A randomized controlled trial of a computer tailored injury prevention program was conducted in the waiting area of a level 1 pediatric trauma center. A computer kiosk was used to randomly assign participants to study groups, collect baseline data and generate tailored or generic reports based on responses to safety assessment items. The kiosk was free-standing in a corner of the pediatric emergency room. Participants self-administered the assessment via a touch screen. No participants had difficulty working the computer. On average the assessment took 12 minutes to complete. The assessment was written at the 6^th grade reading level so as to accommodate participants with low literacy skills. The intervention was planned and implemented to minimize the interruption of flow of the ED. Participants generally completed the assessment prior to being called back from the waiting room. In instances when participants were called back they were able to complete the assessment upon completion of their visit. The intervention group received a personalized and stage-tailored safety report and the control group received a personalized, but otherwise generic report on other child health topics. Telephone follow-up interviews were conducted 2–4 weeks and 4–6 months after enrollment. Home visits were completed for a randomly selected subset (n=100) of parents who completed the 4–6 month follow-up interview (due to cost constraints, home visits were not possible for all participants). Details of the study methods and results from the two-week outcome analysis have been previously reported.¹⁰ Here we provide a summary of the study methods, the outcomes at the 4–6 month follow-up, and the results of the validation subset. The study was approved by the Johns Hopkins Bloomberg School of Public Health’s Institutional Review Board.

Participants were recruited from September 2004 through December 2005 during the PED’s 12 busiest weekly shifts. Triage sheets were used to identify age-eligible children and to avoid approaching the parent of any child whose visit was noted with suspicion of child abuse or neglect. Eligible parents or guardians had to be English-speaking; have a child between 4 months and 66 months of age seeking treatment for any injury or medical complaint, or report having an age-appropriate sibling of the child being seen; live in Baltimore City; and live with the child at least most of the time. Only one child from each family was included. Parents of critical patients were not included in this sample because they bypass they emergency department waiting area were recruitment and assessment took place. Parents of children whose visit was noted with suspicion of child abuse or neglect were not approached for participation.

Following confirmation of eligibility and informed consent, a study recruiter escorted participants to the computer kiosk, where a random number generation program in FileMaker Pro® assigned them to the intervention or control group. All participants then completed an assessment of their current practices and beliefs (focused on safety for the intervention group and general child health for the control group), then received parent reports that were printed at the kiosk. Participants received $10 at the time of enrollment, a $20 gift card by mail after the 4–6 mos telephone interviews, and another $20 gift card at the completion of the home observation.

Estimates for sample size calculations were taken from our previous intervention work, which demonstrated rates of safe poison storage at 10% and working smoke alarms at 80%.^{2 32} The desired sample size of 375 per study group was based on type I error α at .05 and power of 0.80. For 0.10 ≤ p₁ ≤ 0.80 (p₁ is the proportion of safe practices in the control group), we can detect differences of ≥10% in p₂ (proportion of safe practices in the intervention group) with sufficient protection from type I and type II error.³³

The intervention condition -- Safety in Seconds^™ program-- drew primarily on the Precaution Adoption Process Model (PAPM), a stage based behavior change theory asserting that individuals move from being unaware of a problem to planning to change before ultimately adopting and maintaining a new behavior.³⁴³⁵³⁶ Details of the intervention’s theoretical underpinnings have been described previously;¹⁰ here, we summarize its key elements. At the computer kiosk, parents completed a 10–12 minute PAPM stage-based assessment of the three safety behaviors of interest along with other questions used for message tailoring (e.g., sociodemographic characteristics, prevention beliefs, cultural values). Based on responses to the assessment items and the information in the message library created for this project, the computer program printed a personalized, PAPM stage-tailored, four-page safety report immediately at the kiosk.

