6. EBD III Flashcards
Screening using DMFT/DMFS
DMFT/DMFS - the # of decayed, missing, or filled teeth (T) or surfaces (S) in a mouth
● best known of all ____
● deft and defs - primary teeth, “e” is a primary tooth
indicated for ____
● limitations:
○ can be invalid if teeth are lost for ____ reasons (trauma, periodontal disease, ortho)
dental indices
extraction
non-carious
Diagnosis - Specific \_\_\_\_ Can involve additional information (complaint of pain, physical sign) Screening - Population \_\_\_\_
individual
asymptomatic
Heart attack 10 patients are taking a particular blood pressure drug. 6 of those ten have a heart attack. ● \_\_\_\_ = 6/10 = 60% ● \_\_\_\_= 6/4 = 1.5 “probability” Difference is in what is in the \_\_\_\_.
risk
odds
denominator
Rare disease
• Rare disease; odds and risk are\_\_\_\_
similar
Cohort vs. Case/Control studies
• Cohort (and experimental) ○ People with \_\_\_\_ and asking if they develop \_\_\_\_ ○ \_\_\_\_ ratio • Case-control ○ People with \_\_\_\_; and looking back in time to see if they have exposure to risk factor ○ \_\_\_\_ ratio
risk factor outcome/disease risk disease odds
Cohort
Exposed
Unexposted
Who develops the ____
• Start with \_\_\_\_ and \_\_\_\_
outcome
exposed
unexposed
Case control
What was the ____?
____ outcome
____ outcome
• Start with the cases and controls
exposure
cases
control
Event rate - the proportion of patients in a group in whom the event is ____. 40/200 = 20%
observed
Example 1: EER/CER
• Exposed event rate • Control event rate ○ No \_\_\_\_
risk factor
Relative risk
RR = ____ / ____
RR > 1 a person with exposure is at ____ risk(benefit)
RR < 1 a person with exposure is at ____ risk(benefit)
RR = 1 no effect
Likelihood that someone exposed to a risk factor (or treatment) will develop the disease (or experience a benefit) as compared with someone unexposed.
EER
CER
increased
decreased
It’s all relative …
RR = 3
3 times more likely to develop condition in the exposed group
* If baseline is 30% > tripling risk will bring to 90% * If baseline at .1% > would only increase to 0.3% * Depends on the \_\_\_\_!
baseline risk
Relative risk = EER/CER = 8/71 = 0.11. The risk of tooth loss is decreased by ____%
89
Relative Risk/Risk Ratio
Relative risk = risk of outcome in those with ____/
risk of outcome in those without the ____
____ STUDIES AND ____ STUDIES
RR = 1 no change
RR > 1 increased risk of disease among those with exposure RR < 1 decreased risk of disease among those with exposure
exposure
exposure
cohort
experimental
Odds Ratio
Odds ratio = odds of ____ in those with the outcome / odds of ____ in those without the outcome
____ STUDIES.
OR = 1 no change
OR > 1 increased frequency of exposure among cases OR < 1 decreased frequency of exposure among cases
exposure
exposure
case control
7 dental students who went to a bar develop food poisoning. Odds ratio = Odds(cases) = ____ = 2.5 = 8.33
Odds (controls) =____ = 0.3
= 8.33
• Don't need to know everyone who ate fish • Compare those were affected and those who didn’t, and ask who ate the fish ○ Odds of those who were food poisoned over those who weren't
a/c
b/d
Confounding
• Universally, any study that olooks at exposrue of tv and enuropshy outcomes; include things other things that make it more complicated > \_\_\_\_ goes away
association
Potential confounders
● \_\_\_\_sex ● Body mass index at 14 years ● Parental\_\_\_\_ level ● Parental social class ● \_\_\_\_ at 4 years of age ● Mother’s alcohol consumption during pregnancy ● Mother’s \_\_\_\_ use during pregnancy
gender
education
babysitter
tobacco
Tooth loss confounding?
• Potential confounders between crown and tooth loss: ○ \_\_\_\_ • \_\_\_\_ people - adjust for potential confounders; if can't randomize, need to measure \_\_\_\_ status so you can measure the association
money
randomized
socioeconomic
Effect Modification
Definition: when the magnitude of the ____ of the primary exposure on an outcome (i.e., the association) differs depending on the level of a third ____.
* Third variable - makes relatinoship stronger * And no varibale - goes away entirely
effect
variable
• Where ESL - may not be harmful, but can also be beneficial and can ____ them speak more efficiently
help
Effect modification
* No difference bt new \_\_\_\_ and placebo * Thought there would be a \_\_\_\_ * In this study more \_\_\_\_ than men
drug
relationship
women
Effect modification
• Look at mean HDL between gender ○ In men, there is a new drug ○ Statistically \_\_\_\_ • Important in including more women in research studies • Want to weed out \_\_\_\_; but bc the relationship is different bt two groups you want to see the \_\_\_\_
significant
confounder
difference
Herpes Zoster
• Odds ratio ○ People with herpes and those without, and back in time looking at exposures • For RA ○ No diff bt three age strata ○ Age is not \_\_\_\_ ○ No diff bt crude odds and adjusted odds ratio § The same with those exposure and not exposed • For lupus ○ \_\_\_\_ ratio is diff in youngest than in the two other cohorts ○ Crude odds ratio is an average of 3 odds ratios ○ In lupus, age is an \_\_\_\_
effect modifier
odds
effect modifier
Key points:
1. Avoid or adjust for confounding - it ____ true effects
2. Effect modification - look for it and report it
3. A given variable can be a ____, an effect
modifier, both, or ____, depending on the research question …
* Avoid confounding by \_\_\_\_ * Age can be a \_\_\_\_, not always an effect modifier
masks confounder neither randomizing confounder
Types of Studies
- Descriptive- used to quantify ____ status in a community (____ and incidence)
- Analytical - assess the ____ between exposure/disease, cause/effect
• Descriptive ○ Improving \_\_\_\_ ○ Not trying to generalize to other clinics • Analytical ○ Make a \_\_\_\_ statement
disease
prevalence
relationship
quality
• Inferential
○ Want to know about the ____
population
Inferential statistics
“Is it statistically significant?”
