Marital Status and Mortality Among First Acute MI Survivors
Marital Status and Mortality Among First Acute MI Survivors
The population-based Augsburg Myocardial Infarction Registry began continuously registering all cases of coronary deaths and non-fatal AMI in 1984 within the framework of the MONICA (Monitoring trends and determinants on cardiovascular diseases)-project. The registry has been part of the KORA (Cooperative Health Research in the Region of Augsburg) framework since 1995. The data covers the 25–74 year old population in the city of Augsburg and the two adjacent districts located in southern Bavaria, Germany (totalling 600,000 inhabitants). Patients hospitalised in 8 hospitals within the study region and 2 hospitals in the surrounding areas are included. Approximately 80% of all AMI cases in the study region are treated in the study region's major hospital, "Klinikum Augsburg", a tertiary care centre offering invasive and interventional cardiovascular procedures, as well as heart surgery facilities. The study was approved by the Ethics Committee of the Bavarian Medical Association. All participants submitted written informed consent before being enrolled in the study. Methods of case identification, diagnostic classification of events, and data quality control have been described elsewhere.
All patients registered between January 1, 2000, and December 31, 2008, who survived longer than 28 days after a first AMI were initially included into the sample. The follow-up was continued until August 26th 2010. Among 4,405 men and women who survived 28 days after their first AMI, we excluded 298 patients without information on marital status. Additionally, 341 individuals with incomplete covariate data were excluded. The final sample size was 3,766 persons aged 28 to 74 years.
Study participants were interviewed with a standardised questionnaire during their hospital stay after being transferred from the intensive care unit. The interviews were performed by trained study nurses and covered demographic information, risk factors and co-morbidities. Marital status and living arrangements were assessed via interview, and dichotomized as married or unmarried, and living alone or living together with someone, respectively. Applied and recommended compositions of marital status and co-habitation as predictors of long-term disease and mortality outcomes are heterogeneous throughout the literature. Consequently, we let the most suitable explanatory variable composition be defined by the properties of the available data. Socio-economic status (SES) was assessed as the minimal years of schooling required for the highest school degree attained. Based on the German educational system, nine years was defined as low SES while more than nine years was defined as 'higher SES'. Risk factors and co-morbidities were determined either by patient report (smoking) or by chart review (e.g. hyperlipidemia, diabetes, stroke). Information on AMI characteristics and treatment (e.g. AMI type, type of reperfusion therapy) as well as in-hospital complications (e.g. cardiac arrest, pulmonary oedema) were also determined by chart review.
The study's end point was long-term all-cause mortality among patients after their first AMI. Mortality was assessed by checking the vital status of all registered persons in the KORA AMI registry through the population registries, inside and outside the study area in 2010. This procedure guaranteed that the vital status of cohort members who had moved outside of the study area could also be assessed. Death certificates were obtained from local health departments.
