Responsiveness of the Patient-reported Outcomes Measurement Information System (PROMIS) Pediatric Measures to Changes in Disease Status and Quality of Life Among Children and Adolescents With Crohn’s Disease

The study evaluated responsiveness of the PROMIS Pediatric measures to changes in disease activity and HRQOL in pediatric Crohn’s disease. It demonstrates the longitudinal responsiveness of PROMIS Pediatric measures in this disease, thus supporting their use in clinical outcomes research.

Abstract

Background

PROMIS Pediatric domains provide self-reported measures of physical, emotional, and social health in children with chronic conditions. We evaluated the responsiveness of the PROMIS Pediatric measures to changes in disease activity and disease-specific, health-related quality of life (HRQOL) in children with Crohn’s disease (CD).

Methods

IBD Partners Kids & Teens is an internet-based cohort of children with inflammatory bowel disease (IBD). Participants age 9 to 17 report symptoms related to disease activity (short Crohn’s Disease Activity Index [sCDAI]), the IMPACT-III HRQOL measure, and 5 PROMIS Pediatric domains. We conducted longitudinal analyses using mixed linear models to examine the extent to which PROMIS Pediatric measures respond to changes in sCDAI and IMPACT-III.

Results

Our study sample included 544 participants with CD (mean age 13 years, 44% female). All PROMIS Pediatric domains responded to changes in sCDAI, indicating improved physical, emotional, and social health, corresponding to improved disease activity and the converse (P < 0.001). Observed effect estimates ranged from 1.8 for peer relationships to 6.8 for fatigue. Of 246 participants with 2 or more completed reports, disease activity was stable in 527, worse in 72, and improved in 67. Changes in PROMIS Pediatric scores were associated with changes in IMPACT-III (r = −0.43 for anxiety, r = −0.45 for depressive symptoms, r = −0.43 for pain interference, r = −0.59 for fatigue, and r = 0.23 for peer relationships).

Conclusions

This study provides evidence for the longitudinal responsiveness of the PROMIS Pediatric measures to change in disease status and HRQOL in pediatric CD patients.

INTRODUCTION

Patient-reported outcome (PRO) measures are used to assess how patients feel and function in their daily lives. Directly obtained from patients without physician interpretation, they can capture outcomes that matter most to patients.^{1, 2} Hence, PROs serve as essential patient-focused end points in research and clinical care.

Patient-reported outcome measures are particularly important in pediatric Crohn’s disease (CD), a chronic gastrointestinal condition that impacts approximately 58 per 100,000 children in the United States and is associated with significant symptom burden, missed school, and psychological comorbidity.^{3, 4} In addition to the impact of pediatric CD on physical, social, and emotional development during childhood, there are downstream effects that hinder achievement in education, employment, and interpersonal relationships in adulthood.⁵ Tracking and quantifying the impact of pediatric CD on physical and psychosocial well-being can better equip researchers and clinicians to improve patients’ experiences.^{1, 6}

The Patient-Reported Outcomes Measurement Information System (PROMIS) is a system of PRO measures developed with funding by the National Institutes of Health (NIH). PROMIS includes measures of physical, emotional, and social health with strong evidence of reliability and validity in a large number of chronic disease populations.^{7, 8} The PROMIS Pediatric measures include domains geared specifically toward child respondents, with self-report items available for children ages 9 to 17.⁸

PROMIS Pediatric measures have been shown to discriminate between different levels of disease activity and severity in a range of pediatric chronic conditions, including asthma, cancer, chronic kidney disease, obesity, rheumatic disease, and sickle cell disease.^8–12 A cross-sectional analysis by Arvanitis et al demonstrated robust correlations between several PROMIS Pediatric measures and validated measures of disease activity in pediatric CD.¹

