The Influence of Lunar Cycle On Hospital Outcomes Following Ascending Aortic Dissection Repair

The primary objective of this study is to determine whether the findings of Shuhaiber et al. [2013] (see citations) may be confirmed using national audit data from the United Kingdom.

A cohort of all acute ascending aortic dissection repair surgeries will be extracted from the National Adult Cardiac Surgery Audit (NACSA) database. The current dataset version 4.12, available for download from the website of the National Institute for Cardiovascular Outcomes Research (see links), contains 168 fields that detail pre-operative patient characteristics, operative details and post-operative outcomes. The dataset is revised intermittently to meet the requirements of users and to respond to changes in the management of adult cardiac surgery; however, the fields to be analysed in this study have not been subject to any documented revisions. The extracted data are to be cleaned using version 9.9 of the NACSA cleaning algorithm.

Seasons are to be coded in the same way as in the published study: autumn (September 21 through December 20), winter (December 21 through March 21), spring (March 22 through June 21) and summer (June 22 through September 20).

Three definitions are to be used for the phases of the moon. The first definition mirrors that employed in the motivating work: each lunar cycle from new moon to new moon is to be split into four parts and labelled "New", "Waxing", "Full" and "Waning". The second definition centers each of those four labels around the named event: i.e., the "New" period is to be centered around the new moon event itself, instead of starting on the day of a new moon as in the first definition. The third definition will split the lunar cycle into 8 event-centred periods, called "New", "Waxing crescent", "First quarter", "Waxing gibbous", "Full", "Waning gibbous", "Last quarter" and "Waning crescent".

There is no need to explore the individual univariate relationships of predictors to outcomes in this work, given its limited purpose to confirm the earlier multivariate findings. The relationship between the environmental factors (seasons and the lunar cycle) and in-hospital mortality will be examined using multivariate logistic regression, controlling for demographic and mortality risk factors. The specific variables that can be included in common with the earlier work are age, sex, hypertension, diabetes, renal failure, and an indicator of weekend admission. The national audit does not include information on dyslipidaemia, but it is felt that its exclusion from our model will not effect the comparability of our work to the earlier study because none of the earlier published models showed any effect of dyslipidaemia on surgical outcomes. The relationship of the environmental factors to Length of stay (LOS) in survivors will be examined using a multivariate linear regression that includes the same risk factors as in the mortality model. As in the published study, LOS will be transformed to the natural logarithmic scale prior to analysis.