American Journal of Epidemiology Copyright ® 1992 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved

Vol. 136, No. 6 Printed in U.S.A.

Behavioral Risk Factors for Injury among Rural Adolescents

This 3-year, longitudinal, prospective study examined behavioral risk factors for medically attended injuries among a cohort of 758 rural students from Maryland's Eastern Shore region who were 12-14 years of age in 1987. Students were surveyed annually in the eighth, ninth, and tenth grades with a self-administered questionnaire. Information was obtained on the number of injuries experienced, risk-taking behaviors, delinquency, alcohol and drug use, physical exercise and sports, parental supervision, and work experience. Information on the parents' education was obtained from a parental interview. Slightly more than half (53.5%) of the students reported having experienced one or more injuries in the eighth grade as compared with one-third of the students in ninth grade, and 38% of those in the tenth grade. Poisson regression analyses were conducted to examine the association of eighth grade variables with ninth grade injuries and ninth grade variables with tenth grade injuries. Results from these analyses indicated that, in addition to sex and race, a high degree of risk taking, frequent cruising, and having high and low parental supervision in the eighth grade significantly increased the number of injuries in ninth grade. In the tenth grade, risk taking continued to be associated with injuries. In addition, students who reported disciplinary problems in school, working 1-10 hours per week, drinking on 1-2 days during the past month, lifetime use of marijuana equal to 1 -5 occasions, and involvement in sports experienced greater numbers of injuries in the tenth grade. Am J Epidemiol 1992;136:673-85. adolescence; rural population; wounds and injuries

At the time of life considered to be the healthiest, many young people die or suffer lifelong disabilities as the result of injuries. Injuries account for 57 percent of the deaths among adolescents 10-14 years of age, and

79 percent of the deaths among those 15-19 years of age (1). In 1985, adolescents and young adults (15-24 years of age) accounted for 22 percent of all injuries and one-quarter of the total costs of injury (2). Far greater than the number of deaths is the number of young people disabled by an injury. For every injury-related death in Massachusetts among 13- to 19-year-olds, there were an estimated 41 injury-related hospitalizations and approximately 1,110 emergency room visits (3). Epidemiologic studies have described the types of injuries adolescents experience and have also identified sociodemographic risk factors (e.g., race, sex, and age) for injury (4-8). Less is known about the behaviors of adolescents that contribute to the risk of injury, particularly for young people who live in rural areas. While other behaviors

August 1, 1991, and in final form May 29, 1992. 1 Department of Maternal and Child Health, The Johns Hopkins University School of Hygiene and Public Health, Baltimore, MD 2 College of Nursing, Salisbury State University, Salisbury, MD. Reprint requests to Dr. Cheryl S. Alexander, Associate Professor, Department of Maternal and Child Health, The Johns Hopkins University School of Hygiene and Public Health, 624 N. Broadway, Baltimore, MD 21205. The research reported here was supported by grants from the National Center for Nursing Research, from the National Institutes of Health (1R01NU01336-01), and from the W. T. Grant Foundation. The authors wish to thank the personnel of the health and educational departments of the three counties in the study for their collaboration. They would also like to thank Dr. Bernard Guyer for his consultation and review of an early draft.

673

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

Cheryl S. Alexander,1 Margaret E. Ensminger,1 Mark R. Somerfield,1 Young J. Kim,1 and Karin E. Johnson2

674

Alexander et al.

tivities are expected to report a greater number of injuries than their noninvolved peers. The second category of behavior represents an active lifestyle such as athletics and participation in the labor force. We predict that adolescents who are physically active or who are employed are expected to be at greater risk of injury than those who do not participate in sports or do not have jobs. It is also hypothesized that stricter family control, as indicated by parental supervision, will significantly decrease injury risk. MATERIALS AND METHODS Study population

The data for this paper come from a 3year longitudinal cohort study of adolescents from three rural counties on the Eastern Shore of Maryland. The purpose of the larger study was to investigate factors associated with the initiation of tobacco use, drug and alcohol use, and early unprotected sexual intercourse among rural youth. Because injuries represent such high morbidity among adolescents, information was also obtained on injuries among these rural youth. In 1986, all eighth grade students in the three counties were asked to participate in the study for the two succeeding years. Two of the three counties had participated in a previous cross-sectional survey of adolescent health behavior and expressed an interest in having their students enroll in the longitudinal study (14). The third county was selected based on its sociodemographic similarity to one of the counties. The three study counties are representative of rural areas on Maryland's Eastern Shore in terms of population density, major occupations, racial composition, income distribution, and poverty rates. Selected population and sociodemographic information on the counties is presented in table 1. Of the eighth graders eligible for the study, 758 (63.7 percent) participated. Objections to questions about sexual activity seemed to have influenced parental consent for participation in the study. In 1987, consent for their child's partici-

