Family Chaos and Asthma Control

The Asthma Action at Erie Trial, described in a previous publication,¹⁸ involves a partnership with Erie Family Health Centers, a Federally Qualified Health Center serving mainly low-income minority families across multiple sites in the Chicago area. Participants were enrolled as parent-child dyads. Inclusion criteria were as follows: (1) patient at a participating Erie clinic (6 sites; 10% total participants in 3 smaller sites, 35% in 2 middle-sized sites, and 55% from a larger site), (2) aged 5 to 16 years, (3) live with a caregiver at least 5 days per week, and (4) uncontrolled asthma (≥1.25 on the Asthma Control Questionnaire,^19,–21 <20 on the Asthma Control Test [ACT] or Childhood Asthma Control Test^22,–24 [cACT], or a report of at least 1 oral corticosteroid burst for asthma in the past year).²⁵ Families were excluded if they could not speak English or Spanish, they did not have permanent housing, caregivers did not have permanent custody of the child, or the child had significant developmental delays or comorbidities limiting their participation. Recruitment occurred from March 2016 to August 2017. Of 1688 potentially eligible children, 223 families were enrolled and randomly assigned to 1 of 2 community interventions.¹⁸ 

All study procedures were approved by institutional review boards and the Erie Research Committee. Caregivers and children older than 7 years provided signed consent and assent, respectively. Research assistants collected data in participants’ homes via computer. Research assistants asked questions verbally, showing prompt cards with responses. For privacy, caregivers completed depression and PTSD measures using computers. Asthma control was captured in multiple ways. Asthma control was measured by using the ACT (children aged 12 years and older) and cACT (children under 12 years of age).^22,23 Families were also asked the number of days out of the past 14 with activity limitation to capture this powerful domain.²⁶ The Asthma Functional Severity Scale (AFS)²⁷ assessed symptoms over the past year. Families were asked about asthma medications. For children with an inhaled corticosteroid (ICS) controller, an electronic medication monitor was attached for 2 weeks to record daily actuations. Emotional triggers for asthma were asked of the child and/or caregiver via the item, “Do strong emotions trigger [child]’s asthma?” (yes, no, or do not know). Demographics, medications, triggers, and self-reported behaviors were assessed through questions and observations.

Parent depression symptoms were measured via the 9-item Patient Health Questionnaire (PHQ-9).^28,29 Parent PTSD symptoms were captured via the Short Form of the PTSD Checklist–Civilian Version,³⁰ a 6-item measure widely used in trauma research.

Depression symptoms for children ages 7 years and older were assessed by using the Children’s Depression Inventory 2 (Short Form),³¹ a 12-item measure of the affective, behavioral, and cognitive symptoms of depression occurring over the past 2 weeks. The 28-item Child Report of Post-traumatic Symptoms³² assessed broad posttraumatic symptomatology experienced over the past 7 days for youth aged 7 years and older.

Family chaos was examined by using the Confusion, Hubbub, and Order Scale (CHAOS),³³ which assesses the level of commotion, disorganization, and routine within the household (eg, “We are usually able to stay on top of things”) on a yes-or-no scale. Total scores range from 0 to 15; higher scores reflect greater chaos and disorganization. The CHAOS has good validity, reliability, and stability over a 12-month period and thus is considered to capture chronic family conditions.³³ Parent social support was assessed by using the 8-item Patient-Reported Outcomes Measurement Information System³⁴ Emotional, Instrumental, and Informational Support scales, which were developed and validated to measure social function in adults.

Age-specific ACT measures were combined into a single measure to retain all subjects in the analyses by rescaling the ACT to the range of the cACT. Generalized linear regression models (GLMs) examined the key psychosocial predictors, identified a priori as guided by the literature, on asthma outcomes. All models controlled for demographic characteristics identified as significantly related to asthma outcomes by using backward elimination with P < .10 (child age, sex, ethnicity [Hispanic, yes or no], BMI, maternal education, and maternal language). Because of the multisite design, models also controlled for site effects. For brevity, only the effects for psychosocial predictors are discussed. Complete case analyses were implemented because the proportion of missing data was negligible. Residuals from all models were evaluated for normality and outliers; no departures from normality or outliers were identified.

