Assessing Children’s Eudaimonic Well-Being: The PROMIS Pediatric Meaning and Purpose Item Banks

Abstract

Objective

To describe the development of the Patient-Reported Outcome Measurement Information System (PROMIS) Pediatric Meaning and Purpose item banks, child-report and parent-proxy editions.

Methods

Data were collected from two samples. The first comprised 1,895 children (8–17 years old) and 927 parents of children 5–17 years old recruited from an Internet panel, medical clinics, and schools. The second comprised a nationally representative sample of 990 children 8–17 years old and 1,292 parents of children 5–17 years old recruited from a different Internet panel. Item pool evaluation was done with Sample 1 and analyses were used to support decisions about item retention. The combined sample was used for item response theory (IRT) calibration of the item bank. Both samples were used in validation studies.

Results

Eleven items were deleted from the item pool because of poor psychometric performance. The final versions of the scales showed excellent reliability (>0.90). Short form scales (4 or 8 items) had a high degree of precision across over 4 SD units of the latent variable. The item bank positively correlated with extant measures of positive psychological functioning, and negatively correlated with measures of emotional distress, pessimism, and pain. Lower meaning and purpose scores were associated with adolescence and presence of a special healthcare need.

Conclusion

The PROMIS Pediatric Meaning and Purpose item banks and their short forms are ready for use in clinical research and practice. They are measures of children’s eudaimonic well-being and indicative of children’s hopefulness, optimism, goal-directedness, and feelings that life is worth living.

Introduction

Individuals judge their lives as meaningful when life makes sense, has a purpose, and is predictable and goal-directed (Steger, 2011). Meaning and purpose is a component of subjective well-being, a multidimensional concept that includes experienced (emotional states), evaluative (how satisfying life is), and eudaimonic (sense of meaning and purpose) dimensions (National Research Council, 2013). Experienced well-being refers to physical and emotional distress as well as the positive feelings of contentment, happiness, and joy. Evaluative well-being, also called life satisfaction, comprises assessments of life in general and across specific contexts, such as self, family, friends, home, school, and work. Eudaimonic well-being includes appraisals of life as having meaning, purpose, and hope.

Meaning and purpose promotes psychological well-being, positive health-related behaviors, and resistance to the impact of illness. Among adults, meaning and purpose has been associated with reduced risk of mortality (Hill & Turiano, 2014) and better mental health (Krause, 2007). In pediatric samples, meaning and purpose is associated with better academic achievement (DeWit, Woolsey, & Walsh, 2009), less delinquency (Hill, Edmonds, Peterson, Luyckx, & Andrews, 2016), and lower rates of substance use (Harlow, Newcomb, & Bentler, 1986). There are few measures developed specifically to assess meaning and purpose in children and adolescents (Ravens-Sieberer et al., 2014; Hill et al., 2016; Shoshani and Russo-Netzer, 2017). Those available were not informed by children’s or parent’s perspectives on how they evaluate their lives; cognitive testing was not done to ensure item comprehensibility; translatability reviews were not done to ensure cultural appropriateness; and, item response theory (IRT) was not used. It is also unclear how well extant measures assess meaning and purpose across a broad range of the latent variable. Finally, none includes parallel child self-report and parent-proxy versions. The latter is of particular importance in situations when the child is too sick or too fatigued to respond to questionnaires.

To address these limitations in previously developed measures, we used the Patient-Reported Outcomes Measurement Information System (PROMIS^®) mixed qualitative-quantitative approach for person-reported outcome development (Forrest et al., 2012) to create a meaning and purpose item bank. The item pool development and content validation are described elsewhere (Ravens-Sieberer et al., 2014). The meaning and purpose measure is part of a trio of PROMIS Pediatric measures that assess positive psychological well-being—namely positive affect (experienced well-being) (Forrest et al., 2017), life satisfaction (evaluative well-being) (Forrest et al., 2018), and meaning and purpose (eudaimonic well-being). Our objectives were to produce child-report and parent-proxy item banks that reflect children’s perspectives on meaning and purpose and are well understood by children as young as age 8 years-old, unidimensional, free from differential item functioning, and precise across a wide range of the latent variable.

