As part of a community-engaged research partnership comprising investigators from the University of Washington (UW) Pacific Northwest Agricultural Safety and Health Center, the Yakima Valley Farm Workers Clinic (YVFWC), and the Northwest Communities Education Center/Radio KDNA, we conducted a randomized trial from July 2015 to February 2019 to assess the effectiveness HEPA air cleaners to improve the health of children with asthma. The study was conducted in the Lower Yakima Valley, a region largely comprised of immigrant Latino farm workers and their families who participate in the area’s intensive agricultural production. Pediatric asthma is a longstanding health concern in this community. The study protocol and methods are described in detail elsewhere [18].
Recruitment and screening
From July 2015 to November 2017, participants were recruited from children referred to the YVFWC Asthma Education Program, an established CHW delivered program. YVFWC Asthma Education Program CHWs were trained and supervised by UW research staff to conduct screening and data collection for the HAPI Study. Children were eligible to participate if they met the following criteria: 1) aged 6–12 years; 2) live within 800 m of dairy or crop production; and 3) have poorly controlled asthma. The latter was defined as four or more days with asthma symptoms in the past 2 weeks, use of asthma rescue medications for four or more days in the past 2 weeks, hospitalization or emergency department visit for asthma in the past year, or unscheduled clinic visit due to an asthma attack in the past year. Potential participants were excluded if they lived with someone who smoked, had more than one primary residence, planned to leave Yakima for more than 2 months during the study intervention period, or planned to move within 6 months.
Study enrollment and randomization
Of 170 children who were screened, 96 (56%) were eligible (Fig. 1). Seventy-nine children (46%) followed up with an enrollment visit in clinic that included caregiver consent and child assent for study participation as well as the first session of the CHW standard three session asthma education that is available through the YVFWC for this community. The study baseline assessment in the participant’s homes was conducted on average 7-8 weeks later at which time participants were randomized for the trial. Participants were randomized to the intervention (HEPA air cleaner in addition to continuation of the standard asthma health education) or to the control group (continuation of the standard asthma health education alone). Participants were randomly assigned to the intervention group or the control group using sealed envelopes containing group assignment. Envelopes were prepared at the UW and unsealed by a UW study staff member just prior to the baseline assessment.
The UW Human Subjects Division and the YVFWC Research Review Committee approved the study. Participants who agreed to share clinical record data signed Health Insurance Portability and Accountability Act forms.
Intervention
The intervention group received two portable Austin Air Pet Machine 410 HEPA Air Cleaners (Austin Air Systems Ltd., Buffalo, NY). One portable HEPA air cleaner was placed in the child’s sleeping area and a second air cleaner was placed in the home living room. Caregivers were instructed to keep the air cleaner running at the highest tolerable fan level at all times over the study year of follow-up and, if possible, to close the bedroom door. An Onset® HOBO UX90–004 Motor On/Off Data Logger (Onset Computer Corporation, Bourne, MA) that was designed to record the date/time with motor on or off events was placed in the sleeping area HEPA units and in half of the living HEPA units. Use of the air cleaners was also quantified subjectively via caregiver survey conducted at the mid-study and final study visit that recorded the reported use behavior regarding fan speed level used and the number of days each air cleaner was turned off in the prior 30 days. Intervention participants were allowed to keep the air cleaners after study completion. Control group participants were offered an air cleaner after completion of final data collection.
The three asthma education visits conducted by the YVFWC CHWs addressed proper medication technique, medication adherence, asthma trigger identification, and behavioral guidance on trigger control. As noted above, the first asthma education session occurred in the clinic at study enrollment. Importantly, this included review of the participants prescribed asthma medication management plan and steps to ensure access (e.g. refills, replacement) were taken if needed. The second asthma education session occurred during the home visit for the study’s baseline assessment. The last asthma education session was conducted in the home approximately 3 months after enrollment, independent of the study assessments. All participants also received dust mite covers for the child’s pillow and mattress, a cleaning kit with floor and surface cleaners free of irritant or toxic chemicals, a medication storage container, and a peak flow meter.