In addition to the ability of a PRO measure to discriminate between groups known to have different levels of health-related quality of life (HRQOL), high-quality PRO measures must demonstrate responsiveness—or sensitivity to clinical change—before their use in clinical care or as clinical trial end points.^{6, 13, 14} Several recent studies have demonstrated the responsiveness of PROMIS Pediatric measures in a variety of pediatric diseases, including cancer, nephrotic syndrome, sickle cell disease, and asthma.^{6, 15–17} An exploratory longitudinal analysis by Arvanitis et al suggested that PROMIS may respond to changes in disease activity over time in children with CD; however, this study concluded that further longitudinal validation was needed.¹

In the present study, we evaluated the responsiveness of the PROMIS Pediatric measures to changes in patient-reported disease activity and disease-specific HRQOL in children with CD by studying a national, prospective cohort. We hypothesized that PROMIS Pediatric measures would worsen in response to worsening disease and treatment-related symptom burden.

MATERIALS AND METHODS

IBD Partners

IBD Partners is an internet-based cohort of adults living with IBD, originally launched in 2011. The methodology has been described elsewhere.¹⁸ In 2013, IBD Partners Kids & Teens was developed in conjunction with the adult cohort. In brief, IBD Partners Kids & Teens recruits patients with self-reported IBD, along with their parents, to complete surveys related to treatments, health behaviors, disease activity, and other health outcomes. Children and their parents complete follow-up surveys every 6 months. Children ages 9 to 17 complete all symptom and function-related questions themselves, including the PROMIS Pediatric domains of anxiety, depressive symptoms, pain interference (a measure of consequence of pain on various aspects of life), fatigue, and peer relationships.^{1, 18} In the present study, we included children ages 9 to 17 with self-reported CD who completed PROMIS Pediatric measures by 2019.

Measures

For each of the 5 selected PROMIS Pediatric domains, participants completed either 4-item short forms (surveys before May 7, 2017) or computerized adaptive testing (CAT) for surveys after May 7, 2017. Computerized adaptive testing has previously been shown to result in measurement equivalence with fixed length measures but reduce respondent burden.¹⁹ Therefore, the 2 assessment mechanisms were treated equivalently in the present analysis. PROMIS Pediatric measures are calibrated using a T-score metric, with the mean of the original calibration population equal to 50, and the standard deviation in the calibration population equal to 10.^{7, 8} Higher PROMIS Pediatric scores represent more of the domain being measured.⁸ The minimally important difference (MID), defined as a difference in scores that is detectable and important to children and parents, is 3 points for PROMIS Pediatric measures.^{1, 20}

We used the short Crohn’s Disease Activity Index (sCDAI) as a measure of disease activity and the IMPACT-III as a measure of HRQOL, both of which are widely used and previously validated.^21–24 The sCDAI includes questions regarding number of diarrheal stools, abdominal pain, and general well-being over the past 7 days. A sCDAI <150 indicates clinical remission, with a MID of 70 points.²⁴ The IMPACT-III includes 35 HRQOL-related items in 6 domains. Possible scores range from 35 to 175, with higher scores indicating better functioning.²⁵

We used Global Impression of Change (GIC) questions to assess the relationship between self-perceived rating of functional improvement or worsening and changes in corresponding PROMIS Pediatric score.²⁶ The GIC questions asked participants to provide an overall assessment of change in their anxiety, depressive symptoms, pain, and fatigue compared with the previous survey they completed.

We also obtained the following demographic information for the cohort: age, sex, race, ethnicity, parental education level, disease characteristics, state of residence, current and historical medical therapy, and presence of smokers in the household.

Statistical Analysis

We computed summary statistics for patients’ characteristics and for or between all of the measures included in the study for all enrolled children who completed at least 2 surveys. We evaluated PROMIS Pediatric measures responsiveness to changes in sCDAI and IMPACT-III using mixed linear models to adjust for time and account for the clustering of individual participants who completed multiple surveys. We additionally adjusted for age, sex, parental education level, race, and ethnicity in our models.

