Exam notes PDF

Preview

Summary

Exam notes from lectures and weekly quizzes...

Description

Week 1 Epidemiology: Study of things among people - Disease (health problems) - Distribution (how its distributed in population - Determinants (factors that determines distribution of disease) Descriptive Epi. – describing disease in terms of place, time & person. Who, what, where, when? - Looks at trends (changes overtime) Analytic Epi. – investigating courses – risk factors, determinants, associations. Why? - Compares Validity – how accurate measurements/ truth - Does it measure what its suppose to measure - Reliability = consistency/repeatability/reproducible - Random error affect reliability - Systematic error affects validity of results P value – result of statistical hypothesis test - If p0.5(greater than), do not reject - If for chi-squared test not significant conclude that row and column variable are not associated Statistical hypothesis – decision making tool used to confirm or reject hypothesis - Not valid & reliable – can’t reproduce them - Statistical inference – conclusion based on sample about wider population - Primary prevention (stage of susceptibility)– prevent disease before exposure e.g. vaccine - Secondary prev. (subclinical stage) – early detection of disease/before signs & symptoms e.g. screening - Tertiary prev. (Clinical stage) – reduce impact/diagnosed with signs & symptoms e.g. mental health Accurate and reproducible – validity and reliability Categorical data - Nominal – grouping data e.g. eye colours, disease names - Binary/dichotomous – group with two categories only - Ordinal – ordered series of relationships e.g. severity, mildmoderate, first to last External validity – refers to the degree that you can generalize the findings of a study to wider population Continuous data - Interval – number zero is artificial i.e. exists e.g. temp - Ratio – zero is true or absolute e.g. height (cm), body weight (kg), km/per min

Statistical significance - whether the observed results from a sample are real and therefore reflected in the source population ~ if p calc. AR > calc. AR% Total cohort = 600 Lung cancer No Total Exposed 50 150 200 No exposed 40 360 400 -

Q. In a study evaluating the effect of pesticide exposure on the cancer of the salivary glands 1,154 people with salivary glands cancer and 3,072 people without this cancer were selected. Out of those who had cancer of salivary glands 363 reported recent or past exposure to pesticides whereas among those who did not have salivary gland cancer 160 were exposed to pesticides. Calculate the odds ratio from this study. Outcome No outcome Total Exposure 363 160 523 No exposure 791 2912 3703 Total 1154 3072 4226 OR = (363/791)/ (160/2912) = 8.35 Q. In a study evaluating the effect of pesticide exposure on the cancer of the salivary glands 1,638 people with salivary glands cancer and 3,565 people without this cancer were selected. Out of those who had cancer of salivary glands 328 reported recent or past exposure to pesticides whereas among those who did not have salivary gland cancer 190 were exposed to pesticides. Calculate the odds ratio from this study (give your answer with two decimals). Outcome No outcome Total Exposure 328 190 518 No exposure 1310 3375 4685 Total 1638 3565 5203 OR = (328/1310)/ (190/3375) = 4.45 Week 6 Analytic study – Associations & cause and effect (causation) relationship

E.g. In January 2014, a total of 3215 injectable drug users were screened for HIV at a local clinic. Of these, 905 persons were found to be HIV positive. The same group was examined again in January 2016, and 335 new cases of HIV were discovered. It was learned that of the 905 HIV positive persons diagnosed at the 2014 examination, 95 died, all in 2015. Otherwise, all persons examined in 2014 came to the second examination. a. Point prevalence of HIV in the examined group in January 2014. Point Prev. = Total cases / Population at that point (2014) = 905 / 3215 x 100 = 28.15% b. Point prevalence of HIV in the examined group in January 2016. Point Prev. = Total cases / Population at that point (2016) = [(905-95) + 335] / (3215-95) x 100 = 37.34% (Since 95 were not present at the ‘end point’) c. Two-year period prevalence of HIV in the group. Period Prev. = Total cases (incl. dead) / Average population Population (2014) = 3215 Population (2016) = 3215 – 95 = 3120 (905 + 335) / [(3215 + 3120/2)] = 1240 / 3167.5 x 100 = 39.15% d. Cumulative incidence of HIV in the group for 2 years of follow up. CI = 335 / (3215 – 905) x 100 = 14.50% Week 3 Diagnosis – determines health status and factors responsible - Reliability:

