Case-control study of hospital admission withasthma in children aged 5–14 years: relation withroad traYc in north west London

P Wilkinson, P Elliott, C Grundy, G Shaddick, B Thakrar, P Walls, S Falconer

AbstractBackground—Evidence for an associationbetween road traYc pollution and asthmais inconclusive. We report a case-controlstudy of hospital admissions for asthmaand respiratory illness among childrenaged 5–14 in relation to proxy markers oftraYc related pollution.Methods—The study was based on routinehospital admissions data in 1992/3 and1993/4 for North Thames (West) healthregion within the M25 motorway. Caseswere defined as emergency admissions forasthma (n = 1380) or all respiratory illnessincluding asthma (n = 2131), and controls(n = 5703) were other emergency admis-sions excluding accidents. Cases and con-trols were compared with respect todistance of residence from nearest mainroad or roads with peak hour traYc >1000vehicles and traYc volume within 150 mof residence, obtained by GeographicalInformation System techniques. Statisti-cal analysis included adjustment for age,sex, admitting hospital, and a deprivationscore for the census enumeration districtof residence.Results—Adjusted odds ratios of hospitaladmission for asthma and respiratoryillness for children living within 150 m of amain road compared with those livingfurther away were, respectively, 0.93 (95%CI 0.82 to 1.06) and 1.02 (95% CI 0.92 to1.14).Conclusions—This study showed no as-sociation between risk of hospital admis-sion for asthma or respiratory illnessamong children aged 5–14 and proxymarkers of road traYc pollution.(Thorax 1999;54:1070–1074)

Keywords: childhood asthma; road traYc pollution

A number of studies have reported on theassociation between asthma or wheeze symp-toms and road traYc. Positive findings havebeen reported between traYc volume and self-reported wheezing in school children,1 2 be-tween proximity to major roads and prevalenceof respiratory symptoms in children andadults,3–7 and between modelled estimates ofair pollution from traYc and respiratorysymptoms.8 9 An excess of hospital admissionsor symptom reporting near roads, if reflecting acausal association, could either be the result ofexacerbations among susceptible individuals oran increase in the prevalence of the condition.

In the UK, whereas a study in Birminghamshowed a positive association between roadtraYc and hospital admission with asthma inchildren,10 Livingstone et al11 found no associ-ation of asthma diagnosis and treatment withproximity to major roads in London, andWaldron et al12 found a lower prevalence ofself-reported wheeze in children living in elec-toral wards traversed by motorways. In 1995 areport by the Committee on the Health Effectsof Outdoor Air Pollution13 concluded thatavailable evidence does not “support a causa-tive role for (non-biological) outdoor pollu-tion” in the initiation of asthma, but recom-mended further analyses of “health serviceutilization in urban areas . . . in relation to localtraYc density . . .”. We report here such a studyamong children living in north west Londonwhich examines whether the risk of hospitaladmission for asthma is higher in thosechildren living near major sources of road traf-fic emissions.

Figure 1 Identification of study population amonghospital admissions in North Thames (West) health region(numbers of exclusions in parentheses). *Admission forwhich the primary diagnosis was missing/invalid, age wasinconsistent with dates, or first episode of current spell endedbefore study period.

68 365

36 458

33 622

15 502

admissions

11 723admissions

9535

7834

children

Respiratory cases(ICD-9: 460–496)

2131

Asthma cases(ICD-9: 493)

1380

Non-respiratorycontrols

5703

Outside M25(31 907)

Exclusions*(2836)

Non-emergencyadmissions(18 120)

Excludedhospitals(1701)

Accidents(3779)

Thorax 1999;54:1070–10741070

EnvironmentalEpidemiology Unit,Department of PublicHealth and Policy,London School ofHygiene and TropicalMedicine, LondonWC1E 7HT, UKP WilkinsonC GrundyB ThakrarP Walls

