Fall 2023 Flashcards
What is a natural disease?
The usual course of disease over time
Prognostic Factor
Something (environmental factor, personal characteristic, behavior) that can be used to estimate the chance of recovery or recurrence
Risk Factor
Something (environmental factor, personal characteristic, behavior) that increases the chance of developing a disease
Prospective Cohort Study
1- Identify Sample at Risk
2- Asses Risk Factors @ Interest
3- Measure who develops outcome/not
Incidence (rate of developing)
Risk Ratio
Retrospective Case-Control Study
1- Collect data/review chart or records to determine who had exposure/risk factor or not
2- Identify “Cases” and “Controls”
Prevalence
Odds Ratio
Prevalence
Number of people who have the outcome at a given point in time / # of people at risk at that point in time
Incidence
Number of new people who develop the outcome over a period of time / # at risk during that period
Risk Ratios // Relative Risk
Likelihood with which those that have the risk factor will develop the outcome compared to those that don’t have the risk factor (prospective)
Odds Ratios
Likelihood with which those who have a condition (or outcome) have been exposed to a risk factor compared to those who don’t have the condition (retrospective)
Relative Risk - Cross Tabulation
a/(a+b)
_______
d/(c+d)
Odds Ratio - Cross Tabulation
ad
___
bc
RR/OR > 1
RR/OR - 1 = __ * 100% =
___% increase in Risk/Odds
RR/OR < 1
1 - RR/OR = __ * 100% =
___% reduction in Risk/Odds
Rule with RR/OR Confidence Interval
If it includes 1, there is no significance
Unadjusted or Crude Odds Ratio
- 2x2 Table or Logistic Regression
- Accounts for One Relationship
- One Independent, One Dependent
Adjusted Odds Ratio
- Calculations using Logistic Regression
- Accounts for Multiple Variables that Might Affect Relationship
- Multiple Independent Variables
Internal Validity
Relationship between independent (intervention) and dependent (outcome) variable
Tight control of experiment leads to higher internal validity
Efficacy of Treatment
External Validity
Extent to which the independent (intervention) and dependent (outcome) can be generalized outside of the experiment
Tight control of experiment can decrease external validity
Effectiveness of Treatment
Efficacy
Benefit of an intervention in an experimental setting
-Control of many variables
Effectiveness
Benefit of an intervention in a real-world setting
-Less control of variables, bias
Threats to External Validity
Selection
Setting
History
Intervention/Protocol
Construct Validity
Statistical Inference
Estimating characteristics of a population (parameter) from sample data (statistics)
-Population aggregate of persons/objects/events that meet a specific criteria
Probability Sampling
Simple Random
Systemic
Cluster
Stratified Random (Proportional / Disproportional)
Non-Probability Sampling
Convenience (Consecutive / Volunteer)
Quota
Purposive
Inclusion Criteria
Primary traits of the target and accessible population that qualifies them as a subject
Exclusion Criteria
Factors that would preclude someone from being a subject
WEIRD Bias
Western (White)
Educated
Industrialized
Rich
Democratic
Construct Validity
Abstract behavior or event that cannot be directly observed, but is inferred from other relevant observable variables
Operational Definition
The measure(s) used by a particular study to quantify the construct
Parametric Tests
-Estimation of Population
-Assume Normal Distribution
-Often continuous/discrete variable
Ex: T-test, ANOVA
Non-Parametric Tests
-Not based on population parameters
-Assuming Non-Normal Distribution
-Nominal or Ordinal Variables
Ex: Mann-Whitney U Test, Wilcoxon Signed Rank Test, Friedman/Kruskal-Wallis ANOVA, Chi-Square or Fisher Exact Test
ANOVA
More than 2 groups (independent or dependent/repeated measures)
T-Test
Between two groups (independent or unpaired)
Within one group (before & after intervention) (dependent or paired)
alpha
% chance that we found a difference but there wasn’t one (chance of type 1 error)
p
How rarely we would expect a difference this large (found in the study) by chance
Effect Size - Index/Ratio
Difference between groups (treatment + error) / variability within groups (error) = Cohen’s d
Cohen’s d
Index/Ratio
0.2 - 0.5 = small
0.5 - 0.8 = medium
> 0.8 = large **lower variability/rule of thumb too large
Type I Error (alpha)
Conclude an effect is real when it is actually due to chance // found a difference when there really isn’t
-extremely large samples make it easier to find significant effect
1 - .95 = .05(alpha) —> 5% risk of Type 1 Error
Type II Error (beta)
Conclude an effect is due to chance when it is actually real // found no difference when there really is
-not enough subjects per group to detect difference
.8 power = 1 - beta(.20) —> 20% risk of Type 2 Error
a priori
planned comparisons prior to data collection
can correct for Type 1 Error with alternate t-test (Bonferroni)
Post-hoc
unplanned comparisons made after data is analyzed (ANOVA)
can correct for Type I error with alternate t-tests (Tukey’s)
Confounding Variable
Anything other than the independent variable that could potentially affect the dependent variable
Stratification
Equal recruitment in each group to decrease confounding variable (known prior to recruitment)
Blocking
Randomized sequence in pre-determined blocked groups to ensure equal distribution between groups
Bias
Any tendency which prevents unprejudiced consideration of a question
Selection Bias
