脓毒症预警诊断模型建立和初步应用分析

《脓毒症预警诊断模型建立和初步应用分析》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、万方数据解放军医学院硕士学位论文度74．8％时，约登指数最大，预测概率P=7．6‰可以认为此时有最佳预警准确性。3)用人工神经网络模型验证回归模型，得出AUC=96．4％，与Logistic回归模型的AUC相近。4)将30例严重感染患者的脓毒症早期预警相关因素输入该模型中，模型结果与患者是否实际发展为脓毒症进行对比，实际有11例发展为脓毒症。预警模型灵敏度45．5％，特异度84．2％，阳性预测值62．5％，阴性预测值72．7％，诊断符合率70．o％。将11例患者发生脓毒症前24h内的各项生理参数输入预警模型中，分析得出发展为脓

2、毒症的有8例。该8例与实际发生脓毒症相同，灵敏度72．7％。结论：1)本研究发现脓毒症预警诊断因素可能包括皮肤淤点、血小板、降钙素原、肌酐和国际标准化比值。2)本研究建立了脓毒症预警诊断模型，预警模型运算出预测概率P可以判断患者是否发展为脓毒症。3)经验证和临床初步应用，该预警模型灵敏度和特异度较高，对提前判断感染是否发展为脓毒症有较高的准确性，具有一定的·临床价值。关键词：脓毒症；预警模型；Logistic回归分析2万方数据解放军医学院硕士学位论文Establishmentandpreliminaryapplicationo

3、fearlywarninganddiagnosismodelforsepsisAbstractObjective：Sepsisistheacutesystemicinflammatoryresponsesyndromecausedbyinfection，anditisoneofthemaincausesofdeathincriticallyillpatients．Thesepsisclinicalmanifestationsarenotspecificandlackreliablecriteria,SOpatientstend

4、todeterioraterapidlywhichleadstohighmortality．Byanalyzingtheearlyriskfactorsintheearlystageofsepsis，theresearchidentifiedfactorsassociatedwithsepsisandestablishedearlywarningmodelinordertovalidatetheclinicalvalueofthemodel．Methods：Wecollected215casesfromJanuary2010t

5、oOctober2015，including129casesofinfection，86casesofsepsis．Generalcondition,clinicalsymptomsandlaboratoryexaminationwererecorded．Thenweanalyzerelatedfactorsofinfectioncausingsepsis．ThelogisticregressionmodelwithcomputerWasusedtoconstructtheearlywarningmodelofsepsisby

6、screeningouttherelevantfactors．Afterthat，weselected30severeinfectedcasesandinputdatatoverifythemodel，comparingwiththe30casesSOastoanalyzetheclinicalvalueoftheearlywarningmodelofsepsis．Results：1)Thegeneralconditionsof129infectedpatientsand86sepsispatientsrevealedthat

7、thedifferencesofclinicalsymptomsbetweentwogroupsinhypoxemia,hypotension，disturbanceofconsciousness，andskinspotsthewerestatisticallysignificant．SepsisgroupdataWashigherthanthoseininfectiongroup俨<0．05)．Thedifferencesoflaboratorytestsbetweentwogroupsinplateletcount，pro

8、calcitonin，totalbilirubin，urea,creatinine，glucoseandinternationalnormalizedratiowerestatisticallysignificant．Plateletcountsinsepsisgroupwe