Exam 1 Flashcards
3 aspects of internal validity before statistical association:
-check for bias, counfounding/effect modification, statistical significance
Bias
systematic (non-random) error in study design or conduct leading to erroneous results (distorts relationship btw exposure & outcome)
after study end- fix bias?
Nothing you can do to fix it
prospective (prestudy)
adjustment minimize bias (assess to confirm internal validity and conclusions)
components of bias
- source 2. magnitude 3. direction
bias magnitude
how much of an impact bias has on changing odds ratio
bias direction
move Odds Ratio/Risk Ratio away or towards 1.0 (groups equal) enhance or minimizing
main categories of bias
how you measure (collect data) or Selection (ppl) -measurement- related biases -selection-related biases * both result in grps different–dont want– want grps to ne as equal as possible except the one thing you are studying biases make grps different & creates the error
measurement- related biases
way the researcher collects information, or measures/observes subjects which created a systemic difference between groups in quality/accuracy of info.
selection-related biases
way the researcher selects or acquires study subjects which creates a systemic difference in the composition between groups (not from same pop/group)
types of selection bias
health-worker bias self-selection/participant (res ponder) bias control selection bias
selection bias
-selecting study subjects that are not representative of your primary pop. of interest or generates differences in grps being compared -how you pick your people- bias– made groups different (inappropriate) -convenience study: mall on wed.
health-worker bias
environmental employer research -sick & dead not at job site
self-selection/participant (responded) bias
people who want to be in survey -volunteer different than nonresponse
control selection bias
only diff btw grps not clear based on disease definition -call 1st 50, not accounting for if they have a phone/can they answer
recall/reporting bias
subject-related variation -differential level of accuracy/detail in info btw grps - ex: recall after bad event, after disease; diabetics monitor eating more than regular ppl
hawthorne effect
overly encouraging, give more than needed, overly positive/helpful
types of measurement bias
subject related: recall bias, contamination bias, compliance bias, lost to follow up bias observer related: interviewer bias, diagnosis bias,
contamination bias
aspirin v placebo: dont want other drugs in same family (how investigators account for what else taking) -control grp accidentally receive tx of exposed to intervention being studied
compliance/adherence bias
comply, follow instructions- worry one grp different in complying -grps being interventionally studied have different compliances
lost to follow-up bias (attrition)
follow up diff in 1 grp or another; diff withdrawal or lost to follow-up rates or other differences -differential v. non-differential ex: surgery-good pain control drugs vs. placebo– hear from person in pain or drops out
interviewer (proficiency) bias
- not trained in what to do -interviewer expects certain answers -conscious or unconscious –> mask inverstigator so they dont know who meds/who placebo -body language, inflection, expect diff. answers- skewed/false info
diagnosis/surveilance (expection) bias
diff evlauation, classification, diagnosis, observation preconceived expectations (hawthorne-like effects from researchers) expect drug to feel better
controlling for biases
*Select precise, accurate, & medically-appropriate measures of assessment and evaluation/observation -validated screening -specifics of data collection ex: polygraph test *blinging/masking *multiple sources data *randomly allocate observers for data collection *methods to minimize loss to follow up
misclassification bias
error in classifying either disease or exposure status or both -measurment (information/observation) bias put ppl in wrong group
misclassification bias types
non-differential & differential
non-differential misclassification bias
error in both grps equally misclass. of exposure/disease is unrelated to the other move ratio towards 1; attenuates effect of association
differential misclassification bias
error in one grp differently than other put ppl in wrong grps missclass. of exposure/disease is related to the other move ratio in either direction in relation 1.o (inflate or attenuate)
controlling for misclassification
want to minimize error and balance equation -similat tech. to limiting biases- its a measurement-related bias -technology on both grps
**differential misclassification
?
