Levels of Prevention: A Comprehensive Guide to Public Health Interventions
Levels of Prevention:
1
Primary
People healthy, time prevention to reduce risk factor(Healthy take Vitamin C); 2.
Secondary
People have disease, but early they don’t have symptoms, screen for early detect, improve outcome w/ early treatment (black marker test/gargling for sore throat); 3.
Tertiary
Have disease, symptomatic, limit disability, delay progression, improve quality of life (ex.Surgery/Paxlovid for COVID19)
Populations:1.Sample:
recruit in study, 2.
Source
:people eligible to be in study, 3.
Base
:poeple develop the disease in the study(cases), 4.
Target
:how far we can push, broad(ex. California, the U.S.)
Incidence:
# of NEW cases of disease during period of interest
Incidence Rates
#New cases/Total p/t of observation, exclude people have already develop disease or not capable of develop it
Cumulative Incidence Rates:
New cases at specified time period/population @risk at beginning of time period, Q type:
What is the probability will develop/survive breast cancer?
Incidence Density Rates (IDR):
IDR(measure of the instantaneous force of disease occurrence)=# of New disease events/Total person-time at risk(denominator is person/years) Q type:
What is the average rate during a window of time(average month, average annual..)
Person/time denominator Used as the “population at risk” when individuals at risk for event are under observation for different lengths of time due to different time of accrual pr attrition(dying from other causes, moving away, lost followup)
Method 1:
Person-time=Sum different time periods that each individual was known to be at risk(Ex.60sub, each follow 4yrs=240p/y) Ex. 300 people tested, 30 tuned to hypertensive, over 3 yrs 20 initially normo-tensive: CI=20/300-30=74.1 per 1000 estimate annual incidence by dividing 3: 74.1/3=24.7 per 1000 people
Method 2:
Midpoint (estimate person-time but you not actually given follow-up time in study)
=multiply the total # of people at risk throughout the study period by 1/2 of the length of time of the study period (more accurate)= case # * half of time + survival #* full time Ex)begin w/ 300-30(previous turned)-3 yrs(20 incident each yr)=(250x3yrs)+(20 persons x 1.5 yrs)=780p-yrs, 20 incident/780=25.6 per 1000 per yr
Ex)
5000 adults in La Puente, CA, 600 were found to have diabetes, during the next 5 years, 150 more first examined developed diabetes: 150/(5000-600)=0.034, average rate=0.034/5 or [(4,400 – 150 new cases) x 5 yrs] + [150 x 2.5 yrs]=21625—150/21625
Mortality Rates:
Crude=# deaths/Population at risk, CFR=# of deaths due to a specified disease/Total # of cases of same disease (died from that disease), ProportionMR=# deaths due to a specified disease/all deaths in population(cause of all deaths)
Attack Rates:
# of new cases among exposed(or unexposed)/# at risk at start of time period among exposed (or unexposed) Special type of Cumulative Incidence(duration of the outbreak), ex. Food poisoning.
Attack Rate Ratio=Attack rate in exposed/Attack Rate in unexpose
Prevalence:(had):
P=I*Duration=Total # of cases/Total population @ risk (# of existing cases of disease at a point in time or during a period of interest (NEW & OLD))
Age-specific Rates:
# of events in age group/total individual in the age group (10.5 out of every)
Sex-specific Rates:
# of events in gender group/total individual of that gender group (Ex.)
Group of 1000 college students in Southern California screened for smoking status, 500 males and 500 female,145 males report and 95 female (prevalence), 20 male and 12 female became(incidence), Prevalence=145+95/1000=0.24=24%, Sex-specific prevalence Male=145/500=29% Female=95/500=19%
Race-specific Rates:
# of events in race group/total individual in the race group
Cause-specific Rates:
#deaths from specific cause/total population (as a result)
AGE
Age-specific Rates=# of events in the age group/ total # of people in same group
Indirect & Direct Age-adjustment
Direct (NOT a True Rate):
allow to compare summary rates b/t populations that are weighted w/ the same # or percent of people in each age group;
Indirect (RATIO, SMR orSIR)
:SMR=observed # of deaths(cases)/age-adjusted expect # of deaths*100
Calculation for Direct Age-adjustment
: Choose a standard population; and calculate EXPECTED in each age strata=Age-specofoc Rates of study population (Rate) x # of age-strata specific persons in standard population (Weight); Divide total Expected events by total standard population. If 15,000 standard population, and Town A have 0.712 rates and Town B have 2.167 rates, the expected in town A=15000*0.712; town B=15000*2.167, add all rates * population up and divide each to standard population
Survival Analysis and Randomized Trials:
Censoring:
anything other than the event of interest, who didn’t develop the disease of interest during the follow-up period, Censor at the last time we have the information (withdrawal, loss to follow-up, end of study period, incomplete data, non-response) Reasons: Lost to follow-up–lose to track and don’t know of they’re alive or dead (label censored at the last time we observed them to be alive and free of the disease), subjects completes the entire follow-up period under study and never develop the disease of interest, Subjects died of a competing cause of death (die from the disease not interested/studied)
Major Purpose of Key Trial:
1. Randomization
: participants have equal chances being assigned to any given treatment group. Eliminate selection bias(control confounding that you would/would not know to control for), balance the known and unknown prognostic factors among groups to help ensure the groups are comparable of the study.
Intent to Treat Analysis(ITT):
including all participants regardless whether they fully follow the trial protocol; Not using: biased results, because it excludes data for participants who may drop or not fully follow from protocol which might give exaggerated of treatment’s effectiveness.
2. Double-blind
:None participants would know who is receiving treatment or placebo. Minimize bias in assessment of the treatments’ efficacy and side effects, both participants’ and researchers’ expectations would not influence O.
3. Placebo
:control for the placebo effect, where participants experience changes in their condition based on their belief of receiving treatment. Allow researchers to isolate the effect of the study treatment from psychological effects (Phase III: Large scale comparison of effectiveness and safety against standard treatment (or placebo)).
P-value:
>0.05, null is true, we conclude that our data consist with null, we can conclude that any difference between the observed RR and 1.0 could easily have occurred by chance alone.
Study Design Guide:
Conceptual Hypothesis:
Idea proposed by a researcher to explain the occurrence of disease(O) with respect to an exposure(E).
Operational Hypothesis:
The testable prediction that is derived from the conceptual hypothesis.
1
Descriptive Design
No hypothesis in the beginning, hypothesis generating.
