Epidemiology Exam 3

Observational analytic study

-important for those exposures that cannot be ethically assigned (subject participants to smoking, surgery, radiation, etc.) -evaluation of associations between exposure and outcome variables begin with a specific priori hypothesis

Sir Austin Bradley Hill- 9 Criteria that could be used to determine whether statistical associations were causal associations

1) Strength of association 2)consistency of association 3) specificity 4)temporality 5) Biologic gradient 6) Biological plausibility 7) Coherence 8) Analogy 9) Experimental evidence

Types of trials

Clinical trials: used to evaluate the efficacy and safety of a new drug or new medical procedure prophylactic trial- used to evaluate preventative measures therapeutic trial- used to assess new treatment methods Community trial- tests a group intervention designed for the purpose of education and behavioral changes at a population level.

3 Methods of hypothesis formulation in disease etiology (John Stuart Mill)

Method of difference- If a risk factor can be identified in one condition and not the second, it may be the factor that causes the disease Method of agreement-if risk factors are common to a variety of different circumstances and the risk factors have been positively associated with the disease, then the probability of that factor being the cause is high. Method of concomitant variation-the frequency or strength of a risk factor varies with the frequency of the disease or condition.

Field Epidemiology

defined under a set of conditions -the problem is unexpected -a timely response may be expected -the investigation time is likely to be limited because of the need for timely intervention

Blinding

placebo:a substance containing no medication or treatment given to satisfy the patient's expectation to get well single blind-participants are unaware of who is receiving treatment, but the investigators are aware. double blind- neither participants nor investigators know who is receiving treatment triple blind- treatment, research approach, nature of interventions are all kept from investigators and participants and the person analyzing drug

Pilot Study

standard scientific approach that involves a preliminary analysis that can greatly improve the chance of funding for major studies.

Case-cohort study , (nested case-control study)

- a case-control study "nested" within a cohort study A sample of cases and noncases are selected, and their exposure status is compared

Confounding

- occurs when an extrinsic factor is associated with the disease outcome and, independent of that association (several variables include age, gender, education level, smoking status, etc.)

Analytical Epidemiology Studies

-answers questions involving how and why -evaluates a predetermined hypothesis about associations between exposure and outcome variables. -these studies make use of a comparison group

Case-control study

-involves grouping people as cases (persons experiencing a health related state or event) and controls and investigating whether the cases are more or less likely than the controls to have past experiences, lifestyle behaviors, or exposures -allows researchers to evaluate both diseases with long latency periods and evaluate one or more exposure variables associated with a given outcome. steps: 1) establish the diagnostic criteria/case definition 2)Selection of controls (should look like the case subjects but without the disease) The ratio of cases to controls should not exceed 1:4 3) Collect information on exposure status, with emphasis on time window

Risk ratio (relative risk)

-measure of association used in cohort studies. This measure reflects the probability of the health related state or event among those exposed relative to probability of the health related state or event among those not exposed RR= a/ (a + b) c/ (c+ d) Numerator is cumulative incidence rate (attack rate) of disease among exposed. Denominator is cumulative incidence rate (attack rate) of disease among unexposed 1= no association, >1 =positive association, <1= negative association Risk ratios can be expressed as percentages ( % increase or decrease change) % increase change- (RR-1) x 100 for RR>1 %decrease change- (1-RR) X 100 for RR<1

Odds ratio/ relative odds

-measures relative probability of a disease in case-control studies -measures strength of association between exposure and disease variables in case-control studies odds of the disease among exposed individuals/ odds of the disease among unexposed individuals - If the odds ratio= 1, this indicates no association between exposure and disease - If the odds ratio is >1, this indicates positive association between exposure and disease -If the odds ratio <1, this indicated negative association between exposures and disease a*d / b*c a= exposed case b= exposed control c=not exposed case d= not exposed control

Observational exploratory study

-no specific prior hypothesis -a variety of associations are examined

Outbreak

-term typically used when the event is confined to a geographic region

Guidelines for investigating clusters

1) initial response 2) assessment 3)major feasibility study 4)etiologic investigation

