Incidence is the count of new cases of a disease or condition that arise in a population during a specific time period. It is fundamentally about onset: how many people who did not have a condition yesterday have it today, this month, or this year. That distinction between “new” and “existing” is the single most important thing to grasp about the term, because it separates incidence from its close cousin, prevalence, which counts all current cases regardless of when they started.
Incidence Versus Prevalence
The confusion between incidence and prevalence is so common that epidemiology textbooks typically teach them as a pair. Prevalence reflects the number of existing cases of a disease, while incidence reflects the number of new cases and can be reported either as a risk or as a rate.1PubMed. Measures of disease frequency: prevalence and incidence Think of it this way: if a bathtub is filling with water, incidence is the flow from the faucet (new water entering), and prevalence is the total water level at any given moment. The water level depends on how fast the faucet runs and how fast the drain removes water (through recovery, death, or cure). A disease can have high incidence but low prevalence if people recover quickly, like a common cold. It can also have low incidence but high prevalence if people live with it for decades, like type 2 diabetes.
This matters outside academia. When a news headline says “cancer rates are rising,” you need to know whether they mean incidence or prevalence. Rising incidence means more people are developing cancer. Rising prevalence could simply mean people with cancer are surviving longer, which is good news dressed in alarming language. The two measures answer different questions, and swapping them leads to very different conclusions about whether things are getting better or worse.
Two Ways to Measure Incidence
When researchers report incidence, they usually mean one of two things, and the difference is worth knowing because it changes what the number tells you.
The first is cumulative incidence, sometimes called incidence proportion or simply “risk.” This is the fraction of a defined group that develops the condition over a set period. If you follow 1,000 people for a year and 50 develop the flu, the cumulative incidence is 5% over one year. It is straightforward and intuitive, but it only works cleanly when you can track every person for the entire period. If people drop out of the study, move away, or die of something else before the observation window closes, the denominator gets murky.
The second is the incidence rate, also called person-time incidence. Instead of dividing new cases by the number of people, you divide new cases by the total time everyone was observed. A study might calculate incidence rates as the sum of all new episodes of illness divided by the size of the population measured in person-years.2PubMed Central. Calculating incidence rates and prevalence proportions: not as simple as it seems If 1,000 people are each followed for half a year, that is 500 person-years of observation. If 25 develop the disease, the incidence rate is 25 per 500 person-years, or 50 per 1,000 person-years. This approach handles the messiness of real studies better because it accounts for people entering and leaving the study at different times.
The practical takeaway: when you see an incidence figure, check whether it is a proportion (expressed as a percentage or “X per 100,000 people”) or a rate (expressed as “X per 100,000 person-years”). A proportion tells you the chance of getting sick over a defined window. A rate tells you the speed at which new cases appear. Both are called “incidence,” but they are not interchangeable.
Open and Closed Populations
The type of population being studied also shapes how incidence is calculated. A closed population, sometimes called a fixed cohort, starts with a defined group and follows them forward. Nobody new enters. Think of a clinical trial: you enroll 500 participants, and those 500 are the study. Cumulative incidence works well here because the denominator is fixed at enrollment.
But many real-world populations are open, or dynamic. People move into a city, babies are born, residents die or leave. A distinction has been made between open and closed populations because the characteristics of the group can shift depending on the time frame the researcher is examining.3International Journal of Epidemiology. Incidence rates in dynamic populations An occupational cohort, for instance, can be treated as fixed if you define membership as “everyone who ever started work at this factory,” but it is dynamic in terms of calendar time because workers join and leave at different points. For dynamic populations, incidence rates using person-time are the standard approach because they accommodate the constant flux.
This is not just a technical footnote. When public health agencies report the incidence of, say, heart disease in a metro area, they are dealing with a dynamic population. The number they publish depends on assumptions about how they define the denominator, whether that is the mid-year population, total person-years, or person-years restricted to those actually at risk. Those choices can shift the reported incidence meaningfully, even when the raw number of cases stays the same.
Attack Rates in Outbreaks
During disease outbreaks, you will often hear the term “attack rate” instead of incidence. Despite its name, an attack rate is technically a cumulative incidence: the proportion of people in an identified group who become infected by the end of an outbreak. In workplace outbreaks of COVID-19, for example, the attack rate was defined as the proportion of workers in a workplace that became confirmed cases by the end of the outbreak, with the number employed serving as the denominator.4BMJ Open. COVID-19 outbreak rates and infection attack rates associated with the workplace: a descriptive epidemiological study
Attack rates are useful precisely because outbreaks happen fast and in bounded settings: a school, a cruise ship, a factory floor. The population is more or less fixed for the duration, and the time window is short enough that you do not need to worry about person-time adjustments. If 30 out of 200 people at a wedding get food poisoning, the attack rate is 15%. That number instantly communicates severity in a way a person-time rate would not, because the event has a clear start and end.