To facilitate comparison with previous work and determine the relative weights of child-, parent-, and family-level factors, a series of GLMs examined the hypothesized relationships between psychosocial predictors and asthma outcomes through a 2-step process to identify the independent and relative associations with each independent variable. We were interested in child-parent associations and child-child associations separately as well as their joint associations on child outcomes. First, separate models examined the bivariate relationships between each parent- and family-level predictor and the ACT (controlling for significant demographics). Next, we examined the set of significant parent and/or family predictors identified in the bivariate models (P < .05), which allowed for the examination of adjusted effects of each predictor, controlling for all other parent and/or family and demographic variables in the model. Only significant factors were included to enhance parsimony and model fit. This approach was then followed for all child-level predictors. Last, a final GLM including all parent, child, and family factors that were identified as significant in the bivariate models was estimated to examine the relative contributions of each predictor, controlling for all other child-, parent-, and family-level predictors and demographic variables in the model. All models were repeated for the AFS and activity limitation. Additionally, separate logistic regression models examined each of the significant parent-, family- and child-level variables (identified in the GLMs described above) as predictors of emotional triggers of asthma symptoms (dichotomized as yes or no). Because of age restrictions of the child depression and PTSD measures, there were a total of 153 participants in select analyses.

To examine our hypothesis that family factors may mediate the link between parent mental health and child asthma outcomes, we estimated a full serial mediation model with family chaos and emotional triggers as potential mediators of parental depression and child ACT. The GLMs helped identify the key relationships to include in the mediation model (ie, parent depression and family chaos), and emotional triggers were included as an intermediary pathway between psychosocial factors and the ACT. Linear regression was used for family chaos, and logistic regression was used to model emotional triggers (dichotomized as yes or no), controlling for site, child age, and child ethnicity. Indirect effects were estimated as a product of regression coefficients with bootstrapped SEs. Analyses were conducted in SAS 9.4 (SAS Institute, Inc, Cary, NC) and Mplus 8.

Table 1 presents parent demographic characteristics. Caregivers were predominantly women (96.9%) and Hispanic (85.7%). Parents had low educational attainment (66.4% high school diploma or general education diploma or less) and income (62.3% made <$60 000). Child demographics and asthma characteristics are presented in Table 2.

The majority of the sample (55.7%) demonstrated uncontrolled asthma on the ACT. Asthma morbidity was significant, with high rates of oral corticosteroid bursts (60.4%), emergency department visits (38.3%), and hospitalizations (16.7%) for asthma in the past 12 months. Only 82.1% of children had a quick-relief medication, and 44% had an observed ICS controller medication. Children were taking approximately half (mean = 53.4) of prescribed ICS medications via monitor measurement. More than one-third (36.8%) reported strong emotions as an asthma trigger. Asthma outcomes (ACT, 12-month severity, and activity limitations) were moderately correlated (r = −0.448 to −0.511; P < .0001).

Table 3 presents mental health characteristics. Parents reported elevated depression and PTSD symptoms. Specifically, 14.8% endorsed moderate to severe depressive symptoms, 19.3% met criteria for PTSD, and 46.6% reported a traumatic event. Children reported similarly high levels of depression (18% elevated or very elevated), and 51.3% endorsed clinically significant PTSD symptoms. Families reported a wide range of chaos scores (0–13) but mean scores were low (3.2) comparable to other pediatric samples (2.84³⁵ and 3.89³⁶) and the original validation study (3.3³³). Family social support was within the normal range. Table 4 displays Pearson correlation coefficients for the psychosocial variables. Results suggest that psychosocial measures were related but largely distinct constructs.

