A Mixed Methods Analysis of Environmental and Household Chaos: Implications for Childhood Obesity Prevention Through Toddlerhood

Background: Chaos has implications for child health that may extend to childhood obesity. Yet, results from studies describing associations between chaos and childhood obesity are mixed. Challenges to studying chaos-obesity relationships may include inconsistencies in how chaos is operationalized and reliance on caregiver perceptions. Furthermore, multiple pathways may link chaos to obesity, though few have been empirically examined. Methods: We conducted a concurrent mixed methods analysis of quantitative and qualitative data describing home and neighborhood chaos among a diverse cohort of 283 caregiver-toddlers dyads from Ohio. We examined the underlying structure of environmental and household chaos using exploratory factor analysis then sought to validate the structure using qualitative eld notes. We generated total scores for factors of chaos and described their distributions overall and according to cohort characteristics. Additionally, we conducted a thematic content analysis of brief ethnographies to identify potential pathways linking chaos to childhood obesity with the intention to direct future research efforts. Results: Dyads varied according to household composition, income, education, and race/ethnicity. We found evidence for a multi-factor structure for chaos, which included disorganization and neighborhood noise. Household disorganization scores ranged from 8-18 and were on average 11.37 (SD = 2.58). Neighborhood noise scores ranged from 4-12 and were on average 6.93 (SD = 1.89). Both disorganization and neighborhood noise were associated with indicators of socioeconomic disadvantage, such as food insecurity and lower income-to-poverty ratio, though only disorganization was associated with additional social factors within homes, such as caregiver mental health and overall health. Finally, we identied unique themes from brief ethnographies which future contextualize the social and material environments in which chaos was observed, including child behavior and caregiver-child interactions. Conclusions: Chaos is a complex construct composed of multiple factors and the mechanisms linking chaos to childhood obesity may be equally complex. Future studies of chaos-obesity relationships may require greater specicity when operationalizing chaos and empirical study of pathways, like child behavior and caregiver-child interactions, may inform future obesity prevention strategies.

mixed [25,26]. The CHAOS is a parent-reported survey consisting of 15-items which describe various conditions contributing to environmental confusion (e.g., we almost always seem to be rushed or it's a real zoo in our home) [21]. The tool offers a quick, low-cost approach to measuring parents' aggregate experiences within their home. However, the CHAOS may be limited by its reliance on caregiver perceptions, which tend to be subjective and in uenced by factors such as parental coping strategies and personality traits [28]. Additionally, the CHAOS may characterize household and environmental chaos too broadly, inadvertently excluding important subdomains, such as disorganization and instability [22]. One study noted the association between children's exposure to poverty in early-life and academic achievement at kindergarten-age was mediated by chaos, in the form of household disorganization, but not instability [17]. Such ndings suggest chaos as a construct may be more nuanced and aspects of chaos may matter for speci c health outcomes. Therefore, obesity prevention researchers may bene t from revisiting the theoretical structure of chaos and considering more comprehensive measurement tools.

Potential Pathways from Chaos to Obesity
Multiple pathways may connect chaos to child health, but such pathways have not been fully considered in literature examining associations between chaos and childhood obesity risk. For example, the distracting and overstimulating nature of chaotic environments may impede children's developing selfregulatory skills. A recent meta-analysis concluded household chaos and executive functions, like selfregulation, are signi cantly and inversely related [29]. Such relationships are important, as self-regulatory skills, including inhibitory control and emotion regulation, may be protective against childhood obesity development [30].
Parenting practices and parent-child interactions may link chaos to child health [14]; such pathways may extend to child weight outcomes. For example, in chaotic environments, parents may perceive relationships with their children more negatively and exhibit less warmth or enjoyment [31], parents may be less verbally responsive to their children [32], and parents may demonstrate lower levels of responsivity or acceptance towards their children [18]. The quality of parent-child interactions in early-life has been noted as a potentially important determinant of child weight outcomes extending from preschool-age [33] to adolescence [34], but the role of chaos in such associations is poorly understood.
Finally, stress may be another process by which chaos is linked to child weight outcomes. While relationships between stress and childhood obesity are inconsistent, early-life stress is associated with overweight and obesity in adulthood [35]. Chaotic conditions are likely unsettling for young children and exposure to chaos may induce a stress-response that increases children's risk for obesity. Studies examining noise suggest cardiovascular stress indicators and neuroendocrine stress hormones may be sensitive to louder environments [36,37], Additionally, other forms of chaos, such as "emotional chaos", may be associated with diurnal cortisol patterns among young children [23]. Such physiological responses to chaos may be implicated in stress responses that are consequential for child weight outcomes [35], but empirical evidence is limited.