Methods

Overview

Meaning and purpose items were developed using child, parent, and subject matter expert semi-structured interviews, a systematic literature review, readability analysis, translatability review, and cognitive interviews, a set of qualitative processes previously reported (Ravens-Sieberer et al., 2014). Items have no recall period, because evaluations of meaning and purpose are expected to change slowly. Each item starts with “Thinking about my life, …” Items use a 5-point response scale—not at all, a little bit, somewhat, quite a bit, and very much. The parent-proxy edition replaces in the stem the word “my life” with “my child’s life.” Data for this study were collected from two samples, both included children and parents. Data analysis proceeded along three phases: Item Pool Evaluation; Item Bank Calibration; and, Validation.

Sample 1 comprised children (8–17 years old) and parents of children 5–17 years old recruited from an Internet panel, medical clinics, and schools. Sample 2 comprised a nationally representative sample of children 8–17 years old and parents of children 5–17 years old recruited from a different Internet panel. Sample 2 was added to obtain a nationally representative sample on which to base norm-referenced scale scores. We provided no compensation to participants. All the variables and measures for both data collections are shown in Supplementary Table 1. The Institutional Review Board of the Children’s Hospital of Philadelphia (CHOP) approved study protocols (IRB #10-007684 and IRB #13-010404). Parents gave informed consent, and children provided assent for the study.

Sample 1: Internet Panel, Schools, Clinics

Data were collected from April 2011 to March 2012 from 1,895 children aged 8–17 years and 927 parents of children aged 5–17 years. About half (54% of children [n = 1,029] and 51% [n = 475] of parents) were recruited from a convenience Internet panel. Others were recruited from schools (38% of children [n = 729], 34% of parents [n = 311]) or the CHOP clinics (7% of children [n = 137], 15% of parents [n = 141]). A sub-sample of children (n = 101) and parents (n = 632) from the Internet panel completed the items twice about 3 weeks apart (survey interval median = 23 days, SD = 3 days, range = 16–31 days) to evaluate test–retest reliability.

The Internet sub-sample was recruited through Op4G, an online research panel with members who complete research activities on their home computers (see Op4g.com for more information). Parents who consented were sent an electronic link to the questionnaire. After completing their assessment, parents asked their child to complete a questionnaire. The research panel firm did not retain the number invited to participate, which precluded estimation of a participation rate. Parents were sent reminder emails every 3 days or a maximum of 3 email reminders were sent.

School-based data collection occurred in 15 schools located in New Hampshire, Vermont, and Texas. Students in grades 3–12, excluding those in self-contained special education classrooms, were invited to participate. Consent forms were sent home at the start of the school year and returned to the school. Children in 5^th–12^th grades completed self-administered paper-and-pencil questionnaires; the study team read questions to groups of children in 3^rd–4^th grades. Children were given a parent questionnaire and asked to take it home. Parents mailed completed questionnaires to the study team.

For clinic-based surveys, parents and children were recruited in the reception areas. Parents provided consent and responded to questions presented on a tablet computer while they waited in the reception area, at the same time their children completed a questionnaire on tablet computers.

Sample 2: Representative Internet Panel

Data for Sample 2 were collected from June 2014 to September 2014. We enrolled parents of children aged 5–17 years-old from GfK Knowledge Panel, a dual-frame (random-digit dial and address-based) online probability panel (GfK, 2019). Sample 2 provided a nationally representative sample for use in item bank calibration and standardizing scale scores to a national norm. The initial sample weights adjusted for oversampling of individuals living in minority communities and Spanish-language dominant areas, although fluency in English was required for the study. The weights were iteratively adjusted until the sample's distributions of age, gender, race, ethnicity, education, census region, metropolitan area, household internet access, and language (English/Spanish) matched those in the 2013 Current Population Survey. Parents were emailed a link to an online questionnaire. After completing their questionnaire, they asked their children to participate. The child survey was administered as an audio-assisted computerized questionnaire. Children could stop the audio after recording their answer. Data collection continued until balanced age-gender quotas were met.

