Purpose: To determine the influence  of patient characteristics (age, sex, family history, and disease burden) on  medical students' estimation of life expectancy in elderly patients before and after an educational intervention.

Methods: In 2005 and 2006, 107 fourth year medical students were shown 7 case scenarios with varied age, sex, family history, and disease burden (Table 1). For each scenario, students estimated patient life expectancy in years. The lecturer (PAB) then discussed cancer screening strategies in the elderly, using population-level data on life expectancy stratified by age, sex and health status as a guide.   Students again scored the same 7 cases. A mixed model analysis of variance was used to evaluate the contribution of age, sex, family history of cancer, and overall disease burden to students' judgments of prognosis as measured by life expectancy. Results: Student estimates of life expectancy varied considerably and appropriately by disease burden on a univariate analysis and changed minimally post-lecture. A multivariate analysis of age, sex, family history, and disease burden revealed that all four cues were significantly related to estimated life expectancy before the lecture(p<0.001), and the relationship between these cues and life expectancy did not change after the lecture (p>0.169). Discussion: Student estimates of life expectancy incorporate multiple cues, some of which (family history of cancer) are not useful predictors of prognosis in aged populations.  Though student estimates were fairly well calibrated to actual prognosis, more education about probabilistic reasoning using population health statistics is warranted. Table 1 Distribution of Cues and Mean Life Expectancy (LE) Estimates in Cases

Case	Age	Family history	Sex	Disease burden	LE – Pre Lecture	LE-Post Lecture
1	70	+	M	L	14.31	14.37
2	70	+	F	L	11.75	12.47
3	75	+	M	H	4.87	5.29
4	75	+	F	H	5.02	5.86
5	80	-	F	H	4.32	5.22
6	80	-	F	L	12.14	12.13
7	80	+	M	L	7.45	7.08