Grandparents and Injuries

Risk Factors for Unintentional Injuries in Children: Are Grandparents Protective? David Bishai, Jamie L. Trevitt, Yiduo Zhang, Lara B. McKenzie, Tama Leventhal, Andrea Carlson Gielen and Bernard Guyer Pediatrics 2008;122;e980-e987 DOI: 10.1542/peds.2007-2995 The online version of this article, along with updated information and services, is located on the World Wide Web at: http://www.pediatrics.org/cgi/content/full/122/5/e980 PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since 1948. PEDIATRICS is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, 60007. Copyright © 2008 by the American Academy of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275. Downloaded from www.pediatrics.org by on November 4, 2008 ARTICLE Risk Factors for Unintentional Injuries in Children: Are Grandparents Protective? David Bishai, MD, PhD, MPHa, Jamie L. Trevitt, MPPa, Yiduo Zhang, PhDa, Lara B. McKenzie, PhDb, Tama Leventhal, PhDc, Andrea Carlson Gielen, ScD, ScMd, Bernard Guyer, MD, MPHa a Department of Population Family and Reproductive Health and dCenter for Injury Research and Policy, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland; bResearch Institute at Nationwide Children’s Hospital, Columbus, Ohio; cDepartment of Child Development, Tufts University, Boston, Massachusetts The authors have indicated they have no financial relationships relevant to this article to disclose. What’s Known on This Subject There is only conjecture that grandparents have old-fashioned sensibility about child safety and child care practices that could put children at risk. Children who are watched by grandparents are a select group, and the reasons that they enter grandparent care may confound inference. What This Study Adds We found that children who were aged 30 to 33 months, whose mothers work, and who are watched by grandparents have a statistically significantly lower rate of having an injury hospitalization in the previous 12 months. ABSTRACT OBJECTIVE. We sought to identify sociodemographic and familial correlates of injury in children aged 2 to 3 years. METHODS. The Healthy Steps data set describes 5565 infants who were enrolled at birth www.pediatrics.org/cgi/doi/10.1542/ peds.2007-2995 doi:10.1542/peds.2007-2995 Key Words child care, family issues, injury, safety Abbreviation OR— odds ratio Accepted for publication Jul 29, 2008 Address correspondence to David Bishai, MD, PhD, MPH, Johns Hopkins Bloomberg School of Public Health, Department of Population Family and Reproductive Health, 615 N Wolfe St E4622, Baltimore, MD 21205. E-mail: dbishai@jhsph.edu PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275). Copyright © 2008 by the American Academy of Pediatrics in 15 US cities in 1996 –1997 and had follow-up until they were 30 to 33 months of age. Data were linked to medical claims reporting children’s medically attended office visits by age 30 to 33 months. Each claim was accompanied by a reason for the visit. An analytical sample of 3449 was derived from the children who could be effectively followed up and linked to medical charts. Missing data were imputed by using multiple imputation with chained equations. The analytical sample showed no systematic evidence of sample selection bias. Multivariate logistic regression was used to determine the odds ratios of injury events. RESULTS. Odds of medically attended injuries were decreased for children who received care from grandparents. Odds were increased for children who lived where fathers did not co-reside or in households where the parents never married. Statistical results were robust to the addition of a variety of covariates such as income, education, age, gender, and race. CONCLUSIONS. Children are at higher risk for medically attended injury when their parents are unmarried. Having grandparents as caregivers seems to be protective. Household composition seems to play a key role in placing children at risk for medically attended injuries. Pediatrics 2008;122:e980–e987 affect roughly 1 of 4 US children every year.1,2 Current estimates are that in 2000, injuries to those 14 years of age cost $50.6 billion ($11.9 in medical costs and $38.7 in productivity losses).3 Risk factors that are associated with unintentional nonfatal injuries in young children include individual characteristics of children (age, gender, temperament/behavior), the family environment (socioeconomic status, family structure, siblings, parental characteristics, child care arrangements), and community and neighborhood factors; therefore, childhood injuries are predicted by a complex and interacting set of sociodemographic and child-related factors that make it difficult to separate individual effects.4 Much of what has been learned has emerged from small injury registries that cover only a subsample of the population. These are sometimes