Survival analysis : research for decompensation risk factors

Get Complete Project Material File(s) Now! »

Study design

It was an analytical, epidemiological and a prospective cohort study. A Feasibility pilot study was initially led with 64 patients resident in Lanmeur nursing home in 2014 (17). But because of a lack of power in comparability between groups, a second period of inclusion was planned in 2015. A global of 88 patients were included on 127 required for an expected difference of 20% on variables, based on alpha level of 5% and beta level of 20%, using a non symmetrical sample ( 75%-25). The two periods of inclusion lasted six months each and the monitoring of patients was carried out three months of their inclusion. This thesis dealt with the all-cohort pilot study followed up at twelve months of their inclusion.

Study population

The study population was the feasibility pilot study cohort completed by 24 patients of a second period of inclusion to obtain 88 patients. It included all the multimorbid patients (according to the EGPRN definition of multimorbidity) whose GPs was practicing in the university center of Lanmeur and residing in Lanmeur nursing home.
Enrollment criteria were all patients corresponding to the definition of multimorbidity, meaning « any combination of chronic disease with at least another disease (acute or chronic) or a bio psychosocial factor (associated or not) or somatic risk factor ». For the team, consisting of doctors, residential students and researchers in family practice on multimorbidity, a bio psychosocial factor included psychological risk factors, psychosocial risk factors, lifestyle, demographics (age, gender), psychological distress, social-demographic characteristics, aging, beliefs and expectations of patients, physiology, and pathophysiology (12).

Population description

Like previous studies about multimorbidity lead in Brest, the team based statistical analysis on an alpha risk of 5% and a beta risk of 20% to expect a significant differ- ence on the variables of 20%. 127 patients were required. A first period of inclusion lasted from July to December 2014 but decompensated patients were not enough for a significant difference. Thus, another period of inclusion was planned from July to December 2015 using the same method as described above. A bi-dimensional analysis was realized for each variable. The aim was to compare the two patient’s groups « decompensation » and « noting to report », and to verify data quality. A fisher’s exact test, chi2 with or without Yates correction were used for qualitative data , with an alpha level set at 5%. For quantitative data, a Wilcoxon/ Mann-Whitney test was performed to compare median in a non-parametric test. Multidimensional analysis : clustering and classification Patients were also described, whatever their study’s status, by common characteristics. First of all , irrelevant and no discriminating variables were removed. A dendrogram was made, using a multiple correspondence analysis (MCA) discriminating variables in each group, and the technique of hierarchical clustering on principal components (HCPC).

READ  Informative descriptor preservation via commutativity for shape matching

survival analysis

To compare survival time of patient’s groups, estimating its distribution and how does variables impact it, team used a multivariate model called the Cox Model. Its interest was the adaptation of the follow up at each patient.
First of all, a non-parametric estimator called Kaplan Meier was used to compare for each variable overall survival. Then to obtain unrefined hazard ratios (HR), using Cox’s model a multivariate analysis was performed. At last, to compare the association between the variables ( others being equal ) a multivariate analysis was used (the Cox’s model) to calculate the adjusted HR.


Ethics committee of the « Université de Bretagne Occidentale » Faculty of medicine approved the study.

Table of contents :

Abstract .
study design
Study population
Study endpoint
Data cleaning
Statistical analysis
Study population
Status at twelve months
Survival analysis : research for decompensation risk factors
Univariate analysis
Multivariate analysis
Encountered difficulties
Results analysis
Encountered difficulties
Study limitation
Future prospects


Related Posts