Diagnostic Reasoning I

A lecture on Diagnostic Reasoning by Dr. Rajesh Mangrulkar, M.D. This lecture was taught as a part of the University of Michigan Medical School's M1 - Patients and Populations Sequence.

View the course materials:
http://open.umich.edu/education/med/m1/patientspop-decisionmaking/2010/materials

Creative Commons Attribution-Non Commercial-Share Alike 3.0 License
http://creativecommons.org/licenses/by-nc-sa/3.0/

it's a rude awakening and mornig welcome
to do with the uh... third week of

medical school for you you might have
noticed that our

cohort of medical students has that
doubled

in size if you guys have noticed you if
you're here first of all the job finding

this other lecture hall sorry work

jumping you from place to place but this
is part of

trying to revamp

all of our lecture halls and so we had
to do this in a stage process those of

you that understands gantt charts

and construction

will uh... will sympathize but you found
the place

in north lecture hall now are your
colleagues here and suits

they are a great source of advice

and uh... and uh... being able to and
sort of queries about different staff so

now you've got your built-in

counselors uh... here but they're also
going to be very busy as you notice they

probably probably noticed they started
about an hour before you did so

for

that's what you have to look forward to
it here

that that

uh...

hope you guys survived your first quiz
hopefully it wasn't too painful i know

many people ask will they be medical
decision making

stuff on the quiz and the answer to that
was

uh...

remember that genetics in pathology or
the big components of the quizzes and

that we in india and we have assignments
so you turned in your first assignment

last thursday at small groups

uh... i'm in the process of reviewing
them

they will get back to you in your mail
boxes

uh... so you will be have them to study
from

i will also when i return them to you
you should pay attention on c_ tools all

publisher ultimate which will also have
the answer key

to interview uh... assignments

so there's no hidden doctors here it'll
be a turkey will give you an explanation

to each of the questions that way you
can look at your answer what banter in

turkey was any notes that you had

and reconcile and if there are any

other questions certainly we're here to
help you

you have another small group tomorrow
afternoon

that will largely be based on material
that we cover today

uh... hopefully if you guys have had a
chance to get a jump start on that i try

to post in a week in advance

you're more then welcome to

uh... reading ahead is hopefully fine
encouraged if you want to do want to

that's totally fine as well

there should be enough time to be able
to complete that again it's the same

protocol

bring it to your small group still
discuss that with here

co students in your small group silkair
and then returned them

uh... tournament and then i'll be
returning to you in a timely fashion

any questions about the logistics of
this course for this part of this force

so why don't we go ahead and get started

uh...

today we're going to be for we're gonna
be picking up a little bit where we left

off

last time

and he wants start by declaring uh...
any industry relationships

that i might add to contacts that meet
your otherwise

in the answer is

certain factor

uh... dancers is i have not been

to disclose

so nothing that would uh... interfere
with my ability to present you an

objective view of uh... medical decision
made

so our first just returning to where we
left off with their first thread

this is around information retrieval
focusing on asking in acquiring

if you remember and

after last monday we left off by talking
about the way the structure well

foreground question and these are sort
of that

uh... fundamental tools that were
required to

af but good questions get the
information from the literature and then

interpret and apply them as a way to
focus a little bit just to revisit

this tool that we talked about

this was the pico tool that encourage
you to just use it as a tool basically

it's a way of specifying the different
elements in your question there will be

important

tour it ought to be able to sort out the
appropriate answers

and we think it's important to be able
to do with in advance

because by doing it in advance

uh... you're really able to focus on
what you need and what your patients

need

so in order to practice doing this went
to ask you to do is to look at this case

and practice with your partner

jotting down the foreground

uh... question so let me go over the
case real briefly

because we're gonna be using this uh...
moving forward

so this is now this is a forty
two-year-old woman

who comes to her primary-care
practitioners office

for follow-up of her diabetes and you're
the medical student

she's currently on libby ride

ten milligrams twice daily

however her blood sugars still stay
elevated despite being on that

medication delivery right

after you see this patient

you're attending asked whether you think

she should admit foreman to her regimen

pretend like you know what would your
ideas and pretend like you know what

metformin is you may not but these are
both

uh... it is that we use in diabetes

we'll be right is the sole final you
react

and metformin is uh... another agent
that we use

uh... packets dvd

uh... and i cant really pronounce what
that actually stands for it's a very

difficult word to pronounce but

suffice it to say good beer i can see
his your um... insulin

uh... usa secretion but metformin
improves your sensitivity uh... to

insulin is well or that your body makes

so let's say you know all of that

and the question i would ask you to do
is to start putting on

your hat as if you were a clinician just
to get a sense of what

these kind of foreground questions are
about remember therefore important

components

for a foreground question you have to
define the patient population that

you're interested in

you have to define the intervention

that you're interested in

the comparison group that would be part
of the study that might help you answer

the question andy outcomes of interest

that you would be that you in your
patient would be interested in

this will help you then scanned the
literature to figure out what are the

articles that would be most relevant

for this particular case

so i'd like you did it was turned your
partner and come up with one

that would meet all four categories it's
a fairly straightforward exercise but

wanna make sure that you get

such answer practice that's so spent a
couple minutes doing that and we'll talk

about it

a

okay three parking lot with remember
that really

nobody answers here because

every single uh... every single answer
that you would give each of the four

different categories would have some
debate around

whether the article that you retrieved
with those particular characteristics of

the study population

would actually help you answer your
question

or not and the degree to which it would
do that so let's just hear some of the

questions what i'd like you to do is not
the same like he was ex and i was

uh... was why

i'd like to see if you could stated in
the form of a question one that would

contain all four

of those different components

just a little hint

i might ask something like this uh... on
one of the either the assignments

or certainly

while our final examination as well so
who wants to take a stab at it

them

uh...