complemented by studies of large national samples.5 Recent growth in the number of grandparental caregivers has some observers concerned that grandparents will apply a style of child care that is less adherent to modern safety practices.6 The number of grandparents who reside with children has been increasing since the 1970s.7 The US Census Bureau estimated that as of 2000, there were 5.8 million grandparents living with grandchildren, and 31% of these grandparents reported responsibility for grandchildren.8 If there is a systematic problem with the child safety practices of grandparent caregivers, then a relationship between injury rates and grandparental care could be a clue to this and would be a reason for greater concern. To our knowledge, there have been no previous studies of the relationship between grandparental caregivers and child injury rates. e980 BISHAI et al I NJURIES ARE THE leading cause of death for children in the United States.1 Injuries that require medical attention Downloaded from www.pediatrics.org by on November 4, 2008 The objective of this article was to examine individual and family risk factors for injuries in toddlers (2–3 years of age) using a large national survey. We pay particular attention to the potential relationship between grandparental caregivers and child injury rates. Family characteristics and parental safety behaviors are particularly important in the case of toddlers, who depend on effective supervision by their adult caregivers for protection from injuries. We hypothesize that grandparental caregiving would be associated with injury risks in toddlers, and additional sociodemographic characteristics of caregivers would further help to explain toddlers’ injury risks. One of the particular strengths of our study is the presence of data on who was taking care of the child for mothers who worked. Whereas others have already noted that children of employed mothers have lower injury rates9 and that children in formal child care have lower injury rates,10 little is known about the effects of grandparents on child safety. A growing number of “grandparenting workshops” are premised on the belief that older caregivers need to be updated on the latest developments in modern child care, including car seats, lower hot water temperatures, and other child safety practices that emerged after their own children were toddlers. We hypothesized that if grandparents are indeed systematically applying outdated child safety practices, then there would be a positive association between grandparental caregiving and child injury and a negative association between grandparental caregiving and adoption of better safety practices. METHODS Data Source and Study Design Data are from the National Evaluation of the Healthy Steps for Young Children Program, and the sampling strategy is described in detail elsewhere.11 These data describe 5565 infants who were enrolled in 15 US cities in 1996 –1997 and had follow-up until they were 30 to 33 months of age. Telephone interviews were conducted with parents of Healthy Steps families for phase 1 of the National Evaluation as the Healthy Steps children reached 30 months of age. The primary respondent for the interview was the mother or was the guardian or the primary caregiver when the mother was not available. The questionnaire included an update of sociodemographic characteristics of the family. This research was approved by the Johns Hopkins Committee on Human Research. Study Sample Of the 5565 children enrolled, 3737 (67%) were followed up at 30 to 33 months. A total of 2450 children had no missing data on any outcomes or covariates. For each variable, this sample did not differ significantly from the full sample on any variable. An initial analysis was performed without performing any imputation. Subsequently, an iterative multiple imputation procedure was performed by using switching regression and chained equations.12 This process resulted in a sample of 3449 children with some imputed values. The regression analysis was repeated on the imputed sample, yielding coefficients similar in size and significance to those estimated initially. Tables 1 and 2 report the findings from the imputed data set because the imputed values are less likely to be subject to sample selection bias. Dependent Variables Serious child injury as a dependent variable was measured by records of medical use at hospitals, emergency departments, and urgent visits to the practice for each child. These data were gathered by staff from the pediatric practice and were based on both billing records and medical chart review. The original reason that the medical use data were collected was to test the hypothesis that the Healthy Steps Program would lower medical use and costs for sick visits. As part of the Healthy Steps study, office staff prepared electronic files that