**non-differential misclassification
?
confounding variable
3rd variable that distorts an observed relationship btw exposure & outcome (disease) -associated with both but independent of both (not in causal pathway) (something else that gets involved with exposure & outcome; getting in the way)
confounding affect
can over or under estimate an association (RR/OR/HR) and change direction of effect
impact of confounders
intensity/magnitude/strength: association more or less extreme than true association direction: association that moves association in + or - direction
confounding example: coffee and low birth rate
smoking has relationship with both outcome and exposure= confounder
factor effect of confounder
report adjusted odds ratio
testing for confounding
step 1: crude outcome measure of association (OR/RR) btw exposure & outcome - unadjusted [linear 2x2 factoring] step 2: re-calculated outcome measure of association (OR/RR) btw exposure & outcome while statistically controlling the effects of the confounder [add 3rd variable for outcome of interest] step 3: compare the crude vs. adjusted measure of association btw exposure & outcome [compare the 2]
crude & adjusted estimate (RR/OR)
20% is confounding present
purpose of controlling for confounding
to get more accurate estimate of the true association btw exposure & outcome
causal pathway btw exposure & outcome
confounder NOT in pathway
ways to control confounding
can do before!! 1. study design stage 2. analysis of data stage
study design stage
randomization (blocked or stratified) restriction matching
analysis of data stage
stratification multivariate statistical analysis
randomization
hopefully allocates an equal number of subjects with the known & unknown confounders into each intervention grp (make grps equal) strengths: sufficient sample size, randomization makes grps equal weakness:sample size might not be large enough to control for known & unknown confounders, only intervention studies
restriction
limit who can enter study, decrease ability to take results & transfer to rest of community restricted to only subjects who do not fall within pre-specified categories of confounder strengths: straight forward, convenient, inexpensive, does not neg. impact internal validity weakness: sufficiently narrow restriction criteria reduces sample size, residual confounding effects, neg impact external validity
matching
more difficult, hands on researchers activity- difficult to find exactly like pt in other grp selected in matched pairs related to confounding variable to equally distribute confounder among each grp (let in but need one for other grp) strengths: intrusive weakness: difficult to match on everything needed to match, to find same person; overmatching
stratification
statistical analysis of data by evaluating association btw exposure and disease within various strata within confounding variables strengths: intrusive, straight- forward and enhances outstanding of data weakness: impractical
multivariate analysis
statistical analysis of the data by mathematically factoring out the effects of the confounding variables strengths: simultaneously control for mult. confounding variables; beta coefficients converted to OR’s weakness: not understand data
effect modification (interaction)
3rd variable modifies the magnitude of effect of association by varying it within different levels of 3rd variable (effect modifier) MUST report for each strata (informative)
effect modification (interaction) example: mortality odds of newborn’s born as singleton
birth weight (3rd variable) may impact odds of mortality- variable checked for confounding and effect modification–> Unadjusted mortality odds: 1.06 children born as single babies 6% more likely to die adjusted (for birth weight) mortality odds: 1.02 birth weight not confounder NEEDS TO CHANGE MORE THAN 20%
effect modification (interaction) example: mortality odds of newborn’s born as singleton– check strata
reduced risk—> increased risk as weight increased **effect modification is present b/c OR changes acc. diff. strata of effect-modifying variable
effect modifier
changes layers/strata ex: birth weight not confounding- didnt change outcome & exposure BUT layers of weight does change (at least 20%) btw highest and lowest OR
testing for effect modification
Step 1: crude outcome measure of association btw exposure & outcome (OR/RR) Step 2: crude outcome measure of association (OR/RR) btw exposure & outcome FOR EACH STRATA OF EFFECT-MODIFYING VARIABLE step 3: compare stratum-specific measure of association btw exposure & outcome (OR/RR)
point-estimate (RR/OR)
difference by 20% btw lowest & highest strata of effect-modifying variable IF effect present
detecting confounding & effect modification: how does the change in RR/OR change in the presence of confounding & effect modification?
problem we want to eliminate (control/adjust for via several means) in the study
confounding
-crude vs. adjusted
is adjusted >20% difference than crude
control @ beg. & end
natural occurenece that we want to describe and study further
effect modification
not ignore, explain- informative
compating stratum-specific measures of association
is stratum-specific estimates >20% differnece from each other?
practice exercise
confounded?
interacted?
