Use of the PROMIS-29 to Identify Subgroups of Mothers with Chronic Pain

Abstract

Objective:

Children of mothers with chronic pain are at increased risk for poor health, but few studies have examined what characteristics of maternal chronic pain may be associated with children’s risk. This study identified subgroups of mothers based on patterns of pain, physical, and emotional function on the PROMIS-29 and evaluated associations between maternal subgroups and children’s pain and emotional functioning.

Methods:

Mothers with chronic pain (n = 334) completed the PROMIS-29 and reported on pain intensity, pain interference, physical functioning, anxiety, depression, fatigue, sleep disturbance, and participation in social activities. Mothers and their school-aged children also completed measures of child pain and emotional functioning.

Results:

Latent profile analysis of PROMIS domains indicated a 4-class solution (Group 1:13.5%, Group 2: 9.9%, Group 3: 43.5%, and Group 4: 32.9%). Group 4 reported the most severe pain, psychological distress, sleep disturbances, and lowest functioning. Group 1 reported the lowest pain, psychological distress, sleep disturbances, and highest functioning, while Groups 2 and 3 represented moderate symptoms. Groups significantly differed on maternal reports of children’s pain frequency, but not intensity, and children’s self-reported somatic symptoms. Further, child depressive symptoms (mother-proxy and self-reported), anxiety (mother-proxy reported), and pain catastrophizing (self-reported) differed by maternal group.

Conclusions:

Patterns of maternal symptoms and functioning were associated with pain frequency and emotional symptoms in children. Further examination of individual differences in mothers with chronic pain which may confer risk for chronic pain and psychological disorders in children is warranted.

Use of the PROMIS-29 to Identify Subgroups of Mothers with Chronic Pain

Children of mothers with chronic pain are at increased risk for pain, anxiety and depressive symptoms, and poor social functioning according to a recent meta-analysis (Higgins et al., 2015). The relation between maternal chronic pain and children’s physical and emotional health could be explained by a number of factors including genetics, social learning, general parenting and health behaviors, and shared environmental stress (Stone & Wilson, 2016). Additionally, the intergenerational transmission of risk from mothers with chronic pain to offspring may be moderated by maternal chronic pain characteristics (e.g., pain severity, physical functioning, and mental health).

Although maternal chronic pain may increase children’s risk for poor physical and mental health outcomes, it is not a sole determinant of these outcomes. A proportion of children of mothers with chronic pain will never experience clinically elevated levels of pain or internalizing symptoms (Higgins et al., 2015). Thus, rather than looking at presence or absence of maternal chronic pain in isolation it is important to examine parent pain status in conjunction with physical and mental health symptoms to more completely capture risk. Similar approaches have been used to examine subtypes of patients with chronic pain (Banta-Green, Merrill, Doyle, Boudreau, & Calsyn, 2009), pain complaints in large community samples (Hartvigsen, Davidsen, Hestbaek, Sogaard, & Roos, 2013), and youth with chronic pain (Walker, Sherman, Bruehl, Garber, & Smith, 2012a, 2012b) in order to understand patterns of symptoms which may influence individual-level risk. However, no studies, to our knowledge, have examined patterns of symptoms and functioning in parents with chronic pain to evaluate their influence on risk for pain and related outcomes in offspring.

The majority of studies conducted on the intergenerational transmission of risk for chronic pain have focused on the presence of maternal chronic pain as a categorical factor (i.e., whether or not a mother has chronic pain; Higgins et al., 2015). A few studies have identified more frequent parental pain behaviors, greater pain interference, or number of parental chronic pain sites as factors related to increased functional impairment or pain in children (Schanberg et al., 2001; Stone & Walker, 2017; Wilson & Fales, 2015). Further, studies have identified an increased prevalence of psychopathology in mothers of children with chronic pain (Campo et al., 2007; Garber, Zeman, & Walker, 1990). However, no studies have examined how comprehensive patterns of parental physical and emotional symptoms and functioning may influence children’s risk for experiencing pain or elevated anxiety or depressive symptoms.

