Skip to Content
Interactive Textbook on Clinical Symptom Research Logo


Home Button

Statistical Models for Prognostication
Author Bio
Introduction
Predictions: Statistical Models
Insight: Statistical Models
Currently selected section: Ingredients: Statistical Models
Theoretical Aspects
Central Concepts
Regression Models
Problems: Regression
Practical Advice
Example 1
Example 2




Chapter 8: Statistical Models for Prognostication: Ingredients of Statistical Models
        

Types of Outcome Variables

Outcome variables may be of several types. The most important types include continuous, dichotomous, categorical, and survival variables (Altman, 1991).

Continuous variables may be analyzed as interval data, i.e. that a difference in one unit has the same meaning for all values of the variable. An example is temperature, either in C or F, where a one-degree difference has the same meaning at zero or 100 degrees. Measurements of symptoms may sometimes be on a continuous scale. For example, a visual analogue scale (VAS) may be used for the rating of pain, but, whether such a scale truly represents an interval scale is debatable. As long as the researcher is explicit about the assumption made and thoughtful about whether it might be misleading, scales for pain may often be treated as interval scales.

Dichotomous variables take one of two values. For example, a risk factor may be present or not; complications after treatment occur or do not occur; patients survive until 30 days after admission or die before 30 days. The researcher must be thoughtful if a dichotomy is constructed. Often, dichotomies are arbitrary splits of physiologically continuous variables (e.g. age under or over 50 years).

Categorical variables have more than 2 categories. These variables can be analyzed as nominal data, when no ordering is present, or as ordinal data, when an ordering is present. For example, disease category at primary diagnosis might be a nominal variable, and patient satisfaction measured on a 5-point Likert scale might be an ordinal variable.

Survival variables may be used to analyze mortality over time, or the occurrence of complications in time. Survival variables actually consist of two components: a continuous time-to-event variable which indicates a duration, and a dichotomous censoring variable which indicates whether the event occurred, e.g. whether the patient died or had a complication. When the event did not occur, this is registered as "censoring", i.e. that the event had not occurred by the end of follow-up. Censoring may be due to the end of the study period, i.e. related to a certain calendar time. In this case, it may safely be assumed to be unrelated to the occurrence of the event. However, censoring is also coded for patients who are lost to follow-up (e.g. refusal to participate further, lost contact, etc.), which may have several underlying reasons that may be related to the event (e.g. progression of the disease).

Previous Page