The Science Behind DYNHA®

Modern psychometric methods [e.g., Item Response Theory (IRT), Rasch] are statistical models of the relationship between a person's response on an item and his/her score on a concept, or latent trait, being measured. Once a model has been estimated for a pool of items, the model can be used to select items for a specific purpose (e.g., development of a fixed-length short form; development of a dynamic form).

In contrast to traditional psychometric methods that assume that measurement error - the random variation in measurement - is constant across the various levels of a given scale, Rasch and IRT models treat measurement precision as varying over the score range. Thus, a specific estimate of the measurement error can be given for each person at each scale level. According to Rasch and IRT, measurement error is determined by the constellation of the item characteristics (difficulty and discrimination parameters) answered by the respondent. Basically, an item is most informative for people whose level of health is closest to the difficulty of that item. Therefore, measurement precision may be several times greater at levels of the ruler where many item thresholds are lumped together compared to other parts of the scale.

The use of Rasch and IRT models enables the DYNHA® system to appropriately estimate the level of measurement error around an individual's score. In addition, the DYNHA® system has the capacity to use a computer algorithm that will identify the optimal selection of items for a given individual's level of health. The result is a more precise estimate of the individual's health score.
Furthermore, Emedilab‘s partner’s DYNHA® system uses modern psychometric methods to improve the accuracy of scoring for respondents with missing data. Research to date has shown that estimation techniques based on Rasch and IRT models are more precise than current estimation methods that rely on the simple averaging of non-missing data and that a substantial proportion of those currently lost due to missing data can be recovered. While current estimation methods require an individual to have complete responses to at least one-half of the questionnaire items in a particular scale, Rasch and IRT methods can estimate a score with as little as one item response per scale. Consequently, more data can be recovered with the use of Rasch and IRT methods than current methods based on the sum-score approach.

Modern psychometric methods allow for tests to be individually tailored, or adapted, for test respondents. Respondents can be ranked on the same continuum, regardless of whether they have been presented any items in common during assessment. In computer adaptive testing, a computer is used to administer a tailored test (Wainer et al., 1990). Modern psychometric methods, then, effectively complement the development of computer adaptive tests.

Drawing on a pool of items from widely used health surveys (general and disease specific), DYNHA® designs a brief assessment by asking only those questions relevant to the individual respondent's health state. By scoring all responses on a standard metric, results can be compared for those who answer different questions. The brevity of the assessment means that the DYNHA® system determines the scores on the "health ruler" at a fraction of the cost of traditional health assessments.

Item Pool Development

These item pools are developed though fielding of items designed to assess general health, disease/condition impact and demographics. Typically, the SF-8™ survey is presented first, followed by a set of developmental disease/condition impact items and demographic items (items selected from our partner’s Background Information Survey -- i.e., age, gender, chronic disease condition checklist, global disease impact items). The developmental impact assessment includes existing, modified and experimental items for fielding. Existing items are selected from the most widely used survey instruments for a given disease/condition. Each instrument is evaluated in terms of item content, and approximately 20-30% of unmodified items are included in the new item pool in order to anchor the new tool to existing instruments. Modified items alter selected existing items to more closely approximate the standards for survey development. Experimental items are developed by research scientists. Empirical tests are conducted to examine the modified and experimental items prior to inclusion in the final pool.

The developers’ latest initiative in item pool development is comprised of five disease conditions (asthma, congestive heart failure, rheumatoid arthritis, osteo arthritis, and rhinitis), as well as an assessment of patient satisfaction, health risk and global disease impact for close to 30 disease conditions. Data for these conditions is being gathered by a professional polling firm.
Survey access is free for non-commercial use. The information gathered on the site is kept confidential.