Epi Midterm – Flashcards
Unlock all answers in this set
Unlock answersquestion
What is Epidemiology
answer
epi= upon, among demos= people logy = study the study of what is upon the people The study of the distribution (descriptive) and determinants (analytic) of disease frequency in human populations and the application of this study to control health problems (Aschengrau & Seage 2014)
question
Distribution (PPT)
answer
to describe health conditions at the population leve in terms of trends across: -person -place -time
question
Person (individual level characteristics)
answer
-age -gender/sex -race/ethnicity -socioeconomic position
question
Place
answer
- Geographically restricted or widespread - Presence of clusters (single, multiple) - Relation to place-specific factors
question
Time
answer
- Changing or stable over time - Seasonal variation - Recurrent - Related to an event
question
Determinants of Health
answer
• Causes of disease/outcome (different from predictors of who is at risk for disease) • Factors that are capable of bringing about a change in health ? Physical (biological, chemical, genetic) ? Psychosocial (stress, social networks) ? Behavioral (adverse lifestyle factors) • Occur at many and different organizational levels
question
Does Epi study disease among invidiuals?
answer
No just studies disease among human population groups, not inds, lab animals, petri dish often referred to as population medicine
question
Objectives of Epi
answer
• Identify cause of disease • Determine extent of burden of disease • Study natural history and prognosis of disease • Evaluate existing and new preventions and treatments for disease • Provide foundation for public health policies
question
What is Biostatistics Why do we need epi and bio stats?
answer
• The collection, organization, analysis, and interpretation of numerical data pertaining to the biological and health sciences (Pagano & Gauvraeu 2000) To describe health outcomes in large populations, we need to summarize information, often from samples - Statistics - Statistical techniques used by epidemiologists fall under the heading of biostatistics (about biology) Epidemiologic investigations involve descriptive and analytic methods which draw on statistical techniques for describing data and evaluating hypotheses, biological principles and causal theory
question
Measures of Epi: the basics
answer
-Count (frequency) : NUMBER -proportion (percent probability or risk) : Number out of total with characteristic/total number or total at risk...ranges from 0-1 (0-100%) Rate (100 miles per hour, 3 cases per 22 students over 4 months) : denominator contains time (ranges 0-infinity)
question
The 2 measures of frequency: prevalence Incidence
answer
• Prevalence - The proportion of persons with a particular disease in a given population at a given time • Incidence - The proportion or rate of NEW cases of a disease occurring during a specified time period in a population at risk for developing the disease
question
To protect confidentiality
answer
• Do not include participant names in most databases • Assign ID number to represent each individual • Keep a separate list linking ID to name in a secure location, apart from the data, limited access
question
Types of Variables/labeling in SPSS?
answer
• Nominal(categorical) • Categories with no ranking - e.g. religion (1=Catholic / 2=Muslim / 3=Buddhist / 4=Jewish) • Ordinal(categorical) • Categories with ranking - e.g. health status (1=Poor / 2=Fair / 3=Good) • Continuous • numbers used as numbers, e.g. age in years • discrete numbers, e.g. number of children in a family • Open-ended-"string"variable • Text - Other, specify? _____________ • Date,time
question
Databases -nominal
answer
usually assign numeric codes (or values) to represent categorical data Nominal: yes/no (1=yes, 0= no), white/black/asian (1,2,3)
question
What do you conclude about the relationship (association) between DISEASE and OUTCOME ANSWER KEY
answer
Because 100% of those who did not use a cervical barrier became HIV+ compared to only 33.3% of those who used such a barrier, it seems that a cervical barrier offers some protection against becoming infected with HIV.
question
LECTURE 2
answer
Measures of Disease Frequency and Association
question
Proportions Rates
answer
• Proportions - Fraction of the population affected by a particular disease - Usually the numerator is part of the denominator • Rates - Speed the disease is occurring in the population
question
Types or Prevalence and Incidence
answer
Prevalence • Point prevalence or period prevalence • Always a proportion (0-1) Incidence • Incidence proportion (risk) or incidence rate
question
Prevalence
answer
Proportion of persons with a particular disease/condition in a given population at a given point or period in time • Measures burden of disease/condition • How would this be useful for a public health practitioner?
question
Prevalence EQUATION
answer
Prevalence = # of cases (new & old) [A] / # of total population [N] In a population = A/N in a sample a/n not a rate because no measure of time
question
Point Prevalence
answer
# cases (old & new) / Total population (at a point in time)
question
Period Prevalence
answer
(e.g. last year, last 5 years) # of cases (old & new) / Total population during a time period ex) Example: The small town of Statberg had 5 residents with colon cancer at the beginning of 2014 (Jan 1). During the year, 3 new cases were diagnosed. The estimated population of Statberg in July 2014 was 8,389 Period prevalence = All cases during period Population during period = (5+3)/8,389 = 0.00095 = (0.00095)*10,000=9.5 per 10,000 population in 2014 12
question
Incidence Would someone who already had the disease be included in the numerator? denominator?
