Sample characteristics

The ages of those in the sample (n = 4130) ranged from 18–94 with a median age of 54 and more people 60 and older than in younger age groups (Table 1). Forty-nine percent were female, and < 1% were transgender or did not identify as female, male, or transgender. Seventy percent were non-Hispanic White, 12% were Hispanic, 10% non-Hispanic Black, 5% non-Hispanic of another race, and 3% were two or more races. Thirty-four percent reported a high school degree or less, 26% had some college, and 40% had a bachelor’s degree or higher. Annual income was $100,000 or more for 42% of the sample.

Table 1 Sample Characteristics (n = 4,130)PROMIS-16 confirmatory factor analyses

The confirmatory factor analysis model estimated for the PROMIS-16 (without pain intensity) was rejectable statistically (χ2 (12 df = 391.532, p < 0.0001) but fit the data reasonably well: CFI = 0.97 and RMSEA = 0.09. The estimated correlation between physical and mental health was 0.46. Similar results were found for the confirmatory factor analysis model estimated for the PROMIS-16 with pain intensity included in the pain composite indicator. It was rejectable statistically (χ2 (12 df = 355.194, p < 0.0001) but fit the data reasonably well: CFI = 0.97 and RMSEA = 0.08. The estimated correlation between physical and mental health was 0.45.

PROMIS-16 and PROMIS-29 reliabilities

Table 2 provides internal consistency reliability estimates for the multi-item scales and reliability estimates for the summary scores that are weighted combinations of the PROMIS-16 and PROMIS-29 measures. PROMIS-16 reliabilities ranged from 0.73 (sleep disturbance) to 0.95 (mental health summary score) while PROMIS-29 reliabilities ranged from 0.87 (sleep disturbance) to 0.98 (mental health summary score).

Table 2 Product-moment Correlations Among the PROMIS-16 and PROMIS-29 Measures

The average difference between the reliabilities of the PROMIS-16 measure and those estimated from the Spearman-Brown prophecy formula was zero for the four measures used in creating the PROMIS physical and mental health summary scores that were a subset of the corresponding PROMIS-29 measures. However, the reliabilities for the two PROMIS-16 measures with some unique items were 0.04 (sleep disturbance) and 0.07 (ability to participate in social roles and activities) lower than those estimated from the PROMIS-29.

Product-moment correlations among the PROMIS-16 and PROMIS-29 measures

Correlations among the PROMIS-16 and PROMIS-29 measures are shown in Table 2. Correlations within parentheses are provided to show the similarity of associations when the PROMIS pain measure used is the average of pain interference and the pain intensity item. The bolded entries in the correlation matrix, representing corresponding PROMIS-16 and PROMIS-29 measures, ranged from 0.81 (sleep disturbance) to 0.98 (mental health summary score). The correlations are all essentially perfect when adjusted for the unreliability of measurement.

Product-moment correlations of PROMIS summary scores with the PROMIS global physical and mental health scores

The PROMIS-16 and PROMIS-29 physical health summary scores correlated more strongly with the PROMIS global physical health score (r’s = 0.73 and 0.75, respectively) than with the PROMIS global mental health score (r’s = 0.35 and 0.37, respectively). However, as expected based on previous research [8], the PROMIS-16 and PROMIS-29 mental health summary scores correlated more strongly with the PROMIS global physical health score (r’s = 0.72 and 0.75, respectively) than with the PROMIS global mental health score (r’s = 0.65 and 0.66, respectively).

Mean differences between PROMIS-29 and PROMIS-16 measures

Mean scores and mean differences for the PROMIS-29 and PROMIS-16 are given in Table 3. The effect size differences ranged from 0.02 to 0.10 in absolute value. Intraclass correlations between corresponding PROMIS-29 and PROMIS-16 measures ranged from 0.81 (sleep disturbance) to 0.995 (pain).

Table 3 Mean Differences Between PROMIS-29 and PROMIS-16 Measures

The Bland–Altman plot shows scatter bias for the difference between the PROMIS-16 and PROMIS-29 estimated physical health summary score, with the PROMIS-16 estimates overestimating the PROMIS-29 estimate at the upper (better health) end of the distribution (Fig. 1). Scatterbias is also seen for the PROMIS-29 mental health summary score (Fig. 2), but it is not as large as was seen for the physical health summary score.

Fig. 1

Bland–Altman Plot for PROMIS-16 Versus PROMIS-29 Estimated Physical Health Summary Score

Fig. 2

Bland–Altman Plot for PROMIS-16 Versus PROMIS-29 Estimated Mental Health Summary Score

Correlates of mean differences between corresponding physical and mental health summary scores estimated by PROMIS-29 and PROMIS-16

Product-moment correlations indicated significant variation in differences between PROMIS-29 and PROMIS-16 estimated physical health summary scores by age (r = −0.03, p = 0.0398), female gender (r = −0.04, p = 0.0160), education (r = 0.07, p < 0.0001), and income (r = 0.11, p < 0.0001). Differences in PROMIS-29 and PROMIS-16 estimated mental health summary scores varied significantly by female gender (r = −0.04, p = 0.0098), education (r = 0.04, p = 0.0194), and income (r = 0.09, p < 0.0001). While different statistically at p < 0.05, these differences are small.

We illustrate the magnitude of the differences in the PROMIS-29 and PROMIS-16 estimates with income, the sociodemographic variable with the strongest association with the differences. Table 4 provides mean summary score differences by four levels of income. The PROMIS-16 physical health means were larger than the PROMIS-29 means among those with lower income and smaller for those with the highest income. All these mean differences are small.

Table 4 Mean T-score Differences Between PROMIS-16 and PROMIS-29 Estimated Physical and Mental Health Summary Scores by Family Income