Tutorial 5 - Interpretation of Studies Q1 a) Odds Ratio Formula = a(d) / b = 38(36) / 18(12) = 6.

33 - from Case control & Analytical cross sectional study - Hence from this: Those with Hyperlipidaemia hv 6 times risk of getting diabetes compare to those who don't have hyperlipidaemia If repeating the test on other sample from the same population, 95% of getting OR within that range. Relative Risk Formula = p (exposed) / p (unexposed) = a/(a+b) / c/(c+d) = 38/50 / 18/54 = 2.28 - from Cohort study - to Estimate association between exposure & disease; Hyperlipidaemia & diabetic - Hence from this: Those with Hyperlipidaemia have 2 times risk of getting diabetes relative to those who don't hv hyperlipidaemia - more accurate and significant because based on incidence (new cases). Others - calculate Attributable risk a/(a+b) - c/(c+d) = 0.427 = 43% - how many percent we can prevent the disease from happen - in this sample, 43% of the diabetic incidence/cases can be prevented if hyperlipidaemia is treated/prevented. - In this sample, 43% of the diabetic cases is attributable to hyperlipidaemia. b) Measure of strength of association: based on risk Case control - Odds ratio Cohort - Relative risk - association: statistical relationship between two or more variables - Risk: probability conditional / unconditional of the occurrence of event in a time eg: probability of individual developing a diabetic over a time interval among known hyperlipidaemia or no lipidaemia c) Magnitude of association: Odds Ratio = 6 times higher Relative risk = 2 times higher

d) Odds Ratio: Those with Hyperlipidaemia hv 6 times risk of getting diabetes than those who don't have hyperlipidaemia Relative Risk Those with Hyperlipidaemia have 3 times risk of getting diabetes than those who don't hv hyperlipidaemia *more accurate and significant because based on incidence (new cases), temporal relationship * to see the effect of hyperlipidaemia which causes diabetic based on new cases e) Possible biase Case control study - Selection (berkson Hospital bias, self selection, case selection) Information (recall bias, interviewer not follow protocol) Confounder (genetic, diet, exercise, medication) Cohort - Information (Loss to follow up, interviewer) - Confounder f) Independent - Hyperlipidaemia, exposure, variables we want to compare Dependent - Diabetic (outcome), interest Others: Prevalence of exposure in disease and non disease a/(a+c) vs b/(b+d) ----------------------------------------------------------------------Q2 Subject were asked whiter they have had a history of migraine before First, selecting sample a) those who had stroke and b) no stroke (control) Then, history taking a) Case control study They selecting case first, which is those who had stroke and no stroke for control then, they ask history of migraine. (there is no follow up) b) Flow chart (in exam)

c) Measure of risk, strength of association: Odd risk ratio ad/ bc = 2.27 the relationship between the exposure and the disease. d) Those with migraine hv 2 times higher risk of getting Stroke compare to group with never had migraine. There is twice risk of developing stroke in those with migraine comapare to never had migraine. e) Possible bias: Case control study - Selection (berkson Hospital bias, self selection, Information (recall bias, interviewer bias not follow protocol,

Confounder (genetic, htn, medication, hyperlipidaemia, diet, exercise) f) Advantages : For common and rare disease : Rabies (rare) & diabetes (common) : Less expensive and Quicker : Many different exposure can be studied. : Good in small outbreak event, outbreak dengue (when no of incidence is higher than expected value) : Fewer subject - Disadvantages : Bias (Selection, incomplete information, Confounder, recall bias) : Cannot calculate incidence, only odd ratio : Difficult in rare cases : Temporal relation not clear : problem in seeking control group & matching variables (like race, income, origin, diet, lifestyle etc) ---------------------------------------------Q3 Description: Case control study -> investigate relationship between Maternal Tobacco vs Low birth weight Baby How: Selection of cases - must define the target cases (baby born with Low BW and must define LBW) - accessed by interviews, Questionnaires Selection of controls - must define the control cases (baby born with normal, must define normal) Matching - Control must hv similar character like, location/hospital, race, occupation, - purpose to remove the confounder 2 type matching : group individual Analyse - Using Odd ratio (exposure among the disease and non exposure among the disease, compare to control) Low Birth weight Normal (Disease) (Exposure) Smoking Mother a b a+b `Non smoking c d c+d - Odd ratio: [(a/b) / (c/d)] = ad/ bc - 1=likely equal in both, >1 smoking likely causes LBW, <1 smoking protective in LBW for Cohort study* How: Selection of exposed group - by exposure status, ability to

