boss et al,39 using 9 World Health Organization criteria for determining the appropriateness of community-based screening programs, concluded that a number of criteria are not met for asthma detection in this setting. However, several key assumptions require clarification. Those authors suggested that children diagnosed with mild asthma uniformly have good prognoses. They asserted that treatment does not modify the underlying disease process, even among those with moderate disease, and thus early detection in the school setting has little impact on prevention among children ⬎5 years of age. However, it has been reported that as many children with clinical manifestations of mild disease die as those who have more serious disease.40 As shown in our data (Table 7) and reported by others, 41 children with mild disease experienced ED visits and hospitalizations. Even among asthma specialists, there is disagreement regarding the classification of asthma severity,35 which suggests that current methods of assessing disease severity might be inadequate. Furthermore, those with moderate asthma disease do experience profound changes in ED visits and hospitalizations when appropriately treated.41 Although there is no conclusive evidence that the natural history of the disease is changed, there is evidence e314

suggesting that pulmonary function might be permanently lost if antiinflammatory treatment is not initiated within 2 years after diagnosis.8 Boss et al39 also suggested that previously undiagnosed asthma is usually mild. However, Clark et al14 reported similar asthma severity for previously undiagnosed (39%) and previously diagnosed (55%) cases, whereas we found that 21% of previously undiagnosed cases involved persistent asthma. Although the cost/benefit ratio is always a concern for subjects identified at school, costs for inner-city children with asthma might actually be less, rather than more, if patients were given education and appropriately treated, because of fewer ED visits and hospitalizations.42 Yawn et al43,44 suggested that children who are previously undiagnosed and those with the diagnosis of asthma in the school setting infrequently seek medical care recommended by the school, to confirm the diagnosis and receive appropriate therapy. Those authors concluded that a school detection tool might be worthwhile for those previously diagnosed with asthma but not for those previously undiagnosed, in terms of accessing physician care. However, it could also be concluded that, if education and accessibility to better care were available for both suspected and previously diagnosed children, then better physician-based diagnosis and treatment would have been found. Although we agree in principal with the assertion of Clark et al14 that school-based detection of children with asthma would be useful only if effective, affordable, and accessible medical services were available, we think that school-based detection is the first step. The challenge is to provide good ongoing medical care, particularly in inner-city settings, where it is needed most. Several programs, in fact, have begun to reduce the barriers to good care in inner-city environments, with promising results.45–48 CONCLUSION

We found this asthma detection tool to be suitable for detecting children at high risk for persistent asthma across the dominant cultures in Orange County. The school setting seems to be well suited for detecting these children, with the enthusiastic support of school personnel, particularly for the inner-city minority population. An abbreviated modification of the complete 11-element questionnaire with only 3 questions, ie, question 1 asthma, question 4 (exercise-induced asthma symptoms), and question 6 (asthma symptoms during the day in the past 4 weeks), which yielded excellent prediction of persistent asthma, might be more suitable for inclusion in the health survey that all children are required to submit for school. Although there is concern that subjects detected in the school setting might not have adequate access to appropriate care, case detection is an important first step that could lead to earlier diagnosis and treatment of asthmatic children who are most in need. Reducing the barriers to good ongoing medical care in inner-city environments is the next step.

CROSS-CULTURAL ASTHMA CASE-DETECTION TOOL

APPENDIX A: INTERNATIONAL STUDY OF ASTHMA AND ALLERGY IN CHILDHOOD QUESTIONNAIRE

Core questionnaire wheezing module for 6 to 7-year-olds and 13 to 14 year olds 1. Has your child ever had wheezing or whistling in the chest at any time in the past? Yes [ ] No [ ] 2. Has your child had wheezing or whistling in the chest in the last 12 months? Yes [ ] No [ ] If you answered “no,” please skip to question 6 3. How many attacks of wheezing has your child has in the last 12 months? None [ ] 1 to 3 [ ] 4 to 12 [ ] More than 12 [ ] 4. In the last 12 months, how often, on an average, has your child’s sleep been disturbed due to wheezing? Never woken with wheezing [ ] Less than one night per week [ ] One or more nights per week [ ] 5. In the last 12 months, has wheezing ever been severe enough to limit your child’s speech to only one or two words at a time between breaths? Yes [ ] No [ ] 6. Has your child ever had asthma? Yes [ ] No [ ] 7. In the last 12 months, has your child’s chest sounded wheezy during or after exercise? Yes [ ] No [ ] 8. In the last 12 months, has your child had a dry cough at night, apart from a cough associated with a cold or a chest infection? Yes [ ] No [ ]