Steps for conducting a field experiment

1) prepare for field work 2) establish the existence of an epidemic or outbreak 3) confirm the diagnosis 4)establish criteria for case identification 5)search for missing cases 6)count cases 7)orient the data according to person, place, and time 8)classify the epidemic 9)determine who is at risk for becoming a case 10)formulate hypothesis 11)test hypothesis 12)develop reports and inform those who need to know 13)maintain surveillance to monitor trends and execute control and prevention measures 14)carry out administration and planning activities

Causal inference

A conclusion about the presence of a health related state or event and the reason for its existence Direct causal association:has no intermediate factor and is more obvious (trauma on skin results in bruise) Indirect causal association:involves one or more intervening factors and is often much more complicated

Cohort Study

Cohort: a group or body of people. As time passes, this group moves through different and successive time periods Cohort study: generally involves the study of persons who have been exposed and are followed over time with selected health outcomes compared to another group who have not been exposed. Prospective= the predictor variable is measured before the outcome has occurred retrospective= historical cohort is reconstructed with data on the predictor variable (measured in the past) and data on the outcome collected (measured in the past after some follow-up period) Cohort effect: the change or variation in the disease or health status of a study population as the study group moves through time.

Case-crossover study

Each case serves as his or her own control, and the value of a time-dependent exposure in the period is just before the outcome occurred is compared with its value at one or more previous control periods of time appropriate where individual exposures are intermittent, wherein the disease occurs abruptly and the incubation period for detection and induction period are short. -individuals serve as their own controls, the analytic unit is time. The time just before an acute event is the "case" time and is compared to some other "control" time

Etiology

Etiology: the study of the causes of disease and their modes of operation Multifactorial etiology: involves that study of disease arising from many factors at-risk behavior: an activity performed by persons who are healthy but are at greater risk of developing a health related state or event due to that behavior. Predisposing factors: those existing factors or conditions that produce a susceptibility or condition without actually causing it

Bias in cohort studies

Healthy worker effect-> occurs in cohort studies where the workers represent the exposed group and a sample from the general population represents the unexposed group. Workers tend to be healthier than the general population therefore results can be skewed. Loss to follow-up --> researchers lose contact with study participants, resulting in unavailable outcome data for these individuals

Hypothesis testing

Inductive reasoning is the process leading from a set of specific facts to general statements that explain those facts. This is fundamental to the development of hypothesis testing. Ho= null hypothesis (there is no association between x & y) Ha= alternative hypothesis significance level = .05 Selecting an appropriate sample size is necessary to assure sufficient power Select appropriate test statistic and identify degrees of freedom and the critical value Collect data and estimate the measure of association and the test statistic If observed measure exceeds critical value, reject H0 in favor of H1 otherwise do not reject H0.

Ethics in Experimental Research

Institutional Review Boards (IRBs)- assigned at institution level to review plans for research involving human participants The establishment of this board was associated with the Tuskegee Syphilis study which assessed the natural course of syphilis in untreated black males.

Statistical terms

P-value= the probability that an effect at least as extreme as that observed in a particular study could have occurred to chance alone, given that there is truly no relationship between exposure and disease When you accept your null hypothesis, the results are due to chance alone Type I Error-> When Ho is rejected, but Ho is true Type II Error-> When Ho is not rejected, but Ho is false P (Alpha)= P (Type I Error) P (Beta) = P ( Type II Error) Power of a Test is 1- Beta , and is the chance that a given study will detect a deviation from the null hypothesis when one really exists Random error = an incorrect result due to chance Systematic error= incorrect results due to bias

Experimental studies

Researchers evaluate the effects of an assigned intervention on an outcome, the investigators intervene in the study by influencing the exposure of the study participants Experimental studies make use of a comparison group, which allows for testing of specific research hypotheses

