Epidemiology Exam Notes PDF

Epidemiology ============ Measures of association ----------------------- Ratio (relative) measures -- measure of the comparative risk of developing a disease according to the exposure to a suspected risk factor Relative Risk (risk ratio and rate ratio) -- incidence data from cohort or interventi...

Epidemiology ============ Measures of association ----------------------- Ratio (relative) measures -- measure of the comparative risk of developing a disease according to the exposure to a suspected risk factor Relative Risk (risk ratio and rate ratio) -- incidence data from cohort or interventional studies Odds ratio (an estimate of the relative risk) -- prevalence data from case-control and (often) cross-sectional studies What does a ratio tell us? Ratio is a relative measure that tells us how many times more likely it is that someone who is 'exposed' to something will experience a particular health outcome than someone who is 'unexposed' Do not tell us anything about the actual amount of disease occurring in either group -- 5 in 7,000,000,000 is 5 times the risk of 1 in 7,000,000,000 but is still not a lot Difference (absolute) measures -- measure of the impact of a disease on a population by examining the difference in risks (risk difference and rate difference) - Attributable risk - Population attributable risk Only measure risk differences when we measure risk (incidence) -- which studies can do this? **Relative risk** RR = incidence of the disease in the exposed group/incidence of the disease in the unexposed group - Calculated two types of incidences depending on the data available; CI & IR (never use both) - Depending on which is calculated, can calculate risk ratio or rate ratio -- both = 'relative risk'  **2X2 tables** Data is categorised into 4 groups - Exposed and diseased (a) - Exposed and not diseased (b) - Not exposed and diseased (c) - Not exposed and not diseased (d) Example with risk ratio to calculate RR A study was conducted to assess the association between alcohol consumption and liver cancer. 12 600 people who regularly consumed alcohol and 13 800 people who did not drink were followed for 15 years. At the end of the 15 years, there were 241 cases of cancer among the drinks (exposed) and 108 cases among non-exposed.  Calculating RR = (241/12,600) / (108/23,800) RR= 0.01912698/ 0.00782609 =2.44 (times by 100 etc. to get a readable number) (interpretation) = a person who is exposed to alcohol consumption for a period of 15 years has a 2.44 times the risk of developing liver cancer compared to an unexposed person. **Interpretation of the relative risk** Those exposed to (exposure) have (RR number) times the risk of developing (the outcome), compared to the unexposed If the value of RR is = to 1 indicates that the disease and the risk factor are unrelated If it is larger the RR, it is a stronger association between the disease and risk factor Values of RR less than 1 indicate a 'negative association' between the risk factor and disease (that is, the factor may have a protective effect) **Rate Ratio** Two formulas for relative risk Use the other one if we have an incidence rate rather than a cumulative incidence Only ever have: - Ire = incidence rate in the exposed - Iro = incidence rate in the unexposed If working with incidence RATES will NOT use a 2x2 tables 2x2 tables only work for COUNTS not rates Example: A diabetes follow-up study included 218 diabetic women and 3,823 nondiabetic women. By the end of the study, 72 of the diabetic women and 511 of the nondiabetic women had died. The diabetic women were observed for a total of 1,862 person years; the nondiabetic women were observed for a total of 36,653 person years. Calculate the relative risk of death for diabetic women. Ire (for women exposed to diabetes) = 38.6 deaths per 1,000 person-years Iro (for women unexposed to diabetes) =13.9 deaths per 1,000 person-years  From this, we have the incidence rate in the exposed (Ire), and we have the incidence rate in the unexposed (Iro) **Odds Ratio** To calculate RR we need incidence (CI or IR) In a cohort or experimental study, we follow exposed and unexposed cohorts to calculate incidence as new cases occur In a case-control study or cross-sectional study, we start with disease and non-diseased and often calculate the prevalence of exposure in each group No incidence= no risk We must estimate using risk odds A close-up of a text Description automatically generated The odds are defined as the probability that the event will occur divided by the probability that the event will not occur Best formula  Example 340 workers with asthma were compared to 340 workers who had no\ symptoms of asthma. Of the workers with asthma, 47.6 % were exposed to\ isocyanides in spray paint. 