gms | German Medical Science

54. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

07. bis 10.09.2009, Essen

A propensity score based analysis of data on cognitive function in the elderly using the R-package “NonRandom”

Meeting Abstract

Search Medline for

  • Susanne Stampf - University Medical Center Freiburg, Freiburg, D
  • Angelika Caputo - Novartis AG, Basel, CH
  • Martin Schumacher - University Medical Center Freiburg, Freiburg, D

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 54. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds). Essen, 07.-10.09.2009. Düsseldorf: German Medical Science GMS Publishing House; 2009. Doc09gmds102

doi: 10.3205/09gmds102, urn:nbn:de:0183-09gmds1029

Published: September 2, 2009

© 2009 Stampf et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.en). You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.


Outline

Text

Background: The analysis of data from observational studies typically requires an adjustment for systematic differences in treatment groups. This is often done using traditional regression. In addition, propensity score (PS) based methods have been frequently applied to analyse such data in the last decade [1]. Both techniques differ in their handling of covariates and interpretation of parameters to be estimated, but they yield similar estimates for treatment effects if used appropriately [1].

So far, there is no software available for a complete PS based analysis of data from observational studies. We established the R-package “NonRandom” providing a comprehensive tool if stratification or matching by PS is desired. Therefore, we implemented functions to estimate the PS using logistic regression, to stratify or to match observations by PS, to check the balance of covariates using parametric and non-parametric tests or visualization of covariate distributions, and subsequently to estimate the effect of interest. The package will be available under CRAN.

Methods: The R-package “NonRandom” was used to analyse data from the cohort study Hei.DE where the cognitive status in the elderly was investigated by means of telephone interviews [2], [3]. Here, our aim was to investigate the impact of pain medication on the cognitive function measured as a sum score using stratification and matching by PS.

Results: For the experienced user, the R-package “NonRandom” provides an extensive set of functions for an easy and flexible implementation of PS based analyses. All functions are well documented and illustrated by example code. Users who are not familiar with PS methods so far are guided by the package through critical decision points.

In our application, the estimated effects for pain medication using PS stratification and matching are very similar and show a non-significant effect throughout.


References

1.
Stürmer T, et al. A review of the application of the propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared to conventional multivariable models. Journal of Clinical Epidemiology. 2006;59(5):437-461.
2.
Debling D, et al. Assessment of cognitive status in the elderly using telephone interviews. Zeitschrift für Gerontologie und Geriatrie. 2005;38(5):360-367.
3.
Debling D, et al. S08.1: Diabetes and cognitive function in the elderly. Biometrical Journal. 2004;46(S1):16.