← All decks
Anaesthesiology
Experimental Design and Statistics
Built from Barash Clinical Anesthesia, 9th Edition
What’s inside
15 sections · 180 slides
Overview
- Scope of this topic
- Key points of the topic
- Why a clinician needs statistics
Foundations and data types
- Sources of variation in a data set
- Explanatory and response variables
- Classification of data types
- Categorical data subtypes
- Consequences of assigning the wrong data type
- Data types with definitions and examples
Descriptive statistics
- Purpose of descriptive statistics
- Probability density functions
- Importance of the normal distribution
- Measures of central location
- Sample mean as estimator of the population mean
- Variance and standard deviation
- Proportion of data within one, two, and three standard deviations
- Interpreting the standard deviation bounds
Types of research design
- Elements that define a research design
- Longitudinal versus cross-sectional studies
- Prospective (cohort) and retrospective (case-control) studies
- Limitations of case-control studies
- Deliberate intervention versus observational studies
- Concurrent and historical controls
- Newer clinical trial designs
- Classification of clinical research reports
- Reading the classification of research reports
Management of bias
- Definition and importance of bias
- Populations, samples, and strata
- Methods of sampling
- Limitations of convenience sampling
- Fixed dosing versus individualised dosing
- Sources of comparative data for a control group
- Internal mammary artery ligation for angina
- Empirical evidence on historical controls
- Random allocation of treatment groups
- Concealment of random allocation
- Randomised trials versus observational studies
- Blinding of patients and investigators
- Reporting of blinding in research papers
- Design features that minimise bias
Hypothesis testing
- From biologic hypothesis to statistical hypothesis
- Null and alternative hypotheses
- Algebraic statement of statistical hypotheses
- Logic of proof in hypothesis testing
- Level of significance
- Assumptions of a statistical model
- Type I error
- Type II error and statistical power
- Effect of true difference size on type II error
- Sample size formula for comparing two means
- Options for reducing type II error
- Historical neglect of sample size planning
- Errors in hypothesis testing: the two-way truth table
- Reading the two-way truth table
- Steps for accepting or rejecting the null hypothesis
- Probability distributions and degrees of freedom
The P value and the Bayesian alternative
- Common misinterpretation of the P value
- Hypothesis plausibility and the meaning of a P value
- Prior plausibility and post-experiment odds
- Frequentist and Bayesian inference
- Rationale for Bayesian methods
- Barriers to adoption of Bayesian methods
Statistical tests for interval data
- Parametric statistics for interval data
- Structure of a parametric test statistic
- Confidence intervals
- Standard error of the mean
- Central limit theorem
- Standard deviation versus standard error
- Ambiguity of 'mean plus or minus number'
- The Student's t-test
- Paired and unpaired t-tests
- Critical value of the t statistic
- Reporting effect size with a comparison of two means
- Repeated measurement experiments
- Inflation of type I error with multiple t-tests
- Analysis of variance and the F ratio
- Structure of the F statistic
- Multiple comparisons after analysis of variance
- Robustness of parametric statistics
- Nonparametric tests
- Nonparametric equivalents of parametric tests
- Limitations of nonparametric tests
- Choice of statistical test by variable type and number of samples
- Using the test selection table
Binary variables
- Proportions as a measure of central location
- Standard error of a proportion
- The rule of three for zero numerators
- Bayesian credible interval for a zero numerator
- Effect of prior information on the estimated event rate
- Interpreting the prior-informed estimate
- Hypothesis testing in a two-by-two table
- Metrics of binary treatment effect
- Choosing between chi-square and Fisher's exact test
- Measures of binary treatment effect
- Effect sizes in the transient neurologic symptoms trial
- Interpreting number needed to harm
Regression and correlation
- The linear regression equation
- Least squares estimation and testing the slope
- The correlation coefficient
- Worked example of the coefficient of determination
- Four scatter plots from the Anscombe data sets
- What the Anscombe data sets show
- Implications of the Anscombe data sets
Epidemiology and non-randomised studies
- Scope of epidemiology in anesthesia
- Association and causation
- Refutation of observational findings by randomised trials
- Multivariable linear regression
- The multivariable linear model equation
- Logistic regression for binary outcomes
- Confounding and its control
- Stepwise multivariable logistic regression
- Performance of published prognostic models
Propensity score matching
- Purpose of propensity score matching
- Pulmonary artery catheterisation in critical illness
- Baseline imbalance in the catheterisation cohort
- Result after propensity matching in the catheter study
- Limits of simple matching
- Definition and estimation of the propensity score
- Timing rule for covariate inclusion
- How matched pairs are formed
- Practical points about matching
- Covariate balance before and after propensity score matching
- Reading the covariate balance plot
- Balance achieved in the epidural anesthesia cohort
- Other uses of propensity scores
- Reporting requirements for a propensity analysis
- Why propensity matching cannot replace randomisation
- The TITRe2 result and its lesson
Systematic reviews and meta-analyses
- Definition and scope of systematic reviews
- Steps of a systematic review protocol
- Types of bias within included trials
- Focus of bias detection in included trials
- Effect sizes used in meta-analysis
- Forest plot of transient neurologic symptoms after spinal lidocaine
- Anatomy of a Forest plot
- Interpretation of the transient neurologic symptoms Forest plot
- Power of pooled summary statistics
- Sources and updating of systematic reviews
- A systematic review is provisional
- Ongoing developments in meta-analysis methods
Big Data in anesthesia
- Defining attributes of Big Data
- Historical anesthesia studies of Big Data scale
- Search engine data and influenza surveillance
- Perioperative data repositories
- Why anesthesia data is not yet Big Data
Interpreting and using results
- Statistical significance and clinical relevance
- Effect of sample size on interpretation
- Dose-response and generalisability checks
- Hierarchy of evidence for a therapeutic claim
- Resources for study design and statistics
- Reporting guideline organisations
- Statistical training in academic anesthesia
- Summary of the topic
- References
- References (continued 1)
- References (continued 2)
- References (continued 3)
- References (continued 4)
- References (continued 5)
- References (continued 6)
- References (continued 7)
- References (continued 8)
- References (continued 9)
- References (continued 10)
- References (continued 11)
- Barash, Cullen, and Stoelting's Clinical Anesthesia, 9th Edition