answer
Number of new cases of disease commencing during a specified time period in a given population at risk (proportion) or for a given person-time at risk (rate) Measures RISK of acquiring a disease or illness in a population Incidence can be characterized in different ways - Risk (proportion) - Rate Gordis refers to both as "incidence rates" NO Risk is also referred to as "cumulative incidence" or "incidence proportion" • Interpreted as the probability of developing disease • Assumes entire population at risk has been followed for full duration of study • Often used in fixed populations (countries, outbreaks)
question
Measuring Incidence risk. why is it not a rate
answer
Risk = Proportion of population developing the disease over a period of time Not a rate... .why? .because it is a proportion 1. Numerator is a subset of the denominator 2. Denominator is the total population 3. Time element is the time period in which population at risk was observed, e.g. incidence risk of HIV in 2014 15
question
Incidence Rate
answer
Occurrence of newcases of disease that arise during person-time of observation in a period of time New cases of disease / Person-time of observation in population AT RISK for developing the disease
question
Person-time
answer
Sum of individuals units of time that persons at risk for developing disease have been followed • 1 person followed for 1 year = 1 person-year • 1 person followed for 50 yrs = 50 person-yrs • 10 people followed for 5 yrs = 50 person-yrs • 50 people followed for 1 yrs = 50 person-yrs • Apersoncontributestimetothe denominator of an incidence rate until the study ends or he/she: - Develops outcome being measured (e.g. disease) - Dies from other causes - Is lost to follow-up • Is similar to the concept of milesperhour - Measures the rate at which disease is developing during a given time period
question
Incidence Rates
answer
Also called incidence density 2 cases per 49 person-years often presented as 2/49 x 100 = 4.1 cases per 100 person-years need to specify time frame- 4.1 cases of hep c virus infection per 100 person-years in 2013
question
Crude and Category-specific measuressummary measure of health for entire population in a given time period
answer
Crude = summary measure of health for entire population in a given time period (- Overall incidence proportion = 122.9 cases of breast cancer per 100,000 women in the US (SEER, 2003-7) • Category-specific rates/risks - calculated for a population subset in a given time period - Incidence proportion for black women = 118.3 per 100,000 black women in the US (SEER, 2003-7)
question
Benefits/Drawbacks of incidence measures
answer
Cumulative incidence(incidence risk) - Easy to calculate and understand (proportion) - Estimates average risk - Often used in fixed populations Incidencerate(incidence density) - More accurate, especially in dynamic populations - Takes into account different amounts of time observed - Usually need a dedicated epidemiologic study - Less intuitive
question
Common Uses of Prevalence and Incidence
answer
• Prevalence - Planning services or programs - Determining health priorities • Incidence - Studying causes of diseases (etiology) - Evaluating program effectiveness
question
Mortality data: Why?
answer
• Trends in mortality 'rates' provide information on: • Differences in risk of dying by person, place and time • Disease severity • Effectiveness of treatment regimes • Surrogate for incidence for severe diseases
question
historical info on mortality/morbidity
answer
see lecture 2
question
Annual Mortality 'rate'
answer
Annual mortality 'rate' for disease x= [Total number of deaths from disease x / Number of people in the population at midyear] * 100,000 • All cause mortality rate in the US in 2014 = 724.6 per 100,000 population (CDC) • Mortality rate from heart disease (#1 cause of death) in 2014 = 167.0per 100,000 population (CDC) 40 Annual Mortality Rate: Expressed PER 100,00 population
question
Case Fatality Rate
answer
Case fatality 'rate'(CFR)=[Number of people who die after a specific disease onset or diagnosis over a specified period of time / Number of people with that disease during that time period ] * 100 Measure of severity of disease •14%of women diagnosed with breast cancer die within 5 years of diagnosis (American Cancer Society) NOT REALLY RATES- proportion- person time of observation not in denominator CFR expressed as a percentaige
question
Proportionate Mortality Years of Potential Life Lost (YPLL)
answer
Proportionate mortality=[Number of people who die of a specific disease during a specified period of time / Number of deaths during that time period ] * 100 (YPLL)=Measure of premature mortality (standard: <65 or <75 years)
question
Cross tab- 2x2 table
answer
see lecture 2 slides D+ D- Total E+ a b a+b E- c d c+d Tot a+c b+d a+b+c+d
question
Prevalence Ratio