listed charges, International Classification of Diseases, Ninth Revision code, and Current Procedural Terminology procedure code, and an open-ended “reason for visit” was filled out by office staff. Entries in the “reason for visit” field were tabulated and coded as injuries or noninjuries by the investigators. The injury reasons ranged from “cut face” to “drank paint thinner” to “fall from shopping cart.” Noninjury reasons ranged from “fever 3 days” to “turned blue.” Differentiating injuries from noninjuries left little ambiguity and even permitted identification of the cause of each injury; however, only 6 burns requiring hospitalization and 18 poisonings requiring hospitalization were recorded. Efforts to model determinants of these specific events were attempted and later abandoned because of small sample size. One limitation from the office-based approach is that there could have been some injury-related hospitalizations and emergency department visits that never came to the attention of the pediatric office and would be missed in the analysis, such as injuries that occurred while a child was traveling out of state. Conversely, a strength of this approach is that it avoids problems of parental recall or reporting bias, which has proved to be large with periods longer than 5 weeks.13 Independent Variables On the basis of the literature, we included the following child, maternal, and family characteristics in our analysis: child gender; child birth weight; maternal demographics (age race/ethnicity, education status, first-time mother); maternal health (self-rated) physical limitations, sense of competence, Center for Epidemiologic Studies Depression Scale maternal depression score; household income; mother’s behavior (smoking, drinking, illicit substances); mother’s marital status; family structure (father co-resides, stepfather co-resides, members of household, siblings); whether the nonmaternal caregiver was a father, grandparent, other relative, or nonrelative; and whether the family had moved in the previous year. It should be noted that to remain in the Healthy Steps evaluation database, a child had to remain in the same pediatric practice, so the relocation in this PEDIATRICS Volume 122, Number 5, November 2008 e981 Downloaded from www.pediatrics.org by on November 4, 2008 TABLE 1 Descriptive Data Independent Variable No (N Child is male First-time mother Mother’s age, y 15–19 20–24 25–29 30–34 35–49 40 Marital status Never married/not cohabiting to married Never married/not cohabiting at both rounds Never married/cohabiting to married Married to unmarried Partnered to unmarried/unpartnered Parents still partners but not cohabiting both rounds Never married and cohabiting both rounds Other transitions in martial status Mother’s education Less than high school High school graduate Some college College graduate Mother’s race White Black Asian/Native American Other Hispanic Mother’s health Physical limitations Self-rated health is excellent Self-rated health is very good Self-rated health is good Self-rated health is fair Self-rated health is poor CES-D score is 16 While mother works Nobody else watches child Father watches child Grandparent watches child Other relative watches child Nonrelative or child care watches child Other watches child Primary caregiver Father Grandmother Household income $20 000–$49 000 $50 000 Moved in the previous year Family structure Father is not co-residing Stepfather or adoptive father co-residing Biological grandparent co-residing Nonbiological grandparent co-residing Great-grandparent co-residing Child Had Hospital or ED Record of Any Injury in Previous Year, n (%) 3222) Yes (N 227) Total (N 3449) 1594 (49.50) 1565 (48.60) 89 (2.80) 577 (17.90) 794 (24.60) 904 (28.51) 630 (19.60) 226 (7.00) 56 (1.70) 74 (2.30) 111 (3.40) 110 (3.40) 111 (3.40) 132 (4.10) 378 (11.70) 130 (4.00) 318 (9.87) 877 (27.20) 919 (28.50) 1108 (34.40) 1983 (62.32) 724 (22.50) 127 (3.90) 371 (11.50) 587 (18.20) 594 (18.40) 1071 (34.62) 1119 (34.70) 734 (22.80) 209 (6.50) 27 (0.80) 220 (6.80) 196 (6.10) 235 (7.30) 481 (14.90) 170 (5.30) 1013 (31.40) 79 (2.50) 57 (1.80) 419 (13.00) 1025 (34.40) 1408 (43.70) 638 (19.80) 737 (22.90) 56 (1.70) 351 (10.90) 22 (0.70) 31 (1.00) 119 (52.40) 100 (44.10) 4 (1.80) 50 (22.00) 63 (27.80) 63 (27.75) 40 (17.60) 7 (3.10) 3 (1.30) 5 (2.20) 6 (2.60) 13 (5.70) 9 (4.00) 16 (7.00) 40 (17.60) 15 (6.60) 28 (12.33) 74 (32.60) 69 (30.40) 56 (24.70) 141 (63.23) 62 (27.30) 7 (3.10) 15 (6.60) 36 (15.90) 44 (19.40) 77 (34.68) 74 (32.60) 55 (24.20) 19 (8.40) 2 (0.90) 22 (9.70) 14 (6.20) 17 (7.50) 27 (11.90) 18 (7.90) 73 (32.20) 3 (1.30) 7 (3.10) 22 (9.70) 60 (27.90) 88 (38.80) 31 (13.70) 77 (33.90) 4 (1.80) 30 (13.20) 2 (0.90) 2 (0.90) 1713 (49.70) 1665 (48.30) 93 (2.70) 627 (18.20) 857 (24.80) 976 (28.13) 670 (19.40) 233 (6.80) 59 (1.70) 79 (2.30) 117 (3.40) 123 (3.60) 120 (3.50) 148 (4.30) 418 (12.10) 145 (4.20) 346 (10.03) 951 (27.60) 988 (28.60) 1164 (33.70) 2124 (62.38) 786 (22.80) 134 (3.90) 386 (11.20) 623 (18.10) 638 (18.50) 1148 (34.62) 1193 (34.60) 789 (22.90) 228 (6.60) 29 (0.80) 242 (7.00) 210 (6.10) 252 (7.30) 508 (14.70) 188 (5.50) 1086 (31.50) 82 (2.40) 64 (1.90) 441 (12.80) 1085 (34.00) 1496 (43.40) 669 (19.40) 814 (23.60) 60 (1.70) 381 (11.00) 24 (0.70) 33 (1.00) ED indicates emergency department; CES-D, Center for Epidemiologic Studies Depression Scale. e982 BISHAI et al Downloaded from www.pediatrics.org by on November 4, 2008 TABLE 2 Logistic Regression Analysis for Any Hospital Chart of Injury in Previous Year Reported at Age 30 to 33 Months (N Parameter Mother’s demographics Education: college graduate Race: other Age: 40 y Mother’s transitions Stayed partnered Stayed never married Moved in the previous year Family structure Father does not co-reside Caregivers while mother works Nobody else watches child Grandparent watches child Primary caregiver is father Log likelihood Pseudo-R2 CI indicates confidence interval. a Significant at 1%. b Significant at 5%. c Significant at 10%. 3449) Model D Stepwise, OR (95% CI) 0.749 (0.536–1.047)b 0.535 (0.310–0.924)c 0.418 (0.193–0.903)c 1.938 (1.113–3.377)c 0.600 (0.404–0.892)c 1.909 (1.387–2.627)a Model A Bivariate, OR (95% CI) 0.625 (0.458–0.852)a 0.544 (0.319–0.928)c 0.422 (0.196–0.906)c 1.775 (1.037–3.038)c 1.609 (1.125–2.302)a 0.641 (0.434–0.945)c 1.731 (1.299–2.306)a 1.015 (0.580–1.776) 0.769 (0.509–1.163) 1.767 (0.796–3.919) Model B Multivariate, OR (95% CI) 0.713 (0.398–1.276) 0.409 (0.208–0.803)a 0.398 (0.178–0.888)c 1.990 (1.097–3.608)c 1.519 (0.645–3.578) 0.584 (0.388–0.877)a 1.493 (0.694–3.211) 0.932 (0.506–1.719) 0.651 (0.401–1.057)b 2.317 (0.990–5.422)b 797.89 0.0467 Model C Multivariate, OR (95% CI) 0.652 (0.358–1.187) 0.416 (0.211–0.820)c 0.397 (0.178–0.886)c 2.093 (1.147–3.818)c 1.567 (0.665–3.693) 0.584 (0.389–0.878)a 1.516 (0.707–3.250) 0.936 (0.507–1.726) 0.642 (0.395–1.043)b 2.324 (0.993–5.442)b 797.07 0.0477 0.666 (0.437–1.016)b 2.045 (0.907–4.608)b 807.16 0.0356 database would have been confined to local relocation within the catchment area of a pediatric practice. Relocation was more common among families who had made a marital transition and among lower income households. Mothers were asked to identify who the primary caregiver was, and they cited themselves 85% of the time, fathers 2% of the time, and grandmothers 13% of the time. In addition, mothers were asked whether they worked and, if they did, who watched the infant: mother herself, father, grandparents, other relatives, nonrelatives, or others. There was very little correlation between the identity of the “primary caregiver” and the alternative person who watched the child while the mother worked. The raw correlation between grandparents’ being primary caregivers and grandparents’ watching the child while the mother worked was 0.076; the corresponding correlation for fathers was 0.16. Analyses To determine which individual, maternal, and family characteristics affected the likelihood of childhood injury, we used logit models to regress the dichotomous indicator for injury on the independent variables discussed previously. The variables shown in Table 1 were entered into the model 1 block at a time, and the pattern of confounding was explored by observing how significance levels were altered as additional variables were entered. In Table 2, model A shows the results of bivariate models. For model B, multivariate regression models that do not adjust for household income are shown, and model C shows how the results differ when income is controlled. Model D is included to reveal what would emerge in the most parsimonious specification but was not guided by previous literature or theory. The complete models are listed in the Appendix. RESULTS Table 2 presents pertinent correlates of hospitalized injury that were found to be significant and pertinent variables that were found to be insignificant. Child’s birth weight had no statistically significant effect on injury; boys had higher odds of injury at 1.121, but this was not statistically significant. Black and Hispanic children had similar injury rates to white children. Children of mothers who were 40 years old had lower injury rates. There was no statistically significant relationship between income and injury, and including income measures in the model did not alter the effects of race and family structure. Compared with women who stayed married throughout the child’s life, odds for injury were statistically significantly higher for children whose parents never married. Children of women who became divorced or separated had a higher odds ratio (OR) of injury at 1.143, but these results were not statistically significant and are not shown in Table 2. Also not