Comprehensive multidimensional assessment is critical given chronic pain is multifaceted, often co-occurs with other somatic (e.g., fatigue and sleep disturbance) and psychological (e.g., anxiety or depression) symptoms, and can impact functioning across a variety of settings (e.g., work, recreation, social activities) (Arnow et al., 2006; Kelly, Blake, Power, O’Keeffe, & Fullen, 2011; van Weering, Vollenbroek-Hutten, Kotte, & Hermens, 2007). Assessment of all these domains is critical for defining which children may be at greatest risk for developing poor physical or emotional health outcomes. The Patient-Reported Outcomes Measurement Information System (PROMIS) developed by the National Institutes of Health offers a brief (29-item), comprehensive measure to assess patients’ physical and emotional health and functioning across seven domains including pain, depression, anxiety, sleep, and physical functioning (Gershon, Rothrock, Hanrahan, Bass, & Cella, 2010). The PROMIS-29 Profile provides an ideal measure for identifying subgroups of mothers with chronic pain because it is quick to administer, captures a wide range of physical, social, and emotional functioning domains, has strong psychometric properties, is publicly available, and can easily be implemented in both research and clinical settings.

The aims of the current study were two-fold. First, the present study utilized the PROMIS-29 to identify latent underlying classes (referred to as groups) of mothers with chronic pain based on patterns of pain, physical and emotional function. Second, the present study aimed to evaluate associations among maternal groups and physical and emotional functioning in children. Although both physical and mental health conditions in mothers are known to impact children (e.g., Goodman et al., 2011; Korneluk & Lee, 1998; Lewis, Hammond, & Woods, 1993), unique combinations of functioning across physical and emotional health domains may increase children’s risk for problems. For instance, it is hypothesized that maternal groups who have higher levels of psychological problems such as depression and anxiety in combination with elevated pain symptoms will be more likely to have children who experience higher levels of pain and psychological symptoms. The present study evaluated the association between maternal groups and child- and mother proxy-reported pain frequency, intensity, anxiety, and depressive symptoms. Further, the present study evaluated the association between maternal groups and two specific vulnerabilities, children’s somatic symptoms and pain catastrophizing, which may increase children’s risk for the development of chronic pain and associated disability (Jones, Silman, Power, & Macfarlane, 2007; Sullivan et al., 2001; Vervoort, Goubert, Eccleston, Bijttebier, & Crombez, 2006).

Methods

Participants

Mothers with chronic pain and their children (n = 334 dyads) were recruited as a part of an ongoing, longitudinal, multisite study evaluating the impact of maternal chronic pain on children’s emotional and physical functioning as they transition from childhood to adolescence. This baseline sample intentionally focused on children ages 8–12 to capture characteristics of this high-risk sample prior to adolescence when chronic pain rates increase. Eligibility criteria for mothers included: (a) the presence of a current chronic pain condition for at least the past 6 months, (b) pain currently occurs at least weekly, (c) have sought specialty medical care (e.g., physician, massage, acupuncture, etc.) for chronic pain within the past 6 months, (c) living in the same home as a biological child (ages 8–12 years) at least 50% of the time, and (d) ability to complete online questionnaires in English. Exclusion criteria for mothers included: (a) current pain attributable to a life-threatening or chronic illness (e.g., cancer or diabetes), and (b) psychiatric hospitalization within the past 6 months. Children were eligible to participate in the present study if they had the cognitive and English language abilities to complete online questionnaires. If mothers had multiple children in the eligible age range, research staff instructed the mother to select one child to participate. Research staff did not provide any guidelines or criteria for deciding which child participated.

Procedures

Mothers were recruited from three pain clinics at academic medical centers (i.e., Oregon Health and Science University, Stanford University, and University of Washington; 27.2% of sample) and through online social media (e.g., Facebook, Twitter; 51.2% of sample) and other marketing strategies (21.6% of sample) including word of mouth, research databases (e.g., ResearchMatch and local university postings), and advertisements through chronic pain support groups. The institutional review board at each academic medical center where participant recruitment occurred approved the study protocol. Mothers completed an initial contact form indicating interest in receiving more information about the study. Trained research assistants spoke with each interested individual over the phone to provide further information regarding the study, screen for eligibility, and review the informed consent and assent forms. Mothers then completed a second call which administered an extensive interview assessing familial health history and maternal pain characteristics. After completing the family history interview, dyads completed all study questionnaires through Research Electronic Data Capture (REDCap), a secure online data collection site (Harris et al., 2009).

Measures

Maternal pain characteristics.

