Events and Seminars : 2014 Seminars

MAKING CONTINUOUS OUTCOMES CLINICALLY MEANINGFUL

JANET PEACOCK, PH.D.
Professor of Medical Statistics
Department of Primary Care and Public Health Sciences
King’s College, London
THURSDAY, DECEMBER 11
11:00 a.m.–12:00 p.m., CRB 692

Continuous outcomes are sometimes dichotomised by clinicians to help decision-making. For example when systolic blood pressure is dichotomised at 160 mmHg to help decide if treatment should be given. Epidemiologists also dichotomise continuous outcomes such as birth weight reported as<2500g (low birth weight) —-a measure of the percentage of women at high risk of poor pregnancy outcome in a population. Biostatisticians generally criticise the practice of dichotomisation since it discards information, and loses power.

In this talk I’ll describe some collaborative work that provides a way to report a difference in means and its equivalent difference in proportions below (or above) a given cut-point, without losing power. This ‘distributional approach’ promotes clinical usefulness and statistical rigour. I’ll talk about extensions that we’re currently developing to deal with skewed data and to adjust for other factors eg using mixed models.

Refs: Peacock JL, Sauzet O, Ewings SM, Kerry SM. Statistics in Medicine2012; Sauzet O, Peacock JL. Statistics in Medicine. 2014