We estimated predicted change in PROMIS Pediatric measures corresponding to a 70-point change in sCDAI, an accepted threshold of clinical response.²⁷ We next graphically depicted predicted mean change in each PROMIS Pediatric domain corresponding to improved, stable, or worsened disease activity, categorized based on at least a 70-point change in sCDAI. We calculated change in PROMIS Pediatric measures corresponding to a 70-point change in sCDAI utilizing the raw data, as well, and depicted the results graphically. We also used Pearson correlation to evaluate associations between change in IMPACT-III and change in each of the PROMIS Pediatric domains.

Finally, we conducted exploratory analyses of the associations between each participant’s GIC questions and the actual change in PROMIS Pediatric scores. The GIC responses were categorized as worse (including the responses “significantly worse” and “worse”), the same, and better (including the responses “better” and “significantly better”). We determined the mean change and 95% confidence interval (CI) in PROMIS Pediatric score for anxiety, depressive symptoms, pain interference, and fatigue corresponding to a worse, same, or better GIC rating for the respective domain. For these analyses, we excluded GIC reports that referred to a prior survey completed more than 1 year earlier, as we did not think participants could reliably recall their symptoms over an extended time period.

ETHICAL CONSIDERATIONS

All data were prepared and analyzed using SAS v 9.3 (SAS Institute, Cary, North Carolina). The IBD Partners Kids & Teens protocol was reviewed and approved by the institutional review boards of the University of North Carolina at Chapel Hill and Duke University.

RESULTS

We analyzed a total of 544 children and adolescents with self-reported CD from 44 different states across the United States. The mean age was 13.3 years (SD 2.4 years), and mean duration since IBD diagnosis was 3.2 years (SD 2.6; Table 1); 44.4% were female, 90.4% were white, and 6.4% were of Hispanic ethnicity. The majority of children (81.2%) were in remission based on sCDAI at the time of baseline survey. Most participants had ≥1 parent who completed at least some college (97.9% for mother; 97.9% for father). A total of 246 participants completed more than 1 survey, resulting in 689 pairs of consecutive reports. A total of 91 subjects completed 2 surveys, 53 subjects completed 3 surveys, 38 subjects completed 4 surveys, and 64 subjects completed between 5 and 12 surveys. The predictive model utilized all data, including the data from participants who completed only 1 survey (n = 298). Participants who completed more than 1 survey had similar demographic characteristics to the entire cohort (Table 1). Between reports, disease activity worsened in 72 cases, remained stable in 527, and improved in 67 (based on a 70-point change in sCDAI).

TABLE 1.

Demographic Characteristics

Demographicsa	Entire study cohort (N = 544)	Participants who completed >1 survey (N = 246)
Age, mean (SD)	13.3 (2.4)	12.5 (2.3)
Sex, % female	44.4%	42.4%
Raceb
White	90.4%	92.4%
African American	4.2%	3.4%
Asian	1.5%	1.7%
Multiracial	3.6%	2.5%
Other	0.4%	0%
Ethnicity
Hispanic	6.4%	4.2%
Non-hispanic	93.6%	95.8%
Highest parental education (father, mother)
Less than 12th grade	2.1%, 2.1%	0.4%, 0.8%
12th grade	14.2%, 7.0%	8.9%, 5.1%
Some college	17.9%, 19.5%	16.0%, 16.0%
College	36.2%, 40.6%	43.0%, 43.0%
Graduate school	28.7%, 30.5%	30.8%, 35.0%
Unknown	0.9%, 0.4%	0.8%, 0.0%
Gastroenterologist setting
University/academic	23.7%	26.0%
Private practice	24.6%	24.0%
Hospital other than a university hospital	49.3%	47.2%
Other	1.9%	2.4%
Unknown	0.6%	0.4%
Disease characteristics
% in remission (SCDAI <150) at baseline survey	81.2%	83.3%
Years since IBD diagnosis, mean (SD)	3.2 (2.6)	2.9 (2.3)
Number of states represented	44	42
Therapy
Historical therapy
Oral 5-aminosalicylates	57.6%	60.6%
Immunomodulators	71.9%	79.9%
Biologics	66.0%	69.3%
Corticosteroids	83.2%	87.4%
Surgery	15.4%	13.4%
Current therapy
Oral 5-aminosalicylates	25.0%	26.4%
Immunomodulators	42.6%	47.5%
Biologics	58.6%	62.7%
Corticosteroids	14.4%	15.9%
Cigarette smokers in the household
Yes	7.7%	5.7%
No	92.3%	94.3%
Disease-specific measures
Disease-specific activity scores, mean (SD) SCDAI	103.0 (73.0)	101.7 (74.2)
Health-related QoL scores, mean (SD) IMPACT-III	134.7 (24.4)	135.4 (21.5)
PROMIS Pediatric measures
PROMIS Pediatric T-scores, mean (SD)
Anxiety	46.9 (11.0)	45.9 (10.4)
Depressive symptoms	43.2 (8.5)	42.4 (7.9)
Fatigue	48.2 (12.2)	47.2 (11.9)
Pain interference	47.3 (11.2)	46.0 (10.9)
Peer relationships	48.9 (9.6)	49.1(9.4)

