The validity of self-reported seatbelt use: Hispanic and non-Hispanic drivers in El Paso

Abstract

The validity of self-reported seatbelt use among low belt use populations has not been evaluated directly, despite the importance of such data for estimating the effectiveness of community-wide interventions and compliance with state laws. To address this gap in knowledge 612 drivers were recruited from convenience stores located in 12 randomly generated zip code areas located in El Paso, TX. Self-reported seatbelt use was compared with observed seatbelt use in a single sample of participants. Both data sets were collected almost contemporaneously and no participant was aware of having their seatbelt use observed. Hispanic (n=388) and white/non-Hispanic (n=126) drivers over reported seatbelt use by 27 and 21%, respectively. These findings suggest that response bias within low belt use populations may be greater than suggested by state and national data.

Introduction

Motor vehicle crashes (MVCs) are a leading cause of death in the United States, accounting for more than 42 000 fatalities and 3.4 million non-fatal injuries in 1997 alone (USDOT, 1999). Safety belts reduce the risk of serious motor vehicle injury by 40–52% (MMWR, 1993a). Medical costs associated with MVC’s would decline by approximately 4 billion dollars annually if occupants used safety restraints (Miller et al., 1998). Despite their effectiveness, however, safety restraints are used by only 29% of adolescents, 57% of children, and 65% of adult drivers (MMWR (1993b), USDOT, 1997). A 1997 Presidential initiative (NHTSA, 1998) seeks to increase seatbelt use to 90% by the year 2005, and similar goals have been set by Healthy People 2000 and related public health initiatives (DHHS, 1991).

National and local estimates of safety belt use are typically based on one of two assessment strategies: (1) direct observation (DO) by trained coders, and (2) self-report surveys administered to randomly identified households by trained phone interviewers. Although DO provides a valid index of restraint use, DO is subject to several limitations, including (1) restriction of seatbelt coding to day light hours, (2) restriction of seatbelt coding to a single assessment of a driver’s behavior (compared to evaluations of seatbelt use on multiple occasions), (3) difficulty of coding seatbelt use among rear seat occupants, (4) difficulty of determining belt use in vehicles with tinted windows (a growing problem for investigators), and (5) subjective assessments of a driver’s age and ethnicity (e.g. MMWR, 1988).

Self-report surveys avoid the above problems but provide estimates of safety belt use that may be distorted by respondents’ desire to report socially acceptable behavior (Wentland and Smith, 1993). The magnitude of this bias has typically been estimated by comparing rates of seatbelt use obtained from telephone self-report surveys with rates of seatbelt use observed among motor vehicle occupants. The National Highway Traffic Safety Administration, for example, compared rates of seatbelt use obtained from the 1987 Behavioral Risk Factor Surveillance System (BRFSS) telephone survey of 15 states with observed seatbelt use in the same states (MMWR, 1988). Selfreported seatbelt use exceeded observed use by as much as 24%, leading Robertson (1992) to conclude that self-report data obtained from telephone surveys are not predictive of actual seatbelt behavior. More recent data, however, suggest a greater correspondence between observed and self-reported seatbelt use rates. Dee (1998) reports an average discrepancy of 10% between observed use rates and self-reported use rates based on recent BRFSS data, and he attributes the discrepancy between BRFSS survey data and observational data to differences in sampling procedures rather than response bias.

The present study contributes to this debate in several ways. First, we compare self-reported and observed seatbelt use in a sample of drivers who were recruited from a population of drivers with a low rate of seat belt use (Womack, 1997). Drivers with low rates of safety restraint use are of particular interest because they are less responsive to seatbelt laws (Dee, 1998) and, thus, they are more likely to be targeted for seatbelt interventions. Assessing the validity of their self-report data is essential for evaluating the efficacy of large scale interventions directed towards resistant users. Second, the present study overcomes two limitations associated with earlier research: (1) we collect self-report and observational data almost contemporaneously within a single sample of participants, and (2) we collect observational data unobtrusively and thus minimize the likelihood that participants know their seatbelt use has been recorded (which could bias a driver’s self-reported seat belt use). Finally, recent research indicates that Hispanics report lower rates of seat belt use than do White/non-Hispanics and Blacks (MMWR, 1986, Matteucci et al., 1995). It is not known if this finding reflects actual differences in restraint use or, instead, reflects cultural differences in response bias. The present study also investigates this issue.

