Presentation is loading. Please wait.

Presentation is loading. Please wait.

Date: Presenter: Ryan Chen

Similar presentations


Presentation on theme: "Date: Presenter: Ryan Chen"— Presentation transcript:

1 Date: 2017.01.10 Presenter: Ryan Chen
Multilevel mixed-effects parametric survival models Date: Presenter: Ryan Chen

2 Date: 2017.01.10 Presenter: Ryan Chen
Multilevel mixed-effects parametric survival models Date: Presenter: Ryan Chen

3 16 January 2014

4 Objective To investigate temporal changes in survival after acute myocardial infarction (AMI) by early invasive strategy. Background It is not known how survival has changed over time for patients who do and do not receive an early invasive strategy for the management of AMI. This study, therefore, aimed to investigate the temporal trends in hospital treatments, including revascularisation, and 6-month survival rates after AMI.

5 Study design Myocardial Ischaemia National Audit Project (MINAP) between 1 January 2003 and 31 July 2011. multicentre prospective registry of patients hospitalised in England and Wales.

6 Study design-2 We compared survival rates between groups of patients admitted to hospital for 2003–2004, 2005–2006, 2007–2008 and 2009–2010. We based our analyses on the diagnosis made at hospital discharge of the phenotype of AMI. We stratified the management of STEMI by the provision of acute reperfusion therapy (which was defined as PPCI and/or thrombolysis), and the management of NSTEMI by the provision of coronary angiography.

7 Covariates Validated mini-Global Registry of Acute Coronary Events (GRACE) 6-month risk score was calculated. For patients with AMI, treating hospitals were ranked based on quartiles (Q1–Q4) the number of coronary angiograms (but not primary PCI cases) the number of PPCI cases.

8 Statistical methods Survival time was calculated from the date of admission to hospital to the date of death up to 6 months. We identified an elevated risk of mortality within 30 days of AMI, and therefore survival was studied with adjustment for this early mortality using Weibull accelerated failure-time regression models (AFT model). PH model VS. AFT model

9 Statistical methods-2 Models were fitted by age group (<65, 65–80, >80 years), sex, type of AMI (STEMI vs NSTEMI) and by biennial period of admission to hospital (2003–2004, 2005– 2006, 2007–2008 and 2009–2010). Patients were clustered by hospitals; therefore, to allow for variations at the hospital level, random effects (shared frailty) models were fitted. Shared Frailty model VS. Stratified Cox model

10 Statistical methods-3 Regression model parameter estimates were represented by time ratios (TR) and 95% CI and can be construed as the relative change in survival over time. Relative survival as a measure of cause-specific survival. Excess mortality was considered as the mortality attributable to the index AMI.

11

12

13

14

15 CONCLUSIONS Survival rates after AMI have improved.
Whereas survival estimates for STEMI patients who did not receive reperfusion therapy were stable, they worsened for NSTEMI patients not receiving coronary angiography.

16

17 Mixed-effects survival models
contain both fixed effects and random effects. random effects are useful for modeling intracluster correlation. adjusting survivor functions accelerated failure-time (AFT) model Proportional hazards (PH) model. mestreg is suitable only for data that have been stset. appropriate for data exhibiting multiple records per subject and time- varying covariates

18 Mixed-effects survival models

19 Mixed-effects survival models

20 Mixed-effects survival models

21

22

23

24

25 Each test is identified by tid.
Occasion: indicate the times. E.g. 1,2,3,4 Treat: Major independent variable. Pid: Patients’ id.


Download ppt "Date: Presenter: Ryan Chen"

Similar presentations


Ads by Google