Submit Article
Back to Journal

SM Journal of Biometrics & Biostatistics

Mediation Analysis of Co-occurring Conditions for Complex Longitudinal Clinical Data

[ ISSN : 2573-5470 ]

Abstract
Details

Received: 29-Nov-2016

Accepted: 25-Jan-2017

Published: 25-Jan-2017

Research Article | Volume 2 - Issue 1 | Article DOI :

Download PDF

Douglas D. Gunzler1*, Nathan Morris2, Adam Perzynski1, Deborah Miller3, Steven Lewis1 and Robert A. Bermel3

1Case Western Reserve University, Center for Health Care Research & Policy, Metro Health Medical Center, Cleveland, Ohio, USA

2Case Western Reserve University, Department of Epidemiology and Biostatistics, Cleveland, Ohio, USA

3Mellen Center for Multiple Sclerosis Treatment and Research, Cleveland Clinic, Cleveland, Ohio, USA

Corresponding Author:

Douglas D. Gunzler, Case Western Reserve University, Center for Health Care Research & Policy, Metro Health Medical Center, 2500 Metro Health Drive, Cleveland, Ohio, 44109-1998, USA, Tel: 216-778-2764

Keywords

Structural equation modeling; Latent growth modeling; Mediation analysis; Differential item functioning; Multiple indicator multiple cause modeling; Multiple sclerosis

Abstract

Background: Symptoms or test results may be common to two or more co-occurring conditions. This problem of symptom overlap makes it challenging for clinicians to determine a focus for treatment in a patient given changes in the severity of either condition.

Methods: Structural equation modeling methods can be used to disentangle some of the complexities of disease symptom etiology, given co-occurring conditions, and support treatment decision making. These techniques provide the flexibility to deal with specific challenges present in data as extracted from Electronic Health Records (EHR) (i.e. individually varying follow up times, irregular follow up, missingness, systematic error in patient reported outcomes, lack of clear temporal precedence between measures). Specifically, a proposed latent growth modeling approach accounting for differential item functioning along with the Monte Carlo simulation method for assessment of mediation can be used to investigate how one condition leads to a co-occurring condition, adjusted for the overlapping symptoms of both conditions.

Results: This paper uses an example investigating how Multiple Sclerosis (MS) leads to depression in patients in which depressive symptoms overlap with other symptoms of MS, such as fatigue, cognitive impairment and physical impairment to illustrate the methods. It was demonstrated that not adjusting for this overlap can lead to different results.

Conclusions: Developing methods for mediation analysis of co-occurring conditions for more complex longitudinal clinical data as recorded at a typical patient visit can help clinicians make improved use of data bases such as EHR to support clinical decision making in real time.

Citation

Gunzler DD, Morris N, Perzynski A, Miller D, Lewis S and Bermel RA. Mediation Analysis of Co-occurring Conditions for Complex Longitudinal Clinical Data. SM J Biometrics Biostat. 2017; 2(1): 1004.