The "July Effect" and Relative Risk

Fig 1. A typical new intern
on his first day of wards

Click here to listen to a recording of Astrid's 5-minute talk on Relative Risk.

Everybody jokes about the spike in mortality which occurs when all the third-year residents leave and the new interns arrive.  Does the “July Effect” really exist?  Well, the Annals of Internal Medicine just published a systematic review of studies describing trainee changeover on clinical outcomes.  The results were heterogeneous but did support the existence of the “July Effect.”  Based on a couple of higher quality studies, they estimate that the relative risk increase for mortality (RRI) in the peri-changeover period is between 4.3% and 12.0%.

Relative risk (RR, AKA the "risk ratio,") is a way of relating an event to an exposure.  The event is the outcome of interest (e.g. disease or death) while the exposure is something we believe is related to the frequency of the event,  (e.g a treatment, a vaccination, an environmental factor, or, as in our example, admission in July).  The relative risk is simply the ratio of the incidence of the event in the exposed group to the the incidence of the same event in the non-exposed group - anothe rway of saying that is that it's the experimental event rate (EER) divided by the control event rate (CER).

You can easily calculate the relative risk using a table not disimilar to the one used in our last post to calculate sensitivity, specificity and PPV/NPV:

Using this table, RR = [(a / a + b) / (c / c + d)].

This seems pretty simple - but the Annals article reports its findings in terms of relative risk increase.  This is different than the relative risk, and refers to the size of the increase in risk relative to the baseline incidence.  So if the relative risk is (EER/CER), the relative risk increase is [(EER-CER)/CER].

Please remember, both relative risk and relative risk increase are comparative measures - they tell you nothing about the absolute magnitude of the effect.