Weather and household predictors of childhood obesity treatment attendance in low-income urban families

INTRODUCTION

The current standard of care treatment for childhood obesity consists of family-based interventions targeting children’s eating patterns, physical and sedentary activity, and sleep. Family-based interventions can produce meaningful weight loss,^1,2 but their effectiveness is diminished by poor treatment attendance.^3–7 On average, families attend less than 75% of treatment visits, and there is substantial variability across programs and amongst families.^4,5,7

Childhood obesity treatment programs are predominantly available through academic medical centers or universities (rather than community practices), and usually involve frequent clinic appointments. Travel distance,^8,9 caregiver work schedules, and flexible clinic appointment times¹⁰ have been identified as influences on attendance rates in these programs. Attendance is also lower among socioeconomically disadvantaged families,^9,11 which may be attributable to barriers such as limited access to an automobile, or working in occupations with less flexible work schedules. Caregiver characteristics, including ethnicity/race and readiness for change, are weak predictors of attendance.^11,12 Inclement weather (e.g., extreme temperatures, heavy precipitation) may also deter attendance. The notion that treatment attendance suffers during bad weather is clinical lore in some settings, but the relationship between weather and childhood obesity treatment attendance has not yet been quantified.

The objective of this analysis was to identify weather-related and household-level predictors of clinic attendance among low-income families participating in a childhood obesity treatment trial. It was hypothesized that the likelihood of missing a scheduled treatment visit would be higher on days with inclement weather, and higher for families who have longer travel times to the clinic, do not own an automobile, or have limited work schedule flexibility.

METHODS

Description of the parent trial

Data for this analysis originate from an ongoing childhood obesity treatment trial comparing home-based to clinic-based intervention delivery in low-income households (ClinicalTrials.gov: NCT03195790). Families in both arms were offered 18 visits of family-based childhood obesity treatment delivered over 12 months. Eligible participants had at least one child age 6-12 y with a BMI percentile ≥85, and a household income below 200% of the federal poverty ratio. Key exclusion criteria included inability to speak English or Spanish, residing more than 15 miles from the study site in downtown Chicago, IL, and major medical or psychiatric illnesses. The trial’s methodology is detailed elsewhere.¹³ Given our focus on clinic attendance, this analysis only involves data from 98 families randomized to clinic-based treatment. Families were reimbursed for on-site parking or public transportation at each treatment visit. Treatment visits were scheduled as late as 06:00 PM, Monday through Saturday. Due to the impact of COVID-19, we examine attendance only for the 1,036 visits scheduled between September 2017 and March 2020, prior to the onset of the pandemic restrictions in the U.S. The study received ethical approval from the Institutional Review Board of Rush University Medical Center. Adult participants provided written consent, and children provided verbal assent, at the time of enrollment.

Measures

Treatment visit attendance.

All scheduled intervention visits were marked as attended or not attended by a trial interventionist using an electronic data capture form. Consistent with medical center policy for the study site, visits rescheduled within 24 hours of the appointment were counted as missed, whereas visits rescheduled at least 24 hours in advance were not considered missed.

Household variables.

Primary caregivers provided information on their household’s characteristics and resources, including automobile ownership, usual mode of transportation and travel time to the study site, and caregiver employment status and work schedule flexibility.

Weather.

Daily weather data for the study site (Chicago, IL), including precipitation, accumulated snow depth, and minimum, maximum, and average daily temperature, were obtained from the National Oceanic and Atmospheric Administration’s database. “Clement days” (n=149, 14.4%) were defined as having no/trace precipitation, minimum temperature >50 °F, and maximum temperature <85 °F. “Inclement days” (n=171, 16.5%) were defined by any of the following weather conditions considered likely to impact travel via personal automobile or public transportation: >0.75 in precipitation, >1.0 in of snowfall, minimum temperature <15 °F, or maximum temperature >90 °F (see Table 1). All other days were coded as “moderate days”.

Table 1.

Descriptive information and clinic attendance for days on which clinic-based treatment visits were scheduled in Chicago, Illinois, USA, between September 2017 and March 2020. Values are means (standard deviations) unless otherwise noted.

	All days (n=1,036)	Clement days (n=149)	Moderate days (n=716)	Inclement days (n=171)
Precipitation (in)	0.1 (0.3)	0.0 (0.0)	0.1 (0.2)	0.5 (0.7)
Minimum temperature (°F)	42.8 (19.1)	59.7 (4.4)	41.4 (16.4)	33.9 (27.0)
Maximum temperature (°F)	58.8 (21.3)	77.1 (4.4)	57.1 (19.2)	50.2 (28.4)
Average temperature (°F)	50.8 (19.9)	68.4 (3.7)	49.2 (17.5)	42.0 (27.5)
Snowfall (in)	0.1 (0.4)	0.0 (0.0)	0.0 (0.1)	0.5 (1.0)
Accumulated snow depth (in)	0.3 (1.2)	0.0 (0.0)	0.2 (0.8)	1.1 (2.4)
School out of session [n, (%)]	306 (29.5)	77 (51.7)	185 (25.8)	44 (25.7)
Weekend day [n, (%)]	61 (5.9)	8 (5.4)	43 (6.0)	10 (5.9)
Missed clinic visits [n, (%)]	309 (29.8)	60 (40.3)	203 (28.4)	46 (26.9)

Analysis

Mixed effects logit models were used to estimate the odds of a missed clinic visit from household variables alone (model 1), and from household variables and weather (model 2). A three-level categorical variable represented clement, inclement, and moderate weather days. Accumulated snow depth was included as a separate variable, as its impact on travel can last multiple days (independent of other current weather conditions). Both models adjusted for whether the visit occurred on a weekend (vs. weekday), or during a scheduled break in the Chicago Public School System.