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

such as the use of controlled substances, delinquency, and sexual activity have been extensively examined in terms of their antecedents and predictors in adolescents, much less attention has been paid to the antecedents of injury in this group. Given the recognition in the past decade that "accidents" do not occur randomly (9), it seems timely and appropriate to examine adolescent injury within the framework of health behavior. According to this framework, certain adolescent behaviors and family practices are likely to place adolescents at greater risk for the occurrence of injuries. The use of controlled substances, participation in physically risky activities, and the use of an automobile for leisure are all activities that may result in injury (10). Certain family patterns may also enhance or inhibit the opportunity for injury to adolescents. Strict parental monitoring, for example, may restrict adolescents so that they are less likely to participate in activities that are more likely to lead to injuries (11, 12). Finally, certain activities may in and of themselves provide occasions for injuries to occur. Thus, adolescents who participate in sports (10) or who are employed (13) may be exposed to injury more than other adolescents. In this paper, we identify behavioral factors that differentiated injured from noninjured rural adolescents in the eighth, ninth, and tenth grades. Using longitudinal data, we examine the association of eighth grade behaviors and circumstances with injuries in the ninth grade, and of ninth grade behaviors and circumstances with injuries in the tenth grade. Because ethnicity, sex, and family background may predispose adolescents to different injury patterns, we also include race, sex, and the parents' education in the analyses. We hypothesize that two broad categories of behavior will put adolescents at greater risk for injuries. The first reflects a rebellious or risk-taking lifestyle. Thus, students who report alcohol and marijuana use, school discipline problems, riding around in cars several times per week with no particular destination, or engaging in thrill-seeking ac-

Behavioral Risk Factors for Injury among Adolescents

675

TABLE 1. Population and sociodemographic characteristics of study counties: Eastern Shore Adolescent Health Study,* 1987-1989 County B

County C

72.4 28.3 0.3 18.2

69.9 29.7 0.5 19.2

73.1 26.2

99.0 6.7

51.6 12.8

66.2 10.2

3.2 4.5 20.1 5.7 16.1 7.4 43.0

4.6 6.6 24.9 5.9 7.9 14.6 35.5

4.6 3.8

Racial composition (%) White Black Other % of school age population Population density (people/ sq. mile) % unemployed % of employment, by industry Farming Mining, construction Manufacturing Utilities, finance Trade Government Other

0.6 19.4

16.8

4.9 8.2 16.7 45.0

' Source: Maryland Department of Economic and Community Development, Baltimore, MD, 1983.

pation in the study was sought from the parents of all ninth grade students. The decision to recruit additional study respondents was based on two considerations. First, in two of the counties, there were private schools (kindergarten to eighth grade), and students from these schools often entered public high schools after completing the eighth grade. These students would not have been eligible for the study in eighth grade but were eligible in ninth grade. Second, some parents who initially refused consent for participation said they thought that some of the questions in the survey were more appropriate for high school students. We hoped that by giving these parents a second opportunity, we could increase the representativeness of the sample. An additional 120 students were enrolled, 68 percent of those eligible. No additional students were added in the tenth grade.

local interviewers when their adolescents were in the ninth grade. Complete data from eighth and ninth grade questionnaires and parent interviews were available for 612 of the original cohort of 758 eighth grade students. Data from this subsample were used in the analyses of ninth grade injuries. Complete data from ninth and tenth grade questionnaires and parent interviews were available for 632 of the 781 ninth graders participating. Included in this sample were 551 of the eighth grade cohort and 81 of the 120 students who joined the study in the ninth grade. Data from this subsample were used in the analyses of tenth grade injuries. The sample in the eighth grade was 53 percent male and 37 percent black. Sixtyfour percent of the adolescents reported living with two parents and approximately 15 percent were in single-parent households. Half indicated that they lived in small towns, 46 percent in communities but not towns, and 4 percent on farms.

Data collection

Students completed a self-administered questionnaire in the spring of each year during one 45-minute class period. Students who were absent on the day of the questionnaire administration were surveyed at a later date. Parents were interviewed by trained

Measures

Injuries were assessed by asking students whether they had had any serious accidents or injuries during the past year for which they had to be treated by a doctor or nurse.

