Introduction to Epidemiology Course Worksheets and Notes
Introduction to Epidemiology Course Worksheets and Notes PubH 6003
Popular in Introduction to Epidemiology
verified elite notetaker
verified elite notetaker
verified elite notetaker
verified elite notetaker
verified elite notetaker
verified elite notetaker
Popular in Public Health
Test Prep (MCAT, SAT...)
verified elite notetaker
This 126 page Bundle was uploaded by Elizabeth Kapelan on Thursday September 1, 2016. The Bundle belongs to PubH 6003 at George Washington University taught by Dr. Keri Apostle in Fall 2016. Since its upload, it has received 40 views. For similar materials see Introduction to Epidemiology in Public Health at George Washington University.
Reviews for Introduction to Epidemiology Course Worksheets and Notes
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/01/16
PubH 6003Principles and Practice of Epidemiology Worksheet for Lecture on Screening (Week 8) PART 1: RELATIONSHIP BETWEEN DISEASE PREVALENCE, TEST SPECIFICITY AND SENSITIVITY, AND PREDICTIVE VALUE (PV) Calculation of Test Sensitivity, Specificity, Predictive Value Positive (PV+), and Predictive Value Negative (PV) The principle measures of screening test validity are the ones mentioned above. Sensitivity answers the question “if I have the disease, what is the likelihood that I will test positive?” Specificity answers the question “if I don’t have the disease, what is the likelihood that I will test negative for it?” PV+ answers the question “if I test positive for the disease, what is the likelihood that I truly have the disease?” PV asks the question “if I test negative for the disease, what is the likelihood that I truly don’t have the disease?” The general conceptual framework for calculating these four parameters of a screening test is: Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive A B A + B Negative C D C + D Total A + C B + D A + B + C + D Where: Sensitivity = A/A + C Specificity = D/B + D PV+ = A/A + B PV = D/C + D The Predictive Value (PV) of a screening test is dependent on three primary parameters: the test specificity, sensitivity, and the disease prevalence. In the first example, we will assess the relationship between disease prevalence and PV. Example 1: You are given three scenarios; in one, the disease prevalence is 1%; in the second the disease prevalence is 5%; and in the third, the disease prevalence is 15%. If the population size is 10,000 persons, the test Sensitivity is 99% and the test Specificity is 95%, calculate the PV under all three different disease prevalence rates. Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative 1 Total 10,000 N = 10,000 PV+ = Prevalence = 1% Sensitivity = 99% PV = Specificity = 95% Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 5% Sensitivity = 99% Specificity = 95% PV+ = PV = Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 15% Sensitivity = 99% Specificity = 95% PV+ = PV = As the disease prevalence changes, what happens to the values for PV+ and PV? What relevance does this relationship have to the decision as to the choice of disease for which a screening program should be developed? In the second example, we will examine the relationship of the PV to Test Specificity. 2 Example 2: You are given three scenarios; in one, the test specificity is 65%; in the second the test specificity is 80%; and in the third, the test specificity is 95%. If the population size is 10,000 persons, the test Sensitivity is 99% and the disease prevalence = 10%, calculate the PV under all three different test specificity levels. Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 10% Sensitivity = 99% Specificity = 65% PV+ = PV = Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 10% Sensitivity = 99% Specificity = 80% PV+ = PV = 3 Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 10% Sensitivity = 99% Specificity = 95% PV+ = PV = As the test specificity changes, what happens to the values for PV+ and PV? Example 3: For a fixed disease prevalence and test specificity, how can you improve the PV+? We have already seen the effect on PV+ levels when either the disease prevalence or test specificity is changed. The last effect we will examine is the effect on PV+ levels caused by a change in the test sensitivity. In this case, the disease prevalence and test specificity are fixed. Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 10% Sensitivity = 80% Specificity = 95% PV+ = PV = 4 Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 10% Sensitivity = 90% Specificity = 95% PV+ = PV = Results of Screening DISEASE STATUS Test Sick Not Sick Total Positive Negative Total 10,000 N = 10,000 Prevalence = 10% Sensitivity = 99% Specificity = 95% PV+ = PV = Part 2: Use of Multiple Screening Tests; Sequential or TwoStage Screening In sequential or twostage screening, a less expensive or less invasive and more acceptable screening test is generally performed first. Those who screen positive are referred for further testing with a more expensive or more invasive test which may have greater sensitivity and specificity. One goal is that by bringing back for