This was a cross-sectional study to determine the prevalence of and factors associated with safety belt non-use among ED patients at an urban, teaching hospital in Boston, Massachusetts. The facility has an annual volume of over 90,000 visits.

Research staff approached patients in the Adult ED through a process of systematic sampling (see below). Adult patients (age 18 and over) who were deemed "stable" by the ED attending on duty, with non-urgent medical complaints (i.e. patients not triaged to the acute area of the ED), who were able to speak either English or Spanish, and were able to provide verbal informed consent for this anonymous survey, were eligible for participation in the study. Patients were excluded if they were unable to provide verbal consent (including those experiencing altered mental status due to drug/alcohol intoxication, injury, or acute psychiatric illness); in severe pain or distress; if the patient was deemed by ED staff to be too ill to be interviewed (i.e., major trauma or medical illnesses); if the patient was a prisoner in custody, or if the patient had previously completed our survey. The Institutional Review Board at our institution approved this study.

Our survey process consisted of systematically approaching adult ED patients via room-to-room assessment, between the hours of 8 am to 8 pm, Monday through Friday, from February 2004 to April 2004. The systematic sampling method consisted of research assistants walking clockwise from room-to-room in the ED, sampling every patient to identify those that met enrollment criteria. If a patient met enrollment criteria, they were asked to participate in a questionnaire on a range of health and safety issues important to emergency medicine practice. If a patient did not meet criteria, or refused to participate, research staff members went to the next sequential room.

Participants were asked questions read to them by research staff from a questionnaire in either English or Spanish. This was done to expedite survey completion, minimize the risk of missing data, and reduce the likelihood of language or literacy barriers. The survey contained questions regarding participants' demographics, frequency of riding in automobiles, frequency and correlates of safety belt use and safety belt non-use (e.g., listing items under the heading, 'things that make me wear my seat belt' and 'things that make me not wear my seat belt'), and frequency of alcohol consumption. The survey generally required 5 to 7 minutes to complete.

The outcome of interest, safety belt use, was determined by two methods. Using a single, global measure of safety belt use, participants were asked to respond to the following single question based on a 5-point scale: "Think about the times you've ridden in a car in the past 30 days. Overall, how often did you wear a seatbelt? a) Always, b) More than half the time, c) About half the time, d) Less than half the time, or e) Never." According to the Single Measure, the overall safety belt use outcome for analysis was determined by the frequency of participants that answered 'always' on this question. Conversely, any answer other than 'always' on this scale was determined to be non-use. The Single Measure of Safety Belt Use is a validated single-item measure of safety belt use which has been validated, and is currently the standard instrument used for population-based surveys on safety belt use and other health-related risk factors in the US 28293031. These population-based surveys are used to monitor the behaviors that may influence disease or injury, as well as to help set the national health agenda. Despite its wide use in survey-based research, the Single Measure of Safety Belt Use has its limitations. Critics point out that the use of the single measure can overestimate observed safety belt use by 2–27% as compared to direct observation 323334. For example, in a study of self-reported safety belt use compared to directly observed safety belt use, Nelson found that the validity of self-reported safety belt use was best among respondents residing in States with the highest observed prevalence of safety belt use, and conversely, the lowest validity was in States with a low prevalence of safety belt use 32. He suggests that the risk of over-reporting bias due to social desirability would be greatest in areas of lowest observed safety belt use. Also, in the 2003 Motor Vehicle Occupant Safety Survey sponsored by the National Highway Transit Safety Administration, self reported safety belt use using a single measure was 84% 35. However, in a follow-up question, 7% of persons stating they wore their safety belt 'always' stated they had driven unbelted within the past week. Using these two questions in previous years' Motor Vehicle Occupant Safety Survey, the follow-up question reduced the rate of self-report safety belt use by an average of 8–10%.

To minimize the effects of over-reporting in our study, safety belt use was also assessed by a series of nine separate questions (the Multiple Measure of Safety Belt Use) on scenario-specific aspects of transportation (highway, local, daytime, nighttime, driver, front-seat passenger, backseat, short rides, and long rides) (Figure 1). We constructed the Multiple Measure of Safety Belt Use – an instrument consisting of 9 distinct scenario-specific survey questions – via a focus group of public health researchers and emergency physicians. We piloted the questions among a sample of ED patients for content and syntax prior to commencing the study. Participants were prompted to answer these questions using the same 5-point frequency scale used to assess the Single Measure of Safety Belt Use. A description of the Multiple Measure of Safety Belt Use, and a comparison between the Single Measure and the Multiple Measure has been presented elsewhere 35. Briefly, we classified safety belt use by the Multiple Measure of Safety Belt Use as a response of 'always' to all nine scenarios. Participants that responded with an answer of less than 'always' to any one of the above 9 questions were categorized as non-use according to the Multiple Measure.

The explanatory variables that were examined in association with the outcome were: demographics (age, gender, race, education, possession of a valid driver's license); safety questions (history of a car crash, being cited for not wearing a safety belt, drinking alcohol prior to driving a car, riding in a car with someone who drank alcohol before driving); frequency of use of public and private transportation; and attitudes about seatbelt use ('seatbelts are uncomfortable', etc.).