To characterize maternal chronic pain characteristics, maternal pain locations were examined to determine the presence of generalized pain. Maternal pain locations were assessed via a body map designed for integration with the REDCap software. Generalized pain was defined as “pain in at least 4 of 5 regions (left upper, right upper, axial, left lower, and right lower)…jaw, chest, and abdominal pain are not included in generalized pain definition” (Wolfe et al., 2016). Mothers also self-reported age of onset of chronic pain and the duration of chronic pain was calculated by subtracting age of onset from age at the time of study participation.

PROMIS-29 Profile v.2.0.

The PROMIS-29 was designed to assess general physical and emotional health and functioning across 7 domains with 4 questions in each domain. Domains include: Pain Interference, Physical Functioning, Anxiety, Depression, Fatigue, Sleep Disturbance, and Ability to Participate in Social Roles and Activities. An additional pain intensity item provided an 8^th domain for analyses, but was not converted to a T-score. In the present study, all 7 domains exhibited high internal consistency (Cronbach’s α range: 0.83 – 0.94).

PROMIS Parent Proxy and Pediatric Short Forms v 1.1.

Mothers completed the PROMIS Parent Proxy Anxiety 8b and Depressive Symptoms 6b Short Forms regarding their children’s symptoms (Varni et al., 2012). Children completed the PROMIS Pediatric Anxiety 8b and Depressive Symptoms 8b Short forms. The anxiety and depressive symptom short forms asked mothers and children to rate the child’s symptoms over the past week on a ‘0’ (Never) to ‘4’ (Almost Always) scale. Both Parent Proxy and Pediatric Short Forms exhibited high levels of internal consistency in the present study (Cronbach’s α range: 0.89 – 0.94). Correlations between parent-proxy and child self-reported anxiety and depressive symptoms were 0.46 and 0.49, respectively.

PROMIS Scoring.

All PROMIS measures (Cella et al., 2010) are publicly accessible online via http://www.healthmeasures.com. For each PROMIS scale with greater than 50% of items completed, a prorated total raw score was computed based on the number of items completed and converted to T-scores based on the data tables provided through scoring manuals in the Assessment Center. In general, for the PROMIS-29 domains, a T-score of 50 represents the average for the United States general population. However, the T-scores for Ability to Participate in Social Roles and Activities and Sleep Disturbance as well as Pediatric and Parent Proxy scales represent a calibration sample which was enriched for chronic illness (Cella et al., 2010; Irwin et al., 2010).

Children’s pain.

Mothers and children both completed the Children’s Pain Questionnaire which assesses usual pain intensity via a 0–10 NRS, as well as frequency in the past 3 months. Pain frequency was measured on a scale ranging from 1 to 7 with the following anchors: 1 = Not at all, 2 = Less than 1 time per month, 3 = 1 to 3 times per month, 4 = About 1 time per week, 5 = 2 to 3 times per week, 6 = 4 to 6 times per week, and 7 = daily. These items have been validated and used extensively with children (McGrath, 1990; von Baeyer et al., 2009). Correlations between mother proxy-reported and child self-reported pain intensity and pain frequency were 0.44 and 0.60, respectively.

Children’s somatic symptoms.

Children completed the Children’s Somatization Inventory-24 (CSI-24) which asks children to rate how much they were bothered by 24 different somatic symptoms over the past 2 weeks on a 5-point scale ranging from 0-”not at all” to 4-”a whole lot” (Walker, Beck, Garber, & Lambert, 2009). The CSI-24 has strong psychometric properties and has been validated in both community and clinical samples of youth (Lavigne, Saps, & Bryant, 2012; Walker et al., 2009). Alpha reliability for the CSI-24 in the present study was 0.91.

Children’s pain catastrophizing.

Children completed the Pain Catastrophizing Scale for Children (PCS-C; Crombez et al., 2003) which measures the extent to which children engage in negative thinking associated with pain. The PCS-C consists of 13 items rated on a 5 point scale from 0-”not at all” to 4-”extremely.” Items are summed to yield a total score ranging from 0 to 52. The PCS-C exhibited high internal consistency in the present sample (Cronbach’s α = 0.91).