further testing only those who screen positive will reduce the problem of false positives. Consider the following twostage screening problem: Test 1: Screening for Blood Glucose Levels Assume: Disease prevalence = 5% Screened Population = 10,000 persons Sensitivity = 70% Specificity = 80% Results of Screening DISEASE STATUS Test Diabetes No Diabetes Total Positive 5 Negative Total 10,000 Fill in the four cells of the screening table. How many persons would be referred for the second screening test (glucose tolerance test)? Test 2: Glucose Tolerance Screening Test Assume: Sensitivity = 90% Specificity = 90% Results of Screening DISEASE STATUS Test Diabetes No Diabetes Total Positive Negative Total Net Sensitivity = Net Specificity = PART 3: SCREENING FOR PROSTATE CANCER A “REALWORLD” SCREENING SITUATION Background: Prostate cancer is the most common cancer diagnosed in North American men, excluding skin cancers. It is estimated that, in 2004, approximately 230,110 new cases and 29,000 prostate cancer related deaths will occur in the United States. Prostate cancer is now the second leading cause of cancer death in men, exceeded only by lung cancer. Currently, the two most commonly used screening tests are Digital Rectal Examination (DRE) which involves manual palpation of the prostate gland through the lower part of the rectum to assess symmetry, size and presence of abnormal nodules, and the Prostatespecific antigen (PSA) test which measures the level of PSA in blood. PSA is a protein produced by the cells of the prostate gland. PSA test results report the level of PSA detected in the blood. In the past, most doctors considered PSA values below 4.0 ng/ml as normal. However, recent research found prostate cancer in men with PSA levels below 4.0 ng/ml (2). Many doctors are now using the following ranges, with some variation: 0 to 2.5 ng/ml is low 2.6 to 10 ng/ml is slightly to moderately elevated 10 to 19.9 ng/ml is moderately elevated 20 ng/ml or more is significantly elevated 6 There is no specific normal or abnormal PSA level. However, the higher a man’s PSA level, the more likely it is that cancer is present. But because various factors can cause PSA levels to fluctuate, one abnormal PSA test does not necessarily indicate a need for other diagnostic tests. When PSA levels continue to rise over time, other tests may be needed. Scenario: Suppose that 100,000 men were screened for prostate cancer for the first time using the PSA screening test. Of these, 4000 men had a positive result (PSA > 4 ng/ml) on the screening blood test; of those who tested positive, 800 had a biopsy indicating a diagnosis of prostate cancer. Among the remaining 96,000 men who screened negative, 100 developed prostate cancer within the following year and were assumed to be false negatives to the original PSA screen (i.e., PSA levels < 4 ng/ml but still developed the cancer) 1) Set up the twobytwo table for these data: Results of Screening DISEASE STATUS Test Prostate Cancer No Prostate Cancer Total Positive (PSA > 4) Negative (PSA ≤ 4) Total 100,000 2) What is the prevalence rate/1000 of prostate cancer in this population? 3) Calculate and interpret the test sensitivity 4) Calculate and interpret the test specificity 5) Calculate and interpret the PV+ of the test 6) Describe two hidden costs of screening for prostate cancer in this example. 7 7) If the cutoff value for a positive screening value (criterion for positivity) for the PSA was increased from 4 ng/ml to 6 ng/ml, how would that affect the test sensitivity, specificity, and PV+? 8 Week 7: Worksheet on Bias, Confounding, and Effect Modification Selection Bias in a CaseControl Study A casecontrol study was conducted of the association between having a pap smear performed and a reduced risk of cervical caner. The research question was: “Do women with cervical cancer have a lower number of previous pap smears prior to cancer diagnosis compared with healthy controls?” Onehundred cases of primary cancer of the cervix were identified from the hospital. An equal number of controls were selected from the neighborhood adjacent to the hospital. This Table Represents the “True Relationship” between a prior history of pap smears and cervical, based on an unbiased selection of controls Cervical Cancer Neighborhood Cases Controls Had Pap 100 150 Smear Did Not Have 150 100 Pap Smear Total 250 250 The “unbiased” odds ratio = (100 x 100)/(150 x 150) = 0.44 or a 56% reduced risk of cervical cancer among women who had a previous pap smear. One big problem: during the selection of controls, only women who were home at the time of contact were interviewed by the researchers. These women were less likely to work, and less likely to have regular medical checkups and an annual pap smear. Consequently, control selection was biased. They were not representative of the study population that produced the cases, violating one of the cardinal tenets of control selection. Instead, in the biased control selection, 50 fewer controls reported having a pap smear than in the unbiased scenario. On this basis, fill in the 2 x 2 for the casecontrol data with biased control selection. Assume that the cases have not changed with their distribution of pap smears and that the same numbers of controls were selected (250): Represents Biased Control Selection Cervical Cancer Neighborhood Cases Controls Had Pap 100 Smear 1 Did Not Have 150 Pap Smear Total 250 250 1) Calculate the new Odds Ratio 2) Compare this with the unbiased odds ratio of 0.44. Why is there a difference? 