Lecture 5: Epidemiology of affective disorders and risk factors Flashcards
Which are the classes of epidemiology?
- Descriptive
- Analytic
- (teacher adds: treatment)
What are the two types of interviews that can be conducted in epidemiological studies?
- Structured
- Semi-structured
What are the two types of studies performed in descriptive epidemiology?
- Cross-sectional (to measure prevalence)
- Inception cohorts (to measure prognosis)
What are the types of studies performed in analytic epidemiology?
- Cohort (to determine outcome. subtypes: prospective, retrospective, panel)
- Case control (to determine outcome in terms of odd ratio -mostly used in rare disorders)
- (teacher adds: ecological studies -compare clusters)
Outline the types of studies that belong to descriptive & analytic epidemiology?
Descriptive (how the world is - generates hypotheses on risk factors associated with diseases):
- Cross-sectional studies
- Inception cohorts
Analytic epidemiology (test the hypothesis at individual level - i.e. risk factor exposure and disease development):
- Case control studies
- Cohort studies
- (Ecological Studies)
https://pediaa.com/what-is-the-difference-between-descriptive-and-analytic-epidemiology/#:~:text=The%20main%20difference%20between%20descriptive,as%20well%20as%2C%20analyzing%20the
Outline the different types of interview and some pro’s and con’s associated with each?
Structured:
- Can be conducted by non-clinicians
- Increased reliability at expense of validity
- Low observer bias
- Can lead to false positives
Semi-structured:
- Requires clinicians/trained individuals
- Less reliable than structured interviews
- More validity
- Observer bias can occur with semi-structured interviews
Outline how cross-sectional epidemiological studies function?
Measure prevalence in a defined population using a sample from the whole population.
All of the chosen sample is interviewed or a two stage sampling procedure is done
What are the areas to assess the quality of a cross-sectional epidemiological study?
Cleary defined frame
Random sampling - i.e. everyone in the population has a chance to be included
High participation rates
A precise estimates of disease prevalence
If a two stage sampling procedure has been done - is there a screen for negatives?
Give some examples of cross-sectional epidemiological studies?
Epidemiological Catchment Area Study, National Comorbidity Study, UK Adult Psychiatric Morbidity Study
Describe how the UK Adult Psychiatric Morbidity study is conducted?
UK Adult National Psychiatric Morbidity Survey is conducted first by selecting postcodes - identify adults at these addresses - selecting them for participation - asking them to complete a structured interview CIS-R (assess prevalence of individual symptoms)
What have been recent findings from the UK Adult National Psychiatric Morbidity Study?
Falling participation rates
The most common symptoms reported on CIS-R are sleep disturbance, irritability, fatigue, worry
Women have a higher prevalence of symptoms & CMD than men
The rate of CMD is rising
CMD are more prevalent in younger individuals
Those who are unemployed/economically inactive, in receipt of benefits or smoke more (> 15. a day) are more likely to have a common mental health disorder
Most prevalent CMDs are mixed/anxiety depression > GAD > Depression > Phobias > OCD > Panic Disorder
(Higher in women, younger individuals, unemployed/economically inactive and in receipt of benefits, or smokers)
What are inception cohort studies and are they conducted?
Recruit participants at inception with aim of recruiting at same time (i.e. when disorder first recognised)
Follow up participants to identify prognosis
Should assess for a range of outcomes (with result analysis based upon baseline differences)
Critical analysis - look out for:
- High quality inception studies have good follow up rates
- Observer bias
- Representative cohorts (i.e. not recruited from a. tertiary centre)
Give some examples of an inception cohort study
Collaborative depression study
- Patients with MDD treated in an academic centre
- Over 12 years - 59% of patient/weeks had minor, dysthymic or sub-threshold symptoms
- Mediation episode duration 3-6 months
NESDA
How do cohort studies function?
Assess risk of developing a disease based on exposure to a risk factor
Individuals with the disease are excluded & participants are defined by exposure status
Follow up to determine outcome - displayed as relative risk
Outline the types of analytic epidemiological cohort studies?
Prospective - classic design e.g. occupational cohort
Retrospective - opportunistic
Panel - series of cross-sectional based on same subjects
What areas should be assessed in a cohort study?
Follow up rate - should be high
Any healthy worker effect - a form of selection bias that underestimates the prevalence of disease in a set of exposed individuals
Any observer bias
Type I error - false positives? - if accumulating multiple exposures?
Type II error - lack of true positive due to underpowered sample, especially the cases for rarer outcomes that may need a large sample size
Confounding variables?
What is the healthy worker effect?
Form of selection bias that may underestimate prevalence of exposed individuals (exposed workers - typically seen in occupational cohort studies)
Underestimation occurs as the exposed works are compared to the general population
General population includes workers and non-workers.
As non-workers may have high rates of disease outcome (disease may affect them working) - prevalence of disease is higher in population > exposed workers
However better comparison is between exposed workers and workers
How are case control studies conducted?
Recruit participants based on disease status then look for exposure
Works well if rare diseases and can assess for a number of exposures
What areas are particularly relevant for case control studies:
- Selection bias (recruit from 2nd care)
- Recall bias
- Observer bias
- Direction of causation - as recruiting on disease status
- Type I error if multiple exposures
- Types II error if rare exposures
- Confounding
What did Cheasty et al find with a case control format?
Increased odd’s ratio of having depression if exposure to abuse
How are ecological studies conducted?
Test data at a higher order cluster -i.e. postcode, region or country
Plot two variables against each other - suicide rates and unemployment
Allows a fast test of a hypothesis
However:
- Confounding bias - many covariates not controlled for
- Ecological fallacy - i.e in example above one individual who committed suicide may not be unemployed - is this a fair way to test
Define the following concepts
a) Confounding
b) Selection bias
c) Observer bias
d) Recall bias
e) Type I error
d) Type II error
a) Where another variable can cause the outcome and the exposure
b) False outcomes from recruiting a non-representative sample
c) The measurer is influenced in their judgement by prior knowledge of participant status or by their knowledge of the study
d) Participant answers are influenced by their exposure - i.e. depressed mood and recall of +ve/-ve events
e) Type I error - false +ve
d) Type II error -not picking up a true association due to lacking statistical power
Outline some structured/semi-structured interviews
Structured - CIS-R // CIDI
Semi-structured interviews - SCAN, SCID