okay great so here let me let me restate
this

in women with diabetes

what is the effect of metformin plus
glider ride

blood sugar levels compared to drive me
right along

so let's see if we got all the before
components in their is that patient was

women with diabetes your intervention
was metformin it would be right together

versus delivery right alone in the
outcome you're interested in

was improvement in her blood sugar
levels okay

so let's break that down the patient
population is that were women with

diabetes was that it did anyone have
something different

then women with diabetes here

awesome al you specified in age cutoff
so women with diabetes over forty ok any

other

variabilities there

asel na specifying gender so just all
patients with diabetes let's back up and

think about that

limelight it so that the question would
be if we're to get an article

that had

all ahead men and women

in the past study population

would we say that's ok to extrapolate
the results

and then apply to this patient
population

and so you said

potentially that would be ok any reason
to think that verb rationale

what goes through your mind as you're
making that call

but the answer being that date one of
the rationale might be fifty-year

defined studies where you have

both genders represented as opposed to

one gender alone so that's one certainly
one consideration what might be another

consideration you would use

uh... so now you're talking you're
getting even more specific saying

impatience

who are

poorly controlled on good be ride with
diabetes

what about the comparison group purses
uh... verses the intervention so that we

another

shape that you would be there

but that you might use backing up to the
gender issue

the question i would ask you to ask and
we ask this question all the time does

the disease manifest differently or act
differently

in one gender verses another

does the disease manifest if really
inpatient over forty

rice's those that might be younger than
forty

those would be the questions that you
would be asking yourself and remember

you have to build some background
knowledge about this subject before

being able to ask the most sophisticated

foreground question

i'm trying to know a little bit about
the pathophysiology

but it's reasonable to have those
different want uh... those those

different shades of variation

in the patient

the intervention was metformin plus
we'll be right did anyone have a

different intervention

oddly be what we're thinking about the
kind that we're not talking about

stopping the glitter ride and adding
metformin

but in some cases you might be looking
for head to head comparison

and in fact if you're thinking about it
from a drug company standpoint they may

be interested in head to head
comparisons because they're trying to

prove one

at the at the efficacy of one drug over
another

if you're thinking about it from the
standpoint of

of uh... pathophysiology you might be
interested in head to head comparisons

but if you're thinking about it in the
practical sense we often might be

looking at

adding and agent to an existing regimen

those are less commonly found

in the literature on just a tad

and so you may be stuck saying well i
don't have an additional

uh... i don't have a study that shows
the addition of metformin

and so you would have to extrapolate

from head to head comparison

which is tricky did you

certainly beyond the scope of this
course but keep in mind those of the

questions that we entertain as positions
as for reading the literature

now one of the outcomes that was was
mentioned was improvement in blood sugar

so first of all how would we measure
that

but voz votes

okay so you could be looking at that's
the blood sugar which have been one

marker of diabetes hemoglobin a one c
which alert next year

is a eleven for a test that we get that
it looks at patients long-term control

our blood sugar at least medium control
over the past three months

so that be a good outcome

to specify so you could get that level
of specificity and you might find

studies that look at just after larger
verses does that look at a one c and you

would have to decide which ones

for me in my patient now would be
important consideration

any other types of outcomes you guys
looked at

or anyone specified besides blood sugar
improvement

basil side effects might be another
outcome that you would look at what sort

of side effects might be worth we'd be
worried about a few

if you know any

so one that i might be worried about is
would i be dropping this patient's blood

sugar too low

to the point that they have hypoglycemic
events

so that would be something to be
thinking about

but absolutely remember this is a
therapeutic question that your asking

and as if they're peter question there
are unintended or sometimes

uh... known side effects of the
different genes that we

that we use and so looking at the
adverse events

would be another important outcome

so keep in mind this is just an exercise
but to wall

to get us to where we dot where we need
to go so that when we look at the

literature

we know what we're looking at

that is the point of the foreground
question

now this is up there

foreground question before i questions
about their feet

later on in this lecture really talking
about or run questions

as it pertains to diagnostic tests

again to get distinct types of questions

but both equally important here a couple
of examples that we that were uh... that

would be reasonable i think you guys
came in kimba up with these in tight to

die but if that was one thing we didn't
talk about the fact that this woman

is a tight to diabetic

she may be insulin requiring more
insulin

resist uh... you know that we talk about
that one person so i can fight one being