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

County A

676

Alexander et al. stances may be impaired (10); adolescents who use drugs or alcohol may be more likely to disregard physical risks, thus increasing their exposure to injury. We obtained information on the lifetime and 30-day use of a variety of substances, including cigarettes, smokeless tobacco, marijuana, cocaine, barbiturates, and amphetamines. For these analyses, alcohol and marijuana were selected, based on the frequency of their use among this population and on their effects on motor skills and judgment. Only 30-day use of alcohol was surveyed, because pilot data indicated that 30-day use appeared to be a more reliable indicator of alcohol use than lifetime estimates. Given the low frequency of marijuana use reported within the past 30 days, the lifetime measure was used in the analyses. We also included two indicators of active lifestyle among adolescents. One, physical activity, was measured differently for middle school students and high school students because of the absence of formal team sports in the eighth grade. Eighth grade students were asked how many hours per week they spent exercising, not counting physical education classes. In high school, students were asked to report the number of teams on which they played. The second factor, labor force participation, was measured by asking students the number of hours per week they worked at a job for which they were paid. In addition to the adolescent behaviors, we included a parental supervision variable. Parent supervision and surveillance have been suggested as important factors related to childhood injuries (11, 12). Adolescents who are more closely monitored by their parents may have less opportunity for exposure to injury. Students were asked about the extent to which parents set rules for homework, curfew on weekdays and weekends, and the choice of their friends and the places where they were permitted to go. A five-item summated scale representing parental supervision was created. Three indicators of sociodemographic risk for injury were included: the race and sex of the adolescent, and the education of the

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

Students who responded affirmatively were asked to report the number of injuries experienced in the past year. Several adolescent behaviors were examined as factors that may predispose to injury. Adolescents who engage in risk taking or problem behaviors may be more likely to be injured. A measure of risk taking, constructed for the study, was used to assess the extent to which young people engage in physically daring feats (e.g., racing on a bike) and minor delinquent acts (e.g., sneaking out at night). This instrument was originally developed using the responses from focus groups of 8-10 adolescents from Maryland's Eastern Shore. Students were asked to describe "things that teenagers your age do for excitement or thrills." The development, reliability, and validity of the Adolescent Risk Taking Scale (ARTS) have been described in greater detail previously (15). Cruising, defined as riding around in cars with no particular destination, served as an indicator of increased exposure to the opportunity for injury. Students were asked to report their frequency of cruising using five response categories ranging from "not at all" to "several times per week." The leisure use of an automobile by teenagers provides opportunities for unsupervised activities and increased mobility, as well as careless driving. School discipline problems may be related to injury because students with such problems may be more likely to disregard rules, including those related to safety and fighting with other students. School discipline problems were measured in two different ways. In the eighth grade, students were asked to report whether or not they had been in enough trouble in school to be sent to the principal's office. In high school, students reported whether or not they had experienced a detention. Adolescent use of controlled substances may be associated with injuries for several reasons. First, adolescents who use drugs or alcohol may experience impaired motor skills and be injured as a result (10). Second, the judgment of users of controlled sub-

Behavioral Risk Factors for Injury among Adolescents

parent or guardian interviewed. Education served as a proxy for socioeconomic status, which has been associated with injuries in the literature on this topic (16).

The association of injury with adolescent behaviors and sociodemographic variables was examined using contingency tables, multiple logistic regression, and Poisson regression analyses employing maximum likelihood procedures. For the contingency analyses, quartiles were constructed for all scales based on the distributions of the scores for these variables. Crude risk ratios with 95 percent confidence intervals were computed for injury likelihood based on exposure to various behaviors and on sociodemographic characteristics at each grade level. In the multivariate analyses, the injury outcome was considered in two ways. First, comparisons were made between injured and noninjured students using logistic regression analyses. Multivariate analyses of the relation of variables in one year to the likelihood of injury in the next year were conducted using SAS logit procedures (SAS Institute Inc., Cary, NC). Separate predictor models were constructed for estimating ninth and tenth grade injured/noninjured status. Adjusted odds ratios with 95 percent confidence interval levels were computed from the adjusted logistic coefficients. Second, Poisson regression analyses employing the generalized linear model (GLIM) were used to model the effects of behaviors in one year on the numbers of reported injuries in the next. The number of injuries is a count variable that follows the Poisson distribution. An appropriate model for the Poisson variable is the log-linear regression model, where the logarithm of the response variable is linearly regressed on the predictor variables (17). The exponent (antilog) of the estimated regression gives the mean ratio, which is the relative estimated mean number of injuries for that category to the reference category. Thus, the mean ratio from the log-linear model is analogous to the better known odds ratio from the

logistic linear model, except that the mean ratio refers to the estimated mean numbers, while the odds ratio refers to the estimated odds of the outcome. The overall fit of each model was examined using the scaled deviance, which is a chi-square goodness of fit test (17). Models were constructed separately for the number of injuries in the ninth and tenth grades. RESULTS Study attrition