3) What would you do differently in this study to avoid this type of bias? 4) Do you think that selecting controls from the same hospital would have been a better choice of controls and why? Selection Bias in a RetrospectiveCohort Study A retrospective cohort study was conducted of workers who were involved in the manufacturing of milled automobile parts. They were exposed to metal working fluids, which are a potential carcinogen. An unexposed comparison group consisted of persons who worked in the shipping department. Exposure classification was based on the existence of past employment records. Cases of lung cancer were identified in both the exposed and unexposed groups. In the ideal study, all occupational records were identified and available to determine exposure classification. Unbiased association based upon complete identification of records Lung No Lung Cancer Total Cancer Exposed to Metal 50 950 1000 Working Fluids Unexposed to 50 950 1000 Metal Working Fluids Therefore, the “unbiased” cumulative incidence ratio (or relative risk) = (50/1000) (50/1000) = 1.00 demonstrating no relationship between the exposure and risk of lung cancer. 2 An example of selection bias can occur when the number of lost employment records occurs with greater frequency among the exposed workers who did not develop lung cancer. Lets say that 200 records were lost, all among the exposed workers without lung cancer. The biased 2 x 2 table would look like this (fill in the missing blank numbers) Lung No Lung Cancer Total Cancer Exposed to Metal 50 Working Fluids Unexposed to 50 950 1000 Metal Working Fluids Therefore, the biased estimate of risk would be RR = (50/____) / (50/1000) = ______ Compared this with the unbiased RR of 1.00. Why is there a difference? OBSERVATION BIAS: NONDIFFERENTIAL MISCLASSIFICATION Nondifferential misclassification is a form of observation bias where study subjects in both groups (cases and controls, exposed and unexposed) equally misclassify their exposure status or outcome status. The effect is to drive any observable association towards the null hypothesis of no difference (RR or OR = 1). 1. A casecontrol study was conducted to determine the association between use of artificial sweeteners and risk of bladder cancer. The unbiased data with no misclassification is shown in this table: Bladder Cancer Controls Cases Use of Sweeteners 50 20 No Use of 50 80 Sweeteners The “unbiased” odds ratio is (50 x 80) / (20 x 50) = 4.0 Assume that 30% of each group misclassified their exposure. Create the new “biased” 2 x 2 table: 3 Bladder Cancer Controls Cases Use of Sweeteners No Use of Sweeteners Calculate the “biased” odds ratio based on the misclassified data: OR = Compare with the unbiased odds ratio. This shows the effect of nondifferential misclassification. OBSERVATION BIAS: DIFFERENTIAL MISCLASSIFICATION Differential misclassification, which is defined as one group either overorunder reporting exposure, compared with the other study group can occur when information is collected in a nonconsistent manner from your study groups. The effect is to bias the measure of association either towards or away from the null hypothesis. An Example of Differential Misclassification “True” Distribution Exposure Cases Controls Yes 50 20 No 50 80 The “True” Odds Ratio = 50 x 80 20 x 50 = 4.0 Differential Misclassification: Controls have Poorer Recall of Past Exposures than Cases resulting in 25% Underreporting of Exposure by Controls Exposure Cases Controls Yes 50 No 50 The “Biased” Odds Ratio = 4 In this case, the odds ratio is inflated going from 4.0 to 5.7 as a result of under reporting of exposure by the controls but not the cases. Differential Misclassification Can Also Bias Towards the Null “True” Distribution Exposure Cases Controls Yes 50 20 No 50 80 The “True” Odds Ratio = 50 x 80 20 x 50 = 4.0 Differential Misclassification: Controls OverReport Past Exposures Compared to Cases Resulting in 30% Overreporting Exposure Cases Controls Yes 50 No 50 The “Biased” Odds Ratio = Compare the biased and unbiased odds ratio. What has happened in each situation? How do you explain the discrepancy? CONFOUNDING Confounding is defined as a distortion of the measure of an association due to the effect of a third variable, the confounder. In order to be a confounder, the factor has to be associated with the exposure and independently with the risk of disease. That is, it is a separate risk factor for the disease or outcome being studied. For example, take the study of the association between male gender and malaria: Malaria No Malaria Males 88 68 Females 62 82 The Odds Ratio = (88 x 82) / (68 x 62) = 1.71 Or a 71% increased risk of malaria due to male gender. 