Data Analysis

The seven domains of the PROMIS-29 Profile and pain intensity score were entered into a latent profile analysis (LPA; a.k.a. gaussian (finite) mixture model) to identify profiles (classes) of mothers with chronic pain based on patterns of pain, physical and emotional functioning. LP models were estimated using the expectation-maximization (EM) algorithm as implemented in Mplus statistical software version 7.31. This is an iterative estimation procedure in which estimates of within-class parameters (i.e., means and standard deviations of the underlying class distributions) are used to estimate latent class probabilities (that a randomly selected observation is located in class j) and vice-versa until convergence is reached (i.e., there is very little change in the parameter estimates upon repeated steps). For each mother, a probability of belonging to each latent class was assigned, and the highest membership probability was used to make a final class assignment. Identification of the best fitting LP model was based on information criteria where smaller numbers indicate a better fit (e.g., Bayesian Information Criterion), convergence (entropy > 0.80), the proportion of the sample in each latent class (not <5%), posterior probabilities (mean probability of belonging in “most likely” class near 1.0), and the Bootstrapped LRT test (which is used to compare alternative models, e.g., k vs. k-1 classes) (Vermunt & Magidson, 2002). Interpretation of the final LP solution was made by examining the means and standard deviations of the seven domains of the PROMIS-29 and pain intensity score for each class (group). Analysis of Covariance (ANCOVA), controlling for maternal demographics, with LSD post-hoc tests of estimated marginal means were used to examine the association between maternal groups and PROMIS domains, in order to determine statistically significant group differences on domains. Chi-square and ANOVA analyses examined group differences on demographic factors. Two MANCOVA analyses, controlling for child age and sex, examined significant differences between maternal groups on children’s pain, somatic symptoms, and emotional functioning.

Missing data.

Initially, 335 dyads were enrolled, but one mother was excluded from the present study because she did not complete any of the PROMIS-29 Profile items; thus, final sample size for LPA was 334 mothers. Because MANCOVA analyses required both parent-proxy measures and child-report measures to be complete to be included, the sample size is reduced in this portion of the analyses. The final sample sizes for MANCOVAs were 290 for pain and somatic symptoms and 295 for emotional functioning. Mothers with complete versus incomplete data were similar on key demographic variables, pain intensity, and pain interference. However, mothers with incomplete data reported higher levels of anxious and depressive symptoms compared to those with complete data. Thus, data was considered missing not at random and presented data may be biased slightly towards mothers with lower levels of anxiety and depression.

Results

Demographics

On average, mothers were 40.15 years old (SD = 5.80), predominately Caucasian (91.9%), and married (75.2%), and children were 10.37 years old (SD = 1.39), predominately Caucasian (91.6%), and female (51.8%). Close to half of mothers (43.4%) met criteria for generalized pain based on reported pain locations (Wolfe et al., 2016). The majority of mothers reported multiple chronic pain conditions. Most frequently reported chronic pain conditions included fibromyalgia (n = 127, 37.9%), headaches or migraines (n = 116, 34.6%), and back pain (n = 89, 26.6%). On average, mothers had experienced chronic pain for 15.75 years (SD = 9.93) and average age of chronic pain onset was 24.45 years of age (SD = 10.58).

Latent Profile Analysis

LPA constructed from the PROMIS-29 domains indicated models with 3–5 classes fit well (Table 1). The 4-class model was chosen as the best fit based on graphical examination of class means and interpretability. See Figure 1 for means on each PROMIS-29 domain by class (group) and Table 2 for brief descriptions for each group based on the PROMIS-29 domains. Group 1 (13.5% of sample, n = 45) was characterized by lower pain, low anxiety and depressive symptoms, low sleep disturbances and fatigue, and average physical and social functioning. Group 2 (9.9%, n = 33) and Group 3 (43.5%, n = 146) had moderate pain intensity, pain interference, sleep disturbances and fatigue. However, Group 2 reported lower levels of physical and social functioning than Group 3, while Group 3 reported higher levels of anxiety and depressive symptoms than Group 2 (see Table 4 for results from post-hoc analyses examining group differences on PROMIS domains). Group 4 (32.9%, n = 110) reported severe pain, high levels of anxiety and depressive symptoms, severe sleep disturbances and fatigue, and low physical and social functioning.

Table 1.

Fit Statistics for Models with 2–5 Classes based on PROMIS-29 Domains

Classes	AIC	BIC	SABIC	Boot LRT	Entropy
2	16628	16723	16644	<.001	.84
3	16432	16561	16453	<.001	.85
4	16344	16507	16371	<.001	.85
5	16254	16452	16287	<.001	.84

Figure 1.

PROMIS-29 Profile Subgroups in Mothers with Chronic Pain

Note. Pain Intensity represents patient’s 0–10 numeric rating scale pain score scaled to a 0–100 score.