Analysis Phase 1: Item Pool Evaluation

This phase of data analysis used Sample 1 data only for descriptive analyses of items and scales, reliability, dimensionality, monotonicity of items (i.e., increasing item scores are associated with increasing scale scores), and item local independence (i.e., items were weakly correlated, <0.20, after adjusting for the correlation of the scale). Results informed decisions about retention of items. We re-examined the cognitive interviews, previously reported (Ravens-Sieberer et al., 2014), for items that were flagged for deletion.

Each item’s mean, standard deviation, skewness, and percentage with scores at the ceiling (score of 5) or floor (score of 1) were computed. At the scale level, we examined the range and the percentage of individuals at the floor and ceiling. Reliability was evaluated using marginal reliability, an IRT-based estimate that is similar to Cronbach's alpha (Green, Bock, Humphres, & Linn, 1984), item-total correlations, and test-retest correlations.

Unidimensionality, local independence, and monotonicity are prerequisites for the graded response IRT model (Edelen & Reeve, 2007). Unidimensionality was examined using Confirmatory Factor Analysis (CFA) with weighted least squares means, variance adjusted estimator, and an oblique rotation using Mplus 7.2. We evaluated CFA model fit with the comparative fit index (CFI > 0.95 for good fit), the Tucker-Lewis index (TLI > 0.95 for good fit), and root mean square error of approximation (RMSEA < 0.06 for good fit). A criterion of ≥0.60 for CFA factor loadings was used for item retention. If the fit indices did not support the unidimensional model, we planned to examine the residual correlation matrix to identify local dependence—i.e., the residual correlation was ≥ 0.20. Graphs of item mean scores conditional on total test scale score were examined to confirm item monotonicity.

We examined each item for differential item functioning (DIF), which occurs when groups of individuals with equivalent latent variable scale scores answer items differently as a function of another variable (e.g., age). We used a multiple group factor analysis approach (Carle, 2010) to probe for DIF across age, gender, race, ethnicity, and study sample, and determined if statistically significant DIF (detected at the item level) substantively biased scale scores.

Analysis Phase 2: Item Bank Calibration

The two samples were combined for item bank calibration using Samejima’s Graded Response Model (Samejima, 1997). In IRT, calibration estimates discrimination and threshold parameters for each item. The discrimination statistic (also referred to as slope and designated by a) measures how well item responses differentiate respondents by their level of the latent variable (i.e., meaning and purpose). The IRT model produces threshold parameters (referred to as item difficulty, and designated as b), which correspond to the difficulty, or level of the latent variable required to endorse a particular item response. Thresholds indicate the point on the latent variable where a respondent is more likely than not to respond in a higher category. For an item with five response options, the IRT model results in four item threshold statistics.

Ideally, one calibrates items using a representative sample from the population of interest. Although we obtained a representative sample of the general U.S. pediatric population (Sample 2), those children answered an eight-item short form rather than all items in the item bank. Thus, we could not calibrate all the items using just Sample 2 and employed a multiple group IRT approach to conduct calibration analyses (Millsap &Yun-Tein, 2004; Takane and de Leeuw, 1987). This allowed us to use data from both samples, as well as the nationally representative design weights from Sample 2, to estimate item-level calibration parameters. We implemented these analyses in Mplus 7.2 using maximum likelihood estimation with a logit link.

Finally, to create fixed-length short forms, we selected items based on content (representative across the conceptual facets of hope, optimism, goal orientation, and purpose) and IRT item discrimination and threshold parameter estimates. For the 4-item short form, items with high levels of item discrimination were preferred, because this statistic reflects the item’s ability to differentiate among individuals at different levels of the latent variable; the 8-item short form includes all the items in the 4-item form and expands the range of content. Short form and full item bank marginal reliabilities were plotted by the IRT-based scale score to assess precision across the range of the meaning and purpose continuum.

Scoring

PROMIS measures are scored in the direction of their concept’s name, so higher meaning and purpose scores indicate better eudaimonic well-being. After finalizing item parameters, we used Firestar v1.2.2, an R-based software program, and estimated full bank, 8-item Short Form (SF8a), and 4-item Short Form (SF4a) scale scores using Bayesian Expected A Posteriori (EAP) estimation. EAP scoring uses the pattern of responses and the model's parameters to estimate an individual's score, called theta, which is set to mean of 0 with a SD of 1. Theta scores are transformed to T-scores by multiplying by 10 and adding 50. A score of 50 represents the average meaning and purpose level for children in the national sample, and a score of 40 is 1 standard deviation below the national average, and so on. We produced conversion tables that translate the total sum score to a PROMIS T-Score (available on healthmeasures.net).

