Our analysis of implantable cardiac defibrillator (ICD) insertions in NSW 2005–2020 is now out in @theMJA : open access 👇 https://t.co/C0odug4Wn6 led by @DrLinZhu with @AndrewHayen Bianca Blanch, Nathan Engstrom, @jadoust , Chris Semsarian, and me.

For this Systematic Review, we set out to estimate the incremental value of coronary artery calcium scans, beyond that provided by traditional cardiovascular risk scores (e.g. Framingham). @DrLinZhu, @2weekSR @PaulGlasziou @wiserhealthcare @syd_health @Institute4EBH 1/6

Out today in @IntJEpidemiol https://t.co/5FPea3FbS4 with Dr Gabriel Gregory, @DrLinZhu @AndrewHayen Pandemic response led to a drop in Australian deaths in 2020

We are deeply saddened to hear of the passing of Sir David Cox. As well as his immense contributions to statistical research, he'll be remembered as a boundlessly generous and supportive friend to generations of statisticians. https://t.co/pTXVi4xpCl

Measurement variation in fasting plasma glucose and HbA1c can cause prediabetes and diabetes #overdiagnosis when repeat screens are done in healthy people. May also explain some apparent“progression” or “regression” of either condition. Our latest in @JClinEpi👇

So we can expect reporting a conditional (adjusted) OR to has higher power for a given sample. However, I think to adjust or not depends on if the conditional effect is of interest. We should not choose to report a conditional OR just for the purpose of increasing power.

For the measures such as OR, marginal(unadjusted) and conditional(adjusted) estimates are different(unless OR=1) due to a lack of collapsibility(in the absence of confounding). It has been reported marginal OR is always shifted toward the null compared to the conditional OR.

Excited to share our findings on the ACCORDION study: fibrate add-on therapy during the trial period was associated with a legacy benefit of improved survival in diabetic patients with dyslipidemia. @AndrewHayen @KatyJLBell @UTS_PH @UTS_Health https://t.co/3vB12uk5YE