Study Aims and Objectives
We aimed to characterize the home and neighborhood environments of a contemporary cohort of toddlers and explore potential contributors to chaotic environments. Our analyses utilized direct observations of family homes from 283 diverse families. To accomplish our aim, we examined chaos using a concurrent mixed methods research design, as such methods presented opportunities for triangulation, expansion of descriptions of chaos, and may better contextualize the broader social and material environments in which chaos occurs [38]. The objectives for this study were (1) to examine the underlying structure of environmental and household chaos using both quantitative and qualitative data, and (2) to conduct an exploratory analysis of qualitative eldnotes to identify potential pathways that may link chaos to child weight outcomes. We hypothesized that more comprehensive and nuanced assessments of family home environments would provide evidence for a multi-factor structure of chaos, including disorganization, noise, and instability. Additionally, we anticipated that context offered from qualitative eldnotes of family home environments will help inform obesity prevention efforts by underscoring potential pathways linking chaos to childhood obesity risk.

Study Population
Data are from the Play & Grow Study-a prospective cohort study of 299 parent-child dyads from central Ohio. Study design and cohort characteristics for the Play & Grow Study have been previously described [39] but are brie y summarized here. The target population for the Play & Grow Study included 18-month-old children (± 2 months) living in central Ohio. A sampling frame was constructed using patient medical records from Nationwide Children's Hospital (NCH) in Columbus, Ohio. Caregiver-child dyads were enrolled between December 2017 and May 2019. Dyads enrolled included primary caregivers (93% biological mothers) and children who were born singleton with gestational ages between 23 and 42 weeks. Enrolled dyads lived within 15 miles of NCH with no family plans to move beyond that radius during the study and the participating caregiver attested to taking part in the child's meals on a regular basis. Participants were excluded from recruitment if the child had deafness, blindness, food allergies (either child or potential participating caregiver), the child's recorded gestational age > 42 weeks, or if the child was tube-fed or a patient for a clinical feeding disorder.
The Play & Grow Study is ongoing, and we utilized data from the rst and second assessments, which took place when children were approximately 18-and 24-months of age and include caregiver-reported survey data and direct observations of family homes. We limited or analyses to only records with complete data on the variables examined in this study (n = 283). The study was conducted in accordance with the Declaration of Helsinki and the Institutional Review Board of NCH approved study procedures.
Researchers obtained written documentation of informed consent for all subjects.