Outcomes
Asthma-related health metrics were collected over the course of the participants’ year in the study collected during enrollment and trial data collection at baseline, mid-study, and end of study home visits. Metrics included clinically oriented measures related to asthma management (Asthma Control Test [ACT] score, days with asthma symptoms, urgent or unscheduled clinical utilization for asthma concerns, prescription for a course of oral steroids, and lung function. Biomarkers associated with asthma exacerbation, including fractional exhaled nitric oxide (FeNO) and urinary leukotriene E4 concentration (uLTE4) were also assessed.
We selected ACT scores, recent symptom days, unplanned clinical utilization, and uLTE4 (ng/mg creatinine or ng/mg) as primary outcomes. A power calculation estimated a repeat measure design would provide power to observe a 2-point change in the ACT score. We estimated to have adequate power to observe a two-point change in the raw ACT score. The ACT score is a validated composite metric to capture longer term asthma control (four-week referent period), the goal of asthma treatment. This is the one composite measure of asthma that has been identified as a recommended core metric for assessment of asthma control in NIH-initiated prospective trials in children less than age 12 years [19]. Symptom days in the past 2 weeks is a commonly assessed outcome in asthma intervention research and represents participant perception of more recent asthma-related disruption. Clinical utilization is an important measure for understanding individual and public health impacts of asthma morbidity and is arguably less subjective than caregiver or child report of asthma related symptoms and disruptions. Lastly, uLTE4 (ng/mg) was selected as a primary objective measure, because preliminary data in our research group had previously identified suggestive relationships between ambient sources of air pollution in this community and uLTE4 in children with asthma [20].
The ACT provides a composite score of disease activity based on a set of questions regarding symptoms during the most recent 4 weeks. This is a validated instrument available in English and Spanish [19]. Two versions of the ACT Questionnaire were used as appropriate for age: the five-item version for persons aged 12 or older and the seven-item Childhood ACT (C-ACT) version for children aged 4–11. Both questionnaires use 5-point and 6-point Likert scales with higher values indicating better asthma control with a highest possible score of 25 for the ACT and 27 for the C-ACT.
Caregivers were asked to report the total number of days with any asthma symptoms and to report the number of days with specific symptoms (wheezing, woken at night from asthma symptoms, stopped playing due to asthma symptoms, and days of missed school due to asthma symptoms) during the most recent 2 weeks. They were also asked to report number of unplanned or unscheduled visits to a clinic or emergency department for asthma symptoms and treatment.
For uLTE4 assessment, spot urine specimens were collected during each study visit. Aliquoted samples were stored at − 20 °C prior to submitting the urine samples to the National Institute of Health’s Children’s Health Exposure Analysis Resource for quantification of urinary creatinine using a G-EQUAS (http://www.g-equas.de/) proficiency test validated colorimetric method based on Taussky [21] and uLTE4 using an enzyme-linked immunosorbent assay (No. 501060; Cayman Chemical, Ann Arbor, Michigan). Creatinine-adjusted uLTE4 (ng/mg) was also calculated, using uLTE4 (ng/mL) divided by creatinine (mg/mL), for comparison with other studies. ULTE4 is an emerging biomarker in pediatric asthma research correlated with airway inflammation and asthma exacerbation [22].
Oral steroid prescriptions were identified by retrospective medical records review. The number of short-term steroid prescriptions was calculated using the total number of these prescriptions ordered at YVFWC clinics over a participant’s entire period of HAPI enrollment. Exhaled nitric oxide is an FDA approved approach for monitoring airway inflammation among individuals with asthma. Lung function was assessed with the EasyOne spirometer (NDD Technologies, Andover, MA) according to American Thoracic Society guidelines. Participant flow volume loops and volume time curves were reviewed and tests with evidence of cough or poor performance were not used. Participant sessions with an average of at least two exhalations with a forced exhaled volume (FEV) capacity difference of 0.15 L or less were used for analysis. FeNO was measured prior to spirometry via a portable NIOX VERO® device (Aerocrine Inc., Stockholm, Sweden and Aerocrine Corp.).
Covariates
Socio-demographic, health, and psycho-social characteristics assessed by self-report on study surveys included: age, sex (male, female), Hispanic/Latino ethnicity, born in U. S, annual total household income (< $14,999, $15,000 – $29,999, $30,000 – $60,000, > $60,000), public health insurance, time spent in home in past 24 h (hours), home location (rural, on a farm; rural, not on a farm; in town), home within 800 m of pollution sources (farm raising animals, farm growing crops, roads with heavy traffic, unpaved dusty roads), and use of asthma controller medication (inhaled corticosteroid and/or leukotriene antagonist).