measures of association
exposure & disease
exposed v. non-exposed
diseased v. non-diseased
rows: exposure
column: disease
descriptive group comparisons common
absolute differences
relative differences
absoulte differences
SUBTRACTION
- subtracting frequencies
- males had 28 more surgeries
- females had 28 fewer surgeries
relative differences
DIVISION
division (ratio) of frequencies
- males had 2.6 times the number of surgeries compared to females (>160% increase)
- females had just under 40% number of surgeries of males
relative differences
DIVISION
division (ratio) of proportions
*more common
divide proportions than interpret
- males had more than 2.06 times the proportion of surgeries compared to females
- females had just over 45% lower proportion of surgeries of males
absolute differences vs. relative differences
absolute differences will always be smaller then relative differences
EX: absoulte: A-B
20-10=10%
2-1=1%
.2-.1=.1%
relative: B/A
10/20= 50%
or 20/10=2 (100% higher)
1/2=50%
.1/.2=50%
*risk of disease of B is 1/2% A
or group B has 50% rate of disease compared to group A
pharma companies
relative differece because absoulute diff. is always smaller than relative diff.
risk ratio/ relative risk (RR)
used in studies where subjects are allocated based on exposure and outcome
ex: cohort studies
risk- incidence rate (IR), attack rate (part/whole)
ratio of 2 risks from different groups
Ramapril Example
Outcome?– bad (death/MI)
Ramapril: 14% had outcome
17% placebo had outcome
absoulte difference in event rates 3.8%
placebo 3.8% more likely
or
Ramapril had 3.8% absoulte lower event rate
% rates
the risks of event in each group
Risk and Risk ratio
risk is a proportion
probablility of outcome in exposed and in non-exposed
probability of outcome in exposed A/(A+B)
probablility of outcome in non-exposed C/(C+D)
Incidence risk (IR)
Risk is a proportion
risk of outcome in exposed: A/(A+B)
risk of outcome in nonexposed: C/(C+D)
Risk Ratio
ratio of risk from 2 different groups
risk in exposed/risk in nonexposed
2/2 = 1
17/17 = 1
always equalness, no differences
bigger/less = >1
(Ramapril EX) Risk ratio
MI ramipril: 14% MI placebo:17.8%
RR= 14/17.8=.77
RR (.77) < 1
ramapril has a decreased risk
numerator
always study group/ reason for study
interest/reference
Interpreting ratios- risk/odds/hazard
=1 no difference in risk/odds
>1 increased ratio (larger, bigger,above)
>2 “times control”
<1 decreased ratio (from one how much do you have to go down to get to odds ratio below one)
increased RR >1
use decimal value (converted to %) for interpretation
ratio over one-increased
If RR 1.53: then 53% increased risk in comparative grp
“the study grp of interest had a 53% increase. stop when get to 2 –> 2xs, 3xs…”
>2
OR=6.18
comparator group is 6.18 times greater odds
Decreased Ratio < 1
subtract decimal value from 1 (answer converted to %)
HR= 0.73
27% lower probablility of hazard outcome
amount you went down to get to it, CANT JUST GRAB THE DECIMAL
forest plots
plot with indicator line at 1.o
talk about where OR # lands
higher or less than 1
vertical line at 1.0 represents equalness/sameness
ppl who take statins are about 40% lower hazard risk for dying of MI
higher end of 1.o
magnitude btw 1-3
when interpreting ratio’s (RR, OR, HR) looking for…
- direction of words
- magnintude
- group compairison
*target for wrong answers ex: grps backwards
(increased or decreased)
Absolute Risk Reduction (ARR)
attributable risk
Absoulte Risk difference (ARD)
SUBTRACTION
simple absoulte difference (subtraction) in risks
AR
risk difference in outcome among exposed that can be attributed to actual exposure
ex: ramipril 14% placebo 17.8%
ARR= 3.8%
subtract them - claim difference soly to use of ramipril
attribute
risk difference
*** example exam page
relative risk reduction (RRR)
relative risk difference (RRD)
DIVISION
(ARR)/ R exposed
ex: ramipril 3.8%/17.8%=21.3% relative risk reduction in Ramipril (exposed) grp relative to placebo (unexposed)
Number needed to treat (NNT)
how many ppl needed to tx to get benefit
(vital to understand & explain to pts)
1/ absoulte risk reduction [subtract risk then 1/ #]
(1/ARR) ex: 1/0.038= 26.3=27 pts
interpretation: # of pts needed to be tx to experience the studied event outcome (benefitial or harmful)
Ramipril number needed to treat
if ramipril harmful, AR would be negative and NNT would translate “for every 27 pts 1 would be harmed/bad event”
how many to avoid one outcome of death, stroke, MI
1/ARR
Number needed to treat
ex:opiod
NNT easier for ppl to comprehend
for every 4 pts, i given opioid to 1 pt is going to get nausea (seems common)
on average 1 in 4 pts will have nausea
constipation: 1 in 5 got constipatin (common)
NNH
same as NNT
negative event
Number needed to harm
ex: 1 out of 822 had MI
would have to tx 822 pts with this drug before 1 had MI
*** Practice
Rexp=
Rnexp=
ARR=
RR=
RRR=
NNT=
**Pracice with Bivalirudin vs Heparin
Bivalirudin = 190 N=2289
Heparin= 199 N=2281
Rexp=diseased/total= 190/2289=8.3%
Rnexp=nondiseased/total= 199/2281=8.7%
ARR=subtract= 8.7-8.3=.4%
RR=division=8.3/8.7=.954=95.4% <1 so lower risk