● Based on the ____ and the ____ of effect
● Depending on the ____ of data (categorical, continuous), the statistical
tests that are used will vary …
○ Correlation coefficients - used to compare ____ data
○ T-tests - used to compare ____ data
○ Chi-square tests - used to compare ____ data
○ P values - gives us the confidence to accept or reject the ____
* Large sample size - can be significant if not impressive * Small sample - great result, but not enough to generalize
sample size magnitude type continuous continuous categorical null hypothesis
• Increase study size > 40 to 80; ____ will go down; but if there is a bias that’s a ____ error that is irrespective of increasing the sample
randomized error
systematic
Variable
Definition: what is being observed or measured ■ Traits or \_\_\_\_ (age, gender, race) ■ \_\_\_\_ a research is trying to manipulate (DAU attendance) ■ Outcomes that are of \_\_\_\_ (understanding of 4 handed dentistry)
characteristics
dimensions
interest
Types of Data
● Discrete values ○ Daily \_\_\_\_ (yes/no) ○ Sex ○ \_\_\_\_ (male/female/trans/fluid) ● Continuous variables ○ May take any value within a \_\_\_\_ range ○ Age, \_\_\_\_, height ○ \_\_\_\_ of DAU days attended (0% → 100%)
attendance gender defined weight %
Confidence intervals
How we go from the sample → population
• Normal distribution • Alpha is 0.05 ○ Two tailed, so 0.025 • Value falls in \_\_\_\_ > you'll be confidence
range
Hypothesis testing
○ Ho (null) : there is no difference in the ____ of days of DAU attendance between those who do and do not receive kitty cafe tickets
○ Ha (alternative): Those who receive kitty cafe lottery tickets are more likely to ____ DAU
• Continuous variable here
%
attend
P-values
The probability of seeing results like ours if, in
truth, there is no difference.
● Small p (e.g. 0.03) the observed results
would occur ____ if the null hypothesis were ____.
● Large p (e.g. 0.3) the observed results would occur ____ (30% of time) if null hypothesis were ____)
● Based on the p-value, we accept or reject ____.
• \_\_\_\_ p ○ Can reject the null hypothesis • \_\_\_\_ p ○ Cannot reject the null hypothesis
rarely true often true Ho
small
large
The “Magical” .05 level
Most research studies are said to be statistically significant if p
0.05
Type 1 errors (alpha)
At alpha = 0.05, we accept a ____% chance of rejecting ____ when it is actually ____.
5
Ho
true
Type 2 errors (beta)
Failing to reject ____ when it is ____.
Ho
false
T-tests
● Assume the dependent variable has a ____
● ____ test
* Same distribution between both examples * If attendance is greater than 45% in those who didn't have lottery attendance, then there is a diff in attendance bt both groups * Test of means
normal distribution
means
Chi-square tests
● The dependent variable does not have a ____
No variable
normal distribution
5
Fishers Exact tests
● The dependent variable does not have a ____
• If cell has less than \_\_\_\_ > use fishers exact test rather than chi square
normal distribution
5
Correlation Coefficient
● Between two ____ variables (i.e. GRD grade and DAU attendance)
continuous
Calculating Power
Ho (null hypothesis) strength level = 500
Ha (alternative hypothesis) new drug increases strength level
Ho: mean = 500 (no effect)
Ha: mean > ____
* Don't know the reality * If we think our study is true, and matches reality > we're good * And if no effect in both > we're good * If study says no effect but there is an effect > \_\_\_\_ * If study says there is an effect, but there's no reality > \_\_\_\_
500
type 2 error
type 1 error
Calculating Power
• Cut off is 533 ○ Greater than this > intervention has an \_\_\_\_ • Okay with 0.05 error ○ \_\_\_\_ for research studies
effect
standard
• Cut point is 533
• More likely to think that test is not true when it is true
○ ____ error
beta
• Power is related to ____
○ 1 - beta
• Alternative hypothesis
○ Think direction will be in a specific way - ____ tailed
○ If can go either way -____ tailed alt hypothesis
beta
one
two
Calculating Power
B = beta ~ \_\_\_\_ E = \_\_\_\_ size A = \_\_\_\_ N = \_\_\_\_size
power
effect
alpha
sample
By convention, beta < 20% so Power > ____%
* \_\_\_\_ size comes into play * Beta is usually less than 20 and power is greater than 80
80
sample
B = beta
E = effect size
(difference between ____) A = alpha
N = sample size
• Expect big effect size - test pop will \_\_\_\_ from population without intervention a lot > large effect size (more people show up bc of lotto tickets); and vice versa ○ \_\_\_\_ of effect will affect power calculations ○ How well the drug/intervention is designed; should have an idea before going into study about what the ES is
means
differ
size
• Alpha is usually pre-determined
○ If beta is 0.2, and alpha is ____
0.05
• All comes down to ____
○ If expect ES to be largely different; then you don’t need a ____sample to be well powered
○ If smaller sample > less ____ > the sample can allow you to have a good ____ regardless of smaller effect size
• Sample size is KING for power calculations
sample size
large
error
beta
