Health related selection anddeath ratesintheUnited KingdomAtomicEnergy Authority workforce

Followup dataon 37355employees ofthe UnitedKingdomAtomicEnergyAuthority (UKAEA)fortheperiod 1946-79 wereanalysed toinvestigate theextent towhichselection for workonthebasis ofhealth affected subsequent deathrates. Causesofdeathweregrouped into twobroadcategories foranalysis: allcancers andallothercausesofdeath. Evidence foran effect ofselection ofhealthy individuals into the workforcewas soughtprimarilyby examiningstandardised mortality ratios (SMRs)byperiod since recruitment. SMRs for bothcategories wereparticularly lowduring the firsttwo yearsafterrecruitment (SMR = 69,95% confidence interval (CI) 48-97 forallcancers; SMR = 55,95% CI44-69forall othercausesofdeath). SMRs forallcancers didnotincrease significantly withperiod since recruitment (X' fortrend= 04,p = 0-53) but didincrease forcausesofdeathotherthan cancer(X2fortrend= 11.1, p = 0X001). Althoughadjustment forsocial classstrengthenedtheassociation betweendeathfrom causesotherthancancerandperiodsince recruitment (X2fortrend= 18.8, p < 0-001), simultaneous adjustment forallconfounding factors considered (ageatdeath, sex,calendar periodofdeath, geographical location ofthe workforce, andsocial class) produced results broadlysimilar tothoseobtained fromthe unadjusted analyses forbothcauseofdeath categories. SMRs remainedlowevenafter25 yearsoffollow up(SMR = 84,95%CI69-101 forallcancers; SMR = 81,95% CI72-91forall othercausesofdeath).The persistently low SMRs observed inthisworkforce areunlikely tobeduetotheselection ofhealthy individuals