2. Observational: Examine association w/p intervention
Cohort:
Begin 2 population, ascertain E, wait, ascertain Outcome
–
Prospective
At start of study, E: may or may not have occurred, O: Definitely NOT happened
–
Retrospective (at the time study design, everything has already happened): Select after exposure and outcomes have occurred, Assembled from past records ex) Exposure of rocket fuel, people have job before 30 yrs
Calculate: Risk Ratio(RR)=[(A/(A+B)/C/(C+D))]
Or Rate Ratio=(A/Person-time in exposed)/(C/Person-time in unexpo)
Case-control:
Begin w/ Outcome, ascertain E in past
Calculate:
Odds Ratio(OR)
=AC/BD
Matching: in CC study, control for confounding variables, allow for more directly compare cases and controls, by ensuring both groups are similar w/ certain variables (age, sex..) 1.
Individual Matching
:each case matched individually to one or more controls based on specific characteristics (age, gender). Close control of confounding but restrictive and costly; 2.
Frequency Matching
Controls are selected with certain characteristics matches the case with those same characteristics. Less restrictive and flexible
Cross-sectional(Prevalence Studies):
Begin w/ population-ascertain E & O
Calculate Odd Ratio(better)(when O measure is incidence/prevalence)=AD/BC, if odds ratio=2.2, People who with parasites were 2.2 times as likely to be thin than people who didn’t have parasites
Ecological(Correlation Study):
Know the overall group, not know individual (group-level data, no individual)
Ecological Fallacy:
Made incorrect conclusions about individual-level relationships based on group-level data.
ex. If data show countries with higher chocolate consumption with have higher # of heart disease, might conclude eating chocolate contribute to heart disease, ignoring individual data that could suggest other confounding variables.
Clinical Trial:
Calculate:
Rate Ratio=(A/Person-time in exposed)/(C/Person-time in unexpo)=[(A/(A+B)/C/(C+D))]
Experimental
: Assign people to take drug or not
Relative & Attributable Risk Guide:
Absolute Risk:
The probability that an individual will develop a disease or other health condition during a specified time, conditioned on that individual’s not dying from any other condition during the period. E.G 3 case per 100,000 people (AR=0.003%) AR is tiny
Relative Risk:
how different the risk is b/t groups; RR=Risk in exposed population/Risk in non-exposed or reference group
Best estimate of the relative risk: a ratio of two comparable event rates=IDRsmokers/IDR non-smokers
Higher values of RR, farther from the null indicate a stronger association.(both +/-)
95% Confidence Intervals:
– 95% CI excludes 1.0: Reject the null hypothesis:
At the alpha=0.05, our data is not consistent w/ a true RR of 1.0. Reject the null hypothesis; Conclude: the observed RR is statistically significant different from 1.0 at the 0.05 level
-95% CI includes 1.0 (Null): NOT reject the null hypothesis:
Not certain that the true RR is different from 1.0; A RR of 1.0 is one of the values consistent with our data; Conclude: the observed RR is not statistically significant from 1.0 at the 0.05 level.
Attributable Risk (AR): 1. Attributable Risk Difference (Rate difference)
AR=Rate(exposed)-Rate (unexposed);
2. Attributable Risk Percent (AR%) (proportion):
AR%=Rate(exposed)-Rate (unexposed)/Rate(exposed) * 100%=1-1/RelativeR=RR(or Odds Ratio)-1/RR;
Conclusion: 1. In this study, XX(52) lung cancers per year per 100,000 smokers were attributable to smoking. 2. In this study, 89% of the lung cancer among smokers were attributable to their smoking.
Population Attributable Risk (PAR): 1. Population Attributable Risk Difference (Rate difference)PAR=Rate(population) – Rate (unexposed)
2. Population Attributable Risk Percent (PAR%):
PAR%=Rate(population) – Rate (unexposed)/Rate(population) * 100% =Pe(RR-1)/[Pe(RR-1)]+1 Pe=proportion of population exposed.
Conclusion: 1. According to this data, roughly 17 lung cancers per year per 100,000 persons in this entire population were attributable to smoking. 2. In this study, almost 71% of lung cancers in the entire population were attributable to smoking. 3. If smoking were eliminated, almost 71% of the lung cancers in this population could have been prevented.
Number Needed to Treat (NNT)
: # of individuals that would need to be treated to prevent a single case
1. Preventable Risk(PR) or Absolute Risk Reduction (ARR)
=NNT=1/PR or 1/ARR; PR: based on this data, roughly 455 inactive persons would need to have exercised regularly for one year to prevent one CHD death. NNT=1/AR; AR: based on this data, roughly 1786 smokers would need to have been non-smokers to prevent one lung cancer per year
When change in prevalence of an exposure in a population:
Relative Risk not change, it’s a ratio of risks and remains constant as long as relationship b/t exposure and Outcome doesn’t change; ARP not change because it depends on the relative risk. If exposure becomes more common, PAR% increases indicates larger portion of disease burden, if exposure less common, PAR%decrease.
Causality, Bias, Confounding, and Interactions Guide:
Natural of Associations in Studies
1.
Causal:
Exposure directly causes the disease. 2.
Non-Causal
: Association seen due to the behavior or exposure changing after the disease occurred. 3.
Due to chance, confounding, or bias
: random variation, confounding variables related to E and O, or systematic errors in study design or data collection.
Hill’s Criteria for Causation:
Used to assess if an association is likely yo be causal
1.
Strength of Association
: Strong associations are more likely to be causal. 2.
Consistency
: Observed in different studies and populations. 3.
Specificity
: Association is constrained to specific E and O. 4.
Temporality
: The exposure prior to the outcome. 5.
Biological Gradient
: Increase E increases risk. 6.
Plausibility
: Biologically or socially plausible. 7.
Coherence
: Fits with current knowledge. 8.
Experiment
: Causation more likely if evidence is based on Randomized Experiments. 9.
Analogy
: Similar associations can suggest causality.
Misclassification Types:
1.
Differential
: Error varies b/t study groups. 2.
Non-differential
: Error is uniform across study groups.
Impact:
Differential bias results more significantly than non-differential misclassification.
Biases and Issues:
1. Self-selection Bias: Occurs when individuals select themselves into study group, leading to non-random study groups. 2. Survival Bias (selection bias): Occurs when observations are made only on surviving subjects. 3. Temporal Relationship Problems: Incorrect assumptions about the timing of E and O. 4. Overmatching: Matching on variables that are not true confounders, which can reduce variability and obscure true associations. 5. Regression to the Mean: Extreme values tend to be closer to the average on subsequent measures. (Use Randomized controlled Trial)
Confounding:
Ex) age confounds on the association b/t place residence and risk of death; Smoking confounds the association b/t Coffee drinking and heart disease
1)Covariate is the risk factor for the disease (must meet); 2) Exposure and Covariate are associated (tend to occur together)people smoke tend to be more likely to drink coffee; 3) Covariate is not an intermediate b/t E and D (E not cause C) High fat diet is exposure, high obesity is intermediate, cardiovascular disease is outcome
Detection Confounding in data:
1) Modeling-crude and adjusted differ; 2) Stratum specific estimates differ from crude or collapsed estimate like OR crude=2.0 OR age-adjusted=1.2. (After you adjusted the confounder the Odds ratio changes, means you have evidence of confounding in your data)
Direction of Confounding:
RRT (control confounder estimate): True Relative Risk, RR^: Sample estimate of Relative Risk
Controlled 是RRT, RR^没有control的Crude OR (both positive bias)
Positive Bias=E goes up so does C, make the effect estimate smaller than the true effect Ex)
E: alcohol consumption, O: Heart Disease, Confounder: Smoking; smoking and alcohol contribute to increase risk of HD.