Complete ninth grade questionnaire and parent information was obtained for 612 of the original 758 eighth grade students, which represents an 80.7 percent retention rate. The retention rate from the ninth to the tenth grades for students in the study, including parental data, was 80.9 percent. A series of analyses was performed using the chi-square statistic to assess the extent to which students with incomplete information—absence of a survey at either Time 2 or Time 3, or of a parent interview—differed from those with complete information on sociodemographic and behavioral characteristics. In addition, respondents and nonrespondents at the eighth grade recruitment time period were compared based on sex and race. This was the only sociodemographic information made available by the schools. The counties provided us with mean California Achievement Test scores for each school, but we were unable to separate out scores for participants in the study from scores for nonparticipants. Table 2 displays comparison findings. A significantly greater percentage of the students whose eighth and ninth grade data were incomplete reported frequent cruising in the eighth grade, more risk taking, and lifetime marijuana smoking of six times or more. In the tenth grade, students whose data were incomplete were significantly more likely than those for whom complete data were available to have worked more than 11 hours per week, to be female, and to report lifetime marijuana smoking of six times or more, more frequent alcohol use,

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

Statistical analyses

677

TABLE 2. Comparison of percentages of respondents and nonrespondents at recruitment and each followup periodf: Eastern Shore Adolescent Health Study, 1987-1989 Recruitment: Time 1 (8th grade) Variables

Time 2 (9th grade) Time 3 (10th grade) Re^nondents NonresDondents Respondents Nonrespondents Respondents Nonrespondents (n = 758) (n = 1,172) (n = 612) (n = 146) (n = 632) (n=149)

53.0 47.0

51.0 49.0

53.2 46.8

54.7 45.3

50.4 49.6

37.0** 63.0

62.5 37.5

65.4 34.6

63.0 37.0

63.7 36.3

64.6 35.4

57.6 42.4

34.8 65.2

48.6** 51.4

40.4 49.6

51.3* 48.7

26.9 41.6 31.5

33.7* 43.3 23.0

22.8 47.5 29.8

26.9 46.4 26.7

42.5 57.5

56.3** 43.7

51.6 48.4

55.6 44.4

58.0 21.7 20.3

51.3 22.9 25.8

54.6 18.2 27.2

43.1* 23.2 33.7

80.6 12.2

65.7*** 17.4 16.9

73.7 13.2 13.1

59.0*** 14.5 26.5

38.5 47.6 13.9

45.2 41.8 13.0

>3 days Lifetime marijuana use None 1-5 times >6 times Exercise (hours/week) 1 2-5 >6 Number of team sports played None 1-3 >4 Parental supervision (percentile) High (upper 25th) Medium (mid-50th) Low (lower 25th) Employment (hours/ week) None 1-10

7.2 14.3 42.3 43.4

>11

Injuries Yes No

16.2 40.4 43.4

25.7 52.6 21.7

29.0 53.1 17.9

22.1 47.1 30.8

31.3" 49.6 19.1

50.6 38.9 10.5

42.6 41.9 15.5

56.2 24.0 19.8

54.1* 16.6 29.3

47.0 53.0

45.8 54.2

33.5 66.5

38.8 61.2

• p < 0.05; ** p < 0.01 ; • * * p < 0.001 (chi-square test of association). t Respondents were defined as follows: 8th grade—students whose parents consented to their participation in the study and who completed the 8th grade questionnaire; 9th grade—students for whom complete 8th and 9th grade questionnaires and parent data were available; 10th grade—students for whom complete 9th and 10th grade questionnaires and parent data were available. Nonrespondents were defined as follows: 8th grade—students whose parents refused to consent to their participation; 9th and 10th grades—students for whom the questionnaire data at either Time 2 or Time 3 was incomplete, or for whom complete parental data was unavailable. $ Cruising: riding around in cars with no particular destination.

678

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

Sex Male Female Race White Nonwhite Cruising^ (times/week) Several 11 hours per week. These findings are based on crude risk ratios and have not been adjusted for other variables.

