EBD Flashcards
Study type
What it is
Pros
Cons
- Systematic review/Therapeutic guidelines
- systematic review of RCTs
- Highest level of evidence
- Relies on good selection of existing literature
- RCT
- Researcher randomly manipulates level of X
→ evaluate effectiveness of an intervention - Good for confirming casuality with large sample size, good randomisation, good compliance, good follow-up
- Highly cost and resource intensive
May not be ethical - Cohort studies
- Observe X, then measure Y moving forward - prospective
→ evaluate cause, risk - Confounding factors
- Case control
- Based on presence/absence of Y, assess X - retrospective
- Faster and less costly to conduct
- Can track multiple exposures at the same time
- Good for studying disease over long time
- Recall bias, Confounding
- Cross-sectional/
Ecological - Observe levels of X, Y at the same time
- Can do it with existing pt records
- No time component. Confounding
*Animal/lab studies
- Results may not apply 1:1 to humans
* Expert opinion
- Subjective
- Not always evidence-based
Measuring risk
Categorical
Relative
Risk ratio/Relative risk - incidence exposed/incidence unexposed
Measures strength of association
3.3 RR = 3.3x greater risk.
Odds ratio - odds of a case being exposed/odds a control was exposed
Probability of disease occuring under exposure
Prevalence ratio - prevalence of cases in exposed/unxposed
Relative prevalence
Absolute
Risk difference - incidence of exposed subtract incidence of unexposed
For each unexposed case, there is 4 additional cases among exposed.
Measures public health outcome - how many cases could be prevented by eliminating exposure
Numerical
Absolute
Mean difference
Beta coefficient - standard deviation
Types of errors / biases
1) Random error
Occurs by chance due to fluctuations in a population
Increase sample size
2) Systematic error
Confounding bias
Lack of good randomisation, different baseline characterises
Stratification by confounding factor, randomisation decreases confounding - blinding
Blinding in RCTs - helps reduce contamination problems between groups—make sure groups being compared are similar wrt confounding factors
Exchangeability = participants unexposed have same potential outcome had they been exposed
Selection bias
For studies on prevalence: unrepresentative sample → have more representative sample
For studies on casuality: differential loss of follow up → registration protocol/policy to ensure follow-up
Information bias
Differential misclassification of participants (from recall bias)
Have better questionnaires with specific qns, Consider self-administered questionnaires vs interviewers, try to assess past exposure through biomarkers
True / False Positives/Negatives
Sensitivity, specificity, positive predictive value, negative predictive value
True/False = was it correct.
Positive/Negative = Present/Absent
Sensitivity - how good at identifying disease-present cases TP/(TP+FN)
Specificity - how good at identifying disease-absent cases
TN/(TN+FP)
Positive Predictive value/Precision - of cases deemed positive, how often was it actually present? TP/(TP+FP)
Negative Predictive value - of cases deemed negative, how often it was actually absent TN/(TN+FN)
Forest plot
A type of meta analysis.
The graph orders many different studies by confidence interval.The higher the CI, the higher weighting the study impacts on the overall summary average.
Social determinants of health
Implicit bias
Non-medical factors that influence a person’s health outcomes - conditions in which people are born, live, grow and age - inequities relate to unfair and unavoidable differences in health outcome
Implicit bias
Making associations to make sense of information quickly
Upstream: affects perceptions, policymaking and public health institutions → reinforces systemic prejudice
Midstream: affects clinical decisions - dx assuming certain ethnicities have better pain tolerance
Blaming pt attitude for severe disease and poor prognosis
Lifecourse epidemiology
Upstream - Macro
Eg - fluoridated water
Public dental healthcare
Dental subsidies for the disadvantaged
Policies
Health, economic, welfare, tax, housing
Determinants
Income, education, employment, housing
- Role of policy is to change distribution of social determinants for better health outcomes, by modifying the pathways through which social determinants operate to cause inequalities.
Midstream
Availability of healthy diet options in supermarkets
Affordability
Keeping people informed about health choices - labelling, education, advertising
Psychosocial
Control, social support, stress
Behaviour
Diet/nutrition, smoking, alcohol, physical activity
Healthcare
Access, utilisation
Location/distribution of services
Downstream
Genetics - periodontal phenotype
Composition of biofilm - aciduric/acidogenic bacteria proportion
Physiological systems
Biological reaction - like immune system strength
Health
Population of interest being assessed =
Population of interest being assessed = who are we surveying
Providing Evidence-Based Advice
Formulate question with PICO
Search for evidence with databases like Pubmed and Clinical Guidelines
Do not limit advice to just oral health
Connect to services offered in the community
Why Trial registration? =
Why Trial registration? = for transparency in establishing impt characteristics of study. Deviation from protocol must be reported—researchers are held accountable for reporting totality of results instead of cherrypicking results (publication bias)
Outcome =
Outcome = the outcome + How it was assessed (the measurement)
Biological plausibility =
Biological plausibility = [Exposure] is linked to series of biological processes that affect [x] and [y] which increase susceptibility of [Outcome]
Blocked randomisation =
Blocked randomisation = ensures that equal numbers of participants (of each confounding characteristic) are allocated to intervention and control groups
How to assess if randomisation was successful?:
How to assess if randomisation was successful?: Table with baseline info by intervention and control grps, w as much info as possible on participants- look for balance
Consequences of not blinding:
Consequences of not blinding: different management/data collection, pt feels & reports differently