Confounding: The Spoiler of Research Evidence | Maher El-Masri | TEDxUniversityofWindsor

not all that shines is gold as an

not all that shines is gold as an expression that sometimes we use to

expression that sometimes we use to

expression that sometimes we use to refer to things that look real when

refer to things that look real when

refer to things that look real when they’re actually not it might be true

they’re actually not it might be true

they’re actually not it might be true about people it might be true about

about people it might be true about

about people it might be true about places sometimes events but it’s also

places sometimes events but it’s also

places sometimes events but it’s also true about some research not all

true about some research not all

true about some research not all published research is gold and the truth

published research is gold and the truth

published research is gold and the truth is I’m a professor I’m a researcher I do

is I’m a professor I’m a researcher I do

is I’m a professor I’m a researcher I do research and teach research so don’t get

research and teach research so don’t get

research and teach research so don’t get me wrong research is very important I’m

me wrong research is very important I’m

me wrong research is very important I’m a very strong believer in the importance

a very strong believer in the importance

a very strong believer in the importance of research to advance informed decision

of research to advance informed decision

of research to advance informed decision making in practice professions it’s very

making in practice professions it’s very

making in practice professions it’s very important to advance evidence-based

important to advance evidence-based

important to advance evidence-based practice but research is not immune to

practice but research is not immune to

practice but research is not immune to error every research study is prone to a

error every research study is prone to a

error every research study is prone to a degree of error we refer to it as random

degree of error we refer to it as random

degree of error we refer to it as random error or systematic error random error

error or systematic error random error

error or systematic error random error is a fancy word for a chance error we

is a fancy word for a chance error we

is a fancy word for a chance error we assume that it’s benign it does not have

assume that it’s benign it does not have

assume that it’s benign it does not have an impact on our conclusions about

an impact on our conclusions about

an impact on our conclusions about reality and I agree with that we tend to

reality and I agree with that we tend to

reality and I agree with that we tend to be forgiving with regards to random

be forgiving with regards to random

be forgiving with regards to random error systematic error however refers to

error systematic error however refers to

error systematic error however refers to bias and by a

bias and by a

bias and by a is not so benign buyers can impact our

is not so benign buyers can impact our

is not so benign buyers can impact our understanding of reality it can lead us

understanding of reality it can lead us

understanding of reality it can lead us to make inferences about things they’re

to make inferences about things they’re

to make inferences about things they’re not there are not necessarily true

so to understand the issue of bias we

so to understand the issue of bias we need to realize that there are several

need to realize that there are several

need to realize that there are several sources of bias in research the one

sources of bias in research the one

sources of bias in research the one source of bias I want to talk about

source of bias I want to talk about

source of bias I want to talk about today is the confounding bias to better

today is the confounding bias to better

today is the confounding bias to better understand confounding bias we need to

understand confounding bias we need to

understand confounding bias we need to think about who we are as humans we do

think about who we are as humans we do

think about who we are as humans we do not exist in vacuum we are variable clip

not exist in vacuum we are variable clip

not exist in vacuum we are variable clip for instance our health estate if I come

for instance our health estate if I come

for instance our health estate if I come to you and I say your health state is a

to you and I say your health state is a

to you and I say your health state is a function of your dietary lifestyle

function of your dietary lifestyle

function of your dietary lifestyle I’m telling you part of the picture

I’m telling you part of the picture

I’m telling you part of the picture because your health state is not a mere

because your health state is not a mere

because your health state is not a mere function of your diet your diet might be

function of your diet your diet might be

function of your diet your diet might be a small factor that influences your

a small factor that influences your

a small factor that influences your health estate your genes can also

health estate your genes can also

health estate your genes can also influence your health estate your age

influence your health estate your age

influence your health estate your age can influence your health estate your

can influence your health estate your

can influence your health estate your exercise and activity can influence your

exercise and activity can influence your

exercise and activity can influence your health estate therefore it’s important

health estate therefore it’s important

health estate therefore it’s important when I want to understand the impact of

when I want to understand the impact of

when I want to understand the impact of diet on health estate that we also

diet on health estate that we also

diet on health estate that we also account for other factors that might

account for other factors that might

account for other factors that might influence this relationship things don’t

influence this relationship things don’t

influence this relationship things don’t exist in vacuum because if they do I can