Department ofEpidemiology andPublic Health,Imperial CollegeSchool of Medicine, StMary’s Campus,London W2 1PG, UKP ElliottG ShaddickS Falconer

Correspondence to:Professor P Elliott

Received 23 July 1996Returned to author15 October 1996Revised manuscript received19 August 1999Accepted for publication6 September 1999

on May 18, 2022 by guest. P

rotected by copyright.http://thorax.bm

j.com/

Thorax: first published as 10.1136/thx.54.12.1070 on 1 D

ecember 1999. D

ownloaded from

MethodsThis was a hospital based case-control studyusing routine postcoded regional Hospital Epi-sode Statistics for children aged 5–14 yearsliving in the North Thames (West) healthregion within the boundary of London’s orbitalM25 motorway. It was carried out for twoyears, April 1992 to March 1994. The HospitalEpisode Statistics for England allocate eachadmission to the postcode of residence regard-less of where that admission occurred nation-ally. The co-ordinates of the place of residencewere determined from the centroid of the resi-dential postcode (±10 m) using the OrdnanceSurvey’s Address Point. Separate admissionswithin the same individual were linked bymatching date of birth, sex, and postcode.Analyses were also done for all admissions—that is, including individuals admitted morethan once during the study period. Two casegroups, asthma (9th revision of InternationalClassification of Diseases (ICD-9) 493) andrespiratory illness (ICD-9 460–496), and con-trols were identified among emergency admis-sions using primary diagnosis for the first con-

sultant episode of each admission (fig 1). Eachcase required at least one admission during thestudy period with a primary diagnosis ofasthma or respiratory illness. Controls com-prised all other children with an emergencyadmission during the study period (but not forrespiratory illness) with the exclusion ofadmissions for accidental injuries and poison-ing (ICD-9 800–999), the latter being ex-cluded because of possible association betweenroad accidents and residence close to a mainroad. Cases and controls from hospitals admit-ting five or fewer cases of asthma wereexcluded. The primary diagnosis at firstadmission for the control group is shown in fig2.

ROADS AND TRAFFIC VOLUMES

Data giving location of the centre lines of allroads within the study area (±1 m) wereobtained from the Ordnance Survey. TraYcdata were derived from the London ResearchCentre’s road traYc model for London, whichcomprises a set of links between road junctionswith modelled peak hour flows (vehicles/hour)along interjunctional segments of all majorthrough roads.14

Linking of the traYc model to OrdnanceSurvey data was carried out using a Geographi-cal Information System. Once the links werecreated, all main roads (motorways, primaryroads, A roads, B roads) and, in centralLondon, also many minor roads were assignedestimated traYc volumes. For residentialstreets, which generally have low traYc vol-umes, the Department of Transport 1991 traf-fic report was used to assign them an averagepeak hour traYc flow. One of three diVerentaverages was applied depending on whether thestreet was located in outer, inner, or centralLondon.

Measures of exposure to traYc were: (1)simple Euclidean distance to nearest mainroad; (2) distance to nearest road with a mod-elled peak hour traYc volume exceeding 1000vehicles/hour; (3) computed traYc volume(vehicle-metres/hour) along roads within aradius of 150 m of the postcode centroid.Derivation of the latter variable entailed a twostage process: (1) conversion of the roadnetwork model into a raster grid of 10 m ×10 m squares, each with an estimated peakhour flow, and (2) for each case and control the10 m × 10 m cells were selected within thespecified distances and the sum obtained oftheir individual traYc flow values (fig 3). Traf-fic volumes within the diVerent radii were cat-egorised into quartiles but with the cut oVpoints rounded to the nearest convenientnumber.