^aPercentages do not include missing values. For all demographic characteristics, less than 4% of data were missing for each category except for sCDAI score (11.0% overall and 12.2% of participants with >1 survey), remission (extrapolated from sCDAI score), and Pediatric PROMIS score (≤8.5% overall and ≤10.6% for participants with >1 survey for each domain).

^bA single participant could select more than one race.

Responsiveness to Changes in Disease Activity

All PROMIS Pediatric domains were responsive to changes in disease activity as measured by sCDAI, indicating improved physical, emotional, and social health. Predicted mean changes in PROMIS Pediatric scores corresponding to a 70-point decrease in sCDAI are shown in Table 2. Associations between change in PROMIS Pediatric scores and change in sCDAI were statistically significant (P < 0.001) for all domains. Corresponding to a 70-point decrease in sCDAI, effect estimates ranged from 1.8 (95% CI, 1.3–2.4) for peer relationships to 6.8 (95% CI, 6.3–7.4) for fatigue, in line with MID estimates in both adult IBD and other pediatric chronic conditions.^{1, 28} Mean changes in PROMIS Pediatric scores corresponding to a 70-point change in sCDAI utilizing the raw data are shown in Supplemental Table 1 and Supplemental Figure 1.

TABLE 2.

Predicted Change in PROMIS Pediatric Scores_a Corresponding to a 70-Point Decrease in sCDAI

PROMIS Pediatric Domain	Predicted Change in PROMIS Pediatric Score (95 CI)	P
Anxiety	−3.2 (−3.8 to −2.6)	<0.0001
Depressive symptoms	−3.0 (−3.5 to −2.6)	<0.0001
Pain interference	−6.1 (−6.7 to −5.6)	<0.0001
Fatigue	−6.8 (−7.4 to -6.3)	<0.0001
Peer relationships	1.8 (1.3 to 2.4)	<0.0001

^aNegative changes in anxiety, depression, pain interference, and fatigue indicate improvements in health and/or functioning. A positive change in peer relationships indicates improved functioning.

We did not find any significant effect of age, race, ethnicity, or sex on the relationship between PROMIS Pediatric domains and disease activity using linear mixed modeling. Figure 1 demonstrates the predicted mean changes in PROMIS Pediatric scores as a function of change in disease activity, adjusted for age, sex, parental education level, race, and ethnicity.

FIGURE 1.

