- Posts: 75
There is also a question on the Forum providing more detail on how the scoring tables were constructed.
HealthMeasures instruments are scored using item-level calibrations. This means that the most accurate way to score a HealthMeasures instrument is to use the HealthMeasures Scoring Service ( www.assessmentcenter.net/ac_scoringservice ) or a data collection tool that automatically calculates scores (e.g., Assessment Center, REDCap auto-score, Epic PROMIS app). These approaches to scoring employ a common, highly accurate method that uses each participant’s responses to each item administered. We refer to this method as “response pattern scoring.” Because response pattern scoring is the most accurate method of scoring -- more accurate than the use of raw sum score/scale score look-up tables included in Scoring Manuals -- it is preferred. Response pattern scoring is especially useful when there is missing data (i.e., a respondent skipped an item), different groups of participants responded to different sets of items, or you have created a new questionnaire using a subset of items from an item bank.
The HealthMeasures Scoring Manuals include tables that convert a raw summed score from a fixed length short form or scale to a T-score. Creating these tables is a little tricky, though, because there are usually multiple ways respondents could answer items on a short form and obtain the same raw sum score. For example, on a 4-item short form, a person who responds 1, 2, 2, 1 would have the same raw sum score (6) as a person who responds 1, 1, 1, 3. Consequently, the look-up tables were constructed by first identifying the multiple ways a raw sum score could be obtained. Each of these “same raw sum score” response patterns was then scored using the response pattern scoring method as described earlier. The resulting set of T-scores was then averaged to produce the most reasonable T-score for a specific observed raw sum score. Look-up table T-scores are a “second best” method for scoring. They are especially useful when response pattern-based scoring is not available or logistically complex to access. Response pattern scores and look-up table scores will be highly similar. However, they will not be “exactly” the same, as slight differences will be observed between them, more so at the individual score level than at the group score level.