communicate Selection of controlled group - not exposed to the ris, but have similar characteristic like working time, working place etc Collect the data - by medical records, answering questionnaire, physical examination, etc Follow up - in period of time; One in 3 month for 4 times a year Analysis - Calculate the rate of incidence of accident in 3 month/ a year. - Use Estimated Relative Risk (to estimate the association between Night shift and accident; like example, (How much likely those working in night shift to have accident on their first 3 night compare to other controlled group) Risk Ratio - p exposed/ p unexposed = a/ (a+b) / c/(c+d) - Attributable risk (method of attributing the occurrence of a disease to a specific exposure, which may contribute the development of disease) : how many percent we can prevent the disease from happen Incidence - Incidence in the other group - No ODD ratio -----------------------------------------------------------Q4 Question 5 Case control study Bias: Selection: Berkson bias (hospital bias) type of illness may varied depends on the service they offered Information - bias due to incomplete or incorrect during data collection - Interviewer/observer bias: hv preconceived expectation/ solicit (ask for to obtain) - Recall bias - asking - Questionnaire bias Page 106 Confounding (Major) - any associated risk like family hx, diet, other illness, stress, income, drug compliance etc Summarise result: calculate the odds ratio for risk of MI of beer MI Non MI Beer 169 670 No Beer 30 63 ad/bc = 10647/20100 =0.5297 = 0.53

- 1=likely equal in both, >1 smoking likely causes LBW, <1 smoking protective in LBW - Drinking beer have a protective effect on MI. - Those who drink beer have lower risk by half of getting MI than those who never/ don't drink. - Drinking beer have a protective effect on MI but only up to certain point, almost daily/daily, 3 times lower risk if take more than eg twice a day, it will have no effect or may contribute to MI \ Odd ratio adjusted for age & district? - THE SAMPLE IS FOR SPECIFIC AGE & DISTRICT - To reduce confounding bias, since age is a great risk factor for MI, District for environment for diet, culture, race. Other factors wish to adjust for: - Diet factor, race, occupation, income, children, etc ------------------------------------------------------------------------Q5 a) Table Dependent (outcome) Babies with Congenital HD Yes No Total Yes 100 50 150 No 400 450 850 500 500 1000

Drug usage

b) Study design: Case control studies Subject were picked whiter they have baby with CHD or No First, selecting sample a) those who had CHD and b) no CHD Then, drug use history taking. (no follow up) c) avoid misclassification bias * Misclssification bias - occur by putting cases and control in the wrong category Avoid - not based on history taking, but do conformational diagnostic test like CT scan or MRI. confirmed by classified medical practitioner - Start study with those with drug usage, then go for baby born, because drug usage is much common d) Association: based on Odd risk ratio: ad/bc = 100(450) / (400)50 = 2.25 Those with history of drug use during pregnancy have 2 times higher risk of

having baby with CHD compare to those with no drug usage. Researcher must look at presence of bias in studies as well as confounder (like genetic family history, htn or gestational DM mother, other drug usage etc) Or can proceed doing Cohort study/ larger sample e) Advantages : For common and rare disease : Less expensive and Quicker : Many different exposure can be studied : Good in small outbreak event, outbreak dengue (when no of incidence is higher than expected value) : Fewer subject - Disadvantages : Bias (Selection, information, Confounder) : Cannot calculate incidence, only odd ratio : Difficult in rare cases : Temporal relation not clear : problem in seeking control group & matching variables (like race, income, origin, diet, lifestyle etc, same number of people, same population) - overmatching is match all over much the same - match on what we not interested - to remove confounding bias

--------------------------------------------------Q6 a) Case control study Subject were picked first, whether they have MI (hospital admit) or no (outcome) Then, working history is taken (a month before) b) a) Odds Ratio Formula = a(d) / b = 80(215) / 85(100) = 2.02 - from Case control & Analytical cross sectional study - Hence from this: Those exposed with more than 9 hours working time a day hv 2 risk higher of getting Acute MI than those work less than 9hours

Relative Risk Formula = p (exposed) / p (unexposed) = a/(a+b) / c/(c+d) = 2.57 - from Cohort study - to Estimate association between exposure & disease; working more 9h and Acute MI

- Hence from this: Those work more than 9 hours a day have 3 times risk of getting Acute MI relative to who work less 9 hours a day - more accurate and significant because based on incidence (new cases) in this sample, 16% of the Acute MI incidence/cases can be prevented if work less than 9 hours a day.. - In this sample, 16% of the Acute Mi is attributable to those work more than 9 hours a day. d) other test - analytical (can test hypothesis) Cohort Study - An observational study, to obtain link/evidence/associate between the exposure and disease - Its analyze the risk factor; follow the group who expose to the risk factor who do not hv disease for a period of time - Do comparison with the control group; group who not expose to the risk factor of the disease - Those group with same characteristic/experience within a Defined period of time. - Looking for incidence (new cases) - Advantages: Exposure data is more accurate (rare exposure: radiation, : Less information bias (recall and interviewer bias, misclassification bias) : Can examine multiple effect on a single exposure : allow measurement of Incidence of disease in exposed and non exposed group, find out the new cases. Relative risk & attributable risk : provide indication for incubation period (if communicable disease) - Disadvantages: Inefficient in evaluating rare disease : Expensive, time consuming, high effort : Validity can be affected by losses to follow up : sensitive to people, take a long follow up : Non response bias : Large number of subject required, big number of drop up e) Measure of Prevention - reduce working hour a day among worker - educate worker how to manage stress - Diet - Exercise - Treat other associated illnesses if present: HTN, hyperlipidaemia - Smoking -