Advanced

Multivariate linear regression modelling of lung weight in 24,056 Swedish medico-legal autopsy cases

Gustafsson, T.; Eriksson, A. and Wingren, C J LU (2017) In Journal of Forensic and Legal Medicine 46. p.20-22
Abstract

Heavy combined lung weight at autopsy is a non-specific autopsy finding associated with certain causes of death such as intoxication. There is however no clear definition of what constitutes "heavy" lung weight. Different reference values have been suggested but previous studies have been limited by small select populations and only univariate regression has been attempted. The aim of this study was to create a model to estimate lung weight from decedent parameters. We identified all cases >18 years age autopsied at the Swedish National Board of Forensic Medicine from 2000 through 2013, excluding cases with a post-mortem interval >5 days as well as cases with extreme values, totalling 24,056 cases. We analysed body weight, body... (More)

Heavy combined lung weight at autopsy is a non-specific autopsy finding associated with certain causes of death such as intoxication. There is however no clear definition of what constitutes "heavy" lung weight. Different reference values have been suggested but previous studies have been limited by small select populations and only univariate regression has been attempted. The aim of this study was to create a model to estimate lung weight from decedent parameters. We identified all cases >18 years age autopsied at the Swedish National Board of Forensic Medicine from 2000 through 2013, excluding cases with a post-mortem interval >5 days as well as cases with extreme values, totalling 24,056 cases. We analysed body weight, body height, sex, age, BMI, BSA as well as untransformed and transformed lung weight. The analysis was stratified for sex. We evaluated the fit of the models and that the model assumptions were not violated. We set out to apply the model with the highest residual sum of squares to derive limits for heavy lungs. In univariate regression BSA and height showed best performance. The final model included height, weight and age group. After excluding large standardized residuals (>3, <-3) the final model achieved R(2) of 0.132 and 0.106 for women and men respectively. While we managed to create a multivariate model its performance was poor, possibly a fact reflective of the physiological nature of the lungs and in turn its variability in fluid content. Linear regression is a poor model for estimating lung weight in an unselected population.

(Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
Journal of Forensic and Legal Medicine
volume
46
pages
3 pages
publisher
Elsevier
external identifiers
  • scopus:85007489246
  • wos:000397021900004
ISSN
1878-7487
DOI
10.1016/j.jflm.2016.12.001
language
English
LU publication?
yes
id
93f83cfb-cfa5-4c98-8858-109898fbcb80
date added to LUP
2017-01-05 08:08:00
date last changed
2017-11-19 04:36:30
@article{93f83cfb-cfa5-4c98-8858-109898fbcb80,
  abstract     = {<p>Heavy combined lung weight at autopsy is a non-specific autopsy finding associated with certain causes of death such as intoxication. There is however no clear definition of what constitutes "heavy" lung weight. Different reference values have been suggested but previous studies have been limited by small select populations and only univariate regression has been attempted. The aim of this study was to create a model to estimate lung weight from decedent parameters. We identified all cases &gt;18 years age autopsied at the Swedish National Board of Forensic Medicine from 2000 through 2013, excluding cases with a post-mortem interval &gt;5 days as well as cases with extreme values, totalling 24,056 cases. We analysed body weight, body height, sex, age, BMI, BSA as well as untransformed and transformed lung weight. The analysis was stratified for sex. We evaluated the fit of the models and that the model assumptions were not violated. We set out to apply the model with the highest residual sum of squares to derive limits for heavy lungs. In univariate regression BSA and height showed best performance. The final model included height, weight and age group. After excluding large standardized residuals (&gt;3, &lt;-3) the final model achieved R(2) of 0.132 and 0.106 for women and men respectively. While we managed to create a multivariate model its performance was poor, possibly a fact reflective of the physiological nature of the lungs and in turn its variability in fluid content. Linear regression is a poor model for estimating lung weight in an unselected population.</p>},
  author       = {Gustafsson, T. and Eriksson, A. and Wingren, C J},
  issn         = {1878-7487},
  language     = {eng},
  pages        = {20--22},
  publisher    = {Elsevier},
  series       = {Journal of Forensic and Legal Medicine},
  title        = {Multivariate linear regression modelling of lung weight in 24,056 Swedish medico-legal autopsy cases},
  url          = {http://dx.doi.org/10.1016/j.jflm.2016.12.001},
  volume       = {46},
  year         = {2017},
}