Direction:
study finds weaker association b/t alcohol consumption and HD than truly exists, smoking might be positive confounder, because it increase risk of HD in exposed (drinkers) and unexposed (non-drinker) groups to mask the true effect of alcohol.
Negative Bias:
make effect estimate larger than true effect Ex)
E:Regular Exercise, O: Wight loss, Confounder: Diet Quality. Good diet quality associated with more exercise and greater weight loss. If study finds stronger association b/t exercise and wight loss than actually exists, diet quality might be negative confounder, because it independently promotes weight loss and associated w/ more exercise, exaggerating the effect of exercise on weight loss.
Control Confounding:
In Design:
Randomization (Big enough sample size, random assignment of subjects to exposure groups (experiments), balance confounders), Restriction (subjects to one level of a potentially confounding factor (can be used in any design) ex. Coffee drinking on lung disease (restrict smoking, non-smokers)), Matching (on potentially confounding factors (most effectively in case-control));
In Analysis:
Stratification:
Measure the association within each strata of the potentially confounding variable Eg. Measure on smokers vs. Non-smokers, Stratification usually used only 1-2 potential confounders
Crude and stratum OR are different
Statistical Modeling (multivariable statistical methods): Adjusting for potentially confounding factors by statistical analysis or modeling
E.G. OR crude=2.0 OR age-adjusted=1.2 (Figure above)
Effect Modification and Interaction: Effect Modification (population Interaction)
True effect of one exposure on disease outcome caries across level of second exposure, True difference in the strength of association b/t E and O across strata (Levels) of a second factor;
This is NOT a bias, This is a difference in effect depending on the levels of a second factor you want to measure and interpret E.G. Effect of therapy on kids vs. Adults. Positive enhance effects, negative diminish effects
Joint effect of 2 variables on O is much larger or smaller than expected based on their independent effects (Have E 1 or 2 on same person will cause different risk that either larger or smaller than people only have E1 or only have E2) Ex. Someone have lung cancer or someone drinking alcohol and have other people only eat fatty food, people have drinking alcohol and also eat fatty food
Additive Model:The combined effect is the sum of individual effects.–
E1 and E2 acting on your body are same; – When effect of E2 adds to the effect of E1 (Att Risk difference);
– E1 and E2 are essentially an increased exposure to the same thing; Ex. Smoking and asbestos cause same irritation and damage to your lung, even they are not the identical chemical
Multiplicative Model: The combined effect is the product of individual effects. –
one cell cause damage, and one cell can faster the division of cells; When E2 multiplies the effect of E1 (Ratio measure)–
Biological mechanisms through which they influence risk is sufficiently different such that E1 and E2 have greater effects than predicted by simply adding more of same exp.
Homogeneity or heterogeneity of effects
The effect of E1 on O is not same in strata formed by E2 (E2= association modifier)
Comparison b/ observed and expected joint effects of E1 and E2:
interaction present when observed joint effect of E1 and E2 differs from the expected joint effect based on their independent effects.
Assessing interaction based on evaluating joint effects:
RR00: completely unexposed / RR10: people only exposed to E2 / RR01: people only exposed to E1 / RR11: joint exposed
Effect of E1 varies by presence of E2 (interaction)
If joint of E1 and E2 no interaction, then the effect of E1 is the same for people exposed to E2, as for people not exposed to E2. And the effect of E2 is the same for people with or w/o E1
Validity and Reliability of Tests:
Screening tests:
identifying early signs of disease in people w/ no signs or symptoms of the disease
Test Parameters (indices of Validity)
: Involves comparing the results of the test being evaluated with those derived from a definitive diagnostic procedure (gold standard)
False Positives:
Test indicates a person has the disease when they don’t.
False Negatives:
test indicates a person doesn’t have the disease when they do
=proportion of persons w/ disease who test positive / =Probability the test will be positive if truly have the disease / =Test’s ability to identify correctly individuals who truly have the disease
=proportion of persons w/ disease who test negative / =Probability the test will be negative if truly do not have the disease / =Test’s ability to identify correctly individuals who truly do not have the disease
: =proportion of persons testing positive who truly have the disease
: =proportion of persons testing negative who truly do not have the disease
Cut Points:
Sensitivity and specificity depend on cut points
Ex.Blood pressure for hypertension when exceed specific cut points. As the cut points is lowered, more diseased individuals will test positive (assuming high value present disease), so the sensitivity will increase, but more non-diseased individuals will also test positive, so specificity will decrease.
Higher prevalence of disease, higher PVP, b/c more true positive to false positives; lower PVN, b/c fewer true negatives to false negatives
3Prevalence of disease and 4 Test Parameters:
1. Sensitivity and specificity are independent of the prevalence of undetected disease in the population being tested, they Depend on : cut point, extent of overlap 2. Predictive values depend heavily on prevalence of undetected disease in the population being screened: Higher prevalence of disease, higher PVP, b/c more true positive (A) to false positives; lower PVN, b/c fewer true negatives to false negatives (C) As the prevalence of the target (undetected) disease increases (A & C increase), the PVP will increase (larger in A cell, A/(A+B)). As prevalence increases, the PVN will decrease slightly, the C cell will also increase (D/C+D)
Two Stage Testing: 1) Net sensitivity: Sensitivity (Test 1) * Sensitivity (Test 2); 2) Net specificity: Specificity (Test 1) + Specificity (Test 2) [1-Specificity (Test 1)]
Evaluation of Screening Programs: Randomized studies, best and most valid evidence of efficacy of screening program comes from randomized trials w/ cause- and age-specific mortality as the O; Case-control: Individuals w/ and w/o the disease are compared for past screening. Cohort studies: case-fatality or 5-year survival rate
When high Specificity is important (FP is important): If do really expensive and difficult screening, make sure specificity is higher. When False positive lead to harmful follow-up testing, e.G. Biopsy; If do repeated screens on people, so you can only follow up tests for people have the disease in cancer screening, when treatment has sever side effects, expensive and unpleasant (0 False Positive)
When high Sensitivity is important (FN is important): Screening for serious diseases where missing a case (false negative) could be dangerous (e.G., cancer screening, HIV screen) also when disease is fatal; also for early stages of disease screening where early treatment can significantly improves outcomes.
