EpiNotes

= Notes on Epidemiology and Statistics =

We continue with the materials on Epidemiology and Statistics that we posted yesterday. The plan from now on is to post short focused posts on different aspects of health research relevant for the week under discussion on a daily basis. Please join in the discussions and add your views.

There can be one or two posts per day. Today in the first post, I shall discuss my reflections on the paper by Paolo Vineis that I posted yesterday and provide another framework for understanding cause and effect relationships in health care research. This paper is by Sir Austin Bradford Hill, the British Epidemiologist Occupational Physician. I have also included a commentary on the paper by Sander Greenland for your review.

As an accompanying reading material, I have put together some notes on statistics and epidemiology to go with it. We start today's discussions with my reflections on Paolo Vineis' text and then move on to describe how one can compare prevalence and incidence rates and introduce concepts of indirect and direct standardizations, then we talk about probability and some principles of statistics. These notes are fairly low level notes.

Please read these notes and please send your comments if you can. Let's discuss these issues. I shall post some key lessons every day for your reading. Please reflect on these topics, read the additional lessons.

Notes on using the wikispaces
Starting with this note, I have started sending these notes both as email and in a wiki format. To make things easy, I have created separate student accounts for each one of you so that you can enter the wikispaces wiki without any hassle. Just use the user name and password assigned to you and you will be up and going. You can of course create your own username and password for the wikispace if you want as well. If you want your own username, please let me know. Please feel free to add your comments on this space if you want.

My reflections on Paolo Vineis Text on Causality
In this article, Paolo Vineis provides a brief introduction to some of the key issues around cause and effect deduction in medicine and health sciences.

An interesting point in this article is his perspectives on the nature of epidemiology and how cause and effect relationships between entities have defined the scope of epidemiology itself. Vineis defines Epidemiology as a cross between natural and social sciences and indeed has defined the role of bias and confounding much better than social sciences.

As he writes, historically, the notions of causality in general has evolved through the phases of mono-causal theories where a single component cause was considered to be both necessary and sufficient for causation. This was the era of microbiology. For example, small pox virus for small pox (indeed any bacterial/viral/fungal/biological agent) for any defined outcome. Note that such notions of cause and effect relationships still persist in medicine and biology (think of how genes and studies on genetics and genomics are used to make statements about cause of specific diseases).

However, with chronic disease, and diseases such as the ones with strong environmental component, one cannot discuss a single component cause, and one needs to necessarily think of more complex web of causality. Each causal factor or causal variable is like a slice of a pie or a spoke in a wagon wheel (see Figure 1). A really good reading is the work on causality by John Mackie and articles by Sander Greenland on how component causes model have influenced Epidemiology.

So, what is the net take away result in Epidemiology and health research of these issues around finding causal factors for Epidemiological and health resarch related studies? As we shall see, risk factors are calculated by computing Relative Risks, and Odds Ratio of association between an exposure and an outcome (or an intervention and an outcome) and then based on these figures, relative reduction of the outcome risks are calculated. These figures enable one to predict the extent to which the risk of a specific outcome can be reduced in the population (as well as for individuals) if the specific exposure variable could be eliminated. This forms the basis of much public health action.

In the next post of the day, I shall discuss basics of Epidemiology and some issues around statistics. I shall also attach for you an article by Sir Austin Bradford Hill for your review and critique. This has been "wrongly" stated by some as Hill's Criteria.