Outcomes of the bivariate models examining each parent- and family-level predictor of the ACT are presented in Table 5 (top portion). This model summarizes variable associations in terms of estimated β coefficients, which represent the predicted change in outcome per 1-unit increase in covariate value. For example, a 1-point increase in parent PHQ-9 score corresponds to a decrease of 0.2 in predicted ACT scores (β = −.20; SE = 0.06). Parent depression, but not PTSD, symptoms were significantly related to ACT scores. Parent emotional support showed a modest but significant relationship with the children’s ACT scores. Instrumental and informational support were not significant. Family chaos demonstrated a robust effect on ACT scores. With all significant parent- and family-level predictors examined together in the multiple model (Table 5, bottom portion), only family chaos retained significance, suggesting that the associations between parent depression and parent emotional support with the ACT were accounted for by the effects of family chaos. This adjusted model accounted for 16.4% of the variance in ACT scores, with 31.5% of explained variance being attributed to family chaos.

Table 6 presents the outcomes of the models examining child-level predictors. In the separate bivariate models (top portion), child depression and PTSD symptoms were each related to ACT scores, and family chaos was significant. To explore the relative influences of child depression versus PTSD symptoms, a second model examined child depression and PTSD symptoms together on the ACT; only depression remained significant at a trend level (P = .062). The multiple regression model (bottom portion) included child depression, PTSD, and family chaos; only family chaos remained significant. This model accounted for 26.2% of the variance in ACT scores, with family chaos contributing 23.8% of explained variance.

To evaluate the relative contributions of parent versus child influences on asthma, child and parent depression were examined together in a model (N = 153). Child depression retained significance (β = −.11; SE = 0.03; P = .002), but parent depression did not (β = −.11; SE = 0.07; P = .126). Last, when all significant child-, parent-, and family-level predictors were examined in a final multiple regression model (Table 7), only family chaos remained significant. To directly compare the magnitude of effects, semipartial standardized estimates were calculated for each predictor, providing an estimate of the adjusted correlation between the predictor and ACT and controlling for all other variables in the model. Family chaos showed the strongest relationship with the ACT, followed by child depression.

Analyses examining activity limitation revealed an identical pattern of results. Analyses examining the AFS produced the same pattern of results but weaker relationships. Logistic regression models indicated that the odds of emotional triggers were significantly increased with higher parent depression (odds ratio [OR] = 1.13; β = .12; SE = 0.03; P < .001), family chaos (OR = 1.13; β = .12; SE = 0.05; P = .02), and child depression (OR = 1.03; β = .03; SE = 0.02; P = .06).

We explored family chaos and emotional triggers as mediators of the parent depression-ACT association (Fig 1). Findings indicated significant indirect paths between parent depression and the ACT through family chaos and separately through emotional triggers. When controlling for family chaos and emotional triggers in the model, the parent depression-ACT direct path became nonsignificant, suggesting full mediation. The path via family chaos and emotional triggers was not significant. Findings suggest that the effect of parent depression on ACT scores is mediated separately through family chaos (β = −.09; SE = 0.04; bootstrapped bias corrected confidence interval [CI] = −0.17 to −0.02; 31% of total indirect effect) and emotional triggers (β = −0.17; SE = 0.11; CI = −0.48 to −0.03; 60% of total indirect effect).

To explore what the family chaos construct may be capturing, post hoc analyses examined associations between family chaos and asthma management measures and income. Family chaos was significantly and inversely related to parental help with medication, as reported by the parent (r = −0.174) and child (r = −0.176; P = .009). The association with adherence was not significant (r = −0.070; P = .550; n = 75). Family chaos was not related to the presence of a controller (t[221] = −0.83; P = .408) or reliever medication (t[221] = −0.65; P = .517) and was unrelated to income (r = −0.069; P = .404). Additionally, we examined ACT scores at the top and bottom quartiles of family chaos scores. For youth with the greatest family chaos (top quartile), the average ACT was 15.58 (SD = 5.59), whereas for the lowest quartile of family chaos, the average ACT was 18.94 (SD = 6.75). This represents a clinically meaningful difference in levels of ACT scores across ranges of family chaos.³⁷ Nonparametric methods confirmed that ACT scores worsened linearly with increasing levels of chaos.

