Key Measures of Disease Frequency

Why This Matters

Every epidemiological study you'll encounter relies on these core measures, whether you're analyzing an outbreak, evaluating a treatment, or identifying risk factors. You're being tested on your ability to select the right measure for the right question: When do you use incidence versus prevalence? Why would a case-control study report odds ratios instead of relative risk?

These measures fall into distinct categories: frequency measures (how often disease occurs), mortality measures (how deadly it is), and association measures (how exposures relate to outcomes). Don't just memorize definitions. Know what each measure tells you, what study design it belongs to, and when you'd choose one over another.

---

Frequency Measures: Counting Cases

These measures answer the fundamental question: how much disease exists in a population? The key distinction is whether you're counting new cases (incidence) or all existing cases (prevalence). That difference determines everything from study design to resource allocation.

Incidence

• Measures new cases only: the number of people who develop a disease during a specific time period, capturing disease risk
• Expressed as a rate (e.g., 50 cases per 100,000 person-years), which accounts for both population size and time at risk
• Essential for evaluating prevention: if incidence drops after an intervention, you know fewer people are getting sick in the first place

Note the distinction between incidence rate (person-time in the denominator) and cumulative incidence (population at risk in the denominator). Cumulative incidence gives you a proportion (e.g., "5% of the cohort developed disease over 10 years"), while incidence rate gives you a true rate with person-time units.

Prevalence

• Captures total disease burden: all existing cases (new and old) at a single point in time or over a period
• Expressed as a proportion or percentage of the population, providing a snapshot for healthcare planning
• Influenced by three factors: incidence rate, disease duration, and mortality

That third point is worth sitting with. A disease with high incidence but quick recovery will have lower prevalence than one with moderate incidence but a chronic course. The relationship can be approximated as: $\text{Prevalence} \approx \text{Incidence} \times \text{Duration}$. This is why HIV prevalence rose when antiretroviral therapy extended survival, even as incidence stayed stable.

Attack Rate

• A special incidence measure for outbreaks: the proportion of an at-risk population that develops illness during a defined epidemic period
• Calculated as $\frac{\text{new cases}}{\text{population at risk}} \times 100$, typically expressed as a percentage
• Critical for outbreak investigation: comparing attack rates across groups (e.g., those who ate potato salad vs. those who didn't) identifies the exposure source

Despite the name, attack rate is technically a proportion (cumulative incidence), not a rate, since it lacks person-time in the denominator. You'll still see it called a "rate" everywhere in practice.

---

Mortality Measures: Quantifying Death

These measures assess how deadly a disease is, but they answer different questions. Mortality rate tells you about death in the population; case fatality rate tells you about death among those already diagnosed.

Mortality Rate

• Deaths per population over time: typically expressed as deaths per 1,000 or 100,000 individuals per year
• Enables population comparisons: you can compare mortality across countries, time periods, or demographic groups
• Drives public health priority-setting: diseases with high mortality rates command resources and policy attention

Be careful with crude vs. specific mortality rates. A crude mortality rate covers all causes and all ages. Cause-specific and age-specific mortality rates narrow the focus and are more useful for targeted comparisons.

Case Fatality Rate (CFR)

• Proportion of diagnosed cases who die: calculated as $\frac{\text{deaths from disease}}{\text{total cases of disease}} \times 100$
• Reflects disease severity and treatment quality: a high CFR indicates either a lethal disease, poor healthcare access, or both
• Varies dramatically by context: Ebola CFR ranges from 25–90% depending on outbreak setting and healthcare infrastructure; COVID-19 CFR varied by age, vaccination status, and hospital capacity

CFR depends heavily on how thoroughly cases are detected. If mild cases go undiagnosed, the denominator shrinks and CFR appears artificially high. This is why early-pandemic CFR estimates are often inflated.

