Measuring disease

Learning goals

After this lesson, you should be able to:

  • Describe the difference between prevalence and incidence in terms of the concepts of disease risk and disease burden.
  • Evaluate whether information on prevalence or incidence is more relevant to a particular public health question.
  • Compute prevalence, cumulative incidence, and incidence rate from appropriate data.
  • Describe how the prevalence of a disease is affected by incidence and typical duration of disease.





Review of video material

Burden vs. risk

Central question: how “common” is a given disease?

Thinking about the word “common” involves contrasting the notions of disease risk and burden.

  • Risk: How likely is it for an individual to develop this disease within a given time span?
    • Concerned with new disease cases
    • Measured with incidence (cumulative incidence and incidence rate)
  • Burden: How much of the population is living with the disease now?
    • Concerned with existing disease cases
    • Measured with prevalence

Being careful about denominators

With prevalence and incidence measures, we need to be very careful about denominators. Everyone represented by the denominator must be capable of being counted in the numerator–otherwise, the measure gives a distorted view of how “common” the disease is.

Prevalence

Prevalence=# cases at a given time# who COULD be a case at a given time \text{Prevalence} = \frac{\text{# cases at a given time}}{\text{# who COULD be a case at a given time}}

Example

In a recent survey of 1000 people, 11 report having asthma. The estimated prevalence of asthma would be 11/1000 = 0.011 or 1.1%.

Interpretation: An estimated 1.1% of people are living with asthma right now.

Question: Suppose we are trying to estimate the prevalence of COVID at Macalester on January 1, 2024. Do you think that “total # of employees hired and students enrolled on Jan 1” is a good denominator? Explain.

Response

If the goal of estimating the COVID prevalence on that date is to inform immediate campus policy regarding public health measures, the most useful denominator should probably consist of people who are actually using campus. Students who will be on vacation for a while and students who are studying abroad should not be counted in the denominator. Similarly for employees who do not work during January.

Cumulative incidence

Cumulative incidence=# NEW cases that develop over a time span# who COULD be a case in that time span \text{Cumulative incidence} = \frac{\text{# NEW cases that develop over a time span}}{\text{# who COULD be a case in that time span}}

Example

A study followed 1000 children fully for the first 5 years of their lives. 12 develop asthma within that time. The estimated cumulative incidence would be 12/1000 = 0.012 or 1.2%.

Interpretation: We estimate that 1.2% of children will develop asthma in the first 5 years of life.

Incidence rate

Incidence rate=# NEW cases that develop over a time spanTotal person-years at risk \text{Incidence rate} = \frac{\text{# NEW cases that develop over a time span}}{\text{Total person-years at risk}}

  • If individuals in a study are followed up for different lengths of time, it is better to use the incidence rate rather than cumulative incidence so that the total time that individuals were followed up is more accurately represented.
    • Why? A study follows 1000 people for 5 years, but most individuals stop participating in the study by year 2. We’re really not observing the 1000 people for the full 5 years. There’s quite a lot of time in those 5 years that we’re not observing people who may develop disease. Using cumulative incidence would understate the risk of the disease.
Example

A study followed 1000 children for the first 5 years of their lives for a total of 2500 person-years of follow up. 12 develop asthma within that time. The estimated incidence rate would be 12/2500 = 0.0048 cases per person-year.

Interpretation: We estimate that 0.0048 new cases of asthma in children per person-year. Because 1/0.0048 = 208.33, we could also phrase this as 1 case per 208.33 person-years.


Relationship between prevalence and incidence

For conditions that are not very common (low prevalence):

Prevalence≈Incidence×Mean disease duration \text{Prevalence} \approx \text{Incidence} \times \text{Mean disease duration}

Why is this approximate relationship important?

  • Low prevalence doesn’t necessarily mean low risk.
  • Example: Males have a higher melanoma incidence, but males survive a shorter amount of time (see here and here) with it than females. Females might have a higher prevalence of melanoma than males even though their risk is lower.





Activity

Navigate to Moodle, and open the PDF titled 02_measuring_disease_activity.





Broader application of these ideas

I’m currently reading Attached by Amir Levine and Rachel Heller. This is a book about adult attachment theory in romantic relationships, and while it seems to be far removed from epidemiology, I found myself thinking about ideas from class this week when reading the early chapters.

Adult attachment designates three main “attachment styles” or manners in which people perceive and respond to intimacy in romantic relationships, which parallel those found in children: Secure, Anxious, and Avoidant. Basically, secure people fell comfortable with intimacy and are usually warm and loving; anxious people crave intimacy, are often preoccupied with their relationships, and tend to worry about their partner’s ability to love them back; and avoidant people equate intimacy with a loss of independence and constantly seek to minimize closeness. … All people in our society, whether they have just started dating someone or have been married for forty years, fall into one of these categories, or, more rarely, into a combination of the latter two (anxious and avoidant). Just over 50% are secure, around 20% are anxious, 25% are avoidant, and the remaining 3-5% fall into the fourth, less common category (combination of anvious and avoidant). Attached, pages 8-9

These percentages are definitely prevalences, but I wondered about them because another part of the book discusses how people can change their attachment style. I wondered about what population these percentages are actually a snapshot of and whether (through this book!?) those percentages might likely change.

My partner also floated the claim that the dating pool might actually be enriched for people with avoidant attachment styles. (He claimed that whatever the prevalence of the avoidant style in the general population, they form a greater percentage in the pool of people who are currently dating.)

  • This made me think about our bathtub analogy for the relationship between prevalence, incidence, and average disease duration.
  • We can think of the water level as the dating pool. Entry into that pool (incidence) might be due to avoidant-attachers leaving their current relationships (through avoidance) and re-entering the dating pool.
  • Drainage from the pool might be enriched for secure-attachers who find relationships and stick with them, thus leaving the dating pool for a long time (or forever).