Multilevel Modeling of ARDS Mortality Predictors.
Background. The Acute Respiratory Distress Syndrome (ARDS) is associated with acute and persistent lung inflammation. Incidence and mortality is high, extensive complex critical care is often required, and ARDS contributes to a substantial portion of health care costs. Nested data is most often the...
| Publicado en: | Multilevel Modeling of Ards Mortality Predictors pp. 149 p - 150 |
|---|---|
| Autor principal: | |
| Formato: | research Doctoral Dissertation |
| Publicado: |
University of Rochester
2012
|
| Acceso en línea: | Ver este registro en EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=ccm&AN=109858648&site=ehost-live header: @attributes: shortDbName: ccm uiTerm: 109858648 longDbName: CINAHL Complete uiTag: AN controlInfo: bkinfo: isbn: 9781267341228 dissinfo: disstl: Multilevel Modeling of ARDS Mortality Predictors. dissinst: University of Rochester jinfo: jid: IOX9 jtl: Multilevel Modeling of Ards Mortality Predictors maglogo: N pubinfo: dt: 2012 pub: University of Rochester artinfo: ui: 109858648 109858648 2012159482 UMI Order AAI3508406 109858648 ppf: 149 p ppct: 1 formats: tig: atl: Multilevel Modeling of ARDS Mortality Predictors. aug: au: Graves, Brian T sug: subj: Hospital Mortality Models, Theoretical Respiratory Distress Syndrome, Acute Mortality Exploratory Research Human Intensive Care Units Logistic Regression Models, Statistical Mortality Risk Factors Resource Databases Retrospective Design Theory Construction Type I Error ab: Background. The Acute Respiratory Distress Syndrome (ARDS) is associated with acute and persistent lung inflammation. Incidence and mortality is high, extensive complex critical care is often required, and ARDS contributes to a substantial portion of health care costs. Nested data is most often the source for ARDS research requiring specific analytical methodologies to avoid statistical errors. Absence of a mortality prediction model has hindered improvement in ARDS outcomes.Objective. To build and statistically test a more accurate and parsimonious evidence-based model that explains the influence of patient, ICU, and hospital level variables on ARDS patient ICU and hospital mortality.Design and Sample. Retrospective, exploratory, non-randomized analysis of a secondary database of critically ill patients (Project IMPACT Critical Care Medicine) during the first 24 hours of ICU admission. 11,096 patients in 133 ICU's representing 97 hospitals were used in the analysis.Methods. Univariate analysis followed by leveled logistic regression was used to build the final ICU mortality and hospital mortality models. The statistical methods used to analyze the final models were Hierarchical Generalized Linear Modeling (HGLM) and logistic regression.Results. Three-level and full two-level HGLM models could not be analyzed due to sample size. HGLM identified type I error in the hospital mortality model logistic regression analysis ICU type vector Medical Surgical compared to Medical. Hospital level variables were not associated with hospital mortality. ICU and hospital level variables were associated with ICU mortality. Examples include: Surgical and Medical/Surgical ICU's association with decreased patient mortality and Critical Care Provider led ICU models association with increased patient mortality. Increasing APACHE II scores and treatment with inotropes were associated with increased patient mortality.Significance. Knowledge obtained may help: 1) identify more rigorous statistical approaches to prediction modeling in large nested datasets; 2) clinicians identify variables for goal directed therapy; 3) influence specifically targeted therapeutic strategies; 4) identify hospital and ICU level staffing and systems requirements necessary for achieving the best possible outcomes for ARDS patients; and 5) stratify subjects for future ARDS research. Limitations include the retrospective exploratory design, use of a secondary dataset, missing and limited variable specificity. pubtype: Dissertation doctype: research Doctoral Dissertation ougenre: Unknown language: English refInfo: holdings: @attributes: islocal: N |
|---|