Asthma mortality in Canada , 1946 to 1990

OBJECTIVE: To assess the impact of asthma on Canadian mortality rates over a 45-year period.

A STfI~'i!\ IS A COMMON CH RONIC IU:SPIRATORY DISl-:l\SI , that affects millions of people worldwide and infli cts a considcrahle financial burden on soc iety.Altho11gh there arc no definitive estimates of the prevalence of a.,thma based on a single 1111il'onn definition, this chronic illness i., estimated to allecI between l) and 12 million people in Canada and the United States alone ( 1,2).Furthermore, populaiion-hascd studies in A11siralia, the Netherlands and New Zealand have shown the prevalence of hronchial hypcrrcpon sivcncss, a condition which may predispose an individual to asthma.to he between 11 and 24'/r in asymptomatic children and adults (J-5).In the United States, the direct and indirect societal costs of asthma were estimated to exceed US$6 billion in I lJlJO, with the largest direct medical cost being in-palil'lll hospital services ( US$ l .6 hi Ilion ) and the );1rgest indin' .L' I cos I heing reduced prnductivity due to loss of school days (U S$ I billion) (6).
fvlos1 ol'tcn when aslhma is well controlled on ,111 out-p~llil'llt basis.it r,m: ly leads to hospitalization or death.I lowever, reccnl increases in asthma death rates in mosl developed countries (7 -11) suggest that the etiologic !'actors associated with lhis chronic illness are slill not well understood.Changes in asthma mortality rates haw bee n shuwn lo he highest in young adoll'scents and adults.where rates ol' death increased by as much as threefold (7.8, 12), and in New Zealand and the Un ited Kingdo m where the asthma mortality reached epidemi c proportions (7).Se veral hypotheses have been pul forward 10 explain I his worldwide phenomenon.lhe most notable being thal heta-agonist use increases lhe risk ol' death.particularly in those with severe asthma ( I J).Increases in morlality were ais(i hypothesized to be associaled with shit'ling patlerns of physician diagnosis, especially from bronchilis lo aslhma in children and chrnnic obstruclivc pulmonary disease lo asthma in older smoh.ers ( 14-15): availahilily ol' helter diagnostic tests ( 16): changes in the classification of deaths over time: and secular changes in the preva lence and sevc rit v or asthma worldwide ( 17, I 8 ).Overall. in contrast to the declining rates or dea1h for other chronic di seases, asthma death rates ha ve Ix-en shown to he on the rise lhrnughout lhc deve loped world.
The present study was designed to examine whether these recent hislorical increases in asthma mort al ity rat es have had a detrimental impacl on adull health.Relying on mortality data for Canada.we undcnook a longit udinal analysis to delermine the impact of asthma on Canadian mortalit y rates over a rccenl ..J-5-year period sta rti ng in 1945.The aim of this study was lo assess ho w as thma mortality rates haw changed over time and lo measure lhe impact or 1hese changes on overall mortality and li fe expec tanc y in men and W(llllCn.