More than a quarter of a century ago researchers sought to assess the validity of self-reported seatbelt use. Waller and Barry (1969) compared self-reported seat belt use with observed seat belt use within a single sample of participants. Seat belt observations were obtained as vehicles traveled past an observation point. Drivers were subsequently identified through license plate and vehicle registration records and then administered, by mail, a survey assessing seatbelt use. Notably, surveys were mailed to participants several weeks after the observational period, introducing a possible confound into the study. Among individuals who reported ‘always’ wearing seatbelts when driving in town, only 77% were observed to be wearing seatbelts at the time of the study. Similarly, among individuals who reported ‘always’ wearing seatbelts when driving on the highway, only 46% were observed to be wearing seatbelts on the highway.

Fhaner and Hane (1973) examined the correspondence between observed and self-reported seatbelt use in a sample of 105 Swedish drivers. Seatbelt use was recorded for 13 days on a single roadway; drivers were subsequently identified though vehicle registration records and contacted, by telephone, at least two months after the observational period. Here, too, the delay in data collection may have introduced reporting bias. Among respondents who reported using seat belts 60–100% of the time (n=22), only three (14%) were observed not wearing seatbelts at the time of the study.

Stulginskas et al. (1985) assessed the validity of self-reported seatbelt use in a sample of 106 Canadian MV occupants. Belt use was assessed unobtrusively as vehicles entered a hospital parking area. Occupants were subsequently asked their name and then contacted, by telephone, later in the day. Over-reporting of belt use was present in 24% of the respondents.

Streff and Wagenaar (1989) conducted the most rigorous validity study thus far. Seatbelt use of Michigan drivers was observed at 240 intersections that were selected using a stratified probability sampling procedure. Roadside interviews were subsequently conducted with 42% (n=1869) of the sample. Overreporting of seat belt use was minimal among participants who said they ‘always’ wore seatbelts (94% were observed wearing seatbelts). Participants, however, knew they were being observed, a bias which may have minimized overreporting.

The present study contributes to the preceeding research in several ways. First, the study focuses on a population with low rates of safety restraint use: a 1997 study of 18 Texas cities revealed that El Paso drivers had the third lowest rate (74%) of safety restraint use (Womack, 1997). Indeed, El Paso is characterized by many of the risk factors associated with non-use of seatbelts, including low income and low educational levels. Second, participants in the study were unaware of being observed for seatbelt use; and third, self-report and observational data were collected almost contemporaneously.

Six hundred and twelve participants were recruited from the population of motor vehicle drivers entering gas station-convenience stores located within 12 randomly generated zip code areas in El Paso, TX. These data collection sites were selected for several reasons. First, all motor vehicle drivers must purchase gas, thereby increasing the likelihood of observing setbelt use among a heterogeneous sample of patrons and decreasing the likelihood of sampling bias that might be associated with observations collected at sites attracting subpopulations (e.g. bars or churches). Second, the stores were part of a large national chain with franchises located in each zip code area of interest; obtaining permission to collect data at these sites ensured that the data sites were virtually identical in each zip code area. Third, convenience store purchases occur quickly, allowing us to delay approaching potential participants until after drivers entered and exited the store (helping to ensure that drivers were unaware of being observed when they drove into the store parking lot).

Participants were part of a larger study examining factors that influence safety belt use (Byrd et al., 1999). Participants’ ages ranged from 16 to 83 years (mean=36.7, SD=13.1). Seventy two percent of the sample was male. Sixty three percent of the participants identified themselves as Hispanic or Mexican–American, 21% were White/non-Hispanic, and 3% were African American; the remaining 13% of the participants were classified as other.

Participants completed a ten item survey ostensibly designed to assess drivers opinions of Texas roadways. Embedded within the survey was a single item assessing seatbelt use (I wear my seatbelt: (1) always, (2) almost always, (3) sometimes, (4) rarely, (5) never). A Spanish version of the survey was also developed. Five native or bilingual Spanish speakers completed translation of the form into Spanish; a certified English–Spanish translator completed the back-translation of the form into English. The latter translator was unfamiliar with the English version of the survey.

Two independent observers recorded the seatbelt use of each participant as they drove into the parking lots of each convenience store. A third observer recorded the total number of drivers entering each location. Observers were positioned unobtrusively at a distance of approximately 20 feet from the entrance to each store. No driver indicated being aware of having their seatbelt use recorded.

Observations were collected in 2 h time blocks between 08:00 and 19:00 h, Monday through Saturday, during a 3 week period in the summer of 1997. Emergency vehicles, police cars, and delivery trucks were excluded from the study. Observers waited until drivers exited their motor vehicle before inviting them to complete the brief opinion survey regarding Texas roadways. Assistants introduced themselves as students or instructors at a local university. Participants were invited to complete the survey in either Spanish or English.