Distribution of injuries by grade level

TABLE 3.

679

680

Alexander et al.

TABLE 4. Risk ratios for injury in the eighth, ninth, and tenth grades: Eastern Shore Adolescent Health Study, 1987-1989 8th grade

Sex Male Female Race White Nonwhite Parental education High school Cruisingt (times/week) Several 6 times Exercise (hours/week) 4 Parental supervision (percentile) High (upper 25th) Medium (mid-50th) Low (lower 25th) Employment (hours/week) None 1-10 211

10th grade

95% Cl*

RR

95% Cl

RR

95% Cl

.47 .00

1.23-1.76

1.94 1.00

1.53-2.47

1.40 1.00

1.15-1.72

.11 .00

0.92-1.32

1.21 1.00

0.95-1.55

1.02 1.00

0.83-1.26

(3.88 1.00 .09

0.70-1.11

0.91 1.00 0.89

0.66-1.23

0.93 1.00 1.12

0.71-1.23

.39 .00

1.18-1.65

0.97 1.00

0.77-1.22

1.43 1.00

1.14-1.79

.88 .00 (3.45

1.61-2.20

1.42 1.00 0.64

1.12-1.79

1.51 1.00 0.63

1.23-1.85

.58 .00

0.92-1.30

0.34-0.58 1.33-1.87

1.34 1.00

0.70-1.12

0.48-0.85 1.07-1.69

1.32 1.00

0.91-1.37

0.49-0.82 1.08-1.61

1.00 .15 .54

0.95-1.40 1.30-1.83

1.00 1.28 1.03

0.99-1.66 0.80-1.32

1.00 1.37 1.37

1.10-1.71 1.12-1.67

1.00 .31 .39

1.05-1.62 1.08-1.78

1.00 1.43 1.16

1.10-1.87 0.86-1.58

1.00 1.12 1.34

0.88-1.43 1.07-1.69

(3.87

0.73-1.04

.00 .31

1.11-1.56

1.00 1.16 1.73

0.93-1.46 1.36-2.20

1.00 1.35 1.03

1.10-1.65 0.74-1.45

1.19 1.00 3.61 1.00 1.19 1.03

0.99-1.43 0.47-0.80

1.29 1.00 0.91

1.00-1.41 0.78-1.36

1.00 0.85 1.48

* RR, risk ratio; Cl, confidence interval. t Cruising: riding around in cars with no particular destination.

1.01-1.64

0.79-1.27

0.71-1.17

1.01 1.00 1.13

0.91-1.41

0.65-1.13 1.16-1.87

1.00 0.96 1.27

0.75-1.24 1.04-1.36

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

RR*

9th grade

Behavioral Risk Factors for Injury among Adolescents

681

TABLE 5. Adjustedf odds ratios for ninth grade injury using eighth grade predictors and for tenth grade injury using ninth grade predictors: Eastern Shore Adolescent Health Study, 1987-1989 9th grade (n = 545) Prpri i p tor^ rluUlwLUI Q

1

2-5 >6 Parental supervision (percentile) High (upper 25th) Medium (mid-50th) Low (lower 25th) Employment (hours/week) 0 1-10 >11

Model x 2

95% C\%

OR

95% Cl

1.96* 1.00

1.26-3.04

1.25 1.00

0.84-1.85

1.20 1.00

0.75-1.92

0.91 1.00

0.60-1.39

0.59 1.00 0.77

0.34-1.04

1.17 1.00 1.22

0.69-1.97

1.36 1.00

0.87-2.14

1.00 1.00

0.67-1.49

1.33 1.00 1.25

0.80-2.21

1.21 1.00 0.71

0.75-1.95

1.27 1.00

0.81-1.99

1.27 1.00

0.86-1.89

0.50-1.18

0.77-2.04

0.82-1.82

0.45-1.12

1.00 0.80 0.79

0.48-1.33 0.44-1.43

1.00 1.69* 1.74*

1.05-2.71 1.07-2.84

1.00 2.03* 0.99*

1.11-3.71 0.43-2.26

1.00 1.33 0.82

0.78-2.31 0.45-1.50

1.00 1.66* 1.46

1.11-2.40 0.83-2.57

1.37 1.00 1.79

0.71-2.64

1.28 1.00 1.04

0.81-2.03

1.00 1.02 2.37*

0.91-3.53

0.63-1.72

0.94 1.00 0.94

0.67-1.59 1.26-4.45

1.00 1.20 1.32

60.38, df = 1 8 , p < 0.001

•p6 times Number of team sports played None 1-3 >4 Exercise (hours/week)

ORt

10th grade (n = 571)

682

Alexander et al.