5 But, before we accept this finding, we need to determine whether this association is distorted due to the presence of a confounding factor. Outdoor occupation which would result in increased exposure to mosquitoes could be a confounder. In order to assess for the presence of a confounder we need to assess a couple of relationships: 1) Is the factor related to both the exposure and the outcome? Need to stratify the effects of the confounder by the exposure (male or female gender) and the outcome (malaria or no malaria). If the stratumspecific odds ratios are in the same direction (away from the null) and of the same magnitude as the crude odds ratio, then the factor could be a confounder. Occupation versus the exposure Outdoor Indoor Males 68 88 Females 13 131 Stratum specific OR = Occupation versus outcome (malaria) Malaria No Malaria Outdoor 63 18 Indoor 87 132 Stratum specific OR = So both of the OR are in the same direction (away from 1) and of the same magnitude. It looks as though occupation might be a confounder. 2) The second relationship to assess whether occupation is a confounder is to analyze stratify the confounder (outdoor versus indoor occupation) and look at the exposure/outcome relationships within each stratum. Is the Odds Ratio in the same direction and of the same magnitude as the crude odds ratio Outdoor Occupation Malaria No Malaria Males 53 15 Females 10 3 6 Stratum specific OR = Indoor Occupation Malaria No Malaria Males 35 53 Females 52 79 Stratum Specific OR = Therefore, the stratum specific odds ratio shows no association between male gender and malaria within each stratum of occupation. They also differ substantially from the crude OR of 1.71, both in the direction from the null and in the magnitude. You can now say that occupation is a confounder and its effects need to be accounted for in your analysis. 3) Using the MantelHaenszel Technique to calculate an adjusted Odds Ration. One method of adjusting for the effect of a confounder in your analysis is to stratify your data by different levels of your confounder and use a pooled technique to develop a summary odds ratio. The most widely used technique was that developed by Mantel and Haenszel. The general stratification scheme is: Stratum 1 Malaria No Malaria Males A1 B1 Females C1 D1 T1 Where T1 is the total # of cases and controls in the stratum Stratum 2 Malaria No Malaria Males A2 B2 Females C2 D2 T2 Where T2 is the total number of cases and controls in the stratum The MantelHaenzsel formula for the weighted odds ratio is: 7 ∑ (A1 x D1 / T1) + (A2 x D2) / T2 ∑ (B1 x C1 / T1) + (B2 x C2 / T2) = Stratum 1: Outdoor Occupation Malaria No Malaria Males 53 15 Females 10 3 71 Stratum 2: Indoor Occupation Malaria No Malaria Males 35 53 Females 52 79 219 (53 X 3)/71 + (35 X 79)/ 219 14.9 (15 X 10)/71 + (53 X 52)/219 = 14.7 = 1.01 So the adjusted Odds Ratio is equal to 1.01 compared with the crude Odds Ratio of 1.71. This demonstrates that occupation is a confounder and that there is no true association between male gender and the risk of malaria. The association seen in the crude Odds Ratio of 1.71 was distorted by the effect of the confounder, occupation. INTERACTION OR EFFECT MODIFICATION Data from a cohort study conducted in Washington County, Maryland, allows the assessment of the effect of interaction or effect modification. The primary exposure of interest is maternal smoking, both preand postpartum and the primary study outcome is neonatal mortality (deaths occurring before 28 days postpartum). The main premise is that there is an association between maternal smoking and increased risks of neonatal mortality. The third variable of interest is the level of education of the father. The first table presents the study data on maternal smoking: Maternal # Neonatal # Live Births Rate per 1000 Smoking Deaths Live Births Yes 101 4600 22.0 No 121 7934 15.2 8 The Relative Risk (RR) = 22/15.2 = 1.45, indicating a 45% increased risk of neonatal deaths among the offspring of maternal smokers Since the father’s education is a good surrogate for SES, it should be considered another factor in the risk of neonatal mortality. It is grouped into two categories, < 8 grades of education and >/= 9 grades of education. Father’s # Neonatal # Live Births Rate per 1000 Education Deaths Live Births < 9th grade 67 2734 24.5 th 155 9800 15.8 >/= 9 grade The Relative Risk (RR) = 24.5/15.8 = 1.55, indicating a 55% increased risk of neonatal deaths among the offspring of fathers with less than a 9 grade education. We analyzed to see if education was a confounder and it was not. However, education could modify the effect of maternal smoking so it could be an effect modifier. We next stratify levels of education by maternal smoking status: Father’s Maternal # Neonatal # Live Rate/1000 RR Education Smoking Deaths Births < 9 Grade No 32 1967 16.4 1.0 Yes 35 767 46.1 2.8 th >/= 9 Grade No 89 5967 14.9 1.0 Yes 66 3833 17.1 1.1 You can see that when you stratify for father’s education, the increased risk of neonatal mortality is limited to those with less than an 8 grade education RR = 2.8). There is no increased risk of maternal smoking when the father’s education is 9 grade or greater (RR = 1.1). This is an example of interaction or effect modification because the effect (increased risk) is limited to one subgroup and not the other. References Szklo M, Nieto FJ. Epidemiology: Beyond the Basics. Gaithersburg, Maryland: Aspen Publiction;2000:125210. 