Biostats: Disease frequency, screening, & Treatment impact - Beehler Flashcards
What is Prevalence?
Helps understand the disease burden or extent of a health problem.
= (Number of people with a disease at a specific point in time) / (Number of people at risk for the illness at that point in time)
What is Incidence?
Helps understand the risk of a specific health event. Main measure of acute disease.
= (Number of NEW people with disease during a time period) / (Number of people at risk for illness during that time period)
What is Sensitivity?
The probability that a diseased person will be identified correctly by a diagnostic/screening test (AKA true-positive probability or true-positive rate)
= (True positives) / (Total # ill people [true positives + false negatives])
What is Specificity?
The probability that a well (non-diseased) person will be identified correctly by a diagnostic/screening test (AKA true-negative probability)
= (True negatives) / (Total # well people [true negatives + false positives])
What is Absolute risk?
Absolute risk (aka “risk difference”): difference in risk of developing a disease or undesired outcome after treatment
= |CER – EER|
Reduction or increase?
Focus on relationship between CER and EER
What is Relative risk and risk ratios?
Relative risk (aka “risk ratio”): proportion of treatment group risk to control group risk
EER/CER
Risk of bad outcome in treatment group:
…Increases when RR > 1
…Decreases when RR
What is Number needed to treat/harm?
…Treat (NNT)
Number of patients who need to be treated to get 1 additional patient a favorable outcome (like therapeutic effort)
NNT = 1/ARR
…Harm (NNH)
Number of patients who, if they were treated, would result in 1 additional patient being harmed
NNH = 1/ARI
A county in Minnesota has a population of 1,500. In 2013, 180 individuals were diagnosed with type 1 diabetes. Last year, 30 individuals were diagnosed with it. What is the prevalence of type 1 diabetes in this population in 2014?
= (180 + 30) / 1500 = 210/1500 = 0.14
=14%
A county in Minnesota has a population of 1,500. In 2013, 180 individuals were diagnosed with type 1 diabetes. Last year, 30 individuals were diagnosed with it. What was the incidence of type 1 diabetes in this population in 2014?
= 30/ (1500 - 180) = 30/1320 = 0.023
= 2.3%
(remove the people who are no longer new)
What is incidence attack rate?
Type of incidence used when nature of disease is acute & population observed for short period of time (e.g., outbreaks, specific exposures)
Attack rate = # new cases / # exposed
Secondary attack rate = # new cases / (# exposed – primary cases)
Measures person-to-person spread of disease after initial exposure
Within a kindergarten class, 5 of 35 kids develop chicken pox during a 1-week period. In the next two weeks another 10 kids also come down with chicken pox. What are the attack and secondary attack rates of chicken pox in the classroom?
Attack = (5+10)/35 = 15/35 = 0.43
Secondary attack = 10/(35-5) = 10/30 = 0.33
What affects Prevalence & Incidence?
-Duration of illness (longer => higher prevalence)
-Number of new cases (more new cases => higher prevalence)
-Migration
In-migration (ill => higher prevalence)
Out-migration (well => higher prevalence)
Recovery & death => lower prevalence
-Prevention => lower incidence
-Changes in diagnostic criteria or reporting (what is considered a case)
What is the relationship between prevalene and incidence?
Prevalence > incidence if disease is long term (e.g., diabetes)
Prevalence ≈ incidence if illness is acute (e.g., flu)
A group of individuals who were exposed to Lyme disease were screened using a new test developed for early detection. Of the 344 screened, the disease was confirmed in 258. The new test detected 263 cases of Lyme disease, 32 of which were disconfirmed. What is the sensitivity of the new test?
=231/(258) = 0.90
True positives/total #ill
What does it mean to have a highly sensitive (.90) but not highly specific (.63) test?
***High sensitivity => err on the side of over-diagnosing
Identify most or all possible disease cases
Most useful when under-diagnosing may lead to severe consequences (e.g., fast developing cancers)
***High specificity => err on the side of under-diagnosing
Identify most or all well people
Most useful when over-diagnosing may lead to dangerous, painful, or unnecessary treatment
What is predictive value?
-Probability that a test will give the correct diagnosis
-Depends on:
test sensitivity and specificity
prevalence of the disease in the population being tested
-Predictive values will vary from population to population, and study to study
What is the positive predictive value?
Probability that a person who tests positive for a disease truly has it
= True positives / All positives [true + false positives]
What is the negative predictive value?
Probability that a person who tests negative for a disease truly is well
= True negatives / All negatives [true + false negatives]
What affects predictive value?
-Depends on prevalence
-Higher disease prevalence
Higher positive predictive value (i.e., greater chance that positive test result reflects true illness)
Lower negative predictive value (i.e., lower chance that negative test result reflects disease-free status)
-Lower disease prevalence
Lower positive predictive value (i.e., lower chance that positive test result reflects true illness)
Higher negative predictive value (i.e., greater chance that negative test results reflects disease-free status)
Why are risk reduction and number-needed-to-treat important?
Relevant when comparing effects in randomized controlled trials
Interest is understanding risk of treatment vs. no treatment
What is frequency of bad outcomes in group being treated compared to the group not being treated?
What is Control event rate (CER)?
Proportion of control group participants who have a bad outcome after “treatment” (e.g., placebo or no treatment)
For example…
If 10 of 30 control group participants become sicker,
CER = 10/30 = 0.33 = 33% had adverse outcomes
What is experimental event rate (EER)?
Proportion of treatment group participants who have a bad outcome after treatment (e.g., new drug)
For example…
If 4 of 30 treatment group participants become sicker,
EER = 4/30 = 0.13 = 13% had adverse outcomes
What is absolue risk reduction/increase?
…Reduction (ARR)* when CER > EER
Higher rate of adverse outcomes in control group
…Increase (ARI) when CER
What is relative risk reduction/increase?
Reduction/increase: difference in 2 event rates, as a proportion of the event rate in the control group
= 1 – RR
Reduction (RRR) when CER > EER
Increase (RRI) when CER