In this study of urban minority youth with uncontrolled asthma, parents and children both demonstrated elevated rates of depression and PTSD symptomatology that exceed prevalence rates in the general population.^14,38 We hypothesized that parent and child psychopathology would influence asthma outcomes and that family chaos and social support would serve as mechanisms underlying these relationships. Findings provided support for the hypotheses: parent and child depression symptoms were each predictors of worse asthma control, asthma severity, activity limitations, and emotional triggers of asthma when examined independently and showed stronger relationships with asthma than parent or child PTSD. Family chaos also emerged as a robust predictor of asthma outcomes even when controlling for parent and child depression and demographic covariates (eg, education level) as well as a mediator underlying the relationship between parent depression and asthma. Contrary to predictions, social support was less relevant to asthma control. To our knowledge, this study is the first to identify the key role of family chaos as a mechanism linking parent depression and child asthma outcomes.

Consistent with the asthma literature, parents in this study endorsed elevated rates of depression and anxiety.^14,39,–41 Although PTSD prevalence was high in this sample, parent PTSD was unrelated to asthma outcomes. Parent depression, however, showed significant associations with asthma control and severity. Numerous studies suggest that parent depression confers risk for child asthma severity.^42,43 Complex transactional relationships underlying caregiver mental health and child asthma outcomes likely exist. The stress of caring for a child with chronic illness may influence the development of parent depression.^3,44 Alternatively, maternal depression may play a role in the development and exacerbation of child asthma via influences on the infant stress response and immune function early in life.^40,45,–48 Parent depression may also give rise to a set of negative practices and beliefs that interfere with asthma management (eg, nonadherence and negative beliefs about asthma treatment).^44,49 

As with previous literature, we also found that poorer asthma control was related to increased symptoms of child depression^8,50 and PTSD.^4,51 Yet only child depression, not PTSD, was associated with asthma when examined simultaneously. Given that our PTSD measure did not assess a specific trauma event, it may have tapped into more general distress that was highly correlated with depression. Correlation analyses support the strong relationship between these factors (ie, r = 0.640); hence, effects for PTSD may have been accounted for by the depression-asthma relationship. It is also possible that child depression is more influential than anxiety for asthma control, as past work has shown.^5,8,11 Biological models^8,52,–54 support a specific depression-asthma mechanism whereby emotional dysregulation may lead to alterations in the autonomic nervous, hypothalamic-pituitary-adrenal, and immune and inflammatory systems that can impair child pulmonary function, particularly in the presence of stress. Notably, child depression remained significant when controlling for parent depression, suggesting a unique relationship with asthma outcomes.

Our findings also extend the literature on psychosocial mechanisms underlying asthma control. Specifically, mediation analyses suggest that increased parental depressive symptoms lead to greater family chaos, which in turn is associated with poorer child asthma control, although the cross-sectional design limits findings; thus, interpretations are made cautiously. Family chaos is a unique construct encompassing the high levels of stimuli, unpredictability, and lack of routine and rules in the home^33,55 that is distinct from other measures of psychosocial adversity as well as family socioeconomic status.⁵⁶ In the extant literature, family chaos was associated with worse glycemic control in youth with type 1 diabetes mellitus.^57,58 Specific to asthma, other indicators of family organization, including children routinely sleeping away from home⁵⁹ and single-parent households,⁶⁰ were associated with asthma-related readmission, even when controlling for parent psychological distress, in a large sample of youth hospitalized for asthma. Consistent with our work, past research has demonstrated that family chaos is related to parent depression^61,62 and is a mediator of the relationship between parent depression and negative child psychosocial outcomes.⁶² 