shown in Table 2 are statistically insignificant results for mother’s self-rated health, depressive symptoms, and sense of competence. Compared with a situation in which the mother did not work, children whose mother worked and arranged for grandparent caregivers had a statistically significantly lower odds for injury, but child care by other relatives was not statistically significant. Because families are not always in a situation in which family members can provide child care while the mother works, a variable for child care by nonrelatives was included in the model, but it was not statistically significant. In households where the grandmother was listed as the primary caregiver, there was no difference in the odds for injury. Among the 64 households in which fathers were cited as the primary caregiver, the OR for child injury was higher PEDIATRICS Volume 122, Number 5, November 2008 e983 Downloaded from www.pediatrics.org by on November 4, 2008 with an OR of 2.0 in multivariate models; however, in situations in which the father watched the child while the mother worked, there was no statistically significant difference in the odds for injury. Contrary to expectation, there was a protective association between residential relocation since birth and the odds for injury. This finding was robust to including or excluding controls for marital transitions, household income, and insurance. DISCUSSION In this study we examined determinants of childhood injury and parental safety behaviors. We now examine our principal findings in light of previous research in this area. Mother’s Demographics A previous study found that children of adolescents were more likely to have 1 serious injury.14 In contrast, we found that mothers who were older than 40 years had lower odds for children with an injury, but we did not find that younger or adolescent mothers had a significantly higher odds for injury among their children. The literature suggests that parental education is negatively associated with childhood injury. Parents who did not complete high school had children who were more likely to be injured than children of parents with high school degrees.15,16 Our results confirm this by showing a significant relationship between the highest level of maternal education and lower rates of child injury. Although it is often hard to separate the effect of race and ethnicity from other socioeconomic determinants of injury, minority children are at higher risk for injury. In the United States, Native American children have the highest unintentional injury death rate followed by black children.17 Our analysis showed no difference in injury rates for black and Hispanic children in unadjusted models as well as models that controlled for income and other socioeconomic correlates of injury. In contrast to previous studies, 1 distinguishing feature of the sample that we studied was that all of the children in the Healthy Steps sample had regular pediatric care, and this may have partially mitigated the effects of race and ethnicity. Family Composition Previous studies showed that children who have serious injuries are more likely to have parents who are unmarried or single.5,14–16 Dawson18 found that children in divorced families or disrupted marriages had a 20% to 30% greater risk for injury. Our results support the previous findings. Children who lived with parents who were cohabiting and never married had roughly twice the odds for injury, and children without co-resident fathers had a higher odds for injury, although the results were significant only in the bivariate and stepwise models. Source of Child Care Interpreting the relationship between child care and injury is complicated by the presence of 2 alternative ine984 BISHAI et al dicators with minimal overlap. The correlation between who was named as primary caregiver and who watched the child while the mother was working was weak. Having fathers named as primary caregiver was associated with significantly higher rates of injury, but having fathers watch the children while the mother worked did not have a significant association with child injury. Having the grandmother named as primary caregiver was not associated with the odds for injury, but having the grandmother watch the child while the mother worked was associated with fewer injuries (P .1). A safe conclusion from these data is that there is no evidence to support the hypothesis that grandparental care is associated with more childhood injuries. We speculate that the question of who the primary caregiver for a child was could have been decided differently by different respondents (mostly mothers). Some might have interpreted the question to mean “who makes the primary decisions about child care,” others as “who actually delivers care.” For example, in 90% of the 252 cases in which the mother worked and the