answer
Used in studies measuring the exposure and disease at the same time Prevalence in E+ / Prevalence in E- = [a/(a+b) / c/(c+d)] = 15% prev (smokers) / 4.2% prev (nonsmokers) = 3.57 Smokers were 3.57 times as likely to have had prevalent flu symptoms as nonsmokers in Sept. 2015 Ratio - dimensionless, range from 0 to infinity
question
PREVALENCE RATIO ANSWER
answer
Smokers were 3.57 times as likely to have had prevalent flu symptoms as nonsmokers in Sept. 2015
question
Interpreting the Association
answer
Results of this hypothetical survey suggest smoking as a possible cause of flu, but many limitations: • Need to consider what is the biological plausibility and pathways • Need to know what the existing literature has shown • Need more rigorous study design to establish temporality We know: - Usually not possible to establish which came first (temporal association between E and D) • e.g. Did some people stop smoking due to flu symptoms (underestimate true association)? - Measured prevalence not incidence • e.g. Is duration of flu symptoms associated with smoking, rather than development of flu symptoms? Do smokers have a harder time getting over the flu than non-smokers? 49
question
Risk Ratio
answer
Used in studies selecting participants based on the exposure status at risk of developing the disease = Incidence in E+ / Incidence in E- = [a/(a+b) / c/(c+d)] = 15% Risk (smokers) / 4.2% Risk (nonsmokers) = 3.57 Smokers were 3.57 times as likely to develop flu symptoms as nonsmokers during 2015
question
RISK RATIO ANSWER
answer
Smokers were 3.57 times as likely to develop flu symptoms as nonsmokers during 2015
question
Odds Ratio
answer
Used in studies examining rare diseases = Odds in E+ / odds in E- = a/b/c/d = 0.2 odds (smokers) / 0.045 Odds (nonsmokers) = 4.44 The odds of flu among those who smoke is 4.44 times the odds of those who do not smoke during 2015
question
ODDS RATIO ANSWER
answer
The odds of flu among those who smoke is 4.44 times the odds of those who do not smoke during 2015
question
Measures of association for cohort studies
answer
Risk and Risk Ratio
question
RISK RATIO ANSWER & INTERPRETATION (cohort study)
answer
RR= Risk E+/Risk E- RR=17.2/26.3=0.65 Interpretation: The risk of skin cancer, among those exposed to anti-wrinkle cream, is 0.65 less than among those who were unexposed. The cream appears to be protective against skin cancer because the RR or 1, it appears that the use of cream (exposure) results in a higher risk of dementia (disease).
question
RISK RATIO ASSOCIATION ANSWER
answer
The risk of skin cancer, among those exposed to anti-wrinkle, is 0.65 times less than the risk of those unexposed. Because the RR is less than 1, using anti-wrinkle cream is associated with a lower risk for skin cancer (it may be protective). ANSWER 2 The risk of dementia, among those using anti-wrinkle cream, is 8.22 times the risk of those not using the cream. Because the risk ratio is > 1, these results indicate that the use of anti-wrinkle cream is associated with an increased risk of dementia.
question
RISK DIFFERENCE ANSWER & INTERPRETATION
answer
RD=Risk E+ - Risk E- RD=0.172 - 0.263= -0.091. Women using anti-wrinkle cream have a lower (note the negative sign for the RD) risk of skin cancer of 0.091. If the RD was greater than 0, we could see there is an excess risk of disease among the exposed. Since the RD 0 we can say there appears to be an excess risk of dementia among those using the cream.
question
EXPOSURE CAUSE OF DISEASE? YES OR NO? INTERPRETATION ANSWER
answer
There is always the possibility that other causes of skin cancer were distributed unevenly in the exposed and unexposed group, leading to incorrect results in our study (i.e. bias and confounding*). Those who are concerned about wrinkles may also be more health conscious in general and, therefore, have a lower risk of skin cancer (confounding*). Also, there is no information about whether those looking for skin cancer were blind to exposure status. If not, perhaps they looked harder & found more lesions in the exposed, thus under-estimating the true protective effect (i.e. ascertainment bias*). *Note: You will learn more about confounding and bias in future lectures. ANSWER 2 This analysis does not take into account: 1) TEMPORAL ORDER Although this is a cohort study, the temporal order of the cream and dementia association is in question here because the participants were not screened for dementia at the baseline visit. Since your colleague used your cohort (originally assembled to test the association between cream and skin cancer) to test the association between cream and dementia WITHOUT screening for dementia at baseline, we do not know the timing of the onset of dementia. 1) possible LOSS TO FOLLOWUP 2) possible CONFOUNDING especially by AGE; 3) Bias (which is always a possibility in cohort studies).