PATIENTS AND METHODS
Information on asthma mortalily and the population at risk of dca1h was obtained from the Canadian Centre for 1-Icallh Statistics in Oltawa.Asthma deaths hy se x, age-group and year were compiled from S1a1is1ics Canada tabulations lhal mention asthma as the underlying cause of death.Age and Bronchitis, chronic X Hay fever wilh aslhma Late-onset aslhma X ' Data ta ken from reference 19 sex-specific population figures for Canada WLTC obtained from censal and post -ccnsal annu,il L'.stimates.Calculated cause and age-specific rates were ex pressed as rate\ rwr I 00.0()() population.
Five revi sions of the lnl crnalional Classification of Diseases (ICD) were used by Stati,Iics Canada to classify ast hma deaths from 1945 to 1990 ( I l)).Changes in Ille diagnostic categories included under rubrics of aslhma over lhis lime period are shown in Talilc I. Out of the 19 condilions !isled.only two -bronchial :1sthma and hay !'ever -WCI\' included in all five rnisions: 12 llf thL'.fl) cundilion., li sted were included in only one (lUI of till' live re vis iuns or !CD. Asthma mort ality rates were as.sessed by comparing changes in age-specific and crude and standardi zed ralcs of death over nine five-year periods.Standardi;,ed ra1cs were ca lculated by directly adjusting lo lhc C11wdi:1 11 agc-s1111clurc in I lJ7 I to I lJ7) period age-specific death rates.Once ralcs were slandardin~d.period standardi;cd mortalily ratios (SMR) were conslructcd by cu111p:1ring aslhma mortalily ralL'S in 1hc period 1971 lo I l)75 wilh rail's for the other eight periods.Slatistical significance was assessed by determining whclher the ex peeled number of deaths, if' l l)7 [ lo 197) rales applied, exceeded the observed numhL:r of dcalhs at a V/r and I %, le vel or signil1cancc (20).
The impact of asthma on ovnall mortalit y was assessed us ing twu measures.The rirst.potential years or life Josi  For PYLL index , si gn ificant differences in the observed and ex pec ted number of death s at a 0.05 and 0.01 leve l were determined using a x 2 test.The second measure, lif'c expec tan cy lost due to asthm a, was eva luated al hirth by const ru ct ing cau se de let ed life tables using stand ard de mographic tech ni ques (21)(22)(23).These tabl es were designed to me as ure the im pact of removin g deaths attributabl e to asthm a on li fe ex pectancy at bi rth.A signifi cant loss in life ex pect ancy was determ ined from the ratio of the expected loss to the corresponding standard error around this estimate.If thi s critical ratio exceeded a threshold val ue of 2.33, then the loss in life expectancy was deemed significant at a 0.0 I level (21 ).
A more detailed description of how the SMR, PYLL and PY LL index, and life expectancy lost were calculated appears in Appendix I.

RESULTS
A total of 12,0 IO male and 8486 female asthma deaths were recorded in Canada fro m 1946 to 1990.The majority of these deaths occurred in men and women above the age of 64 years.ln men.56.5 % of all as thma deat h., occurred among  Male death rate, declined from 5.82 J ca ths per I 00.000 population in 1951 to I q55 to 1.54 deaths per I 00.0()0 population in I lJ86 to I l)()O: fe male death rate declined from .LHl to 1.74 deaths per I 00,(H)O popul ation over the same time interval (hgurl' I ).The gap hctwee n se x-specific rates was larges t in 1951 to I l)55 at 2.52 deaths per I 00.000 population.;111d s1rnllesl in 197(1 to 1980 at 0.02 deaths per 100.()00population.
P:1tterns of age-specific mortality for men and women from age s O to 85 years and over were exa mined for three periods: I l)5 I to I lJ55.I 971 to 1975, and 1981 to I lJ85 (Figun: 2).Age-s pecific mortality rates in 1951 lo 1955 were highest at all ages.ex cept for ages 15 to 24 years when death rates for the 1981 to 1985 period were of greater magnitude.The largest reduction in age-specific rates ove r time ol•currcJ !'or those aged less than one year.where r:1tes declined from 2.50 deaths per 100.000popuhition in 195 I lo J l)55 to lU)5 deaths per I 00,000 population in j l)