Predicted mean change in PROMIS Pediatric domains according to change in Crohn’s disease activity as measured by the short Crohn’s Disease Activity Index. Improved, stable, and worsened were defined as a 70-point decrease in sCDAI, less than a 70-point change in sCDAI, and a 70 point increase in sCDAI, respectively. The sign of change for peer relationships was inversed for this figure, such that a higher score indicates worsened peer relationships. For all other domains, higher scores indicate more of the domain being measured. The threshold used to indicate change in disease activity was a ≥70 point change in sCDAI.

Responsiveness to Changes in HRQOL

Changes in PROMIS Pediatric scores were also associated with changes in IMPACT-III scores in both our correlation analysis and linear mixed-modeling analysis (P < 0.001 for all PROMIS domains; Table 3 and Fig. 2). Pearson correlation coefficients are as follows: r = −0.43 for anxiety, r = −0.45 for depressive symptoms, r = −0.43 for pain interference, r = −0.59 for fatigue, and r = 0.23 for peer relationships.

TABLE 3.

Change in PROMIS Pediatric Scores Corresponding to Change in IMPACT-III_a

PROMIS Pediatric Domain	Crude r (Pearson correlation coefficient)	Crude P	Mixed Linear Model P
Anxiety	−0.43	<0.0001	<0.0001
Depressive symptoms	−0.45	<0.0001	<0.0001
Pain interference	−0.43	<0.0001	<0.0001
Fatigue	−0.59	<0.0001	<0.0001
Peer relationships	0.23	<0.0001	<0.0001

^aIMPACT-III is a 35 item health-related quality of life measure. Possible scores range from 35 to 175, with higher scores indicating better functioning.

FIGURE 2.

Mean change in PROMIS Pediatric scores according to change in IMPACT-III. IMPACT-III is a 35-item health-related quality of life measure. Possible scores range from 35 to 175, with higher scores indicating better functioning.

Exploratory Analyses of Global Impression of Change

A total of 94 GIC reports were obtained from 66 participants. Table 4 shows the mean change in PROMIS measures corresponding to GICs categorized as worse, stable, or better. Most participants rated their GIC as stable. As expected, mean change in PROMIS Pediatric scores did not significantly differ from zero in these patients. Respondents who rated their functioning as improved had a corresponding improvement in the respective PROMIS Pediatric score in all domains, although only pain interference had a statistically significant mean change of −2.68 points (95% CI, −4.56 to −0.80). Respondents who rated their functioning as worsened had a corresponding worsening in the respective PROMIS Pediatric score in all domains except pain interference, although only depressive symptoms had a statistically significant mean change of 9.89 points (95% CI, 3.94–15.84).

TABLE 4.

Change in PROMIS Pediatric Scores Corresponding to Global Impression of Change Responses

	GIC Question Responsea
GIC Question vs PROMIS Pediatric Domain	Worse		Same		Better
	Mean Change in PROMIS Pediatric Domain (95% CI)	N	Mean Change in PROMIS Pediatric Domain (95% CI)	N	Mean Change in PROMIS Pediatric Domain (95% CI)	N
Rate your worried or nervous feelings now vs PROMIS Pediatric anxiety	3.68 (−3.38 to 10.73)	8	0.40 (−1.24 to 2.04)	52	−0.66 (−2.92 to 1.59)	34
Rate your sad or unhappy feelings now vs PROMIS Pediatric depressive symptoms	9.89 (3.94 to 15.84)	8	0.33 (−1.30 to 1.97)	54	−1.32 (−3.34 to 0.70)	32
Rate your pain now vs PROMIS Pediatric pain interference	−3.15 (−18.17 to 11.86)	8	−0.66 (−2.11 to 0.79)	38	−2.68 (−4.56 to −0.80)	47
Rate your tired feelings now vs PROMIS Pediatric fatigue	5.05 (−2.84 to 12.93)	13	−0.78 (−2.78 to 1.21)	44	−1.93 (−5.39 to 1.54)	36

^aA single participant was allowed to contribute multiple responses to this analysis if they completed >1 GIC survey response.