Scenarios for Maximizing PVP: When specificity of the test is high and prevalence of the disease is high. B/c fewer False Positive (B) occur relative to True Positive (D), improve the likelihood of a positive result is a True Positive
Reliability of Tests: Validity=measure of accuracy in relation to the “truth”. Reliability=Repeatability
Evaluation of Screening Programs Guide:
Could observe benefit from screening and early detection if, all or most clinical cases of the disease go through a detectable preclinical phase, and in absence of intervention, most cases in preclinical phase progress to a clinical phase. However, some cases progress too rapidly through detectable preclinical phase to actually be detected by screening or some cases of preclinical disease never progress to clinical disease
Feasibility: =PVP can be increased by 1) increase prevalence of detectable preclinical disease in the population, e.G. By screening only individuals who are at high risk for developing the diseases on the basis of age, medical history, etc. 2) increasing the specificity of the screening test
Ramifications: Reducing # of FP often leads to fewer individuals identified by the test, which can be beneficial in reducing unnecessary follow-up treatments but may also miss some true cases if specificity is not perfectly high.
ALWAYS Wrong to compare Outcomes from a group of screen-detected cases w/ those from a group of clinically-detected cases
Source of Bias in Evaluating Screening Effectiveness:
1) Lead Time Bias:Tendency for cases detected by screening to appear to live longer than cases detected clinically; “Lead Time”—time by which screening test advances the date of diagnosis from the usual symptomatic phase to an earlier pre-symptomatic phase. (Occurs when screening detects a disease earlier than it would have been detected because of symptoms, making it seem like survival time has increased when it has not actually affected the course of the disease.)
2) Length Bias: Tendency for cases w/ a long detectable preclinical phase of the disease to be overrepresented among screen-detected cases compared to symptom-detected cases. E.G. Cases of rapidly progression (very aggressive) disease usually pass through the detectable preclinical stage too quickly to be picked up by most screening programs. (Screening is more likely to detect slower-progressing diseases because they remain detectable for longer periods, potentially skewing effectiveness data since these cases often have better prognoses.)
3) Self-Selection Bias: tendency for persons who choose to participate in screening programs to be different from those who don’t participate in a number of ways that affect survival. – Those participate in screening programs are generally healthier and have other positive health behaviors, e.G. Don’t smoke, not overweight. – Screened are more likely follow medical regimens that may be after a positive screening test. (Individuals who choose to participate in screening programs might differ in health awareness or risk factors from those who do not, affecting the apparent efficacy of the screening.)
4) Overdiagnosis: tendency for some small proportion of non-diseased individuals to be falsely diagnosed as having the disease, so screen-detected cases are diluted w/ individuals who don’t in fact they have the disease. This bias can be limited by rigorously standardizing the procedures leading to a positive diagnosis. (Screening may identify diseases that would not have caused symptoms or death (especially in cancer screenings), leading to unnecessary treatment for such diseases.)
Relationship between sensitivity, specificity: Increasing sensitivity generally decreases specificity and vice versa, high sensitivity is important to identify possible cases to prevent spreading, but may increase FP; High specificity reduce FP but may increase FN.
ACUTE DISEASE EPIDEMIOLOGY GUIDE:
Incubation period:
First exposure to onset of clinical symptoms. It varies by disease and is critical for understanding disease dynamics, determining quarantine measures, and estimating the time of exposure. Incubation
Median incubation: After the first exposure, will be the peak of the curve
Types of Epidemics:
1)Common Source Epidemic:
Involves group of persons get sick after being exposed to a common influence like water supply or food, the E can be continuous or intermittent. Graph: Plateau-like pattern or prolonged tail
2)Point Source Epidemic (population exposed at one point in time):
Persons are exposed over a brief time to same source, such as single meal or event, the # of cases rises rapidly to the peak and falls gradually. (e.G. Food poisoning, gas lead, nuclear accident) Graph: The # of cases rise abruptly and fall again in a log-linear fashion.
3)Person-to-person:
Results from transmission from one person to another. Typically, # of cases increases more gradually and may involve multiple waves of infection. Graph: Progressively taller peaks, indicating waves of infections
CyclicTrend: Secular Trend: (long trends over decades/centuries in one direction)
Steps in Outbreak investigation:
1)Prepare for Field Work: Gather necessary tools and supplies, ensure logistical arrangements.
2) Establish the Existence of an Outbreak: Verify the diagnosis, confirm that the reported cases exceed the expected number. 3)Verify the Diagnosis: Ensure that cases are correctly diagnosed.
4)Construct a Working Case Definition: Define what constitutes a case for the investigation. 5)Find Cases Systematically and Record Information: Active surveillance to identify and document cases. 6)Perform Descriptive Epidemiology:
Describe in terms of time, place, and person. 7)Develop Hypotheses: Based on the integration of all data collected. 8)Evaluate Hypotheses: Implement analytical methods, e.G., case-control studies. 9)Refine Hypotheses and Carry Out Additional Studies: As necessary, adjust hypotheses and conduct further investigation. 10)Implement Control and Prevention Measures: Based on the findings, suggest and facilitate control measures. 11)Communicate Findings: Inform local health authorities, stakeholders, and possibly the public.
Completion of an Outbreak Investigation: 1) Conform all cases identified and documented; 2) Ensure hypothesis has been tested properly and completely; 3) Implement and verify the effectiveness of the control measures; 4) Prepare final report with suggestions; 5) Follow-up ensure measures are sustained
Calculation Attack Rates: Attack Rates=confirmed cases/total number of people at risk * 100
Basic Concepts, Rates, and Age-Adjustments Guide:
Levels of Prevention:
1.Primary: People healthy, time prevention to reduce risk factor(Healthy take Vitamin C)
2.Secondary:
people have disease, but early they don’t have symptoms, screen for early detect, improve outcome w/ early treatment (black marker test/gargling for sore throat);
3.Tertiary:
have disease, symptomatic, limit disability, delay progression, improve quality of life (ex.Surgery/Paxlovid for COVID19)
Populations:
1.Sample: Recruit in study,
2. Source: People eligible to be in study,
3. Base:
people develop the disease in the study(cases), 4.Target:how far we can push, broad(ex. California, the U.S.)
Rates Formulars:
Incidence:
# of NEW cases of disease during period of interest
Incidence Rates:
#New cases/Total p/t of observation, exclude people have already develop disease or not capable of develop it
Cumulative Incidence Rates:
New cases at specified time period/population @risk at beginning of time period;
Q type:
What is the probability will develop/survive breast cancer?
Incidence Density Rates (IDR):
IDR(measure of the instantaneous force of disease occurrence)=# of New disease events/Total person-time at risk(denominator is person/years) Q type:
What is the average rate during a window of time(average month, average annual..)