DISCUSSION
Our study demonstrates that there has been a nearl y threefold decline in asthma dealh rate s since the earl y I 1 )50s.This decline is apparent for both men and women ancl at mos t ages.except hL•twcen 15 and 24 years.Murcovcr.the los., ol life expectancy ;1l hinh attributable to asthma !or mL•n ,1nd women is negli gible over the course or the study.Over:111 1hc Jecline in ast hma mortality appears to be linl,.,L•dlo the u111vergence or male and female rates in the 1970, and I l)80s and with changes in the pallern or age-spec ific mortality ll\L'r l i lllt'.
Contrary to our L :xpectation the analysis shows that even tho ugh asthma dl'ath rates increased slightly in the late 1970s, rates of death s have sliarply declined over this 45-year period.Based on curre nt available research (8) we would have expected that increase, in asthma dl'aths in the 1970s were part of a longer hi storical trend.T his historical pattern of mortality has been especially evident in Ne w Zealand and Australia, where rates of death for those aged fi ve to 34 years have been shu wn to haw increased notably si nce 1940 ( I 0, 11 ).However.our data do not support ,tH.:11 ,1 trend and suggest that rece nt moderate increases in asthma mortality arc actuall y part of an overall decline in mortality.
The effect or asthma on life expectancy is likely to be eve n smaller tha n we have estimated, for two reasons.The lirst is a difficulty in validating that a death allributed to asthma 011 a death certificate is actually due to asthma.Al though the coding or ast hma 011 Jcallt ec nif'icates in New Zea land .thL' United Kingdom and Australia has bcL•n show n to have a sensitivity of between 87 and 97 %, the spec ificity of these coding practi ces has been demonstrated to be only between 59 and 69% (24-26 ).Based on the prevalence or the disease these speciliL•ities ,mJ scnsitivi ties lead lo a positive predictive value or a certified asthrna death actually being caused by asthma to avi:ragc 500 overall.anJ to range from 40 tn 84 7r dcpe11ding on age (2 6) .Asthma appears to be erroneously owr-represented a • the underlying cause or death n1ainly because or other comorbid con diti ons.particularly chronic obstructive lung di sease and cardiovascular disease.being present at the time or death (26,27 ).
The second reason fo r the low impac t or asthma on Ii re ex pectancy is that the patlern or asthma morta lity in the data available for analysis is part iall y a result or a coding arti fac t brou •ht about by re visions in the lCD.The diagnostic categories incl uded under as thma have been altered substanti ally over the rnursc or thi s study.In CanaJa.the overal l rate of asthma mortality appears to increase in the 1950s and 1970s at a time when major changes in the condi tions incl uded under the ru bric of asthma we re made to ICD-6 and ICD-9.Bridge-colling exercises in the United Kin gdom and Uni ted States s ugges t that with ICD-9 at least, re visi ons in the rnnditions incluJeJ under asth ma resul ted in an art ifi cial increase in O\'eral l asthma mortality of between 28 and 35 % (2 8.29).However.th is increase was shown to be substant ially reduced in you ng ad ult s and adol esce nt s. an age groupi ng in ,vhi •h the Jiagnosis o r asth ma is mos t firmly established (9.29).Thus. the true de ath rate due to asthma in the older age groups may be e ven lower than we have esti mated here .
Overall.we have de monstra ted that ast hma deat h rates in Canada have ge neral ly decl ineJ over this period.There has been a nearly threefold dec line in asthma death rates since the 1950s and .except fo r the pc1fod of early ad ul thood.this dec line in rates is apparen t for all age groups.We t'u rll1e r demonstrate that the impact of asth ma mortal ity on overall mortality and lire expecta ncy is almost negli gible.Finally, based on our res ults.we believe that the impact of new therapi es will need In be measured against other indicators 1h,rn r;1tc ur asthma death.,.

ACKNOWLEDGEMENTS:
The auth ors arc indebted to the anadian Centre fo r I le.where Qi repre\ents probability of death excludi ng cause: Q represents overa ll pro bability or death at a give n age inte rval: and Ii represe nts the propurlinn nr death s ,1ttrib-u1ed to causes other than the excluded cause or death.

