EBM revision Flashcards
What do you need to consider when you want to treat a patient with a drug?
- benefits vs harms
- costs
- patient preference
- ethics
- can you use the drug in your patient population?
What is the need for randomisation?
Both the known and unknown characteristics of the study population that might affect the outcome of trial are distributed by chance. This reduces allocation bias as well.
What do these levels of measurement mean?
- Nominal
- Ordinal
- Interval
Nominal- categorical
Ordinal- rank-ordered
Interval- presented as ranges
What are the main biases associated with RCT?
- Allocation bias
- Performance bias
- Detection bias
- Performance bias
- Reporting
What is allocation bias?
Systematic difference that exists in the baseline characteristics between the groups in the trial.
How to reduce allocation bias?
Randomisation using a random sequence generator, allocator concealment
What is performance bias?
Systematic difference in the care given and external factors that may affect outcome that are not the intervention of interest.
How to reduce performance bias?
Blinding the participants and personnel
What is detection bias?
Systematic difference in how outcomes are measured/
How to reduce detection bias?
Blind the outcome assessor
What is attrition bias?
Systematic differences in the groups that have withdrew from the study compared to those who stayed on
What should you be looking for in terms of attrition bias?
incomplete outcome data
What is reporting bias?
Systematic differences between reported and unreported findings
What should you be looking for in terms of reporting bias?
Proof of something that was measured but not reported
What factors may lead to allocation bias if not blinded?
Not wanting to disadvantage patients who are old/young, poor health, vulnerable patients
What are the ethical issues that could be raised from an RCT?
Could we possibly giving patients a drug that could cause a lot of s/e compared to standard treatment and one that may not be as effective?
Are we disadvantaging the control group if this drug is actually more effective
How do we reduce cross-over?
Cluster randomisation
2 Methods of analysis of outcome:
- intention to treat: based on the initial allocation of treatment.
- according to treatment received: treatment received coherent with the protocol of the trial which they were originally allocated to.
Where can you look for study population?
List/register- GP lists, cancer and professional registers, hospital and clinic registers
What are the 3 modes of randomisation?
- Simple
- Cluster
- Stratification
How is simple randomisation conducted?
using random number generator
How is cluster randomisation conducted?
perform in schools, hospitals, gp practices
What is the downside of cluster randomisation?
Special analytical tests need to be conducted to account for the similarities within the groups.
How is stratification randomisation conducted?
Greater proportion of minority groups are recruited to increase power and sample size
What are the non-random methods?
- Systematic
- Convenience
- Snowball
- Street survey
What is the systematic method?
every nth selected
What is the convenience method? and the problem with this?
volunteers recruited –>volunteer bias
What is snowball?
recommendation of a friend from a pre-set requisite
What is street survey?
Open polls and market research
What is bias?
Systematic error in the study design, conduct or analysis that results in deviation from the truth
What is a random error?
All measurements deviate from the centre of target in different directions due to unknown and unpredictable changes in the experiment
What is sampling bias?
Systematic difference in the characteristics of those selected into the study compared to those who weren’t.
What is spectrum bias? Which study is it seen in?
Specific groups inappropriately excluded. DTA
What is interviewer bias?
Error due to interviewer’s subconscious/conscious gathering of data that differs systematically between cases and controls
What is recall bias? When can it be avoided?
incompleteness/inaccuracy recall of prior exposure of interest between cases and controls. Avoided in cohort study
What is recording bias?
Cases have a more detailed recording of data compared to controls.
What is verification bias? what study is it seen in?
Systematic error in DTA where not all participants received both the index and reference standard.
What is the main thing to note in qualitative research?
Reflexive stance of the researcher, their background, beliefs, personality influences their relationship and interaction with participants and may influence the outcome of the study.
What is a confounding factor?
An additional factor that can independently have an effect on the exposure of interest and with the outcome of the study.
Which study design theoretically does the confounding factor not influence very much?
RCT because known and unknown confounders are evenly distributed through randomisation.
How can confounders be adjusted in studies other than RCT?
Matching studies in case control and cohort trial.
What are the 9 criteria for causality?
- Temporality: time sequence exposure before effect
- Strength: stronger the association, more you believe it
- Biological gradient: dose-response relationship
- Consistency: many different studies coming to similar conclusions
- Plausibility: biological mechanism either known/postulated that may support association
(TCBPS) - Coherence: Should not contradict current knowledge
- Experiment/Reversibility: when cause removed, disease does not occur
- Specificity: cause associated with one disease only (rare)
- Analogy: clear cut analogy present
What is incidence?
Total number of new cases during a specified period in a specified population.
What is prevalence?
Total number of individuals with the disease at a particular time
How to measure incidence rate?
New cases/person-time accumulated
Measure incidence risk:
new cases/number at risk
Measure prevalence:
Incidence x duration of disease
Measure point prevalence:
number at some time point/total population at risk at that point
Measure period prevalence:
number over a period/total population seen over that period
What are the 2 ways effect can be expressed?
