Dartmouth‐Hitchcock Medical Center, Dartmouth Geisel School of Medicine, Lebanon, NH, USA
Introduction
The World Health Organization defines an outcome measure as a “change in the health of an individual, group of people, or population that is attributable to an intervention or a series of interventions.”1 Outcome measurement is becoming increasingly important in an effort to document the value of interventions to both patients and society. Harvard Business School professor and healthcare policy expert Michael Porter has proposed a framework that defines value as health outcomes achieved relative to the costs incurred.2 Outcome measures are critical to both clinical research and public health, because this is the primary driver influencing one’s ability to answer important questions in a reliable manner. Outcome measures serve as the target that both researchers and healthcare organizations or governments monitor or attempt to modulate in an effort to improve the quality and/or cost of care. The past 30 years have seen a rise in interest in the measurement of the outcomes of medical care, to the extent that an “outcomes movement” has been described and been labelled “the third revolution in healthcare.”3 Types of outcome measures can be seen, broadly, as biophysical measures, like morbidity and mortality, or patient‐based measures that incorporate a patient’s subjective experience of illness.
Top three questions
What is an outcome measure?
What properties of outcome measures do I have to know?
How should I choose an outcome measure?
Question 1: What is an outcome measure?
An outcome measure is a measure of the health of an individual, group of people, or population. There are several broad categories of outcome measure:
Biophysical outcome measures: these are objective health measures. Some common examples include morbidity, mortality, complication rate, and quality of reduction.
Patient‐based measures: these outcome measures incorporate a patient’s subjective experience of illness. These may be generic or disease/joint‐specific.
Generic outcome measures are not specific to any one disease or anatomic location. They can be used to compare across or within specific pathologic conditions. These have the potential advantage of being more able to measure downstream consequences of a treatment or condition that permits comparison to other unrelated conditions. They also may measure the side effects of complications of a treatment or condition that occur in a different anatomic location. Two common examples of a generic patient‐reported outcome measure are the Short Form 36 (SF‐36) and the Patient‐Reported Outcomes Measurement Information System (PROMIS) Physical Function. Preference‐weighted outcome measures are a subset of generic outcome measures. These outcome measures weight items or dimensions differently to account for how people value a health state, rather than assigning equal weight to each dimension or item included in the outcome measure. Measuring preference‐weighted generic outcome measures is becoming increasingly important in an effort to optimally distribute resources in a resource‐constrained environment. The most commonly used preference‐weighted generic outcome measures are the EuroQol five‐dimensional questionnaire (EQ5D) and the six‐dimensional health state short form (SF6D).
Disease‐ or joint‐specific outcome measures may be more sensitive measures of the specific disease or joint being assessed. For example, joint‐specific outcome measures have been shown to be more sensitive to arthroplasty procedures compared to generic outcome measures.4–7 Common joint‐specific outcome measures in orthopedic surgery include the Western Ontario and McMaster Universities Arthritis (WOMAC) Index, Harris Hip Score (HHS) and the Hip Disability and Osteoarthritis Outcome Score (HOOS) Knee Injury and Osteoarthritis Outcome Score (KOOS).
Question 2: What properties of outcome measures do I have to know?
There are several properties and attributes that have to be considered when choosing appropriate outcome measures.
Reliability: the degree to which a score or other measure remains unchanged upon test and retest, across different interviewers or assessors, or across items on the same test.
Forms of reliability include:
Internal consistency: measures whether several items or questions that propose to measure the same general construct produce similar scores.
Test–retest reliability: measures whether the same person receives the same score on the same test at different time points.
Intra‐rater reliability: the degree of agreement among repeated administrations of an outcome measure assessed by a single rater.
Inter‐rater reliability: the degree of agreement between different administrations of the same outcome measure.
Measured by statistics:
Kappa: a statistic which measures inter‐rater agreement for categorical items.
Intra‐class correlation coefficient: descriptive statistic which describes how strongly units in the same group resemble each other.
Validity: the degree to which a measure or tool actually measures what it is intended to measure.
Variability: distribution of values associated with an outcome measure in the population of interest.
Broader range of values shows more variability.
Ceiling and floor effects are a measurement limitation that occurs when the highest and lowest possible scores on an outcome instrument does not reflect the true range of the domain being tested. This can lead to a mismatch between the distribution of responses and the true distribution of the concept of interest in the population. For example, if a patient reported outcome (PRO) instrument only assesses physical activities that are easy to perform, and the majority of the population scores perfect, then the instrument will not reflect the true distribution of physical abilities.
Responsiveness: ability to detect change in the underlying construct, even if changes are small.
Minimally important difference (MID).8,9
Smallest change in an outcome that a patient would identify as important or meaningful.
This is an important property because, given a large enough sample size, statistical significance between groups may occur with very small differences that are clinically meaningless.9
Only gold members can continue reading. Log In or Register to continue