exist in vacuum because if they do I can

exist in vacuum because if they do I can easily convince you that the shoe size

easily convince you that the shoe size

easily convince you that the shoe size of a school-age child will predict his

of a school-age child will predict his

of a school-age child will predict his or her level of education and if things

or her level of education and if things

or her level of education and if things exist in very

exist in very

exist in very you I can also easily convince you that

you I can also easily convince you that

you I can also easily convince you that those who carry a cigarette lighter are

those who carry a cigarette lighter are

those who carry a cigarette lighter are at higher risk for acquiring lung cancer

at higher risk for acquiring lung cancer

at higher risk for acquiring lung cancer how come let’s look at the example of

how come let’s look at the example of

how come let’s look at the example of shoe size and level of education when

shoe size and level of education when

shoe size and level of education when pigs exist in vacuum that I will support

pigs exist in vacuum that I will support

pigs exist in vacuum that I will support that but what’s missing is the age of a

that but what’s missing is the age of a

that but what’s missing is the age of a child as a matter of fact an older child

child as a matter of fact an older child

child as a matter of fact an older child will have a larger shoe size and an

will have a larger shoe size and an

will have a larger shoe size and an older child will also have a higher

older child will also have a higher

older child will also have a higher level of education age causes both the

level of education age causes both the

level of education age causes both the larger shoe size and the advancement in

larger shoe size and the advancement in

larger shoe size and the advancement in education and that’s the classic

education and that’s the classic

education and that’s the classic definition of confounder it’s a variable

definition of confounder it’s a variable

definition of confounder it’s a variable that confounds the cause and the effect

confounders are two types they can be

confounders are two types they can be complete and they can be partial I

complete and they can be partial I

complete and they can be partial I mentioned the example of cigarette

mentioned the example of cigarette

mentioned the example of cigarette lighter and lung cancer that doesn’t

lighter and lung cancer that doesn’t

lighter and lung cancer that doesn’t make any sense to anybody and that’s why

make any sense to anybody and that’s why

make any sense to anybody and that’s why it’s a complete confounded relationship

it’s a complete confounded relationship

it’s a complete confounded relationship completely confounded relationship why

completely confounded relationship why

completely confounded relationship why because it’s a function of smoking when

because it’s a function of smoking when

because it’s a function of smoking when we ignore smoking we in fear that having

we ignore smoking we in fear that having

we ignore smoking we in fear that having a cigarette lighter will cause someone

a cigarette lighter will cause someone

a cigarette lighter will cause someone to increase his or her risk for lung

to increase his or her risk for lung

to increase his or her risk for lung cancer now let’s put smoking in the

cancer now let’s put smoking in the

cancer now let’s put smoking in the equation we know that smokers are more

equation we know that smokers are more

equation we know that smokers are more likely than non-smokers to carry a

likely than non-smokers to carry a

likely than non-smokers to carry a cigarette lighter and we know that

cigarette lighter and we know that

cigarette lighter and we know that smokers are at higher risk to develop

smokers are at higher risk to develop

smokers are at higher risk to develop lung cancer control for smoking and this

lung cancer control for smoking and this

lung cancer control for smoking and this relationship will vanish it doesn’t

relationship will vanish it doesn’t

relationship will vanish it doesn’t exist

confounders can also be partial

confounders can also be partial for instance again smoking and lung

for instance again smoking and lung

for instance again smoking and lung cancer but now the smoking status is the

cancer but now the smoking status is the

cancer but now the smoking status is the exposing factor smoking is not the only

exposing factor smoking is not the only

exposing factor smoking is not the only thing that increases one’s risk for lung

thing that increases one’s risk for lung

thing that increases one’s risk for lung cancer I’ll take one possible confounder

cancer I’ll take one possible confounder

cancer I’ll take one possible confounder there could be several but let’s

there could be several but let’s

there could be several but let’s consider age the older you are the older

consider age the older you are the older

consider age the older you are the older people are the more likely they’ve

people are the more likely they’ve

people are the more likely they’ve become smokers and then the older you

become smokers and then the older you

become smokers and then the older you are the more likely you are to develop

are the more likely you are to develop

are the more likely you are to develop lung cancer when we adjust for age maybe

lung cancer when we adjust for age maybe

lung cancer when we adjust for age maybe the relationship between smoking and

the relationship between smoking and

the relationship between smoking and lung cancer worse t the difference is

lung cancer worse t the difference is

lung cancer worse t the difference is the margin of the relationship fins it

the margin of the relationship fins it

the margin of the relationship fins it becomes smaller so to understand the

becomes smaller so to understand the

becomes smaller so to understand the real relationships about things we have

real relationships about things we have

real relationships about things we have to keep in mind that relationships could

to keep in mind that relationships could

to keep in mind that relationships could be confounded and we shouldn’t ignore

be confounded and we shouldn’t ignore

be confounded and we shouldn’t ignore that now you might say I’m bringing

that now you might say I’m bringing

that now you might say I’m bringing these fictitious examples because I’m

these fictitious examples because I’m

these fictitious examples because I’m trying to be extreme with regards to

trying to be extreme with regards to

trying to be extreme with regards to these explanations truth is the research

these explanations truth is the research

these explanations truth is the research data publications and reports are

data publications and reports are

data publications and reports are plagued with confounded relationships in

plagued with confounded relationships in

plagued with confounded relationships in this era of commercialization in this

this era of commercialization in this

this era of commercialization in this academic atmosphere where we have to

academic atmosphere where we have to

academic atmosphere where we have to publish or perish sometimes we become or

publish or perish sometimes we become or

publish or perish sometimes we become or we succumb to the temptation and ignore

we succumb to the temptation and ignore

we succumb to the temptation and ignore confounded relationships I’m gonna share

confounded relationships I’m gonna share

confounded relationships I’m gonna share with you an example of a study that I

with you an example of a study that I

with you an example of a study that I found a little troubling so the study is

found a little troubling so the study is

found a little troubling so the study is about tree pollen and the increased risk