Selecting non-equivalent groups
RCT
Maturation Bias
Time on the health condition recovery (natural history)
RCT
Concealment
When the person who is screening the participants for eligibility does not know the randomization sequence
Performance Bias
When participants or personnel perform differently based on knowing what group they are in
Detection Bias
When the personnel assessing the outcomes perform differently based on knowing what group participants are in
Blinding
Reduces bias by keeping researchers and/or participants ignorant of group assignments and/or research hypotheses (single/double/assessor)
Attrition Bias
Effects of loss of participants during the trial (drop outs)
Standardized Reporting
Reporting number loss to follow-up at each time point (intention to treat analysis performed)
Intention to Treat
Data analyzed according to group assignment, regardless if subject completes study or “switches” group
Attrition > 20% then question results
Testing Bias
Confounding effects of repeated testing or use of an outcome
Instrumentation Bias
Poor reliability of tests/outcomes can influence outcomes
Standardization
Standardize # of Tests/Order or Valid Outcomes/Standardize Procedures
PICO
Population
Intervention
Comparison
Outcome
Health Condition
Medical Disease, Disorder, Syndrome, or Injury
Participation Restriction
Difficulty with Involvement in life situation or social role
Activity Limitation
Difficulty with Execution of a task or action by an individual
Impairments
Problems in body structure(s) and/or function(s)
Personal Factors
Any identifiable element intrinsic to a person (including physical and psychological co-morbidities) that influences the way that person conducts his or her life
Environmental Factors
Any identifiable element extrinsic to a person (including physical and psychological co-morbidities) that influences the way that person conducts his or her life
Background Searching
-General Information
-Novice on Topic
-General Question
Foreground Searching
-Specific Clinical Question
-Comparison of Intervention
-Properties of Test and Measures
-Prognostic Factors
-Answers clinical question applied to patient
Where to do Background Searches
APTA
Physio-pedia
RehabMeasures
UptoDate
Clinical Key
Where to do Foreground Searches
PubMed
PACO
Person
Assessment (Test/Measure)
Comparison Assessment
Outcome
PECO
Person
Exposure (Risk/Prognostic Factor)
Comparison Exposure
Outcome
What sections of a research paper are most objective?
Methods & Results
What sections of a research paper are most subjective?
Abstract
Intro
Discussion
Conclusion
Cross Sectional Design
-Observations/Surveys/Measurements made at one time point (or multiple, but no intervention)
-Correlation/Association of 2 Variables
Advantages/Limitations of Cross Sectional Design
Advantages:
-Efficient
-Inexpensive
-Minimal Loss to Follow Up
Limitation:
-No Time Component
-Correlation/Association Only (No Causation)
Prospective Cohort Design
-Identify sample, “Follow” over time
-Common design for prognosis
-Emphasis on “Risk of Developing”
Advantages/Limitations of Prospective Cohort Design
Advantages:
-Establish Temporal Association
-Short Intervals of Exposure to Outcome & Common Conditions
-Multiple Exposures
Limitations:
-Expensive
-Long Follow Up (Attrition)
-Not Ideal for Rare Outcomes
Retrospective Design
-Identify Outcomes and “Go Back” to Determine Exposures
-Emphasis on odds of having a past exposure given already having the condition
Advantages/Limitations of Retrospective Design
Advantages:
-Less Expensive
-Little/No Attrition
-Better for Rare Outcomes
-Relatively Easy to Collect Data
Limitations:
-Dependence on Medical Records/Patient Reporting
-Recall Bias
-Less Certain for Temporal Relationship
-Prevalence not Incidence
Randomized Control Trial Design
-Studying Intervention Efficacy
-Compares at least 2 groups
-Randomly assigned groups
-Intervention Implemented, while controlling other variables
Advantages/Limitations of RCT
Advantages:
-Controls/Reduces Bias
-Creates Similar Groups at Baseline
-Conclusions to Infer Cause/Effect
Limitations:
-Costly & Time Intensive
-May not be Generalizable
-Attrition Risk is High
Quasi-Experimental Design
-Lack Random Assignment/Comparison Group
-Or only one group gets intervention (pre/post)
-Increased Bias to RCT
Case Report or Case Series Design
-In-depth study/description of a patient
-Rare condition, Combination of Conditions, Novel Treatment
-Insight into Clinical Decision Making
Parameters
Measured Characteristic of a Population
Statistics
Measured Characteristic of a Sample
Continuous
-Data that takes on any value along a continuum
-Infinite Number of Fractional Values
Ex: Strength, Distance, Weight
Discrete
-Can only be described in whole units
-Cannot be divided further
Ex: People, Number of Groceries you buy
Dichotomous
Can only have two values
Ex: Heads/Tail, Yes/No
Ratio
Numbers represent units with equal intervals (related to zero)
Ex: distance, age, time, weight, strength, ROM
Interval
Equal intervals between numbers (not related to zero)
Ex: Calendar Years, IQ, Degrees
Ordinal
Numbers indicate rank order of observations (unequal intervals)
Ex: MMT, Pain Scale
Nominal
Numerals represent category labels only
Ex: Sex, Nationality, Blood Type, Clinical Diagnosis
Independent Variable
-Can predict or cause a given outcome
-What is being controlled/manipulated
Dependent Variable
-Response that varies with changes in the independent variable
-The Measured Outcome
Mean
Sum of Scores / # of Scores (Average)