RRR= 1-.954=0.046=4.6% risk reduction relative to unexpected risk
NNT= 1/ARR= 1/.004= 250
tx 250 ppl with Bivalirudin in order to prevent 1 person from death or MI
Risk ratio
part/whole
odds ratio
occurance/nonoccurance
odds & odds ratio (OR)
odds is a ratio
used in studies where subjects are allocated based on disease presence (y/n) and evaluated for exposure
odds of exposure vs. odds of not exposed (cases)
A/C
odds of exposure vs. odds not exposed (controls)
B/D
odds of exposure in cases
odds of exposure in controls
A/C
B/D
Odds ratio (OR)
division of odds
(A/C)/ (B/D)
**Practice
odds of exposure in cases= A/C=457/362=1.26
odds of exposure in controls=26/85=0.305
odds ratio (OR)=1.26/0.305=4.13
practice HPV
Not pt over whole; odds of occuring/odds of not oc
odds of exposure in cases= 37/63=0.587
<1
odds of exposure in controls=11/189=0.058
< likely
odds ratio (OR)= 0.587/0.058= 10.1
10xs more likely to be involved; those with ph cancer are more likely to have had HPV
interpreting OR
looking at all diabetics in the US
were US diabetics more likely to be on a statin or not & characteristcs btw statins and nonstain users
*one grp designated as reference group (odds ratio compared to 1)
compared to female US diabetics, males 38% more likely to be tx with statin to lower cholesterol
OR=1.38: more bc >1 38 into percentage
compare 2 groups and describe likelihood of outcome in 1 group compared to another
all “ratios”
RR,OR,HR
(all ratios) if ratio is 1.0
outcome is equally likely for both groups
(all ratios) if ratio is > 1.0
outcome in more likely to occur in main study (comparison) group
(all ratios) if ratio is < 1.0
outcome is less likely to occur in the main study (comparison) group
association
connection, linkage, NOT cause
(relationships OR/RR/HR) btw exposure & disease
cause
precursor event/condition/characteristic required for the occurance of the disease
main types of associations
- artifactual (false)
- non-causal
but has a role (smoking & lung cancer)
- casual
(one of (not only) causal process)
artifactual associations
can arise from significant bias and or extensive confounding
(false associations)
non-causal associations
- disease may cause the exposure
- disease & exposure both associated with a third factor (confounding)
non-causal associations-disease may cause the exposure
chicken before egg or egg before chicken
RA leading to physical inactivity
(Disease [RA] may cause exposure [inactivity])
non-causal association-disease & exposure both associated with a third factor (confounding)
third factor-
positive association shown btw coff & CHD or Downs & birth order
koch’s postulate’s for infectious disease
if implying causative must have 4 things *limitations
- Must be present in every case of disease
- Must not be found in cases of other diseases or healthy individuals
- Must becapable of isolation, culture and reproducing disease in experimental animals
- Must be recovered from experimentally-induced diseased animals
koch’s postulate’s limitstions
-Disease production may require presence of “co‐factors” that postulates don’t address
-Viruses can’t be cultured similar to bacteria
-Not all viruses/bacteria induce clinical disease
•Carrier & Sub‐clinical disease
Mill’s Canons
cause of any effect must be consist of a constellation of concepts that act in concert
(ex: heart disease- overweight,cholesterol,hypertension)
types of causal relationships
sufficient cause
necessary cause
component cause (risk factor)
Suffcient Cause
set of minimal conditions/events that inveitably produce disease
- if present the disease will always occur
- rare/genetic abnormalities
- can have multiple risk factors/component causes that induce disease
necessary cause
cause precedes a disease
cause must be present for disease to occur, yet the cause may also be present without the disease occuring
EX: Tb
- carriers, have in lungs but not sick
- sick- it has to be in lungs
component cause (risk factor)
characteristic that if present and active increases the probability of a particular disease
(increase likelihood of outcome)
ex: High HDL
some pts primed/susceptible to disease before component causes induce disease (multi-factoral)
multi-factoral
some pts primed/susceptible to disease before component causes induce disease
multiple causation
to examine the influece of a single factor, necessary to adjust/control for the effects of other factors
3 ways to control/adjust variables
- restriction
- matching
- stratification
keep other factors out from study
only <65
no smokers
resriction
similar characteristics in each grp
age, gender, disease, smoking status
matching
categorize patients on exposure levels or disease severity or other important pt characteristics
stratification
Hill’s Guidelines
how close to causation is an association
- strength
- consistency
- temporality
- biologic gradient
- plausibility
the higher the # of criteria met, when evaluating an association, the more likely it may be causal
Strength
size of association (RR,OR,HR)
bigger the association, bigger the relationship
ex: smokers 20xs greater risk of lung cancer