Data Collection
We utilized Rapid Assessment Procedures (RAP) [40] to simultaneously collect quantitative and qualitative data describing neighborhood and household conditions. RAP make use of traditional anthropological techniques, such as participant observation, interviewing, and analysis of quantitative data, over a shortened and more focused period of eldwork [40]. Typically, RAP are implemented by multidisciplinary teams across multiple sites, include prompt turn arounds on data analyses, and are participatory in nature [40,41]. While RAP may never meet the methodological standards sought by most anthropologists, researchers across disciplines using RAP increasingly recognize the approach's ability to offer meaningful insight in complex social and material settings. For example, RAP have been used in disciplines including health education [42], pandemic response in clinical settings [43], and health information technologies [44]. Our RAP consisted of quantitative audits of neighborhood and household characteristics and participant observation techniques in neighborhoods and family homes.
Audit of Neighborhood and Household Environment. As part of the second wave of data collection (home visits) when children were aged 24-months, teams of trained research staff conducted mixed methods audits of neighborhood and household conditions. We designed a novel data collection tool by adapting existing environmental audits and questionnaires focused on neighborhood and household conditions [21,22,[45][46][47]. A total of 32 items were selected for the audit to describe environmental conditions. Items were organized in relation to neighborhood features, neighborhood disorder, household features, and household disorder.
Fieldnotes of Neighborhood and Household Environments. We supplemented quantitative neighborhood and household audits with a rapid participant observation to describe conditions and interactions research staff observed in participants' neighborhoods and homes. Participant observation is a traditional anthropological technique often used when researchers aim to develop an understanding of participants' lived experiences amidst natural settings [38]. Because our research study lacked the time and resources to conduct extensive eldwork typically associated with participant observation, we adapted key features of the anthropological technique to be implemented over numerous home visits lasting approximately 100 minutes each. During our visits, neighborhoods were observed for approximately 10 minutes, prior to the start of home visits, and homes were observed for the remainder of the scheduled visit (approximately 90 minutes). Staff were permitted to interact with participants in ways that helped build rapport as they implemented other study protocols but prioritized acting as an objective observer. Staff were trained to write descriptive notes to illustrate the physical and social environments they observed and practice critical self-re exivity by writing eldnotes re ecting on their experiences. Following visits, study staff returned to research o ces where they logged their neighborhood and household audit and wrote a brief ethnography using their recorded eldnotes.

Research Staff Training and Reliability
Prior to data collection, research staff received a half-day training involving a two-hour classroom session (discussing skills and techniques of ethnography [48]) and a two-hour eld practice component. A second classroom-based review session was additionally conducted once data collection was underway. Photos of varying neighborhood and household conditions were rated and discussed. Detailed descriptions of each rating were provided. Based on group consensus, de nitions for ratings and descriptions were recorded and organized in a manual for reference and future trainings. During the eld component, trainees traveled to the home of a research team member where each trainee completed and discussed the observation form. Trainees, who consisted mostly of college-educated, white, middle-class females under age 40 years, were required to demonstrate adequate inter-rater reliability (≥ 80%) from a minimum of ve observations before they were certi ed to collect data.

Data Analyses
Analyses followed a mixed method design. Quantitative data from neighborhood and household audits were rst analyzed to describe levels of chaos present in households during home visits. We then sought to validate and contextualize ratings of chaos using the brief ethnographies.
Quantitative Analysis. To describe levels of environmental and household chaos, we selected 21 items that were most relevant to environmental and household chaos from the audit of neighborhood and household conditions and respondent surveys ( Table 1). Items from caregiver surveys were included to supplement measurements of household instability (often characterized by changes in parental romantic relationship status, household moves, changes in income or parental employment, and disruption to family routines) [15], as such indicators are not possible to observe during a 100-minute home visit. We reviewed the distributions of responses for the initial 21 items and chose to exclude two due to little variability in item responses. Thus, we sought to empirically derive measures of environmental and household chaos from a total of 19 items (Table 1). We developed scales describing chaos using exploratory factor analyses (EFA) [49] with unweighted least squares and oblique (Promax) rotation methods. All items considered for the EFA were ordinal or binary. Therefore, our factor analysis was based on polychoric correlations, rather than Pearson's correlations [50]. We chose to employ unweighted least squares for ordinal indicators, because it has been shown to be robust to smaller sample sizes, skewed data, and provides greater accuracy and less variability in estimates, when compared to diagonally weighted least squares [51].
Factor extraction was informed by a scree plot [52] and our theoretical understanding of chaos. Our nal factor structure required factors to have a minimum of three items, as fewer than three items generally results in weak factors [52]. Following previously published work, we assigned an item to a factor if the primary loading was ≥ 0.50 [51] and the item did not cross-load (loading was < 0.32 for other factors) [52]. Finally, we generated factor scores by summing the items assigned to each factor.
Variables to describe cohort characteristics were predominantly derived from the caregiver survey administered at the baseline assessment (when children were approximately 18-months old). Descriptive statistics (means, standard deviations (SD), and P values from one-way ANOVA) described how measures of chaos distribute across characteristics of the sample, including child, household, and caregiver characteristics. Quantitative analyses, including the EFA, were conducted using SAS (version 9.4, SAS Institute, Cary, NC).
Qualitative Analysis. Due to the large number of households visited by researchers during the 24-month assessment, we chose to examine and compare qualitative records from a randomly selected subset of families. To do this, we categorized factor scores into quartiles and randomly selected records from the highest quartile and records from the lowest quartile of factor scores. A thematic content analysis [53] was conducted using the brief ethnographies to describe participants' immediate neighborhood and home environments. Informed by our theoretical interest in chaos, we used a deductive approach to develop codes, though an initial round of open coding was completed to assess patterns in the data and codes missing from our a priori coding structure [54]. A nal codebook was constructed with code de nitions to ensure consistency across coding and coding was completed by one researcher. We coded records until thematic saturation [38] was achieved and codes were managed electronically using QSR NVivo (Version 12, QSR International, Victoria, Australia). In total, 87 records were coded for our thematic content analysis.