9 TYPES OF EPIDEMIOLOGIC STUDIES Has exposure already occurred? No Yes (Controlled (Uncontrolled Assignment) Assignment) EXPERIMENTAL OBSERVATIONAL STUDIES STUDIES Sampling with Sampling with regard to disease regard to exposure CASECONTROL CONCURRENT COHORT STUDY STUDIES NONCONCURRENT COHORT STUDY 1. To study the possible protective effect of administering immune globulin to prevent nosocomial infections in premature infants with very low birth weights, investigators conducted a doubleblind study involving neonates weighing 5001500 grams at birth. A total of 584 neonates were randomly assigned to receive periodic intravenous infusions of either immune globulin or a placebo. Nosocomial infection during the next 56 days was assessed. What kind of study is this? _____________________ Results: 287 infants received immune globulin and 297 received placebo. The investigators found that 70 infants who received immune globulin experienced at least one proved infection compared to 104 infants who received placebo. Nosocomial infection Yes No Immune globulin Relative risk = ________________ Placebo 2. Investigators tested the prognostic value of plasma renin activity (a hormone produced by the kidneys) as a possible predictor of heart attack in patients with high blood pressure. They grouped the renin level of their 1700 study participants into 3 profiles: high renin group, normal renin group, and low renin group. They followed the patients for 8 years and found a gradient in the incidence of heart attack, with the highest rates in the high renin group. What kind of study is this? __________________________ 3. In November 1987, an airplaneassociated outbreak of influenza occurred among Navy personnel stationed in Key West, FL. The illnesses occurred among members of a 110person squadron 3648 hours after returning to Key West aboard multiple airplanes from a 10day deployment to Puerto Rico. Fortyone squadron members became ill. The outbreak started with one member who became ill shortly after arriving in Puerto Rico. He transmitted the infection to several other members in the barracks in Puerto Rico. Massive transmission apparently occurred aboard a DC9 airplane on the return trip to Key West. Illnesses were confined almost entirely to enlisted personnel, who stayed in a single barracks in Puerto Rico and who flew aboard the DC 9. Officers, largely spared of illness, flew in separate airplanes to and from Puerto Rico and stayed in separate quarters in Puerto Rico. Investigators used flight records for the return flight to Key West to confirm airplane assignments. They also interviewed and bled all 110 squadron members to determine who became infected. What kind of study is this? ____________________ Results: Of the 90 enlisted persons aboard the DC9 airplane, 40 became ill with influenza. Only 1 of the 20 officers who flew in smaller airplanes became ill. Influenza Yes No DC9 Relative risk = ________________ Other 4. In July of 1987, an outbreak of an acute neurologic illness occurred in Champerico, a small town located on the Pacific Coast of Guatemala. One afternoon, 6 people arrived at the local health clinic complaining of headache, vertigo, paresthesias, and generalized weakness. One child's illness rapidly progressed to respiratory paralysis and death. Within hours, >100 residents were seen staggering out of their houses to seek medical attention. Laboratory analysis of appropriate specimens yielded saxitoxin, the toxin responsible for causing paralytic shellfish poisoning (PSP). To determine what factor was responsible for making people ill, a total of sixtytwo persons with PSP were interviewed along with 43 controls, for a total of 105 study subjects. Ill and nonill persons were not matched on any factors. Study subjects were interviewed regarding their food consumption during the week before becoming ill (cases) and the previous two weeks (controls). What kind of study is this? ____________________ Results: 56 cases reported eating clams while only 7 controls stated they ate clams. Cases Controls Yes Ate clams No Odds ratio: _______________ 5. Matched casecontrol study. A large outbreak of cholera occurred in Portugal in 1974, involving over 2,000 ill persons. To evaluate possible modes of disease transmission, investigators identified 44 people with cholera; for each case, they selected one control matched by age. Thus, there were 44 pairs of cases and controls. The investigators asked cases and controls about the foods and beverages consumed during the week before the case became ill. In the end, they found no pairs in which both the case and control reported eating raw cockles (a type of clam), 11 pairs where the case reported eating cockles but the control hadn’t, 2 pairs where the case had