Epidemiology: Key Concepts and Measures
Proportion: Numerator is always a subset of the denominator; dimensionless (0 to 1 or 0% to 100%).
Rates: Describe changes in one quantity per unit of time. Unit = 1/time. Range (0-infinity).
Risk
Proportion | Dimensionless (can be expressed as a percent) | Appropriate for fixed populations with minimal losses to follow-up because we assume everyone was followed for a specific period.
- New Cases: Numerator. New, non-existing cases. For diseases that can occur more than once, it is the first occurrence of the disease.
- Population at Risk: Those in the denominator can’t already have the disease, be immune, and should have the relevant organs.
- Follow-up: Period population at risk is monitored the same.
Incidence Interpretation (must know the time period of the study)
Rates
Units 1/time | Numerator for risk and rate are the same; the difference is the denominator.
- Population can be dynamic.
- Don’t assume the same follow-up time for all members of the population.
Person-time: The amount of time an individual remains in the study population; can be person-days, weeks, months, or years.
Total person-time is the sum of all study time contributed from all individuals in the study population.
Example: 500 HIV- men were recruited between 1985 and 1991. The men were followed and at risk for a total of 2200 person-years, and 46 new cases were observed.
Rate = 46/2200 person-years = 0.021/year | 0.021 x 100 = 2.1 cases per 100 person-years
Risk and Rates
Risk of disease: Probability of developing the disease/time | Refers to a specific population over a specific time | Average probability of developing the disease given in group x.
Risk = Incidence rate * time (CI = ID * t)
Rate of disease: Force of morbidity | IR expressed as the number of events per unit person-time. For example, the rate of AIDS in Kenya per 1000 population per year.
Prevalence
Number of existing cases/population | Determines the distribution of current disease, service planning, and resource allocation; cannot separate the influence of incidence and survival.
Prevalence approximates incidence density multiplied by duration:
Prevalence = ID x D – if the disease is rare | P/(1-P) = ID x D (when not rare) ID is the incidence rate
D = P/ID
- Point prevalence: # existing cases at one specific point in time.
- Period prevalence: # cases during a specific time interval.
Incidence: Evaluates the effectiveness of a program.
Population
- Fixed: Survivors of the Hiroshima blast (you’re not getting any more). Enrollment can be a defining moment for entering a fixed population.
- Dynamic: Residents of a town, gym members, or members of a health insurance plan (variable).
- Steady State: Equal in | Equal out (death = birth).
Component Causes
- Necessary cause: Must always precede an effect (HIV is necessary for AIDS). A cause doesn’t need to be the *only* cause in order to be necessary.
- Sufficient cause: Causes the effect, but is not the only thing to cause the effect (smoking can cause lung cancer, but so can coal dust).
X + Z -> Y (X is necessary but not sufficient to cause Y because it also needs Z, or Y will not occur) X is necessary but not sufficient.
Each wedge is a component cause, and the whole circle is a sufficient cause. A component cause is a necessary cause if it is in all the sufficient causes.
Relative Measures
Represent the magnitude and direction of the relationship between exposure and disease.
- All ratios (0-infinity and no dimensions) (numerator not required to be related to the denominator) (null = 1).
Risk ratio: Relative Risk (RR) = Risk among exposed [a/(a+b)] / Risk among unexposed [c/(c+d)] RR = [a/(a+b)]/[c/(c+d)] must represent the magnitude of association.
Interpretation: The risk of (disease) among (exposed population) is (answer) times the risk of (disease) among (unexposed population) in this population over the study period.
Prevalence ratio: Same formula as RR | Difference due to the difference between risk and prevalence | Prevalence includes new and existing cases. PR = prevalence among exposed/ prevalence among unexposed. The difference between PR and CIR is that CIR has some implied timeframe for the number of developed cases. Interpretation: The prevalence of (disease) among (exposed population) is (answer) times the prevalence of (disease) among (unexposed population) in this population over the study period.
Incidence rate ratio: Rate among exposed / rate among unexposed.
Interpretation: The rate of (disease) among (exposed population) is (answer) times the risk of (disease) among (unexposed population) in this population over the study period.
Odds ratio: OR = p/ (1-p) | Odds = number of people with event/ number of people without event.
Disease OR: Odds of disease among exposed/ odds of disease among unexposed (a/b) / (c/d).
Interpretation: The odds of (disease) among (exposed) is (answer) times the odds of (disease) among (unexposed).
Exposure OR: Odds of exposure among disease/ odds of exposure among non-diseased. OR = (a/c)/(b/d).
Interpretation: The odds of (disease) among (exposed) is (answer) times the odds of (disease) among (unexposed).
Disease and Exposure OR are mathematically equivalent. When the disease is rare, OR will be close to RR or IRR.
Absolute Measures
Interpretation implies a measure of impact and a measure of association. Null = 0.
Risk difference: Risk among exposed – risk among unexposed Re = a/ (a+b), Rue = c/(c+d).
Interpretation (association): The difference in the risk of (disease) between (exposed) and (unexposed) is (answer). | (Exposed) have (answer) more risk of (disease) as compared with (unexposed).
Interpretation (causal): There are (number of) excess cases of (disease) per 1000 (exposed population) attributable to (disease) | If we assume a causal relationship between (exposure) and (disease), (number of) cases of (disease) per 1000 (exposed) would be eliminated if (exposed) had not (exposed themselves (i.e., smoked)).
Incidence rate difference: IRD = rate among exposed – rate among unexposed. Interpretation: The difference in the rate of (disease) between (exposure) and (non-exposed) is (answer).
Measures of Population Impact
- Attributable proportion exposed
- Attributable proportion total
