Consider two health workers presenting findings from the same community in northern Ghana. The first reports that 38% of women of reproductive age have iron deficiency anaemia - a figure drawn from a cross-sectional survey conducted during the dry season, when dietary diversity is lowest and the pool of untreated cases has accumulated over months. The second reports that the rate of new anaemia diagnoses in the antenatal clinic is 14 per 100 woman-years of observation - a figure drawn from longitudinal surveillance that tracked women from first antenatal visit through delivery. Both numbers describe the same underlying condition in overlapping populations. They are not comparable, they are not interchangeable, and they address quite different policy questions. The first is a prevalence estimate; the second is an incidence rate. Understanding what each measures, how each is calculated, and when each is appropriate is one of the foundational competencies of epidemiological practice.

This article works through the incidence vs prevalence distinction with precision, covers the mathematical relationship between the two measures, and draws on examples from Sub-Saharan African nutrition surveillance to ground the concepts in practice.

Prevalence: Measuring the Burden at a Point in Time

Prevalence is the proportion of a defined population that has a given condition at a specified moment or during a specified period. It is a proportion - a dimensionless number between zero and one, usually expressed as a percentage or per some standard population size (per 1,000, per 100,000).

Point prevalence is measured at a single moment: the proportion of individuals with the condition on a given day or during a brief survey period. Most national nutritional surveys - the Demographic and Health Surveys (DHS), the Multiple Indicator Cluster Surveys (MICS), and the national micronutrient surveys conducted in SSA countries - are designed to estimate point prevalence. When a DHS report states that 29% of children under five in a given country are stunted, it is reporting point prevalence: that proportion of children who, at the time of the survey, had height-for-age Z-scores below −2.

Period prevalence is the proportion with the condition at any point during a defined period - say, the proportion of a population that experienced an episode of severe acute malnutrition at some point during a given calendar year. Period prevalence is larger than point prevalence for the same condition and period because it captures both those who were ill at the period’s start and those who developed the condition and may have recovered during the period.

Prevalence is the appropriate measure when interest lies in the current burden on health services, the scale of need for treatment or support, or the proportion of a population affected by a condition at a given time. A policymaker deciding how many therapeutic feeding centres to fund in a region needs prevalence data; incidence alone, without knowing how long cases remain untreated and how quickly new cases arise, cannot answer that question ( Szklo & Nieto, 2014 ).

Incidence: Measuring the Rate of New Disease

Incidence quantifies the rate at which new cases arise in a population that is at risk of developing the condition. Two formulations of incidence are used in practice, and conflating them is a common source of confusion.

Cumulative Incidence (Risk)

Cumulative incidence is the proportion of an initially disease-free population that develops the condition during a defined observation period. If 1,000 children free of anaemia at birth are followed for one year and 80 develop anaemia, the cumulative incidence is 80/1,000 = 0.08, or 8%. Cumulative incidence is a proportion and requires that the observation period be specified alongside the estimate - an 8% one-year cumulative incidence is a fundamentally different number from an 8% five-year cumulative incidence.

This measure assumes complete follow-up of the entire population. When some participants are lost to follow-up, withdrawn, or censored before the observation period ends, cumulative incidence cannot be calculated directly and must be estimated using survival analysis methods (the Kaplan-Meier estimator, for example).

Incidence Rate (Incidence Density)

The incidence rate accounts for variable follow-up by dividing the number of new cases by the total person-time at risk contributed by all study participants:

$$\text{Incidence Rate} = \frac{\text{Number of new cases}}{\sum \text{Person-time at risk}}$$

Person-time at risk is the sum of each individual’s time under observation from study entry until outcome occurrence, loss to follow-up, or end of study, whichever comes first. Its units are person-time (person-years, person-months, or person-days). The resulting incidence rate carries units of cases per person-time - for example, 14.3 cases per 100 person-years - and is a rate rather than a proportion.

The person-time formulation is more versatile than cumulative incidence when follow-up is variable, as is almost always the case in population-based research. It makes full use of each participant’s observation time rather than discarding censored observations. HDSS systems, which continuously monitor defined populations and record dates of events with precision, are ideally suited to generating person-time denominators and therefore incidence rates ( Streatfield et al., 2014 ).

