exam2
case-fatiality rate
# of deaths / # of people with the disease
Case Fatality rate (CFR)
- among those diagnosed, who dies? example: Assume a population of 100,000 people of whom 20 are sick with melanoma, 18 of which die. Mortality rate? 18/100,000 = 0.018% Case-fatality rate? 18/20 = 90%
factors in expressing disease frequency
-number of ppl affected by health phenomenon -size of population giving rise to the cases (reference population= population at risk) -the time period in which the population is followed
indirect method
-population of interest -death rates in the standard population (don't have the population rate just population) *to get expected
counts
= number of cases or health events Can be important especially with infectious disease
Attack rate (AR)
Alternative form of cumulative incidence Used for diseases observed in a population for a short time period. Not a true rate because time dimension often uncertain. Example: Salmonella gastroenteritis outbreak
cumulative incidence-- why does the denominator have to be "at risk" individuals?
An example might help: Uterine cancer Numbers would be low if we included men (Exam question)
CFR vs mortality rate
CFR = deaths over disease Mortality rate = deaths over those at risk
review
CI = cumulative incidence (MUST be during a specific period of time, closed population assumed) ID = incidence density (open population assumed, calculate person-time) Attack rate = incidence rate over short period of time Prevelance = cross-section of a population
Two Types of Incidence Measures
Cumulative Incidence: Over a defined period of time, OR... Incidence Density: Over person-time (used when individuals can't be observed for the full time period)
Proportionate mortality
Deaths from CVD/all deaths X 100
Example of crude mortality rate
Example: Number of deaths in the United States during 2003 = 2,448,288 Population of the U.S. as of July 1, 2003 = 290,810,789 crude death rate= 2,448,288/290,810,789 = 841.9 per 100,000
how to calculate indirect age adjustment
First step, if you're interested, is to calculate age-specific numbers within each age group These numbers would be given to you You're taking the number of people in your population (firefighters for example) Then you multiply by the standard death rate This would also be given to you Or if you know the actual values, then you can calculate the age-adjusted rates using the direct method The result is the number of expected deaths in your population of firefighters, if your population had the same death-rate as the standard population example: if there 7,989 people in the population of interest in age 15-24 and the death rates (per 100,00) in standard population is 81.5, what is the expected number of deaths? 81.5/100,000 = .000815 .000815 x 7,989 = 6.5 deaths add all the expected deaths from all the age groups... observed deaths/ expected deaths example: Total population size = 230,109 persons Adjusted (expected) death rate = 987.9/230,109 = 429.3 deaths per 100,000 Unadjusted (observed) death rate = 502/230,109 = 218.2 deaths per 100,000 Ratio 218.2/429.3 = .508 X 100 = 50.8% Adjusted death rate = expected death rate Unadjusted death rate = actual death rate Then take observed over expected
If prevalence of disease is low:
Higher proportion of false positives Effect on PPV? Lower proportion of false negatives Effect on NPV? When prevalence of a disease falls, the PPV falls, and the NPV rises.
how to interpret SMR
If the observed and expected numbers are the same, the SMR would be 1.0, indicating that observed mortality is not unusual. An SMR of 2.0 means that the death rate in the study population is two times greater than expected. What does it mean if the SMR = 0.5 or 50%? the observed mortality rate falls below what is expected based on the standard population
incidence density
Incident rate when follow-up is incomplete In a dynamic or open population, individuals enter population over time, become lost, etc. Length of follow-up not uniform Does not make assumption of complete follow-up
indirect method
Indirect method may be used if age-specific death rates of the population for standardization are unknown or unstable Used to study difference in mortality rate between general population and specific occupational group So if you want to know if firefighters have a higher mortality rate than the general population, you would generally use this method The standardized mortality ratio (SMR) can be used to evaluate the results of the indirect method.
continue example
Let's look at the same data that we looked at before Of the 180 that screened positive, 80 actually have the disease So the PPV is 44% Of the 820 that screened negative, 800 of them do not have the disease So the NPV is 98%
age specific mortality rate
Number of deaths due to particular disease/ population size at midpoint of time period x100,000 Every person in the denominator must be at risk of entering the group in the numerator!
Relative survival rate
Observed survival in ppl w/ the disease/ Expected survival if disease were absent
Prevalence vs incidence?
