Preference-based Scores

Preference-based scores provide an overall summary of health-related quality of life on a common metric.

Preference-based scores summarize multiple domains into a single score anchored at 0 (as bad as dead) and 1 (perfect or ideal health). The advantage of preference-based measures is their ability to prioritize health interventions by overall impact and to measure changes in health for comparative-effectiveness and cost-effectiveness analyses. For example, preference-based scores can help estimate the relative value of two treatments with different outcomes (see image below).

Different Types of Summary Measures

Different approaches are used to describe overall health-related quality of life. One strategy is to use a profile measure (e.g., the PROMIS-29). Profile measures assess multiple domains of health such as physical functioning, depression, and pain. Profile measures then provide multiple scores – one for each domain assessed. Each score provides information to understand health in detail and together, the scores provide a broader, more comprehensive description of health.

A second strategy is to use a global measure (e.g., PROMIS Global Health). A global measure combines information about multiple domains of health into a summary score. For example, the PROMIS Global Health scale produces a score for physical function that is based on 4 items about fatigue, pain, and physical function and a score for mental health that is based on 4 items about overall perceptions of quality of life, mental health, social activities and relationships. Global measures can be helpful when comparing large groups. They may be less helpful in providing actionable information for an individual patient in a clinical encounter. It is also difficult to tease apart the contribution of specific domains like fatigue versus pain, particularly when interpreting change over time. Different global health measures utilize different metrics unlike preference-based scores which all use 0 to 1.

A third strategy is to use a preference-based score. A preference-based score begins with the assessment of multiple domains of health. However, unlike a profile, a preference-based score combines multiple domains into a single number that ranges from 0 (“dead”) to 1 (“full health”). This score quantifies the value that individuals place on different states of health. Preference-based scores represent explicit trade-offs between different levels of health in different domains. Preference-based scores can be used in cost-utility analyses and to estimate quality-adjusted life years (QALYs). The images below show examples of their use.

Multiple methods are used to construct preference-based scores including time tradeoff, standard gamble, and others (e.g., discrete choice). Scores are based on preferences for different health states (i.e., different combinations of health profile scores) and are often developed by combining responses from a large nationally-representative sample. Preference-based scores are also called health utility scores when certain constraints are met. A 2020 JAMA publication provides a short overview of health state utility assessment. Learn more>>

Both profile and preference-based scores are important in monitoring health outcomes. It is important to note that preference-based scores are not reflective of preferences for an individual patient and should not be used for individual care decision making; much like democracy, where the aggregated outcomes of an election do not necessarily represent any individual’s preferences, an aggregated preference-based score does not necessarily represent how an individual person would make treatment decisions.

PROMIS-Preference (PROPr) Score

The PROMIS-Preference (PROPr) score is a generic, societal, preference-based summary score (see Dewitt et al., 2018). It is based on PROMIS scores for Cognition (Cognitive Function or Cognitive Function Abilities), Depression, Fatigue, Pain Interference, Physical Function, Sleep Disturbance, and Ability to Participate in Social Roles and Activities. There is evidence for cross-sectional validity including demographics, chronic health conditions, and social determinants of health (see Hanmer et al., 2018 and Hanmer 2021). Learn more about PROPr at proprscore.com.

PROPr Requires Multiple PROMIS Domains

A PROPr score is calculated from the following 7 PROMIS domains:

PROMIS Cognitive Function v2.0 or Cognitive Function – Abilities v2.0
PROMIS Depression
PROMIS Fatigue
PROMIS Pain Interference
PROMIS Physical Function
PROMIS Sleep Disturbance
PROMIS Ability to Participate in Social Roles and Activities v2.0

You can use PROMIS computer adaptive tests (CATs) or short forms for every domain, or use the two profile measures that include all necessary domains:

PROMIS-29+2 Profile v2.1 (PROPr)
PROMIS-16 Profile v2.1 (PROPr)

Missing Cognitive Function
The PROMIS-29, PROMIS-43, PROMIS-57, and PROMIS Profile CAT-29 include all the domains needed to calculate a PROPr score except for Cognitive Function. However, you can estimate a Cognitive Function score and use the estimate to then calculate a PROPr score. Estimating Cognitive Function is okay to do for calculating a PROPr score, but not for other purposes. The best option is to use a measure that includes all 7 domains.

Calculate a PROPr Score

Option 1: Automatic Scoring
If you administer measures via a digital platform supported by the Assessment Center API, a PROPr score can be calculated. Contact api@assessmentcenter.net for support.

Option 2: Manual Scoring

Step 1: Access or calculate theta values for all PROMIS domains.

- Identify if your scoring process produces theta values. The HealthMeasures Scoring Service, for example, includes theta values in the scored files it produces. See Scoring Instructions for more information>>
- If your scoring process only produces T-scores, manually calculate theta values. Theta values can be calculated from T-scores using the formula: theta = (T – 50) / 10
- If you used a PROMIS Profile that is missing Cognitive Function, estimate a Cognitive Function theta. This involves multiplying constants with theta values and summing them together. Note that you will use the 0-10 response score for PROMIS Pain Intensity (see Dewitt, Jalal, & Hanmer, 2020, Value in Health, 23).
  - Cognitive Function – Abilities theta = 0.00943 + (-0.037 * Depression theta) + (0.118 * Physical Function theta) + (-0.223 * Sleep Disturbance theta) + (0.0505 * Ability to Participate in Social Roles and Activities theta) + (-0.168 * Anxiety theta) + (-0.00599 * Pain Intensity score)

Step 2: Access PROPr scoring software in SAS, Stata, or R.

- Navigate to https://github.com/janelhanmer/PROPr.
- Download the code named “MAUT” (Multi-Attribute Utility Theory).
- Apply the code to your theta values to produce PROPr scores.

Other PROMIS Preference-based Summary Scores

EQ-5D preference-based health index scores can be estimated from the PROMIS Global Health scale. This uses scoring for the United States. To learn more, see Thompson et al et al (2017). Scoring instructions are included in the PROMIS Global Scoring Manual.

EQ-5D preference-based health index scores can also be estimated from the PROMIS-29 profile. This uses scoring for France, Germany, and the United Kingdom. See Klapproth et al (2020).

Health Utilities Index Mark 3 (HUI-3) preference scores can be estimated from the PROMIS-29 v2.0 Profile (preferred) or PROMIS Global Health scale. For more information and scoring equations, see Hays et al (2016).

Neuro-QoL™ Preference-based Score

A preference-based score for individuals with multiple sclerosis based on Neuro-QoL cognitive function, depression, fatigue, mobility, upper extremity function, and ability to participate in social roles and activities measures was published in 2020 (see Matza et al). Scoring code in R and SAS is available in the publication’s online Appendices 2 and 3.

Matza, L. S., Phillips, G., Dewitt, B., Stewart, K. D., Cella, D., Feeny, D., . . . Revicki, D. A. (2020). A Scoring Algorithm for Deriving Utility Values from the Neuro-QoL for Patients with Multiple Sclerosis. Medical Decision Making, 40(7), 897-911. https://pubmed.ncbi.nlm.nih.gov/33016238/