autoimmune

more likely to happen upon early onset

tight too

uh... not necessarily requiring insulin

uh... and so you might want to
distinguish that studies that you're

looking at

but then there's a question is metformin
good write better than the bread alone

in lower bucks ordered that was
something that you guys came up with and

here's the side-effect question among
women with type two diabetes

are there more instances of low blood
sugar of insufficient on both metformin

it would be right

when compared with we'll be right along

so that you guys came up with this
fairly straightforward exercised but

keep in mind level of specificity is
important because you're gonna be faced

with a ton of different studies

you're gonna have to work through to
figure out which are the ones that are

most relevant tumi and the patient that
you're working with what are the

important outcomes to that may be very
different

and the outcomes that you are interested
in and so it's important vehicle to

specify that of prop yes

went

at

so it literature doesn't exist uh...
what do you do that is probably

uh... the hardest question we face as
clinicians

there is a when i'm gonna ask you to i'm
not gonna be able to answer that

completely

but as we build our course over the next
three years what you're going to get a

sense of

is that there is a hierarchy of evidence

meaning that as

certain studies take on certain study
characteristics both in terms of how

they're designed and how well their
implemented the level of evidence that

they provide become stronger or weaker

and so it's not necessarily do i have

is there no studies out there

what are their studies out there that
are less well done and so what do i do

with those

compared to studies that are really well
done

keeping in mind that the studies that
are really well done are actually less

frequently encounter

and those studies that are not so well
done or studies that are

observation as opposed to control
clinical trials

so the answer your question is i don't
have the answer now but that is exactly

going to be the point of the next three
years

what do you do when you're faced with a
quandary we're gonna try to party

conceptual model that will on full

that you'll get more comfortable

there many sources the foreground
questions these are the pics precise

questions if you go to medline four

or that you will go to a practice
guidelines or you'll use evidence-based

databases and all of these sorts of

of uh... of resources will be introduced
to you

as we go through

so building on those questions then

we do need to then figure out what is
the data that we're getting and how do

we interpret

and so as i mentioned therapy types of
questions are

are very important part

of what we do is the nation's they seem
to be

the most natural thing you would assume
positions do which is prescribed

treatment

and figure out whether it's

making a difference or not

however it is not the most fundamental
thing that we do in fact

therapeutic questions are very
sophisticated questions

that we actually don't cover much about
their pete this year

we do it in your second year as you
start tackling pathophysiology

this year

the question we want you to start
thinking about

what is going on

is that you all a g of why what i'm so
the the clinical manifestations

of the disease

and so that really focuses on diagnostic
reasoning and diagnostic test and so

we're gonna spend the rest of our time
today focusing on diagnostic tests

so here are the learning objectives for
the rest of today by the end by the end

of this lecture and honestly by the end
of your small groups tomorrow because

you're gonna have to work through these
concepts in small groups it's gonna be

mainly a superficial

covering today of the concept

and the diving into it with your simon
in small groups are really where the

learnings going to happen

but by the end of this series of
sessions

you should be able to do some

fairly basic diagnostic

diagnostic question formulation

you should be able to define and uh...
calculate

sensitivity specificities and the
predicted values for different

diagnostic test for an introduce that
concept you today and tomorrow

and you should be able to explain how
risk factor

that god can thrive prior probabilities

and how this concept relates to
prevalence

and then finally you should be able to
modify probabilities from test results

through on number of different
mechanisms

we introduce you to the concept obeys
probabilistic reasoning as a way of

modifying probabilities over time

today we're gonna use

to buy two tables which is a of more
straightforward

easy to conceptualize and visualize way
of modifying probabilities

but you'll get more and more practice
with using that keep in mind that it is

just as good as base so these are
presented to you as you should be one

over the other represented the us
items to put in your toolbox in portland

out

at different points when the need arises

there also other tools that are
available that will introduce to you one

called likelihood ratios

and you'll get a chance to practice
these

in your assignments and in your smokers

so just like we had the odyssey as a
case for introducing you to use

probabilistic reasoning this time really
immerse yourself in a clinical case and

i want you to think about this again
just with uh...

limited amount of knowledge you have
about this condition of this disease

whatever you have this fall

but bring to bear your experience as we
work through this case during the

remaining a bit the remainder of this
like

so the cases is sixty year old man

who does not have heart disease

who presents with sudden onset of
shortness of breath

dismissed is a term that we use

so here is a politically description of
the problem that you can see

yesterday after he flew in from
california the day before

he'll blokes suddenly in the middle of
the night at three in the morning with

sudden shortness of breath

so we woke up gasping for air at three
in the morning

and he tells you

that you ask them one question which is

well was it bad when you were lying down
and that's why you setup or was it worth

when you or is it about the same
regardless of whether you're setting up

ur lying down

because the people maybe was worse than
usual you said you know what

actually is it

any different online down if i'm sitting
up

i'm still

feeling short of breath and it won't be
up in the middle of the night

all right so that is your initial chief
complaint

this is something that you'll get used
to uh... as you do your data uh...

queries from uh... from patients at your
data gathering

from from different

but you being a good clinician

you start asking some follow-up
questions just like a good clinician

uh... or good mccain sort of building on
the mechanic model that we introduce you