Risk ratio or risk difference
What are case reports/series?
Detailed report of either individual or series of patients
What are the advantages of it?
- Good for unusual disease
- Forming hypothesis
What are cross-sectional studies?
Used to determine point prevalence of a particular study.
Find sample population from target pop–> collect data
What are the advantages of cross-sectional studies?
Cheap, quick
What are the disadvantages of cross-sectional studies?
- Cannot determine causality ( time-sequence, biological gradient, coherence, strength, plausibility)
- Unable to interpret survival because they are most likely to be less aggressive diseases
What is an ecological study?
Measure of rate of death/disease in populations at risk.
What are the advantages of an ecological study?
- Quick
- Cheap
- Good to generate hypothesis
What are the disadvantages of an ecological study?
- Confounding factors
- Sampling bias
- Information bias
- Unclear causality relationship
- Not on individual level
How do you analyse ecological studies?
Scatter plot, Pearson’s correlation
What do the following r values for Pearson's correlation mean? r=1 r=0.8 r=0.5 r=-1
correlation present
strong correlation
moderate correlation
negative correlation
What is a cohort study?
Groups of individual defined on the basis of their exposure to risk factor then followed over time to observe if they develop the disease
Cohort studies can be conducted in 2 methods. What are they?
- Prospective: exposure measured at baseline and followed up
- Retrospective: trace back for exposure info
What are the advantages of prospective to retrospective studies?
exposure accurately measured, confounders can be assessed
What are the disadvantages of prospective to retrospective studies?
it is expensive, time consuming
What are the main advantages of cohort studies?
- Minimise recall bias
- Good causal effect
- Can monitor many diseases
- Can be used for rare exposures
What are the main disadvantages of cohort studies?
- Not useful for rare diseases
- Expensive and time consuming
- Attrition bias
What is a case control study design?
Aetiological factors (the cause of disease), incidence and prevalence cases
What are the advantages of case control?
- Study rare disease
- Explore multiple risk factors
- Good for diseases with long latency periods
What are the disadvantages of case control studies?
- Selection bias
- Causality cannot be formed
- Single disease
- Cannot estimate absolute risks
What types of analysis conducted in case control studies?
- matched: conditional logistic regression
- unmatched: logistic regression
What methods are used to adjust for confounders?
Cox regression, randomisation, stratified analysis, matching
What are the strengths of qualitative evidence?
- real world patient
- high level of detail and content
- capture experience and understanding
What are the weaknesses of qualitative evidence?
- generalisable with caution
- need to trust researcher
- cannot predict causation
What method of data collection used in qualitative research?
Iterative method where collection and analysation of data occurs over and over again until saturation is reached and emerging themes are identified
What tool do we use to appraise qualitative research?
CASP tool
What are the 2 types of errors that may occur in qualitative research?
I- stating something that is not there–> little evidence can be found on how that conclusion was achieved
II- Ignoring something that is there
What does the Diagnostic Test Accuracy study encompass?
Comparison of disease state estimated by the index test (new test of interest) compared to reference standard
What are the 2 important information one needs to know about a patient before recruiting them?
- Presentation of disease
- Prior tests
What are the 3 main types of bias present in DTA?
- Spectrum bias: inappropriately excluding patients that are difficult to diagnose
- Verification bias: patients who received the index test may not always get the reference standard test
- Review bias: Interpretation of index test results not independent of standard ref test results
What is sensitivity good for?
Ruling out a disease
What is specificity good for?
Ruling in a disease
What could the possible effects of test to the population be?
- Safety
- Anxiety
- Accuracy
- Replicability
What is the study design for effectiveness?
RCT
What is the study design for aetiology?
case control, cohort, ecological
What is the study design for frequency?
Cross-sectional, cohort
What is the study design for experience?
qualitative study
What is the study design for diagnosis?
Special cross sectional
What are the 5 steps to evidence in clinical practice?
- Study question- PICO
- Find robust evidence in terms of guidelines, clinical evidence, systematic review, primary studies and if not last resort is information from specialists
- Appraise the evidence using a tool like CASP or Agree II framework
- Is this applicable to your own clinical practice?
- Reflect by taking part in audits to assess performance of clinical team against evidenced based measurable data.
What is purposive sampling? What study is this usually seen in?
Sample chosen based on the characteristics of the population and the objective of the study. Qualitative study.
How do you define sensitivity?
How accurate a test is in individuals with the disease and the magnitude of false positives.
How do you define specificity?
How accurate a test is in individuals without the disease and magnitude of false negatives.
Where would you search for evidence for quantitative study design?
NICE guidelines
Medline- systematic reviews
Primary literature- EMBASE, medline
What is cochrane library and embase not good for?
Observational and qualitative studies
Where would you find qualitative evidence?
MEDLINE, EMBASE, Sociological abstracts, Social service abstracts, PscyInfo
Qualitative health research, Social science and medicine
