Epi test 3 study guide PDF

Preview

Summary

Exam 3 Study Guide for Dr. Spitler, spring 2017...

Description

HLTH 3800 Epidemiology Spitler Spring 2017

Measures of Association: calculating the relative risk for each exposure Attributable risk = the difference in disease frequencies between to populations; 2x2 Tables  a = # exposed, disease  b = # exposed, no disease  c = # not exposed, disease  d = # not exposed, no disease Relative Risk (RR) - often used as a generic term for all measures of association. Rate Ratio (RR) - calculated using incidence rates or incidence density. Risk Ratio (RR) - uses cumulative incidence. Odds Ratio (OR) - approximates RR Standardized Mortality Ratio (SMR) - observed vs. expected with external reference group Proportionate Mortality Ratio (PMR) - based on the use of proportion rather that a rate, used when you have no external reference group True relative risk can only be calculated in studies where we can determine incidence only if we FOLLOW THE EVENTS AS THEY OCCUR Case control does NOT permit “true risk” calculation How to control confounding:  In the design of the research o Restriction (limiting eligibility) o Matching  In the analysis of the data o Stratification (analysis by strata) o Multivariate analysis (analyzing the impact of contributing factors) Interaction – incidence rate of disease in the presence of two or more risk factors differs from the incidence rate expected to result from their individual effects  Positive: the effect is greater than expected o Factors combine to produce a stronger impact than each one would have by itself  Negative: the effect is less than expected o Factors cancel each other out

Etiologic fraction – estimates the proportion of the rate in the exposed population that is due to the exposure  Relative risk in the non-exposed population is not zero  Remove cases that would have occurred without the exposure

Case Control Studies   

Calculate odds ratio, instead of relative risk Ideal for long-latency diseases and rare diseases Addresses shift from acute to chronic diseases with long latency periods

Crucial task: identification of cases and controls; prevent misclassification bias (weakens ability to find a cause)  Find cases and obtain info regarding exposure in the past to some assumed causative factor  Comparison group = controls who are similar but do not have the same outcome of interest  Cases and controls should come from similar populations; demonstrate that cases were like controls EXCEPT for the supposed cause  The only difference SHOULD be some characteristic that is related to the development of disease  determine how cases became cases  Already know disease outcomes, find difference between the two  Considerations for controls: o Source of the cases o Availability of comparable information for cases and controls o Practical economic limitations  Controls are not selected to represent the total, non-diseased population o They are intended to represent people who could have become cases if they had developed the disease o Should be alike to cases in terms of:  Age  Race/ethnicity  Gender  Social factors

1. Review the design strategy o Decide on the source of cases and controls o Select cases based on disease status o Determine past exposure 2. Select study participants o Selection controls = major task o “Matching” in some designs o 3 types of controls:  Hospital-based – don't represent non-diseased population - Patients at the same hospitals admitted for other reasons than the outcome of interest - Major criteria: no known or suspected disease association with the outcome of interest (don’t use emphysema patients in a lung cancer study) - Advantages:  Easy, inexpensive  Cases and controls are likely to come from same “reference population”  Captive audience – recall is high since controls are “sick” and thinking about disease - Disadvantages:  May have some negative health habits that overlap with case habits  Non true representatives of a “non-case” healthy person  Difficulty in deciding hat criteria to use to select hospital controls  General population – difficult to find - Use non-hospitalized persons as controls - Surveys, random-digit dialing, voter registration lists, school rosters, neighborhoods - Advantages:  Useful in studying local environmental problems  Avoids the “sick role” problem - Disadvantages  Poor cooperation (low response and participation rates)  Those available may vary in some systematic way from the norm  Recall bias  “Special groups” – coworkers, friends, relatives - Advantages:  Share habits, diet, environment  Willing to help due to relationship with case - Disadvantages:  May be so similar that they “wash out” exposure effect

 Friends and relatives may have also been exposed, but have not yet developed symptoms 3. Determine exposure status o Techniques must be systematic and consistent o Development of common definition and checklist o Selection criteria must be the same for cases and controls (except disease outcome); should differ only in terms of exposure to the suspected causal factor o How to avoid drawbacks for each type of control  Multiple control groups  Match cases to controls on specific variables that may confound the relationship (ex: age) - Pair matching = individuals - Group matching = match by percentage or proportion in given control categories - Problems with matching  Practical problem: too similar  Conceptual: can’t study what you have matched for - Rules for matching  Not on any variable that may be studied later  Systematic, same criteria and questionnaire, same procedures  1:1  1:4 (no more than 4 controls per case)  more controls = more statistical power 4. Data analysis o Odds ratio – the measure of association for a case-control study  Rate cannot be studied because the investigator determines the rate o Can calculate an odds ratio when:  Controls are representative of the general population  The assembles cases are representative of all cases  The frequency of the disease in the population is small 5. Look for errors a. Misclassification