Analysis Phase 3: Validation

Using data from Sample 2 only, we conducted known-group validity analyses using multivariable regression to test the hypotheses, suggested by prior literature on subjective well-being that meaning and purpose decreases with children’s age and presence of a long-term health condition, and is not different by gender, race, or ethnicity.

Convergent and discriminant validity were assessed using child-reported data from both samples. We contrasted the meaning and purpose item bank and short form scores with other measures of positive psychological functioning (convergent validity) and emotional distress, pessimism and pain (discriminant validity) using Pearson’s correlation coefficients. Measures used for convergent validity analyses included the NIH Toolbox Positive Affect Fixed Form v2.0 (Salsman et al., 2014), NIH Toolbox General Life Satisfaction Fixed Form v2.0 (Salsman, et al., 2014), and the Children’s Hope Scale (Snyder et al., 1997). Discriminant validity was examined using the Youth Life Orientation Test-Pessimism Subscale (Ey et al., 2005), PROMIS Pediatric Short Form v1.0—Anger 6a (Irwin et al., 2012), PROMIS Pediatric Short Form v1.0— Anxiety 8a (Irwin et al., 2010), PROMIS Pediatric Short Form v1.0—Depressive Symptoms 8a (Irwin et al., 2010), and the PROMIS Pediatric Short Form v1.0—Pain-Behavior 8a (Jacobson et al., 2015).

Results

The socio-demographic characteristics of the child and parent samples are shown in Supplementary Table 2. All analyses were replicated for the parent-proxy banks, and can be found in the Supplementary Tables 3–5 and Supplementary Figure 1. Descriptive statistics, reliability, validation, and IRT calibration results for the parent-proxy item bank were consistent with the child self-report edition.

Phase 1 Results: Item Pool Evaluation

Seven items (goal for future, plans for future, set goals for future, expect the best to happen, optimistic about future, optimistic, and expect good things to happen) were deleted because of age-related DIF. We also re-reviewed the cognitive interviews for each of these items and found that their understandability among 8–12 year-old children was borderline; for example, younger children had trouble comprehending the word optimism. Two items (long-term goal and long-term goals) were deleted because of local dependence with other items. Low item discrimination led to the deletion of two items (look forward to doing fun things, expect things to go the way I want them to). These deletions left 44 items in the item pool.

Item-level floor effects were minimal, while ceiling effects were more common (Table 1). The item-total correlation range was 0.57 to 0.83, and the item-level test-retest intraclass correlation coefficients ranged from 0.67 to 0.89, indicating strong item stability over a 3-week period. Using data from the 44 items, the CFA model fit statistics supported unidimensionality (CFI 0.95, TLI 0.95, and RMSEA 0.07). The factor loadings ranged from 0.67 to 0.90. These findings provide support that the item bank measures a unidimensional factor.

Table I.

PROMIS Pediatric Meaning and Purpose, Child Self-Report Edition, Item-Level Descriptive Statistics, Reliability, and Factor Loadings; Data are from Sample 1

