Comparative statistical methods in the analysis of burn victims
| Autor: | J. Apesos, B.K. Dawson, E.J. Law |
|---|---|
| Rok vydání: | 1980 |
| Předmět: |
Multivariate statistics
medicine.medical_specialty Burn injury Univariate analysis business.industry Mortality rate Univariate Poison control General Medicine Critical Care and Intensive Care Medicine medicine.disease Linear discriminant analysis Probit model Emergency Medicine Medicine Surgery Medical emergency business Intensive care medicine |
| Zdroj: | Burns. 6:181-189 |
| ISSN: | 0305-4179 |
| DOI: | 10.1016/0305-4179(80)90063-7 |
| Popis: | Data on 70 burn patients aged 60 years or more who were treated at a regional burn centre were analysed using both univariate and multivariate methods. The univariate analyses provide information on ways in which individual patient characteristics are independently related to mortality and morbidity. These analyses result in simplistic understandings of the role each variable plays in subsequent patient outcomes but no information is provided on how these measures interact with one another. Even probit analysis, while providing useful estimates of mortality rates, traditionally considers only one variable at a time or must duplicate the procedure for separate subgroups of the sample. However, discriminant analysis permits the assessment of the overall contribution of various factors that are mutually present. Additionally, these techniques provide a mechanism to predict individual outcomes for each patient. The present study demonstrates that discriminant analysis of information easily available at the time of burn injury is useful in predicting mortality but not morbidity. The fact that morbidity is not predicted suggests the search for other factors that predispose a patient to subsequent complications. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|