23.8 % of the healthy workers were also exposed to isocyanides. Construct a 2x2 table and calculate the odds ratio To figure out to place in table must 47.6 (exposed)/ 100 x 340 (total workers) = 162 23.8 (unexposed)/100 x 340 (total workers) = 81 **Calculation OR from 2x2**\ OR= 162\*259/81\*178 OR= 41,958/14,418 OR= 2.91 Interpretation from example = the exposed had 2.91 times the odds of developing an outcome **Interpretation of OR** Like RR, but replace risk with odds Those exposed to (exposure) had (OR) times the odds of developing (outcome), compared to the unexposed - OR \> 1 = increased odds in the exposed - OR \< 1 = decreased odds in the exposed - OR = 1 = no difference **Attributable risk** An absolute (rather than a relative) measure AKA risk difference, excess risk or rate difference Tells us how much additional disease is occurring among those exposed to something compared to those who are unexposed and therefore How much disease among those who are exposed that could potentially be prevented among removing the exposure Two possible calculations - Risk difference - Rate difference - Subtracting each other **Attributable risk in a 2x2**  AR= (241/12,600) -- (108/13,800) AR= 1.01912698 -- 0.00782609 AR = 0.0113 (x10n) AR = 1.13 cases per 100 persons Has units that don't disappear when you subtract Attributable risk using IR Example from diabetic example Ire (for diabetic women) = 38.6 deaths, per 1,000 person-years Iro (for nondiabetic women) = 3.9 deaths per, 1,000 person-years A close up of a word Description automatically generated AR= 39.6 deaths per 1,000 person-years -- 13.9 cases per 1,000 person-years AR= 24.7 cases per 1,000 person-years **Interpreting the AR** Example 1: AR = 1.13 cases per 100 years Workers exposed to disease had 1.13 additional cases of outcome per 100 workers compared to those who were not exposed Example 2: AR= 24.7 cases per 1,000 person-years 24.7 cases of death per 1,000 person-years in the exposed group can be attributable to exposure e.g. could potentially prevent 24.7 additional cases per 1,000 person-years among the exposed if we could remove the exposure **Population attributable risk** Tells us how much additional disease is occurring in the total population (exposed and unexposed) compared to a group who are unexposed and therefore How much disease in the total population could be potentially prevented by removing the exposure  PAR describes disease burden among exposed as well as non-exposed people (total population) - Additional (PAR) of (outcome) in the total population can be attributed to (exposure) AR = disease burden among exposed only - Additional (AR) of (outcome) among the exposed can be attributed to (exposure) Relative measures indicate the strength of association between exposure and outcome -- which helps determine causation Absolute measures such as PAR tell us how much disease we could realistically prevent by preventing exposure **Learning objectives** **Epidemiology what is it?** The study of distribution and determinants of health-related states or events in specified populations, and the application of this is study is to control health problems The study of what is 'upon' the people - Who gets disease - What diseases do they get - Where does disease occur (and/or come from) - When does disease occur - Why does disease occur **Components of epidemiology** Descriptive epidemiology -- the pattern of disease & the frequency Analytical epidemiology -- identifies possible causes for why particular health-related diseases occur Interpretative epidemiology/applied -- development of programs and services to address health problems What are? Exposures: the risk factors Outcomes: the disease, injury, health state Case definitions: starting point of detecting a disease Natural history of disease and risk: the progression of a disease over time without treatment **Difference between:** Counts: measure of disease is a 'count' of the total number of cases Proportions: expressed as a % describes a component of the population that's diseased Rates: includes a component of time. The frequency of disease over a specific number of time that the population was under study **Prevalence and incidence** Prevalence = how many people have the disease/condition OR how many people have the exposure Measurement of the population at risk  Incidence = how many people develop a disease/condition every year, month, week Number of the population at risk who developed the disease over a given period of time (new cases) A white background with black text Description automatically generated  **Observational studies** Cohort designs and interventional studies: define a population at risk Follow exposed or unexposed over time to count the new cases Case-control and (analytical) cross-sectional studies: identify people with an outcome and determine the prevalence of exposures among people without and with the outcome **Types of surveillance** Passive -- dependent on the discretion of the health care provider Active--specific data from healthcare providers& usually a sample **Sources of error** Random -- can't be controlled by can be reduced Systematic -- not caused by chance, when there is a difference between the characteristics of the sample population Bias - when there is a difference between the characteristics of the sample population **Ways to control