Quantitative Findings
Of the 299 caregiver-child dyads in the cohort, eld observations were completed at 283 family homes. Compared to families with complete observation data, the proportion of families missing eld observations was higher among those who had the lowest household income, lowest educational attainment, were food insecure, and whose primary respondent identi ed as non-Hispanic Black (data not shown).
The scree plot suggested two to four factors would optimally t our data. We examined the factor loadings for each of the three structures and determined a two-factor solution was best. Eight items were assigned to the rst factor. Factor one was labeled household disorganization and included items describing interior household conditions and household dynamics, such as interior noise, clutter, commotion, overcrowding with furniture, excessive telephone use, communication between household members, and overall preparedness for the study visit. Our second factor, labeled neighborhood noise, consisted of four items describing the types and amount of noise heard outside participants' homes ( Table 2). Scores were generated by summing the items assigned to each factor and internal consistency estimates were calculated (Cronbach's Alpha = 0.73 and 0.67, respectively). Observed scores for household disorganization ranged from 8-18 and the mean score was 11.37 (SD = 2.58). Observed scores for neighborhood noise ranged from 4-12 and the mean score was 6.93 (SD = 1.89). Characteristics of the cohort were provided in Table 3.  Note: Description of disorganization and neighborhood noise were derived from assessments in family homes when children were approximately 24-months of age. Higher scores are indicative of higher levels of chaos. SD = standard deviation; Household disorganization scores summarizes ratings assigned to eight items: (1) interior noise rating, (2) cluttered interior (y/n), (3) commotion (y/n), (4) interruptions (y/n), (5) preparedness rating, (6) loud speaking (y/n), (7) crowded with furniture (y/n), (8) excessive telephone use (y/n). Neighborhood noise scores summarized ratings from four items: (1) exterior noises heard from inside the home (yes/no), (2) the volume of exterior noises heard from inside the home, (3) the rating of noise heard while outside the home, and noise pollution (y/n). Food insecurity was assessed using the U.S. Department of Agriculture Food and Nutrition Service's guidelines for measuring food security. We coded families as having food insecurity if they indicated experiencing any level food insecurity in the 12-months prior to them completing our surveys. Income-to-poverty ratio was calculated using the 2018 U.S. poverty guidelines according to household size. Depression symptomology was determined using the Center for Epidemiological Studies Depression Survey, with a score ≥ 16 indicated symptoms of depression.

