Risk modelling study for carotid endarterectomy G.Kuhan,E.D.Gardiner*,A.F.Abidia,I.C.Chetter,P.M.Renwick,B.F.Johnson, A.R.WilkinsonandP.T.McCollum Academic Vascular Unit, Hull Royal In®rmary and *Applied Statistics Centre, University of Hull, Hull, UK Correspondence to: Professor P. T. McCollum, Academic Vascular Unit, Hull Royal In®rmary, Anlaby Road, Hull HU3 2JZ, UK e-mail: peter.mccollum@virgin.net) Background: Theaimsofthisstudyweretoidentifyfactorsthatin¯uencetheriskofstrokeordeath followingcarotidendarterectomyCEA)andtodevelopamodeltoaidincomparativeauditofvascular surgeonsandunits. Methods: A series of 839 CEAs performed by four vascular surgeons between 1992 and 1999 was analysed.Multiplelogisticregressionanalysiswasusedtomodeltheeffectof15possibleriskfactorson the 30-day risk of stroke or death. Outcome was compared for four surgeons and two units after adjustmentforthesigni®cantriskfactors. Results: Theoverall30-daystrokeordeathratewas3´9percent29of741).Heartdisease,diabetes and stroke were signi®cant risk factors. The 30-day predicted stroke or death rates increased with increasingriskscores.Theobserved30-daystrokeordeathratewas3´9percentforbothvascularunits andvariedfrom3´0to4´2percentforthefourvascularsurgeons.Differencesintheoutcomesbetween thesurgeonsandvascularunitsdidnotreachstatisticalsigni®canceafterriskadjustment. Conclusion: Diabetes,heartdiseaseandstrokearesigni®cantriskfactorsforstrokeordeathfollowing CEA.Theriskscoremodelidenti®edpatientsathigherriskandaidedincomparativeaudit. Paper accepted 16 August 2001 British Journal of Surgery 2001, 88, 1590±1594 Introduction Stroke is the third most common cause of death in the industrializedworld.Thebene®tofcarotidendarterectomy CEA) in the prevention of stroke in patients with a symptomatic high-grade internal carotid artery stenosis 70±99percent)hasbeenprovenbyrandomizedcontrolled trials 1±3 .ForCEAtobeworthwhile,thelong-termbene®t strokeprevention)mustoutweightheimmediateoperative riskstrokeanddeath).Currentopinionsuggeststhatmost symptomatic patients will bene®t from CEA if the combined 30-day stroke or death rate is less than 6 per cent 4 .Sincetheemergenceofclinicalgovernancetherehas been a strong emphasis on the comparative audit of outcome. Crude mortality and morbidity rates have been used widely to compare surgical outcome, but adjustment bycase-mixisnecessaryifmeaningfulcomparisonsaretobe made 5 . Thepublished30-dayriskofstrokeordeathafterCEA varies from 2 to 20 per cent 6 . The reasons for this wide variation in outcome are not immediately obvious. Risk modellingmayidentifythecauses,andmaybeusedtoset clinical standards for vascular surgeons and units. Risk modellingusingmathematicaltechniqueshasbeenusedin cardiology and cardiac surgery 7,8 . The aims of this study were to identify the risk factors that in¯uenced outcome following CEA and to develop a model that accurately predictedoutcometoaidincomparativeaudit. Patients and methods Aseriesof839CEAsperformedbyfourvascularsurgeons fromtwovascularunitsAandB)fortheinterval1992±1999 was available for analysis. Unit B had entirely prospective data41´3percentofthedata)whileunitAhadprospective andretrospectivedata58´7percent).Duringthestudyone surgeonsurgeon4)movedfromunitBtoAandperformed 31CEAs7´1percentofCEAsperformedinunitA).The datawerecollectedonpreprintedsheetsandlaterentered into a vascular database based on Access 97 Microsoft Corporation,Redmond,Washington,USA). The median age of the patients was 68 range 38± 86)yearsand60´4percentweremen.Dataon67possible The Editors have satis®ed themselves that all authors have contributed signi®cantly to this publication Original article 1590 ã 2001 Blackwell Science Ltd British Journal of Surgery 2001, 88, 1590±1594