To assess factors associated with seatbelt non-use, chi-square test was performed for categorical variables. Variables that were statistically significant at the p < 0.05 level from exploratory analysis were tested via univariate logistic regression. Statistically significant variables from univariate logistic regression (p < 0.05) were used to construct a multivariate logistic regression model. Backwards likelihood ratio testing was employed for model selection. Data was analyzed using STATA/SE 8.0 for Windows (Stata Corporation, College Station, TX).

Of 478 individuals approached, 39 (8%) patients refused, 58 (12%) were excluded, and 381 (80%) patients completed the survey. Patients were excluded for the following reasons: pain/too sick (18), language barrier (22), in custody (6), altered mental status (9), previously enrolled (3). There were 183 (48%) African Americans, 10 (3%) Asians, 2 (1%) Native American, 151 (40%) Whites, and 34 (9%) Other patients enrolled in the study. In terms of ethnicity, there were 77 (20% of the total sample) Hispanic or Latino, 19 (5%) Haitian, and 20 (5%) Cape Verdean. For the purpose of this paper, the term 'Hispanic' will be used to refer to those persons self-described as 'Hispanic or Latino'. Of the 77 Hispanics in the study, 32 (8% of total sample) completed their surveys in Spanish. Although not generally included, categories for 'Haitian' and 'Cape Verdean' ethnicity were added because they represent some of the main ethnic groups represented at our institution. The mean age of study participants was 39 years. Patients were 52% male and 48% female (two surveys had missing responses to gender), and 78% of respondents had at least a high school education. There were no statistical differences between participants and those that elected not to participate in terms of age or sex, however, enrollment was only 36% among Asians, compared to 100% for Whites, 87% for African-Americans, and 87% of other races (p < 0.001). Also, Hispanics were less likely to enroll than patients of other ethnicities (79% Hispanic vs. 100% Cape Verdean, 95% Haitian, p = 0.007).

Table 1 shows the association between demographic characteristics and responses to the Single Measure and the Multiple Measure of Safety Belt Use. Women and ethnicity were statistically associated with a response of 'always' wearing a safety belt. Table 2 shows the results of selected questions on safety belt use. Some items associated with belt use were: a history of being in a car crash, receiving a citation for not wearing a safety belt, as well as the quantity and frequency of alcohol use.

Table 3 shows the variance between the Single Measure and the Multiple Measure of Safety Belt Use. The greatest variance between an answer of 'always' on the Single Measure and the specific scenarios of the Multiple Measure of Safety Belt Use was being a back seat passenger (147 of 187, 79% overall agreement); the greatest agreement between the Single Measure and specific scenarios was in driving (137 of 142 respondents able to operate a vehicle, 97% overall agreement). According to the Single Measure of Safety Belt Use, 192 of 382 respondents (51%) stated the 'always' wear safety belts. However, using the 9 components of the Multiple Measure of Safety Belt Use, only 137 of 382 (36%) indicated they 'always' wear safety belts. The converse situation was not observed, where respondents reported less than 'always' to the Single Measure, but 'always' to all nine items of the Multiple Measure.

Reasons were elicited for participants' safety belt use and non-use, and were compared to responses on the Single Measure and Multiple Measure of Safety Belt Use. The statements that "seatbelts are too uncomfortable" and "I just forgot" were among those most strongly associated with safety belt non-use (i.e. "things that make me not wear my seatbelt") on both the Single Measure and the Multiple Measure of Safety Belt Use (p < 0.001). The statements, "I like taking risks" and "I've got too many other things to think about" were strongly associated with non-use according to the Single Measure (p < 0.001). The statements "I've got too many other things to worry about" and "friends make fun of me when I wear a seatbelt" were also associated with non-use according to the Multiple Measure (p = 0.009 and p = 0.016, respectively). The statement that "wearing my seatbelt is my usual habit" was among those that were most strongly associated with safety belt use (i.e., "things that make me wear my seatbelt") according to both the Single Measure and the Multiple Measure of Safety Belt Use (p < 0.001). The statement "I'm afraid of being killed in a car crash" was also associated with safety belt use according to both the Single Measure and Multiple Measure of Safety Belt Use (p = 0.005 and p = 0.009, respectively) The statements "I wear my seat belt because it's the law", "my parents made we wear a seatbelt" and "everyone I know wears a seatbelt" were all strongly associated with safety belt use according to the Single Measure (p < 0.001, p = 0.002, p = 0.003, respectively).

The results of a multivariate logistic regression model of factors associated with less than 'always' wearing a safety belt according to the Single Measure of Safety Belt Use are shown in Tables 4. Male gender, maximum alcohol consumption of greater than 5 drinks in a single episode, riding with others that drink and drive, ever receiving a traffic citation for not wearing a safety belt, the belief that safety belt use is "uncomfortable", and that "I just forget to use my seatbelt" were risks for safety belt non-use, while "it's my usual habit" was protective. Table 5 shows the results of a multivariate logistic regression model predicting less than 'always' wearing a safety belt according to the Multiple Measure of Safety Belt Use. Risks included: male gender, use of public transportation greater than 10 times per month, riding with others that drink and drive, being given a traffic citation for not wearing a safety belt, the belief "wearing a seatbelt is uncomfortable", that "I've got too many other things to think about", and "I just forget to use my seatbelt". As in the previous model, a response that safety belt use is a usual habit was protective against non-use.