question
LECTURE 3
answer
Cross-sectional and Ecological Studies
question
Case Reports
answer
Case report: Detailed report profiling ONE case • Example: 40-yr old premenopausal woman with pulmonary embolism 5 weeks after starting oral contraceptives Outcome: Pulmonary embolism (PE) Exposure: Oral contraceptive (OC) use Can you say much from this case report about whether or not OC use is associated with PE? NO
question
Case Series
answer
Case series:Report describing characteristics of more than one patient with a given disease • Example: 5 young, previously healthy homosexual men diagnosed with Pneumocystis carinii pneumonia at 3 LA hospitals (MMWR 1981) Outcome: Pneumocystis carinii pneumonia (PCP) Exposure: Sexual behavior Can you say much from this case series about whether or not sexual behavior is associated with PCP? 5 case series = aggregation of case reports. several cases of a particular disease. all present same characteristics and same risk factors. we DON'T KNOW much between exposure and outcome
question
Benefits and Drawbacks of Case Reports and Case Series
answer
Importance • Early detection of unusual occurrences or presentations • Preliminary source of information for potential exposures Limitations • Do not have comparison group to ascertain the presence of exposure (s) • Usually very small sample sizes • Often do not systematically collect exposure or other information but rely on medical charts they don't have a comparison group, usually small, helpful for prelim info about exposures effecting pop'n
question
Difference between Observational and Experimental
answer
Observational: • Exposure is not under the control of the researcher • Used when assignment of exposures is not ethical, too expensive, or otherwise difficult • Introduces challenges such as how to accurately measure exposure ; avoid important differences between your exposed and unexposed groups of people Experimental • Researcher "assigns" exposure (or treatment) to a group of study participants • Assignment is randomized • Maximizes control over influence of other factors, improves measurement of exposure • Expensive, can be unethical
question
Descriptive vs. Analytic Studies
answer
Descriptive • Describe burden and patterns of health (person, place and time) • Usually not conducted to make direct inferences about causality • Often provide first important clues about possible causes of disease Analytic • Explicit a-priori hypothesis stated ; tested • Examines associations between exposures and outcomes to understand causes and prevent diseases • Uses appropriate comparison groups • Exposure groups • Outcome groups
question
A-priori
answer
Based on theoretical deduction rather than empirical observation. "sexuality may be a factor, but it cannot be assumed a priori"
question
Descriptive Studies
answer
• Describe disease distribution or characteristics of individuals with disease by person, place and/or time • No a-priori hypotheses (used to generate hypotheses) • Case series ; case reports • Cross-sectional and ecological studies
question
Cross-Sectional Design
answer
snapshot of what happened particular time. Collect info about exposure and disease at the same time. Ask about exposure and disease status among a sample of individuals in a defined population (e.g., a representative sample of the general population) Outcome and exposures in population individual assessed at a given point in time - A SNAPSHOT • Frequencies (PREVALENCE) of disease/outcome of interest measured and examined by exposure, demographic and other subgroups • Extremely useful for public health planning and hypothesis generation • Cannot determine whether exposures preceded disease and thus cannot infer causality NO TEMPORAL ORDER
question
Examples of Cross-Sectional Studies
answer
NationalHealthSurveyAct,1956 • On-going series of representative cross- sectional national surveys since 1960 •Objective data obtained from physical examinations; self-report data from interviews • Repeated cross-sectional surveys can be powerful to evaluate program and policy implementation
question
Strengths and Weaknesses of Cross-Sectional
answer
Strengths: - Compared to other study designs, fast ; cheap - Determines burden of disease for public health planning - Good for hypothesis generation - Can look at multiple outcomes and exposures Weaknesses: - No ability to confirm temporal order for the exposure-disease relationship and thus cannot rule out reverse causality • Not always true (e.g. genes, birth exposures) - Cannot calculate incidence (only prevalence) • Tend to find cases with long duration (survival bias)
question
Prevalence Ratio for Cross-Sectional
answer
Used to measure associations in cross- sectional studies = Prevalence in E+ / Prevalence in E- = [a/(a+b) / c/(c+d)] = 15% prev (smokers) / 4.2% prev (nonsmokers) = 3.57 Smokers were 3.57 times as likely to have had prevalent flu symptoms as nonsmokers in Sept. 2015
question
INTERPRETATION of MEASURE ASSOCIATION PREVALENCE RATIO for Cross-sectional studies
answer
P=1 -Probability of having the disease are identical in both groups during time t; NO RELATIONSHIP between exposure and outcome P;1 Probability of having the disease is greater in exposed group than in unexposed group during time t; POSITIVE relationship between exposure and outcome P;1 Probability of having the disease is lower in exposed group than in unexposed group during time t; PROTECTIVE or INVERSE relationship between exposure and outcome Results of this hypothetical survey suggest smoking as a possible cause of flu, but many limitations: Need to... • consider what is the biological plausibility and biological pathways • know what the existing literature has shown • more rigorous study design • Usually not possible to establish which came first (temporal association between E and D) - e.g. Did some people stop smoking due to flu symptoms (underestimate true association) • Measured prevalence not incidence - e.g. Is duration of flu symptoms associated with smoking, rather than development of flu symptoms? Do smokers have a harder time getting over the flu than non-smokers?
question
Difference between Cross-sectional and case series
answer
if you cannot fill a 2x2 table study is probably case series.