STATISTICAL METHODS

The main prior hypothesis related to the risk ofhospital admission within 150 m of a mainroad. Simple and multiple logistic regressionwas used to examine the association betweenhospital admission for asthma/respiratory ill-ness and the various markers of traYcexposure. Analyses were carried out for boysand girls separately and combined, but only the

Figure 2 Primary diagnosis at first admission of control group: 1 = digestive systemdisorder; 2 = infection; 3 = musculoskeletal disorder; 4 = nervous system disorder; 5 =genitourinary disorder; 6 = disorder of the blood forming organs; 7 = disorder of the skinand subcutaneous tissue; 8 = endocrine or nutritional disorder; 9 = mental disorder; 10 =neoplasm; 11 = ill defined symptoms and signs; 12 = other.

1500

1000

500

01 2 3 4 5 6 7 8 9 10 11 12

Diagnostic group

Freq

uen

cy

Figure 3 Exposure classification of postcode locations: traYc volumes were computed alongroad segments within 150 m of each point (10 × 10 pixels) giving the grey scale rangingfrom zero (black) to >50 000 vehicle-metres/hour (white). © Crown & Ordnance Survey.

Vehicles per hour

< 250250–500500–1000> 1000

Postcodecentroid

0.5km

Relation between hospital admission with asthma and road traYc 1071

on May 18, 2022 by guest. P

rotected by copyright.http://thorax.bm

j.com/

Thorax: first published as 10.1136/thx.54.12.1070 on 1 D

ecember 1999. D

ownloaded from

results of the combined analyses are tabulated.Odds ratios are given together with 95% confi-dence intervals. The logistic model assumesthat risk falls exponentially with distance;further analysis using the non-linear binaryregression model of Diggle and Rowlingson15

was done which, though similar to the logisticmodel, has the advantage that risk at distance isallowed to converge to background. For bothmethods, tests for improvements in model fitwere based on comparisons of likelihood ratios.We estimated that the study had approximately90% power to detect a diVerence of 20% in therisk of asthma admission between children liv-ing within 150 m of a main road and those liv-ing further away.

Adjustment for hospital in the regressionanalyses allowed for possible variation betweenhospitals in completeness of data and diagnos-tic coding. To control for possible socioeco-nomic confounding, each child was assigned aCarstairs deprivation score for the census enu-meration district in which he or she lived.16

This is based on four variables from the 1991census small area statistics: overcrowding,social class, unemployment, and access to a car.Twenty one children lived in enumeration dis-tricts with insuYcient population to compute aCarstairs score and were excluded from analy-sis where socioeconomic adjustment was re-quired.

ResultsAfter exclusions, data were available on 1380cases of asthma, 2131 cases of respiratoryillnesses (which includes the asthma cases),and 5703 non-respiratory controls (fig 1 andtable 1). Eighty four percent of the childrenhad been admitted to hospital only once duringthe study period. Asthma and respiratory casestended to be younger than the controls with ahigher proportion of boys. Median deprivationscores, distributions of distance from a mainroad, and estimated traYc volumes within150 m were all similar for cases and controls(table 1).

There was no association between residencewithin 150 m of a main road and risk of hospi-tal admission for asthma (table 2). Thecorresponding odds ratios for respiratoryadmission were 1.02 (95% CI 0.92 to 1.13) forthe unadjusted analysis and 1.02 (95% CI 0.92to 1.14) after adjustment for age, sex, admit-ting hospital, and deprivation. There was nosignificant interaction between the eVect ofhospital and distance from roads. Similarresults were obtained for distance to a roadwith a peak hour traYc volume exceeding 1000vehicles, and when distance was treated as acontinuous rather than a dichotomous variable(data not tabulated). For example, in the fullyadjusted analysis, the odds ratio for asthmaadmission for each 1000 m nearer to a mainroad was 0.96 (95% CI 0.75 to 1.23). Testsbased on the method of Diggle and Rowlingsongave concordant results (p = 0.9 for the eVectof distance from a main road on risk of asthmaadmission in the fully adjusted model).