- Perfect agreement & correlation – line of equality as line of best linear fit - Can’t have perfect agreement & no correlation - Possible – perfect correlation & no agreement; perfect linear - Line of equality – if results from first tests agrees with second or equal – perfect agreement between two results - Sensitivity (true positives/ all with disease x 100) – identify people who really have disease (high = more) - Facecal-oral route – contamination e.g. salmonellosis - Vector – e.g. mosquitos for malaria - Objects – fomites (non-living things) - Carrier – e.g. airborne - Asymptomatic – carries without knowing - In prodromal phase – signs and symptoms - Convalescent – while recovering - Genetic – mutation transmitted - Vertical – mother to infant - Horizontal – person to person - Steps of outbreak investigation 1. Prepare for field work 2. Establish the existence of an outbreak - confirm 3. Verify the diagnosis 4. Construct a working case definition – confirm by lab test 5. Find cases systematically and record info. 6. Perform descriptive epidemiology 7. Develop hypothesis 8. Evaluate hypothesis epidemiologically 9. As necessary, reconsider & re-evaluate hypothesis 10. Compare & reconcile with lab results 11. Implement control and prevention measures 12. Initiate or maintain surveillance 13. Communicate findings Epidemic Curve - Point source type outbreak - One-off exposure - Extended source where exposure may spread over time - Propagative type outbreak – continuous/from person to person - Serial intervals – time between onset of disease in one generation of cases and onset of the next Identifying responsible item using attack rate ratio

- Cumulative Incidence = new cases / no of people at risk at start = (90 / 600) x 100 = 15% 15 new cases per 100 at-risk people - Risk for smokers (CI for smokers) = new cases among smokers ÷ at-risk smokers 50÷200 = 0.25 or 25% - Risk for nonsmokers (CI for nonsmokers) = 40÷400 = 0.1 or 10% - RR = Risk for smokers/Risk for non-smokers = (50/200) ÷ (40/400) = 2.5; Interpretation: Smokers have 2.5 times higher risk of developing lung cancer than non-smokers - AR (risk difference) = Risk in exposed – risk in unexposed = 25 10 = 15%; Interpretation: 15% risk of lung cancer is attributable to Smoking. AR% = Risk in exposed – risk in unexposed Risk in exposed = 25 - 10 / 25 = 60%; Interpretation: 60% cases of lung cancer among smokers are due to smoking Measure of association = RR (association) & AR (prevention) Confidence interval – parameter; the smaller the better - E.g. Null hypothesis: RR=1, or OR = 1, AR = 0 or opposite - If p-vale >0.05, CI will include 1 = not statistically significance - E.g. RR = 1.08 (95% CI 0.75 – 1.23) = No stat. - RR = 1.79 (95% CI 1.56 – 1.93) = yes p-value greater and CI Q. Incidence rate ratio = IR exposed / IR non-exposed. What will be the IRR if IR for exposed is 12 new cases per 10,000 personyears and IR for non-exposed is 6 new cases per 10,000 personyears. Interpretation: 12/6 = 2 Q. Assume there is a mid-year population of 100,000 people of whom 35 are sick with pneumonia, and in one year, 26 died. The mortality rate will be 26/100,000. Q. Assume there is a mid-year population of 100,000 people of whom 35 are sick with pneumonia, and in one year, 26 died. The case fatality rate will be 26/35. Q. Lead time is the ‘time interval between detection of a disease due to screening and its detection or diagnosis due to symptoms’. Patient A is diagnosed with prostate cancer at 50 years of age through Prostate Specific Antigen (PSA)-based screening. He then undergoes treatment but ultimately progresses and dies at 60 years of age. Another patient (patient B) who does not undergo screening is diagnosed due to advance symptoms at age 58. He then under-goes treatment and dies at age 60? The lead time for Patient A, who underwent screening is 8 years Week 7 Case-control study: - Cases = with condition/outcome - Controls = without condition