http://www.pediatrics.org/cgi/content/full/114/3/e307

e315

APPENDIX B: ASTHMA HEALTH SCREENING SURVEY ACKNOWLEDGMENTS This study was made possible by the support of the Ahmanson Foundation, Children’s Hospital of Orange County Foundation, and GlaxoSmithKline Pharmaceutical Company. We thank the board-certified allergists who participated in the project, namely, Sherwin Gilman, Leo Cummins, Steven Weinstein, William Berger, Mark Ellis, Chao-I Lin, Charles Groncey, Barry Josephson, and Donald Gjesdal. We express our appreciation to the school nurses and other school personnel of the Orange County Unified School System, without whom this study would not have been possible. Thanks also go to Rhonda Robles, Alma Carrasco-Gomez, and Marla Nathan for manuscript preparation.

REFERENCES 1. Centers for Disease Control and Prevention. Surveillance for asthma: United States, 1960 –1995. MMWR Surveill Summ. 1998;47:1–27 2. Grant EN, Daugherty SR, Moy JM. Prevalence and burden of illness for asthma and related symptoms among kindergarteners in Chicago public schools. Ann Allergy Asthma Immunol. 1999;83:113–120 3. Joseph CLM, Foxman B, Leickly F, et al. Prevalence of possible undiagnosed asthma and associated morbidity among urban school children. J Pediatr. 1996;129:735–742 4. Perien JM, Homer CJ, Berwick DM, et al. Variations in rates of hospitalization of children in three urban communities. N Engl J Med. 1989; 320:1183–1187 5. Weiss KB, Gergen PJ, Hodgson T. An economic evaluation of asthma in the U.S. N Engl J Med. 1992;326:862– 866 6. Akinbami LJ, Schoendorf KC. Trends in childhood asthma: prevalence, healthcare utilization and mortality. Pediatrics. 2002;810:315–322 7. Glasgow NJ, Ponsonky AL, Yates RE, et al. Asthma screening as part of a routine school health assessment in the Australian Territory. Med J Aust. 2001;174:384 –388 8. Agertoft L, Pedersen S. Effects of long term treatment with an inhaled corticosteroid in growth and pulmonary function in asthmatic children. J Respir Med. 1994;88:373–381 9. Burney P, Chinn S. Developing a new questionnaire for measuring the prevalence and distribution of asthma. Chest. 1997;91(suppl 6):79S– 83S 10. Burr ML. Diagnosing asthma questionnaire in epidemiological surveys. Clin Exp Allergy. 1992;22:509 –510 11. Sockrider MM, Tortolero SR, Bartholemew LK, et al. Pilot study of a screening questionnaire for asthma. Pediatr Asthma Allergy Immunol. 2001;15:15–24 12. Woolcock AJ. Epidemiologic methods for measuring prevalence of asthma. Chest. 1987;91(suppl):89S–92S 13. Pearce N, Pekkanen J, Beasley R. Role of bronchial responsive testing in asthma prevalence surveys. Thorax. 2000;55:352–354 14. Clark NM, Brown R, Joseph CL, et al. Issues in identifying asthma and estimating prevalence in an urban school population. J Clin Epidemiol. 2002;55:870 – 881 15. Blanc ADJ. So You Have Asthma? Springfield, IL: Charles C Thomas; 1966 16. Redline S, Larkin EK, Kercsmar C, et al. Development and validation of school-based asthma and allergy screening instruments for parents and students. Ann Allergy Asthma Immunol. 2003;90:516 –528 17. Cane RS, Ranganathan SC, McKenzie SA. What do parents of wheezy children understand by wheeze? Arch Dis Child. 2000;82:327–332 18. Steering Committee. Worldwide variations in the prevalence of asthma symptoms: the International Study of Asthma and Allergies in Childhood (ISAAC). Eur Respir J. 1998;12:315–335 19. Kelly YJ, Brakin BJ, Milligan PJ, et al. Clinical signs of cough and wheeze in the diagnosis of asthma. Arch Dis Child. 1996;75:489 – 493 20. Wolf RL, Berry CA, Quinn K. Development and validation of a brief pediatric screen for asthma and allergies among children. Ann Allergy Asthma Immunol. 2003;90:500 –507 21. Britton J, Lewis L. Objective measures for the diagnosis of asthma. BMJ. 1998;317:227–228 22. Asher MI, Anderson HR, Beasley R, et al. International Study of Asthma and Allergies in Childhood (ISAAC): rationale and methods. Eur Respir J. 1995;8:483– 491