DISCUSSION

Adolescent injury is not a random event. The findings presented here suggest that there are transitions in the lives of adolescents that have implications for their risk of injury. Injuries were more frequently reported among students in their last year of middle school than in their first year of high school. At 13 years, the average age of eighth graders, the size and development of adolescents vary widely (18). These developmental differences may influence injury occurrence. Additionally, the range of activities across grades may differ, contributing to the observed differences in injury rates by grade level. Fewer behaviors were associated with injuries in the ninth grade than in the tenth. Sex and race, two risk factors for injury identified in epidemiologic studies (4-8), were significant predictors of injury rates in the ninth grade. Once students were in high school, their drug use, risk taking, discipline problems, and participation in sports were the significant predictors of tenth grade injuries. These findings suggest that among high school age adolescents, the effects of sex and race on injury may be mediated by behavior. Rather than focusing on sex per se, researchers should give greater consider-

ation to behaviors that place males at risk for injury. Participation in the labor force was associated with the likelihood of injury in the ninth grade and with greater numbers of injuries in the tenth grade. Occupationally related injuries have received little research attention, especially among adolescents; however, data from Sweden indicate a higher incidence of injuries among employed adolescents as compared with nonemployed adolescents (13). While we are unable to determine from these data the number of injuries that occurred as a result of job-related activities, the positive association between the number of hours worked per week and the number of injuries supports the need to examine the prevalence, nature, and specific causes of occupational injuries for young adolescents. The positive association between high parental supervision in the eighth grade and the number of ninth grade injuries was unexpected. We had hypothesized that strictly supervised students would have a reduced risk of injury. Findings from the Poisson regression analyses of ninth grade injuries indicated, though, that both high and low parental supervision were significantly associated with greater numbers of injuries. It may be that strict parental supervision is a response to the rebellious behavior of adolescents, the same young people who are at risk for injury. In a study of parental discipline and adolescent sexual behavior, Miller et al. (19) found a curvilinear relation, with sexual activity rates highest among adolescents with permissive parents, intermediate among adolescents with strict parents, and lowest among those who reported moderate parental supervision. They speculate that very low or very high levels of parental control are not effective in shaping adolescent behavior. Given the emphasis that some investigators and clinicians are placing on the importance of parental monitoring for adolescent behaviors such as delinquency (20, 21) and substance use (22), further exploration of parent supervision and its association with injury in adolescents is warranted.

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

that, in addition to the effects of sex and race, high risk taking, frequent cruising, and limited as well as strict parental supervision in the eighth grade significantly increased the mean number of injuries in the ninth grade (table 6). In the tenth grade, risk taking continued to be associated with injuries. In addition, students with the following characteristics experienced a greater number of injuries in the tenth grade: employment of 1-10 hours per week, discipline problems in school during the ninth grade, drinking on 1-2 days during the past month, lifetime use of marijuana on one to five occasions, and involvement in sports. Being male continued to be associated with significantly higher numbers of injuries when other variables were controlled.

Behavioral Risk Factors for Injury among Adolescents

683

TABLE 6. Adjustedt mean ratios for ninth grade injury using eighth grade predictors and for tenth grade injury using ninth grade predictors: Eastern Shore Adolescent Health Study, 1987-1989 9th grade (n = 520) Predictors

Scaled deviance

95% CI*

10th grade (n = 551) Mean ratio

95% Cl

1.54* 1.00

1.17-2.02

1.29* 1.00

1.10-1.52

1.55* 1.00

1.17-2.06

0.96 1.00

0.96-1.14

0.76 1.00 0.76*

0.54-1.08

1.02 1.00 0.98

0.82-1.26

1.39* 1.00

1.08-1.80

1.10 1.00

0.94-1.29

1.63* 1.00 1.13

1.22-2.16

1.62* 1.00 1.12

1.35-1.95

1.26 1.00

0.60-0.98

0.83-1.54 0.97-1.64

1.23* 1.00

0.83-1.14

0.92-1.36 1.05-1.44

1.00 0.76 0.75

0.56-1.02 0.55-1.03

1.00 1.36* 1.11

1.13-1.64 0.91-1.35

1.00 1.23 0.69

0.90-1.68 0.41-1.14

1.00 1.83* 0.78

1.53-2.20 0.61-1.01

1.00 1.29* 1.33*

1.04-1.61 1.13-1.57

1.24 1.00 1.19

0.82-1.88

1.42* 1.00 1.38*

1.09-1.84

1.00 1.09 1.06*

0.78-1.82

1.04-1.84

1.09 1.00 0.91

0.84-1.40 0.51-2.79

1.00 1.39* 1.19

724.69, df = 501

•p6 times Number of team sports played None 1-3 >4 Exercise (hours/week) 1 2-5 >6 Parental supervision (percentile) High (upper 25th) Medium (mid-50th) Low (lower 25th) Employment (hours/week) 0 1-10 >11