Item stem	Mean (SD)	Floor (% not at all)	Ceiling (% very much)	Item-total correlation (r value)	Test-retest reliability (intraclass correlation)	CFA factor loadings
I expect amazing things to happen to me.	4.03 (1.10)	3.4	44.4	0.72	0.74	0.79
I expect things to work out for the best.	4.26 (0.92)	1.1	50.6	0.74	0.75	0.80
I expect to have a job in the future.	4.64 (0.72)	0.7	74.5	0.62	0.84	0.76
I expect to have a family in the future.	4.43 (0.93)	1.6	65.1	0.57	0.82	0.67
I expect to be successful in the future.	4.42 (0.86)	1.2	60.6	0.79	0.78	0.88
I expect to enjoy my future life.	4.48 (0.81)	1.1	62.5	0.79	0.70	0.87
I expect to have a long life.	4.40 (0.89)	1.3	60.9	0.68	0.68	0.76
I expect to have success in the future.	4.39 (0.86)	1.1	58.0	0.81	0.76	0.89
I expect to achieve what I want in life.	4.27 (0.93)	1.4	52.1	0.80	0.69	0.87
I look forward to what will happen in the future.	4.36 (0.91)	1.3	57.7	0.79	0.69	0.85
I expect to succeed at what I try to do.	4.33 (0.87)	0.9	53.6	0.78	0.70	0.85
When bad things happen, I expect them to get better.	4.10 (1.01)	2.5	44.1	0.68	0.74	0.74
I have hope.	4.33 (0.92)	1.4	56.4	0.79	0.85	0.86
I am full of hope.	4.22 (0.98)	1.5	51.1	0.79	0.74	0.86
I always have hope.	4.11 (0.96)	1.4	42.2	0.70	0.80	0.77
I feel hopeful about my plans for the future.	4.37 (0.89)	1.2	58.2	0.78	0.77	0.86
I feel hopeful about my future.	4.36 (0.90)	1.4	57.2	0.83	0.82	0.89
I am positive about my future.	4.27 (0.93)	1.3	52.1	0.83	0.76	0.88
I can do almost anything if I have enough faith in myself.	4.33 (0.92)	1.6	55.7	0.76	0.73	0.83
I have goals for myself.	4.29 (0.95)	1.4	55.1	0.73	0.73	0.83
I make plans for my future.	4.21 (1.02)	2.4	52.5	0.72	0.77	0.82
I have things I want to do in life.	4.49 (0.79)	0.8	63.6	0.71	0.65	0.81
I have things I need to do in life.	4.43 (0.85)	1.4	60.7	0.72	0.66	0.80
I have things I want to accomplish in life.	4.41 (0.86)	1.0	59.5	0.75	0.71	0.84
The things I have done in the past will help me in the future.	4.10 (1.03)	2.4	45.0	0.65	0.76	0.72
I expect to achieve my goals.	4.31 (0.90)	1.2	54.1	0.75	0.70	0.83
I know where I am going in life.	3.96 (1.11)	3.8	40.6	0.71	0.78	0.79
I can reach my goals in life.	4.31 (0.88)	1.2	52.4	0.78	0.77	0.85
I want to make the most out of my life.	4.54 (0.80)	0.9	67.9	0.75	0.72	0.84
My life is filled with important things.	4.28 (0.91)	1.2	52.1	0.79	0.72	0.85
My life is important.	4.49 (0.87)	1.3	67.0	0.76	0.75	0.85
I want to do what is important.	4.36 (0.86)	0.9	55.9	0.67	0.67	0.74
I try to find meaning in life.	4.09 (1.05)	3.0	45.8	0.64	0.81	0.73
My life has meaning.	4.34 (0.96)	2.0	59.2	0.80	0.80	0.89
My life is filled with meaning.	4.13 (1.01)	2.2	46.1	0.77	0.89	0.85
I try to find purpose in life.	4.12 (1.04)	2.7	47.0	0.62	0.80	0.70
My life has purpose.	4.29 (0.97)	2.1	56.1	0.81	0.78	0.90
My life is filled with purpose.	4.21 (0.99)	2.2	51.2	0.82	0.80	0.89
I have a clear purpose in life.	3.98 (1.10)	3.4	42.3	0.77	0.68	0.84
I know what makes my life meaningful.	4.08 (1.08)	3.3	46.3	0.74	0.70	0.81
My life is filled with things that interest me.	4.25 (0.89)	1.0	48.8	0.71	0.81	0.78
I have a reason for living.	4.43 (0.92)	1.9	64.1	0.76	0.75	0.86
I am satisfied with my purpose in life.	4.20 (0.99)	2.2	49.9	0.80	0.81	0.86
People will remember me when I die.	4.16 (1.05)	2.7	50.8	0.66	0.76	0.72

Note. CFA = confirmatory factor analysis.

Although we found statistically significant item-level DIF across each of the socio-demographic variables and samples, the differences between scale scores accounting for DIF and ignoring DIF (i.e., impact) were small (all <0.1 SD units). Scores ignoring DIF were highly correlated with scores incorporating DIF (>0.99) and the scatter plots of scores ignoring and adjusting for DIF showed a nearly perfect linear relationship with no heteroskedasticity.

