Exam 1 Flashcards
Nominal
Qualitative or categorical
Examples: Ethnicity, race and gender
Nominal in medicine
categorical variable with only two categories
Example: Outcomes of medical treatment or surgical procedure: succesful/not
Ordinal
ranks-order cases in terms of quantity of a particular characteristic.
Examples: 1st, 2nd, 3rd….
Ordinal Scales in Medicine
Determines a patients amount of risk or the appropriate type of therapy.
Percentages and proportions are often used
Numerical Scales
Continuous-quantitative, interval, dimensional
Examples: height, weight, IQ, Age
Numerical Scales-Discrete
A variable in which possible scores have limited values.
Examples: Numbers of brothers, sisters, operations, and bone fractures
Descriptive Statistics
Central tendency gives us an idea of the typical value of a distribution
Measures of the middle -center of distribution
Dispersion (or Variabiliity)
tells us how different the data is from eachother, how spread out it is
Inferential Statistics
tells us if the difference we see is likely caused by chance (if real or not)
if descriptive statistics are probably accurate
if sample represents population
Variance
A variable that causes variance: experimentation or observation
Mean
a measure of central tendency
Average, Xbar
mode
a measure of central tendency.
the value occurs frequently
median
a measure of central tendency
the middle observation or the point at which half the observations are smaller and half are larger
A score at 10th percentile is
higher than 10% of all scores
A score at 75th percentile is
higher than 75% of all scores
The median is the score at
the 50th percentile
mean > median
positively skewed
mean < median
negatively skewed
Standard Deviation
used to measure dispersion with medical and health data
average of the spread of the observations around the mean
average distance of scores from their own mean
Variance (mean square)
the mean of the squared deviations
Correlation Coefficient
How strong is the relationship between two continuous variables?
if slope of SD line is positive
R is positive
Postive Correlation
always lie between 0 and 1
if r is close to 1=strong relationship
Negative correlation
Always lie between -1 and 1=if SD line goes down and r<0
if r=-1, perfect negative relationship
Randomized Control Trials
measures primary outcome of randomly assigned participants
participants have an equal likelihood of being assigned to intervention
strongest study desgin
required for FDA and NDA
In RCT
blinding is important
Validitty
the extent to which an instrument measures what it is intended to measure
degree to which findings are correct
Internal Validity
The outcome of interest (dependent variable) caused by the treatment (independent variable)
How to strengthen internal validity?
include a control group
random assignment-equally distributed across groups
why is internal validity important?
establishes cause and effect relationship
strong evidence of causality
low degrees of internal validity=little or no evidence of causality
threats to internal validity
RAMSHIT
regression, attrition, maturation, selection, history, instrumentation, testing
Selection Bias
systematic error in the estimate of the effect due to procedures used to select subjects or factors that influence study participation
differences in patients baseline characteristics in two groups can lead to
selection bias
Suppose that a weight loss drug is given to individuals who volunteered to be part of a weight loss program and that the comparison condition includes only individuals who were not volunteered in the weight loss program.
* What differences related to selection would you expect between the two groups?
The volunteering individuals might have more motivation to eat healthier, exercise more often, or otherwise differ from non-volunteers in ways that might affect their weight loss achievement. So individuals who were given the drug that volunteered might have lost more weight, even without the new drug because of their motivation.
history
changes in the outcomes of a study due to the occurrence of external events during the course of the study
Maturation
normal changes in study participants over time
Attrition/Experimental Mortality
caused by differential drop out of patients in treatment and control groups in RCT
Testing
changes in outcomes due to repeated prior assesments
instrumentation
changes in the outcomes due to instrumentation or technique used to measure the outcome
Regression
shift in the initial extreme measures towards the mean or average in subsequent measures due to statistical variability
extreme groups
External Validity
the extent which the results of a study can be generalized to other populations or settings
threats to external validity
treatment interaction with subject selection, settings of the study, historical factors
External Validity examples:
interactions with treatment and:
subject selection
pre-testing–leads to it cannot be generalized
settings
setting-could have low external validity
study in past may not apply to future
Hawthorn Effect
modifications in a study subject’s behavior because of the fact that she is being studied or observed
Multiple treatments
multiple treatments can have a significant effect on the results
Bias
IPADS
investigator, performance, attrition, detection, selection
What is Bias?
systematic error in study design in the way subjects are selected, measured and analyzed leading to incorrect findings
Investigator Bias
can minimize by binding
errors in study design, implementation, or analysis
after study-ascertain bias could be present
ascertain bias
due to differences in assessing or analyzing outcomes by the researcher due to awareness of which participants received the active versus control interventions
Performance Bias
due to systematic differences in care between treatment groups or in exposure to factors other than the intervention being studied
attrition bias
can be minimized to intent to treat
dropouts of patients
if the data is analzyed only including smaller groups (excluding the drop-outs)
detection bias
can be minimized by the use of non-study personnel to assess patient outcomes
overestimation
when the investigator is aware of the study treatment and makes an assessment of the outcome
selection bias
can be minimized by random assignment
preferential enrollment of specific patients into one treatment group over another
randomization
assigning patients randomly/by chance
highly effective in reducing biases and confounding factors
confounding factors
a factor that is associated with both exposure (treatment) and the outcome
influences treatment
simple randomization
random number generator to allocate participants
leads to unequal sample sizez
block randomization
process of dividing subjects into a specified number of blocks to be randomized at the beginning of the trial
ensuring groups are equal
stratified randomzation
ensures balance of participants for predefined strata based on prognostic factors such as disease severity
like age, race, gender, and disease severity differences -in small samples stratification helps more than in large groups
single blind
only 3 categories of individuals (usually participant) is unaware of the intervention assignment
double blind
both participants and investigators are unaware of the randomization schedule
for studies involving investigational agents (phase 3 trials), at least 2 DB trials are required for the drug to be approved
Triple Blind
most objective design-patients and investigators are blinded, as well as the external group of individuals monitoring the study
Open Label
least objective label-a study that involves unblinded participants, investigators, and assessors. EVERYONE is AWARE
statistical power
likelihood to detect an affect in a sample, if the effect truly exists in the population
studies typically have 80% power
inclusion criteria
the characterisitics the invesitgator is most interested in studying
exclusion criteria
factors that would confound or impair the ability to interpret the study results
placebo control
group of patients only receive an inert pill that includes all the extraneous conditions except the active ingredients
active control
known or accepted standard of care in a RCT
Historical Control
external group who were observed at different times
internal valditiy concerns=need large sample size
non-inferiority trials
seeks to determine whether a new therapy is no worse than a standard therapy
does not asess if one is better than the other.
they see if they are equivalent
parallel study design
each subject is randomized to either treatment group or placebo group only
strong design and shorter time period needed but requires a large sample size
cross over design
subjects receive intervemtions on a specified sequence of events (washout period)
most statistical power with fewer subjects but need to enroll patients with stable disease states
factorial randomized trials
multiple dose levels and multiple drug regimens
cluster randomization
selection of a specific group of subjects for randomization such as those enrolled in a clinic or hospital
adaptive designs
changes the condition of study plan over time based on the results of the preliminary analysis at interim points of time
can reduce number of subjects needed
Drug Efficacy Study
determines the effects of intervention under tight control
examples: BP, seizures, survival , quality of life (survey-yes/no)
Drug effectiveness
determines the effects of the intervention under the conditions that the drug is most often used in the clinical setting