about tree pollen and the increased risk

about tree pollen and the increased risk of nonviolence

of nonviolence

of nonviolence suicide and

suicide and

suicide and women so the title of the study is that

women so the title of the study is that

women so the title of the study is that three Pauline Peaks are associated will

three Pauline Peaks are associated will

three Pauline Peaks are associated will increase nonviolent suicide in women the

increase nonviolent suicide in women the

increase nonviolent suicide in women the study intrigued me I actually went and

study intrigued me I actually went and

study intrigued me I actually went and read the study and I found that the

read the study and I found that the

read the study and I found that the authors of the study examined all kinds

authors of the study examined all kinds

authors of the study examined all kinds of relationships they examined men and

of relationships they examined men and

of relationships they examined men and women they examined nonviolent and

women they examined nonviolent and

women they examined nonviolent and violent suicide and they also included

violent suicide and they also included

violent suicide and they also included all types of pollens but the one single

all types of pollens but the one single

all types of pollens but the one single significant relationship in the study

significant relationship in the study

significant relationship in the study suddenly made itself to the title of the

suddenly made itself to the title of the

suddenly made itself to the title of the study I want you to think what makes

study I want you to think what makes

study I want you to think what makes three Pauline lead women to learn violet

three Pauline lead women to learn violet

three Pauline lead women to learn violet suicide sometimes we know the confounder

suicide sometimes we know the confounder

suicide sometimes we know the confounder and other times we may never know it you

and other times we may never know it you

and other times we may never know it you might just be left for guessing so

might just be left for guessing so

might just be left for guessing so confounding is a problem and we have to

confounding is a problem and we have to

confounding is a problem and we have to handle it but how do we handle it we

handle it but how do we handle it we

handle it but how do we handle it we need to be aware that research studies

need to be aware that research studies

need to be aware that research studies are two types they’re either

are two types they’re either

are two types they’re either observational in nature or experimental

observational in nature or experimental

observational in nature or experimental in nature we refer to them in clinical

in nature we refer to them in clinical

in nature we refer to them in clinical settings as clinical trials and

settings as clinical trials and

settings as clinical trials and observational studies we can never

observational studies we can never

observational studies we can never completely eliminate confounding because

completely eliminate confounding because

completely eliminate confounding because we can only control for what we know

we can only control for what we know

we can only control for what we know however we have to be real about what we

however we have to be real about what we

however we have to be real about what we know if I know that variables X Y & Z

know if I know that variables X Y & Z

know if I know that variables X Y & Z confound the relationship in my study I

confound the relationship in my study I

confound the relationship in my study I need to see evidence that the

need to see evidence that the

need to see evidence that the researchers have controlled for them

researchers have controlled for them

researchers have controlled for them they have accounted for them otherwise

they have accounted for them otherwise

they have accounted for them otherwise you are left to doubt the level of

you are left to doubt the level of

you are left to doubt the level of evidence if the study is a randomized

evidence if the study is a randomized

evidence if the study is a randomized control trial or experimental in nature

control trial or experimental in nature

control trial or experimental in nature we have to understand that by virtue

we have to understand that by virtue

we have to understand that by virtue over

over

over damas Asian meaning we give people equal

damas Asian meaning we give people equal

damas Asian meaning we give people equal chance to be assigned to one or the

chance to be assigned to one or the

chance to be assigned to one or the other group in the study we equate the

other group in the study we equate the

other group in the study we equate the groups on all confounders the ones we

groups on all confounders the ones we

groups on all confounders the ones we know and the ones we may not know by

know and the ones we may not know by

know and the ones we may not know by virtue of randomization we accomplish

virtue of randomization we accomplish

virtue of randomization we accomplish that so confounding seems to be a list

that so confounding seems to be a list

that so confounding seems to be a list of a problem and well conducted

of a problem and well conducted

of a problem and well conducted randomized control trials where do we go

randomized control trials where do we go

randomized control trials where do we go from here we have to become informed

from here we have to become informed

from here we have to become informed consumers and utilizers of research

consumers and utilizers of research

consumers and utilizers of research research is very important but only when

research is very important but only when

research is very important but only when it’s will conducted and only when the

it’s will conducted and only when the

it’s will conducted and only when the evidence in it is made of diamond we

evidence in it is made of diamond we

evidence in it is made of diamond we have to be the diamond that resists the

have to be the diamond that resists the

have to be the diamond that resists the pressure of temptation if we are

pressure of temptation if we are

pressure of temptation if we are researchers we have to resist the

researchers we have to resist the

researchers we have to resist the pressure of ignoring confounding because

pressure of ignoring confounding because

pressure of ignoring confounding because it suits our presumed ideas and if we

it suits our presumed ideas and if we

it suits our presumed ideas and if we are utilizers we have to resist the

are utilizers we have to resist the

are utilizers we have to resist the pressure of ignoring confounding because

pressure of ignoring confounding because

pressure of ignoring confounding because the results suit my liking thank you

the results suit my liking thank you

the results suit my liking thank you very much

very much

very much [Applause]