to a week ago

so you ask him what other symptoms were
you feeling at the time

now as you get more sophisticated you
will be asking more specific questions

right now

sort of the about what i would expect

and what seems to be able to do what
else we feeling at the time

well he says does not testing

he doesn't have a leg pain he doesn't
notice any swelling of his leg

he says idea just come back from a long
plane ride he flew in from california

solicit was about five hours

nonstop

and he's had no problems like this
before that he knows

this shortness of breath

he takes one aspirin every day and he
does a smoke a pack of cigarettes

everyday

so that's kind of the next phase of
diagnostic

intake

so the question that faces all of us
just like the mechanic faces

is to be able to build a differential
diagnosis you remember me mentioning

well we as clinicians

still differential diagnoses all the
time

and basically differential diagnosis is
a list

of possibilities with associated likely
that's with associated probabilities

so as we mentioned last time if you can
p of a particular condition you'd be

saying that the likelihood that

this particular condition is the reason
for the shortness of breath

is x percent that's how you would write
that down

and what you would do is you place it in
descending order of likelihood and you

would be talking about wine

if you could get this

conceptual approach

jupe all your differential diagnoses

you will do well in medicine because
this is how we talk

now we don't do it necessarily in such
mathematical discrete models

is exactly what we talk when we talk
with another position we like you know

what

i think this patient has colon cancer
it's definitely more likely that he's

got colon cancer than he has hemorrhoids

that's kind of how we speak in barely in
formal settings

when we're discussing

uh... the case uh... a particular
patients case in trying to understand it

yala g

what we're really doing though is that
words generating a differential

diagnosis and it is the regardless of
what field you're going to do

you'll be doing this over and over again

so you get to your first one today

so now what i'd like you to do

is think about

what possibilities may be going on with
this particular time

talk it over with your partner list two
or three things that might be going on

and i will talk about what i think might
be going on

e

okay productive well on the five because
it's probably arctic every different

about yourself

at this stage that you're at

but it's not wrong to try

so let's just hear something you don't
have to give me a probability or

anything like that

just give me some possibilities of what
might be going on

all materialism versus congestive heart
failure okay so we've got pulmonary

embolism

congestive heart failure what else do we
go

sleep apnea okay great other things yeah

guilty gear emphysema while you guys and
he you've gone through medical school

before these are great coverage is a
great differential diagnosis absolutely

ep

events

hotbed i'd i didn't ask him about scuba
diving and whether he had uh...

at whether he had done that but
certainly that would be a follow-up

question that we would act

these are these are great things now
let's peace apart with these things me

first of all what is a pulmonary
embolism

which had pulmonary embolism you know
what it ps

unexplained

so o'clock in the long

the person was sitting in mobile for a
period of time which allows rumbled sis

to develop especially in the lower
extremities which can then migrate up

and gets stuck in the long

you stated that his shortness of breath
with sudden and so you started unpack

the whole concept of rationale and my
thinking that

versus some of the other things because
often times when you get a pe it

suddenly breaks off part of your of the
clock suddenly breaks off and goes into

the pulmonary basketball

now you'll understate you'll start
developing the language around that but

that's a great way to explain that you
also mentioned congestive heart failure

now what is congestive heart

art which on the spot but i was a good
differential i'm coming back into our

world seven you know

all right so congestive heart failure
being basically eight at in enhanced

pulmonary vascular

in the palm area vascular system because

the heart is not able to

nearly as good injection of the blood
through the periphery and so things back

up for a bride of the reasons either
though

orgasm pump is well it's two-step you
got valvular leakage

and it backs up into the pulmonary
vasculature you increase the pressure it

you cause pulmonary edema which gives
you offensive shortness of breath but as

you mentioned oftentimes that'll be
accompanied by peripheral oedema because

of the heart backs up from left
ventricular failure right ventricular

failure

than you actually have

summit denying your lower extremities

she doesn't according to him we haven't
done the examination you so that's the

caveat here but that may be a reasonable
thing to come up with uh... to uh... to

come up with in your differential and
you put it lower because there were some

aspects that weren't necessarily as

consistent with that bag no's' right

so that's a that's a great way to
approach the differential

i would mention obstructive sleep apnea

overhear yet

what's obstructive sleep apnea might
like that that here

e

pacbell now that site so what let me
just uh... unpack what you're saying

there

you're commenting on the fact that this
guy woke up in the middle of the night

with shortness of breath now oftentimes
people who

have obstructive sleep apnea don't
actually wake up

but they have apnea episodes which means
that they

stopped breathing

for a period of time and sometimes i can
cause them to suddenly startle and wake

up they don't necessarily wake up short
of breath but they can't

absolutely until anyone that wakes up in
the middle of the night

i'm certainly thinking about sleep apnea

it's probably one of the most under
diagnosed conditions in this country

i didn't know the reggie white died of
it but he certainly would be pat rescue

uvic because people who have are

uh... and i don't know how it will be c
was but large

body habit is certainly is a risk factor
for obstructive sleep apnea

also can lead to right sided congestive
heart failure so might also be connected

to one of the diagnoses that we heard
early on

so keep in mind that sometimes diagnoses
are completely independent of themselves