How Incidence Data Gets Collected
Incidence figures do not appear out of thin air. They come from surveillance systems, and the quality of those systems determines how trustworthy the numbers are. Cancer incidence in the United States, for example, relies on a combination of the CDC’s National Program of Cancer Registries and the National Cancer Institute’s SEER program, together covering the entire country.5Preventing Chronic Disease. CDC’s Public Health Surveillance of Cancer Cancer registries act as a census of cancer cases, allowing researchers to define and monitor incidence at local, state, and national levels, investigate treatment patterns, and evaluate prevention efforts.6PubMed Central. The history and use of cancer registry data by public health cancer control programs in the United States
Globally, population-based cancer registries play a unique fundamental role in cancer surveillance, providing indicators of incidence and survival that risk factor surveys alone cannot deliver.7Epidemiologic Reviews. A Global Cancer Surveillance Framework Within Noncommunicable Disease Surveillance: Making the Case for Population-Based Cancer Registries In countries without strong registry infrastructure, incidence data is patchy or estimated from small regional samples, which means the numbers for certain diseases in certain regions carry wide uncertainty. When you see a global incidence map with neat color coding, keep in mind that the precision behind each country’s shade varies enormously.
For infectious diseases, surveillance tends to rely on mandatory reporting by clinicians and laboratories. If a doctor diagnoses measles, they are legally required to report it to public health authorities in most countries. These reports are aggregated into weekly or annual incidence counts. The system works well for dramatic, easily diagnosed conditions but underestimates diseases with mild or ambiguous symptoms that people never bring to a doctor.
When Rising Incidence Is an Illusion
One of the most common misinterpretations of incidence data involves screening. When a new screening test is introduced or an existing one becomes more widely used, more cases get detected, and incidence appears to jump. But the jump may reflect better detection, not a true increase in the underlying disease. Estimating trends in cancer incidence is often difficult because changes in screening or other detection processes over time can artificially inflate or deflate the observed numbers.8PubMed. Estimation of Cancer Incidence Trends Adjusted for Changes in Screening and Detection Processes
This problem has a formal name: surveillance bias. It arises when differences in how often a condition is found stem from changes in the method of detection rather than from a real change in risk.9PubMed Central. Screening and Surveillance Bias in Cancer The classic example is thyroid cancer. Incidence rates in many countries have climbed steeply over recent decades, but mortality has barely budged. The most likely explanation is that ultrasound technology improved and became more widely used, catching small, slow-growing tumors that would never have caused symptoms. The “epidemic” is largely an epidemic of detection.
Breast cancer in France followed a similar pattern. Research suggested that the reported “epidemic” of breast cancer could largely be explained by screening increasing the diagnosis of slowly progressing, non-life-threatening cancers and by misdiagnosis among women without progressive disease.10PubMed Central. An investigation of the apparent breast cancer epidemic in France: screening and incidence trends in birth cohorts This does not mean screening is worthless. It means that incidence figures from periods of expanding screening need to be interpreted with caution. A spike in incidence right after a new screening program launches is expected and does not, by itself, prove the disease is becoming more common.
The Lookback Window Problem in Chronic Disease
For chronic diseases tracked through medical records or insurance claims, the definition of a “new” case depends on how far back you look. If you define a new diabetes case as someone with a diagnosis code who had no diabetes code in the prior year, you will count many more “new” cases than if you define it as someone with no diabetes code in the prior ten years. The shorter your lookback window, the more existing patients get misclassified as new cases, inflating incidence.
Research using electronic health records and claims databases has shown that extending the lookback window from one year to the patient’s full medical history can reduce measured incidence by roughly two to three times, depending on the disease.11Dove Press / PubMed Central. Measuring prevalence and incidence of chronic conditions in claims and electronic health record databases This is a methodological choice, not a flaw, but it means that two agencies reporting incidence for the same disease in the same country can publish different numbers simply because they used different lookback periods. If you are comparing incidence figures across reports, check whether they define “new case” the same way.