Control group participants completed an assessment at the kiosk, the same length as that completed by the intervention group. Items included sociodemographic characteristics and questions about four child health-related topics: development, sleep, neighborhood safety, and dog bites. The control group then received a personalized kiosk-generated report, which used the same four-page template and contained generic information on the four child health-related topics. We did not assess the child safety seat, smoke alarm, and poison storage behaviors for the control group at baseline for two reasons. First, the use of a randomized design allowed us to assume equivalence of the two groups. Second, the safety behavior assessments were an inseparable component of the intervention itself (i.e., the assessments determined the message content) and were not a baseline measure in the traditional sense of a pretest/posttest design. Therefore, we concluded that it was neither necessary nor appropriate to assess the safety behaviors for the control group.

To check randomization and equivalence of the groups at the 4–6 month follow-up, we compared sociodemographic characteristics between the intervention and control groups at enrollment and at the follow up interview and between the home observation sub-sample and the follow-up sample. Knowledge outcomes were compared between study groups using t-tests of the total mean percent correct scores for each safety topic area and for all ten items. For the self-reported behavioral outcome analyses, we used ordinal regression for the quartiles of child safety seat use and logistic regression for the dichotomous poison storage and smoke alarm use variables. For the comparisons between reported and observed behaviors, we calculated positive predictive value (PPV) and negative predictive values (NPV) for smoke alarms and poison storage. PPV was calculated as the proportion of those who reported the safe behavior that were observed to be safe and the NPV was calculated as the proportion of those who reported the unsafe behavior that were observed to be unsafe. Fisher’s exact test was used to determine whether these values differed between the intervention and control groups.

Figure 1 describes recruitment, enrollment, and 4–6 month follow-up results. Of the 1,412 parents with age-eligible children who were approached, 239 (17%) were ineligible, 201 (14%) refused to participate, and 69 (5%) were missed by the recruiters due to limited time or multiple patients presenting at the same time. No significant differences were found between those who enrolled and those who refused in child’s age or reason for visit, which were the only data available for comparison. A total of 901 parents were enrolled (n=448 intervention group, n=453 control group). Follow-up rates at 4–6 months were 80% in both the intervention (n=359) and the control group (n=361).

No differences were observed at study enrollment between intervention and control groups on child’s age and sex, reason for ED visit, respondent’s relationship to the study child, ethnicity/race, marital status, employment, mother’s age, and education; nor were there any differences in rates of completing the follow-up interview. (data not shown). Regardless of study group, participants who reported higher income (per capita income >$5000) (37%) at enrollment were more likely to complete the follow-up compared with those with lower income (income <$5000) (63%) (data not shown).

At the follow-up interview, no differences were found between the study groups on sociodemographic variables or reason for visit, and no differences were found between families who received home visits and those who did not on these variables (data not shown). Since there were no differences between study groups, Table 1 presents the sample characteristics for the group that completed follow –up. Children were typically between 1 and 2 years old (42%) with slightly more boys (51%) than girls. The majority of respondents were African-American (93%), mothers (90%), not married (69%), between 20–29 years of age (55%), with a high school degree (74%) and an annual per capita income of <$5,000 (63%). Most PED visits were for medical complaints (72%) rather than an injury.

The intervention group scored significantly higher than the control group on knowledge related to smoke alarms, poison storage, and on the total knowledge score (Table 2). Percent correct scores ranged from a low of 50% for car seat knowledge to a high of 83% for poison storage. The total knowledge scores, while significantly different between groups were less than 75% correct.

Comparison between observed and self -reported behaviors are available for n=98 smoke alarm observation and n=85 poison storage observations. Those not reporting having any poison at the 4–6 month follow-up were not asked if they had a locked place. Table 3 displays the PPV and NPV for smoke alarm and poison storage by study group and for the total sub-sample with home visits. Overall, PPVs ranged from 15% for those who reported having a locked place to store poisons to 44% for those who accurately reported having a working smoke alarm on each level. NPVs ranged from 32% for those who reported not having a working smoke alarm on all levels to 100% for those who reported not having a locked place to store poisons. No statistically significant differences (Fischer’s exact test, p>.05) were found between study groups for PPV and NPV for either smoke alarms or poison storage.