Relevant explanatory variables were determined through a literature search and grouped into: socio-demographic factors, patient history, behavioural factors, clinical parameters, treatment and in-hospital complications. After initial descriptive analyses, all explanatory variables were subjected to bivariate Log-Rank tests against 10.7-year survival. Variables which were significant at an alpha-level of 0.2 were included into a multivariate Cox regression model fitting. Marital status was the primary, binary, explanatory variable of interest. Co-habitation was examined in sensitivity analyses. Variation in the time of study entry was adjusted for by inclusion of the variable "recruitment day". Recruitment day was defined as the number of consecutive days between December 31, 2000 and the recruitment date. Initial model fitting was concluded via manual backward elimination using a sample of n = 3,398 complete cases. Marital status, sex, age-group and recruitment day were forced into the model. Based on the results of bivariate Log-Rank tests, SES, hypertension, angina pectoris, hyperlipidemia, stroke, diabetes, obesity, ST-elevation myocardial infarction, bundle-branch-block, cardiac arrest, pulmonary oedema, and reperfusion therapy were introduced into the model fitting procedure. Among these, obesity, SES, and ST-elevation were removed from the model as they failed to show significant effects on 10.7-year survival. Finalized model fitting was computed using the maximal available sample of n = 3,766 after omitting SES which reduced missing values by 368. Finalized model fitting yielded results identical to the initial analyses. The proportional hazards assumption was examined in the full model using the correlation of Schönfeld residuals against observation-time and squared observation-time for each explanatory variable, respectively. Violations of the proportional hazards assumption were observed for reperfusion therapy and hypertension. Interaction terms with observation time showed p-values of <0.0001 and 0.0154 respectively. Time-dependencies were incorporated into the model by introducing significant interaction terms. Multicollinearity in the covariate structure of the fully adjusted model was ruled out as variance inflation factors were below 1.2 for all explanatory variables. All explanatory variables in the full model were tested for interaction with marital status. Both minimally and fully adjusted models were run as analyses stratified by variables which significantly interacted with marital status (age-group, hyperlipidemia). In a final step, fully adjusted, stratified models were run for different survival cut-offs from one to ten years. All tests within the multivariate model fitting were conducted at an alpha level of 0.05.
Several sensitivity analyses were performed. First, a model without the information loss produced by dichotomization of the continuous variable age was compared with the main analyses. Second, the fully adjusted model was rerun stratified only by hyperlipidemia with age introduced as a continuous variable and the inclusion of an interaction term between age and marriage, as shown in Figure 1. Third, since a combination of marital status with co-habitation produced strata with too infrequent events for stable multivariate analyses, the stratified, fully adjusted model was rerun with co-habitation instead of marital status as the main explanatory variable to confirm any observed associations.
(Enlarge Image)
Figure 1.
Associations between age in years and the effect of marital status on 10.7-year survival – fully adjusted model stratified by hyperlipidemia.
All Cox modelling procedures were performed with the exact method assuming an existing order in tied measurements. All statistical analyses were performed using SAS software, release 9.2 (SAS Institute, Cary, NC).
Methods
The population-based Augsburg Myocardial Infarction Registry began continuously registering all cases of coronary deaths and non-fatal AMI in 1984 within the framework of the MONICA (Monitoring trends and determinants on cardiovascular diseases)-project. The registry has been part of the KORA (Cooperative Health Research in the Region of Augsburg) framework since 1995. The data covers the 25–74 year old population in the city of Augsburg and the two adjacent districts located in southern Bavaria, Germany (totalling 600,000 inhabitants). Patients hospitalised in 8 hospitals within the study region and 2 hospitals in the surrounding areas are included. Approximately 80% of all AMI cases in the study region are treated in the study region's major hospital, "Klinikum Augsburg", a tertiary care centre offering invasive and interventional cardiovascular procedures, as well as heart surgery facilities. The study was approved by the Ethics Committee of the Bavarian Medical Association. All participants submitted written informed consent before being enrolled in the study. Methods of case identification, diagnostic classification of events, and data quality control have been described elsewhere.
Sample
All patients registered between January 1, 2000, and December 31, 2008, who survived longer than 28 days after a first AMI were initially included into the sample. The follow-up was continued until August 26th 2010. Among 4,405 men and women who survived 28 days after their first AMI, we excluded 298 patients without information on marital status. Additionally, 341 individuals with incomplete covariate data were excluded. The final sample size was 3,766 persons aged 28 to 74 years.
Data Collection and Endpoints
Study participants were interviewed with a standardised questionnaire during their hospital stay after being transferred from the intensive care unit. The interviews were performed by trained study nurses and covered demographic information, risk factors and co-morbidities. Marital status and living arrangements were assessed via interview, and dichotomized as married or unmarried, and living alone or living together with someone, respectively. Applied and recommended compositions of marital status and co-habitation as predictors of long-term disease and mortality outcomes are heterogeneous throughout the literature. Consequently, we let the most suitable explanatory variable composition be defined by the properties of the available data. Socio-economic status (SES) was assessed as the minimal years of schooling required for the highest school degree attained. Based on the German educational system, nine years was defined as low SES while more than nine years was defined as 'higher SES'. Risk factors and co-morbidities were determined either by patient report (smoking) or by chart review (e.g. hyperlipidemia, diabetes, stroke). Information on AMI characteristics and treatment (e.g. AMI type, type of reperfusion therapy) as well as in-hospital complications (e.g. cardiac arrest, pulmonary oedema) were also determined by chart review.