e316

23. Burney PGJ, Laitenen LA, Perdrizet S, et al. Validity and repeatability of the IUALTD (1984) Bronchial Symptoms Questionnaire: an international comparison. Eur Respir J. 1989;2:940 –945 24. Jenkins MA, Clarke JR. Validation of a questionnaire and bronchial hyper-responsiveness against respiratory physician assessment in the diagnosis of asthma. Int J Epidemiol. 1996;25:609 – 616 25. Sole´ D, Vanna KT, Yamada E, et al. International Study of Asthma and Allergy in Childhood (ISAAC) written questionnaire: validation of the asthma component among Brazilian children. J Investig Allergol Clin Immunol. 1998;8:376 –382 26. Shaw R, Woodman K, Ayson M, et al. Measuring the prevalence of bronchial hyper responsiveness in children. Int J Epidemiol. 1995;24: 597– 602 27. Bauer EJ, Lurie N, Yeh C, et al. Screening for asthma in an “inner-city” elementary school in Minneapolis, MN. J Sch Health. 1999;69:2–16 28. Jones CA, Morphew T, Dyer M, et al. A school based identification process for identifying “inner-city” children with asthma: the Breathmobile Program. Chest. 2004;125:924 –934 29. Hall CB, Wakefield D, Rowe TM, et al. Diagnosing pediatric asthma: validating the Easy Breathing Survey. J Pediatr. 2001;139:267–272 30. Joseph CLM, Foxman B, Leicky FE, et al. Sensitivity and specificity of asthma definition and symptoms using a survey of childhood asthma. J Asthma. 1999;36:565–573 31. Frank TL, Frank PI, McNamee R, et al. Assessment of a simple scoring system applied to a screening questionnaire for children ages 5–15 years. Eur Respir J. 1999;14:1190 –1197 32. National Asthma Education and Prevention Program. Expert Panel Report: Guidelines for the Diagnosis and Management of Asthma Update on Selected Topics: 2002. J Allergy Clin Immunol. 2002;110(suppl): S141–S219 33. Parks DB, Ahrens RC, Humphrey CT, et al. Chronic cough in childhood: approach to diagnosis and treatment. J Pediatr. 1989;115:856 – 862 34. Braganza S, Sharif I, Ozuah PO. Documenting asthma severity: do we get it right? J Asthma. 2003;40:661– 665 35. Baker KM, Brand DA, Hen J. Classifying asthma: disagreement among specialists. Chest. 2003;124:2156 –2163 36. Diette JB, Skinner EA, Nguyen TTH, et al. Comparison of quality of care by specialists and generalist physician as usual source of asthma care for children. Pediatrics. 2001;108:432– 437 37. Department of Health and Human Services. Strategies for Addressing Asthma Within a Coordinated School Health Program. Atlanta, GA: Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion; 2002 38. Gerald LB, Redden D, Turner-Henson A, et al. A multi-stage asthma screening procedure for elementary school children. J Asthma. 2002;39: 29 –36 39. Boss LP, Wheeler LS, Williams PV, et al. Population-based screening or case detection for asthma: are we ready? J Asthma. 2003;40:335–342 40. Robertson CF, Rubinfeld AR, Bowes G. Pediatric asthma deaths in Victoria: the mild are at risk. Pediatr Pulmonol. 1992;13:95–100 41. Adams RJ, Fuhlbrigge A, Finkelstein JA, et al. Impact of inhaled antiinflammatory therapy and hospitalization in emergency room department visits for children with asthma. Pediatrics. 2001;107:706 –711 42. Clark NM, Feldman CH, Evans D, et al. The impact of health education on frequency and cost of health care used by low income children with asthma. J Allergy Clin Immunol. 1986;78:108 –115 43. Yawn VP, Wollan P, Scanlon P, et al. Are we ready for universal school-based asthma screening? An outcomes evaluation. Arch Pediatr Adolesc Med. 2000;156:1256 –1262 44. Yawn VP, Wollan P, Scanlon P, et al. Outcome results of a school-based screening program for under-treated asthma. Ann Allergy Asthma Immunol. 2003;90:508 –515 45. Brito A, Wurm G, Delamater AM. School based identification of asthma in a low income population. Pediatr Pulmonol. 2000;30:297–301 46. Adams EK, Johnson V. An elementary school-based health clinic: can it reduce Medicaid costs? Pediatrics. 2000;105:770 –788 47. Kelly CS, Morrow AL, Shelts J, et al. Outcomes evaluation of a comprehensive intervention program for asthmatic children enrolled in Medicaid. Pediatrics. 2000;105:1029 –1035 48. Federico MJ, Liu AH. Overcoming childhood asthma disparities of the inner-city poor. Pediatr Clin N Am. 2003;50:655– 675

CROSS-CULTURAL ASTHMA CASE-DETECTION TOOL