not eaten cockles whereas the control had, and 31 pairs where neither the case nor control had eaten cockles. Control Ate Did not clams eat clams __________________ | | | Ate clams | | | | | | Case | || | | | Didn't eat clams | | | |________|_________| Matched odds ratio = Description, Strengths and Weaknesses of Different Epidemiology Study Designs Study Description Strengths Weaknesses Design Case Study Snapshot description of a health problem in an In-depth description Conclusions limited to individuals individual; Provides clues to new Cannot be used to establish a cause- Qualitative descriptive research diseases effect relationship Identifies potential areas of research Ecologic Aggregate data; no information on individuals Takes advantage of existing Susceptible to confounding Prevalence of potential risk factor compared with data Exposures and disease or outcomes not rate of outcome Quick and inexpensive measured on same individuals Can be used to evaluate Ecologic fallacy (i.e., and error that programs, policies, or occurs if one mistakenly assumes that regulations because the majority of a group has a Allows estimation of effects risk factor, that the risk factor is not easily measurable for associated with the outcome. individuals Cross- Variables measured at one point in time. A snapshot Control over study population No data on the time relationship Sectional of a population at a single point in time. No Control over measurements between exposure and disease distinction between potential risk factors and Several associations between development outcomes variables can be studied at the Potential bias from low response rate same time Potential measurement bias Short time period required Higher proportion of long-term Complete data collection survivors or those with chronic Exposure and outcome data conditions collected for same individual Not feasible for studying rare exposures or outcomes Does not yield incidence data; measures disease/condition prevalence Study Description Strengths Weaknesses Design Case- Measures presence of risk factors/exposures for Effective for rare Limited to one disease Control persons with a condition (cases) compared with diseases/outcomes Does not yield incidence rates persons who do not (controls). Sometimes called a Requires less time and money Less effective than a cohort study at retrospective study because you are assessing than a cohort approach establishing time sequence of events exposures/risk factors prior to the time period in Yields the odds ratio which Potential recall and interviewer bias which the case was diagnosed measures the association Possible selection bias between exposure and disease Potential survival bias Can measures multiple risk factors/exposures Case- Measures exposure frequency during a window Controls for fixed individual Does not automatically control for Crossover immediately prior to an outcome event (hazard characteristics that may confounding from time-related factors period) compared to exposure frequency during a otherwise confound the Design is used under very specific control time or times at an earlier period. Measured association conditions in the same individual (each case acts as their own Effective at studying the control) effects of short-term exposures on the risk of an acute event Nested A case-control study conducted within a cohort. Has the scientific benefit of Non-diseased persons from whom the Case- To carry out a nested case-control study, records of the cohort design. controls are selected may not be Control interest must be available from before the outcome Smaller sample size required. representative of the original cohort Study condition occurred. Avoids selection bias because of deaths or loss-to-follow-up Less prone to recall bias among cases Very efficient-you only need to measure the exposures on your cases and controls, not the entire cohort. Worksheet for the INDIRECT STANDARDIZATION Method A B C D E F Age Groups US Death Rates Population Expected # of Deaths (per 100,000) (in millions) Florida Alaska Florida Alaska < 5 years 251.1 0.85 0.06 2134 151 5 – 19 47.1 2.28 0.13 1076 61 20 – 44 161.8 4.41 0.24 7135 388 45 – 64 841.9 2.60 0.08 21,889 674 > 65 Years 5,104.8 2.20 0.02 112,305 1,021 Total Expected # of Deaths 144,539 2,295 Total Observed # of Deaths 131,044 2,064 To use the Indirection Standardization method to ageadjust rates, you need two pieces of information: 1) the agespecific mortality rates for your standard population. In this example, we are using the agespecific mortality rates for the US population (Column B) 2) the number of persons in each age group for your two populations (Columns C and D) 3) You multiple the standard mortality rates in each age by the number of persons in the same age group (Column B x Col C for each age group) to calculate the expected numbers of death for each age group in both Alaska and Florida 251.1/100,000 x 850,000 = 2134 expected deaths for Florida, ages < 5 4) Do this for each age group in both Florida and Alaska and sum the expected # of deaths for all age groups. 