Analytical

Present = cross-sectional  Investigate the association between potential causative factors and disease outcomes  Do not know past or future  Usually done when a problem is being investigates for the first time  Establish presence of relationships between certain factors and the disease outcome  “First line” of analytical approaches  Snapshot  Simple, quick  Exposition of incidence rates Past = retrospective  Determining differences in lifestyle, diet, family history, environmental exposures, genetics, etc  Two groups: o Those with the disease outcome of interest o Those without the disease outcome of interest  Interviews or surveys—subjects are asked about potential risk factors  Advantages: o Suitable for rare diseases o Inexpensive, short o Minimal ethical problems o Small number of subjects needed o Subjects need not volunteer  Disadvantages: o Selection and memory bias o Inconsistency in definition of disease or symptoms over time o Cannot determine incidence o Limitations in the data available

o Relative risk approximation Future = prospective  Cohort and experimental trials  Know potential risk factors, look at future outcomes  Follow subjects over time, compare exposed/risk factor groups to those without  Advantages: o Less variability and bias than other designs  No recall or memory necessary o Determines incidence o Accurate estimation of relative risk o Good for the study of rare exposures  Disadvantages: o Consistent disease and symptom definitions o Longer time, expensive o Common diseases only o Dependent on volunteers, high drop-out rates, large # of subjects o Hawthorne Effect : people knowing that they’re in the study may affect their natural behavior  Experimental studies o Give researcher control over subject activity and exposure o Drug trials, community intervention trials o Randomized Clinical Trial  Selection of subjects minimizes impact of extraneous factors (less bias)  Subjects are assigned to different treatment conditions and then compared in terms of outcome of interest Sampling  Surveys and questionnaires  Cross-sectional and some types of case-control and cohort studies  Before conducting a survey: o Define the types of questions to be asked o Define sampling strategy o Design and test the questionnaire o Train field workers and data collection personnel o Define techniques for cross-validation o Define the nature of the final analysis  Target population reflects the main question of interest  “all subjects”  Sampling unit: o The individual o A family o A group  Sampling frame: o Existing list or directory o New listing



Time: o Cross-sectional or follow-up o How many subjects are needed for how long

Stratification  Sample size is proportional to strata size  Equal sample sizes from all strata  Sample size inversely proportion to strata variability  Stratified samples are preferred over random o Tend to be better estimators of population trends and event o May be the only way to ensure a representative sample Trend studies  Follow-up surveys may be used to study health and disease trends over time  Multiple surveys are more than one point in time o More than 2 time periods to establish true trend data  Cohort study: o Each subjected is interviewed 3-4 times o Results in depletion or losses to follow-up  Panel study: o Each subject is interviewed 2 times o Each time data is collected, a subset of subjects is interviewed o Different group selected each time Study Designs and Data Analysis  Case-control, cohort, experimental: data presented and analyzed in 2x2 table form  Contingency table  Relationship between a disease outcome and an exposure can be analyzed using combination of cross-products or marginal totals to calculate the risks or odds of the outcome

Cohort Studies “Prospective,” “follow-up,” “incidence studies” Begin with exposure and work forward toward an outcome (case-control is opposite) Greatest advantage = ability to confirm causal associations  Follow groups over time and clearly establish temporal sequence  calculation of true risk Major features  Exposure to passible causative agents is not under the researcher’s control (natural)  Subjects area divided into two groups based on their past history of exposure  Focuses on disease development o Establish which factors are linked with the development if a disease o Each group must be free of disease/outcome at the beginning of the study Advantages:  Time sequence is established  ability to attribute causation  Good for rare exposures (ex: occupational risks), target particular diseases/exposures  Examination of multiple effects of a single exposure  “Captive” populations o Industry or work site o Easy access to subjects or records  Few ethical concerns o Do not withhold treatment o Do not artificially expose to harm  Large sample sizes are required  solid, credible findings  Useful for health service and intervention program development and planning  No memory or recall bias  Influences or confounding variables can be estimated, studied, and controlled  Logistic advantages – some groups are easier to get information on that others o Professional groups, students, participants in medical plans, occupation groups Disadvantages:  Large study population needed, not easy to find  Expensive, time o Very expensive for rare diseases  Requires long-term commitment from subjects, dropout rates  Unforeseen events may affect the outcome  Study population I often not representative of the general population  Limited to a particular disease or exposure  High time, effort, and coordination demands  Data analysis demands strict adherence to systematic information gathering  Difficult to obtain controls if the exposure if popular  May be difficult to achieve “blindness” for subjects and researchers Selection of comparison group  Major principle: comparison groups should be as similar as possible to the exposure except for the exposure under study