bias & error** Bias -- memory aids, data collection tools, clear and concise wording for questionnaires high participation rates Error- big sample size, more precise tools, repeating study **Validity and reliability** Validity- was what was measured what you wanted to measure Reliability -- was the information reliable **Observational studies and experiments** Observational: Case reports Ecological (correlation) study Cross-sectional (prevalence) study Case-control Cohort Experimental/interventional: Actively intervenes to change something and determine whether the intervention/ experiment affects health outcome Has a control group e.g. trying immunisation on sick people **Main types of epidemiological studies** Case reports/ series (descriptive, observational) - Detailed reports based on clinical observations - Descriptive - No comparison group - Generated when an individual has a health issue of interest - Useful for first investigation of the health issue Ecological (correlation) studies (descriptive, observational) - Existing population data is used - Descriptive - Plots the prevalence of an exposure and health outcome in populations or groups against exposure - Does not examine individual data - Good for hypothesis for other tests - Prone to bias Cross-sectional (prevalence surveys) - Assesses the individual's health status with the presence or absence of exposures and/or disease - Descriptive or analytical - Conducted at a single point in time - Only represents a particular population at one point in time - Population health surveys is an example Case-control studies (analytical, observational) - Individuals with the outcome of interest and control group without outcome studied for comparison - Looks for past exposures Cohort (analytical, observational) - Classified on the presence or absence of an exposure to determine the outcome - Followed for period of time - Compares exposed and unexposed - Directly measures incidence and risk (RR and AR) Interventional studies (analytical) - Controls the exposure status of participants/ communities assigned to treatment or control group - Evidence of casual association - Ethical implications Randomised controlled trials (RCTs) - Determines the efficiency of a preventative Casual association - Goal is to determine whether the relationship between exposures and health outcomes is casual or otherwise - Demonstrates to variables that correlate or are associated **Bradford Hill causation criteria** Strength of association - Association between risk factors and health outcome - Common methods: odds ratio in a case-control study - Relative risk in cohort study - Example risk ratio of 6 is stronger than a relative risk of 1.5 Consistency - Findings have been replicated by studies in different places, times, people etc. Specificity - Factor leads to one effect rather than multiple - Additional support to casusual associtation Temporal relationship - Exposure must occur prior to the onset of the health outcome - Established by cohort and experimental studies Biological gradient - Findings are consistent with current knowledge Coherence - Cause and effect relationship Experimental evidence - The association is confirmed through controlled experiments Sufficient causal model - Component cause: a factor that contributes to disease - Necessary cause -- it is required to cause disease - Sufficient cause -- group of components that will produce disease **Statistical interference** Random error Systematic error Cofounding -- factor being studies that is also associated with other diseases **Critical review steps** A list of scientific research methods Description automatically generated with medium confidence **Data collection** Secondary data: - Death certs - Pathology reports - Hospital and medical records - Surveillance data Primary data (for the study) - Surveys - Interviews - Biological samples - Focus groups - Examinations **NHMRC values** Research merit and integrity Justice Beneficence Respect **Prevention of disease** Primary: aim to prevent disease occurring in first place e.g. vaccinations Secondary: reduce morbidity and mortality e.g. screening **Screening** - Likelihood of developing disease - Do Not diagnose, lead to further investigations **Feasibility of a test** - Used by 2x2 to calculate - Positive predictive value (PPV) probability of a person with a positive test result being diseased - Negative predictive value (NPV) probability of a person with a negative result being diseased =a/(a+b) A close up of black text Description automatically generated = d/(c+d) **Accuracy** - 2x2 table - Sensitivity = probability of a diseased person having a positive result - Specificity = probability of a non-diseased person having a negative result  = a/(a+c) = d/(b+d) **Factors to consider before implementing a screening program**  - Must be measured against gold standard - Those with disease should test pos - Those without should test neg - Must be safe - Simple and cheap - Disease should be severe and perceived as a health problem - Must understand the natural history