Metrics
An item bank is a large collection of questions or symptoms that represent the manifestation of a latent construct, disorder, or trait. The key difference between item banks and classical symptom scales is the application of Item Response Theory (IRT) models to generate information about the statistical relationship between a person’s underlying disorder severity score (or latent trait score) and the probability of endorsing a particular response option on each of the symptom indicators (i.e. the item parameters).
Our item banks seek to substantially improve the relevance and content validity of traditional self-report symptom scales developed using classical test theory. Content validity is maximised by collating items and developing the bank using a systematic process that seeks to cover all aspects of the construct as well as address a wide range of severity.
Currently we have developed item banks for eight common mental disorders: social anxiety disorder, panic disorder, obsessive compulsive disorder, post-traumatic stress disorder, adult attention deficit hyperactivity disorder, drug use, psychotic-like experiences, and suicidal thoughts and behaviours. We have also validated the NIH PROMIS Depression and Anxiety item banks in the Australian population and provide re-calibrated item parameters based on a weighted sample of an Australian community sample.
The individual item banks, scoring programs, and manuals are freely available for non-profit research or clinical settings.
We use the term "short scales" here to describe any form of brief scale (typically less than 10 items) that was developed for a specific purpose. Commonly, we develop short scales that are derived from the Mental Health Metrics item banks using data-driven techniques, such as IRT, regression, or decision trees.
Short scales seek to maximise efficiency and therefore reduce the unacceptable level of respondent burden that is often observed in research and clinical settings. From our item banks, we provide two types of short scales developed using two types of methods.
The first type seeks to provide efficient scoring of disorder severity along a dimension, in essence attempting to replicate the scores derived from the item bank using as few items as possible. Two methodological approaches are applied to achieve this: static administration, where a single brief set of optimal items are selected and administered to each response; and adaptive administration, where a computer is used to tailor the administration of items to each individual until a pre-determined stopping rule is met.
The second type seeks to provide efficient screening of disorders based on the presence of a few key items that are highly related to a diagnostic decision (such as criteria described in the American Psychiatric Association's Diagnostic and Statistical Manual 5th Edition). Again, short screening scales can be administered statically or adaptively. The static versions are developed using a range of regression or machine learning prediction analyses while the adaptive versions are developed using decision trees or a process called stochastic curtailment.
The individual short scales, scoring problems, and manuals are freely available for non-profit research or clinical settings. Additional short scales developed using a variety of alternative approaches are also available and links are provided to obtain a copy from the appropriate source.
In mental health there is a large number of scales available that supposedly measure the same thing but the scales differ in terms of length, response options, and content. These differences often make direct comparisons between different scales very difficult and in some cases inappropriate. In order to draw comparisons between scores, a common scale or metric is required so that a score on one scale can be converted or re-scored to match scores of another scale.
Statistical approaches developed primarily in educational testing are suitable to adjust scores from a variety of mental health scales to form a common and comparable metric. These approaches generate what is known as "cross-walk" tables and scoring programs to easily convert group mean scores or individual scores from one scale to another. These tables facilitate merging datasets or comparing different clinical samples that have been assessed and scoring using different scales.
The cross-walk tables and R-scoring programs for a variety of common metrics to equate different scales are freely avialable for non-profit research or clinical settings.
In addition to the item banks, short scales, and cross-walk tables, the Mental Health Metrics website also provides a repository for a variety of purpose-built scales and diagnostic interviews to measure a range of different psychopathological constructs and mental or substance use disorders. Each of these scales or diagnostic interviews has been developed in collaboration with members of the Mental Health Metrics research team.
Information about each of the scales and diagnostic interviews is provided where appropriate and the scales and diagnostic interviews are free to use for non-profit research or clinical settings.