Qualitative Findings
Household Disorganization. Concurrent with our quantitative ratings of disorganization, family homes with the highest levels of disorganization were described by researchers as environments where household members often spoke loudly or over one another or households were ill-prepared for the study visit (e.g. families were late for the visit, did not follow instructions for visit preparation, or it was apparent participating caregivers did not communicate with household members that the study visit had been scheduled to occur). Furthermore, home interiors were often cluttered, and households were tumultuous, which was often attributed to crowding (i.e., more people that the space appeared to accommodate) and heavy foot tra c in and out of the home. Example passages from ethnographies include, "The rst thing I noticed upon walking in the front door was…how cluttered everything was… There were boxes of diapers, laundry, miscellaneous papers… crammed against walls and made it very di cult to get around… There were several sofas crammed into the small space, one of which was over owing with laundry and blankets. We had to step around several kid's toys scattered across the living room oor. The kitchen… counter surfaces were completely covered in dishes, food, papers, laundry…" Brief ethnographies also revealed themes that helped contextualize the social and material environments observed in both disorganized and organized homes (Table 4). In nearly all ethnographies from homes with high disorganization, staff wrote about children's behaviors or energy levels. In many cases, researchers described children as being "very active" or having "a lot of energy". Sometimes study staff noted children "running around the home" or climbing on study materials or furniture. Some children were described as "crying/screaming out of excitement…" or "demand[ing] a lot of attention from [study] staff and [caregivers]". Such descriptions of child behavior were often coupled with staff observations of situations when caregivers attempted to bring order to the home environment unsuccessfully using approaches, such as speaking at elevated volumes or yelling across rooms to get a child's attention. In contrast, descriptions of child behavior were mostly absent from ethnographies from homes with the lowest disorganization. However, among the most organized homes, researchers portrayed moments when caregivers were observed using speci c behaviors or strategies to successfully mitigate potential disorganization within the home. This was often noted when caregivers were described as "encouraging" and "supportive" of their children as they worked through study tasks or when they responded to energetic or uncooperative children in ways that successfully calmed the child without introducing additional sources of disorganization. In one household, a mother was described as asking her excited daughter to "pause, take a deep breath, and then speak…" so she could better understand her needs.
In homes with high disorganization, staff frequently described caregivers as passive towards other household members or unengaged in visit activities. For example, one caregiver was described as "mostly passive" when she did not intervene when one child in the household was "throwing things or picking up and swinging around… three young [household] pets". Another caregiver was described as only directly engaged with the participating child when "[complying] with the visit activities," but not during other points of the study visit, such as transitions between activities or during preparation for mealtime. Furthermore, ethnographies from the most disorganized homes illustrated strained interactions between caregivers and children. Speci cally, caregivers visibly, or audibly, expressing frustration towards their child's actions (e.g., "mom shout[ed] 'why?!' several times… when [child] was crying and refusing to be measured on the stadiometer"). Other study staff wrote that caregivers "seemed to be stressed" or were "visibly frustrated" when their child struggled during the visit. These descriptions from highly disorganized homes directly contrasted with descriptions of interactions between parents and children in organized homes. In the most organized homes, staff described interactions between caregivers and children using language that was mostly positive in sentiment. For instance, one dyad seemed to enjoy shared activities as the "child giggled, or mom laughed in response to something… the child was doing".
Another family was observed "communicat[ing] a lot with one another and happily discuss[ing] their many shared interests." One dad made comments to research staff about "how [he and the mother] were proud of the kids". Research staff suggested that such interactions made homes feel "calm and relaxing".
Competing caregiver responsibilities were noted in households described as highly disorganized. In such cases, caregivers shared with study staff they were rushing to get to or come from places like work, school, doctor's visits, or were balancing other family priorities, such as medical challenges. For example, one mother shared that she "scheduled a far-away appointment 90 minutes after the set visit start time "Though no cars drove through the lot during our observation, there was a constant stream of tra c down that main road and accompanying sounds of tra c. Police and ambulance sirens persisted for a few minutes and were intermittently present throughout the entire observation." Sources of neighborhood noise were similar across households, despite the ratings of neighborhood noise assigned by researchers. Sources included car tra c, such as distant highway noise or engines, sirens (e.g., ambulance or police sirens) and alarms (e.g., beeping from construction vehicles), air tra c, speech from people in the neighborhood, and other noises, such as dogs barking. However, among homes described as having the lowest neighborhood noise, researchers more frequently described noise as being noticeable to a lesser degree. They did this by conditioning their descriptions using words like "faint" or "mu ed" or described noise as being "sporadic", rather than persistent.
We noted two themes that were unique to households in the loudest neighborhoods (assigned to the highest quartile of neighborhood noise) (Table 4). First, in especially noisy neighborhoods, researchers more frequently described being able to hear loud ambient noises from the surrounding neighborhood while inside the participant's home. These descriptions often included noises such as vehicles driving by, airplanes, and sirens from emergency response vehicles. In fewer instances staff described hearing neighbors talking or yelling outside. The second theme, unique to the noisiest neighborhoods, was the presence of loud music from passing vehicles. Researchers described neighborhood environments where drivers' music was so loud, they could "feel the vibrations" from the music being played. Loud music from cars was observed when researchers were outside homes and inside homes.
We also noted two themes that were unique to the quietest neighborhoods (assigned to the lowest quartile of neighborhood noise). Researchers described neighborhood settings where geography appeared to play a role in attenuating environmental noise. Examples of geographic features include trees surrounding the neighborhood which buffered against noise, distance from the metropolitan airport, or a large parking lot or dead-end street separating the participant's home from major sources of tra c noise. Finally, in homes that were describe as being the quietest, research staff described an overall lack of loud ambient noise. In such cases, researchers often noted the inability to hear exterior noise while inside the home. Child Behavior I endorsed commotion in the home because the [target child] had a LOT of energy and was running around the home from one end to the other, grabbing toys, riding his truck, and crying/screaming out of excitement/motioning for staff or parent to do things for him.