The range in T-scores for the full item bank spanned 5.5 standard deviations (10.9–66.2). Fewer than 1% of individuals had a floor effect, although ceiling effects were seen among 10%. Marginal reliability of the item pool was 0.98 and test–retest reliability was 0.75. The correlation between child-report and parent-proxy scale scores was 0.54.

Phase 2 Results: Item Bank Calibration

The item with the best discrimination, providing the greatest level of information about meaning and purpose, was expect to have success in the future (Table 2). The range of threshold parameters was −3.66 (expect to have a job in the future) to 0.07 (know where I am going in life). Items selected for the 4-item and two 8-item short forms are shown in Table 2. The marginal reliabilities of the short forms and full item bank (Supplementary Figure 2) show acceptable levels of precision across 5 SDs for the item bank, 4.5 SDs for SF8a and 4.0 SDs for SF4a.

Table II.

Item Response Theory Item Parameters for the 44-Item PROMIS Pediatric Meaning and Purpose Item Bank, Child Self-Report Edition, and Short Form Item Assignment; Data are from the Combined Sample

Item stem	Short forms	Item discrimination	Item thresholds
			b1	b2	b3	b4
I expect amazing things to happen to me.		2.25	−2.57	−1.81	−0.98	−0.06
I expect things to work out for the best.		2.41	−3.16	−2.21	−1.31	−0.25
I expect to have a job in the future.		2.10	−3.66	−2.75	−2.06	−1.06
I expect to have a family in the future.		1.61	−3.53	−2.57	−1.74	−0.78
I expect to be successful in the future.		3.35	−2.84	−2.20	−1.43	−0.54
I expect to enjoy my future life.	SF8a	3.29	−2.92	−2.33	−1.55	−0.57
I expect to have a long life.		2.06	−3.26	−2.45	−1.55	−0.58
I expect to have success in the future.		3.66	−2.79	−2.14	−1.39	−0.46
I expect to achieve what I want in life.		3.19	−2.79	−2.05	−1.24	−0.31
I look forward to what will happen in the future.		3.09	−2.83	−2.10	−1.36	−0.47
I expect to succeed at what I try to do.		2.91	−3.06	−2.20	−1.40	−0.35
When bad things happen, I expect them to get better.		1.93	−2.94	−2.14	−1.18	−0.03
I have hope.		3.12	−2.77	−2.05	−1.31	−0.43
I am full of hope.		2.99	−2.77	−1.89	−1.19	−0.27
I always have hope.		2.10	−3.19	−2.16	−1.16	0.031
I feel hopeful about my plans for the future.		3.24	−2.82	−2.17	−1.37	−0.48
I feel hopeful about my future.	SF8a, SF4a	3.32	−2.68	−2.08	−1.33	−0.40
I am positive about my future.	SF8a	3.58	−2.70	−1.96	−1.23	−0.31
I can do almost anything if I have enough faith in myself.		2.74	−2.84	−2.13	−1.36	−0.41
I have goals for myself.		2.63	−2.87	−2.09	−1.25	−0.40
I make plans for my future.		2.46	−2.67	−1.99	−1.19	−0.33
I have things I want to do in life.		2.55	−3.27	−2.51	−1.68	−0.64
I have things I need to do in life.		2.47	−2.99	−2.42	−1.56	−0.56
I have things I want to accomplish in life.		2.79	−3.04	−2.26	−1.47	−0.51
The things I have done in the past will help me in the future.		1.89	−2.98	−2.13	−1.16	−0.07
I expect to achieve my goals.		2.63	−3.06	−2.25	−1.32	−0.36
I know where I am going in life.	SF8a	2.29	−2.43	−1.74	−0.84	0.07
I can reach my goals in life.	SF8a, SF4a	2.94	−2.86	−2.16	−1.33	−0.29
I want to make the most out of my life.		3.04	−3.01	−2.34	−1.61	−0.75
My life is filled with important things.		2.97	−2.93	−2.10	−1.27	−0.30
My life is important.		2.97	−2.87	−2.20	−1.52	−0.73
I want to do what is important.		2.00	−3.52	−2.58	−1.55	−0.42
I try to find meaning in life.		2.01	−2.78	−2.05	−1.16	−0.12
My life has meaning.		3.44	−2.54	−2.00	−1.28	−0.50
My life is filled with meaning.	SF8a, SF4a	2.92	−2.54	−1.92	−1.09	−0.19
I try to find purpose in life.		1.83	−2.97	−2.15	−1.22	−0.15
My life has purpose.	SF8a, SF4a	3.43	−2.51	−1.96	−1.19	−0.43
My life is filled with purpose.		3.46	−2.47	−1.89	−1.10	−0.28
I have a clear purpose in life.		2.72	−2.38	−1.73	−0.82	−0.01
I know what makes my life meaningful.		2.52	−2.45	−1.82	−1.01	−0.13
My life is filled with things that interest me.		2.29	−3.30	−2.35	−1.37	−0.21
I have a reason for living.	SF8a	3.04	−2.66	−2.13	−1.48	−0.72
I am satisfied with my purpose in life.		3.16	−2.52	−1.93	−1.13	−0.24
People will remember me when I die.		1.82	−2.96	−2.11	−1.17	−0.24

