Scoring Instructions

TIP: It is preferable to use 1) an administration platform that automatically calculates scores, 2) the HealthMeasures Scoring Service, or 3) the Assessment Center API. These use the most accurate scoring method as each response to every item is used to calculate a person’s score. Scoring Manuals are a good second option for self- and proxy-report measures. They convert the raw sum on a measure to a final T-score.

The HealthMeasures Scoring Service

The HealthMeasures Scoring Service is a free, web-based application that scores an Excel file of raw participant responses and returns by email a file with calculated T-scores for all measures.

Watch a brief demonstration of the HealthMeasures Scoring Service.

The HealthMeasures Scoring Service is especially useful when:

• You administered a short form, but not all participants answered every question
• Different groups of participants responded to different items
• You have administered a new custom short form using a subset of items from one of the HealthMeasures (e.g., PROMIS ® ) item banks

How to use the HealthMeasures Scoring Service

• Download the Input Template .
• Add your data (all respondents’ individual answers to questions) for one HealthMeasure to the template (multiple timepoints are okay).
• Determine how much data is missing. In order to produce a precise score, a respondent must answer at least 4 items in measures that include 4 or more items. If a precise score is required, remove respondents who answered less than 4 items from the input template and do not calculate a T-score.
• Upload your spreadsheet to the HealthMeasures Scoring Service (if this is your first time, register as a new user).
• Access your email to receive a spreadsheet with calculated scores.

• HealthMeasures Scoring Service tutorial
• Tutorial to score:
  • PROMIS Profiles
  • PROMIS Global
  • PROMIS Psychosocial Illness Impact

  The Assessment Center API

  The Assessment Center API is a licensed web service that scores raw participant responses to produce T-scores. It also uses the most accurate scoring method (i.e., response-pattern scoring). It is installed on your desired server.

  It is especially useful when:

  • You want your data to remain behind your firewall
  • You administered a short form, but not all participants answered every question
  • Different groups of participants responded to different items
  • You have administered a new custom short form using a subset of items from one of the HealthMeasures (e.g., PROMIS) item banks

  The API can be hosted on a statistician’s computer and called as a web service from the desired statistical package for scoring.

  • SAS has Proc SOAP to allow the calling of a web service, which can also be hosted within the SAS environment.
  • R can use the httr library to call a web service. SPSS Modeler can call a web service via CLEF.
  • STATA and other software packages have similar capabilities.

  Administration Platforms that Calculate Scores

  Multiple administration platforms (e.g., REDCap, Epic) include HealthMeasures and automatically calculate scores.

  TIP: Select measures labeled with “auto scored” in REDCap.

  Scoring Translations

  Translations of self- and proxy-report HealthMeasures use the same scoring as the English measures. Follow the same instructions for scoring using the HealthMeasures Scoring Service, an administration platform that autoscores (e.g., NIH Toolbox app in Spanish), or score by hand using the Scoring Manual.

  Score Precision

  Significant Figures
  T-scores and associated Standard Errors (SE) should be reported as integers. Additional decimal places may be retained in interim computations but the reported value should be rounded to an integer to maintain the appropriate level of significant figures.

  Handling Missing Data
  Response pattern scoring, available in the HealthMeasures Scoring Service and automatically computed in many digital administration platforms can produce a T-score for a respondent as long as 1 item has been answered. However, for short forms with at least 4 items, a T-score may not be sufficiently precise if a respondent has answered fewer than 4 items. Depending on how the score will be used, you may want to discard the score or proceed with caution. For short forms with 1-3 items, all items should be answered to produce a sufficiently precise score.

  Using a raw summed score to T-score conversion table to compute T-scores by hand requires that all items are answered in order to produce a valid score.

  Calculate a Confidence Interval
  Use the T-score and Standard Error (SE) to calculate a confidence interval. A 95% confidence interval is common. A 95% confidence interval means there is a 95% probability that the true T-score is within this range. The formula for a 95% confidence interval is:

  For example, if T=52 and SE=2, the lower boundary of the confidence interval is (52 – (1.96*2) = 48 and the upper bound is (52 + (1.96*2) = 56.