Passive or Unengaged Caregivers
Mom and children were still in pajamas and there was clutter on the oor making it di cult to navigate our things inside... Children's faces were dirty, and they went back and forth from living room to kitchen and were occasionally shouting. Mom spent most of our visit looking at her phone and playing with the youngest baby. She called the oldest over once or twice, but for the most part, he was interested in and getting into our materials. Mom did not pay much attention to what was going on in her home. She provided minimal engagement with the children and even less with us.

Strained Parent-Child Activities
Even when the non-preferred activities stopped, child continued to want to run around or sit in the puzzle box and resisted when we tried to direct his attention elsewhere. Mom became even more visibly frustrated during the meal but would quickly get over it and laugh it off at times. The more mom reacted to the child's outbursts, the more chaotic the environment felt.

Strategies to Create Order
For book, child was very interested in the measurement tools still and our suitcase in the kitchen. He became very upset when Mom would not let him come to the kitchen. She became louder and louder in her requests to sit and look at the book. She sent Dad to the kitchen for a piece of candy in hopes that child would calm down. He did not. Eventually she gave in and let him come to the kitchen to see. She brought him back but was not able to get him to focus on the book much. The puzzles were a similar scenario. Mom got louder and louder to try to keep him engaged in the activity, but he still ran to the kitchen several times and she would yell for him to come back.

Competing Caregiver Responsibilities
Dad was very di cult to get a hold of due to his crazy work hours. Mom tried to reschedule and cancel the visit on his behalf multiple times. After some prompting and a few Saturday calls, dad con rmed a date and did the survey before the visit.

Low Disorganization
Factor Positive Parent-Child Interactions The home feels extremely inviting and a nurturing place for young children to grow. Mom listened to child throughout their meal and asked child questions as if they were engaged in a discussion. Mom genuinely laughed with child and genuinely seems to enjoy her time with child-and child with mom. Mom calmly redirected behaviors when child behaved less favorably. All interactions were positive and in a respectable tone.