RESULTS

Characteristics of participating families are shown in Table 2. Most households owned an automobile (84.7%) and usually traveled by automobile to the clinic (80.6%). Travel times averaged over 30 minutes each way (60 minutes round-trip), with an interquartile range of 15-45 minutes (30-90 minutes round-trip). Of 1,036 visits scheduled, 309 (29.8%) were missed.

Table 2.

Household and child characteristics of families randomized to clinic-based childhood obesity treatment (N=98).

	M (SD)	N (%)
Household characteristics
Income:poverty ratio*	1.1 (0.5)
Household size	4.4 (1.4)
Travel time to medical center (minutes)	31.6 (19.1)
Owns a working automobile		83 (84.7)
Usual mode of transportation to medical center
Automobile		79 (80.6)
Public transportation/other		19 (19.4)
Flexibility in caregiver’s work schedule
None / a little		9 (9.2)
Some		24 (24.5)
A lot		33 (33.7)
Does not work outside the home		32 (32.7)
Child characteristics
Age (years)	9.8 (1.8)
BMI z-score	2.1 (0.5)
Female gender		42 (42.9)
Race/ethnicity
Black		46 (46.9)
Hispanic/Latino		46 (46.9)
Other		6 (6.1)

^*Income:poverty ratio is the ratio of combined household income to the U.S. federal poverty guideline for household size at the time of enrollment.

In model 1, none of the household characteristics examined were related to the odds of missing a scheduled treatment visit, including car ownership (OR=0.81, 95%C.I.: 0.41, 1.60) or travel time (OR=0.99, 95%C.I.: 0.98, 1.01). Attendance did not differ for households that traveled to clinic by public transportation vs. automobile (OR=1.12, 95%C.I.: 0.62, 2.01), or who had more flexible works schedules (χ²=3.72, p=.29). Visits were slightly more likely to be missed when scheduled during a school break (OR=1.36, 95%C.I.: 1.00, 1.84), but not on weekends vs. weekdays (OR=0.97, 95%C.I.: 0.49, 1.95).

The unadjusted probabilities of visits being missed were 40.3% during clement weather, 26.9% during inclement weather, and 28.4% during moderate weather (Table 1). In model 2, which included weather variables in addition to household characteristics and scheduling variables (weekend vs. weekday, school in vs. out of session), the odds of a missed visit was significantly higher on days with clement weather vs. moderate weather (OR=1.70, 95%C.I.: 1.15, 2.51). Attendance did not differ on days with inclement vs. moderate weather (OR=0.93, 95%C.I.: 0.62, 1.40), or vary with accumulated snow depth (OR=1.01, 95%C.I.: 0.90, 1.14). The association between attendance and weather was robust to a variety of changes in the criteria used to define clement and inclement days (results not reported).

DISCUSSION

The main finding from this analysis is that low-income families’ childhood obesity treatment visit attendance was unrelated to household characteristics and resources, but was significantly lower on days with clement weather. These results do not align with prior studies that identified logistical factors such as travel distance and scheduling flexibility as influences on attendance.^8–10,14 Though participating families reported lengthy travel times to the clinic, the fact that most owned and traveled by automobile to treatment visits may have facilitated attendance. Roughly one-third of the sample had limited flexibility in their work schedule, but this too was unrelated to attendance. Overall, attendance in this sample of low-income families was comparable to that in previous childhood obesity treatment trials,^4,5,7 and the specific ways in which they have been able to navigate logistical barriers to treatment visit attendance warrant further study.

The view that inclement weather discourages attendance was not supported. It might be expected that caregivers are less willing to travel significant distances with their children in inclement weather because it is unpleasant and contributes to increased travel time. However, weather has shown inconsistent associations with attendance at healthcare appointments in prior analyses.¹⁴ Favorable weather has been linked to greater emergency department visits, possibly due to a higher rate of injuries related to outdoor recreational activities on days with pleasant weather.¹⁵ Our observation that attendance was lower on clement days suggests that families may prioritize enjoying pleasant weather with their families over coming to a clinic setting for weight management. Perhaps this is understandable, especially in the Midwest where pleasant weather can be a rare occurrence. Given the significant geographic variability in weather across the U.S., it may be worth exploring its influence on treatment attendance in various regions. Potential strategies to increase childhood obesity treatment attendance during nice weather may include pre-emptively scheduling around days when nice weather is forecasted, conducting visits by telehealth, and scheduling visits in the morning so that the family can enjoy the rest of the day outside.

A strength of this study was the availability of detailed information on treatment visit attendance, daily weather, and household characteristics and resources in the parent trial. The composition of the study sample (predominantly low-income, minority, and urban families in the Midwest) was an additional strength. Attendance in this trial has been similar to prior studies,^4,5,7 though treatment attendance in clinic trials (including those that provide transportation/parking reimbursement) may not correspond with attendance in clinical programs. Specific reasons for missed clinic visits were not collected from participants, but could have provided additional insights. This trial (like many others) was partially suspended at the onset of the COVID-19 pandemic (March 2020), and data from this period were not analyzed. We note that all data were collected at a single geographic location, and the influence of weather on attendance may vary based on regional climates. The minimum intervention dosage necessary to achieve meaningful weight loss remains unclear,¹⁶ so it is difficult to determine the degree to which weather-related decreases in attendance affect weight loss outcomes. Once concluded, the parent trial will be positioned to answer this question.

In conclusion, treatment attendance among the low-income families in this trial was comparable to that reported in previous clinic-based childhood obesity interventions, and was unrelated to the presence of logistical barriers. The observation that attendance suffered when the weather was nice may reflect families’ preference to enjoy time outdoors on those days.