Bias Confonder PG Lecture Series

Dr. J Kishore Professor MAMC, New Delhi; drjugalkishore@gmail.com

Bias and Confounders

J KishoreMBBS, MD, PGCHFWM, PGDEE, MSc, MNAMS, FIPHA

What is Research?

• Research is a quest for knowledge through diligent search or investigation or experimentation aimed at the discovery and interpretation of new knowledge.

• Lets take one example

Increased Susceptibility

Ingestion of Cholera vibrio

Cholera

Exposure to contaminated

Effect of cholera toxins on bowel

wall cells

Genetic Factors

Poverty

Malnutrition

Crowded housing

Risk factors for cholera Mechanisms for cholera

Healthy

PersonAgent

Exposure to contaminated water

Effect of cholera toxins on bowel wall cellsGenetic Factors

Poverty

Malnutrition

Crowded housing

What are the causes of Susceptibility of healthy

What are Mechanisms

Disease Complications

What Complications?Why?

How can be prevented?

“An event, condition or characteristic that preceded the disease event and without

which the disease event either would not have occurred at all or would not have

occurred until some later time.”

Rothman, 1998

Necessary, sufficient and component causes

Observed association, could it be:

Assessing the Assessing the relationship relationship

between a possible between a possible cause and an cause and an

outcomeoutcome

Selection or measurement bias

Confounding

Chance

Causal

Probably Not

Fact = Truth + Error

(Systematic) (Random)

Bias Chance

Research Method•Type of study•Randomization•Stratification•Blinding

Statistics• p value• confidence interval• Sample size

Validity of Research

• It is the degree to which a research /test / survey measures what it is intended to measure.

• It is the ability of research to detect the truth (disease/health event/finding).

THREATS TO VALIDITY

A study’s internal validity, or how close its findings are to the TRUTH, can be compromised by three things….

• BIAS• 3rd VARIABLES (CONFOUNDING or

INTERACTION)

• CHANCE

Relationship between bias and chance

80 90Diastolic Blood Pressure (mm Hg)

No. of Observations

True BloodPressure (Intraarterial)

Blood Pressure (Sphygmomanometer)

Non Differential Misclassification

Chance

FaultyInstrument/

Observation

Sample Sample

All patients with the conditions of interest

Study on Sampled Patients

Selection

MeasurementConfounding

Chance

CONCLUSIONGeneral Population Having Normal &

Diseased persons

Internal Validity and Generalizability (External Validity)

Internal ValidityExternal Validity

External Validity/Generalizability

Internal Validity

Cross Sectional Studies

Case Control Studies

Cohort Studies

Non-Randomized Studies

Randomized Controlled Trial

What is bias?

• Prejudice; deviation from truth; the concept of bias is the lack of internal validity or incorrect assessment of the association between an exposure and an effect in the target population. Result of such studies can not be extrapolate to general population. There is no external validity without

internal validity.

Types of biases

• There may be more than 100 biases, however, for the understanding they are simply divided in to:

• Selection Bias

• Information Bias

References: J Kishore. A Dictionary of Public Health 2007

Systematic Errors

Selection Bias

• Ascertainment Bias is systematic error resulting from failure to identify equally all categories of individuals who are supposed to be represented in a group, e.g., study based on specialty hospital.

• Non-Random Sampling bias, e.g. non-representative sample.

• Healthy worker effect is another selection bias which may take place when employed are compared with unemployed.

A faulty assumption that occurs because there are systematic differences

in characteristics between those who are selected for study and those who are not.

Selection Bias

Information Bias

• occurs during data collection. The main types of information bias are:

• 1. Misclassification Bias

• 2. Observer/interviewer Bias

• 3. Recall Bias

• 4. Reporting Bias

• 5. Other information biases: Hawthorne effect, loss to follow up,

MISCLASSIFICATION BIAS

Misclassification Bias: the erroneous classification of an individual, a value, or an attribute into a category other than that to which it should be assigned

• often results from an improper “cutoff point” in disease diagnosis or exposure classification;

• Originated when sensitivity and or specificity of the test is low.