question
Ecological Studies
answer
• Assess exposure and disease associations using population-level rather than individual- level data • Unit of analysis=populations or groupsof people, not individuals • Often based on surveillance data • Often require age standardization to make appropriate/fair comparisons Examine associations between: 1. Group exposures (proportion exposed, mean exposed) -single payer or multipayer health care system -neighborhood SES 2. Group outcomes (Incidence proportion, incidence rate, mortality rate) -mortality rates from acute myocardial infraction ALWAYS GROUP LEVEL DATA...we don't know ind level data just totals
question
Ecological: Units of Analysis
answer
Geographic - census tract - county, city or state - country Institutional - schools - hospitals Time - annual - seasonal
question
Example of Ecological Study
answer
•Does presence of a national pension plan lower fertility rates in a country? - Compare countries with a pension plan to those without in terms of fertility rate (# children per woman of reproductive age) - Not interested in association between how women receiving pension plans differ from those who do not in terms of number of children. Interest is whether policy has the net effect of lowering fertility at the national level
question
Exposure Measures in Ecological Studies
answer
Three general types of group-level exposure measures are used in ecologic studies: 1) Aggregate data 2) Environmental measures 3) Global measures Aggregate: Summation of individual exposure measures • Mean fat intake • Proportion of the population living below the federal poverty level • Counts of cigarette packs sold • Literacy rate • Counts of health club memberships Environmental: Exposures describing geographic area where groups live (could be measured for individuals, but easier for groups) • Air pollution • Temperature, humidity, sunshine, etc. • Land-use mix Global: A ubiquitous, homogeneous exposure per group (everyone is similarly exposed). No analog at the individual level • Seatbelt law • Speed limit • Food laws (e.g. transfat ban, calorie labeling) • Type of healthcare system • Population density
question
Ecological Studies measure...
answer
-disease distribution across groups or time (surveillance) -Type of disease may vary.. prevalence, incidence or mortality rates, counts of disease, change in disease, dichotomous (presence of human cases of west nile in a state yes/no) -type of measure impacts how it can be analyzed and interpretted
question
Correlation Coefficient
answer
Correlation coefficient or r - Tests hypothesis of whether a linear relationship exist as a measure of association between E and D - Two types of 'r' to use, depending on the outcome's distribution • PEARSON - A parametric test (outcome must be continuous ; should be NORMALLY distributed) • SPEARMAN - A non-parametric test (outcome can be discrete or ordinal) • For a perfect linear relationship with positive slope, r = 1 • For a perfect linear relationship with negative slope, r = -1 • For no linear relationship, r = 0 Guidelines: • 0.00 - 0.25 Little or no relationship • 0.25 - 0.50 Fair • 0.50 - 0.75 Moderate to Good • ;0.75 Good to Excellent
question
Limitations of Ecological Studies
answer
Temporal order cannot be determined when based on prevalence measures • Which came first, presence of supermarkets or income/racial composition of population? Cannot rule out reverse causality Group-level analysis • Can only interpret the results at the group level, not at the INDIVIDUAL level • Who actually shopped at the larger markets?
question
Ecological Fallacy
answer
Incorrectly interpreting that a group-level association exists at the individual level • There may be something that explains the group level association that was not accounted for, and thus results are not true at the individual level • Because ecological studies often use data collected for other purposes (e.g. surveillance), you may not have ideal group level data or information on all factors you want to account for
question
Strengths and Weaknesses Ecological Study
answer
Strengths - Strong design if exposure of interest is group level (e.g. policies, environmental factors,other global measures) - Fast ; easy, can often use existing data (surveillance data) - Sometimes measure socially unacceptable exposure more accurately at group level • Smoking - ask people how much they smoke or measure cigarette sales in the state • Diet - ask people how many Big Macs they eat or measure how many Big Macs are sold Weaknesses - Cannot establish temporal order if based on prevalence measures - Cannot make conclusions about individual level exposure and risk of outcome (ecological fallacy) USEFUL To... • To generate or to test etiologic hypotheses • To evaluate the impact of intervention programs or policies
question
1.Describe the variables in terms of the mean, standard deviation and range of fluoride content, rate of cavities, and rate of oral cancer.
answer
•Fluoride mean= 0.70; SD= 0.75; Range= 0 - 2.60 •# Cavities per 100: Mean = 537.38; SD = 243.39; Range = 236.00 - 1037.00 •Oral Cancer rate per 1,000: Mean = 2.10; SD = 0.94; Range = 1 - 4
question
Histograms and what they show eco studies
answer
The histogram below shows a count of the various data points for the variable # of cavities. The distribution of # of cavities appears to be bimodal, with most values falling within the 200 and 700 range. Therefore, the distribution is not a normal distribution (does not appear to be in the shape of a normal curve
question
ECO STUDy CORRELATION
answer
The Pearson correlation coefficient is -0.857. Since this is a negative correlation, we can say as the fluoride content of the water supply increases, the number of cavities per 100 children decreases. In other words, there is an inverse relation between fluoride content in the water and number of cavities. This is a strong correlation because the coefficient is -0.857 which is close to 1.0. Please note: The Pearson correlation is a parametric test meaning you would report it only when you have a normal distribution. According to the histogram above, # of cavities is not a normal distribution. Therefore, you would report the Spearman correlation above (-0.9). The Spearman correlation is non-parametric meaning you do not need a normal distribution to report it. Since the histogram above is not in the shape of a normal distribution, we report the Spearman correlation, which does not require assumptions of normally distributed data . It is similar to the results of the Pearson correlation. Spearman correlation = -0.9. Negative, strong correlation The scatterplot above confirms this in that there is a fairly linear line (dots closely nicely around a line and are not dispersed all over the graph) indicating a strong relationship..