5) Divide the Observed # of deaths by the Expected # of Deaths and multiply by 100 to calculate the Standardized Mortality Ratio (SMR), which is the ratio of Observed to Expected numbers of Deaths. A SMR greater than 100 means that there is an excess of deaths compared to the standard population. A SMR less than 100 means that there are fewer deaths than expected. Therefore, the SMR for Florida = 131,044/144,539 x 100 = 90.7 The SMR for Alaska = 2,064/2,295 x 100 = 89.9 REVIEW SHEET FOR FINAL EXAMINATION PUBH 6003: PRINCIPLES AND PRACTICE OF EPIDEMIOLOGY The final exam will contain 50 multiple-choice questions that cover materials over the entire semester (i.e., the exam is comprehensive). You will have 3-hours to complete the exam once you start. You may access the final exam following the end of your final live session (week 10). Your section leader will post the password for the exam to the wall of your section. You will have up to one week following your final live session to access and complete the exam. The final exam will be open book and open notes (including the formula and definitions s heet that was provided to you earlier in the course). Good luck and we all hope that you enjoyed the course and that it was a positive learning experience for you. 1). You will be responsible for reviewing the following course materials: • Chapters 1 - 16 and the end-of-chapter questions in the textbook by Aschengrau and Seage. • Week 1- the lecture materials and familiarity with the approach John Snow used to investigate the London cholera epidemic (from class exercise #1). • Week 2- the lecture materials on rates and measures of association, worksheet on calculation and definition of different public health rates, how to directly and indirectly standardize rates. • Week 3- the lecture materials covering the interpretation of vital statistics and trends(real and artifact), strengths and weaknesses of death and birth certificates and sources of public health data.Principles of outbreak investigation and the essential tasks in an investigation. The types of epidemic curves and the formation of a case definition. • Week 4 – the lecture materials on Infectious Disease Epidemiology. Know the key concepts and terminology as described in the lecture and in the readings. • Week 5 - the lecture materials on the design of ecologic, cross-sectional and case-control studies. Review the article on public health surveillance and go over the lecture materials on surveillance. Know the different types of surveillance systems, what is passive and active surveillance, the role of the case definition in a surveillance system. • Week 6 - the lecture materials on design of cohort and experimental studies. Know how to identify each of the different study designs and the strengths and weaknesses of each type of design. 1 • Week 7 – review the lecture materials on bias and confounding; Review the worksheet on bias, confounding, and effect modification and be sure that you understand all of these concepts.Review the reading and lecture materials and the handout on analyzing epidemiologic data and how to calculate an Odds Ratio for a matched case-control studyand how to interpret a p-value and confidence interval. • Week 8 - review the lecture materials and the worksheet on disease screening. Know how to set up your 2 x 2 table to calculate Predictive Value Positive and Negative if given the disease prevalence and test sensitivity and specificity. Know how to calculate and interpret sensitivity and specificity and net sensitivity and specificity. Know the types of biases that can occur in screening program (volunteer, lead-time, and length). • Week 9 - Review the lecture materials on disease causation. • 2). You should be familiar with the following epidemiologic concepts, materials, and calculations: • The definition of epidemiology • The differences between epidemiology and clinical medicine • Some key historical figures in epidemiology • The different types of prevention (primary, secondary, tertiary) and examples of each • The concepts and uses of descriptive epidemiology • The role of classifying events by person, place and time • The difference between a fixed and dynamic population • The role of, importance, and what makes up a case definition • Define and recognize prevalence, cumulative incidence, and incidence. Know how they are used and interpreted in public health. • Know how to define and calculate various rates • The relationship between incidence, prevalence, and disease duration • The concept and calculation of person-time of observation and how it is used to calculate incidence rates • How to calculate measures of comparison of disease frequency (relative risk, risk ratio, rate ratio, rate differences, and attributable proportion • How to interpret these measures • The purpose, calculation and interpretation of standardized rates • Know how to use the direct and indirect methods of rate standardization • How to interpret mortality and morbidity rates both within and between countries 2 • Know what factors influence the accuracy and completeness of public health rates; for example, the accuracy of