Exam Two - Clinical Prediction Rules Flashcards
CPR use a combination of clinical findings to predict:
- presence of a disorder
- likelihood of a treatment outcome
- most likely prognosis
What is the goal of CPR?
intended to simplify and increase the accuracy of clinicians’ diagnostic and prognostic assessments
CPR is used in combination with….
clinical intuition!
- not a replacement for experience
- reduce the influence of bias in clinical judgements
CPRs are especially useful when…
- decision making is complex
- clinical stakes are high
- opportunities exist to avoid unnecessary tests without compromising patient care
CPRs quantify what three aspects of our practice?
- diagnosis
- prognosis
- intervention
Diagnosis CPRs are frequently designed to maximize _______ in order to better rule ______ the diagnosis
sensitivity
out
A _________ diagnostic CPR rules out the fracture
negative
A _________ diagnostic CPR does not rule in the fracture because its not __________
positive
specific
What question does intervention CPRs seek to answer?
what patient characteristics predict a positive outcome following an intervention?
Nearly all intervention CPRs were derived using a ________ study design. Is that good?
“single arm”
nope because there’s no control group or covariation of cause and effect
Single arm study designs seek to answer:
what is similar among responders that is different from nonresponders?
- use multiple logistic regression
describe pathway of single arm intervention CPR
get group, then measure outcome of interest and possible predictors, then apply intervention, then measure change in outcome of interest, the separate the responders from nonresponders
many PT intervention CPRs define “improvement” using the ______- scale
Global Rating of Change (GROC) scale
Prognosis CPRs predict the…
natural course of a condition, or its risk for development
What is a prognostic factor?
a characteristic that increases or decreases a person’s risk for a certain outcome
What are a few types of prognostic factors?
- demographic
- disease-specific
- co-morbidities
- other (insurance, access to healthcare)
What are favorable prognostic factors?
- increase likelihood of a positive event or decrease the likelihood of an adverse event
What are unfavorable prognostic factors
- increase the likelihood of an adverse event or decrease the likelihood of a positive event
nearly all prognostic CPRs were derived using a _____ study design. Is this good?
single arm
yes, it’s good. you aren’t giving an intervention, you’re just letting time pass before you measure outcomes of interest again
describe pathway of single arm prognosis CPR
get participants, then measure possible predictor variables, then let time pass (no intervention), then evaluate outcome of interest again, then separate participants into yes/no
Why are we concerned with intervention CPRs not having a control group?
with no control group, the identified predictors may simply indicate which patients will get netter over time (alone)
If we do a two arm study design for intervention CPRs, what experimental design does it mimic?
RCT
- allows us to test intervention effectiveness
- allows ID of exam findings that predict success with that intervention
What are some cons to a two arm study design for intervention CPRs
more subjects, more expensive, more complicated
Typically, CPR results include:
1 - sensitivity
2 - specificity
3 - likelihood ratios
why is sensitivity useful for intervention CPRs?
negative test results for CPRs with high sensitivity are useful for ruling out a condition or outcome
why is specificity useful for intervention CPRs?
positive test results for CPRs with high specificity are useful for ruling in a condition or outcome
why are likelihood ratios useful to have for intervention CPRs?
it tells us to what extent we should shift our suspicion for a condition or outcome
What is the process of CPR development?
level four - derivation
level three - narrow validation
level two - broad validation
level one - impact analysis
step one of derivation of CPR
define the condition or outcome of interest
- consider measurement validity
step two of derivation of CPR
compile a list of variables that may predict the condition or outcome of interest (do a lit review, expert opinion, or focus groups)
important that all logical potential predictors are included BUT you need at least 10 subjects per possible predictor
T or F: authors need to provide a rationale for predictor variables?
true, because otherwise there is an increased risk of spurious findings that cannot be explained
step three of derivation
measurement of predictors variables and track outcomes
consider
- reliability/validity of predictor variables
- subject inclusion/exclusion criteria
- masking to limit bias
step four of derivation
collect/analyze data
prospective - measure predictor variables and track subject outcomes
retrospective - evaluate subject outcomes and predictor variables measured previously
How do you determine which combination of variables best predicts the outcome of interest?
multiple logistic regression
what does narrow validation do?
demonstrates that repeated application leads to the same results in the same setting
what does broad validation do?
answers how much you can generalize the CPR
what does the impact of the CPR answer?
how helpful is this clinically? does it even matter? who cares? is it useful
widespread use of CPR is not technically recommended until:
1 - it has been validated using a prospective design in more than one setting (level 2)
2 - an impact analysis complete (level one). it may show that CPR doesn’t make a difference in patient outcomes or cost
When is it acceptable to use CPRs before they’ve been appropriately validated?
1 - under unique clinical circumstances where clinician uncertainty is greater
2 - when the PT setting, presentation, and expertise is similar to the CPR
3 - the magnitude of the effect of the CPR is especially great/high