Prevalance = snap-shot (no measure of risk)Is not used as a measure of risk of disease; Estimating the frequency of an exposure {flow down river} Incidence = includes only new cases and also includes time so it's a measure of risk {flow down waterfall}
Adjusted rates
Problems comparing crude rates among populations: Groups differ with respect to underlying characteristics that affect overall rate of disease (especially age, sex, and race) and so you may be making an unfair comparison between populations Use adjusted rates instead of crude rates Comparing death rates between 2 or more populations when population age structures may differ. Age may CONFOUND the relationship
incidence
Quantifies number of new cases of disease that develop in a population at risk during a specified time period Three key concepts: New disease events, or for diseases that can occur more than once, usually first occurrence of disease Population at risk (candidate population) - can't have disease already, should have relevant organs Time must pass for a person to move from susceptible to diseased (follow-up) * closed population is assumed
standardized mortality ratio (SMR)
SMR= observed deaths/ expected deaths x100 Sample calculation: The number of observed deaths due to heart disease is 600 in a certain county during year 2003. The expected number of deaths is 1,000. The SMR = (600/1,000 x 100) = 60% What does 60% mean? (40% less than expected number of deaths)
maximize positive predictability
Screen high-risk populations More cases Higher proportion of true positives
RELATIONSHIP TO ERROR TYPES
TYPE 1 - False positive = incorrectly rejecting the null TYPE 2 - false negative = failure to reject a false null
Specificity
The ability of the test to correctly identify those who do not have the disease
Sensitivity
The ability of the test to correctly identify those who have the disease
year of potential life lost (YPLL)
Those that die younger represent a greater loss of productive years. (Pre-determined age at death) - (age at death) 65 years is often used as pre-determined age how to calculate: YPLL = Σ (E - ai) where: Σ = sum of E = the endpoint of interest ai = the midpoint of individual age category age population deaths 0-9 3,610 3 10-19 2,500 2 20-39 5,125 3 40-59 4,950 8 60-69 3,945 12 70 &over 2,900 13 YPLL= [3(70 - 5] + [2(70 -15)] + [3(70 - 30)] + [8(70 - 50)] +[12 (70 - 65)] = 645 years of potential life lost **Don't count deaths that occurred in those 70 years and older
rates
Time is an intrinsic part of denominator for true rate Otherwise the measure is a frequency Term is most misused (Get used to it) Should specify if measure represents events or people Ex. Number of motor vehicle injuries/1,000,000 passenger miles driven in 2008 Implicitly an annual rate
Crude Mortality Rate
Total number of cases in the population / the total number of individuals in that population at a specific time period
example of specificity and sensitivity
We have a hypothetical population of 1000 people 100 have the disease 900 do not We have a test that yield positive/negative results A dichotomy The antibodies are present or they are not We want to know: How good is the test? How good at ID'ing those that had the disease? Of the 100 with the disease, 80 were id'd as positive 80/100 = 80% sensitivity How good at ID'ing those that do not have the disease? 800/900 = 89% specificity
direct method
What is our standard population? Calculate a standard population Can be any population Often-times the national population is used In this case, it's the sum of the early and late periods Based on the crude mortality rates, how many people would we expect to die in our standard population? Apply the age-specific crude rate to the standard population How many total people die in the standard population? Sum the number of hypothetical deaths What is the new mortality rate? Divide the sum by the standard population hypothetical deaths/sum of standard population
practice problem: A city contains 100,000 people (45,000 males and 55,000 females), and 1,000 people die per year (600 males and 400 females). There were 50 cases (40 males and 10 females) of lung cancer per year, of whom 45 died (36 males and 9 females). What is the crude mortality rate? Sex-specific mortality rate for males? Cause-specific mortality rate for lung cancer? Case fatality rate for lung cancer?
What is the crude mortality rate? 1,000/100,000 X 1,000 = 10 per 1,000 Sex-specific mortality rate for males? 600/45,000 X 1,000 = 13.3 per 1,000 for males Cause-specific mortality rate for lung cancer? 45/100,000 X 1,000 = .45 per 1,000 Case fatality rate for lung cancer? 45/50 X 100 = 90%
scenario
You decide you want to know what the risk is of developing prostate cancer in men age 50-55 That population is your blue box But some of them will already have prostate cancer If you include them, your incidence rate will be inflated (Exam question) So instead you take them out and only study men age 50-55 for the next year
Example of age specific mortality rate
age-specific mortality rate due to pneumonia in 2003= 190,000/10,000,000 = 190 per 10,000
crude vs age-adjusted
crude methods: overall population & age-specific age-ajusted methods: direct (have population numbers) & indirect (don't have population numbers)
mortality
deaths from the disease
morbidity
disease occurrence
prevalence and how it effects specificity and sensitivity
example: Let's take a sample population of 1000 people Prevalence is 50% Thus 500 people have the disease Sensitivity and specificity are 50% Therefore, of the 500 that screened positive, 250 have the disease So the PPV is 50% 50% prevalence is rare, so let's look at an example where the prevalence is 20% Which would still be high! We keep sensitivity and specificity at 50% Now the PPV is 20% (100/500) This illustrates our previous point that as prevalence falls, so does PPV If we increase sensitivity to 90%, what happens Keep in mind, now we're getting better at detecting those that have the disease Now the PPV is 31% A modest increase Now we've taken sensitivity back down to 50%, but raised specificity to 90% Now PPV is 56% A huge increase Why? Because, with a low prevalence, most of our population is to the right of the vertical line (800 people) And they don't have the disease So if we get better at detecting the people that don't have the disease, then our PPV will rise
direct method
know the -population - death rate
Are age-adjusted rates the "real" rates?
no, These are hypothetical death rates that would have occurred in each period IF each period had the age distribution of both periods put together. Adjusted rates are good only for comparison -- alone they are meaningless.
Cause-Specific Mortality rate
number of deaths due to particular disease/ population size at midpoint of time period x 100,000
proportion
numerator is subset of denominator, often expressed as a frequency or a percentage Multiplied by 10X where X = 1000, 10,000, etc. Ex. of frequency: 53 deaths per 100,000 population Ex. of percentage: 67% of adult American population is considered overweight or obese A proportions tells us what fraction of the population of interest is affected (EXAM QUESTION)
Relative risk (RR)
provides a direct measure of association between exposure and outcome [A/A+B]/[C/C+D] = Ie/Io yes no total rate yes A B A+B A/(A+B) no C D C+D C/(C+D)
studying rare exposures: But what type of cohort is best-suited for this?
special cohort not population-based cohort
basic model of disease
susceptible<-> diseased -> not at risk<->
Negative Predictive value (NPV)
the proportion of individuals screened negative by the test who do not have the disease (d/c+d)
Positive Predictive value (PPV)
the proportion of individuals screened positive by the test who actually have the disease (a/a+b)