Person/time denominator Used as the “population at risk” when individuals at risk for event are under observation for different lengths of time due to different time of accrual pr attrition(dying from other causes, moving away, lost followup)
Method 1:
Person-time=Sum different time periods that each individual was known to be at risk(Ex.60sub, each follow 4yrs=240p/y) Ex. 300 people tested, 30 tuned to hypertensive, over 3 yrs 20 initially normo-tensive: CI=20/300-30=74.1 per 1000 estimate annual incidence by dividing 3: 74.1/3=24.7 per 1000 people
Method 2:
Midpoint (estimate person-time but you not actually given follow-up time in study)
=multiply the total # of people at risk throughout the study period by 1/2 of the length of time of the study period (more accurate)= case # * half of time + survival #* full time Ex)begin w/ 300-30(previous turned)-3 yrs(20 incident each yr)=(250x3yrs)+(20 persons x 1.5 yrs)=780p-yrs, 20 incident/780=25.6 per 1000 per yr
Ex)5000 adults in La Puente, CA, 600 were found to have diabetes, during the next 5 years, 150 more first examined developed diabetes: 150/(5000-600)=0.034, average rate=0.034/5 or [(4,400 – 150 new cases) x 5 yrs] + [150 x 2.5 yrs]=21625—150/21625
Mortality Rates
Crude=# deaths/Population at risk
CFR=# of deaths due to a specified disease/Total # of cases of same disease (died from that disease)
ProportionMR=# deaths due to a specified disease/all deaths in population(cause of all deaths)
Attack Rates:
# of new cases among exposed(or unexposed)/# at risk at start of time period among exposed (or unexposed) Special type of Cumulative Incidence(duration of the outbreak)
ex. Food poisoning
Attack Rate Ratio=Attack rate in exposed/Attack Rate in unexpose
Prevalence (had)
P=I*Duration=Total # of cases/Total population @ risk (# of existing cases of disease at a point in time or during a period of interest (NEW & OLD))
Age-specific Rates:
# of events in age group/total individual in the age group (10.5 out of every)
Sex-specific Rates:
# of events in gender group/total individual of that gender group
Ex) Group of 1000 college students in Southern California screened for smoking status, 500 males and 500 female,145 males report and 95 female (prevalence), 20 male and 12 female became(incidence), Prevalence=145+95/1000=0.24=24%, Sex-specific prevalence Male=145/500=29% Female=95/500=19%
Race-specific Rates:
# of events in race group/total individual in the race group
Cause-specific Rates:
#deaths from specific cause/total population (as a result)
AGE:
Age-specific Rates=# of events in the age group/ total # of people in same group
Indirect & Direct Age-adjustment:
Direct (NOT a True Rate)
Allow to compare summary rates b/t populations that are weighted w/ the same # or percent of people in each age group;
Indirect (RATIO, SMR orSIR)
SMR=observed # of deaths(cases)/age-adjusted expect # of deaths*100
Calculation for Direct Age-adjustment:
Choose a standard population; and calculate EXPECTED in each age strata=Age-specific Rates of study population (Rate) x # of age-strata specific persons in standard population (Weight); Divide total Expected events by total standard population. If 15,000 standard population, and Town A have 0.712 rates and Town B have 2.167 rates, the expected in town A=15000*0.712; town B=15000*2.167, add all rates * population up and divide each to standard population
Survival Analysis and Randomized Trials:
Censoring:
anything other than the event of interest, who didn’t develop the disease of interest during the follow-up period, Censor at the last time we have the information (withdrawal, loss to follow-up, end of study period, incomplete data, non-response)
Reasons: Lost to follow-up–lose to track and don’t know of they’re alive or dead (label censored at the last time we observed them to be alive and free of the disease), subjects completes the entire follow-up period under study and never develop the disease of interest,
Subjects died of a competing cause of death (die from the disease not interested/studied)
Major Purpose of Key Trial:
1. Randomization:
participants have equal chances being assigned to any given treatment group. Eliminate selection bias(control confounding that you would/would not know to control for), balance the known and unknown prognostic factors among groups to help ensure the groups are comparable to the study.
Intent to Treat Analysis(ITT):
including all participants regardless whether they fully follow the trial protocol; Not using: biased results, because it excludes data for participants who may drop or not fully follow from protocol which might exaggerate treatment’s effectiveness.
2. Double-blind:
None participants would know who is receiving treatment or placebo. Minimize bias in assessment of the treatments’ efficacy and side effects, both participants’ and researchers’ expectations would not influence O.
3. Placebo:
control for the placebo effect, where participants experience changes in their condition based on their belief of receiving treatment. Allow researchers to isolate the effect of the study treatment from psychological effects (Phase III: Large scale comparison of effectiveness and safety against standard treatment (or placebo)).
P-value:
>0.05, null is true, we conclude that our data consist with null, we can conclude that any difference between the observed RR and 1.0 could easily have occurred by chance alone.
Study Design Guide:
Conceptual Hypothesis:
Idea proposed by a researcher to explain the occurrence of disease(O) with respect to an exposure(E).
Operational Hypothesis:
The testable prediction that is derived from the conceptual hypothesis.
1. Descriptive Design:
No hypothesis in the beginning, hypothesis generating. EX)By describing a small cases series, the authors are not testing a hypothesis, rather they are describing the …..
2. Observational: Examine association w/p intervention
Cohort:
Begin 2 population, ascertain E, wait, ascertain Outcome
– Prospective:
At start of study, E: may or may not have occurred, O: Definitely NOT happened Ex)The exposure to alcohol assessed at the enrollment through self-administered baseline questionnaire, and the participants were followed forward in time for O incidence. It’s…B/c participants were followed forward in “real time” not through previous calendar time.
-Retrospective (at the time study design, everything has already happened): Select after exposure and outcomes have occurred, Assembled from past records ex) Exposure of rocket fuel, people have job before 30 yrs/ Ex)subjects were selected for E status and then followed forward (retrospectively) in time for incidence of the outcome. This is.. B/c the cohort has been followed through past calendar time for episodes of the O and using previously collected data.
Calculate:
Risk Ratio(RR)=[(A/(A+B)/C/(C+D))]
Or Rate Ratio=(A/Person-time in exposed)/(C/Person-time in unexpo)
Case-control:
Begin w/ Outcome, ascertain E in past.
Ex) Nested case-control study:
subjects were selected from participants in a cohort based on outcome status. The E information for both case and controls was obtained from the cohort study baseline questionnaire and E frequencies were compared statistically to look for association b/t E and O.
Ex) Case-control:
Subjects were selected based on Outcome status. The E for both cases and controls was obtained from….And E frequencies were compared statistically to look for association b/t E and O.
Calculate:
Odds Ratio(OR)
=AC/BD
Matching: in CC study, control for confounding variables, allow for more directly compare cases and controls, by ensuring both groups are similar w/ certain variables (age, sex..)