father watched the child, the mother stated that she is the primary caregiver, and the same was stated for 81% of the 508 cases in which the grandmother watched the child during the mother’s work hours. Of the 64 men described as primary caregiver, only 24 of them watched the child while the mother worked. Because the respondents who stated that fathers were primary caregivers seem to have been influenced by psychosocial circumstances other than the time that fathers spent watching children, we suspect that it is these other, unspecified circumstances that are more strongly associated with higher rates of child injury and not the actual process of fathers’ watching children. The OR in Table 2 compares a situation in which grandmothers watch a child while the mother works with a stay-at-home mother. To compare grandmothers with other relatives and child care, we ran an additional model similar to model C (data not shown), in which other relatives/child care was the excluded category. This model demonstrated an OR of 0.71 with (P .109) for the protective effect of grandmothers relative to child care or other relatives. According to the literature, better supervision of children is associated with fewer injuries,19 but it is still not clear what kind of child care situations provide the best form of supervision. What is known is that children in child care centers rarely sustain severe injuries.20 Children who spent more time in nonparental child care were slightly less likely to have an unintentional injury after adjustment for other known risk factors and characteristics of child care centers.10 Child care centers are highly regulated for safety in the United States, but the protective effect extends to less regulated family child care environments as well.10 In a study that compared children who received home care, center-based care, and other forms of out-of-home child care, the rate of minor injuries was highest in center-based care, but there was not a significant difference among the 3 types of care for severe injuries.21 We did not find any signif- Downloaded from www.pediatrics.org by on November 4, 2008 icant associations between child care centers or nonrelative child care. There are several limitations in this study. It is unknown who was watching the child at the time of the injury, and it is possible that some injuries that did not come to the knowledge of the pediatric office occurred. The 7% incidence of injuries is lower than the 11% rate for children who were younger than 15 reported by the Centers for Disease Control and Prevention.22 Unfortunately, despite the sample size of 3449, there is still not sufficient sample size to break down the effects of grandparental care by race/ethnicity or by the number of children being watched by 1 caregiver. The data do not allow us to know how old the grandparents were, and this could affect the results. Other unmeasured aspects of the family may still confound this relationship. CONCLUSIONS Our study of a nationwide survey of children who were enrolled at birth and followed until age 30 to 33 months found that having grandparents as caregivers was protective, cutting the odds for injury roughly by half compared with having a stay-at-home mother. We cannot claim that this association is causal, although it remains robust after controlling for socioeconomic status, race, ethnicity, parental safety behavior, and measures of family composition. To our knowledge, this is the first analysis to look at an association between grandparental caregivers and child injury. Our results provide evidence that children who are cared for by grandparents have a lower incidence of a medically attended injury. It may still be the case that the households that choose grandparents as caregivers selectively choose only grandparents who will enforce healthier child safety behaviors; however, we note that when households “choose” relatives other than grandparents to look after their children, the risk for child injury increases slightly. Although additional studies of how households choose relatives to watch their children and the actual caregiving styles of grandparents are warranted, for now there is no evidence that grandparental care places children at higher risk. ACKNOWLEDGMENT This work was supported by Maternal and Child Health Bureau grant R40MC05475. REFERENCES 1. Fingerhut LA, Warner M. Injury Chartbook: Health, United States, 1996 –97. Hyattsville, MD: National Center for Health Statistics; 1997 2. Kogan MD, Overpeck MD, Fingerhut LA. Medically attended nonfatal injuries among preschool-age children: national estimates. Am J Prev Med. 1995;11(2):99 –104 3. Corson K, Gerrity MS, Dobscha SK. Screening for depression and suicidality in a VA primary care setting: 2 items are better than 1 item. Am J Manag Care. 2004;10(11 pt 2):839 – 845 4. Dal Santo JA, Goodman RM, Glik D, Jackson K. Childhood 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. unintentional injuries: factors predicting injury risk among preschoolers. J Pediatr Psychol. 