Table 2.

Descriptions of Maternal Groups Identified by LPA

Group	Description based on PROMIS-29 Domains
1	Low pain and pain interference, low anxiety and depression, low sleep disturbances and fatigue, high physical and social functioning
2	Moderate pain and pain interference, low anxiety and depression, moderate sleep disturbances and fatigue, low physical and social functioning
3	Moderate pain and pain interference, high anxiety and depression, moderate sleep disturbances and fatigue, moderate physical and social functioning
4	Severe pain and pain interference, high anxiety and depression, high sleep disturbances and fatigue, low physical and social functioning

Table 4.

Mother PROMIS Domain Means by Group

PROMIS-29 Domains	Group 1 M (SD)	Group 2 M (SD)	Group 3 M (SD)	Group 4 M (SD)	F value	p-value	Group Comparisons
Pain Intensity	4.60 (1.78)	5.76 (1.82)	5.64 (1.44)	6.85 (1.24)	20.02	<.001	1<2&3<4
Physical Functioning	49.40 (6.93)	37.20 (5.26)	40.12 (4.52)	34.72 (4.36)	74.06	<.001	1>3>2&4
Anxiety	50.23 (6.50)	48.45 (6.22)	61.20 (5.48)	64.48 (6.50)	90.04	<.001	1&2<3<4
Depression	45.79 (6.08)	43.84 (4.12)	58.44 (4.84)	63.03 (5.95)	145.96	<.001	1&2<3<4
Fatigue	53.31 (5.42)	63.15 (6.50)	62.38 (6.01)	71.76 (4.59)	96.19	<.001	1<2&3<4
Sleep Disturbance	50.87 (5.96)	55.68 (6.77)	56.16 (5.18)	64.06 (6.74)	47.62	<.001	1<2&3<4
Social Roles	49.36 (5.17)	38.33 (5.65)	41.73 (3.98)	34.06 (4.74)	94.73	<.001	1>3>2>4
Pain Interference	57.44 (4.65)	63.92 (4.38)	63.24 (4.02)	71.18 (4.26)	94.76	<.001	1<2&3<4

Note. F statistics from ANCOVA analyses controlling for maternal demographic factors (age, annual income, marital status). Raw means are reported in table. Group comparisons indicate significant differences on LSD post-hoc tests of estimated marginal means at the p < .05 level.

Results from chi-square tests indicated maternal groups significantly differed on some demographic and pain characteristics (Table 3). Mothers in Group 4 were less likely to be employed and reported lower income brackets compared to mothers in Group 1 (p < 0.05). A greater proportion of mothers in Group 4 were recruited via social media, compared to the proportion of mothers in Group 1 (p < 0.05). Further, a greater proportion of mothers in Group 2 and Group 4 reported generalized pain compared to the proportion of mothers in Group 1 (p < 0.05).

Table 3.

Maternal Demographic Factors by Group

Demographic Variable	Group 1 (n = 45)	Group 2 (n = 33)	Group 3 (n = 146)	Group 4 (n = 110)	Group Comparisons
Maternal age in years, M (SD)	41.76 (5.81)	40.18 (5.67)	39.98 (5.86)	39.82 (5.07)	ND
Marital status, % (n)     Married     Divorced or separated     Remarried     Never married	86.4% (38) 11.4% (5) 2.3% (1) 0.0% (0)	81.8% (27) 15.2% (5) 0.0% (0) 3.0% (1)	81.9% (113) 10.1% (14) 2.9% (4) 5.1% (7)	73.7% (73) 19.2% (19) 3.0% (3) 4.0% (4)	ND ND ND ND
Employment status, % (n)     Employed, full or part time     Unemployed, by choice     Unemployed     Receiving disability	75.6% (34) 15.6% (7) 2.2% (1) 6.7% (3)	54.5% (18) 27.3% (9) 6.1% (2) 12.1% (4)	52.1% (75) 25.7% (37) 11.8% (17) 10.4% (15)	27.5% (30) 20.2% (22) 21.1% (23) 31.2% (34)	1>3;2&3>4 ND 1<4 1&3<4
Annual household income, % (n)     $25,000 or less     $25,001 - $49,999     $50,000 - $79,999     $80,000 - $119,999     $120,000 - $149,999     $150,000 or more	2.3% (1) 11.4% (5) 13.6% (6) 34.1% (15) 9.1% (4) 29.5% (13)	9.1% (3) 18.2% (6) 21.2% (7) 30.3% (10) 6.1% (2) 15.2% (5)	8.3% (12) 22.9% (33) 20.8% (30) 29.9% (43) 6.9% (10) 11.1% (16)	21.2% (22) 29.8% (31) 16.3% (17) 21.2% (22) 6.7% (7) 4.8% (5)	1,2,&3<4 ND ND ND ND 1>2&3
Recruitment source, % (n)     Pain clinic (OHSU, Stanford, UW)     Social media     Other (word of mouth, support groups, other websites)	44.4% (20) 33.3% (15) 22.2% (10)	33.3% (11) 39.4% (13) 27.3% (9)	25.3% (37) 50.7% (74) 24.0% (35)	20.9% (23) 62.7% (69) 16.4% (18)	1 > 4 4 > 1 ND
Widespread pain, % (n)	22.2% (10)	56.3% (18)	39.4% (56)	60.4% (61)	4 >1&3; 2>1
Number of pain locations, M (SD)	5.0 (3.71)	8.83 (5.47)	6.47 (4.56)	9.48 (4.86)	4&2>1&3
Years since chronic pain onset, M (SD)	13.87 (10.29)	15.97 (9.57)	15.63 (9.28)	16.56 (10.73)	ND