(
PYLL). was used to determine the i111pac1 or as1 h111a on infant.child and young and middle-aged ,idu ll mo rtality by Can Respir J Vol 2 No 1 Sp ring 1995

Can Respir J Vol 2
No 1 Spring 1995 Asthma mortality in Canada

TABLE 1 Conditions included in the coding of asthma deaths in the International Classification of Diseases (ICD) Revision, 1948-1977 ICD revision '
XAllergic.any ca use

TABLE 2
Asthma mortality rates and standardized mortality ratios (SMR) forCanada, 1946Canada,  -1990 ' The number of deaths expected if 1971 to 1975 rates for Canada applied; t Rates are dea ths per 100,000 population; t Ratio of the number of dea ths observed to the number expected: §Significantly different from 1 at P-c0.01 calculating the potential number or years of life Inst hcf()rl' age 75 years.PYLL were calcu lated by multiplying the total nu mber of age-group deaths by the average number of years re mainin g to age 75 years.Time trends were examined usi ng the PYLL index, which is a ratio of the observed to expected PY LL if 1971 to 1975 asthma mortality rates applied .
Life expectancy at birth (ex°) and loss due to mortality attributed to asthma in Canada, by sex 1946-1 990 In compa ri son, male life expec tancy al birth rose by nearly eigh t yea rs from 65 .7 to 73.5 years, and J'cmalc life expectancy inc reased by alntosl 12 years t'rom (1l) .6 lo 81.4 years over this 45-yl'ar period .
j l)7 I to 1975 fur men and women.respectively.In 1951 to 1955.PY LL was 3.23 times higher fo r men and 2.b5 times higher for women lhan expected if 1971 lo 1975 sex-specific asthma r:1tcs applied.In till' most current period.19:-:;6to1990.PYLL t'igurcs I'm men and women ;ire of till' same 111agnit11Je:1l l5,8-Hland 15.--11(1.respcctivcly.Fur all periods and either sex. the loss in life expectancy at birth allrihuted lo as thma was arounJ one month and was 64 TABLE3no t signific:1nl al a O.Cl I level (Tabl e .l).

E
1 1th Stati stics.Statisti cs Canada in Ott awa for th eir ass istance.This work was supported by the Nationa l Health Rescan:h De velopment Program of the De partme nt of Nati onal Hea lth an d We lfa re throu gh a National He.11th Research Scholar Award to Dr MlJnt ancr.and a Na tional Health Research Sc ien ti st A ward to Dr Schechter.Adams PF.Benson V. Curre nt est imates from the Na tio nal I kalth Inte rvie w Survey 1989 .Nation al Cent er for Health Statisti cs.Yi ta i I kalt h Stat Seri es I 0. No 176.1990.' Wig le DT.Pn:v,d cnce of selected chronic diseases in Canad,1.1978-79.Ch ronic Dis an 198'.:;.3:9..,. Salmm.: CM .P •at JK.Brillon WJ.Woolcock AJ.Bro nchial hypL'ITespnnsivrnL'SS i11 t1 o po pul ati ons of Aust ralian school L'i1ildrc11.I. ffrla1ion t1) rc~pi ratory sympt oms and diagnosed a,lhma.Clin Alkrg.y1987; 17:27 1-8 1. 4. SL•,irs MR. l:l urrnws B. Flanne ry EM .Hcrbi,on GP.Hew itt CJ.HDldaway MD. !{elati on between airway res ponsiveness and SL'J'lllll lg.I : in d1 ild rL ' ll with a,thma and in apparen 1ly normal ch ildren.N Eng.I .I Med 199U25: 1067-7l.'.i.Rijcken ll.Schouten JP.Weis~ ST.SpeiLcr FE. van dc r Lcnde Can Respir J Vol 2 No 1 Spri ng 1995 di represents nhservcJ deaths in age group i and Ci re presents expected J caths in age grnup i. Confidence intervals around the observed SM R values where c;tlcu lateJ using multipliers based on Byar•s apprnximatinns or the exact Poisson limits ( 14).Potential years of life lost (PYLL ) PYLL = Id; x (75 -Y;) where lli represents number of deaths in age group i and Yi rt.'pre sen ts age at midpoint of age group i. Potential years of life lost index (PYLLI ) Ie; x 175 -Y,> where O represents obser\'cd PYLL E rep rc:sent~ expected PY LL i r Canadian age-spec ific death rates are appl ied : di represents number of deaths in age group i: L'i represe nts expec ted deaths in age group i: and Yi represen ts age at midpoint of age group i.