HSCI 330 Lecture Notes - Lecture 4: Statistical Parameter, Relative Risk, Cohort Study
27 May 2018
School
Department
Course
Professor
Module 4: 7-1 Validity
Primary Objective of Epidemiologic research: obtain a valid estimate of an effect measure of interest
Validity= The general issue of whether or not there are imperfections in the study design, the methods of data
collection, or methods of data analysis that might distort the conclusions made about an exposure-disease relationship.
Valid= no imperfections
Bias= Imperfections with distorted results from correct conclusion; A measure of the extent of distortion of conclusions
about an exposure-disease relationship.
3 general types of validity problems:
Berkson demonstrated case-control studies in exclusively hospital settings are subject to a type of seletion ias called
BERKSON’S BIAS
Arises b/c patients with 2 disease conditions or high-risk behaviours are more likely to be hospitalized than those with
only 1 condition
Those patients are over-represented in the study population when compared to community population
Information bias → misclassification of a toxic trace; bias could be avoided using laboratory techniques developed to
measure TCDD from blood serum
Using biologic markers in epidemiologic research will reduce misclassification and improve accuracy of study
measurements
Confounding bias → a distortion in the results caused by failure to take into account such additional variables
Concern with whether all relevant variables were taken into account
Need to ensure that any observed effect could not be explained by other additional variables
Summary:
Validity issues are illustrated by:
❖ Hospital-based case-control studies Berkso’s seletio ias.
❖ Job misclassification to assess TCDD exposure (information bias).
❖ Control of relevant variables in the Sydney Beach Users Study (confounding).
Center of the target (epidemiologic): A measure of effect, either a prevalence ratio or a prevalence odds ratio (i.e., the
blue dot or bulls eye), for the association between 2 variables in the target population
Validity VS. Precision
Validity and precision concerns 2 different sources of inaccuracy that can occur when estimating an exposure-disease
relationship (systematic error=validity VS. random error=precision)
❖ Validity concerns systematic error whereas precision concerns random error.
❖ Systematic and random error can be distinguished in terms of shots at a target.
Systematic error: a difference between what an estimator is actually estimating and the effect measure of interest.
Random error: a difference between any estimate computed from the study data and the effect measure actually being
estimated.
VALIDITY
PRECISION
-Systemic Error affects validity
- does not consider statistic inference, but rather
methodologic imperfections of the study design or
analysis.
-Random error affects the precision
-concerns statistical inferences about the parameter of
the population actually being aimed at.
-concerns how much individual variable there is from
shot to shot
-precision reflects SAMPLING VARIABILITY
-concern methodologic imperfections for study design or
analysis
find more resources at oneclass.com
find more resources at oneclass.com
Study Questions (Q7.1)
Consider a cross-setioal study to assess the relatioship etee aliu itake high ersus lo i oe’s diet ad the
prevalence of arthritis of the hip in women residents of the city of Atlanta between the ages of 45 and 69. A sample of
female hospital patients is selected from hospital records in 1989, and the presence or absence of arthritis as well as a
measure of average calcium intake in the diet prior to enter the hospital are determined on each patient.
1. What is the target population in this study?
❖ Women residents of the city of Atlanta between the ages of 45-69
2. What does the center of the target (i.e., the bulls-eye) represent in epidemiologic terms?
❖ Measure of effect, prevalence ratio or prevalence odds ratio for the target population
3. What do we mean by random error associated with this study?
❖ It is the difference between the odds ratio in the hospital sample and the odds ratio in the population of
hospital patients from which the sample is selected
4. What do we mean by systematic error associated with this study?
❖ Whether or not the odds ratio in the population of hospital patients is being sample is different from the
odds ratio in the target population
A Hierarchy of Populations
❖ There are a variety of populations to consider in any epi study.
❖ The sample is the collection of individuals from which the study data have been obtained.
❖ The study population is the collection of individuals that our sample actually represents and is typically those
individuals we can feasibly study.
❖ The source population is the group of restricted interest about which the investigator wishes to assess an
exposure-disease relationship. (ex. Specific city, community or occupation at risk for being a case)
o ALSO called the study base or target population
❖ The external population is a group to which the study has not been restricted but to which the investigator still
wishes to generalize the study results.
Study Questions (Q7.2)
Consider an epi study carried out in New York City (NYC) to assess whether obesity is associated with
hypertension in young adults. The investigators decided that it was not feasible to consider taking a sample from
among all young adults in the city. It was decided that fitness centers would provide a large source of young
NYC adults. A sample of subjects is taken from several randomly selected fitness centers throughout the city
and their blood pressure is measured to determine hypertension status.
1. What is the source population for this study?
2. What is the study population in this study?
3. Does the sample represent the study population?
4. Does the study population represent the source population?
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Primary objective of epidemiologic research: obtain a valid estimate of an effect measure of interest. Validity= the general issue of whether or not there are imperfections in the study design, the methods of data collection, or methods of data analysis that might distort the conclusions made about an exposure-disease relationship. Bias= imperfections with distorted results from correct conclusion; a measure of the extent of distortion of conclusions about an exposure-disease relationship. Berkson demonstrated case-control studies in exclusively hospital settings are subject to a type of (cid:858)sele(cid:272)tion (cid:271)ias(cid:859) called. Arises b/c patients with 2 disease conditions or high-risk behaviours are more likely to be hospitalized than those with only 1 condition. Those patients are over-represented in the study population when compared to community population. Information bias misclassification of a toxic trace; bias could be avoided using laboratory techniques developed to measure tcdd from blood serum. Using biologic markers in epidemiologic research will reduce misclassification and improve accuracy of study measurements.