Statistical Inference

Sample= a subset of items that have been taken from the population reasons to use samples instead of populations: -samples can be studied more quickly -less expensive -can be more accurate -reflect characteristics more appropriate for studying certain disease in a population Statistical inference: a conclusion made about a population based on sampled data Confidence Interval: a range of reasonable values in which a population parameter lies, based on a random sample from that population

Stages for testing new therapies

Stage 1- unblinded, uncontrolled study with typically less than 30 patients. The purpose of phase 1 trials is to determine the safety of a test in humans (may have advanced disease and exhausted all other options) Phase II - relatively small (up to 50), randomized blinded trials that test tolerability, safe dosage, side effects, and how a body copes with a drug Phase III- typically much larger and may involve thousands of patients Phase IV- large studies (could be random or nonrandom) conducted after the therapy has been approved by the FDA to assess the rate of serious side effects and explore further therapeutic uses

Controlling for bias

To avoid confounding, the level of potential confounding variable can be restricted such that there is no longer an association between the exposure and the confounding variable Matching: strategy in which the distribution of potential confounding factors are forced to be similar between cases and controls (selecting control groups of similar age/gender to limit confounding of these factors) OR= b/c

Effect modification

When an association between an exposure and disease outcome is modified by a level of extrinsic risk factor beyond random variation, the extrinsic variable is called an effect modifier. This influences the association between two other variables in an informative way

Rate Ratio in Cohort studies

When the total time that exposed and unexposed persons are at risk is available rather than the total number of subjects in the two groups, substitutions are necessary. When the denominator in the calculations involved person-time, we use the word "rate" instead of "risk" Rate Ratio= a/ PTe (person time exposed) c/ PTo (person time unexposed)

Web of Causation

a graphic, pictorial, or paradigm representation of complex sets of events or conditions caused by an array of activities connected to a common core, experience, or event. (may not lead right to cause) Decision tree: a flow chart that visually presents a process through which lines and symbols lead to proper decisions and understanding of the role of certain risk factors in webs of causations.

Disease clusters

an unusual aggregation, real or perceived, of health events that are grouped together in time and space and are reported to a health agency. Examples of clusters can involve injury or death related to accidents, natural disasters, political and social upheaval, food poisoning caused by improper food handling cluster investigation: involves reviewing unusual numbers of health related states or events, real or perceived, grouped together in time and location. Cluster investigations are performed to confirm reported disease cases, identify whether the number of cases is above what is expected, and identify causal relationships

attack rate

appropriate statistic for investigating disease outbreaks because it describes rapidly occurring, new cases of disease in a well-defined population over a limited time period

Risks

attributable risk- then a causal assumption is made between an exposure and outcome, the difference in risks is called attributable risk, which is the absolute risk in the exposed group attributable to the exposure (difference in attack rates (risk difference) or person-time rates (rate difference)) Ie-Io Attributable risk percent- can be calculated with the Ie or Io or the risk ratio. Population attributable risk- PTt - PTo

Methodological design

between group design (strongest methodological design) - outcomes are compared between two or more groups of people receiving different levels of the intervention within-group design- the outcome in a single group is compared before and after the intervention. (susceptible to time related factors) Natural experiment- an unplanned type of experimental study in which the levels of exposure to a presumes cause differ among a population in a way that is relatively unaffected by extraneous factors so that the situation resembles a planned experiment.

Epidemic types

common source-starting at specific point through intermittent or continuous exposure to a source over days, weeks, or years propagated-spread gradually from person to person, or a result of common source of exposure that is then spread secondarily from person to person mixed epidemic- involves a combination of both types of epidemics. These are common source epidemics that turn into propagated epidemics.

Group randomization

groups or naturally forming clusters (rather than individuals) are randomly assigned the intervention. There is a greater feasibility to deliver intervention at group level, although individual randomization is more efficient.

