Educational attainment and differences in relative survival after acute myocardial infarction in Norway: a registry-based population study

Background Although there is a broad societal interest in socioeconomic differences in survival after an acute myocardial infarction, only a few studies have investigated how such differences relate to the survival in general population groups. We aimed to investigate education-specific survival after acute myocardial infarction and to compare this with the survival of corresponding groups in the general population. Methods Our study included the entire population of Norwegian patients admitted to hospitals for acute myocardial infarction during 2008–2010, with a 6- year follow-up period. Patient survival was measured relative to the expected survival in the general population for three educational groups: primary, secondary and tertiary. Education, sex, age and calendar year-specific expected survival were obtained from population life tables and adjusted for the presence of infarction-related mortality. Results Six-year patient survivals were 56.3% (55.3–57.2) and 65.5% (65.6–69.3) for the primary and tertiary educational groups (95% CIs), respectively. Also 6-year relative survival was markedly lower for the primary educational group: 70.2% (68.6–71.8) versus 81.2% (77.4–84.4). Throughout the follow-up period, patient survival tended to remain lower than the survival in the general population with the same educational background. Conclusion Both patient survival and relative survival after acute myocardial infarction are positively associated with educational level. Our findings may suggest that secondary prevention has been more effective for the highly educated.

average follow-up of 3.4 years per person. I really enjoyed reading the paper. I have few questions for authors. 1. In results section authors mentioned that there are no indications of a socioeconomic gradient to the differences in relative survival for the age groups 40-49 and 50-54 years and, if anything, the pattern is mixed in these groups. How these findings can be explained? It suggest that AMI patient survival in younger age groups do not depend on education level? 2. Also, authors reported that expected survival rate is always higher for higher levels of education but without explanation. What are possible reasons for higher expected survival rate for higher levels of education, or opposite what are possible reasons for lower expected survival rate for lower levels of education? 3. Survival of AMI patients depend on many factors, not only on education. Is education independently associated with survival after AMI? Did treatment of AMI patients differ between educational groups? Is it possible, for instance that AMI patients with higher education get better medical treatment? Is there possibility for authors to adjust for comorbidities, revascularization or localization of MI? It should be discussed thoroughly and mentioned in study limitations

GENERAL COMMENTS
The authors use the first AMI in the study period if an individual had more than AMI. However,it is not clearfrom the description, if the authors are modelling incident AMI (i.e. first ever AMI) or first in the study period.
Did authors use only main hspital diagnosis or all diagnoses including supplementry diagnoses?
Due to lack of survival data for the years 2011-13 model-based predicted survival was used for the analysis, where the prediction was performed based on age, sex and education. This is considered a limitation. The research objective is relevant, however, data are not available to support this analysis As far as I understand data before 2008 in the Norwegian Patient Register was de-identified. Therefore, before 2008 data from the Norwegian National Patient Register could not be linked with other data sources. This is a limitation of the study as incident AMI in the study period (from 2010) is only possibe to identify beased on prevoius AMI's in the period 2008-9.
Authors refer to results in table 1 as rates, however, it seems to be proportions and not rates.

Patricia Morton
Rice University USA REVIEW RETURNED 15-May-2017

Methodological Review
This paper utilizes novel data to investigate the relationship between educational attainment and AMI survival. Although I applaud the authors' use of population data, I have some additional questions about the methods presented and ask for a bit more clarification.

Main comments
The authors utilize Poisson regression and cubic spline curve to predict survival estimates for 2011-2013 (p. 7). Since the authors have population data from years prior, why was a Bayesian smoothing technique not considered to generate these data?
On a related note, why did the authors decide on using a 6-year follow-up period, especially when data are not available past 2010? Is this related to mean AMI survival periods? Why is 2008-2010 used as AMI baseline?
For each of the results and tables presented, the authors discuss differences across survival parameters. However, it seems that they did not conduct any statistical tests for the parameters to determine whether these differences are statistically significant (e.g., is relative survival for the primary educational group actually different from other groups?). Did the authors test across parameters?
I think it would be helpful to stratify all of the Table 1 information by sex as well (replacing Table 1 or adding an additional table). The protective effect of education can vary by sex, as does AMI risk and survival. Also, depending on the results of stratifying Table 1 for men and women, perhaps the authors should also stratify the remaining tables by gender. This would aid in the paper's overall contribution as well as the discussion during which the authors mention how AMI survival differs for men and women (p. 14).
Given that occupational stress (not education) is highly correlated with cardiovascular disease and AMI, I think this paper would benefit from a stronger argument for the use of education instead of occupation.
Did the authors consider adjusting survival for experiencing multiple AMI, as this might influence survival and preventive subsequent AMI health behaviors may vary by education level.

Minor comment
Last sentence in Results section of Abstract is somewhat confusing.
Disclaimer: I was asked to review the methods of this paper. Therefore, I have not reviewed any conceptual and/or theoretical component of this paper.

VERSION 1 -AUTHOR RESPONSE
Author response -Version 1 We wish to thank the reviewers for the constructive comments and BMJ open for the opportunity to revise and resubmit our manuscript.
Editorial Requirements 1. We have added a sub-title describing the study design. 2. The mentioned sections have been removed.
Response to reviewer #1, Sean Randall's comments: Thank you for the positive remark.