revelry talked about independent
independent events

sometimes different items in your
differential are actually related to

other items that you differential so we
have to keep that in mind as well

by legal said feel pd in the back and
now possible emphysema what is that

and why might he be at risk for them

front his

baton so

clearly he's got to look just to
reiterate what you're saying

he's a smoker

long-term smoking can cause destruction
and inflammation to the pulmonary

bronchus tree

andy alveolar

uh... components of it what you learn
about this year and next year

uh... destruction of the albee ally is
typically what we see as a mechanism

towards emphysema

which is one manifestation clinical
manifestation of seo pd and certainly

can impair oxygen

exchange which would make you short of
breath and destiny

and maybe he's in the early stages maybe
this is just his first manifestation of

that

you know he said he has ended for breath
before

but your focusing on his rest

dot keep that in mind

some people were focusing on how he
presented waking up in the middle of the

night

some people are focusing on risk

both are absolutely critical as you get
as you generate your differential

diagnosis

risk

drives the order often of the things in
your differential diagnosis

but how u manifest

also changes the order and what you're
thinking about

in a different likeness

both of the things that we need to walk
in with

and as we think about observational
studies tying this to what doctor grover

was talking about last week

observation als studies give us
information about race

they give us information about risk
factor

that contribute to different clinical
case uh... disorders

so very important to trying title of
this to get and question over here some

yes

back gop_ d is chronic obstructive
pulmonary disease

it is the long-term manifestations of
tobacco use

it can manifest through either emphysema
where you have destruction of the

alveolar tissue

and impaired oxygen exchange as a result

you can also manifest as what we call
chronic bronchitis where you have a

tremendous amount of inflammation

in the broncos and with mucus production
which can also cause impaired oxygen

i mean

uh... oxygenation

so sorry i didn't identified annual

learn all of these concepts as you move
on

so you're doing a great job building
your first differential diagnosis now

the problem is is i also said

we have to start assigning probabilities
to get to these different

uh... clinical manifestations of just
give you an idea of what my list was

before we get started

that's right

congestive heart failure

and that the amount exacerbate shin

and i also thought about asthma which
could be another

different manifestation that is not
necessarily smoking related

but also uh... can contribute to airway
inflammation

i didn't have obstructive sleep apnea in
my differential that doesn't mean that

it's wrong to put it there

um... many people would and what you'll
find is a different clinicians will

bring their different behind the scenes
to their differential diagnosis

of the question is what do you do when
they're all these behind seas out there

and how do we assign probabilities to
the different uh... clinical

uh... entities in your differential

or sometimes we can do it by a gut
feeling

based on what we know about the disease
based on what we know about the patient

based on what we know about the for
respects

and so here's my gut feeling
differential diagnosis

i put pe at the top

i could see a jeff next

at thirty uh... and i put emphysema
thirteen asthma fourth

and i tried to make it all add up to a
hundred percent so it was nice to meet

so i could forty thirty twenty ten

there is at a right answer you don't
know

that's just my gut feeling but what it
does tell you is i didn't think that

pete was so overwhelmingly likely and i
would put that at seventy percent and

everything else down at the bottom

there may be clinical conditions where
you do that but keep in mind that each

one of these diagnosis

had both its

can uh... consistent features

and something that we're just kind of
atypical

why wouldn't it be a pe well he said
he didn't have a new leads well he'll

appear on a plane ride for five hours
most people don't get a ddt_

well maybe ever

symbol that looks like

y c h f not bad

because there's you know there's some
things that are consistent with a but he

doesn't have flirts from any of you

until their fingers that will sort of
make you head

and these numbers are part of making

are are my manifestation of making

uh... contribute communicating the head

there a couple of other things you can
do with this

so for example remember

you can combine probabilities of
different events

so what is the probability that
shortness of breath is due to be there p

foresee a chap

given these a particular numbers when
you guys think

seventy percent

if the two are mutually independent
events

meaning that they're not dependent on
each other not meaning that if you happy

you are not more likely to get seja

or vice versa

because if there is that overlap remote
from what we understand about the

disease and you can't combine them by
adding so yes the answer to that would

be seventy percent but remember a thing
in mind that provided that both don't

happen simultaneously

just as an intellectual exercise if you
thought that there might be a ten

percent overlap

in the likelihood that you have both p n
c h f bowing out of the same time

meaning that they are dependent and then
sport

that they're both likely

uh... the it's possible that you're
having both of them happen at the same

time then that you could combine the two
by saying what is the likelihood that

you have either p or c h f

but it wouldn't be seventy percent

would actually be sixty percent because
there's also ten percent chance

but both are happy happening
simultaneously

but the good intellectual exercise to go
through again the numbers art as

important as the concept that it goes
down

when but to events for depend

keep in mind that in medicine

their are absolutely independent events

and their absolutely

that will be dependent and has you go
through your blocks and understand the

diseases you'll get an appreciation

berwyn things are determined and when
they read

so basically what we're doing is worth
actually creating

prior probability

before we have done any further down the
track gathering with this patient

re-affirm a couple of questions that we
would ask

well before we do any further exit we
can do the physical exam yet

we didn't do any real specific
questioning within the week testing yet

we're generating

it prior probability remember we did
that with the woman with the brc_ a

haitian that we were concerned about

we had a prior probability with her
walking in

do when she was twenty years old

well this is a prior probability that's
based on a number of different factors

that i doubt that i mentioned

on the other hand you can actually
generated prior probability

using some tools that are out there is
what don't want to demonstrate obscene

as some of the tools that might be
helpful to you

so why is there are these things called
clinical prediction rules

mechanical production rules are ways of
using the literature and what we know

about the literature estimate s

about particular disease

and

the uh... the way that we did there are
a number of them that are out there

pulmonary embolism is one of those
clinical diseases that actually has a

number of different political
predictions and i want to sort of show

you

what uh... what one uh... looks like

so what you have here to really think
that your slides that you can use as

well

this is mad calc three counts

this is on the uh... you guys have
access to this and uh... they're unknown

bura different uh...