question
Lecture 4
answer
Cohort Studies
question
Cohort
answer
a population group with a common defining characteristic: -age cohort: a group of people within an age range (0-5 years) -educational cohort: group of people who studied at the same time (graduated high school the same year) -occupational cohort: group of people who have the same or similar occupations
question
Cohort studies in Epi
answer
-designated group of individuals who are followed over a period of time -cohort could be assembled from: - a defined population (based on some geographic, demographic, or other common characteristic), irrespective of exposure status. -Exposure-based (based on whether or not they were exposed to a risk factor)
question
Classification of Epidemiologic Study Designs
answer
Observational studies: exposure is not assigned by a research analytical: tests a-priori hypothesis about risk of disease or (health outcome) in exposed versus unexposed ind'ls Cohort= observational Observational= cross sectional, ecological, cohort, and case control
question
Descriptive vs. analytic
answer
Descriptive Studies (case report, case series, cross sectional and ecological): -describe patterns of health phenomena (person place and time) -no a-priori (pre articulated) hypothesis -cannot make direct inferences about causality -often provide first important clues about possible causes of death Analytic Studies (cohort studies) -explicit a-priori hypothesis stated and tested -examines associations btwn exposures and outcomes to undesrtand causes and prevent disease -uses appropriate comparison groups: exposure groups & outcome groups
question
Cohort study Design
answer
collects information on exposure of interest at beginning- all participants must be disease free D- upon enrollment (or are excluded) Followup= calculate which measure of occurrence?= incidence of disease
question
Types of Cohort Studies
answer
1.Prospective • Recruitment & baseline assessment of exposure status occur IN PRESENT • Participants followed into the future to determine disease risk 2. Retrospective • Recruitment & baseline assessment of exposure status gathered USING RECORDS FROM THE PAST • Participants are followed forward from that time to determine disease outcome Prospective: 1. recruit cohort, baseline assessment of exposure status Retrospective 1.Recruit cohort, baseline assessment from PAST records rest of the study design is the same... Exposed, unexposed, disease no disease
question
retrospective cohort studies
answer
assemble cohort and determine prior exposure status from records collected IN THE PAST Follow participants forward and assess health outcomes: -through some point in the past using exisiting records -up to the present (often defined as retrospective cohort) or -Into the future (often defined as historical prospective or ambispective)
question
Measuring Exposure and Disease in Cohort Studies
answer
exisiting records -Medical, employment, etc. +'s:OFTEN CHEAPER and more rapid new data collection options -self reported (interviews) -examination based (physical exam, lab tests, biological specimens) -environmental monitoring results from new data collection are often high quality but expensive
question
Follow up in Cohort Studies
answer
length of followup can range from days (infectious diseases) to several decades (chronic diseases) the longer the time period, the more difficult to maintain contact with the cohort and the more expensive the study -loss to followup can reduce validity of results if outcomes are different among those lost compared with those tracked
question
Strengths of Cohort Study
answer
-clear time-order relationship btwn the exposure and the disease -can directly measure incidence -good for rare exposures -good exposure assessment (prospective) -possible to look at multiple diseases (compare risk of many disease outcomes for any given exposure)
question
Weaknesses
answer
-expensive and lengthy (especially prospective cohorts) -inefficient for rare diseases (need to wait a long time for enough cases to accrue) -drop out or "loss to followup" decreases the sample sizes (lowers statistical power) and may bias results -as in all observational studies, results are still affected by unmeasured confounding
question
Measures of association
answer
question: is the exposure associated with disease method: compare disease occurrence in exposed to the one in unexposed two approaches: -take the RATIO of the two numbers -take the DIFFERENCE of the two numbers
question
Ratios characterize RELATIVE difference HOW TO INTERPRET INCIDENCE FOR COHORT STUDIES
answer
in cohort studies typically involves: incidence (risk) of disease in exposed / incidence (risk) of disease in unexposed interpretation: the incidence of disease in exposed people is [ratio] times (greater/higher/lower) than in unexposed people. can be calculated using: 1. cumulative incidence (or risk) OR 2.incidence rate (or density) -person time in denominator
question
Incidence Summary
answer
Cumulative Incidence or risk = -easy to calculate and understand -often used in fixed population -can be less accurate if assumptions of completeness of followup not met Incidence Rate or Density -more accurate, esp in dynamic pops -often no feasible to calculate -less intuative
question
Putting Cohort Data into 2x2 table using cumulative incidence or risk
answer
see slide cumulative incidence or risk among E+ = a/(a+b) cumulative incidence among E- = c/(c+d)
question
Calculations: Risk Ratio
answer
Risk Ratio = R(E+)/ R(E-)= a/(a+b) / c/(c+d)
question
