Prognostic Studies Flashcards
Area under the curve
measure of the accuracy of a quantitative diagnostic test
ROC Curves Axis
Y = true positive
X = false positive
ROC Curve
a plot of sensitivity against 1=specificity, used to determine the most effective cut-off score to make a diagnosis/prognosis
defines cutoff scores, helps predict the outcome of interest
the more area under the curve, the better it is at predicting the outcome of interest
What is the significance of RR/OR?
RR and OR are estimates and are subject to error
you need to have a p value or confidence interval to determine that the result is not due to chance
Odds Ratio
proportion of participants who have the outcome of interest compared to the proportion of participants who do not have the outcome of interest
comparing columns
Relative Risk
proportion of participants who have the risk factor and the outcome compared to the proportion of participants without the risk factor who have the outcome
comparing rows
Relative Risk ratio calculation
(Outcome +, Risk factor +)/ (Outcome +, Risk factor +) PLUS (outcome -, risk factor +)
DIVIDED BY
(outcome +, risk factor -)/(outcome +, risk factor -) PLUS (outcome -, risk factor -)
Clinical Significance
<1 = decreased risk or odds of developing outcome
(>) 1 = increased risk or odds of developing outcome
= 1, 50/50 chance
Risk Ratio
predicts the probability of an outcome when a risk factor is present vs not present
used in longitudinal cohort studies
Logistic Regression
used to predict a categorical outcome variable
Ex: fall vs no fall, discharge vs no discharge
Multiple Regression
used when there are multiple variables contributing to an outcome variable or when the outcome is a continuous variable
uses a weighting process
y=a+bx1+bx2+bx3…
B is equivalent to
slope of the line
regression coefficient
A is equivalent to
intercept of the regression line at y
also known as a regression constant
Y is equivalent to
outcome of interest or the value you are trying to predict
Simple Linear Regression
exam of two variables that are linearly related to determine how well a variable predicts (x) an outcome (y) variable
can be positive or negative
y = a+bx
Standard error of the estimate
represents the average error of prediction for the regression equation
Confidence Intervals
provides info about the accuracy of the prediction
P-value
estimates the probability that chance contributed to the prognostic equation
does not indicate clinical meaningfulness
r²
coefficient of determination
represents the robustness of the regression model
proportion of variance that is shared by two variables
Regression
Used to make predictions from one or more variables about the outcome of interest
uses a regression analysis to find the line of best fit to predict the outcome variable
the closer R2 is to 1, the more robust the prediction
Correlation can be useful for
developing prediction or prognostic models
Correlation does not equal
causation
Very high correlation
.9 - 1.0
High correlation
.7 - .89
Moderate correlation
.5 to .69
Low correlation
.26 to.49
Little or no correlation
0 to .25
Negative correlation
as one variable increases, the other decreases
Positive correlation
as one variable increases, so does the other
Correlation
a measure of the extent to which two variables are associated. calculated as a correlation coefficient
can either be positive or negative
Interpret the results of prognostic study
what statistics were used to determine the prognostic statements?
Were the stats appropriate (correlation/regression)?
Study Process
Were evaluators masked/blinded to reduce bias?
Was the study time frame long enough for the participants to experience the outcome of interest?
Was the monitoring process appropriate?
Were all participants followed to the end of the study?
Outcome measures and factors associated w/these measures
Were end points clearly defined?
Are the factors associated with the outcome well justified in terms of the potential for their contribution to prediction of the outcome?
should be systematic, precise, relevance, defined, reliable, valid
Determining quality of prognostic study
study design
study sample
outcome measures and factors associated w/measures
study process
Is there a defined, representative sample of pts assembled at a common point that is relevant to the study question?
Were they randomized into groups vs part of a cohort or case-control group?
Were they assembled at a common point?
Are there enough pts?
Study participants must not already have the study outcome
Types of prognostic studies
Cohort
Case Control
Cross-Sectional
Classification measures
sensitivity
specificity
predictive values
clinicians need to verify the chosen thresholds is relevant for their pt
Discrimination is the same as
accuracy
Calibration is the same as
reliability
Performance measures
Prognostic Models must demonstrate: Calibration, Discrimination
without these, model may be inaccurate and can mislead decision-making of PT
Clinical Decision Rules
suggest a course of action
treatment-effect modifiers are used to build a CDR
guides clinicians in their decision-making and care pathways by predicting treatment response
represented as regression/classification tree, score chart rules, or survival groups/meta-models
Clinical Prediction Rules
estimates the probability of future outcomes
prognostic factors help to build it
should be consistent to be helpful
presented as regression formula/calculator, nomogram or table/score chart
Treatment-Effect Modifiers
Characteristics that predict a pt’s response to a treatment
influence the relationship between an intervention and an outcome
used to create clinical decision rules
Prognostic factors
characteristics found in individuals with a disease
can influence the course of the disease
used to create clinical prediction rules
Risk Factors
characteristics found in individuals that are healthy
factors that increase the risk of developing a health condition
important to PTs who have an interest in prevention
Dimensions of Prognosis
Risk Factors
Prognostic Factors
Treatment-effect modifiers
Why is it beneficial to target prognostic factors?
improve treatment decision-making process
personalize rehab approaches
enhance pt outcomes
help better align clinical practice with value-based care principles
Contemporary view of prognosis
prognostic-related findings must inform clinical decision making to enhance patient health outcome
Traditional prognosis method
prognosis is driven by pathoanatomical diagnosis to guide treatment and guide prognosis. has variable outcomes
Prognosis definition
predicting which patient will have the best outcome or respond to a particular treatment
traditional and contemporary views