2004;29(4):273–283 Russell KM. Preschool children at risk for repeat injuries. J Community Health Nurs. 1998;15(3):179 –190 Levey B. When grandparents risk junior’s safety. Washington Post. December 20, 2001;sect C:11 Casper L, Bryson K. Co-resident Grandparents and Their Grandchildren: Grandparent Maintained Families. Washington, DC: US Census Bureau; 1998:26. Population Division Working Paper Simmons T, Dye JL. Grandparents Living With Grandchildren: 2000. Washington, DC: US Census Bureau; 2003 Gordon RA, Kaestner R, Korenman S. The effects of maternal employment on child injuries and infectious disease. Demography. 2007;44(2):307–333 Schwebel DC, Brezausek CM, Belsky J. Does time spent in child care influence risk for unintentional injury? J Pediatr Psychol. 2006;31(2):184 –193 Guyer B, Hughart N, Strobino D, Jones A, Scharfstein D, Healthy Steps Evaluation Team. Assessing the impact of pediatric-based developmental services on infants, families, and clinicians: challenges to evaluating the Healthy Steps Program. Pediatrics. 2000;105(3). Available at: www.pediatrics.org/cgi/ content/full/105/3/e33 Royston P. Multiple imputation of missing values. Stata J. 2004;4(3):227–241 Warner M, Schenker N, Heinen MA, Fingerhut LA. The effects of recall on reporting injury and poisoning episodes in the National Health Interview Survey. Inj Prev. 2005;11(5): 282–287 Kahlmeier S, Schindler C, Grize L, Braun-Fahrlander C. Perceived environmental housing quality and wellbeing of movers. J Epidemiol Community Health. 2001;55(10):708 –715 Sanmuganathan PS, Ghahramani P, Jackson PR, Wallis EJ, Ramsay LE. Aspirin for primary prevention of coronary heart disease: safety and absolute benefit related to coronary risk derived from meta-analysis of randomised trials. Heart. 2001; 85(3):265–271 Durkin MS, Davidson LL, Kuhn L, O’Connor P. Low-income neighborhoods and the risk of severe pediatric injury: a smallarea analysis in Northern Manhattan. Am J Public Health. 1994; 84(4):587–592 Rubsam JM. Identification of risk factors and effective intervention strategies corresponding to the major causes of childhood death from injury. J N Y State Nurses Assoc. 2001;32(2): 4–8 Dawson DA. Family structure and children’s health: United States, 1988. Vital Health Stat 10. 1991;(178):1– 47 Morrongiello BA, House K. Measuring parent attributes and supervision behaviors relevant to child injury risk: examining the usefulness of questionnaire measures. Inj Prev. 2004;10(2): 114 –118 Alkon A, Ragland DR, Tschann JM, Genevro JL, Kaiser P, Boyce WT. Injuries in child care centers: gender-environment interactions. Inj Prev. 2000;6(3):214 –218 Kotch JB, Dufort VM, Stewart P, et al. Injuries among children in home and out-of-home care. Inj Prev. 1997;3(4):267–271 Bergen G, Chen L, Warner M, Fingerhut LA. Injury in the United States: 2007 Chartbook. Hyattsville, MD: National Center for Health Statistics; 2007 Johnston B. Family life events and child injury risk. Presented at: Ambulatory Pediatrics Association; May 5, 2002; Baltimore, MD Downloaded from www.pediatrics.org by on November 4, 2008 PEDIATRICS Volume 122, Number 5, November 2008 e985 APPENDIX Logistic Regression for Complete Models (N Outcome Variable: Any Hospital Chart of Injury in Previous Year Child’s characteristics Birth weight, oz Birth weight, oz2 Male Positive behavior score (sum ) 30–33 mo Mother’s demographics First-time mother Education High school graduate Some college College graduate Race Black Asian/Native American Other Hispanic or Latino Age 15–19 20–24 25–29 35–39 40 Health status Sense of competence (sum ) Sense of efficacy (sum ) CES-D score (sum ) Physical limitation(s) Self-rated health Very good Good Fair Poor Household income $50 000 Mother’s transitions Never married/not cohabiting to married Never married/not cohabiting to never married/cohabiting Never married/cohabiting to married Married to unmarried Partnered to unmarried/unpartnered Stayed partnered Stayed never married Other transitions in martial status Moved in the previous year Family structure Father does not co-reside Stepfather or Adoptive father co-resides Biological grandparent co-resides Nonbiological grandparent co-resides Great-grandparent co-resides No. of male adults ( 18 y) in the household No. of female adults ( 18 y) in the household Caregivers while mother works Nobody else watches child Father watches child Grandparent watches child Other relative watches child Nonrelative or child care watches child Other watches child Primary caregiver Father Grandmother 3449) Model B Multivariate, OR (95% CI) 1.051 (0.980–1.126) 1.000 (1.000–1.000) 1.121 (0.849–1.480) 1.145 (1.020–1.284)a 0.839 (0.618–1.138) 