We assessed asthma-related stress using the caretaker’s response (strongly disagree, disagree, not sure, agree, strongly agree) to the following statement: My child’s asthma has caused stress in my family. A caretaker was categorized as having asthma-related stress if they answered “agree” or “strongly agree” to the statement. Child atopy was assessed at enrollment via skin prick test on the forearm using six common aeroallergens (cat, dog, mouse, mixed mold, dust mite, and cockroach). Participants were categorized as atopic if they had a wheal size equal to or greater than three millimeters above the negative control at 15 min after the test, a positive response to histamine, and a negative response to negative control.
At the enrollment visit research staff collected weight in kilograms using a Health-O-Meter scale (model 599KL) and collected height in centimeters using a single measurement with a wall-mounted stadiometer. Body mass index (BMI) was calculated as kilograms of weight divided by height in meters squared. Participants were categorized as overweight or obese if their BMI was at or above the 85th percentile for their age and sex [23, 24]. Height data after enrollment were attained through medical records review of all YVFWC visits during the HAPI enrollment period. We excluded YVFWC visit heights that were less than the HAPI enrollment height value and heights that were more than 25 percentiles different than the enrollment height-for-age percentile. We used each participant’s average age-for-height percentile from all heights after exclusions to estimate heights at the baseline, mid-study, and final study visits.
Descriptive analysis
To examine the effectiveness of randomization, we assessed the comparability of demographic, psycho-social, and health-related characteristics of intervention and control group participants at randomization (study baseline) using chi-square test, analysis of variance, and Student’s t-test. Because all participants received an initial asthma health education session at enrollment, which addressed asthma symptom management and access to medications as prescribed, we also reviewed the health outcomes values at enrollment in addition to the trial baseline visit along with statistical comparison of the groups’ changes from enrollment to baseline.
Health outcomes were characterized using continuous and dichotomous values as follows. For the C-ACT and ACT questionnaires, participants were categorized as having poorly controlled asthma if they had a score of 19 or lower [25]. To evaluate absolute score differences from participants with adult (age > 12 year) and child versions of the ACT together, a standardized ACT score was created by dividing the participant’s total score by the total possible score to calculate a score percent. As noted above, the study was designed to detect a 2-point change in the raw ACT score. Of note, a 7.4-point change in the standardized ACT score is equivalent to a 2-point change in the raw score for the child version of the ACT and an 8-point change in the standardized ACT score is equivalent to a 2-point change in the raw score for the adult version of the ACT. Participants were categorized as having recent asthma symptoms if they answered “yes” to any of the asthma symptoms categories. Creatinine-adjusted uLTE4 was defined as high if it was at or above the baseline median value of 1.35 ng/mg. FeNO concentration was categorized as elevated if the FeNO value indicated an intermediate or high level of pulmonary inflammation according to American Thoracic Society guidelines [26]. For children less than 12 years old, FeNO was categorized as low (< 20 ppb), intermediate (20–35 ppb), or high (> 35 ppb) and older children were categorized as low (< 25 ppb), intermediate (25–50 ppb), or high (> 50 ppb). All spirometry results were interpreted using the multi-ethnic Global Lung Initiative reference eqs. [27]. FEV1 divided by forced vital capacity (FEV1/FVC) and FEV1 percent predicted value were calculated using age and estimated height on the date of spirometry. FEV1 percent predicted value was categorized as low if it was less than 80% of the predicted value as well as whether it was less than the lower limit of normal (LLN). LLN was defined as a z-score less than negative 1.645. FEV1/FVC was categorized as below normal if it was less than 0.80. Below normal forced expiratory flow (FEF) 25–75 was categorized as less than 60% of predicted value as well as whether it was below the lower limit of normal (LLN).