Summary of Findings
With the goal to inform future childhood obesity prevention strategies, we examined detailed descriptions of environmental and household chaos using a concurrent mixed methods approach. Our analyses reexamined the underlying structure of chaos using EFA and found evidence for a two-factor solution consisting of disorganization and neighborhood noise, but not instability. Moreover, qualitative eldnotes describing family homes and neighborhoods supported factors identi ed in the EFA. Finally, we sought to describe the context in which observed chaos occurred, and in doing so, identi ed codes describing social and material conditions that varied according to the level and type of observed chaos.
Our multi-factor structure of chaos aligns with methodological approaches and results described by Vernon-Feagans et al (2012). In their analysis, ten indicators of chaos were collected via ve direct observations of participant homes over children's rst three years of life; factor analyses identi ed two factors: disorganization and instability [22]. Our EFA did not identify instability as an independent factor contributing to chaos. However, methodological differences in study designs may explain such discrepancies. While the indicators selected to represent instability in our study closely aligned with those previously assessed,(e.g., residential moves) [22], we conducted a single-assessment when children were approximately 24-months of age, rather than multiple assessments over time. Unlike other aspects of chaos, which tend to occur regularly, instability often occurs periodically. Therefore, our lack of support for instability as a factor of chaos may be due to our limited assessment timeline. Additionally, the sample examined by Vernon-Feagan et al (2012) was drawn from a target population comprised of families living in low-income rural regions of the U.S. [55], which differs from families included in the Play & Grow cohort. Such differences, especially regarding socioeconomic position, may contribute to instability being unidenti able in our EFA. Still, our ndings build upon previous work by suggesting chaos may be comprised of important subdomains. Thus, assessments of environmental and household chaos may require greater nuance when discussed as a potential correlate of childhood obesity.
Our scale describing disorganization closely aligns with what Matheny and colleagues labeled environmental confusion, in the development of the CHAOS [21]. This suggests the CHAOS may provide a foundation for developing measurement tools designed for structured, direct observations of disorganization in family homes. We believe direct observations may be necessary to avoid potential bias often associated with parent-reported measures [56]. For example, one study examining parent and adolescent perceptions of household chaos using the CHAOS (N = 261 parent-adolescent dyads) found perceptions of chaos in shared home environments were only moderately correlated (r = 0.32), implying individual differences in perceptions of chaos [57]. Another analysis examined associations between maternal personality characteristics and perceptions of chaos using the CHAOS (N = 94). Results indicated mothers with high stimulus sensitivity perceived home environments as more chaotic than what was objectively measured by trained observers [28]. While parent-reported measures offer quick, cost-effective alternatives to direct observations, disentangling caregiver characteristics from measures of chaos may be impossible without more objective assessments. Still, direct observations conducted by trained researchers also have notable shortcomings, including vulnerability to bias resulting from observers' personality, knowledge, beliefs, and experiences. We were mindful of this limitation when designing our data collection procedures. To mitigate potential bias in our direct observations, staff were trained to collect both descriptive and re ecting eldnotes which facilitated staff engagement in re exivity as they assigned ratings. Never-the-less, individual biases may have played a role in our observations.
We identi ed neighborhood noise as a unique factor contributing to our ratings of chaos. In early research investigating associations between chaos and child development, noise was the primary aspect of studied [14], but current de nitions of chaos include little speci city around types and sources of noise [14,21]. Interestingly, one item describing the level of interior noise was highly correlated with our factor of disorganization, but minimally correlated with our factor of neighborhood noise. Such distinction may suggest noise typologies are a necessary level of nuance for measuring environmental and household chaos, with different implications for intervention development. For example, our quantitative analysis found neighborhood noise was associated with indicators of socioeconomic disadvantage, such as not living in a single-family home and income-to-poverty ratio but was not associated with characteristics more closely linked to social environments within homes (e.g., number of household members or caregiver health). Therefore, neighborhood noise may be one aspect of chaos more closely tied to structural disadvantage and may require multifaceted interventions designed to address a variety of upstream social inequalities.