0.959 (0.583–1.578) 0.885 (0.524–1.495) 0.713 (0.398–1.276) 0.934 (0.629–1.386) 0.776 (0.346–1.743) 0.409 (0.208–0.803)c 1.137 (0.711–1.816) 0.502 (0.160–1.571) 1.014 (0.634–1.621) 1.048 (0.713–1.540) 0.872 (0.575–1.323) 0.398 (0.178–0.888)a 0.974 (0.931–1.020) 1.002 (0.943–1.065) 1.171 (0.709–1.934) 1.018 (0.705–1.468) 0.950 (0.680–1.327) 1.011 (0.698–1.463) 1.194 (0.684–2.085) 0.818 (0.178–3.752) Model C Multivariate, OR (95% CI) 1.052 (0.981–1.128) 1.000 (1.000–1.000) 1.121 (0.849–1.480) 1.142 (1.018–1.281)a 0.835 (0.615–1.133) 0.944 (0.573–1.554) 0.854 (0.504–1.448) 0.652 (0.358–1.187) 0.957 (0.643–1.425) 0.781 (0.348–1.753) 0.416 (0.211–0.820)a 1.152 (0.720–1.843) 0.495 (0.158–1.550) 1.057 (0.657–1.701) 1.080 (0.732–1.592) 0.862 (0.568–1.307) 0.397 (0.178–0.886)a 0.973 (0.930–1.019) 1.003 (0.944–1.066) 1.176 (0.711–1.943) 1.015 (0.703–1.464) 0.956 (0.684–1.336) 1.033 (0.712–1.498) 1.244 (0.710–2.181) 0.857 (0.187–3.935) 1.266 (0.882–1.818) 0.834 (0.233–2.980) 1.129 (0.402–3.173) 0.880 (0.367–2.110) 1.682 (0.665–4.252) 1.143 (0.416–3.136) 2.093 (1.147–3.818)a 1.567 (0.665–3.693) 1.809 (0.799–4.095) 0.584 (0.389–0.878)c 1.516 (0.707–3.250) 1.140 (0.373–3.486) 0.651 (0.320–1.325) 0.966 (0.204–4.570) 0.526 (0.109–2.540) 1.114 (0.720–1.723) 1.389 (0.868–2.224) 0.936 (0.507–1.726) 0.877 (0.493–1.560) 0.642 (0.395–1.043)b 1.252 (0.710–2.207) 0.890 (0.619–1.281) 0.512 (0.155–1.693) 2.324 (0.993–5.442)b 0.910 (0.564–1.468) Model D Stepwise, OR (95% CI) Model A Bivariate, OR (95% CI) 1.002 (0.995–1.009) 1.000 (1.000–1.000) 1.125 (0.860–1.473) 1.141 (1.022–1.275)a 0.834 (0.636–1.093) 1.293 (0.969–1.725)b 1.094 (0.816–1.467) 0.625 (0.458–0.852)c 1.296 (0.957–1.756)b 0.775 (0.358–1.680) 0.544 (0.319–0.928)a 0.846 (0.586–1.221) 0.631 (0.230–1.735) 1.295 (0.934–1.795) 1.175 (0.869–1.588) 0.880 (0.619–1.252) 0.422 (0.196–0.906)a 0.965 (0.927–1.005)b 0.991 (0.938–1.048) 1.464 (0.924–2.321) 1.064 (0.756–1.496) 1.052 (0.249–4.451) 1.317 (0.807–2.150) 1.084 (0.791–1.485) 0.909 (0.682–1.211) 0.816 (0.619–1.075) 0.757 (0.235–2.438) 0.958 (0.383–2.394) 0.761 (0.331–1.750) 1.719 (0.951–3.104)b 1.157 (0.579–2.314) 1.775 (1.037–3.038)a 1.609 (1.125–2.302)c 1.683 (0.969–2.923)b 0.641 (0.434–0.945)a 1.731 (1.299–2.306)c 1.014 (0.364–2.822) 1.246 (0.835–1.858) 1.293 (0.302–5.533) 0.915 (0.218–3.848) 0.689 (0.517–0.916)a 1.281 (0.974–1.685)b 1.015 (0.580–1.776) 1.029 (0.617–1.717) 0.769 (0.509–1.163) 1.546 (0.933–2.546)b 1.034 (0.775–1.379) 0.533 (0.167–1.701) 1.767 (0.796–3.919) 0.718 (0.457–1.128) 1.144 (1.025–1.277)a 0.794 (0.598–1.053) 0.740 (0.536–1.047)b 0.535 (0.310–0.924)a 0.418 (0.193–0.903)a 0.969 (0.969–1.010) 0.815 (0.228–2.911) 1.090 (0.388–3.060) 0.853 (0.356–2.041) 1.609 (0.637–4.064) 1.098 (0.400–3.011) 1.990 (1.097–3.608)a 1.519 (0.645–3.578) 1.728 (0.765–3.907) 0.584 (0.388–0.877)c 1.493 (0.694–3.211) 1.156 (0.378–3.540) 0.655 (0.322–1.331) 0.970 (0.205–4.597) 0.515 (0.107–2.483) 1.131 (0.732–1.750) 1.403 (0.877–2.246) 0.932 (0.506–1.719) 0.875 (0.492–1.556) 0.651 (0.401–1.057)b 1.266 (0.718–2.231) 0.914 (0.636–1.312) 0.527 (0.160–1.743) 2.317 (0.990–5.422)b 0.918 (0.569–1.481) 1.938 (1.113–3.377)a 0.600 (0.404–0.892)a 1.909 (1.387–2.627)c 0.666 (0.437–1.016)b 2.045 (0.907–4.608)b e986 BISHAI et al Downloaded from www.pediatrics.org by on November 4, 2008 APPENDIX Continued Outcome Variable: Any Hospital Chart of Injury in Previous Year Log likelihood Pseudo R2 CI indicates confidence interval; CES-D, Center for Epidemiologic Studies Depression Scale. a Significant at 1%. b Significant at 5%. c Significant at 10%. Model A Bivariate, OR (95% CI) Model B Multivariate, OR (95% CI) 797.89 0.0467 Model C Multivariate, OR (95% CI) 797.07 0.0477 Model D Stepwise, OR (95% CI) 807.16 0.0356 Downloaded from www.pediatrics.org by on November 4, 2008 PEDIATRICS Volume 122, Number 5, November 2008 e987 Risk Factors for Unintentional Injuries in Children: Are Grandparents Protective? David Bishai, Jamie L. Trevitt, Yiduo Zhang, Lara B. McKenzie, Tama Leventhal, Andrea Carlson Gielen and Bernard Guyer Pediatrics 2008;122;e980-e987 DOI: 10.1542/peds.2007-2995 Updated Information & Services References including high-resolution figures, can be found at: http://www.pediatrics.org/cgi/content/full/122/5/e980 This article cites 15 articles, 9 of which you can access for free at: http://www.pediatrics.org/cgi/content/full/122/5/e980#BIBL This article, along with others on similar topics, appears in the following collection(s): Office Practice http://www.pediatrics.org/cgi/collection/office_practice Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: http://www.pediatrics.org/misc/Permissions.shtml Information about ordering reprints can be found online: http://www.pediatrics.org/misc/reprints.shtml Subspecialty Collections Permissions & Licensing Reprints Downloaded from www.pediatrics.org by on November 4, 2008