Food borne illnesses

illnesses arising from consumption of contaminated or spoiled foods and liquids food borne illnesses are classified in three ways: food infections, food poisoning, and chemical poisoning Food infection: a result of the ingestion of disease causing organisms (pathogens) such as bacteria and microscopic plants and animals (salmonella, Giardiasis, Shigellosis, typhoid fever) Food poisoning: the result of performed toxins that are present in foods prior to consumption: these toxins are often the waste products of bacteria. (Staphylococcus and Botulism) Chemical poisoning: chemical borne food illnesses. Some chemical agents that are beneficial and are essential nutrients in diet can cause food borne illnesses if consumed in large dosages (zinc, vitamin a , niacin )

Nonrandomization

large research populations are not always available, especially in the clinical setting research is expensive, and funds may not be adequate for research procedures Randomization cannot be applied if an entire population is to be affected or subjected to the treatment A convenience sample may be chosen when randomization is not feasible.

Statistical challenges in cluster investigations

most cluster analysis involve post hoc rather than priori hypotheses. The chance of occurrence in the random variation of disease may be the sole explanation for the unusual events Rates may have the danger of being overestimated because of "boundary shrinkage" of the population where the cluster is presumed to exist

Sentinel events

occurrences of unexpected health related events that occur from specific recognizable causes, the adverse health outcome has a known cause

Misclassification

occurs when either exposure or disease status is inaccurately designed. Almost all studies experience some level of this type of bias non differential misclassification (random)= levels of misclassification is the same between cases and controls (always results in underestimated odds ratio) non-random misclassification: levels of misclassification between controls and cases are different. This may over/under estimate the true association

A causal mechanism that requires the joint influence of multiple components includes factors that are predisposing, enabling, precipitating, and reinforcing

predisposing: factors or conditions already present in host that produce a susceptibility or disposition to a disease without actually causing it reinforcing: have the ability to support the production and transmission of the disease or conditions, support and improve the populations health status, and help control diseases and conditions enabling: can affect health through an environmental factor in either a positive or negative way. These factors include services, living conditions, societal support, skills, and resources that facilitate a health outcome's occurance precipitating factors: essential to the development of diseases, conditions, injuries, disabilities, and death.

Designing a Randomized Controlled Trial

protocol- detailed written plan of the study that helps investigator organize, clarify, and refine certain aspects of the study. 1) Select an intervention 2) Assembling the Study Cohort inclusion/exclusion criteria 3)Measuring baseline variables identifying information 4)Choosing a comparison group 5)Ensuring compliance 6) Selecting the outcome/ end point the power of the study is greater when the outcome is continuous instead of dichotomous

Fish Bone diagram (cause-effect diagram)

provides a visual presentation of all possible factors that could contribute to a disease, disability, or death. This type of diagram can assist epidemiologist in defining, determining, uncovering, or eliminating possible causes

Randomization

random assignment- chance is the only factor that determines group assignment -randomization balances out the effect of confounding (controls for both known and unknown confounding factors) -eliminates bias resulting from patient or physician selection

Special Types of Randomized Study designs

run-in design-> all participants in a cohort are places on a placebo and followed for some time period. Those who remain in the study are then randomly assigned to either the treatment or placebo (disadvantage= at time of randomization participants may not reflect target population) Factorial design-> two or more series of treatments are tried in all combinations. This allows investigators to address the efficacy of two interventions in a single cohort of participants. Participants are placed in 4 groups and given different combinations of 2 treatments and a placebo randomized matched-pairs --> subjects are grouped in pairs based on some variable (sex, age, race, etc.) Within each pair, subjects are randomly assigned either treatment or control.

bias

systematic error in the collection or interpretation of epidemiological data selection bias: selection of cases/ controls for a study that is based in some way on exposure. Recruiting all cases in a population avoids selection bias. Berkson's bias: hospital/ patient selection bias (could under/over- estimate based prevalence based on particular patients at the time) Prevalence- incidence bias (Neyman's bias)- form of selection bias in case-control studies attributed to selective survival among the prevalent cases (mild, clinically resolved, or fatal cases being excluded from the case group) Observation bias: can result from the differential accuracy of recall between cases and contols (recall bias) or because of differential accuracy of exposure information because an interviewer probes cases differently than he or she does controls (interviewer bias)