DISCUSSION

PRO measures that accurately capture a child’s functional status as his or her disease worsens or improves can be powerful tools for the clinician and researcher alike. These instruments can improve physician-patient communication and enable providers to tailor a therapy to a patient’s individual experiences and outcomes. They also act as valuable end points in clinical, observational, comparative effectiveness, and health services research, reflecting outcomes that matter to patients.^{1, 2} Before employing these tools, however, we must show that they both consistently measure the intended outcome and that they do so dynamically, changing in response to evolving clinical status.

In this study, we evaluated the responsiveness of the PROMIS Pediatric measures to changes in disease activity and HRQOL in a large cohort of children and adolescents with Crohn’s disease. We found that PROMIS Pediatric measures respond to changes in disease status and HRQOL in a clinically meaningful manner. Our results support the use of PROMIS Pediatric measures in clinical outcomes research in pediatric CD.

Our findings add to the growing body of literature supporting the responsiveness of PROMIS Pediatric measures across a wide range of pediatric chronic conditions. Reeve et al found that PROMIS Pediatric scores worsened from baseline to disease event and improved from event to remission in cohorts of children with cancer, nephrotic syndrome, and sickle cell disease, respectively.⁶ In another study, PROMIS Pediatric measures tracked closely with changes in the Paediatric Asthma Quality of Life Questionnaire (PAQLQ), a legacy measure of HRQOL in asthma.¹⁶ Taken together, these studies support more widespread adoption of PROMIS Pediatric measures, including use in conditions such as CD that were not originally included in the development and validation cohorts.

Although all 5 PROMIS Pediatric domains responded significantly to changes in disease status and HRQOL, peer relationships had the least robust response. This finding that peer relationships remain more stable than other domains in the face of changing disease severity is consistent with prior work in other pediatric chronic conditions and an earlier exploratory study in IBD.^{1, 11, 17} This stability is likely beneficial and could serve an important role in modifying the impact of chronic disease.

There are a number of strengths to this longitudinal study. We included a large, geographically diverse population of pediatric CD patients. This facilitated the conduct of well-powered, longitudinal assessments. Additionally, in contrast to most pediatric IBD studies conducted at tertiary referral centers, the direct-to-patient nature of this internet-based cohort allowed us to include patients treated in a variety of practice settings.

A notable limitation of our study is that the volunteer nature of our cohort potentially limits generalizability. For example, the majority of our participants were white and from households with well-educated parents who completed at least some college. Future studies of PROMIS and other PROs in children with more diverse backgrounds and lower socioeconomic status will be helpful to confirm our findings. In addition, our study evaluated responsiveness of PROMIS Pediatric measures to changes in symptom-based disease activity indices rather than endoscopic measures of disease activity. We hope that these findings will encourage the use of PROMIS measures in clinical trials that also incorporate endoscopic end points. Such future studies would provide the ability to study the responsiveness of PROMIS pediatric measures in relation to the achievement of mucosal healing. Including patients with self-reported CD rather than physician-confirmed CD could have introduced bias into our study. However, a validation study of the adult IBD Partners cohort demonstrated high reliability of self-reported IBD status, with over 95% of participants accurately reporting their disease type.²⁹ A final limitation of our study was that the small sample size of our GIC analyses limited both the precision and statistical power. Additionally, GIC items with a long interval between reporting periods may reduce accuracy, as respondents may have difficulty recalling their mental state or functioning from 6 to 12 months prior. Thus, we consider the GIC analyses here as exploratory, reinforcing the responsiveness of PROMIS Pediatric measures demonstrated in our primary analyses.

CONCLUSION

In summary, this study provides evidence for the longitudinal responsiveness of PROMIS Pediatric measures in relation to disease status and quality of life in pediatric patients with CD. By capturing the patient experience and changing in response to evolving disease, these PRO instruments can serve as powerful end points in patient-centered research.

Supported by: Research reported in this publication was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under award number U19AR069522 and the Crohn’s & Colitis Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Foundation.