The study's end point was long-term all-cause mortality among patients after their first AMI. Mortality was assessed by checking the vital status of all registered persons in the KORA AMI registry through the population registries, inside and outside the study area in 2010. This procedure guaranteed that the vital status of cohort members who had moved outside of the study area could also be assessed. Death certificates were obtained from local health departments.
Statistical Analyses
Relevant explanatory variables were determined through a literature search and grouped into: socio-demographic factors, patient history, behavioural factors, clinical parameters, treatment and in-hospital complications. After initial descriptive analyses, all explanatory variables were subjected to bivariate Log-Rank tests against 10.7-year survival. Variables which were significant at an alpha-level of 0.2 were included into a multivariate Cox regression model fitting. Marital status was the primary, binary, explanatory variable of interest. Co-habitation was examined in sensitivity analyses. Variation in the time of study entry was adjusted for by inclusion of the variable "recruitment day". Recruitment day was defined as the number of consecutive days between December 31, 2000 and the recruitment date. Initial model fitting was concluded via manual backward elimination using a sample of n = 3,398 complete cases. Marital status, sex, age-group and recruitment day were forced into the model. Based on the results of bivariate Log-Rank tests, SES, hypertension, angina pectoris, hyperlipidemia, stroke, diabetes, obesity, ST-elevation myocardial infarction, bundle-branch-block, cardiac arrest, pulmonary oedema, and reperfusion therapy were introduced into the model fitting procedure. Among these, obesity, SES, and ST-elevation were removed from the model as they failed to show significant effects on 10.7-year survival. Finalized model fitting was computed using the maximal available sample of n = 3,766 after omitting SES which reduced missing values by 368. Finalized model fitting yielded results identical to the initial analyses. The proportional hazards assumption was examined in the full model using the correlation of Schönfeld residuals against observation-time and squared observation-time for each explanatory variable, respectively. Violations of the proportional hazards assumption were observed for reperfusion therapy and hypertension. Interaction terms with observation time showed p-values of <0.0001 and 0.0154 respectively. Time-dependencies were incorporated into the model by introducing significant interaction terms. Multicollinearity in the covariate structure of the fully adjusted model was ruled out as variance inflation factors were below 1.2 for all explanatory variables. All explanatory variables in the full model were tested for interaction with marital status. Both minimally and fully adjusted models were run as analyses stratified by variables which significantly interacted with marital status (age-group, hyperlipidemia). In a final step, fully adjusted, stratified models were run for different survival cut-offs from one to ten years. All tests within the multivariate model fitting were conducted at an alpha level of 0.05.
Several sensitivity analyses were performed. First, a model without the information loss produced by dichotomization of the continuous variable age was compared with the main analyses. Second, the fully adjusted model was rerun stratified only by hyperlipidemia with age introduced as a continuous variable and the inclusion of an interaction term between age and marriage, as shown in Figure 1. Third, since a combination of marital status with co-habitation produced strata with too infrequent events for stable multivariate analyses, the stratified, fully adjusted model was rerun with co-habitation instead of marital status as the main explanatory variable to confirm any observed associations.
(Enlarge Image)
Figure 1.
Associations between age in years and the effect of marital status on 10.7-year survival – fully adjusted model stratified by hyperlipidemia.
All Cox modelling procedures were performed with the exact method assuming an existing order in tied measurements. All statistical analyses were performed using SAS software, release 9.2 (SAS Institute, Cary, NC).
Source...