Concept: Prevalence and Incidence

A. Definitions

1. Prevalence - the measure of a condition in a population at a given point in time (in this document referred to as point prevalence ).

Prevalence can also be measured over a period of time (e.g. a year). This second type of prevalence is called period prevalence ; it is a combination of point prevalence and incidence. Period prevalence is the most common measure of prevalence used in MCHP studies.

Prevalence data provide an indication of the extent of a condition and may have implications to the provision of services needed in a community. Both measures of prevalence are proportions - as such they are dimensionless and should not be described as rates (Friis & Sellers, 1999 ).


Formula:


* = during specified time period


2. Incidence - the number of new occurrences of a condition (or disease) in a population over a period of time.

The incidence rate uses new cases in the numerator; individuals with a history of a condition are not included. The denominator for incidence rates is the population at risk.


Formula:


* = during specified time period


Often expressed as X cases per a given population base (e.g. 10,000 or 100,000)

Even though individuals who have already developed the condition should be excluded, incidence rates are often expressed based on the average population rather than the population at risk.

In the case of chronic conditions, where most people appear to be at risk, the distinction between populations at risk and the whole population appears to be less critical (Friis & Sellers, 1999 ).

Incidence rates are useful in determining the causality of diseases. This type of measure has not previously been used in MCHP studies, but is starting to be used to define and compare disease specific cohorts (e.g. patients with acute myocardial infarction).


• The relationship between prevalence & incidence: An illustration
  
  
• A related issue: Measuring the burden of a disease .

B. Issues

When calculating incidence and prevalence, make sure that the data (individuals) are unduplicated during the period of measure.

In both incidence and prevalence a clear definition is required for the condition. This includes the detection as well as how to deal with ongoing or chronic conditions where there may be issues of latency, remission, cure/treatment, and recurrence. Answering the simple questions 'What is a condition?' and 'When is a condition?' before starting a study is important.

1. Incident Cases

An incident case can be determined based on the first occurrence in the data of a condition (e.g. first diagnosis) or of an event representing the condition. True incidence is often difficult to get since access to earlier data that might contain an indication of the condition may not be available.

Researchers must define a period of time in which to identify new cases. This period of time must be long enough to capture the condition, especially in the case of a rare diseases or diseases with a low rate of diagnosis. The disease may still be common but the diagnosis may not be made or occur in administrative data unless there is a comorbid or aggravated condition. Appropriate definition of incidence will be sensitive to the condition being studied.

2. Prevalent Cases


• Point prevalence - When calculating point prevalence, several years of data can still be used to identify the individuals having a condition. In this case a single point in time would be used to identify all of those individuals in the cohort that showed the condition at that time. There are still issues around conditions that are recurrent or of a finite length (i.e. not chronic) that have to be resolved.
  
  
• Period prevalence - Along with the issues of 'What' and 'When,' researchers must also answer ' Where '. This is important when using period prevalence since a single individual may have multiple ages and residence locations - even be lost to follow-up.
  
  The time period must also be determined. As with incidence, a long enough period must be used to identify the condition, but not long enough to exaggerate the actual prevalence in the population. A period that is too long will be more affected by mortality and other loss to follow-up issues. Most of the studies involving the occurrence of a disease at MCHP use a measure of period prevalence usually looking over two or three years of data (Robinson et al, 1997 ).

The primary issues that need to be resolved here are related to changes over time - e.g. changes in age and residence. MCHP researchers have taken several approaches to this problem. The two most common methods for identifying the residence location are:


1. The age and location of residence at the first or incident case in the study.
   
   
2. The most frequent residence location, usually RHA or other large area and age that matches the population denominator in the rates measure. The age in this case is generally set as the age during the central year of the study.

Issues related to age become especially problematic when there are limits on the data with regard to age. Which age is chosen to determine the cut off - age at start of study, age at incident, or age must be within the given range for the whole period of the study.

3. Calculation of Prevalence or Incidence

When calculating prevalence, researchers must also deal with the denominator issue - who is in the population? This includes questions of risk - is the measure directed at the whole population or only those at risk? In the case of period prevalence, the denominator may have changed over time due to loss to follow-up or migration.

Three alternative suggestions for use when calculating population based prevalence/incidence rates have been made when dealing with MCHP data. Typically three years of data have been used to identify conditions within the MCHP data, but other time periods can also be used. Researchers and programmers may find point or one year prevalence easer to calculate than longer periods.


1. Cohort type study using the whole population found to be resident in Manitoba across all three years and falling within the age limits defined for the study for the whole time.
   
   
2. Use of a single date (e.g. Dec 31) within the middle year for the population and identifying the location of residence by the most frequent occurrence and the age as of the population date. This the most common method used at MCHP.
   
   
3. Define the population based on the average number of individuals across each of the three years.

4. Sensitivity Testing

Currently no systematic comparison at MCHP has been made to determine the difference between each method but Roos et al. (1999) (Appendix A.)) showed very small difference between different denominator calculations. This same paper indicates that changes in the methods for screening diagnosis may have a greater impact. Unpublished work done as part of the First Nations deliverable showed a difference of approximately 3% between two methods of identifying diabetics (19% - 22%) across all of Manitoba. It is likely that the latter two methods for identifying the population will over-estimate the prevalence rate.

Researchers should be careful not to confuse a 3 year case definition with a 3 year prevalence measure. Three years of data (e.g. physician claims) can be used to identify cases. This information could then be used to calculate either point or one year period prevalence. Point and one year period prevalence is more commonly used as a measure of prevalence in the literature.

More Information:

Schoenback & Victor J (2000). Fundamentals of Epidemiology and Chapter 5. Measuring disease and exposure.