The Relationship Between Incidence and Prevalence

Incidence and prevalence are not independent. Their relationship is captured by a simple equation that applies under steady-state conditions - that is, when prevalence is stable over time and the population is not changing dramatically:

$$P \approx I \times D$$

where P is prevalence, I is incidence rate, and D is the average duration of the condition. This relationship, sometimes called the prevalence-incidence relationship or the steady-state approximation, has several important implications.

A condition with high prevalence may have either high incidence or long duration (or both). Iron deficiency anaemia, for instance, has high prevalence in SSA not only because new cases arise frequently - due to low dietary iron bioavailability, high parasitic infection burden, and elevated physiological requirements during pregnancy - but also because cases persist for months or years in the absence of effective treatment. The global burden of anaemia, estimated at 1.93 billion affected individuals in 2013, reflects both a high rate of new-case development and a long average duration of untreated disease ( Stevens et al., 2013 ).

Conversely, a highly lethal condition or one that resolves rapidly may have low prevalence even with high incidence. Severe acute malnutrition (SAM) in emergency settings may have high incidence - many new cases per week - but relatively low point prevalence if affected children either recover quickly through treatment or die. A surveillance system that measures only prevalence in such a setting may underestimate the true burden of acute disease.

An intervention that reduces duration - by improving treatment uptake and adherence - will reduce prevalence even if it has no effect on incidence. An intervention that reduces incidence - such as iron-rich food supplementation or malaria prevention - will eventually reduce prevalence, but with a lag determined by the average duration of existing cases ( Rothman, 2002 ).

Steady State and When the Approximation Breaks Down

The prevalence-incidence equation applies reliably only when the population is in a steady state: new cases accumulating at the same rate as old cases are resolved (through recovery or death), and no large exogenous shocks are altering the pool of cases. This condition is rarely met perfectly in practice.

During the acute phase of a famine, SAM incidence rises sharply before treatment programmes can scale up, and prevalence climbs because the new-case rate exceeds the recovery rate. When an effective intervention is rolled out rapidly - as happened with Community-based Management of Acute Malnutrition (CMAM) expansion in the Sahel - prevalence falls faster than incidence alone would predict, because case duration is shortened by treatment. Interpreting changes in prevalence over time requires attention to which component - incidence or duration - has changed ( Sankoh & Byass, 2012 ).

Population migration introduces additional complexity. If a region receives a large influx of food-insecure migrants, prevalence of undernutrition may rise sharply even without any change in the local incidence rate, simply because the incoming population carries a higher burden of pre-existing disease. HDSS systems that carefully track in-migration and out-migration can distinguish these effects; surveillance systems that do not are vulnerable to misinterpretation.

Applications in Sub-Saharan African Nutrition Surveillance

Iron deficiency anaemia illustrates the incidence vs prevalence distinction with particular clarity in SSA contexts. Regional estimates from the Global Burden of Disease study and the WHO Global Nutrition Report indicate that SSA carries the highest age-standardised prevalence of anaemia among women of reproductive age worldwide, with point prevalence estimates exceeding 40% in many West and Central African countries ( Kassebaum et al., 2014 ).

Cross-sectional DHS data, which generate prevalence estimates, have been the primary tool for tracking anaemia burden at country level. These surveys measure haemoglobin concentration in a representative sample, classify results against WHO cut-offs (< 120 g/L for non-pregnant women; < 110 g/L for pregnant women), and report the proportion below the threshold. The resulting prevalence estimate is operationally useful for planning treatment programmes and assessing national burden, but it cannot distinguish between a population in which anaemia develops rapidly and resolves quickly (high incidence, short duration) and one in which anaemia develops slowly but persists for years (lower incidence, long duration). The two populations require entirely different interventions.

HDSS sites have provided a partial solution by generating longitudinal haemoglobin data on defined cohorts, allowing calculation of incidence rates and recovery rates from prospective data. The Kintampo and Navrongo HDSS sites in Ghana, for example, have contributed data on anaemia trajectories through pregnancy and the postpartum period that cross-sectional surveys cannot replicate ( Streatfield et al., 2014 ).

The 2013 Lancet series on maternal and child nutrition synthesised both prevalence and incidence data to estimate the global burden of stunting, wasting, and micronutrient deficiency and the fraction attributable to inadequate dietary intake, infectious disease burden, and suboptimal breastfeeding practices ( Black et al., 2013 ). The series noted explicitly that reliance on cross-sectional prevalence estimates - while unavoidable given data constraints - limited the ability to characterise the dynamics of nutritional status over the life course.