• 2 types of misclassification bias– differential (systematic)– non differential (random)

Differential misclassification

• Misclassification is different in groups to be compared

• e.g. recall of exposure in disease patients and control. Disease patients will try to recall the exposure more than control group. This will result in greater association in disease and exposure, i.e., moving away from null hypothesis.

Differential Misclassification

"True Situation"CasesControls Total

Exp. 85 40 125Not Exp. 15 60 75Total 100 100 200

85X60/15X40 OR= 8.5

30% unexposed misclassified as exposed in cases, 50% of exposed misclassified as unexposed in controls

Cases ControlsExp. 85 + 5 20Not Exp. 10 60 + 20Total 100 100

90X80/10X20 OR=36.0Bias away from the null (1.0)

MISCLASSIFICATION BIAS

• Non-differential Misclassification Bias

rate of misclassification does NOT differ between study groups (Cases and Control)– all study groups and control equally

susceptible to inaccuracy in classification, e.g., faulty instrument applied to all disease and control group

– dilutes the true association– RR or OR shifts towards null (1.0)

Non-Differential Misclassification

"True Situation"CasesControls Total

Exp. 85 40 125Not Exp. 15 60 75Total 100 100 200

85X60/15X40 OR= 8.5

50% of exposed misclassified as unexposed

Cases ControlsExp. 43 20Not Exp. 15 + 42 60 + 20Total 100 100

43X80/57X20 OR=3.0

Bias towards the null (1.0)

CONTROLLING FOR MISCLASSIFICATION BIAS

• improving sensitivity and specificity of diagnostic tests– raises or lowers the “cutoff point” for

diagnosis

• increasing the completeness of medical records

• multiple questions that ask same information – acts as a built in double-check

• multiple checks in medical records – gathering diagnosis data from multiple

sources

Confounding

Now Lets take one situation

Research QuestionIs down syndrome associated with

birth order?

Cases of Down syndroms by birth order

1 2 3 4 5

Birth order

Cases per 100 000 live births

• Could there be another factor (confounding) responsible for Down syndromes?

What is “Confounding”?

• From the Latin confundere, to mix together

• “The distortion of the apparent effect of an exposure (e.g. Birth Order) on risk (e.g. Down Syndrome) brought about by the association with other factor[s] (e.g. age of mother) that can influence the outcome”

• A Dictionary of Epidemiology by John Last, 1995.

Cases of Down Syndrom by age groups

0100200300400500600700800900

< 20 20-24 25-29 30-34 35-39 40+

Age groups

Cases per 100000 live

births

What is “Confounding”?, continued...

• Confounding Principles:– Indirect (spurious or weaker than we think)

association (A1) between exposure (E) and disease (D) that depends on A2 [and is weaker than direct association (A3) between confounder and disease]

– Confounder (C) must associate with both E & D

Birth Order(E)

Down syndrome(D)

Age of mother(C)

0100200300400500600700800900

Cases per 100000

1 2 3 4 5

Birth order

Cases of Down syndrom by birth order and mother's age

EXPOSURE

(coffee drinking)

DISEASE

(heart disease)

CONFOUNDING VARIABLE

(cigarette smoking)

Does night light increase the risk of Myopia?

Nightlight exposure before age 2 yrs

Myopia children

Myopic Parents

Quinn et al 1999

Zadnik et al 2000, Gwiazda et al 2000

DiseaseOutcome

Exposure

?Confounder

Common feature of Confounder

EXAMPLE:Is maternal smoking a risk factor of perinatal death?Is the association confounded by low birth weight?

Perinatal mortality

Maternal smoking

?Low birth

weight

OR RATHER:Is low birth weight the reason why maternal smoking is associated to higher risk of perinatal death?

Perinatal mortality

Maternal smoking

Low birthweight

BUT THERE COULD BE AN ADDITIONAL QUESTION:Does maternal smoking cause perinatal death by mechanisms other than low birth weight?

Perinatal mortality

Maternal smoking

Low birthweight

Direct toxic effect?