their number of different medical
calculators here

for p what you would do is you would
enter in the day the for the particular

patient

and then it would give you step here in
the lower right hand corner cc

so let's do that for this particular
case

based on what we know right now

based on what we were going out the
sixties right so we put the sixties

we also know that he's made

so we put that

and then you'll see a whole series of
risk factors here

let's say we don't deal with uh... let's
go down and see what he has

we know he's got the acute onset which
means some that's another

word that you learn

sudden onset of this issue

we click on that

notice what's happening to the risk as
we go through this

just being sixty years old and mail

and then adding acute onset of dyspnea

this risk number goes up to fifty
percent

really jumps up there

and if we added that he might have been
immobilized with saying this

let's say he sat

in a plane

completely comatose for five hours
didn't move out what

that would probably qualify as
immobilization

let's say we clicked that

you'll see that his risk of having a
pe with all of this

is now sixty percent

so my gut feeling of forty percent

probably and underestimation now this is
a gut feeling this is based on

spa operational stocks that are out
there and putting them into a

mathematical model

you can get these things for your
handheld device

you can put them on the web a lot of
these are being integrated into the

electronic health record for positions

so that these

guides can be placed right at the point
of care

i'd encourage you to think about these
and try to explore some of these because

they were a number of these for a number
of different conditions out there

but we're gonna return to this as we go
through so right now

we're starting around sixty percent

as our prior probability

but you realize also that there's a
number of questions here that we just

don't know the answer to the right legal
fees at a fever we don't know we know

that he doesn't have a history very
vascular disease

we don't know if he passed out we don't
know if he's actually got one-sided

lights while

so there's more data we need to get

so let's gather

so more date

so here's some more information then i'm
gonna throw at you based on this

particular case

you talk about family history find out
that he actually

has had

he has a family that is that the ddt_
in the past

pretty for it

you do a physical exam on him and you
find a his blood oxygen saturation is

normal on room air

so these oxygen eighty five

checkers respiratory rate at sixteen
that's generally that's a little fast

but it's probably okay but

his pulse rate is a hundred and buy it
now

united now that a hundred five ezell
elevate we would call that eka kartik

so he's technopark

and you examine his loans as you will
learn to do this year and you'll find

when you became his patients' lungs that
he has crackles

ntsb's

user crackles are sort of

they sound like crackers

we've ever listen to rice crispy

that's what it sounds like a bit sounds
with inspiration you would get crap that

usually indicates that there's some
degree of a team up in the long some

degree of swelling in the long and as
the albee a liar trying to expand

they pop open up a very easy way but
enough

tough way because there's a lot of
surface tension at each of you

level because of the fluid in the
interstitial space

he's also got leases leases are
indications of airway obstruction small

airways obstruction in the long

but you'll also notice he doesn't have a
problem which means that he doesn't have

this sort of uh... inflammation in the
pleural space which will learn about

and he doesn't have evidence of
consolidation consolidation would give

you some clues that he might have an emo

we don't have either of those

but you also examine is like and you
find that even though he didn't think it

was a swollen it absolutely is well

and you feel of any

below his knee

most of the time we should be able to
feel the danes attorney but if they're

inflamed possibly because of a clock

you might feel below the knee

you get a chest x-ray that's normal

and you get the cagey in its shows that
his heart's going fast but nothing's

besides that

so now what we do is we go back

to the clinical prediction

calculator and cd about entering in this
data

so the additional data that will get

he not only is uh...

sixty years old and mailing it to keep
this thing out and let's say he's

immobilized he was a mobile as

we now find that he's gotten lateral
excellent

but that he also has leases

contracts

no notice what happens to pay attention
to the number as i enter the zip

i started out with sixty percent by
aditi lateral leg swelling what we're

getting really close for x seventy five
percent

but this person's that p

upholding a bit

remember we're talking about as long as
even though

the race

his lead

because if you've got to cut your leg
you're more likely to throw that but we

have a pulmonary embolism those two are
not

independent features

those who are dependent effects

so it dries it up to seven six percent
but knows what happened when i clicked

on visas and practice

that number within seventy five percent
of the thirty five percent

so what in your mind

must you be thinking

the presents of crackles and we use is
actually make

pulmonary embolism last like

in fact pulmonary embolism the number
one clinical finding in the long for p

is nothing

subtotal e normal pulmonary exam

just because you have a normal palm
trees and doesn't mean you don't have a

disease

as an important consideration but keep
in mind that this is a way of

quantifying

how much

dropped there is so we ever did all this
data and we're now at around thirty five

percent

so turns out i was about

or a percent was pretty close to thirty
five pst

there are interactional terms that are
built into the mathematical model here

so if you look above the rest

and sold those interactions are built
into how he calculates the factor some

which then translates to what the risks

so yes there are interaction terms
between these things

and it's calc it's all done in the
background

clinician you know i'm not gonna do that

and i'm certainly not to be able to look
at a study be able to know how to do

that that's why these calculators to be
very helpful

it'll give you an idea are we weigh off

or are we about right is this guy is

was my initial gut feeling over on
target yeah

was kind of on target even when we did
more data gathering

but you also get a sense

uh... why how different features
increase and decrease the likelihood

which is a good learning tool

rate

so the question is of why is

smoking and wake nodding here

the answer is is that smoking and wait
for actually not considered risk factors

for p

when they've looked at these studies
what you might think of boy there smoker

they're more likely to have a clock

into israel

just because you're a smoker doesn't
mean that you're more likely to have a

pe

now certain smokers do carry increase
from body chris

if you're a young woman who smoking on
or contraceptives

you already higher risk of developing a
lower extremity ddt_

at is clear

but that's not necessarily what we're
talking about here

over the lot big picture it doesn't
contribute risk

it also could be because the studies
were done but in this case is because

they're not respect

uh...