Note. For categorical variables, group comparisons indicate significant differences on z-test comparisons of proportions at p < 0.05 level. For continuous variables, group comparisons indicate significant differences on LSD post-hoc tests at p < 0.05 level. ND = no significant differences between groups.

Relation of Maternal Subgroups to Children’s Pain and Emotional Functioning

Children’s pain and somatic symptoms.

A MANCOVA analysis, controlling for child age and sex, examined differences between maternal groups on mother proxy- and child self-reported pain frequency and intensity and child-reported somatic symptoms (see Table 5). Results indicated significant between-group effects on univariate tests for mother-reported child pain frequency (F[3, 284] = 3.03, p = 0.03, η_p² = 0.03) and child self-reported somatic symptoms (F[3, 284] = 2.69, p = 0.047, η_p² = 0.03). Post-hoc tests on estimated marginal means indicated mothers in Group 4 reported their child experienced significantly more frequent pain than mothers in Group 1 and Group 3. Children of mothers in Groups 2, 3, and 4 reported significantly greater somatic symptoms than children of mothers in Group 1. The overall multivariate tests were non-significant for group as children’s pain intensity did not differ by group based on child and mother report (Pillai’s Trace = 0.68, p = 0.19).

Table 5.

Children’s Symptoms by Maternal Subgroup: MANCOVA Results

	Group 1 M (SD)	Group 2 M (SD)	Group 3 M (SD)	Group 4 M (SD)	F value	p-value	Group Comparisons
Pain and Somatic Symptoms
Mother proxy-report:
Child pain frequency*	2.86 (1.69)	3.13 (1.52)	3.14 (1.42)	3.60 (1.57)	3.03	0.03	4>3&1; 1=2=3
Child pain intensity	2.73 (2.27)	3.00 (2.17)	3.02 (2.24)	3.46 (2.23)	1.49	0.22	NA
Child self-report:
Child pain frequency*	2.16 (1.77)	2.17 (1.73)	2.31 (1.76)	2.66 (1.75)	0.83	0.48	NA
Child pain intensity	3.47 (2.40)	3.14 (2.76)	3.38 (2.23)	3.55 (2.37)	0.19	0.91	NA
Child somatic symptoms	9.09 (9.58)	14.41 (13.26)	14.81 (13.07)	15.95 (12.60)	2.69	0.04	2,3,&4>1
Emotional Functioning
Mother proxy-report:
Child anxiety	45.37 (10.43)	47.83 (9.98)	53.29 (9.44)	54.07 (10.36)	9.52	< .001	3&4>1&2
Child depression	43.42 (7.24)	45.98 (6.55)	50.31 (8.81)	52.13 (9.86)	11.23	<.001	3&4>1; 1=2; 2=3
Child self-report:
Child anxiety	45.69 (9.85)	47.74 (10.18)	48.95 (9.66)	49.41 (10.31)	1.48	0.20	NA
Child depression	44.37 (8.47)	46.83 (9.69)	48.30 (8.91)	50.50 (10.03)	4.72	< 0.01	3&4>1; 4>2; 2=3
Child pain catastrophizing	7.90 (6.72)	12.56 (10.21)	12.38 (10.05)	13.01 (9.96)	3.27	0.02	2,3,&4>1

Note. F values are based on univariate tests. Group comparisons indicate significant differences on post-hoc tests at the p < .05 level. NA = not applicable because the omnibus test was non-significant.

