Study Population
We recruited children from a large, urban pediatric emergency department (PED) in Baltimore, Maryland. Families were enrolled between January and December 2012 if they met the following enrollment criteria: (1) child aged from birth to 7 years, (2) child had a PED visit that was not a follow-up visit, (3) child was discharged home, (4) home address in Baltimore City or County, (5) parent/guardian spoke English, (6) child lived with the parent/guardian most of the time, and (7) the injury occurred in the home where the child lived most of the time (cases). Controls met all of the above inclusion criteria except that their chief complaint for the PED visit was for illness-related symptoms, not an injury. Participants were matched on variables associated with injury risk including, age, gender race, and type of housing during recruitment.
Chapter 11
168
Recruitment
We recruited parents in person in the PED or by mail or phone if the child visited the PED during hours when study staff were not available or were discharged before study staff approached the parent. In the PED, potentially eligible children were identified by reviewing the PED tracking board. Parents of age-eligible children were approached if the child presented with a chief complaint consistent with a home injury and one of the following four injury categories: 1) fall; 2) struck by/against; 3) fire/burn; or 4) cut/pierce or one of the following illness-related chief complaints: 1) fever; 2) wheezing; 3) vomiting; 4) seizure; 5) ear pain; 6) difficulty breathing; 7) cough; 8) rash; 9) abdominal pain; or 10) congestion. For parents who were missed in the PED, potentially eligible children were identified by reviewing patient discharge records. A child was excluded from the study if suspicion of abuse was noted on the tracking board or in child’s medical record.
CHASE Tool Training and Inspection Protocol
Data collectors were trained to inspect and code items according to a standardized inspection protocol for both CHASE and HQS. A field inspection guide was developed with pictures to demonstrate the pass and fail criteria. Data collectors completed human subjects training, standard data collection training, and ten hours of training in conducting the home inspection protocol. Data collectors were observed completing the inspection protocol by the research team prior to being eligible to complete it on their own.
In-Home Data Collection Protocol
A team of two data collectors completed the home visits within one to eight weeks following the PED visit. The home visit included an interview with the parent/guardian who accompanied the child to the PED and an inspection of the home with the CHASE tool. Parents were informed about the study at the time of initial contact and written informed consent was obtained from the parent/guardian at the time of the home visit. The inspection involved completing both the HQS and the CHASE by observing each floor of the household, including specifically selected rooms: kitchen, living room, (or room where the child spent the most time), child’s bedroom, and bathroom most often used by the child. Data collectors also looked for (and tested) smoke alarms on every floor, including attics and basements whenever possible.
Child Housing Assessment for a Safe Environment (CHASE)
Measures
Socio-Demographics
The in-home parent interview assessed demographic information, including parent self-reported race and ethnicity, parent education level, and estimated household income. We classified families as being above or below the Federal Poverty Level (FPL) based on the reported household income and the number of people supported with that income.
Household Characteristics
The home was classified based on parent self-report during the recruitment process into one of four housing categories: (1) row house, town house, or duplex, (2) detached single family home, (3) apartment in a house, (4) apartment in a building.
Housing Inspection Measures
Data collectors were kept blinded about the case/control status of enrollee households. When they reached the homes, they were instructed not to ask about the child’s case/control status. Each data collector completed both HUD’s Housing Quality Standards (HQS) inspection form,40 and the CHASE tool. A total of 20 HUD HQS sub-domains were included. HQS subdomains cover a comprehensive group of measures related to the adequacy and structural integrity of the home including an inspection of the condition of windows, floors, walls, ceiling, plumbing, stairs, cooking facilities etc. HQS subdomains also include examination for electrical hazards, lead based paint, security risks and smoke alarms. Exterior items (I.e. roof, gutters, chimney) from the HQS were not included in our inspection because our focus was on in-home injuries. A total of 25 CHASE items within 12 sub-domains were also inspected. HQS and CHASE items were coded as pass or fail based on the study protocol and the existing HQS standards. Failing any item within a subdomain resulted in a failure on that subdomain (e.g., failing on a book case, entertainment center hazard resulted in failing the sub-domain “tipping hazards” on the CHASE; any broken window resulted in failing the corresponding sub-domain “window condition” subdomain on the HQS).
Statistical Analysis
All statistical analyses were performed using SPSS statistical software version 25. Frequency distributions were used to report on the pass/fail rate on the CHASE and HQS. Univariate comparisons of sociodemographic characteristics between the cases (injured) and controls (sick, not injured) were made using the chi-square statistic for categorical variables. The primary analyses were the comparison between the cases and controls of the CHASE sub-domain and the HQS sub-domain score, using the average pass rate across all sub-domains.
Chapter 11
170
Children were matched on age, gender and type of housing. Matched conditional logistic regression was performed in SPSS, which is analogous to a paired t-test. Each estimated beta coefficient is interpreted as a standard regression estimate. Separate models were run for the CHASE sub-domains and the HQS sub- domains. The matched logistic regression models were adjusted by including education level, poverty status, and rental status, such that the resulting regression estimate accounts for these key covariates.