Mean ratio

684

Alexander et al.

and injury occurrence (e.g., eighth grade behaviors predicting ninth grade injury). Thus, the findings of this study are probably conservative estimates of the influence of behavior on adolescent injuries. Findings from this study of injuries among rural youth underscore the importance of behavioral and developmental factors in adolescent injuries, particularly as young people make the transition between middle school and high school. Exposure seems to be related to injury. Risk taking, frequent cruising, employment, and sports all present opportunities for injury. Behaviors such as drug use and school discipline problems also relate to injuries. Adolescents who engage in these behaviors may be less cautious in other aspects of their lives, and therefore put themselves at risk for injury. Initial participation rates and subject losses over the 3-year period do, however, limit the generalizability of these findings. Students who enrolled in the study did not differ significantly in race and sex from those whose parents refused participation. However, based on our observations as well as reports from the schools, we expect that nonparticipants came from politically and religiously conservative families; adolescents from such families may be less likely to engage in risky behaviors. If injury rates for nonparticipants were the same as, or greater than, the rates for study participants, the exposure-specific injury rates of nonparticipants would be higher than those of participants. Given these conditions, our findings would underestimate the "true" relation between risky behaviors and injury. In contrast, if the injury rates for nonparticipants were lower than those of participants, the effect of nonparticipants on the association between risky behaviors and injury would be minimal. Because we do not have data on either behaviors or injuries for nonparticipants, we can only speculate about the nature of this potential selection bias. Students lost to the study through attrition tended to exhibit more delinquent and risky behaviors than those who were retained. Because the incidence of injury and number of

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

Two measurement issues exist that make these injury data different from those reported in other studies of injury and that may affect their interpretation. All data used in the analyses were based on self-reported information. Adolescents were asked to report only those injuries that required medical treatment. Asking about medically attended injuries is one way to capture more severe injuries and to reduce recall bias since, presumably, adolescents would remember injuries that required a visit to a physician. A longitudinal study found that adolescents were able to report accurately approximately 60 percent of emergency room attended injuries, based on two-year recall and medical record validation (23). Recall bias in our study may be less a problem, since students were asked to report injuries in a 12-month rather than a 24month period. The time elapsed since an injury was shown to be a significant factor in recall in the New Zealand study. A second measurement issue in this study concerns variation in the timing of behaviors and injury events. Students were asked to report injuries that had occurred during the past year, but the time frame for questions regarding other behaviors, such as risk taking, began with the start of the school year. Still other behaviors, including the use of alcohol and marijuana, were reported on the basis of 30-day and lifetime use, respectively. Furthermore, since the report of injury was not specifically connected to these behaviors, we cannot delineate causal relations. We were able, though, to examine the effect of behaviors reported in one year on injury reports in the subsequent year. This strategy improves on cross-sectional studies, in which associations between behaviors and injury are examined using data collected during the same time period. This crosssectional analysis is problematic, since it is possible that injuries could precede the behaviors. However, if behaviors precipitate injury, then longitudinal analyses such as ours may underestimate the true effect of behavior on injury because of the increased time period between behavioral measures

Behavioral Risk Factors for Injury among Adolescents

REFERENCES 1. National Center for Health Statistics. Vital statistics of the United States, 1982. Mortality, Vol. 2, Part A. Washington, DC: US GPO, 1986. (DHHS publication no. (PHS)86-1122). 2. Rice DP, MacKenzie EJ, et al. Cost of injury in the United States: a report to Congress. San Francisco, CA: Institute for Health & Aging, University of California, San Francisco, and Baltimore, MD: Injury Prevention Center, School of Hygiene and Public Health, The Johns Hopkins University, 1989. 3. Gallagher SS, Finison K, Guyer B, et al. The incidence of injuries among 87,000 Massachusetts children and adolescents: results of the 1980-81 Statewide Childhood Injury Prevention Program Surveillance System. Am J Public Health 1984;74: 1340-7.