INTERPRETING RISK RATIOS
answer
R(E+) = R(E-) RR= 1 Exposure NOT associated with disease RR; 1 Exposure associated with greater risk of disease (may be a risk factor/cause of disease) RR;1 Exposure associated with lower risk of disease (may protect against disease) INTERPRETATION: (People with exposure) Smokers were 1.61 times (RR) More Likely (more, less) to develop CHD (disease) than nonsmokers (unexposed) over the study period
question
Incidence Rate (Incidence Density)
answer
denominator = person time contributed while in the study and AT RISK for disease a study participant is no longer considered AT RISK after they: -get the disease being studied -are lost to followup or drop out of study -die of some other cause Person Time each year a subject is followed is a "person year" PY -sum person-years of follow-up...5+10+15+10. =40 PY how many lung cancer cases per person year of followup 4/40 = .25 per 1 person year or .25 x 100 = 25 per 100 person years
question
Incidence Rate or Density
answer
Typically measured over a period of time -lung cancer study measured an incidence rate of 40 per 100,000 person years in a 25 year time frame -if study cannot capture date when disease occurs, then assuming the midpoint btwn two measurements in the time period is appropriate
question
Risk Ratio (RR) vs Rate Ratio (also RR)
answer
-when incidence is calculated using person-time, then the relative measure of association is -Rate Ratio OR -Incidence Rate Ratio (IRR) 2x2 table = dev disease persontime E+ 30 54,000 E- 60 51,000 Incidence Rate(IR) among E+ = 30/54,000 = 5.5 per 10,000 person years Incidence Rate (IR) among E- = 60/51,000 = 11.2 per 10,000 person years Incidence rate ratio (IRR) = 5.5/11.2 = .47
question
INTERPRETTING INCIDENCE RATE RATIO
answer
Incidence Rate E+ = For every 10,000 PM women using estrogen therapy, approximately 5.5 developed CHD over the course of the study Incidence Rate E- = For every 10,000 PM women not using estrogen therapy, approximately 11 developed CHD over the course of the study Incidence Rate Ratio: Hormone users developed CHD at almost half of the rate of nonusers over the study
question
Differences characterize the ABSOLUTE difference
answer
in cohort studies, this typically involves: incidence in exposed-incidence in unexposed interpretation: amount or excess of disease in exposed group attributed to the exposure. can be calculated using -cumulative incidence (risk) OR -incidence rates (or density) assumes cause-effect relationship exisits between exposure and disease
question
INTERPRETING Risk Difference (attributable Risk)
answer
If R(E+) = R(E-), then RD = 0 NO difference if RD; 0: excess of disease risk is attributable to the exposure RD;0: exposure protects or lowers the risk of the disease -measure the excess of disease attributable to exposure among the exposed
question
Attributable Risk Percent (AR%)
answer
RD or AR measure actual amount or excess of disease attributable to the exposure AR% estimates the proportion of disease among exposed due to exposure (% that could be prevented by eliminating exposure) AR% = [((AR or RD) / R(E+)] x 100 will equal exact number of cases attributed to exposure
question
INTERPRETTING AR%
answer
-cumulative incidence- never eat fish (E+) apprxomaitely 5% of adults who never ate fish developed a stroke during the study cumulative incidence -eat fish almost daily: approximately 3% of adults who ate fish almost daily developed a stroke during the study the risk of stroke is 1.75 times (RR) greater among adults never eating fish compared with those eating fish almost daily on average, 40% of stroke cases among the exposed were attributed to never eating fish AR%=[RD/R(E+)] x 100 = (.02/.05) = .4 x 100 = 40%
question
When to use Absolute Measures of Association
answer
very important for public health planning and policy making can assess amount of disease incidence we can remove by targeting a specific exposure allows for allocation of scare resources (time energy money and poltical capital) to most impactful efforts
question
Risk of Exposure Risk of unexposure
answer
R(E+)=a/(a+b) x 100 R(R-)=c/(c+d) x 100 when you make a 2x2 table in SPSS and you add go to cells and select under percent- rows and total you will see the same risk % in the table
question
Risk Ratio
answer
RR= R(E+)/R(E-) interpretation: the risk of (disease), among those exposed to (exposure), is (risk ratio) (less or greater) than among those who were unexposed. The (exposure) appears to be (protective or ? ) against (disease) because the RR (, =) 1. Among those who were (exposed), there is a (lower or higher) risk of (disease)
question
What would you conclude about the association between exposure and disease
answer
the risk of (disease), among those exposed to (exposure), is (RR) times (less or more) than the risk of the unexposed. Becuase the RR is (less or more) than 1, using the (exposure) is associated with a (lower or higher) risk for (disease)
question
Risk Difference
answer
RD=R(E+) - R(E-)
question
Risk Difference INTERPRETATION
answer
People using the (exposure) have a (lower or higher) risk of (disease) of (risk difference). If the RD was greater than 0, we could see there is an excess risk of disease among the exposed. Since RD;0, we say that the exposure is preventative of the disease/there is a lower risk of disease among the exposed. Interpretation: Those using the (exposure) (REDUCE or INCREASE...reduce for negative sign for RD) their risk for (disease) by (Risk difference) compared with those not (using exposure).
