Psychometric Considerations in Developing PROMIS® Measures for Early Childhood

Abstract

Objective: The early expression of lifespan health and disease states can often be detected in early childhood. Currently, the Patient-Reported Outcome Measurement Information System (PROMIS®) includes over 300 measures of health for individuals ages 5 years and older. We extended PROMIS to early childhood by creating developmentally appropriate, lifespan coherent parent-report measures for 1-5-year-olds. This paper describes the psychometric approaches used for these efforts.

Methods: 2 waves of data from parents of children ages 1-5 were collected via 2 internet panel companies. Wave 1 data (n = 1,400) were used to evaluate item pool unidimensionality, model fit, and initial item parameters. Combined data from wave 1 and wave 2 (reference sample; n = 1,057) were used to estimate final item parameters. Using item response theory methods, we developed and tested 12 item pools: Global Health, Physical Activity, Sleep Disturbance, Sleep-related Impairment, Anger/Irritability, Anxiety, Depressive Symptoms, Positive Affect, Self-Regulation, Engagement, Family Relationships, and Peer Relationships.

Results: Wave 1 analyses supported the unidimensionality of Physical Activity, Positive Affect, Anger/Irritability, Anxiety, Depressive Symptoms, and Global Health. Family Relationships and Peer Relationships were combined to form "Social Relationships"; Sleep Disturbance and Sleep-related Impairment were combined to form "Sleep Problems." Self-Regulation was divided into "Flexibility" and "Frustration Tolerance"; Engagement was divided into "Curiosity" and "Persistence." Short forms were developed for item banks with more than 10 items; and.

Conclusions: Using rigorous mixed-methods, we successfully extended PROMIS to early childhood (1-5-year-olds). Measures are now publicly available in English and Spanish (www.healthmeasures.net).

Keywords: infancy and early childhood; measure validation; preschool children; quality of life; research design and methodology; statistical approach.

Figure 1.

Domains and number of items across wave 1 and wave 2 testing. ^aFamily Relationships and Peer Relationships were combined into one Social Relationships item bank. Sleep Disturbance and Sleep-related Impairment were combined into one Sleep Problems item bank. Self-regulation was separated into Flexibility and Frustration Tolerance calibrated scales. Engagement was separated into Curiosity and Persistence calibrated scales. ^bFull-length: number of items included in the final PROMIS EC measures. Measures with 10 or more items are named “item banks.” Measures with less than 10 items are named “calibrated scales.” ^cTesting: number of items used for wave 2 testing.

Figure 2.

Analysis flowchart. ^aDIF on parent’s sex was evaluated. However, because there were significantly fewer fathers versus mothers who completed the testing, evaluation of DIF on parent sex was considered exploratory and not included in the flowchart.

Figure 3.

Information function curve of each measure (red solid line). (a) Measures with Positive Direction: Higher Scores Represent Better Health (X-axis: T score; Y-axis: Information function). Note. X-axis represents T scores with mean = 50 and standard deviation = 10. Y-axis represents information function at the measure level. An information function of 3.3, 10, and 20 is corresponding to a reliability of 0.70 (green dashed line), 0.90 (navy blue dashed line), and 0.95 (purple dashed line), respectively. (b) Measures with negative direction: higher scores represent worse health (X-axis: T scores; Y-axis: information function). Note. X-axis represents T scores with mean = 50 and standard deviation = 10. Y-axis represents information function at the measure level. An information function of 3.3, 10, and 20 is corresponding to a reliability of 0.70 (green dashed line; the top line), 0.90 (navy blue dashed line; the middle line), and 0.95 (purple dashed line; the bottom line), respectively.

Figure 3.

(Continued)