death certification, accuacy of the census, case definition, etc. • Know how vital events (births and deaths) are registered and how they are used in characterizing the public health of a state or country • Know how underlying cause of death is assigned and why this is important • Be familiar with some important sources of public health data (Table 4-1, pages 94-95 in Aschengrau and Seage) • Know the principles and use of descriptive epidemiology • Know the strengths and limitations of mortality data • Know how to categorize and interpret data organized by person, place and time • Know how to interpret both tabular and graphical data • Know the different types of epidemic curves, how they are calculated, and how to interpret them • Know how to calculate and interpret the secondary attack rate • Know why new infectious diseases “emerge” or old diseases “re-emerge” • Define and be able to calculate the attack rate, vaccine coverage, vaccine efficacy • Define virulence and how it is measured; define pathogenicity, infectivity, and immunogenicity • Define and give examples of genetic drift and genetic shift • Define the characteristics of and how interactions between the agent, host, and environment affect infectious disease transmission • Define endemic, epidemic, and pandemic • Define how new epidemics develop • Know the specific steps taken in an Outbreak Investigation and the order in which they are normally performed • Know how to calculate an attack rate and an attack rate difference and how to interpret what they mean • know the difference between observational and experimental studies • Know the difference between descriptive and analytical epidemiologic studies • Be familiar with the factors that influence the choice of a study design • Know the characteristics, principles, and the strengths and weaknesses of the ecologic and cross-sectional study designs • Be able to correctly identify a study using an ecologic, cross-sectional, or case-control design • Know to calculate and interpret the measures of association that are derived from a cross-sectional and case-control study(Prevalence Ratio, Prevalence Odds Ratio, and the Odds Ratio) • Know the issues in selecting cases and controls for a case-control study design • Know potential sources of cases and controls for a case-control study • Know under what conditions you might conduct a case-control study 3 • Be able to describe what a nested case-control and a case-crossover study design is why these designs are used • Be able to define, calculate and interpret an Odds Ratio • Know the strengths and weaknesses of the case-control design • Know to calculate and interpret the measures of association that are derived from a cross-sectional and case-control study (Prevalence Ratio, Prevalence Odds Ratio, and the Odds Ratio) • Be able to define, calculate and interpret an Odds Ratio • Distinguish between the different types of experimental studies: individual trials, community trials, therapeutic trials, and preventive trials. • Be familiar with the reasons for randomization, blinding (either single, double, or triple), the use of a placebo, the effectsof non-compliance, why you would use exclusion criteria in enrolling patients in a trial, issues of generalizability of trial results, the role of informed consent, and what factors determine sample size for a trial. • Know what defines a cohort and how to correctly identify a cohort. Distinguish between the different forms of cohort studies (retrospective, prospective, and ambidirectional), under what conditions you would select one design over another, the key features of conducting cohort studies includinghow to select your study groups, problems in assessing exposure, follow-up issues, and how to calculate person-time. Also be familiar with the strengths and weaknesses of cohort studies. • Know the major types and identify examples of biases including selection bias (selection of cases or controls, surveillance bias, non-response bias, los-tto- follow-up bias) and observation bias (recall bias, interviewer bias,and misclassification). Know these types of bias can occur in a study and how to minimize their effects in study design and/or implementation; know the sources of differential and non-differential misclassification and how it affects your measures of association (RR and OR). • Define and be familiar with examples of confounding; know how to assess for the presence of confounding in a study; know how a confounder can affect your measure of association (OR or RR); know the methods for controlling for the effects of confounding including matching, study restriction criteria, randomization, data stratification, and multivariate analysis. • Know how to interpret a p-value and a confidence interval. • Know what chance, statistical precision, and random error is • Know the process of hypothesis testing and how to interpret a p-value and a 95% confidence interval, and whatthese statistical parameters tell you about statistical significance and precisions of estimates • Know how to interpret an O
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'