^*Child pain frequency is measured on a 1 to 7 scale, where 2 = “less than 1 time per month”, 3 = “1 to 3 times per month”, and 4 = “about 1 time per week.”

Children’s emotional functioning.

A MANCOVA analysis, controlling for child age and sex, examined significant differences between maternal groups on mother proxy- and child self-reported anxiety and depressive symptoms and child self-reported pain catastrophizing (see Table 5). Results indicated significant differences between maternal groups for variables of children’s emotional functioning (Pillai’s Trace = 0.15, p < 0.001, η_p² = 0.05). Specifically, univariate tests indicated mother proxy-report of children’s depressive symptoms significantly differed by maternal group (F[3, 289] = 11.23, p < 0.001, η_p² = 0.14) and children’s self-report (F[3, 289] = 4.72, p = 0.003, η_p² = 0.05). Post-hoc tests indicated mothers in Groups 3 and 4 reported significantly greater child depressive symptoms compared to Groups 1 and 2. Children of mothers in Groups 3 and 4 reported significantly greater depressive symptoms compared to Group 1. Further, mother proxy-report of children’s anxiety significantly differed by maternal group (F[3, 289] =9.52, p < 0.001, η_p² = 0.09), but not children’s self-report (F[3, 289] = 1.48, p =0.22, η_p² = 0.02). Specifically, mothers in Group 4 perceived their children as experiencing significantly more anxiety than mothers in Group 1. Although children’s self-reported anxiety was non-significant, children’s self-reported pain catastrophizing significantly differed by maternal group (F[3, 289] = 3.27, p =0.02, η_p² = 0.03). Specifically, children of mothers in Groups 2, 3, and 4 all reported significantly higher pain catastrophizing compared to children of mothers in Group 1.

Discussion

Results provide initial support for utilizing the PROMIS-29 to identify underlying subgroups of mothers with chronic pain conditions. In the current sample, this empirically driven approach was able to identify latent subgroups of patients across physical and psychosocial domains of functioning. Notably, the PROMIS system offers the possibility of capturing functioning across many domains of physical and emotional health and functioning with relatively low response burden. A few studies have used PROMIS measures in adults with chronic pain conditions (e.g., Sturgeon, Dixon, Darnall, & Mackey, 2015), and in children with chronic pain related to medical conditions (e.g., Brandon, Becker, Bevans, & Weiss, 2017; Dampier et al., 2016; DeWalt et al., 2015) and musculoskeletal pain (Kashikar-Zuck et al., 2016), but to our knowledge no previous studies have identified naturally occuring subgroups of patients based on a number of key domains assessed via the PROMIS measurement system. This type of approach might be a particularly useful way of characterizing functioning among cohorts of individuals being followed prospectively in studies of health outcomes, or in high-risk samples where research seeks to identify individuals or offspring with varying levels or types of risk for poor outcomes.

Results of the LPA identified underlying patterns of functioning in this sample of mothers with chronic pain, suggesting that the presence of maternal chronic pain in and of itself does not fully capture level of risk for children of these mothers. In fact, mothers with chronic pain present with different profiles of functioning that represent distinct patterns of risk for their children. Approximately one-third of mothers in this treatment-seeking group were classified in Group 4 which reported the highest levels of pain and psychological symptoms and the lowest levels of physical and social functioning. This group with the highest level of comorbidity also experienced the highest levels of sleep disturbance and fatigue, indicating that sleep disturbance may be perpetuating severe pain and psychological comorbidities in this subgroup of mothers. Unknown is the extent to which sleep disturbances in mothers with chronic pain may contribute to sleep difficulties in their children which could confer risk for the development of psychopathology or chronic pain (James & Hale, 2017).