Parametric Analysis
-Continuous Variables
-Non-Continuous but Normal Distribution
Median
Divides a rank-order distribution into equal parts
Non-Parametric Analysis
Point Estimate
A single statistic that estimates population parameter
Variability for Mean
Standard Deviation
Variability for Median
Range (difference between lowest and highest values)
Interquartile Range (divides into four equal quartiles)
Confidence Interval
-A measure of variability of the mean
-Estimates a range of scores that can contain the “true” population mean
-Specified Certainty (typically 95%)
Type I Error
Acceptable Risk of Falsely Concluding that an effect is real
Alpha ~ 0.05 … 5% chance a difference was found, when there really wasn’t
Probability / P-Value
Probability of finding an effect as big as the one observed in the study BY CHANCE
P < Alpha
Significant Effect
P > Alpha
No Significant Effect
Correlation + 2 statistics used
A measure of the relationship between two variables
Pearson product moment correlation (r) = normally distributed
Spearman’s rho (p) = not normally distributed
Positive and Negative values for Correlation
+ = Directly or Positively related
- = Inversely or Negatively related
Simple Linear Regression
Establishes relationship between two variables as a basis for prediction
(scatter plot, regression equation, evaluate)
R^2 (Coefficient of Determination)
Accuracy of prediction of linear regression equation
0 = no accuracy
1.0 = full accuracy
Multiple Linear Regression
Established the predictive relationship between a set of independent variables and one dependent variable
Outcome Measure vs. Diagnostic Test
Measure & Track Changes
Identify, Diagnose or Discriminate
Validity
The extent to which a test is measuring what it is intended to measure
Convergent
Two measures reflecting the same construct will have similar results in the same group
or
The ability of a test to measure same construct in different groups
Discriminant
Two measures reflecting different constructs will have different results when measured in same group
or
One measure having different results in different groups
Concurrent
Measures compared with each other at the same time
Often to establish properties of newer, more feasible measure
Reliability
The extent to which a measure is consistent and free from error
Test-Retest Reliability
measures the stability of an instrument or test (consistency)
EX: identical test performed on two different occasions
Rater Reliability:
Intra vs. Inter
Intra - stability of data recorded by a single rater across multiple trials
Inter - stability of data recorded by multiple raters when measuring the same variable
Intraclass Correlation Coefficient (ICC)
A single index that reflects both correlation and agreement among ratings
0 - not reliable
1- perfectly reliable
Reference Standard
The “Gold Standard” of an existing diagnostic tool
Validity is to Reliability
as
Diagnostic is to ______
Outcome Measure
Sensitivity
-The probability that someone with the condition will test positive
-The ability of the test to be positive when the condition is present
Specificity
-The probability that someone without the condition will test negative
-The ability of a test to be negative when the condition is absent
What type of error is correlated to sensitivity?
Type I Error
False Positives
What type of error is correlated to specificity?
Type II Error
False Negatives
Utilizing the Receiver Operating Characteristic (ROC) Curve with sensitivity and specificity, what constitutes excellent vs useless?
Area under the curve
.9-1.0 = excellent
<.6 = useless
Positive Predictive Value
Probability of a. person testing positive and having the condition
(true positive / (true + false positive))
Negative Predictive Value
Probability of a person testing negative and not having the condition
(true negative / (true + false negative))
SPin
With high specificity, a positive test rules IN the diagnosis
SNout
With high sensitivity, a negative test rules OUT the diagnosis
Positive Likelihood Ratio
How many times more likely a positive test will be seen in those with the disorder than in those without the disorder
(Sensitivity/1-Specificity)
(True Positives / False Positives)
Negative Likelihood Ratio
How many times more likely a negative test will be seen in those without the disorder than in those with the disorder
(1-Sensitivity/Specificity)
(False Negatives/True Negatives)
Closer or Further from 1.0 indicates more important results for Likelihood Ratios?
Further
Narrative Review
Summary of literature, often single author, does not address question rather it is a background on a topic, potentially biased
Systematic Review
Usually PICO formatted question, systematic search, structured appraisal, critical assessment and evaluation of research
Meta-Analysis
Systematic review using quantitative methods to combine results (pooling)
Publication Bias
The publication or non-publication of research findings, depending on the nature and direction of the results
(not publishing because the outcome was not interesting or in the direction desired)
Two tools for assessing risk of bias
PEDro Scale (Physiotherapy Evidence Database)
Cochrane Risk of Bias Tool
Heterogeneity
The scale of differences between study’s ….. populations, treatment groups, design of study, attrition, length of follow up, treatment, outcomes, interventions
Homogeneity and chart used
When studies have common ______.
Forest Plot used to compare different values amongst the individual studies.