Applications of Mixed Models for Investigating Progression of Chronic Disease in a Longitudinal Dataset of Patient Records from General Practice
|Zalihe Yarkiner1*, Gordon Hunter1, Rosie O’Neil1 and Simon de Lusignan2|
|1School of Mathematics, Faculty of Science, Engineering and Computing, Kingston University, Penrhyn Road, Kingston-Upon Thames, KT1 2EE, United Kingdom|
|2Department of Health Care Management and Policy, Faculty of Business, Economics and Law, University of Surrey, Guildford, GU2 7XH, United Kingdom|
|Corresponding Author :||Zalihe Yarkiner
School of Mathematics, Faculty of Science
Engineering and Computing
Kingston University, Penrhyn Road
KT1 2EE, London, United Kingdom
Tel: +44 (0) 7842 772 172
E-mail: [email protected]
|Received June 11, 2013; Accepted September 28, 2013; Published September 30, 2013|
|Citation: Yarkiner Z, Hunter G, O’Neil R, de Lusignan S (2013) Applications of Mixed Models for Investigating Progression of Chronic Disease in a Longitudinal Dataset of Patient Records from General Practice. J Biomet Biostat S9:001. doi: 10.4172/2155-6180.S9-001|
|Copyright: © 2013 Yarkiner Z, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.|
The field of longitudinal analysis is a rapidly developing and increasingly important area of statistical modelling. This is in response to the increasing availability of longitudinal data across many fields and recognition of the rich resources such data might provide. However, a lag between the development of statistical methodologies and their applications to substantive problems has been identified, but current advances in novel longitudinal methods aim to redress this imbalance. Longitudinal data often presents repeated response measures, but these data are often unbalanced in relation to number of and intervals between measures. Although Linear Mixed Models provide a framework which can accommodate such unsystematic response patterns, such models become unreliable when responses do not approximately follow a normal distribution. Extensions of Linear Mixed Models to Generalized Mixed Models allow the analysis of such non-normal outcomes via appropriate transformations of the response. These models, which are based on a repeated measures structure within a two-level multilevel framework, allow both random and systematic effects to be studied simultaneously. Although these are well-established, they are only recently being applied in the medical and social sciences.
Here, applications of these models are illustrated by analysing the progression of Chronic Kidney Disease (CKD) over time, and in relation to the impact of known co-morbidities. The data are taken from routinely collected patient records from a representative sample of UK General Practices (GPs). The aim is to use the longitudinal aspects of the data to further understanding of the early indications and the nature of the progression of CKD. The methodologies should be applicable to other chronic illnesses, which are primarily managed at the GP level.
The results of our models concur with previous research, in regard to the associations between individual comorbidities and CKD. Furthermore, our models evaluate the impact of combinations of these co-morbidities on the rate of progression of CKD, as measured by repeated estimated glomerular filtration rate (eGFR) readings. Our results provide evidence that this methodological approach is a useful and appropriate mechanism for investigating dynamic relationships within health-related data, and that such routinely collected data can be useful in epidemiological research.