question
Can you say definitively that exposure is not a cause of disease
answer
there is always a possibility that other causes of disease were distributed unevely in the exposed and nonexposed group, leading to incorrect results in our study (bias and confounding). Those who are concerned about exposure (wrinkles) may also be more health conscious in general and therefore have a lower risk of (disease) skin cancer (confounding*). Also, there is no information about whether those looking for skin cancer were blind to exposure status. If not, perhaps they looked harder and found more lesions in the exposed, thus underestimating the true protective effect (ascertainment bias).
question
Risk Difference ; interpretiation
answer
RD=R(E+) + R(E-) women using anti-wrinkle cream have an excess risk of dementia of .665. Since the RD;0 we can say there appears to be an excess risk of dementia among those using the cream.
question
Do you think exposure caused disease? what alternate explanations might there be for your findings?
answer
This analysis does not take into account: 1) temporal order! although this is a cohort study, the temporal order of the cream and dementia association is in question here because the participants were not screened for dementia at the baseline visit. Since your colleague used your cohort (originally assembled to test the association between cream and skin cancer) to test the association between cream and dementia WITHOUT screening for dementia at baseline, we do not know the timing of the onset of dementia. 1)possible loss to follow-up: 2) possible confounding, especially age; 3) Bias (which is always possibility in cohort studies)
question
Lecture 5
answer
Case Control Studies
question
Design of Cohort Studies
answer
Enroll Cohort (D-) -Exposed -Disease -No Disease -Unexposed -Disease -No Disease
question
Cohort Studies
answer
Measure of disease occurrence -incidence (risk or rate) Measure of Association -risk ratio (RR), rate ratio (IRR) (odds ratio, OR, also possible but not common) -risk difference (RD), rate difference (IRD)
question
Interpretation RR
answer
Risk of (disease) in (exposure people) = people who (exposure) are (number) times as likely or greater to experience (disease) as people who do not (exposure)
question
Interpretation Review
answer
RR ; IRR- the risk or rate of developing the disease in the exposed is (RR) times greater or as likely as in the unexposed, or exposed have (RR) times the risk of developing the disease as unexposed. RD or IRD- Excess or prevention of disease attributed to the exposure (Number or % of cases among the exposed)
question
Case Control Study
answer
select groups based on DISEASE status -cases-people with diseases (preferably incident cases) -controls-people without disease determine PAST EXPOSURE for both (questionnaire, medical records,etc) compare level of EXPOSURE among cases to controls
question
how is a case control study different than a retrospective cohort study?
answer
both use information on disease status in the present and look at exposure in the past BUT case control study selects who will be in the study by DISEASE status and estimates frequency of exposure retrospective cohort study selects who will be in the study by exposure status or by being part of a population cohort and estimates frequency of disease
question
Properties of a well conducted case control study
answer
efficient sampling selection of cases selection of controls measurement of exposures
question
Case Control Sampling
answer
case control studies oversample D+ and undersample D- esp rare disease
question
Case Control Study is efficient
answer
cohort study -most particpants do not get disease -need enough cases of disease to reliably estimate risk, but also increase number of non cases(ie controls) case control studies oversample those with disease and take fewer non-cases -exact sampling fraction is often not known -overall much smaller sample size(efficient, costs less, if done correctly will get same OR as cohort study)
question
Case Selection
answer
need a clear and operable case definition -preferably restrictive to minimize erroneous inclusion of "cases" without disease -may need to losen if #of cases is limited sources may include hospital databases disease registries, surveys, death certificates incident cases preferred over prevalent ones to assess predictors of disease development not duration
question
Control Selection
answer
purpose of control group is provide information on level and distribution of exposures in population giving rise to cases guilding principle: represent same source population from which cases come if a person in control group had the disease, would he or she end up being a study case? Must be selected INDEPENDENT of exposure status (to try and accurately represent exposure levels in source population)
question
Control Selection and Sampling
answer
selected from a population based source (telefphone directories) from clinical settings (hospitals) or a social network (friends of cases) -each has advantages and disadvantages in terms of cost, susceptibility to bias, and likelihood of a good response rate number of controls selected per case can be increased to icnrease power to detect associations (typically range from 1:1 to 4:1)