Statistical analysis
For all outcomes with repeated measures (ACT, symptom days in prior 2 weeks, uLTE4, FeNO, and spirometry), we examined the difference between control and intervention participants for changes in these outcomes averaged over the mid-study and final visits using generalizing estimating equations (GEE) in linear models which account for within-subject correlation [28]. This GEE approach analyzing continuous outcome variables was considered the primary analytical approach. The study was designed to capture the mean outcome improvement over a year with repeat measures. Binomial, Gaussian, and Poisson versions of GEEs were used as appropriate for bivariate, continuous, and count variables. In secondary analysis GEE models examined dichotomous outcome variables for ACT using a clinically defined cutoff score, FeNO, symptom days, uLTE4, and spirometry as described above in the Descriptive Analysis section. GEE models included an interaction term for intervention with visit. Total number of unscheduled clinical utilization visits and steroid prescriptions over the study year were analyzed in linear regression models with robust standard errors. All models were adjusted for the baseline value of the outcome variable and adjusted for a priori determined covariates that are known to be strongly associated with pediatric asthma outcomes including sex, age, controller medication use, and season [29].
We also sought to characterize the effect of the intervention on achieving the overarching long-term goal of optimal asthma care, which is to achieve optimal asthma control, prevent symptoms, and enable patients to live without functional limitations or risk of adverse events (severe exacerbations, urgent clinical care). This was a secondary analytical strategy in which we categorized participants based on “ever” (suboptimal) versus “never” (optimal) having the poorer measure of asthma health status based on each dichotomous outcome after randomization. We estimated the incidence rate ratio (IRR) of suboptimal outcome during the year of follow-up for intervention children compared to control children using multivariable-adjusted Poisson regression models with robust standard errors adjusted for baseline value of the outcome metric, sex, age, controller medication use, and season. To reduce heteroscedasticity in the residual variance for the continuous outcomes standardized ACT score, FeNO and uLTE4 (ng/mg), we used the log-transformed values. To allow for log-transformed analysis for standardized ACT scores with a zero value the analysis used:
$$\mathit{\log}\left( standardized\ ACT\ score-\left(5/\left(100- standardized\ ACT\ score\right)-5\right)\right)$$
The percent of participant measurements unavailable for analysis did not significantly differ between intervention and control groups and was less than 7% for all primary outcome variables except for uLTE4 (ng/mg). At least one uLTE4 value was not included in the analysis for 35% of participants. This reflected agreements for the National Institutes of Health’s Children’s Health Exposure Analysis Resource laboratory processing of urine that were dependent on sending samples prior to completion of the final study visits for 12 participants. In addition, values were out of range for six participants, and samples were damaged or could not be collected for eight participants. Complete FeNO data was not available for analysis after enrollment for 13% of participants and steroid prescription and unplanned clinical utilization data over the study year was not available for 7 % of participants, due to loss to follow-up for five participants. At least one FEV1 value after enrollment was not included in the analysis for 43% of participants and at least one FEF25–75 and FVC values after enrollment was not included in the analysis for 53% of participants. Of the spirometry values not included in the statistical analysis, 34 participants had at least one value excluded due to inadequate quality and six participants were missing due to testing problems or missed visits. Spirometry is a supplemental analysis in this study given the small sample size. There was no missing data for the covariates used in the fully adjusted models. All statistical tests were 2-tailed and used a threshold of significance of p < 0.05. Statistical tests were conducted using Stata version 14.2 software (StataCorp, College Station, TX).
Sensitivity analyses
In sensitivity analyses, first to understand the intervention effect among children with more severe asthma, we examined the subset of participants who used a controller medication at baseline (N = 67). Second, to investigate the role of intervention adherence, we conducted a sensitivity analysis in subset that excluded seven intervention participants who reported turning off the HEPA air cleaner more than 2 days during the month prior to their mid-study or final study visit. Next, we conducted a sensitivity analysis adjusting for nine study participants (6 intervention group participants and 3 control group participants) who also participated in a community-based home weatherization program that became available during the HAPI Study period, because in some cases the weatherization program included individualized components that overlapped with the HAPI study intent to reduce indoor asthma triggers. Components of the weatherization program including HEPA furnace filters and vacuum cleaners, air duct sealing, roof repair, exhaust fan replacement, and carpet replacement with laminated wood flooring. Lastly, we looked at effects on ACT score in the subset (N = 69) that completed the C-ACT at baseline, which did not require standardization for interpretation. Sensitivity analyses followed the same analytic models as the primary analysis except we did not examine spirometry outcomes in sensitivity analyses due to the small sample size.