1. Individual Matching:
each case matched individually to one or more controls based on specific characteristics (age, gender). Close control of confounding but restrictive and costly
2. Frequency Matching:
Controls are selected with certain characteristics matches the case with those same characteristics. Less restrictive and flexible
Cross-sectional(Prevalence Studies)
Begin w/ population-ascertain E & O, Ex) The subjects were selected b/c they are employees at the specific companies, not b/c the E or O status. In addition, it’s cross-sectional study b/c both the E & O were measured at same time with same instruments.
Calculate Odd Ratio(better)(when O measure is incidence/prevalence)=AD/BC, if odds ratio=2.2, People who with parasites were 2.2 times as likely to be thin than people who didn’t have parasites
Ecological(Correlation Study):
Know the overall group, not know individual. Ex)the data for both exposure of interest an outcome of interest represented country level information. Measures on both E & O were not collected on same individuals, but rather statistics from the larger entities as a whole were compared to assess the correlation of interest.
Ecological Fallacy:
Made incorrect conclusions about individual-level relationships based on group-level data. Ex. If data show countries with higher chocolate consumption with have higher # of heart disease, might conclude eating chocolate contribute to heart disease, ignoring individual data that could suggest other confounding variables.
Clinical Trial:
Calculate:Rate Ratio=(A/Person-time in exposed)/(C/Person-time in unexpo)=[(A/(A+B)/C/(C+D))]
Experimental:
Assign people to take drugs or not. Ex) since subjects were randomly assigned to one of two exposure groups (i.E. Therapy or placebo) for the purpose of determining the differential effects of these exposure on disease outcome. Subjects were followed …Time period.
Relative & Attributable Risk Guide:
Absolute Risk:
The probability that an individual will develop a disease or other health condition during a specified time, conditioned on that individual’s not dying from any other condition during the period. E.G 3 case per 100,000 people (AR=0.003%) AR is tiny
Relative Risk:
how different the risk is b/t groups; RR=Risk in exposed population/Risk in non-exposed or reference group=Observed/Expected
Best estimate of the relative risk: a ratio of two comparable event rates=IDRexpose/IDR non-exposure
Higher values of RR, farther from the null indicate a stronger association.(both +/-) The proportion of expose will not affect RR
95% Confidence Intervals:
– 95% CI excludes 1.0:
Reject the null hypothesis:
At the alpha=0.05, our data is not consistent w/ a true RR of 1.0. Reject the null hypothesis; Conclude: the observed RR is statistically significant different from 1.0 at the 0.05 level
-95% CI includes 1.0 (Null):
NOT reject the null hypothesis:
Not certain that the true RR is different from 1.0; A RR of 1.0 is one of the values consistent with our data; Conclude: the observed RR is not statistically significant from 1.0 at the 0.05 level.
Attributable Risk (AR):
1. Attributable Risk Difference (Rate difference): AR=Rate(exposed)-Rate (unexposed)
2. Attributable Risk Percent (AR%) (proportion): AR%=Rate(exposed)-Rate (unexposed)/Rate(exposed) *100%
=1-1/RelativeR
=RR(or Odds Ratio)-1/RR
Ex)If given chart with exposed(100) population and unexposed(10), the AR=IR(exp)-IR(non-exp)=90 per…, AR%=90/100=90%
Conclusion:
1. In this study, XX(52) lung cancers per year per 100,000 smokers were attributable to smoking
2. In this study, 89% of the lung cancer among smokers were attributable to their smoking
Population Attributable Risk (PAR):
1. Population Attributable Risk Difference (Rate difference)PAR=Rate(population) – Rate (unexposed)
2. Population Attributable Risk Percent (PAR%) PAR%=Rate(population) – Rate (unexposed)/Rate(population) * 100%
=Pe(RR-1)/[Pe(RR-1)]+1, Pe=proportion of population exposed.
When proportion of population expose becomes smaller, PAR% become smaller
Ex)PAR=0.1(100)+0.9(10)=19 per.., PAR%= 1 step)19 per – 10 per=9 per, 2 step)9 per / 19 per=47%
Conclusion:
1. According to this data, roughly 17 lung cancers per year per 100,000 persons in this entire population were attributable to smoking.
2. In this study, almost 71% of lung cancers in the entire population were attributable to smoking
3. If smoking were eliminated, almost 71% of the lung cancers in this population could have been prevented
Number Needed to Treat (NNT):
# of individuals that would need to be treated to prevent a single case
1. Preventable Risk(PR) or Absolute Risk Reduction (ARR)
=NNT=1/PR or 1/ARR; PR: based on this data, roughly 455 inactive persons would need to have exercised regularly for one year to prevent one CHD death.
NNT=1/AR; AR: based on this data, roughly 1786 smokers would need to have been non-smokers to prevent one lung cancer per year
When change in prevalence of an exposure in a population:
Relative Risk not change, it’s a ratio of risks and remains constant as long as relationship b/t exposure and Outcome doesn’t change; ARP not change because it depends on the relative risk. If exposure becomes more common, PAR% increases indicates larger portion of disease burden, if exposure less common, PAR%decrease.
Causality, Bias, Confounding, and Interactions Guide:
Natural of Associations in Studies:
1. Causal: Exposure directly causes the disease
2. Non-Causal: Association seen due to the behavior or exposure changing after the disease occurred
3. Due to chance, confounding, or bias:
random variation, confounding variables related to E and O, or systematic errors in study design or data collection.
Hill’s Criteria for Causation:
Used to assess if an association is likely to be causal
1. Strength of Association: Strong associations are more likely to be causal
2. Consistency: Observed in different studies and populations
3. Specificity: One E leads to a very specific O.
4. Temporal Relationship: The exposure prior to the outcome
5. Dose-Response:
3 levels of exposure, as exposure changes, the outcome risk could change in same direction or reverse direction.
6. Biological Plausibility: Biologically or socially plausible with what you are saying and expect
7. Consideration of Alternate Explanations: Consider all of them and all seems to support your conclusion
8. Cessation of Exposure: When risk is high when you smoke, when quit smoking risk goes down
9. Consistency with Other Knowledge:
all the other pieces of evidence align with what you found and your conclusion.
Misclassification Types:
1. Differential:
Misclassification within one group is more likely than in the comparison group (can spuriously increase or decrease estimates of an observed association).
2. Non-differential:
Misclassification is equally likely in the study groups being compared (bias usually toward Null).
Impact: Differential bias results more significantly than non-differential misclassification
Biases and Issues:
1. Self-selection Bias:
Occurs when individuals select themselves into study group, leading to non-random study groups.
2. Survival Bias (selection bias):
Occurs when observations are made only on surviving subjects.