*stong association is neither necessary or sufficient for causality & weakness of association is neither necessary nor sufficient for absence of causality
consistency
reproducability
repeated observation of association in different populations under different circumstances in different studies
ex: smoking & CHD
*may still obscure truth-Menopausal Hormone Therapy (consistency can still be wrong)
temporality
necessity that the cause precede the effect/outcome in time
time/order - proximate cause (short term interval)
- distant cause (long term interval)
induction period & latent period
Ex: (med) side effects & rns / food poisoning
testicular cancer
longer the time period harder to connect
biological gradient
observation of gradient of risk (dose-response) associated with the degree of exposure
prove gradient- down causal pathway
minimum cumulative exposure before disease pop up
threshold effect
plausibility
biological feasibility the association can be understood and explained
make any sense? is event biologically plausible if really true?
plausibility decision on criterion-based from prior beliefs, which may be flawed or incomplete
ex: stomach ulcers
pitfalls in causal research
Bias
confounding
effect modification
synergism
synergism
interacting of 2 or more presumably-causal variables so that the combined effect is clearly greatrer than the sum of the individual effects
EX: health compromising behaviors on preterm births
multiple risk factors, when present put together, worse in risk than alone
induction period
component cause to disease onset
latent period
disease onset to diagnosis/clinical presentation
(somethings take time to develop)
example of temporality
studies show higher lung cancer rates among former smokers during 1st year cessation than current smoking
infer: cont. smoking must decrease risk of cancer
actuality: those who stop b/c early symptoms of illness
(already diseased or already sick and cont. to smoke)
distribution of diease
frequencies & patterns of disease occurance
frequencies of disease occurrences
counts in relation to size of population
patterns of disease occurances
person, place, time
3 Ws: who, where, when
Descriptive Epidemiology
descriptive epidemiology
who, where, when
used to know if location is experiencing disease occurance more frequently than usual
surveillance systems
passive
active
syndromic
passive surveillance system
relies on healthcare system for required reportable diseases
active surveillance systems
public health officials into community to search for new cases
syndromic surveillance system
pre-defined symptoms reported or evaluated
bio survelliance
certain symptoms connected to bad diseases
case definition
most critical element that must be defined
(diagnostic criteria)
*set of unifrom criteria used to define a diseas for public health survellance
change over time as learn more about disease
must accurately define & execute how
confirmed v. probable
CSTE
NNDSS
councile of state and territorial epidemiologists
CDC’s national notifiable diseases surveillance system
classic epi curve
visual depiction- represent data (derived from line table)
who when where
1.pattern of spread (shape)
- common or point source (continuous & intermittenet)
- propagated source
- magnitude of impact
- sentinel case/ peak/ outliers
- time trends
- start/stop/duration
helps to form hypothesis
occurrence of disease clearly in excess of normal expectancy
increased # of disease above what is customary
epidemic
epidemic limited to a localized increase in occurrence of disease
-more concentrated
outbreak
(cluster)
constant presence of a disease within a given area/pop. in excess of normal in other areas
always high for area
ex: HIV in africa
endemic
epidemic occuring over a very wide area involving a large number of people
epidemic across globe
(1918 flu; swine flue h1n1)
pandemic
epi curve pattern of spread: common or point source (continous & intermittent)
not person-to-person
from a common, single point source for the outbreak
epi curve pattern of spread: propagated
person-to-person spread
epi curve magnitude of impact
sentinal case/ peak/outliers
time trends (rate of occurance)
start/ stop/ duration
epi curve helps to form thypotheses on:
routes of transmission
probable exposure periods
incubation period (help id/eliminate causes)
one blob of increasing and decreasing disease
Common or point source outbreak
continous
NOT repeated or propagated (or spread)
NO sentinel/ index case (ppl together and exposed to something common) (no blob early as index/start case)
common location, see single pop up
common or point source
outbreak continuous
NOT repeated or propagated
Semtinel/index case
common or point source
intermittent outbreak
IS repeated
propagated (persontoperson or source of illness still there)
ex: local lake, getting same exposure over and over
propagated transmission
infected subjects infect others who spread infection
Index case
(see incubation period)
incubation period
(dont know when building retrospectivley, can change time)
build from initial exposure
NO index case bc initial exposure not a person
ex: state fair - what did you eat? rides? ??