Among those who could care less about My Chicken Rules, a total of 14,691 adults there were 154 cases of binge eating. Calculate relative risk to see if those who watched MCR were more likely to binge eat than those who did not watch the Disease p ram. + Total + 579 5582 6161 113 14129 14242 RR = 1567 / 6311) = 23.69 154 / 14691 Q. A three-year prospective cohort study investigated the relationship between caffeine intake and coronary heart disease among adults between 18 to 80 years of age. Among 4,330 adults who reported high regular caffeine intake, 752 developed coronary heart disease during this three-year follow up. Among those who had low caffeine intake, a total 5,724 adults there were only 387 new cases of coronary heart dis e. Calculate attribute risk: Disease + Total Q. A three-year prospective cohort study investigated the + 752 3578 4330 relationship between caffeine intake and coronary heart disease 387 5337 5724 among adults between 18 to 80 years of age. Among 3,446 adults AR = 752 387 eggs who reported high regular caffeine intake, 708 developed 4330 ¯ 5724 coronary heart disease during the three-year follow up. Among AR = 17.37 – 6.76 = 10.61% those who had low caffeine intake, a total of 5,203 adults there Q. A three-year prospective cohort study investigated the were only 362 new cases of coronary heart disease. What relationship between caffeine intake and coronary heart percentage of cases among exposed are due to their exposure disease among adults between 18 to 80 years of age. Among (you eed to pick one between RR, AR%, AR). 2,743 adults who reported high regular caffeine intake, 650 Disease developed coronary heart disease during the three-year follow + Total up. Among those who had low caffeine intake, a total of 5,112 + (exposed) 708 2738 3446 adults there were only 436 new cases of coronary heart - (no exposed) 362 4841 5203 d ase. What percentage of cases among exposed are due to AR% = 708/3446 x 100 = 20.55% egg Disease t r exposure? 362/5203 x 100 = 6.96% + Total AR% = 20.55 – 6.96 egg eat ate eel + 650 2093 2743 20.55 = 66.13% 436 4676 5112 Q. A study conducted for six months investigated if there was an AR% = 650 / 2743 = 23.70% association between watching My Chicken Rules and binge eating 436/5112 = 8.53% and in a sample of adults between 20 to 29 years old. Among 6,311 AR% = 23.70- 8.53 eggs ate eels who adults who regularly watched the entire 2017 season of this show 1,567 experienced indulging in regular binge eating from time to 23.70 = 64.01% time during this follow up period.

Type 1 Error – mistakenly believe have found significant result or difference – falsely positive result (reject null hypothesis correct) Type 2 Error – mistakenly believe no difference – falsely negative result (accept null hypothesis incorrect) Q. If a study had 85% power, the probability of making type 2 error will be 15% Temporal relationship/sequence – time Case-control – identify and describe source population Week 8 Experimental research - Types of trials: - Therapeutic trial – treatment to prevent death or improve health - Preventative trials – procedures (for healthy or higher risk individuals - Community trials – assessed at entire communities - Field trials – non-institutionalized people in general population - Clinical – test in specific way to evaluate its safe and efficacy Within group design – one group – before and after results Between group design – two or more groups compared Randomised controlled trials (RCT) – 2 or more times Efficacy – ideal condition; Effectiveness – real condition Quasi-experimental (non-randomised design) – allocated - Community intervention trials – defined geographically (towns, regions); socially determined (workplaces, schools) - Uncontrolled Before-after studies – no control group; scores recorded at baseline and then at the end - Controlled Before-after studies – controlled; scores recorded at baseline and then at the end for both groups. Differences attributes to intervention - Time series design – see if intervention had effect significantly greater than trend. - Not random sample – random assigned into groups Internal validity – accuracy of findings for person who are investigated External validity – extent to findings accurately apply who not investigated Number Needed to Treat (NNT) – inverse of AR - Event rate in control group = 35/100 - Event rate in experimental group = 20/100 - Difference = (35/100) – (20/100) = 0.15 - NNT = Inverse of the difference 1/0.15 = 6.666666 or 7 - Interpretation: We will need to treat 7 people to prevent one episode

Selection bias – from any errors in study and/or from factors affecting study participation (sample/subject) Information (measurement) bias – occurs during data collection Week 9 Causality Hill’s Causation Guild lines - Strength of association – size of association as measured by appropriate statistical tests - -Stronger association, more likely relation of A to B is casual - Biological gradient (dose-response relationship) = WEEK 6 - Lack of temporal ambiguity - exposure always precedes outcome - If A believed to cause disease, is clear A must necessarily always precede occurrence of disease - Specificity of association – if exposure is associated with only one disease or disease associated with only one factor - Experiment – condition can be altered by appropriate experimental regimen - Plausibility – agrees with currently accepted understanding of pathological processes – knowledge/mechanism - Consistency of findings – association is consistent when results are replicated in studies in different setting using different methods - Coherence of evidence – compatible with existing theory and knowledge - Analogy – prior causality results with similar exposure = new findings Week 10 Bias – preference; intentional or unintentional - Overestimated or underestimate true value Random error – random variation => fluctuations around true value from faulty equipment - Non-predictable; different each test - Doesn’t affect average, only variability around average Confidence intervals; (1.5, 3) most precise -closest to 1 - Small samples, wider CI - Larger samples. Narrow CI Systematic error – predictable and consistent from true value from faulty equipment. - Incorrectly calibrated or equipment used incorrectly - Affects the average - Inter-rater reliability – agreement between same test – similar result when done by multiple people