3. Temporal Relationship Problems:
Incorrect assumptions about the timing of E and O.
4. Overmatching:
Matching on variables that are not true confounders, which can reduce variability and obscure true associations.
5. Regression to the Mean:
Extreme values tend to be closer to the average on subsequent measures. (Use Randomized controlled Trial)
Confounding:
When an association is partly or completely created by a third factor.
Ex) age confounds on the association b/t place residence and risk of death; Smoking confounds the association b/t Coffee drinking and lung cancer
1)
Covariate is the risk factor for the disease (must meet); 2)
Exposure and Covariate are associated (tend to occur together)people smoke tend to be more likely to drink coffee; 3)
Covariate is not an intermediate b/t E and D (E not cause C) High fat diet is exposure, high obesity is intermediate, cardiovascular disease is outcome
Detection Confounding in data:
1) Modeling-
Crude and adjusted differ
2)
Stratum specific estimates differ from crude or collapsed estimate like OR crude=2.0 OR age-adjusted=1.2.
(After you adjusted the confounder the Odds ratio changes, means you have evidence of confounding in your data)
Direction of Confounding:
RRT (control confounder estimate): True Relative Risk, RR^: Sample estimate of Relative Risk
Controlled 是RRT, RR^没有control的Crude OR (both positive bias)
Positive Bias=E goes up so does C, make the effect estimate smaller than the true effect
Ex) E: alcohol consumption, O: Heart Disease, Confounder: Smoking; smoking and alcohol contribute to increase risk of HD. Direction: study finds weaker association b/t alcohol consumption and HD than truly exists, smoking might be positive confounder, because it increase risk of HD in exposed (drinkers) and unexposed (non-drinker) groups to mask the true effect of alcohol.
Negative Bias:
Make effect estimate larger than true effect
Ex) E:Regular Exercise, O: Wight loss, Confounder: Diet Quality. Good diet quality associated with more exercise and greater weight loss. If study finds stronger association b/t exercise and wight loss than actually exists, diet quality might be negative confounder, because it independently promotes weight loss and associated w/ more exercise, exaggerating the effect of exercise on weight loss.
Control Confounding:
In Design:
Randomization (Big enough sample size, random assignment of subjects to exposure groups (experiments), balance confounders), Restriction (subjects to one level of a potentially confounding factor (can be used in any design) ex. Coffee drinking on lung disease (restrict smoking, non-smokers)), Matching (on potentially confounding factors (most effectively in case-control));
In Analysis:
Stratification
Measure the association within each strata of the potentially confounding variable Eg. Measure on smokers vs. Non-smokers, Stratification usually used only 1-2 potential confounders
Statistical Modeling (multivariable statistical methods): Adjusting for potentially confounding factors by statistical analysis or modeling
E.G. OR crude=2.0 OR age-adjusted=1.2 (Figure above)
Validity and Reliability of Tests:
Diagnostic Tests:
used for persons in whom disease is suspected due to symptoms, physical findings, or laboratory findings.
Screening tests (asymptomatic people):
identifying early signs of disease in people w/ no signs or symptoms of the disease
Test Parameters (indices of Validity):
Involves comparing the results of the test being evaluated with those derived from a definitive diagnostic procedure (gold standard)
False Positives:
Test indicates a person has the disease when they don’t.
False Negatives:
test indicates a person doesn’t have the disease when they do
=proportion of persons testing positive who truly have the disease
=proportion of persons testing negative who truly do not have the disease
Cut Points:
Sensitivity and specificity depend on cut points
Ex.Blood pressure for hypertension when exceed specific cut points.
As the cut points are lowered, more diseased individuals will test positive (assuming high value present disease), so the sensitivity will increase, but more non-diseased individuals will also test positive, so specificity will decrease.
(When cutpoint higher the specificity higher which means False negative will be greater)
Ex)When give you a disease prevalence, and sensitivity and specificity, use the population * prevalence%=people who actually diagnose disease (A+C), use sensitivity % * (A+C)=A; use specificity % * (total population-(A+C))=D
Prevalence of disease and 4 Test Parameters:
1.
Sensitivity and specificity are independent of the prevalence of undetected disease in the population being tested, they Depend on :
Cut point, extent of overlap
2.
Predictive values depend heavily on prevalence of undetected disease in the population being screened:
Higher prevalence of disease, higher PVP, b/c more true positive (A)
to false positives;
lower PVN slightly, b/c fewer true negatives to false negatives (C) As the prevalence of the target (undetected) disease increases (A & C increase), the PVP will increase (larger in A cell, A/(A+B)).
As prevalence increases, the PVN will decrease slightly, the C cell will also increase (D/C+D)
Higher prevalence of disease, higher PVP, b/c more true positive to false positives; lower PVN, b/c fewer true negatives to false negatives
Increase PVP:
1)Test in a high-risk population where prevalence is high; 2)Increase specificity
Scenarios for Maximizing PVP:
When specificity of the test is high and prevalence of the disease is high. B/c fewer False Positive (B) occur relative to True Positive (D), improve the likelihood of a positive result is a True Positive
Two Stage Testing:
1) Net sensitivity: Sensitivity (Test 1) * Sensitivity (Test 2)
2) Net specificity: Specificity (Test 1) + Specificity (Test 2) [1-Specificity (Test 1)]
Evaluation of Screening Programs:
Randomized studies, best and most valid evidence of efficacy of screening program comes from randomized trials w/ cause- and age-specific mortality as the O
Case-control:
Individuals w/ and w/o the disease are compared for past screening.
Cohort studies
Case-fatality or 5-year survival rate
Effective Screening Program:
1) Constitute significant
When high Specificity is important (FP is important): If doing really expensive and difficult screening, make sure specificity is higher. When False positive lead to harmful follow-up testing, e.G. Biopsy; If do repeated screens on people, so you can only follow up tests for people have the disease in cancer screening, when treatment has sever side effects, expensive and unpleasant (0 False Positive)
When high Sensitivity is important (FN is important):
Screening for serious diseases where missing a case (false negative) could be dangerous (e.G., cancer screening, HIV screen)
Also when disease is fatal
Also for early stages of disease screening where early treatment can significantly improve outcomes.
Reliability of Tests:
Validity=measure of accuracy in relation to the “truth”. Reliability=Repeatability
Evaluation of Screening Programs Guide:
Could observe benefit from screening and early detection if all or most clinical cases of the disease go through a detectable preclinical phase, and in absence of intervention, most cases in preclinical phase progress to a clinical phase. However, some cases progress too rapidly through detectable preclinical phase to actually be detected by screening or some cases of preclinical disease never progress to clinical disease
Feasibility:
=PVP can be increased by
1) increase prevalence of detectable preclinical disease in the population, e.G. By screening only individuals who are at high risk for developing the diseases on the basis of age, medical history, etc.