Standardized Mortality Ratio (SMR)

• Compares observed vs. expected deaths: calculated as $\frac{\text{observed deaths}}{\text{expected deaths}} \times 100$, where expected deaths come from applying a reference population's age-specific rates to your study population
• Adjusts for confounders: controls for age, sex, or other demographic differences that would otherwise skew comparisons
• Interpretation is straightforward: SMR > 100 means more deaths than expected; SMR < 100 means fewer; SMR = 100 means mortality matches the reference population exactly

SMR is a form of indirect standardization. You'll also encounter direct standardization, where you apply your study population's age-specific rates to a standard population structure. Both methods accomplish the same goal of making fair comparisons across populations with different demographics.

---

Association Measures: Linking Exposure to Outcome

These measures quantify the relationship between an exposure (risk factor) and an outcome (disease). The measure you use depends entirely on your study design. This is a high-yield testing concept.

Relative Risk (RR)

• Compares incidence between exposed and unexposed groups: calculated as $RR = \frac{\text{incidence in exposed}}{\text{incidence in unexposed}}$
• Used in cohort studies where you follow groups over time and can calculate true incidence rates
• Interpretation: RR = 1 means no association; RR > 1 means exposure increases risk; RR < 1 means exposure is protective

For example, if lung cancer incidence is 200 per 100,000 person-years among smokers and 10 per 100,000 among non-smokers, $RR = \frac{200}{10} = 20$. Smokers have 20 times the risk.

Odds Ratio (OR)

• Compares odds of exposure between cases and controls: calculated as $OR = \frac{ad}{bc}$ from a 2×2 table (where a = exposed cases, b = exposed controls, c = unexposed cases, d = unexposed controls)
• Required for case-control studies: because you select participants based on disease status, you cannot calculate true incidence, so you cannot calculate relative risk directly
• Approximates relative risk when the disease is rare (generally under ~10% prevalence in the source population), making interpretation similar

The rare disease assumption matters. For common outcomes, OR will systematically overestimate the strength of association compared to RR. If prevalence is 30% and the true RR is 2.0, the OR might be around 2.5 or higher.

Attributable Risk (AR)

• The excess risk due to exposure: calculated as $AR = \text{incidence in exposed} - \text{incidence in unexposed}$
• Answers a different question than RR: tells you the absolute increase in disease attributable to the exposure
• Directly informs intervention impact: if you eliminate the exposure, AR estimates how many cases you'd prevent in the exposed group

RR and AR can tell very different stories. A risk factor might double your risk (RR = 2), but if baseline risk is tiny (say, 1 per 100,000), the absolute excess is only 1 per 100,000. Both the relative and absolute perspectives matter for clinical and policy decisions.

---

Population-Level Impact Measures

These measures extend individual risk to the entire population, helping policymakers decide where to focus resources for maximum public health benefit.

Population Attributable Risk (PAR)

• Estimates disease burden attributable to exposure in the whole population: accounts for both the risk associated with exposure and how common the exposure is
• Calculated as $PAR = \text{incidence in total population} - \text{incidence in unexposed}$, or using the formula $PAR = \frac{P_e(RR-1)}{1 + P_e(RR-1)}$ where $P_e$ is exposure prevalence (this version gives the population attributable fraction, the proportion of total cases attributable to the exposure)
• Guides public health priorities: a modest risk factor that's extremely common (like physical inactivity) may have higher PAR than a strong risk factor that's rare

---

Quick Reference Table

---

Self-Check Questions

1. A chronic disease has moderate incidence but patients live with it for decades. Would you expect prevalence to be higher or lower than incidence, and why?

2. You're investigating a foodborne outbreak at a wedding. Which measure would you calculate to compare illness rates between guests who ate the shrimp and those who didn't?

3. Compare and contrast relative risk and odds ratio: In what study designs is each appropriate, and under what conditions does OR approximate RR?

4. A risk factor doubles disease risk (RR = 2) but only 5% of the population is exposed. Another risk factor has RR = 1.3 but 60% of the population is exposed. Which likely has higher population attributable risk, and what does this tell you about intervention priorities?

5. A question presents mortality data from two countries with very different age distributions. Which measure would allow a valid comparison, and what does it adjust for?