probable exposure period
if know minimum & maximum incubation time- can build probable exposure period
ex: 6 days +/-3
rubella from disney
further back in time, harder for ppl to recall
(relative) measures of disease frequency
Ratios
proportions
rates *time period important, need all same to compare
divison of 2 unrelated numbers
(num not part of denom)
ratios
division of 2 related numbers
num is part of denom
proportions
division of 2 #s with time in denom
rates
gender surgical ratio-general
ratio of female COB to male COB who have undergone surgical procedure
female/# male
scientific academic acumen
proportion of students with undergrad science gpa 4.o
students with 4.0/all students
biopsy ratio
female biopsy/# female students
natural history of disease timeline
stage of susceptibility
stage of subclinical disease
stage of clinical disease
stage of recovery, disability or death
factors in comparing measures of disease frequency between groups
- # people affected (frequecy)
- size of source pop. or at risk
- length of time
** need to standardized, similar denominator
Comparing populations examples
breast cancer in 2 counties
100 cases in A
75 cases in B
population A 50,000
population B 5,000
which country appears to have most cancer? A
which county appears to have a higher rate? (rate= time in demon)
which appears to have a higher rate? (missing time)
100/50000=.2% vs. 75/5000=1.5% B
Cases in A 1 yr Cases in B 3 years which higher rate of cancer? MAKE SAME DENOM.
B
disease rate
of events/ equal # person years
standard baseline comparison= standardized rate
rates of disease MUST
population size & time period
must be equal to adeuately & appropriately compare frequencies btw grps
incidence
INclude
inclusion of new cases
calc. those at risk
Prevalence
existing cases of disease + new cases of disease
PREVious
take everyone
indicence & prevalence are both
proportions & factors in the at risk or base population (denom)
measures of disease frequency: incidence
risk, attack rate, cimulative incidence
new cases of disease/ 3 persons at risk for the disease (or in pop)
- num and denom same time frame
- NOT dynamic populations
-subtract immunized grp from total then take out those who already got disease
***denom only those at risk
ex: attack rate of nausea & vomitting= everyone N&V/ate pizza
dynamic populations (fluctuations) denom (risk grp) used as:
pop at start of year
average pop over the year
pop at mid year
**** subtract out already diseased or immune
new cases of disease/person-time at risk for disease
incidence rate
-useful when everyone followed the same amount of time
person-time
100 person years: contributed 100 person years to population
100ppl for 1yr, 10ppl for 10yrs; 1 person for 100yrs; 25ppl for 4yrs each
4o pack years: packs/day X #yrs
incidence density
new cases/ TOTAL PERSON-TIME of pop at risk
fluid time
contribute to denom
* ? repeated disease
measures of disease frequency: prevalence
existing cases of disease/# persons in pop
num & denom time frames same
denom: with disease + at risk
point-prevalence (pt in time- Dec 31st) & period-prevalence (period of time- 1 yr)
practice- gonorrhea
incidence rate/100,000: 297/100,000
733151 new/246552000 pop X 100,000= 297
incidence rate/1,000: 2.97/1000
733151 new/246552000 pop X 1000= 2.97
*practice-gastroenteritis
Of 46 that got sick, 18 ate tuna salad, 5 ate chocolate. attack rate of chocolate?
attack rate: 46/75= 61%
chocolate attack rate: 5/75???