Internal control/comparison groups as well as external o Varying levels of exposure Multiple exposures  Comparison group must not be exposed to any of them  May want to use multiple comparison groups to increase validity of comparisons  More exposure reduces the amount f people you can find for your control Data collection  Every person  Different methods of analysis o Study of risk of exposure o Compare incidence and prevalence rates o Calculate relative risk 

Experimental Clinical trials  Randomly assign subjects to exposure conditions and follow over time  Begin with two groups that differ in terms of their exposure to an “experimental” condition and a control condition, follow those two groups over time  Greatest advantage: ability of the researcher to directly manipulate the experiences/treatments received by the two groups o Follow over time and clearly establish temporal sequence  attribute causation  Exposure to possible causative agents is under researcher’s control  Random assignment o Each group is compared in terms of the same outcome of interest  Focuses on variation in treatment or intervention impact  Comparative results of treatment of intervention are followed form the onset of the trial, to permit early assessment of impact  “Stopping rules” o If preliminary results show dramatic positive or negative effects, it can be terminated early and the results can be published  May be either planned or unplanned cross-over o Planned: treatment schedules that change subjects to balance design o Unplanned: crisis or adverse event that necessitates their placement in standard therapy  Best test of cause-and-effect  Advantages: o Randomization o Homogeneity within groups o Blinding to reduce bias  Disadvantages: o Often unethical

o Expensive o Time-consuming o Impractical o Limited generalizability 1. Formulate the major research question (hypothesis) o Interventions to be compared, nature of outcomes o # subjects in each group o Eligibility requirements 2. Measuring primary end point o Which end points are most clinically important? o Quality of life – ability to perform certain tasks o Survival o Complications o Intermediate measures - % of patients who have recurrence of symptoms 3. Measuring the efficacy of the intervention

Experimental designs  Randomly assign subjects to exposure conditions and follow over time  Direct control of all factors involved o Quasi-experiments: controls all assignment, but does not control all factors  Practical and ethical limitations  Quasi-experimental is used for most controlled intervention studies

Bias:

Any systematic error in the design, conduct, or analysis of a study that results in a mistake estimation of an exposure’s effect on the risk of disease Selection bias: if the way in which the controls, or the exposed and non-exposed individuals were selected affects the representativeness of the sample and creates the appearance of an association where one may not exist  Non-random samples  Patterns of invitation and acceptance for participation in research  Subconscious preference pattern in the selection of respondents Surveillance bias: is a population is monitored over time, disease ascertainment may be better in the monitored population than in the general population. This may introduce a surveillance bias that can lead to an erroneous estimation of the risk in the general population  “You find what you are looking for”  If you have been “sensitized” to look for or worry about a condition or exposure, you may be more diligent than if it had not been brought to your attention Misclassification bias: inaccuracies in data collection may assign individuals to the wrong category  Differential misclassification between groups  Can occur across all groups due to some problem with data collection methods Information bias: the way information is abstracted from medical record, employment record, or the way an interviewer asks questions  Surrogate interviews  Recall bias  Interviewer bias Non-response bias: those who do not respond to surveys or interview opportunities or refuse to give their consent to research may differ in significant ways from those who participate in research “Wish” bias: subjects who have developed a disease may be attempting to answer the question ‘why me?’ and seek to show that the disease is not their fault  May deny certain exposures related to lifestyle  Over-emphasize workplace exposures (lawsuits) “Health Worker” bias: subjects who are, as a group, healthier than the population as a whole Incidence-Prevalence bias: if a group is investigated a significant amount of time after exposure to a cause or after the disorder has developed, those who have died and those who have recovered quickly will be “missed.”

Berkson’s bias: the spurious association found between some characteristic and a disease, resulting from admission rates to hospitals or any other setting in which the research takes place, that differs for those with and without the disease  In the general population, there may be no association at all, but there appears to be an association due to differential rates of admission to hospitals or medical treatment facilities Volunteer bias: to be ethical, most studies permit patients to refuse participation. Results of any study with “volunteer” participants are based on the assumption that those who volunteered are similar to those who did not volunteer (similar to participation bias) Hawthorne Effect: the phenomenon that occurs when a subject’s performance changes simply because he or she is being observed or studied Confounding  We may or may not be observing a true association between two variables if there is a THIRD variable that affects the two that we are loking at  A “confounder,” factor X of the relationship between variables A and B exists if: o Factor X is a known risk factor for disease B o Factor X is associated with factor A but is not a result of A  For a variable to be considered a potential confounder, it must satisfy two conditions o Association with the disease of interest in the absence of the exposure o Association with the exposure but not as a consequence of the exposure Ways to reduce bias:  Design o Randomization o Restriction o Matching  Analysis o Stratification o Multivariate techniques...