1 - Variation, Diag. Testing, Decision Analysis Flashcards
Define continuous scale. Give an example.
A scale used to measure a numerical characteristic in which fractional values can occur
–Ex. body temperature
Define dataset.
A collection of data organized into observations and variables
Define discrete scale. Give an example.
A numerical scale using only whole numbers
–Ex. number of pregnancies
Define distribution.
The values of a characteristic or variable along with the frequency of their occurrence
May be based on empirical observations or may be theoretical probability distributions (normal, binomial, chi-square)
Define histogram.
A graphical display of a distribution, illustrating how frequently each value occurs
Define mean.
A measure of central tendency
The sum of the values divided by number n in the sample
Define measures of central location/tendency. Give an example.
Index or summary numbers that describe the middle of a distribution
–Ex. mean, median, mode
Define measures of spread. Give an example.
Index or summary numbers that describe the spread of observations about the mean
–Ex. range, standard deviation
Define median.
A measure of central tendency
The middle observation (the one that divides the distribution of values into two halves)
Equal to the 50th percentile
Define mode.
A measure of central tendency
The most commonly observed value of a distribution
Define nominal scale. Give an example.
The simplest scale of measurement
Used for characteristics that have no numerical values
Aka categorical or qualitative scale
–Ex. race, gender
Define normal distribution. What percent fall within 1 SD of mean? 2 SD?
A symmetric, bell-shaped probability distribution with mean u and standard deviation sigma
If observations follow a normal distribution, 68% fall within 1 SD of the mean and 95% of observations fall within 2 SD of the mean
AKA Gaussian distribution
Define ordinal scale. Give an example.
Used for characteristics that have an underlying order to their values
Numbers used are arbitrary
–Ex. Apgar scores
Define population.
The entire collection of observations or subjects that have something in common and to which conclusions are inferred
Define range.
The difference between the largest and the smallest observation
Define skewed distribution.
A distribution in which there are a relatively small number of outlying observations in one direction only
- -If outlying distributions are small, skewed left/negatively skewed
- -If outlying observations are large, skewed right/positively skewed
Define standard deviation.
The most common measure of dispersion or spread
Can be used with mean to describe distribution of observations
=Square root of variance
Define variance.
The square of the standard deviation
A measure of dispersion in a distribution of observations in a population or sample
The sum of squared deviations of the observations from their mean, divided by n-1
Define baseline analysis.
In a decision analysis, the expected value of each strategy calculated using best estimates of each probability and utility
Define chance node.
Intermediate branch in decision tree from which chance events occur
Define decision analysis.
A formal, quantitative approach to examining trade-offs when making patient-care decisions
Define decision node.
Proximal branch in decision tree that specifies clinical strategies under consideration
Define decision tree.
A diagram used in decision analysis to illustrate the possible clinical options and outcomes
Define expected value.
The relative value of clinical strategy, often expressed in quality-adjusted life-years
Define false-negative rate.
Probability of a negative test result in a patient who has the disease being tested for
Equal to 1 - sensitivity
Define false-positive rate.
Probability of a positive test result in a patient who is free of the disease being tested for
Equal to 1 - specificity
Define fold back.
In a decision analysis, the process of calculating expected values by summing outcome utilities, each weighted by its probability of occurrence
Define gold standard test.
A diagnostic test used to ascertain the true disease status when estimating the sensitivity and specificity of another diagnostic test
Define negative predictive value.
Probability of patient does not have the disease being tested for following a negative diagnostic test
Define operating characteristics.
The accuracy parameters of the diagnostic test: sensitivity, specificity, positive predictive value and negative predictive value
Define positive predictive value.
Probability that a patient has the disease being tested for following a positive diagnostic test
Define post-test probability.
An estimate of the probability a patient has a given disease after the results of a diagnostic test are known
AKA posterior probability
Define pre-test probability.
Estimate of the probability a patient has a given disease prior to the use of the diagnostic test
AKA prior probability
Define probability revision.
The process of computing positive predictive value and negative predictive value from the pre-test probability
Define quality-adjusted life-years (QALYs).
Common measure of utility based on multiplying life expectancy by a quality-adjustment factor
Define sensitivity.
Probability of a positive test result in a patient who has the disease being tested for
AKA true-positive rate
Define sensitivity analysis.
Process of testing stability of a decision analysis by allowing input probabilities and utilities to vary
Define specificity.
Probability of a negative test result in a patient was free of the disease being tested for
AKA true-negative rate
Define terminal node.
Most distal node in decision tree
Represents a final clinical outcome or condition
Define threshold value.
In a decision analysis, the value of any input probability or utility that makes the expected values of two clinical strategies equivalent
Define utility.
The relative value of a clinical outcome in a decision analysis
Describe the three basic types of clinical data and the concept of a distribution.
NOMINAL DATA: categorical scale with no particular ordering (positive or negative, ABO blood typing)
ORDINAL SCALE: a categorical scale with an order (not necessarily equal stages) among the categories (stage I-IV cancer)
NUMERICAL DATA: discrete = counting of objects (number of pregnancies); continuous = includes fractions (temperature, height)
DISTRIBUTION: the pattern of the values of a variable; captures how frequently each value occurs
What things should you consider when deciding to use mean or median?
Skewed: median is closer to center (mean affected by extreme values
Mean easier to work with
Median tougher to compare
Understand the concept of sampling and estimation of population parameters.
a
Define and contrast qualitative versus quantitative assessments of clinical uncertainty.
QUALITATIVE:
- -Likely, probable, suspicious
- -Unlikely, possible, can’t rule out
- -Lack precision, open to interpretation
QUANTITATIVE: [probabilities]
–Assessed on continuous scale of 0 (impossible) to 1 (certain)
What questions should be asked about the decision model in clinical decision analysis (structuring the clinical problem as a decision tree)? (2)
Were all relevant strategies included?
Did the authors include all important outcomes
What type of test would you want to use to rule-in a disease? Rule-out?
RULE-IN = high PPV = highly SPECIFIC
RULE-OUT = high NPV = highly SENSITIVE
SPIN = specific test to rule in SNOUT = sensitive test to rule out
What questions should be asked when assigning PROBABILITIES to all potential clinical outcomes included in a clinical decision model? (2)
Are probabilities based on reliable source studies?
Can results found in the literature be generalized to the patient population considered in the decision analysis?
What questions should be asked when assigning UTILITIES to all potential clinical outcomes included in a clinical decision model? (3)
How reliable are the source studies?
Were appropriate methods used to estimate life expectancy?
Are quality of life estimates consistent with reader’s clinical experience?
What questions should be asked about the analysis and interpretation of a clinical decision model? (2)
Is the expected benefit from the favored strategy clinically important?
How stable are the results when probability and utility estimates are allowed to vary