2) Increasing the specificity of the screening test
Ramifications:
Reducing # of FP often leads to fewer individuals identified by the test, which can be beneficial in reducing unnecessary follow-up treatments but may also miss some true cases if specificity is not perfectly high.
ALWAYS
Wrong to compare Outcomes from a group of screen-detected cases w/ those from a group of clinically-detected cases
Source of Bias in Evaluating Screening Effectiveness:
1) Lead Time Bias:
Tendency for cases detected by screening to appear to live longer than cases detected clinically; “Lead Time”—time by which screening test advances the date of diagnosis from the usual symptomatic phase to an earlier pre-symptomatic phase. (Occurs when screening detects a disease earlier than it would have been detected because of symptoms, making it seem like survival time has increased when it has not actually affected the course of the disease.)
2) Length Bias:
Tendency for cases w/ a long detectable preclinical phase of the disease to be overrepresented among screen-detected cases compared to symptom-detected cases. E.G. Cases of rapidly progression (very aggressive) disease usually pass through the detectable preclinical stage too quickly to be picked up by most screening programs. (Screening is more likely to detect slower-progressing diseases because they remain detectable for longer periods, potentially skewing effectiveness data since these cases often have better prognosis.) Decrease time between screening (Screen more drequently)
3) Self-Selection Bias:
tendency for persons who choose to participate in screening programs to be different from those who don’t participate in a number of ways that affect survival. – Those participate in screening programs are generally healthier and have other positive health behaviors, e.G. Don’t smoke, not overweight. – Screened are more likely follow medical regimens that may be after a positive screening test. (Individuals who choose to participate in screening programs might differ in health awareness or risk factors from those who do not, affecting the apparent efficacy of the screening.)
4) Overdiagnosis:
tendency for some small proportion of non-diseased individuals to be falsely diagnosed as having the disease, so screen-detected cases are diluted w/ individuals who don’t in fact have the disease. This bias can be limited by rigorously standardizing the procedures leading to a positive diagnosis. (Screening may identify diseases that would not have caused symptoms or death (especially in cancer screenings), leading to unnecessary treatment for such diseases.)
Relationship between sensitivity, specificity: Increasing sensitivity generally decreases specificity and vice versa, high sensitivity is important to identify possible cases to prevent spreading, but may increase FP; High specificity reduce FP but may increase FN.
ACUTE DISEASE EPIDEMIOLOGY GUIDE:
Incubation Period:
Incubation period:
First exposure to onset of clinical symptoms
It varies by disease and is critical for understanding disease dynamics, determining quarantine measures, and estimating the time of exposure.
Median Incubation incubation= 5 days
Types of Epidemics:
1)Common Source Epidemic:
Involves group of persons get sick after being exposed to a common influence like water supply or food, the E can be continuous or intermittent. Graph: Plateau-like pattern or prolonged tail
2)Point Source Epidemic (population exposed at one point in time): Persons are exposed over a brief time to same source, such as single meal or event, the # of cases rises rapidly to the peak and falls gradually. (e.G. Food poisoning, gas lead, nuclear accident) Graph: The # of cases rise abruptly and fall again in a log-linear fashion.
3)Person-to-person:
Results from transmission from one person to another. Typically, # of cases increases more gradually and may involve multiple waves of infection. Graph: Progressively taller peaks, indicating waves of infections
CyclicTrend:
Secular Trend:
(long trends over decades/centuries in one direction)
Steps in Outbreak investigation:
1)Prepare for Field Work: Gather necessary tools and supplies, ensure logistical arrangements
2) Establish the Existence of an Outbreak: Verify the diagnosis, confirm that the reported cases exceed the expected number.
3)Verify the Diagnosis: Ensure that cases are correctly diagnosed
4)Construct a Working Case Definition: Define what constitutes a case for the investigation
5)Find Cases Systematically and Record Information: Active surveillance to identify and document cases
6)Perform Descriptive Epidemiology: Describe in terms of time, place, and person
7)Develop Hypotheses: Based on the integration of all data collected
8)Evaluate Hypotheses: Implement analytical methods, e.G., case-control studies
9)Refine Hypotheses and Carry Out Additional Studies: As necessary, adjust hypotheses and conduct further investigation.
10)Implement Control and Prevention Measures: Based on the findings, suggest and facilitate control measures.
11)Communicate Findings: Inform local health authorities, stakeholders, and possibly the public.
Completion of an Outbreak Investigation:
1) Conform all cases identified and documented
2) Ensure hypothesis has been tested properly and completely
3) Implement and verify the effectiveness of the control measures
4) Prepare final report with suggestions; 5) Follow-up ensure measures are sustained
Calculation Attack Rates:
Attack Rates=confirmed cases/total number of people at risk * 100
Who decides which diseases should be legally reportable?:
Each state government selects the diseases that are reportable in that state; CDC makes recommendations; Epidemiologists from every U.S. State convene annually to review their respective lists and consider additions or deletions from the common list of conditions to be reported to CDC.
Effect Modification and Interaction:
Effect Modification (population Interaction):
True effect of one exposure on disease outcome caries across level of second exposure, True difference in the strength of association b/t E and O across strata (Levels) of a second factor;
This is NOT a bias,
Joint effect of 2 variables on O is much larger or smaller than expected based on their independent effects (Have E 1 or 2 on same person will cause different risk that either larger or smaller than people only have E1 or only have E2) Ex. Someone have lung cancer or someone drinking alcohol and have other people only eat fatty food, people have drinking alcohol and also eat fatty food
Additive Model: The combined effect is the sum of individual effects
– E1 and E2 acting on your body are same; – When effect of E2 adds to the effect of E1 (Att Risk difference); – E1 and E2 are essentially an increased exposure to the same thing; Ex. Smoking and asbestos cause same irritation and damage to your lung, even they are not the identical chemical
Multiplicative Model: The combined effect is the product of individual effects. –
One cell cause damage, and one cell can faster the division of cells; When E2 multiplies the effect of E1 (Ratio measure) – Biological mechanisms through which they influence risk is sufficiently different such that E1 and E2 have greater effects than predicted by simply adding more of same exp.
Homogeneity or heterogeneity of effects:
The effect of E1 on O is not same in strata formed by E2 (E2= association modifier)
Comparison b/ observed and expected joint effects of E1 and E2:
interaction present when observed joint effect of E1 and E2 differs from the expected joint effect based on their independent effects.
Assessing interaction based on evaluating joint effects:
RR00:
completely unexposed / RR10:
people only exposed to E2 / RR01:
people only exposed to E1 / RR11:
Joint exposed
Effect of E1 varies by presence of E2 (interaction)
If joint of E1 and E2 no interaction, then the effect of E1 is the same for people exposed to E2, as for people not exposed to E2. And the effect of E2 is the same for people with or w/o E1