The WHO’s global nutrition surveillance framework, outlined in the Global Nutrition Report, recommends that countries routinely collect both cross-sectional prevalence data (through DHS-type surveys) and longitudinal incidence data (through HDSS or linked cohort platforms) to enable the kind of trend analysis that policy requires ( WHO, 2020 ). The two data streams answer different questions and should be treated as complementary rather than redundant.

Why the Distinction Matters for Policy

Public health policy depends on clear thinking about which measure is relevant to a given decision. The incidence vs prevalence distinction shapes at least three major policy questions.

Resource allocation for treatment: The number of people currently requiring treatment for a condition is a function of prevalence, not incidence. Procurement of therapeutic foods, staffing of nutrition rehabilitation units, and logistics planning all depend on knowing how many cases exist at a given time.

Evaluating preventive interventions: An intervention that prevents new cases - such as iron fortification of staple foods, or malaria chemoprevention in pregnancy - reduces incidence. Its impact on prevalence will lag behind its impact on incidence by a period roughly equal to the average duration of disease. An evaluation conducted before this lag has elapsed may underestimate the intervention’s true effect.

Distinguishing treatment effects from prevention effects: A decline in prevalence over time could reflect improved treatment (shorter duration per case) or improved prevention (lower incidence), and the policy response to each is different. Separating these effects requires data on both incidence and duration - data that point prevalence estimates alone cannot provide.

For further methodological context on the epidemiological methods that generate these measures, see Epidemiology: Definition, Core Methods, and Applications and the discussion of surveillance infrastructure in Implementing HDSS .

Limitations

Several limitations affect the incidence and prevalence estimates discussed in this article.

Cross-sectional surveys that generate prevalence estimates are typically conducted during defined survey windows, often during the dry season when dietary diversity is lowest and anaemia prevalence highest. Seasonality means that a single-round survey may not represent the annual average burden. Surveys conducted in different seasons produce non-comparable prevalence figures, complicating trend analysis.

Haemoglobin-based anaemia diagnosis conflates iron deficiency anaemia with anaemia from other causes (malaria, sickle cell disease, folate deficiency). Haemoglobin concentration alone overestimates iron deficiency anaemia prevalence in high-malaria settings. Point-of-care ferritin and C-reactive protein measurement would improve specificity but is not yet standard in large-scale surveys.

Incidence data from HDSS sites reflect populations in defined surveillance areas that are not representative of national populations. Generalisability from HDSS to national estimates requires careful consideration of selection differences. The person-time calculation underlying incidence rates assumes that the hazard of developing the outcome is constant over time within exposure strata - an assumption that may not hold during seasonal transitions or acute shocks.

Frequently Asked Questions

What is the key difference between incidence and prevalence? Incidence measures how quickly new cases arise - it is a rate of change, quantifying the flow of new disease into the pool of affected individuals. Prevalence measures the existing stock of cases at a point in time. Incidence requires a disease-free population at the start and a defined observation period; prevalence simply counts who has the condition now. Both are essential, but they address different questions and should not be used interchangeably.

Why can iron deficiency anaemia have very high prevalence but moderate incidence? Prevalence equals incidence multiplied by average duration (under steady-state conditions). Iron deficiency anaemia in many SSA settings has long average duration because access to treatment is limited, dietary sources of bioavailable iron remain low, and the underlying drivers - parasitic infections, repeated pregnancy, low dietary diversity - persist chronically. Even a moderate rate of new case development, compounded over years without effective treatment, produces high point prevalence.

How is person-time calculated in an incidence rate? Each participant contributes time to the person-time denominator from the moment they enter the study until they develop the outcome, are lost to follow-up, or the study ends - whichever comes first. Someone followed for 18 months without developing the outcome contributes 1.5 person-years. Someone who develops the outcome after 6 months contributes 0.5 person-years and counts as one case in the numerator. Summing across all participants gives the total person-time denominator.

How does the prevalence-incidence relationship guide intervention design? The relationship P ≈ I × D tells policymakers that prevalence can be reduced either by lowering incidence (preventive interventions) or by shortening duration (treatment interventions). Understanding which pathway is more feasible - and more cost-effective - in a given context requires knowing both the incidence rate and the average case duration. Surveys that measure only cross-sectional prevalence cannot supply this information, reinforcing the value of longitudinal surveillance platforms.