Probability, heuristics and post-test probability

probability is a number between 0 and 1 that

expresses an opinion about the likelihood that a current state exists or a future event will occur

to decide between options, one doesnt need to know a diagnosis

but needs an estimate probabilities of condition

below the test threshold

probability so low you wont test for it

above the treatment threshold

probability so high that you should start treatment and tests would delay

probability is the key to

interpreting diagnostic information - it always happens in the context of what we know about the patient

at initial stages of history taking it is ok to group probability

broadly

greater precision matters

when making a treatment decision

the probability of a disease is low

do nothing, do not test or treat

probability of the disease is intermediate

get more info

probability of a disease is high

treat without obtaining more info

sources of data to estimate probability

personal, published, patient centered

past experiences and skills for estimating probability is the

most vulnerable to cognitive biases - mastery requires a variety of experience

second most important to influence estimates

published data to estimate

patient centered probability

patient may have clinical finding that seems important in their circumstance but not present in medical literature (history)

sensitivity

proportion of subjects with the condition who have a positive test (true positive test)

specificity

proportion of subjects without the condition who have a negative test (true negative test)

spPin

with high specificity, positive test rules in the disorder - used to confirm leading hypothesis

SnNout

with high sensitivity, a negative test result will rule out disorder - used to exclude active alternative

senstivity and specificity do not take into consideration

pretest probability of the conditions presence

likelihood ratios

used to quantity shifts in probability once the diagnostic test results are known

positive likelihood ratio

odds favoring the condition given a positive test results

negative likelihood ratio

odds favoring the condition given a negative result

likelihood ratio of 1

useless test/no change

a large change in a LR-

closer to 0

a large change in a LR +

>/=10

pretest probability combined with a likelihood ratio can drastically change the

post test probability

dual process theories and behavior

system 1 is based on intuition and gut feeling with minimal thoughts, system 2 is logical and rule-based - sometimes behavior is driven by system 1 and confirmed by system 2

heuristics

mental shortcuts that can get us to a solution quickly, but can also occasionally send us off course

representitive heuristic

process for categorizing something by how closely is resembles a parent population, leads to ignoring prior probability of disease, uses clinical clues that dont predict disease and redundant predictors, can mistake change as meaningful when it may be part of presentation

to adjust representative heuristic

use published accounts of typical features of disease to widen perspective and know prevalence in your clinical setting

availability bias

judging probability of an event by how easily its remembered

base rate neglect

underestimating the probability of an event in favor of specific case information (ignoring how common a condition is)

before attempting to rule in an alternative hypothesis

you should first rule out the leading hypothesis

confirmation bias

looking for signs/symptoms that confirm a diagnostic hypothesis or interpret clinical findings only to support the hypothesis - insufficient data gathering

premature closure

tendency to stop too soon without appropriate consideration of alternative possibilites

bias correction

diagnostic time out, think systemically, was list of diagnoses cut short because of bias, how did you get there - metacognition

red flags

best tool for high consequence, low frequency events - but NOT informative as single findings

for red flags to be meaningful

they must be combined with variables associated with risk, a cluster of red flags

general red flags

symptoms of any kind that present bilaterally

constitutional symtpoms

group of signs and symptoms that indicate the presence of systemic illness - not specific to any body system (general red flags)

constitutional symptoms include

fever, chills, sweating, nasuea, loss of appetite, vomiting, fatigue, weakness, irritability, insomnia, diarrhea, weight loss, dizziness, fainting, pallor, change in voice (should raise awareness of if others are occuring, but not meaningful in isolation)

why to screen red flags

low rate of routine red flag assessment in primary care settings, changes between consultation and our eval, some serious conditions contraindicate PT, therapeutic value in early interventions, ignorance is not bliss

using history for diagnosis

data clusters in medical history records can give information (questionnaire of 23 questions)