Odds ratios for hospital admission withasthma by traYc volume within 150 m of thehome (table 3) suggested some heterogeneityof risk across exposure categories in the age,sex, hospital, and deprivation adjusted model,and there was a borderline significant trend (p= 0.06) of decreasing risk with increasing traYcvolumes—that is, in a direction opposite to thathypothesised. Results for hospital admissionwith respiratory illness (not tabulated) show asimilar, but not statistically significant, pattern.Analyses based on all admissions—that is,allowing more than one admission per child—yielded results very similar to those of table 3,with evidence of a decline in risk of hospitaladmission with asthma at postcodes of highertraYc volume (p = 0.02 for trend before and

Table 1 Characteristics of study population

Asthma cases(n = 1380)

Respiratory cases(n = 2131)

Controls(n = 5703)

Mean (SD) age (years) 8.4 (2.8) 8.4 (2.9) 9.5 (2.9)Percent male 63.3% 60.4% 55.9%Median Carstairs deprivation score (IQ range) 0.25 (−1.64 to 2.29) 0.19 (−1.66 to 2.16) 0.03 (−1.84 to 1.95)Distance in metres from major road (%)

0–74 16.6% 16.9% 16.5%75–149 21.2% 22.9% 22.8%150–299 28.5% 26.9% 27.2%300+ 33.7% 33.3% 33.5%

Median distance from main road (IQ range) 201.8 (105.1 to 370.7) 195.9 (100.1 to 373.9) 196.2 (102.6 to 378.5)TraYc volume within 150 m (vehicle-metres/h)

<1500 28.7% 28.8% 27.5%1500–14 999 25.7% 24.0% 23.4%15 000–49 999 24.3% 24.7% 27.3%50 000+ 21.2% 22.5% 21.8%

Table 2 Numbers (%) of children with at least one hospital admission for asthma bydistance of residence from main road

Cases of asthma Non-respiratory controls

Living within 150 m of main road 522 (37.8%) 2240 (39.3%)Living >150 m from main road 858 (62.2%) 3463 (60.7%)

Odds ratios† (95% confidence intervals): unadjusted = 0.94 (0.83 to 1.06), p = 0.32; adjusted‡ =0.93 (0.82 to 1.06), p = 0.27.†Odds of hospital admission with asthma in children living within 150 m of main road comparedwith odds for those living further away.‡Adjusted for age, sex, hospital, and deprivation.

Table 3 Odds ratios with 95% confidence intervals for hospital admission for asthma bytraYc volume within 150 m of place of residence

Volume within 150 m UnadjustedAdjusted for age, sex, hospitaland deprivation

<1500 1.0 1.01500–14 999 1.06 (0.90 to 1.24) 1.03 (0.87 to 1.22)15 000–49 999 0.86 (0.73 to 1.00) 0.80 (0.68 to 0.95)50 000+ 0.93 (0.79 to 1.11) 0.88 (0.74 to 1.06)Statistical tests:

Heterogeneity (3 df) ÷2 = 6.83; p = 0.08 ÷2 = 10.4; p = 0.02Trend (1 df) ÷2 = 2.38; p = 0.12 ÷2 = 3.58; p = 0.06

1072 Wilkinson, Elliott, Grundy, et al

on May 18, 2022 by guest. P

rotected by copyright.http://thorax.bm

j.com/

Thorax: first published as 10.1136/thx.54.12.1070 on 1 D

ecember 1999. D

ownloaded from

after adjustment for age, sex, hospital, andsocioeconomic deprivation).

DiscussionThis study of 7834 children from north westLondon showed no association between risk ofhospital admission with asthma or respiratoryillness and proxies for traYc related pollutionat the place of residence. This accords withresults of a recent study of general practitionerdiagnosis of, and treatment for, asthma in rela-tion to proximity of residence to busy roads ineast London, but contrasts with the findings ofother studies.1–10