Implications for Childhood Obesity Prevention
Our content analysis of qualitative eldnotes contextualized disorganization and neighborhood noise within family environments and alluded to possible pathways by which chaos may in uence obesity [58].
For example, while our staff were not trained in rigorous protocols for coding children's behaviors [59], nearly every qualitative assessment of highly disorganized homes included staff descriptions of children's behaviors. Parent-and teacher-reported characteristics of child behavior, such as being impulsive, playing carelessly or recklessly, and or becoming out of control relative to playmates, are included in validated measures of children's self-regulation [58]. In toddlerhood, self-regulation develops rapidly [60] and poor self-regulation in early life may in uence short-term obesity development [30], and obesity in adulthood [61]. To the best of our knowledge, only one study has examined the joint effect of chaos and self-regulation on child weight outcomes [26]. In this study, 132 parent-child dyads from a Head Start program in Michigan provided caregiver-reported perceptions of household chaos and children's self-regulation was evaluated according to their performance in the snack delay task when children were approximately 24-months Results indicated a three-way interaction of household chaos, child self-regulation, and child sex on BMI z-score when children were approximately three years old.
Speci cally, for boys with moderate to low self-regulation, BMI z-score was higher when exposed to greater levels of household chaos; no association was found among girls [26]. The association between self-regulation and childhood obesity has been shown to differ among boys and girls [62] and selfregulation may be a promising target for childhood obesity prevention efforts [30]. However, our observed differences in child behavior-related eldnotes between organized and disorganized homes builds upon emerging evidence suggesting chaos may act as a moderator for obesity outcomes relative to children's self-regulation. Therefore, chaos should be considered in future interventions designed to improve child self-regulation.
Fieldnotes describing interactions between caregivers and children from our rapid participant observations diverged between households with the highest and lowest ratings of disorganization.
Coding and characterizing aspects of parent-child interactions requires years of training and expertise not acquired by our research staff. Therefore, our descriptions of caregiver-child interactions do not replace more rigorous, objective protocols employed by developmental and behavioral scientists [63]. However, the consistency and contrary nature of language used by staff to describe caregiver-child interactions according to levels of disorganization suggests caregiver-child interactions may be important context for studies of chaos-obesity relationships. Moderated parent-child interactions and relationship quality have been linked to childhood obesity development. For example, in a nationally representative sample of U.S. children, toddlers demonstrating low security of attachment to their primary caregiver also had greater BMI at preschool age [33]. Other studies examining parent-child food-related interactions suggest interactions characterized by low responsivity may also increase children's risk for developing obesity by promoting obesogenic eating behaviors [64]. Improving parent-child relationship quality and interactions may be a promising childhood obesity prevention strategy [30]. However, previous literature underscores the deleterious effects of chaos on parent-child relationships [14,18,31,32]. Therefore, obesity prevention strategies designed to promote high quality parent-child relationships and food-related interactions may also require intervention components which address chaos. To support such intervention development, future research is needed to examine whether parent-child dyadic pathways mediate or moderate relationships between chaos and childhood obesity risk.
Our study has limitations that must be considered. Observations of chaos were conducted during a single visit in participant's homes. As some aspects of chaos may be acute while others are chronic, we may not have observed the true variation of environmental and household chaos. Furthermore, the presence of study staff and execution of study protocols during the visit may have contributed to an unusual home environment that factored into our staff's ratings. Future studies incorporating objective measure of chaos should strive for repeated assessments to ensure what is measured is "typical" for households.
Our data collection tool for measuring household chaos was novel and data collection relied on multiple research staff members who were predominantly non-Hispanic white, middle-class, women. Though protocols and trainings attempted to overcome systematic error, without more rigorous testing of psychometric properties, construct validity of our tool may be limited, and observations may incorporate individual bias from observers. Furthermore, most observers interacted and built rapport with families at previous assessments. It is known whether these previous interactions in uence ratings and eldnotes.
Finally, though we included two items on household routines in our quantitative assessment, household routines were largely neglected from our observations of chaos. Family routines may be key aspects of chaos with important implications for childhood obesity [65,66]. Future research should combine factors, such as disorganization and noise, with measure of family routines to understand how best to operationalize chaos.

Conclusions
Chaos represents a complex, multifaceted risk factor with research implications spanning various disciplines [67], including public health research focused on early childhood obesity prevention.
Unfortunately, empirical evidence examining chaos-obesity relationships in childhood is limited by heterogenous de nitions and subjective parent-reported instruments. As obesity prevention researchers look to family home environments as preferred settings for prevention efforts [68], more contemporary measures, such as those relying on direct observations which account for multiple underlying factors of chaos, may yield valuable insight on factors contributing to a global public health issue. We demonstrated methods for improving measures of chaos and underscore potential mechanisms linking chaos to early childhood obesity, such as child self-regulation and caregiver-child interactions, which deserve greater attention in the literature.

Availability of Data and Materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics Approval and Consent to Participate
The study was approved by the Institutional Review Board at Nationwide Children's Hospital (IRB16-00826) and participants provided written informed consent.

Consent for Publication
Not Applicable