Note. SF4a: 4-item meaning and purpose short form; SF8a: 8-item meaning and purpose short form that includes the 4-item form and 4 additional meaning and purpose items.

Phase 3 Results: Validity

We fit a multivariable regression model, which regressed meaning and purpose on covariates. Results indicated a significantly higher meaning and purpose for younger children and African–Americans (Table 3). We observed lower meaning and purpose for children with a special health care need, and did not observe gender, socio-economic, or ethnicity differences.

Table III.

Known-Group Validity Analyses using Multivariable Regression; Data are from Sample 2

Covariate	Beta coefficient (SE)	p value	95% CIs
Intercept	48.9 (0.6)	< 0.001	46.6 to 50.1
Age, years
8–12	1.6 (0.8)	0.049	0.1 to 3.2
13–17	referent
Gender
Female	−.01 (0.8)	0.993	−1.7 to 1.6
Male	referent
Race
White	referent
African-American or Black	2.3 (0.9)	< 0.05	0.6 to 4.1
Asian or Pacific Islander	2.1 (1.1)	0.055	−0.1 to 4.3
Other	1.2 (1.3)	0.349	−1.3 to 3.7
Ethnicity
Hispanic or Latino	0.4 (1.0)	0.691	−1.6 to 2.5
Not Hispanic or Latino	referent
Family income
<$40, 000 per year	0.7 (0.7)	< 0.321	−0.7 to 2.2
$40, 000 or more per year	referent
Special healthcare need
Yes	−2.6 (0.6)	< 0.001	−3.8 to −1.3
No	referent

The item bank and short forms showed excellent convergent validity with measures of subjective well-being (positive affect, life satisfaction, and children’s hope)–Table 4. Large negative correlations were observed with pessimism and depressive symptoms (about −0.5), intermediate negative correlations were observed for anger (about −0.3), and small negative correlations for anxiety (about −0.1).

Table IV.

Convergent and Discriminant Validity, Child Self-Report Edition

	Item bank, 44 items	SF8a, 8 items	SF4a 4 items
Convergent validity, Pearson’s r
Children’s hope scale (Sample 2, n = 455)	–	0.66	0.60
Positive affect (Sample 1, n = 225)	0.70	0.66	0.64
Life satisfaction (Sample 1, n = 225)	0.68	0.65	0.63
Discriminant validity, Pearson’s r
Youth life orientation test-pessimism subscale (Sample 2, n = 487)	–	−0.54	−0.52
Anger (Sample 1, n = 255)	−0.35	−0.32	−0.28
Anxiety (Sample 1, n = 255)	−0.16	−0.13	−0.12
Depressive symptoms (Sample 1, n = 255)	−0.45	−0.42	−0.41
Depressive symptoms (Sample 2, n = 996)	–	−0.53	−0.49
Pain-behavior (Sample 2, n = 992)	–	−0.28	−0.26

Note. SF4a: 4-item meaning and purpose short form; SF8a: 8-item meaning and purpose short form that includes the 4-item form and 4 additional meaning and purpose items.