so here's the question if you are
thinking that there is a nother

possibility on your differential
diagnosis that is either equally or more

likely does it drive this down the
absentee answers absolutely it does

and there are limits to these clinical
prediction

so this is done in a vacuum their other
clinical prediction rules

where you can actually have a button
that says is an alternative diagnosis

equally or more likely when you do that
people drop your s

because it knows that there may be other
things going up this prediction will

didn't do that

station had a family history of a ddt_
but not a personal history dvd if you're

interested in what that would have done
you could do it if he had a dvd in the

past

it drives up from thirty five to fifty

yet family history of some of his family
with d

all these are great questions but the
most important thing to keep in mind is

you don't just have to go on gut feel

at these clinical prediction rules for
common diseases which are the ones that

can help you understand

are out there

the clinical manifestations of a
particular disease

so what does this have to do with what
we do

well keep in mind we're now

we are now at

um having completed our data gathering
things were based

with the prior probability of about
thirty five percent

at the station is happy

and so what we need to do you think
about

do we need to get a test

because the question that should be
going through your mind is the thirty

five percent high enough at allstate
yahoo dot abt and got a pulmonary

embolism

we're gonna treat you as such

hopefully thirty five percent is too low
for even you

to say boy i don't think that i would
point to treat based on that certainly

would be low for most of the nation's we
need to probably get some sort of text

so what this is a how those bridges over
into diagnostic test

so here is the question what we do with
this number

well if there was a test that existed
that could rule in pulmonary embolism as

the diagnosis

with a hundred percent sense certainty

we would be saying that probability of
this of a patient having a p

due to the

at the test is positive

is a hundred percent

and the question i would ask you is what
is the stressful

what we call this test the boot stain

the gold standard generally exist for
more most conditions where they're is

where works for the past week after the
finding of the best will stand at the

top

uh... it may not be the best test

that we can envision

but it's the best test

that is available

and the question that i would ask is if
you use the gold standard tests and you

found it

the test was positive

is the probability of an alternative
diagnosis remember that a test test

that's a gold standard is only the gold
standard or particular diagnosis

let's say there's another diagnosis

that you're interested in

is it zero

if the test is positive

the answer is no

because

sometimes you have two things going on

sometimes there are vents

that are related to each other dependent
on each other

for sometimes you just have bad luck

as a patient

and you have both pulmonary embolism and
cha_

but the most important to keep thing to
keep in mind

is that we are always looking to see
what the gold standard is

but we can't always used to go see

so what is that not stick testing people
but not the testing

the the goal of diagnostic testing is to
help us modify probabilities

we talked about how

we modify probabilities based on history
taking physical exam other simple tests

complex s like the ones we're gonna talk
about

will help modify probabilities as well

but they cost a lot of money

and so we need to approach argues a
diagnostic tests

judicious

so in order to understand what we're
looking for work in a study about

diagnostic test we need to understand

how people related items to test

really does relate instead of

the patient

uh... population were interested in

what we're really trying to use the
following

a disease state

what is the disease that we're trying to
diagnose

in this particular case if ur interested
in another test to use

it would be the disease state would be a
pulmonary embolism

intervention

would actually be the test itself so
what tester we interested in his room

looking through the literature

the comparison group is not another
therapy but it was in the therapeutic

questions

it's the gold standard

so you're trying to compare

a test of interest

against what is the best test that's out
there

the problem with the best test that's
out there is that it's frequently

infeasible

sometimes dangerous even though it's the
best s so we're looking for an

alternative test that we can use i can
help us with our page so that those

favoritism

and then the outcome of interest that
were interested in is the performance of

the text

that's the fundamental

and you will be able to

do this

over and over again as you're looking at
different studies and we've this is part

