MRCPsych Paper B - Critical review Flashcards
Type of randomization - using odd/even numbers of DOB
Quasi-randomization
*because it is not reproducible
Type of randomization - coin toss
not randomization
Type of randomization - computer generated random list
Simple randomization
Type of randomization - blocks of 6
Block randomization
Type of randomization - randomize clusters or catchment zones, not individuals
Cluster randomization
Type of randomization - next allocation is based on the characteristics of what is already present
Minimisation
How to control confounders?
Restriction
Matching
Randomisation
Stratification
Multivariate analysis
Type of bias - difference between hospital/admitted population and community population
Berkson bias
Type of bias - using incidence rather than prevalence in determining causation computations
Neyman bias
Type of bias - exclusion of comorbidities in a non-representative sample
Diagnostic purity bias
How to control measurement bias?
Blinding
How to control analysis bias?
Intention to treat analysis
Statistical study to quantify heterogeneity
Chi-square - Cochran’s Q(uantify)
*Q(uantify)
Statistical study to differentiate between heterogeneity and chance
I2 statistic
Economic study - used to decide how to spend the allocated money effectively
Cost benefit
Economic study - used to compare just the costs
Cost minimisation
Economic study - used when both interventions result in an outcome that can be measure using the same clinical scale
Cost effectiveness
Economic study - used when considering the holistic effect of what is being measured (DALY, QOL, life utility units)
Cost utility
Economic study - used when different interventions AND different outcomes are being measured using natural units
Cost consequences
Sensitivity analysis - effect of resampling the data set
Bootstrapping
Sensitivity analysis - the uncertainty in each unit is assumed to possess a probability distribution while estimating the predicted effects
Monte-Carlo simulation
What is PDSA?
Plan Do Study Act
What is FOCUS?
Finding - process to improve
Organizing - a team
Clarifying - current knowledge
Understanding - causes of variation
Selecting - improvement process
What is FADE?
Focus - define the problem
Analyze - data collection for current knowledge and root cause of the problem
Develop - action plan
Execute - implement
Evaluate - measuring and monitoring system
Arrangement of central tendency in a negative/left skew?
Mean Median Mode
*alphabetical
Arrangement of central tendency in a positive/right skew?
Mode Median Mean
Std. error of mean (SE) is always __ to Std. deviation
smaller
*SE < SD
The rectangle of a box and whisker plot pertains to what?
50th to 75th percentile
The line cutting through the rectangle of a box and whisker plot pertains to what?
median value
Percentage of data that lie within 1 SD
68%
*Shaq Nash vs 2 = 34 + 13 + 2
Percentage of data that lie within 2 SD
95%
*Shaq Nash vs 2 = 34 + 13 + 2
Percentage of data that lie within 3 SD
99%
*Shaq Nash vs 2 = 34 + 13 + 2
What is a Type 1 error
False positive, α
What is a Type 2 error
False negative, β
How to increase the power of a study?
increase significance level
increase sample size
increase effect size
decrease variability
use a one sided test
Value of no difference of OR
1
Value of no difference of RR
1
Value of no difference of ARR
0
Value of no difference of RRR
0
Value of no difference of NNT
∞
How can you decrease the width of the CI?
Use a smaller degree of CI (95% to 90%)
Reduce SD by stratifying
Increase sample size = N
Effect size scores and interpretation
0.5 ± 0.3 = medium (vs small, large)
*effect size score is that amount of SD above the average person in the control group
How to correct for multiple testings?
- Bonferroni correction - adjust the p-value/ significance levels
- Family-wise error - probability one of the data sets is a Type 1/ False Positive error
- False discovery rate - control the PROPORTION of Type 1/ False positive error
Average duration of test-retest correlation used in psychiatry
2-14 days
Arbitrary cutoff for Cronbach’s alpha
0.70
Type of validity - agreement between instruments that measure the same construct
Convergent validity
Type of validity - disagreement between two scales measuring different constructs
Divergent validity
Type of validity - sensitivity to change
Experimental validity