Block by adjustment

Confounding

Criteria for a Confounding Factor ( Rothman and Greenland, 1998):

1.A confounding factor must be a risk factor for the disease.

2.A confounding factor must be associated with the exposure under study in the source population (the population at risk from which the cases are derived). Typically, (for a rare disease) check this condition by looking for an association in the control group.

3.A confounding factor must not be affected by the exposure or the disease. In particular, it cannot be an intermediate step in the causal path between the exposure and the disease.

Control of ConfoundingApproaches

1. Design Aspects of a StudyRandomization – used in experimental design to allocate individuals

into experimental groups.Restriction – selecting subjects with equal values for variables

which might be confounders (can use complete restriction or partial restriction known as matching).

2. Analytic MethodsStratification – using homogeneous categories; using Direct, Indirect stratification, Mantel Haenszel tests,

Model Fitting – such as regression models-Linear regression, Logistic Regression, etc.

Coping with confounders

Matching (mostly in case control studies)

Selection of cases and controls with matching values of the confounding variable

Pair wise matchinge.g in coffee drinking study as a predictor of MI, each case (a

patient with MI) could be matched with one or more controls that smoked roughly the same amount as the case (10-20 cigarettes/day)

Matching

Advantages: Can eliminate influence of strong confounders Can increase precision (power) by balancing the number

of cases and controls in each stratum May be sampling convenience making it easier to select

controls

Matching

Disadvantages Time consuming Requires early decision as to which variables are

predictors and which are confounders Requires matched analysis Creates the danger of over matching( matching on a

factor which is not a founder, thereby reducing power)

Comparing control of bias in RCT & NRS

Source of bias RCTs Cohort/other NR studies

Comments

Selection bias Randomi-zation

Control for confounders

Many other study-specific threats

Performance bias

(Exposure)

Double-blinding

Exposure measurement

Misclassifica-tion/ non-comparability

Attrition bias Complete F/U

Complete F/U What amount is critical?

Detection bias

(Outcome)

Masked assessmt

Masked assessment

Misclassifica-tion

Hill criteria - used to evaluate possibility of causation;

(from Rothman, Modern Epidemiology, 2nd ed., 1998)

1. Strength of association (magnitude of effect): Measured as relative risk or Odds ratio?

2. Consistency - more an argument against associations due to chance. Has effect been seen by others?

3. Specificity - idea that a cause should lead to a single effect - not tenable.

4. Temporality: Did exposure precede outcomes?5. Biologic gradient (dose-response)6. Biologic plausibility: Is the association make sense?7. Coherence (also biologic)—with natural history of

disease/condition. Is the association consistent with available evidence?

8. Experimental evidence: Has a randomized controlled trials been done?

9. Analogy — parallels exposure-outcome relationship for similar exposure and/or disease. Is an association similar to others?

It’s the beginning !We have a

long way to goThanks- Dr. J Kishore

Bias Confonder PG Lecture Series

Documents

Lecture notes I: Measurement...

Physics 101: Lecture 3, Pg 1 Kinematics Physics 101: Lecture...

Bias and Variability Lecture 28 Section 8.3 Tue, Oct 25,...

Physics 1304: Lecture 18, Pg 1 AC Circuits Physics 1304:...

CONFOUNDING & BIAS LECTURE [M.Tevfik DORAK] · PPT file ·...

Physics 207: Lecture 27, Pg 1 Dec. 1 Physics 207: Lecture...

Omitted Variable Bias Methods of Economic Investigation...

Introductory PG Lecture - How To Make Your Customers Convert...

Lecture 11: Externalities and Public...

Last lecture summary Bias, Bessel's correction MAD Normal...

Bias Sam Bracebridge. By the end of the lecture fellows will...

Physics 101: Lecture 22, Pg 1 Physics 101: Lecture 22 Sound....

PG Lecture Hall (Admin) - University of...

Spring 2007EE130 Lecture 17, Slide 1 Lecture #17 OUTLINE pn....

Media Bias Lecture

Lecture 17: Prevention of bias in RCTs