of the assignment

for two more

so let's practice

as we've seen the sixty location without
heart disease is presenting with some

nonsense shortness of breath

we're considering a p

that would be our disease or our p

protest and ask us to consider is a test
collabera profusion scheme

cowbell asian perfusion scan

is something that's been around for a
long time

it basically

in the test a patient in hale's

or radio nuclei partner

and we see where that radio nuclei
particle goals

and in addition the patient gets
injected with that same radio nuclear

particle

and we see based on the blood flow

that particle goes

we take pictures

simplistic explanation but i thought you
really need to know at this point

and what we're looking for is

where does the air go

that blood doesn't go

but that there's a place where the air
goes that the blood doesn't go that

would likely be where o'clock would be

api

but that's what happens is that the
clock obstructs blood flow to that

particular place

now the gold standard

four diagnosis of a pulmonary embolism
something we call pulmonary and

geography

which is where you were actually
injecting died and watching where the

blood blow goes

you see a picture of a b q scan on the
top there

on the bottom is a very grainy picture
of a pulmonary angiogram

just keep in mind that an angiogram

more costly

slightly more dangerous

the gold standard

but we wouldn't because it's costly and
more dangerous we wouldn't offered to

all patients

because if we did that

the risk that we would in anyone who we
suspect has a p

the amount of complications and cost
that we would entail

would be mass

that's what we're looking to see is is
of the few scan which is actually

fairly

really fairly

at not terribly dangerous

whether that would be good enough
compared to the gold standard for many

of the geography

what were interested in is diagnostic
performance

so that is the pico recognize the test

but before thinking about it became
scandal then elation perfusion scamper

first question we have to ask

kena beat you scan actually even be used

now you will be asking this question as
a clinician

but before you can even go to market

as a possible test

there's some fundamental questions about
a diagnostic test

that require some definitions that you
should be aware

and they thought this on to concerts
accuracy and precision in order for a

test to be used

in the literature and study for possible
use it needs to be accurate

and it needs to be precise or whining
about it

what accuracy means at the results of
the test

corresponds consistently with from
result

not going to result in terms of
diagnosing the disease

but the correct about

meaning that it'd be q skin if i object

the radio nuclei and i say it goes to
the long

if you go to the law

and if i inhale the radio nuclei and i
say he should be inhaled into the

alveolar space

it should actually be inhaled into the
alveolar sticks

fundamental

but keep in mind that there is a clone
of work that happens

prior to attest going to study

that requires this to happen

and there's a lot of science behind us

so needs to be accurate enemies

besides we're learning about the size
twelve decision means that if you do

with over and over again on the same
patient you'll get the same result

with a reliable test

the repeated values on the same sample
resume results in the same down

so in the same patient you do it once
you do it again five minutes later you

you five minutes later

missing uh... results actually happened

that's the preciseness

you need to have both

they're actually three different

possibilities that tend to happen

when early

phase studies are done and diagnostic
tests

the first is the one that you want

want to be highly accurate and highly
precise pointed out that you're good to

go

ready to go to step two and study its
characteristics

however

you can have something very precise

but be completely inaccurate

that's what's represented conceptually
by the bulls on where you have a lot of

different numbers there clustering all
around the same area

but it's actually not doing what you
think it's doing

and then there is sometimes when you
have

reasonable accuracy it's all clustered
around the bulls eye

but low precision

their hat you get different

answers each time you do it

so what do you do in these different
situations

well the first into a single one step to
your butt

the second one you gotta think about
calibration

and you need to reset or maybe you need
to relook at the reagents veteran vault

at a particular test itself

in the last one unfortunately you gotta
start over

if you've got a bunch of if you're a
test is not precise how usable could

actually be

you have to actually get the precision
death

again this may seem fundamental in
foundational

but there are so many things that don't
make it to market for even testing

because they don't meet these criteria

important for you to keep him

once you make it past that first phase
then you can decide

diagnostic performance it what the
diagnostic performances the o of the

pico for diagnostic tests

and so fundamentally

i would ask you to think about two
things

any good diagnostic study does this

they take a well-defined group of people

who are at risk for a particular
condition

a whole population of them and they
expose them to

the experimental tests

and the gold st

everyone in the study needs to have both

and if you compare the test results

experimental tests

and the gold standard

keep in mind you don't want that patient
population to be

everyone having the disease a hunter
percent of them having the disease

you need to have a fair number

of people who don't have the disease in
order to test

the test characteristics

well that's the fundamental premise of
any good diagnostic study and you'll be

able to recognize he's

what we can do is determine the strength
of the association

between the study results of the
diagnostic test

of interest and the gold standard
missiles he just comparing the two how

well do they compare against each other

the strength of all of that statistical
significance is the degree of

correlation begin the accuracy or not
yet received the test results all the

two different tests that are on in this
particular set

clinical significance is another factor
we're not going to talk about that

right now

well what we need to do is focus on
statistical significance

so before we break i'm just going to

present to you with the way that i'm
gonna ask you

to think about

and represent that data

from a diagnostic study that does
exactly what

what i'd just out

a diagnostic study you are looking at

the

performance of the test of interest

and the gold standard he's y

t

better out there to buy two table

to be able to represent

how those the study participants a sort
themselves based on how the test of

interested

and how the gold standard label

across the top you always

you always uh... the columns are
represented by those that tested

positive

for the disease based on the gold
standard medical center

work theoretically calling a
hundred-percent

therefore each identified those in the
population who have the disease

and those that test negative by the gold
standard

will be in the second call

but all of these populations

in the study

there will be some of those that also
tested positive based on the

experimental test you're interested

and some that test lead

and sometimes they're going to agree

with what the gold standard says

such as inbox a and boxy

sometimes they're going to disagree

with what the gold standard set

based on box b and boxy

for simplicity safe keep in mind that we
arent we are

assuming that the gold standard is a
hundred percent

so we're just gonna assume that whatever
the gold standard said is true

and we're comparing it against test of
interest

each of these boxes have different names
that are going to come get introduced to

you

does that work

disease are are labeled correctly by
experimental tasks

as having the disease based on the gold
standard are called true positives

their inbox it

those that are correctly labelled as not
having the disease by the experimental

test relative to the gold standard are
considered true naked

those that are

falsely labeled as having the ditsy

based on the tax

but they actually build because the gold
standard sent me one

are called false positives

and those that are falsely labeled at
not having the disease

when they actually do

are called false names

the standard nomenclature

gotta remember it

fairly straightforward false positives
false names

these areas of agreement these are the
areas of disagreement

what we're going to do no

is picked up

with our

figuring out how we use this to define
test characteristics

so let's take a five minute break

and then we will come back and talk
about

the test characteristics better rise
from this to my teeth