Seven cases of Hep A among 70 kids attending a child care center.Each infected child from different family. The total number of persons residing in the homes of the 7 affected families was 32. One incubation period later, 5 family members of the 7 infected children also developed hepatitis A
secondary attack rate:
total–> 32
- 7 <– already have
= 25 (denom) at risk
new cases–> 5 (nem)
5/25=20%
practice-death
incidence rate of death:
2 deaths/19 person years = 0.11 deaths/ (1) person years
11 deaths/100 person years
practice- incidence of disease
incidence of disease from 1990-1996:
NO TIME IN INCIDENCE JUST RATE
so cases/population= 2/6= .33=33%
incidence rate of disease
cases/ person time= 2/ (2+2+3+7+2+6)= 22 –> 1/11 =.09/person years
or 9/100 person years
*** ? practice- incident: just count; incidence: new cases/total pop; incidence rate: time (person years)
calc incidence of disease after Oct 1?
of 21 ppl, 4 already had at start, 7/21 vaccincated?
how many incident cases of death occurance?
cures?
what is denom in calc prevelance of disease on April 1st?
period prevalence of disease the whole year?
point prevalence Oct 1?
April 1?
incidence of disease: case/total pop= 6/? 21-4= 17ppl
21-4-7= 10ppl (denom)
death occurance: 5 deaths in 1 year
cures: 2cures in 1 yr
prevelance: 21 (everyone in study)
period: 21 (ever been pt of group)
point whole yr: (cases at pt) new + existing/ all cases= 10/21=.48
pt oct: 2/21
pt april: 6/20
rates
person time
practice- STD
period prevelence: 2 months
180cases/300 total= .60=60%
father of epidemiology
John Snow
public health-discipline basic science studies
distribution
determinants
of disease in populations to control disease and illness and promote health
epidemiology
subspecialties- disease, exposure, population, combined
differences in disciplines
basic science: cells tissues biochem proccesses understanding disease mech/ pharam, microbio
clinical science: sick pts/ improving diag. and tx, internal med
public health: populations/communities/ preventing disease and promo health/ epidemiology policy and mgmt
epidemiolgist
surveillance and descriptive epi
inference
analytical epi
community intervention
objectives in Epi
id patterns in disease occurence
det. extent of disease
id causes of or risk factors
natural course of disease
evaluate effectiveness of measures
public health policy
POPULATIONS
Epidemiological assumptions
disease occurence is NOT RANDOM
SYSTEMATIC INVESTIGATIONS of different populations (method)
MAKING COMPARISONS (data)
distribution of disease
frequencies of disease occurences
- counts in relation to size
patterns of disease occurences
- person place time
DESCRIPTIVE EPI
who where when
determinants of disease
- factors of susceptibility/ exposure/ risk
- etiology/ cause of disease
- mode of transmission (source)
- social/environmental/biologic elements that determine occurrence/presence of disease
ANALYTICAL EPI
Why, How
core functions of epi
public health surveillance
field investigation analytical studies
evaluation
linkages
policy development
Public health surveillance
to portray the ongoing patterns of disease occurrence so that investigation, control, and prevention measures can be developed and applied
- registries
skills: data collection, data mgmt, DATA INTERPRETATION, presentation
field investigation
to determine the sources of vehicles of disease or to learn to simply learn more about the natural history, clinical spectrum, descriptive epi, & risk factors of disease
ex: ebola, SARS
analytical studies
to advance the info (hypotheses) generated by descriptive epi techniques
comparison group
skills: design, conduct, analysis, cmun findings
evaluation
determine, systematically & objectively, the relevance, effectiveness, efficiency and impact of activities with respect to established goals
linkage
to collaborate (link) with other professionals
policy development
provide input, testimony, & rec. regarding disease control and prevention strategies, reportable disease regulations and health care policy
the epi approach
counting (frequencies): cases/events and describing person, place & time
*case/study definitions!
dividing (percentages): 3 cases/ denom to calc rates, ratios, & proportions *Standardized and unstandardized pop. sized
comparing: change in disesase over time, absolute & relative differences, statistical differences