Discussion

The PROMIS Pediatric Meaning and Purpose item bank includes 44 items that assess children’s evaluations of life as having purpose, goals to pursue, and a positive future. A child self-report edition can be used for children 8–17 years-old, while a parent-proxy edition with comparable psychometric properties can be used for children 5–17 years old. Development of the item pools involved input from children, parents, and content experts, a systematic literature review, readability analysis, translatability review, and cognitive interviews (Ravens-Sieberer et al., 2014). In this study, we conducted item and scale level classical test and modern measurement analyses following PROMIS methodological guidelines (see healthmeasures.net). Several items were removed because they had age-related DIF, local dependence with other items, or poor item discrimination. The final item pools demonstrated unidimensionality and item local independence, important assumptions of IRT modeling. The item pools were calibrated using the graded response model, and 8-item and 4-item short forms were developed.

Of the 44 items in the item bank, 12 assess optimism, 7 hope, 10 goal orientation, and 15 purpose. The item banks and short form scales have excellent reliability across a wide range of the latent variable, as assessed by marginal reliability, and excellent test–retest reliability. The 8-item short form has superior precision at the high end of the latent variable compared with the 4-item, less ceiling effect, and a greater measurement range. Choice of form will therefore balance precision with efficiency (number of items) and content.

Our results provide support for the construct validity of the measure, which showed convergent validity with extant measures of positive psychological functioning, and discriminant validity with measures of emotional distress, pessimism, and pain. We found lower meaning and purpose among children with a chronic illness, while no differences were detected by gender and ethnicity. Interestingly, we found higher levels of meaning and purpose in African–American compared with white children, a finding that merits further research. This study did not include an assessment of the associations between meaning and purpose and measures of disease activity, which is an important attribute of the clinical validity of the measure and should be evaluated in future research.

Much of the data for this study was collected from two Internet panels. Advantages of Internet panels include the efficiency with which large amounts of data can be collected, the accessibility of diverse populations, and standardization of the data collection process (Hays, Liu, & Kapteyn, 2015). Because not all individuals have access to home computers and participants of Internet panels tend to have a higher socio-economic status than the general U.S. population (Craig et al., 2013), we cannot say that the study samples were nationally representative; rather, it is fair to say that the measures have been standardized to a national, highly diverse sample. Children and parents completed the questionnaires on the computer, which calls into question the possibility of a mode-of-administration bias should the questionnaire be administered on paper. However, a meta-analysis found that the difference between paper-and-pencil versus electronic questionnaires is negligible (Rutherford et al., 2016).

The meaning and purpose measures can be used in research on the etiology, course, and outcomes of pediatric conditions. Clinicians and researchers can use them to evaluate the impact of interventions on children’s eudaimonic well-being. In addition, the PROMIS Life Satisfaction measure can be used to assess evaluated well-being, and the PROMIS Positive Affect measure to examine experienced well-being. Each of these measures has 4- and 8-item short forms, allowing for all three dimensions of positive psychological well-being to be assessed with as few as 12 items total, which takes < 2 min for a child to complete. Consistent with the resilience paradigm, the positive psychological well-being measures complement the conventional negative outcomes of depression and anxiety by providing evaluations of children’s strengths and well-being.

Funded by the NIH, the measures and their scoring manuals are freely available and ready for use in practice and research settings (see healthmeasures.net for the measures and manuals). Scales can be scored using a total summed score to PROMIS T-score conversion table or using the preferred IRT-based scoring method. Although we developed and evaluated two short forms, any batch of items from the item bank can be used to create customized short forms and scored on the T-score metric. For example, a clinician may want to select items based on their content.

Interesting directions for future research will address how meaning and purpose changes over the pediatric life course and in response to altered health states and healthcare interventions. Health challenges may compromise some children’s optimism and future orientation. For others, a renewed and strengthened sense of meaning and purpose may emerge. Positive, growth-oriented responses to stress and trauma associated with having a serious health condition may include increased appreciation of life and optimism about the future and perceived purpose in life (e.g., in advocating for oneself or others with similar conditions). These positive growth experiences have been observed among adolescent survivors of cancer (Barakat, Alderfer, Kazak, 2006). Research that advances understanding of how health and developmental challenges interact with children’s meaning and purpose will have important clinical applications (Kilmer et al., 2014).