Approximately 13% of the sample (Group 1) reported functioning well across all domains, suggesting that the experience of chronic pain does not negatively impact functioning for all mothers. Further, children of mothers in Group 1 reported the lowest somatic symptoms, pain catastrophizing, and depressive symptoms, indicating average maternal physical and emotional funcitoning in the presence of chronic pain may serve as a protective factor for their children. Of note, Group 1 is also comprised of the highest percentage of mothers recruited from academic medical center pain management clinics; it is possible that these patients are receiving more comprehensive or more effective pain treatment than participants in the other groups. Understanding more about what biopsychosocial or treatment factors might contribute to positive function despite chronic pain will be an important future direction, as this information can inform interventions that have the potential to influence pain experiences across multiple generations.

Groups 2 and 3 represent mothers who are functioning similarly in the moderate range in terms of pain intensity, pain interference, sleep disturbance, and fatigue, but show distinct patterns across other domains, with Group 3 being significantly higher in anxiety and depression, and Group 2 having significantly lower physical and social function. These distinct patterns could present unique patterns of risk for children, and may represent patient subgroups who could benefit from tailored or distinct treatment approaches in the context of multidisciplinary pain management. Overall, the LPA results highlight the diversity of maternal functioning contexts that the children of women with chronic pain experience.

In this sample, maternal subgroup membership was associated child pain frequency (mother report) and somatic symptoms (child report), but not with child pain intensity. Although mother report of child pain frequency was significantly related to group, the highest frequency of pain reported (Group 4) was slightly under 1 time per week, indicating that the majority of the children in the sample are not experiencing pain very frequently. Groups 2, 3, and 4 reported greater somatic symptoms compared to children in Group 1, indicating potentially heightened risk for chronic pain development.

Although subgroup membership was associated with maternal and child reports of child depressive symptoms and maternal reports of child anxiety as measured via PROMIS, it was not associated with child self-report on anxiety. Of note, previous research with PROMIS pediatric scales has found substantial discrepancies between parent and child reports, including on anxiety symptoms (Varni et al., 2015). While chronic pain in a parent or child is associated with greater agreement between parent proxy and child reports on child pain interference items (Varni et al., 2015), it is unknown whether parent or child chronic pain status might impact discrepancies between child and parent report of anxiety or depressive symptoms, but this might be examined in future work.

Subgroup membership was also associated with child pain catastrophizing, with the children of mothers in Group 1 reporting lower pain catastrophizing than the other subgroups. This suggests that for the children of mothers in Group 1, having a mother who is functioning in the average range and modeling behaviors consistent with activity engagemet despite pain may be a protective factor that buffers against pain catastrophizing. Overall, this pattern of results provides some initial support for a model of intergenerational transmission of chronic pain in which child vulnerabilities, such as pain catastrophizing, are influenced by parental pain and functioning (Stone & Wilson, 2016). Future longitudinal analyses with this sample will be able to identify how child vulnerabilities contribute to child pain experiences over time.

The present study had a number of strengths including the focus on a clinically relevant, treatment seeking sample of mothers with chronic pain, utilization of the PROMIS measures, and novel application of latent profile analysis to identify naturally occurring maternal subgroups. However, it is important to note some limitations regarding the generalizability of the present study. First, analyses of missing data indicated mothers who did not complete all study questionnaires reported more anxious and depressive symptoms than mothers with complete data. Thus, the data presented may be slightly biased towards mothers with lower anxiety and depressive symptoms. Second, the data for the present study is cross-sectional and conclusions cannot be made regarding the timing or directionality of effects. The larger study specifically selected the 8 to 12-year-old period in order to capture youth living with a mother with chronic pain prior to adolescence, when the incidence of pain increases. Children were not selected based on experiencing pain themselves; more frequent pain and higher somatic symptoms in this sample may represent youth with greater vulnerability for developing pain problems in the future. Future longitudinal data will determine the extent to which children’s symptoms in middle childhood confer risk for the development of chronic pain. It is also possible that other factors that differ between these maternal subgroups may drive the associations between subgroup and child pain and function. For instance, Group 4 had the highest rates of generalized pain, lowest annual income, and were less likely to be employed. Shared genetic vulnerability for centralized pain or shared stress associated with sociodemographic factors could partially account for heightened risk for poor outcomes in children of mothers in Group 4.

In summary, the PROMIS-29 provided a low-burden means of assessing symptoms and functioning in a sample of mothers with chronic pain conditions, and allowed for identification of underlying subgroups representing distinct patient profiles. Furthermore, these subgroups differed in mother and child perceptions of children’s pain-related and psychological functioning, suggesting that distinct profiles of maternal functioning may confer additional specific or general risk for poor outcomes in children.