How Incidence Guides Health Policy
Incidence data is the backbone of resource allocation in public health. If you want to know where to build new clinics, where to target vaccination campaigns, or how to divide a limited prevention budget across regions, you need to know where new cases are appearing fastest. In HIV prevention across sub-Saharan Africa, researchers have argued that using the heterogeneity of incidence across regions to inform resource allocation could enhance the impact of limited funding, directing combination prevention methods where the epidemic is most active rather than spreading resources evenly.12The Lancet HIV. Optimum allocation of resources for combination HIV prevention among sub-Saharan Africa’s general populations: a mathematical modelling analysis
Prevalence alone cannot do this job. A region with high prevalence but low incidence may have many people living with a chronic condition who need treatment, but the outbreak is not growing. A region with moderate prevalence but high incidence is where the disease is actively spreading and where prevention dollars are most urgent. Policy decisions based on prevalence when they should be based on incidence, or vice versa, can send resources to the wrong places.
Insurance and pharmaceutical industries also depend on incidence projections. When a drug company estimates the market for a new therapy, they start with how many new patients will be diagnosed each year. When an insurer sets premiums, incidence forecasts for expensive conditions feed directly into pricing models. The numbers ripple far beyond academic journals.
Mapping Incidence Geographically
Knowing that a disease has a certain incidence nationally is useful, but diseases rarely spread evenly across geography. Spatial epidemiology takes incidence data and maps it to specific locations, looking for clusters where new cases concentrate. In studies of typhoid fever, for instance, researchers used spatial scanning tools to detect statistically significant hotspots and coldspots of incident cases across study sites, revealing that transmission clustered tightly in certain neighborhoods rather than spreading uniformly.13The Journal of Infectious Diseases. Geographically Weighted Regression Modeling of Spatial Clustering and Determinants of Focal Typhoid Fever Incidence
These geographic analyses help public health agencies focus interventions where they will do the most good. If typhoid clusters around specific water sources, those sources can be targeted for treatment. If lung cancer incidence is elevated in a particular industrial corridor, that pattern can prompt environmental investigations. Without spatial resolution, the incidence number is an average that can mask the places where the real problem lives.
Competing Risks and Why Simple Incidence Can Mislead
In populations where people face multiple possible outcomes, simple incidence calculations can overestimate the risk of any single one. Consider an older population being studied for both heart disease and cancer. If you calculate the incidence of cancer while ignoring that some participants died of heart disease before they could develop cancer, you will overstate the cancer risk. Those who died of heart disease were no longer “at risk” for cancer, but a naive calculation treats them as if they were.
The statistical tool designed to handle this is the cumulative incidence function, which estimates the probability of a specific event while accounting for the fact that other events can remove people from the at-risk pool. Methodological work in this area has focused on ensuring that the estimated probabilities across all competing event types add up correctly, so that the total risk does not exceed 100%.14arXiv. Revisiting the Cumulative Incidence Function With Competing Risks Data For the general reader, the takeaway is simpler: when a study reports the incidence of a disease in an older population, ask whether competing causes of death were accounted for. If not, the number may look scarier than the real-world risk.
Incidence in Case-Control Studies
You might assume that incidence can only be measured by following a group forward in time, but certain study designs can estimate incidence-related measures even when working backward from diagnosed cases. In case-control studies that sample controls from the same dynamic population that produced the cases, the resulting estimates can be interpreted as incidence rate ratios rather than simple odds ratios.15American Journal of Epidemiology. Do Case-Control Studies Always Estimate Odds Ratios? This matters because incidence rate ratios are often more directly useful for understanding how much a risk factor speeds up the appearance of new cases. Research suggests this type of case-control design is the most common and also the most frequently misinterpreted, with analysts often treating the results as odds ratios when they are actually estimating something closer to an incidence rate comparison.
Wastewater and the Future of Incidence Tracking
Traditional disease surveillance depends on people visiting a doctor, getting tested, and having the result reported. Every step in that chain is a place where cases can be missed. Wastewater-based epidemiology offers a way around this bottleneck. By analyzing sewage for traces of pathogens, researchers can monitor the spread of infectious disease and the emergence of new outbreaks at the community level in something close to real time.16PubMed Central. Future perspectives of wastewater-based epidemiology: Monitoring infectious disease spread and resistance to the community level
This approach gained widespread attention during the COVID-19 pandemic, when wastewater monitoring detected surges in viral RNA days before clinical case counts climbed. It does not replace traditional surveillance, but it captures a signal from people who never get tested, including those with mild or no symptoms. For diseases where underreporting is the norm, wastewater data can provide a more honest picture of incidence than case reports alone. The method also has potential for tracking antimicrobial resistance, another domain where knowing the rate of new emergence matters enormously for public health planning.
As these systems expand, the definition of a “case” may shift. Instead of relying solely on a clinician’s diagnosis or a lab-confirmed test, future incidence estimates could incorporate environmental signals, creating a layered picture that combines individual case reports with population-level biological data. The core concept stays the same: how many new occurrences, in how many people, over what period of time. The tools for answering that question are simply getting sharper.