- Intra-rater – agreement between results done multiple times by same person - High inter-rater – observed result on repeated tests by different people to be related/similar to one another Misclassification Bias – wrong category - Non-differential – no difference between groups affected by misclassification e.g. inaccurate medical records - Effects/consequences = bias towards null – underestimate - Decreases ‘strength’ of association between exposure and outcome - Differential – rate of misclassification is different for two groups - One or the other, OR one affected more than the other - Bia towards or away from null – under or overestimate - Strong association show up as weak or no association Surveillance bias/ascertainment bias/ detection bias - Follow up on some more closely than others Recall bias – differential recall/ different level of recall between cases and control Reporting bias – results are reluctant to report an exposure due to attitudes, perceptions. Over or under report Observer/interviewer bias – information collected differently between case and controls/ exposed and non-exposed Confounding – mixing of effects/ mixed results - Partially or entirely due to third factor is associated both with exposure and outcome - Be independent risk factor of disease - Associated with exposure - Not intermediate between exposure and outcome Association is real but non-casual confounding Association is NOT real bias Random sample : pick random people to participate in study Randomisation: assign willing participates into 2 groups Prevent confounding - Study design: - Restriction: limit selection - Randomization: do in experimental studies - Matching: selecting patients so potential cofounders are identically distributed between groups e.g. age and gender - Study analysis: - Restriction: restricting analysis to certain groups or exclude - Stratification: divide into singular/ subgroups; groups-specific - Standardization - Statistical modelling

Crude – estimate/measure is calculated without controlling or adjusting for other factors Adjusted – RR estimates the effect of exposure after adjusting or removing effects of cofounding factors

-

Advantages: clear temporal sequence - Can study multiple outcomes from single exposure - Minimise bias in ascertainment of exposure – prospective - IR and RR can be calculated - Disadvantages: expensive & time consuming - Require availability of adequate records – retrospective - Not suitable for rare disease - Not suitable for long time - Loss of follow up - Participants with known exposure status are followed over time and the development of new cases of the disease are records RCT: A clinical trial in which participants are randomly allocated to a treatment and a control; involves concurrent enrolment and follow-up of both groups; gold standard in testing the efficacy of an intervention (therapy/intervention) - Advantages: randomisation balances prognostic factors across groups - Detailed information can be collected on baseline and subsequent - Dose levels can be pre-determined - Binding reduces bias - Disadvantages: subject exclusion - Long period of time required - Large number of subjects - Expensive - Ethical issues - Lack compliance with treatment Critical appraisal – examine evidence to judge trustworthiness, value and relevance in particular content Critical review Introduction 1. Does the lit review establish a clear need for the study 2. How does this study differ from the previous investigations? 3. Is the problem, purpose & hypothesis stated clearly? 4. Are the hypothesis developed from previous literature or from sound clinical observation? 5. Are the hypothesis plausible? 6. Is the potential significance of the study described? Study participants - Are the sampling methods clearly described?

- Random sample; Recruitment; Sample size; Informed consent Inclusion and Exclusion criteria - What limitations may occur because of these criteria? - Assignment to the groups Matched or paired, control group, randomisation - Are all relevant characteristics of Study participants described e.g. age, gender, education, disease or health status etc. - Did investigators controlled for the biases that can affect selection of study participants? If so how? Study Design - Is the study design appropriate, given the objective of the study? - Data sources appropriate (e.g. death certificates, medical records, surveys, self reported data, questionnaires, occupational histories etc) - Are the independent and Dependent variables operationally defined? - How were these variables measured? - Is the reliability of measurements established? - How was patient compliance monitored? - How were patients followed? - How was patient attrition handled? - Were the study participants treated identically? - How were the measurement biases controlled? Procedures - Source and type of equipment - Explanations / references for unusual study methods - Procedures: standardized and consistent. - Blinding: - If blinding was not used, what biases may threaten the study? - Quality assurance procedures: multi centre study, or if more than one individual measured and evaluated the outcome? - Contraindications: modification or discontinuation of the treatment. - Instructions given to the participants quoted or summarized. - Do the ...