Biostatistics Flashcards
Continuous data
values that change by the same amount
2 types: ratio and interval data
Ratio vs interval data
Interval data has no meaningful zero (e.g. celsius)
Ratio data has a meaningful zero (zero equals none, e.g. BP)
Discrete (categorical) data
Nominal, ordinal
Have categories
Differences between ordinal and continuous data
In ordinal data, the categories do not increase by the same amount
Measures of central tendency
Provide simple summaries of the data
Mean, Median, Mode
Spread (variability) of data
Range
Standard deviation (how dispersed the data is from the mean)
Gaussian distributions
Normal (Bell-shaped)
Large sample sets of continuous data
Characteristics of a gaussian distribution
Symmetrical curve
Median, Median and Mode are all the same
68% of the values fall within 1 SD, 95% of the values fall within 2 SDs
What causes skewed (not symmetrical) distributions?
small sample size +/- outliers
When is the Median a better measure of central tendency?
When there are outliers on a small # of values. In this case, the outliers can have a big impact on the mean.
alpha
error margin
commonly 5% (0.05)
Determining CI in ratio data vs difference data
Difference data (mean): not significant if the CI crosses 0 (e.g. change in FEV1)
Ratio data (OR, HR, RR): not significant if the CI crosses 1
CI & precision
Narrow CI = high precision
Wide CI = less precision
Type 1 error
False positives
When the alpha is set as 0.05 and the study reports a P value < 0.05, it is statistically significant and the probability of making a Type I error is 5%
**CI = 1 - alpha (type 1 error)
Type II error
Denoted as beta (B)
power
Power to avoid a type II error
How is power determined?
of outcomes
Difference in outcome rates
Significance (Alpha) level
Relative risk
risk in treatment group/risk in control group
Interpret a RR of 57%
Patients in the treatment group were 57% as likely to have the outcome as the placebo group
Relative risk reduction
1-RR
Interpret a RRR of 0.43
Patients in the treatment group were 43% less likely to have the outcome than the placebo group
Absolute risk reduction
(% risk in the control group) - (% risk in the treatment group)
Interpret a ARR of 12%
It means that 12 out of a 100 patients will benefit from the treatment
Number needed to treat
1/ARR
Always round up
NNH
Always round down
In what type of studies is odds ratio used?
Used in case-control studies to estimate the risk of unfavorable events
OR calcuation
AD/BC
Hazard ratio
Used in survival analysis
Rate at which an unfavorable event occurs over a short period of time
HR formula
HR (treatment group)/HR(control group)
Interpret a HR of 2
Means that there are twice as many deaths in the treatment group
Interpret a HR of 0.5
Means that there are half as many deaths in the treatment group
When are parametric tests appropriate?
When data is normally distributed
When are nonparametric tests appropriate?
When data is NOT normally distributed
Parametric test for 1 group (continuous data)
One-sample T-test
parametric test for 1 group (with before & after measurements) (continuous data)
Dependent/paired t-test
parametric test for 2 groups (continuous data)
Independent/unpaired student t-test
parametric test for ≥ 3 groups (continuous data)
ANOVA (or F-test)
Discrete/Categorical test for 1 group
Chi-square test
Discrete/Categorical test for 2 groups
Chi-square test or fisher’s exact test
Correlation vs regression
Correlation - one variable
Regression - used to assess multiple independent variables or if there is a need to control many confounding factors
Sensitivity
The true positive
100% sensitive test will be positive in ALL patients with the disease
Specificity
The true negative
100% specific test will be negative in all patients without the disease
Sensitivity formula
A (positive, have condition)/ A + C (negative, have condition)
Specific formula
D (Negative, no condition)/B(positive, no condition) + D
Interpret: sensitivity of 28%
Means that only 28% of the patients with the condition will have a positive result, but, 72% of patients with the disease can test negative (missed diagnosis)
Interpret: specificity of 87%
Test is negative in 87% of patients without the disease, but 13% without the disease can test positive (incorrect diagnosis)
Intention-to-treat vs per-protocol study
ITT: includes data for all patients originally allocated to each treatment group even if the patient did not complete the trial according to the study protocol
PP: analysis completed for only those that completed the study according to the protocol
Equivalence vs non-inferiority trials
Equivalence: same effect as reference
Non-inferiority: no worse than reference
Boxes in a meta-analysis
Show the effect estimate Size of the box correlate with the size of the effects from the single study shown
Diamonds in a meta-analysis
Represent pooled results from multiple studies. Wider the diamond, the less reliable the study results.
Horizontal lines in a meta-analysis
Illustrate the length of the confidence interval. The longer the line, the wider the interval and less reliable the study results
vertical solid line in a meta-analysis
Line of no effect
“0” for difference data
“1” for ratio data
Case-Control study: description
Compares patients w/ a disease (cases) to those without the disease (control). The outcome of the cases and controls are already known, but the researcher looks back in time to see if a relationship exists b/w outcome and risk factors
Case-Control study: benefits
Data easy to collect
Good for looking at outcomes when intervention is unethical
Cheap
Case-Control study: limitations
Cause & effect cannot be reliably be determined
Cohort study: description
Compares outcomes of a group of patients exposed and not exposed to a treatment. Researcher follows both prospectively or retrospectively to see if they develop the outcome.
Cohort study: benefits
Looking at outcomes when the intervention is unethical
Cohort study: limitations
influenced by confounders
more time-consuming and expensive than a retrospective study
Cross-Sectional survey
Estimates the relationship b/w variables & outcomes (prevalence) at one particular time (cross-section) in a defined population
Case report and case series
Case report - unique condition/ADR that appears in a single patient
Case series - same but in a few patients
Case reports/series: benefits
Can identify new diseases, drug side effects, or potential uses
Generates hypothesis that can be treated with other study designs
Case reports/series: limitations
Conclusions cannot be drawn from a few cases
Pharmacoeconomic research
Identifies, measures and compares the costs (direct, indirect & intangible) and the consequences (clinical, economic and humanistic) of pharmaceutical products and services
Cost-effectiveness analysis
Used to compare the clinical effects of 2+ interventions to the respective costs
Cost-minimization analysis (CMA)
Used when 2+ interventions have demonstrated equivalence in outcomes and costs of each are compared
Cost-utility analysis
Includes a quality-of-life component of morbidity assessments, using common health indices such as QALYs and DALYs
Cost-benefit analysis (CBA)
Comparing benefits & costs of an intervention in terms of monetary units (dollars)
ECHO model
Economic, clinical and humanistic outcomes
what is the purpose of pharmacoeconomic studies?
Guide optimal healthcare resource allocation
Incremental cost-effectiveness ratios
Represent the change in costs & outcomes when 2 treatment alternatives are compared
ICR formula
(C2-C1)/(E2-E1)