Parents with depressive symptoms may experience executive functioning deficits and high levels of stressors that interfere with the establishment of consistent routines, stability, and organization in the home⁶² that are essential for asthma management.^63,64 Indeed, we found that family chaos was related to lower parental medication management. Family chaos may also affect asthma via biological mechanisms. For example, family chaos and poor family routines predicted systemic inflammation and proinflammatory cytokine production in a study of healthy adolescents³⁵ and youth with asthma, possibly via influences on medication use.⁶⁵ Hence, our results are corroborated by research demonstrating the relevance of family chaos for multiple child health indicators. Interestingly, findings highlighted emotional triggers as a separate pathway linking parent depression and child asthma. Emotional triggers have been shown to mediate the child depression-asthma association,⁸ and our work extends this literature by demonstrating the importance of parental depressive symptoms for such triggers.

Strengths of this study include the focus on a high-risk population recruited and enrolled in a community setting. Additional strengths include the use of data from both children with asthma and their parents and the replication of findings with several measures of asthma control. Limitations must also be noted. The cross-sectional design precludes inference on the direction of effects and causality. Moreover, the magnitude of change in asthma outcomes vis-à-vis levels of family chaos are difficult to assess. Hence, prospective studies are necessary to confirm findings. Our measures of mental health, albeit well validated, did not provide clinical diagnoses, and findings may not generalize to clinical samples. Additionally, correlations between mental health variables may have influenced our analyses. Combining the cACT and ACT for analyses has not been examined in previous work. Because of age limits on the child measures, our sample was restricted. The sample was predominantly Hispanic, limiting generalizability to other populations. Although we focused on youth with uncontrolled asthma, the sample was not severely symptomatic and may better reflect a community sample. Last, family chaos captures only portions of the family environment, and a broader assessment of family functioning may aid in the interpretation of results.

Our results identify psychosocial pathways worthy of exploring in longitudinal and possible qualitative studies to understand the multifactorial etiology of pediatric asthma. Findings highlight the importance of assessing and addressing child and parent depression and family chaos to enhance asthma outcomes in high-risk samples. Effective, well-validated measures for screening youth and parent depression in primary care settings with minimal burden are readily available (eg, the Children’s Depression Inventory 2 [Short Form]³¹ and PHQ-9^28,29) and are increasingly being promoted by clinical guidelines.⁶⁶ Additionally, the CHAOS³³ may be an effective clinical discussion tool for illuminating areas of risk within the household. Integrated behavioral interventions that target a reduction in family chaos via development of family routines, structure, and predictability, particularly around medication,⁶³ can help to optimize asthma care.

This study was funded by the National Institutes of Health and the National Heart, Lung, and Blood Institute (R01HL123797). We thank the staff who collected these data: Gizelle Alvarez, Daisy Cintron, Jazmin Morales, and Genesis Rosales. We also recognize the Asthma Action at Erie Trial Steering Committee who are not authors: Michael Berbaum, Ana Cesan, Hannah Chi, Andrea Fragoso, Denise Guerrero, Melissa Hernandez, Steven Rothschild, and Angkana Roy. The Chicago Asthma Consortium Community Advisory Board provided oversight for the study. We thank everyone at Erie Family Health Centers. This includes medical directors, providers, and staff at the Erie West Town Health Center, Erie Evanston–Skokie Health Center, Erie Helping Hands Health Center, Erie Humboldt Park Health Center, Erie Westside Health Center at Laura S. Ward Elementary School, and Erie Johnson School-Based Health Center. Finally, we thank all the families who made this research possible.

ACT

Asthma Control Test

AFS

Asthma Functional Severity Scale

cACT

Childhood Asthma Control Test

CHAOS

Confusion, Hubbub, and Order Scale

CI

confidence interval

GLM

generalized linear regression model

ICS

inhaled corticosteroid

OR

odds ratio

PHQ-9

9-item Patient Health Questionnaire

PTSD

posttraumatic stress disorder