In common with all but two of thesestudies,8 9 we used a proxy for exposure to roadtraYc based on proximity to a major road orestimated traYc volumes rather than modelledestimates of ambient air pollution. Such studiestake no account of vehicle type, periods ofacceleration and deceleration, and vehiclespeed, all of which aVect emission levels.Furthermore, ambient pollution concentra-tions depend on atmospheric chemistry, dis-persion processes and meteorological condi-tions, as well as emissions. Whereas it seemsprobable that proximity to a main road is a rea-sonable surrogate for the outdoor concentra-tions of most traYc related pollutants, for oth-ers such as ozone it will be a poor markerbecause photochemical oxidation leads toreduced levels near traYc sources.17

In none of the studies to date, including ourown, have data on indoor air pollution or inte-grated personal exposures been considered.Categorisation of individuals based on place ofresidence clearly makes the unrealistic assump-tions that (1) children are not exposed in thehome to major indoor sources and (2) they donot move about within the urban environment.In comparison with the few other reportedstudies, we have attempted as far as possible torefine the measures of exposure based ondistance methods by including a traYc volumemeasure for each individual based on Geo-graphical Information System techniques. Ourmethods are still likely to have led to appreci-able, but unquantified, misclassification ofexposure for individuals. Misclassification isalso possible in the categorisation of cases andcontrols, although a recent study of hospitaladmissions in the North Thames (West) healthregion18 reported 88% agreement betweenlocal and external coders in the three digitcoding of primary diagnosis of asthma. Suchmisclassifications will lead to underestimates ofany true association between asthma admis-sions and road traYc.

A number of potentially important con-founding factors including parental smokingwere not considered in the analyses as they areunavailable in the routine data. Nonetheless,control for such confounding was achieved tosome extent by adjusting for Carstairs depriva-tion index which has been shown to predictsmoking patterns among individuals in NorthWest Thames.19 It is unclear to what extentconfounding may have contributed to the posi-tive associations reported in other studies.

Two of the positive studies related topre-school age children, in whom place of resi-dence is more likely to reflect true exposure totraYc related pollutants as, in this age group,children spend more of their time close tohome. These very young children may be moresusceptible to the eVects of air pollution thanolder children, although the spectrum ofdisease is diVerent and will include a number ofrespiratory conditions other than true asthma.

There was a suggestion of an inverse associ-ation between asthma and traYc density. Whilethis may be a chance finding, there is apossibility of selection bias. Families with asth-matic children may choose to live away frommajor roads because of concern for theirchildren’s health.

Although there may be reasons why we failedto find an association between road traYc andrespiratory health if one exists, this result is notexceptional.20 Recent evidence from a panelstudy in Holland21 found that only thesubgroup of children with bronchial hyperre-sponsiveness and high serum concentrations ofIgE appeared susceptible to air pollution,which suggests an interaction between pollu-tion and allergic response. Further work isneeded to disentangle this interaction and theunderlying mechanisms, but the balance ofevidence remains that outdoor air pollution hasat most a modest eVect on asthmatic symp-toms.

In conclusion, based on routine data sourcesand with acknowledged limitations in exposureclassification, the study found no evidence ofan association between hospital admissions forasthma and respiratory illness in school agedchildren in relation to road traYc. Improvedmarkers of exposure to traYc related pollutantsfor use in epidemiological studies arerequired.22

This study was supported by a British Lung Foundation/3MHealth Care Research Fellowship grant. The authors are grate-ful to Ordnance Survey for supplying high resolution road topo-graphy data for the London area (OSCAR Asset Manager), andto David Hutchinson of the London Research Centre forproviding the road traYc model for London. Work in this reportis copyright Crown and Ordnance Survey.

1 Wjst M, Reitmar M, Dold S, et al. Road traYc and adverseeVects on respiratory health in children. BMJ1993;307:596–600.

2 Weiland SK, Nundt KA, Rueckmann A, et al. Self-reportedwheezing and allergic rhinitis in children and traffic-densityon street of residence. Ann Epidemiol 1994;4:79–83.

