Stats/Primary Literature Flashcards
How can you overcome random error?
Increase sample size
What type of trial is impacted the most by selection bias?
Case-control study
How can selection bias be reduced in observational studies?
Propensity score matching
What are disadvantages of case report/case series?
-Difficult to generalize results
-Report may not truly be characteristic of the disease
-Does not determine causality or association between exposure + outcome
Case Report/Case Series
Describes experience of a single patient or small group of patients with an unusual or rare disease or symptoms
Cross-Sectional Study/Survey
Identify the prevalence or characteristics of a condition in a group of individuals and evaluate associations
Snapshot of a population at one point in time
What are disadvantages of cross-sectional study/survey?
-Low response rates
-Responding population may be biased
-Does not usually target acute, rapidly fatal, or rare diseases
-Cannot establish temporal or causal relationships
Case-Control Study
Provides an efficient means to determine the association (or exposure) between the risk factor and the outcome of interest
Retrospective
Cannot determine the actual risk of the outcome (RR) but can determine an estimate risk (OR)
What are disadvantages of case-control study?
-Can have poor data quality
-Matching is often flawed
-Indirect measurement of risk
-Susceptible to bias or confounding
-Temporal relationship is not always known
Cohort Study
Healthy subjects are followed over time to try to correlate exposure (or risk factors) and disease occurrence
Identifies the relationship between exposure and outcome
Strongest design to establish cause-and-effect relationship for risk
What are disadvantages of cohort study?
-Only risks evaluated at baseline can be used
-Cannot investigate a rare disease
-Take a long time to complete
What does prevalence measure?
Occurrence of new and existing cases of a disease (or event) at a specified point or period in time
What does incidence measure?
Occurrence of new cases of a disease (or event) during a time interval
When is an OR or RR considered not statistically significant?
When 95% CI includes 1.0
Randomized Controlled Trial
Subjects are randomly assigned to intervention or control group
Gold standard for study interventions and determining causal relationships
When would a crossover RCT be appropriate?
When chronic or highly variable diseases (ex: migraines) are being evaluated
What are disadvantages of RCT?
-Time-consuming and expensive
-Adherence to study protocol can be poor
-Incomplete follow-up from subjects
-Guidelines may have changed since the initiation of trial
-Subjects may differ from the population of interest
Superiority Trial
Designed to detect a difference between experimental treatments
Equivalence Trial
Designed to confirm the absence of a meaningful difference between treatments; neither better nor worse (both directions)
Non-Inferiority Trial
Designed to investigate weather a treatment is not clinically worse (not less effective) than an existing treatment
Usually requires larger sample sizes
Intention-to-Treat
Includes everyone in the trial who was initially randomized regardless of treatment adherence or dropping out
Similar to routine clinical practice
Per-Protocol
Only evaluates those who completed the trial and adhered to the protocol
Cost-Minimization Analysis (CMA)
-Differences in cost among comparable therapies are evaluated
-Only useful to compare therapies that have similar outcomes (no outcomes measured)
-Primary focus is $$$
Cost-Effectiveness Analysis (CEA)
-Outcome: clinical units or cost per unit health outcome (ex: years of life saves, number of symptom-free days, blood glucose)
-Useful to measure the cost impact when health outcomes are improved
Cost-Utility Analysis (CUA)
-Assigns utility weights to outcomes so that the impact can be measured in relation to the cost ex: quality-adjusted life-years)
-Compares outcomes related to mortality when mortality may not be the most important outcome
Cost-Benefit Analysis (CBA)
-Monetary value is placed on both therapy costs and beneficial health outcomes
-Allow analysis of both the cost of treatment and the costs saves with beneficial outcome
What should not be reported with ordinal data?
Mean + SD
What type of data would it be appropriate to use mean?
Continuous and normally distributed (parametric)
What type of data would it be appropriate to use SD?
Continuous and normally distributed (parametric)
What type of data would it be appropriate to use median?
Ordinal and continuous (including non-parametric)
What type of data would it be appropriate to use mode?
Nominal, ordinal, and continuous
What is the relationship between mean and median in normally distributed data?
They will be roughly the same
What does a 95% CI mean?
There is a 95% change that the range includes the true population value
When is a CI for continuous variables not statistically significant?
When it contains 0
What is the null hypothesis?
No difference between groups being compared
What is the alternative hypothesis?
There is a difference between groups being compared
What is a Type I Error?
Concluding there is a difference when there is truly no difference (falsely rejecting the null hypothesis)
What is a Type II Error?
Concluding there is no difference when one truly does exist (falsely accepting the null hypothesis)
Test used for: non-parametric, nominal data w/ 2 independent groups
Chi-squared OR Fischer exact test
Test used for: non-parametric, nominal data w/ 3+ independent groups
Chi-squared OR Fischer exact test
Test used for: non-parametric, nominal data w/ 2 dependent groups
McNemar test
Test used for: non-parametric, nominal data w/ 3+ dependent groups
Cochran Q test OR Mantel-Haenszel test
Test used for: non-parametric, ordinal data w/ 2 independent groups
Mann-Whitney U test OR Wilcoxon rank sum test
Test used for: non-parametric, ordinal data w/ 3+ independent groups
Kruskal-Wallis test
Test used for: non-parametric, ordinal data w/ 2 dependent groups
Wilcoxan signed-rank test
Test used for: non-parametric, ordinal data w/ 3+ dependent groups
Friedman ANOVA
Test used for: parametric, continuous data w/ 2 independent groups
Student t-test
Test used for: parametric, continuous data w/ 3+ independent groups
ANOVA
Test used for: parametric, continuous data w/ 2 dependent groups
Paired t-test
Test used for: parametric, continuous data w/ 3+ dependent groups
ANOVA