4. Rivara FP, Mueller BA. The epidemiology and causes of childhood injuries. J Soc Issues 1987;43: 13-31. 5. Runyan CW, Gerken EA. Epidemiology and prevention of adolescent injury: a review and research agenda. JAMA 1989;262:2273-9. 6. Chalmers DJ, Cecch J, Langley JD, et al. Injuries in the 12th and 13th years of life. Aust Paediatr J 1989;25:14-20. 7. Paulson JA. The epidemiology of injuries in adolescents. Pediatr Ann 1988; 17:84-96. 8. Bass JL, Gallagher SS, Mehta KA. Injuries to adolescents and young adults. Pediatr Clin North Am 1983;32:31-9. 9. Baker S, O'Neill R, Karp R. Injury fact book. Lexington, MA: Mass Books, 1984. 10. Runyan CW, Gerken EA. Injuries. In: Hendee WR, ed. The health of adolescents: understanding and facilitating biological, behavioral, and social development. San Francisco, CA: Jossey-Bass Publishers, 1991:302-33. 11. Peltzer K. Accidents/injuries of children in surgical admissions at the University Teaching Hospital in Lusaka, Zambia: a psycho-social perspective. Cent AfrJMed 1989;35:319-22. 12. Carrigan L, Heimbach DM, Marvin JA. Risk management in children with burn injuries. J Burn Care Rehabil 1988;9:75-8. 13. Jacobsson B, Schelp L. One year incidence of occupational injuries among teenagers in a Swedish rural municipality. Scand J Soc Med 1988; 16: 21-5. 14. Alexander CS, Klassen A. Drug use and illnesses among eighth graders in rural schools. Public Health Rep 1988;394-8. 15. Alexander CS, Kim YJ, Ensminger M, et al. A measure of risk taking for young adolescents: reliability and validity assessments. J Youth Adolesc 1990; 19:559-69. 16. Rivara FP. Epidemiology of childhood injuries. I. Review of current research and presentation of conceptual framework. Am J Dis Child 1982; 136: 399-405. 17. Frome EL. The analysis of rates using Poisson regression models. Biometrics 1983,39:665-74. 18. Tanner JM. The course of children's growth. In: Hill JP, Sheldon J, eds. Readings in adolescent development and behavior. Englewood Cliffs, NJ: Prentice-Hall Inc., 1971:6-22. 19. Miller BC, McCoy JK, Olson TD, et al. Parental discipline and control in relation to adolescent sexual attitudes and behavior. Journal of Marriage and the Family 1986;48:503-12. 20. Wilson H. Parental supervision: a neglected aspect of delinquency. Br J Criminol 198O;2O:2O3-35. 21. Lober R, Dishion T. Early predictors of male delinquency: a review. Psychol Bull 1983,94:68-99. 22. Ensminger M, Brown, C, Kellam S. Sex differences in antecedents of substance use among adolescents. J Soc Issues 1982;38:23-42. 23. Langley JD, Cecch, JC, Williams SM. Recall of injury events by thirteen year olds. Methods Inf Med 1989;28:24-7. 24. Peterson L, Farmer J, Mori L. Process analysis of injury situations: a complement to epidemiological methods. J Soc Issues 1987;43:33-44.

Downloaded from https://academic.oup.com/aje/article-abstract/136/6/673/55277 by Tulane University user on 17 January 2019

injuries reported by students lost through attrition did not vary from that reported by students who remained in the study, our findings may overestimate the relation between behavior and injury. Given the assumption that we underestimated the association between risky behaviors and injury based on recruitment selection and overestimated this association based on loss at follow-up, it is possible that selection bias operated in both directions. Without additional information about nonparticipants, we cannot estimate the magnitude of the bias nor fully specify its direction. Research is now needed to delineate the mechanisms by which behaviors and environments interact to place young adolescents at the risk of injury and how these interactions differ for rural, urban, and suburban youth. A next step in this research would be to use process analysis of injury events for adolescents. Process analysis is a qualitative approach that attempts to specify the environmental and behavioral precursors and consequences of both injuries and near injuries (24). This type of analysis may yield important data on the linkage between behavioral risk factors and injury suggested by the research reported here. Such information in combination with descriptive data is essential for determining appropriate interventions.

685

Behavioral risk factors for injury among rural adolescents.

This 3-year, longitudinal, prospective study examined behavioral risk factors for medically attended injuries among a cohort of 758 rural students fro...
907KB Sizes 0 Downloads 0 Views