3 Oosterlee A, Drijver M, Lebret E, et al. Chronic respiratorysymptoms in children and adults living along streets withhigh traYc density. Occup Environ Med 1996;53:241–7.

4 Van Vliet P, Knape M, De Hartog J, et al. Motor vehicleexhaust and chronic respiratory symptoms in childrenliving near freeways. Environ Res 1997;74:122–32.

5 Nitta H, Sato T, Nakai S, et al. Respiratory healthassociated with exposure to automobile exhaust. I. Resultsof cross-sectional studies. Arch Environ Health 1993;48:53–8.

6 Murakami M, Ono M, Tamura K. Health problems of resi-dents along heavy-traYc roads. J Human Ergol 1990;19:101–6.

7 Nikai S, Crest J, Nitta H, et al. Respiratory health associatedwith exposure to automobile exhaust. III. Results of across-sectional study in 1987, and repeated pulmonaryfunction tests from 1987 to 1990. Arch Environ Health1999;54:26–32.

8 Norsk Institut fur Luftforskning (NILU). The health eVects oftraYc pollution as measured in the Valerenga area of Oslo.Summary report. Lillestrom: Norsk Institut fur Luftforskn-ing, 1991.

9 Pershagen G, Rylander E, Norberg S, et al. Air pollutioninvolving nitrogen dioxide exposure and wheezing bronchi-tis in children. Int J Epidemiol 1995;24:1147–53.

Relation between hospital admission with asthma and road traYc 1073

on May 18, 2022 by guest. P

rotected by copyright.http://thorax.bm

j.com/

Thorax: first published as 10.1136/thx.54.12.1070 on 1 D

ecember 1999. D

ownloaded from

10 Edwards J, Walters S, GriYths RK. Hospital admissions forasthma in preschool children: relationship to major roads inBirmingham, United Kingdom. Arch Environ Health 1994;49:223–7.

11 Livingstone AE, Shaddick G, Grundy C, et al. Do peopleliving near inner city main roads have more asthma needingtreatment? A case control study using routine general prac-tice data. BMJ 1996;312:676–7.

12 Waldron G, Pottle B, Dod J. Asthma and motorways: onedistrict’s experience. J Publ Health Med 1995;17:85–9.

13 Committee on the Medical EVects of Air Pollutants. Asthmaand outdoor air pollution. London: HMSO, 1995.

14 London Research Centre. London Energy Study. Road Trans-port Methodology. London, 1993.

15 Diggle P, Rowlingson B. A conditional approach to pointprocess modelling of elevated risk. J R Statist Soc A1994;157:433–40.

16 Carstairs V, Morris R. Deprivation: explaining diVerences inmortality between Scotland and England. BMJ 1989;299:886–9.

17 Advisory Group on the Medical Aspects of Air PollutionEpisodes. First report. Ozone. Department of Health,London: HMSO, 1991.

18 Dixon J, Sanderson C, Elliott P, et al. Assessment of thereproducibility of clinical coding in routinely collected hos-pital activity data: a study in two hospitals. J Publ HealthMed 1998;20:63–9.

19 Kleinschmidt I, Hills M, Elliott P. Smoking behaviour canbe predicted by neighbourhood deprivation measures. JEpidemiol Community Health 1995;49:S72–7.

20 Burney P. Air pollution and asthma: the dog that doesn’talways bark. Lancet 1999;353:859–60.

21 Boezen HM, van der Zee SC, Postma DS, et al. EVectsof ambient air pollution on upper respiratory symptomsand peak expiratory flow in children. Lancet 1999;353:874–8.

22 Elliott P, Briggs D, Lebret E, et al. Small area variations inair quality and health (the SAVIAH study): design andmethods. Epidemiology 1995;6:S32.

1074 Wilkinson, Elliott, Grundy, et al

on May 18, 2022 by guest. P

rotected by copyright.http://thorax.bm

j.com/

Thorax: first published as 10.1136/thx.54.12.1070 on 1 D

ecember 1999. D

ownloaded from