Referee 3's Review (Merete Osler):

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication, the manuscript has been revised in light of these comments and to address other editorial requirements.
*********

This interesting correlation study addresses a relevant public health issue, whether educational inequalities in mortality from all and specific causes of death in adults aged 25-64 years have increased in US from 1993 to 2001. Thus, it is an update of earlier studies.

The study is based on information on level of school education obtained from death certificates and from census surveys for the background population.
The main finding is that educational inequalities in mortality have increased during the study period.

The data has been analysed with relevant statistical methods and the authors describe and discuss the data restrictions made and study limitations due to changes over time (in ICD-coding and the classification of education) or differences in how complete the two data sources are and how they have classified education.

However, there are some additional points I would like the authors to consider:

1. The Introduction is very brief. It would be nice to have some more information on why it is of interest to monitor educational differences in mortality and the background for the focus on the 7 specific causes of death. Thus, it is not clear why the authors focus on death from cancer, heart disease, stroke, accidents, HIV, diabetes and COPD. Is it based on any a prior hypotheses? Further, it would be interesting to see data for some specific cancers rather than for total cancer.

2. The authors comment on excluding 7 states where more than 20% of education data was missing and briefly discuss how this might affect the results. However, the authors also restrict analyses to deaths in individuals aged 25-64. The majority of deaths occur in individuals aged 65 years and older. The authors state that educational attainment is more accurate in individuals aged 25-64, but the generalisability of their findings to the overall population that dies is unclear.

3. Table 1 gives number of deaths and death rates in relation to level of education in 4 categories and a rate ratio (+? 95%CI) for the lowest (< 12 y) vs. highest (16+ y) educated groups in 1993 and 2001 by race and gender. It is fine to see the graded relation between education and all-cause mortality. It would also be nice to have a measure of the absolute difference (rate difference) in addition to the rate ratio. Table 2 and 3 provides nearly same information for the 7 specific causes of death. It would provide a better overview if these two tables just gave the figures for the highest and lowest educated groups, a rate ratio and rate difference. The more detailed information could be available in Tables on the web

The figure does not add much information.

I find Table 4 rather hard to interpret. What is the purpose of showing contributions to increasing mortality when the trend is a decrease?

The Discussion of study findings is very general. The authors do not adequately comment on the fact that inequalities differ for the specific causes according to gender and race. This study actually reveals that mortality disparities are probably explained by a variety of factors that contribute to different degrees.

A study limitation to be discussed is that autopsy rate influence the validity of the specific diagnosis. Does this rate differ in relation to education and over time?

The conclusion is very general. It does not apply to the group of black women.