[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
November 2006

The Association Between Parents' Past Agricultural Injuries and Their Children's Risk of InjuryAnalyses From the Regional Rural Injury Study–II

Author Affiliations

Author Affiliations: Regional Injury Prevention Research Center, Division of Environmental Health Sciences, School of Public Health, University of Minnesota, Minneapolis (Drs Carlson, Alexander, and Gerberich; Ms Langner; and Messrs Ryan and Mongin); Child Health Evaluation and Research Unit, Department of Pediatrics, University of Michigan, Ann Arbor (Dr Gurney); and Division of Education and Research, St Mary's Duluth Clinic Health System, Duluth, Minn (Ms Renier).

Arch Pediatr Adolesc Med. 2006;160(11):1137-1142. doi:10.1001/archpedi.160.11.1137

Objective  To examine associations between parents' and children's agricultural injuries in a cohort of farming and ranching households.

Design  Analyses from a population-based, nested case-control study.

Setting  The 1999 Regional Rural Injury Study–II, involving a cohort of 3765 agricultural households. Demographic, injury, and exposure data were collected for household members for 1 year.

Participants  A total of 203 injured children (cases) and 755 randomly selected control children were identified for the study.

Main Exposure  Children's risk of injury was estimated in reference to individual and combined parental injury experience. Two periods were evaluated, separately and in combination.

Main Outcome Measures  Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated by using logistic regression; directed acyclic graphs guided selection of potential confounders.

Results  When controlling for potential confounders, children whose fathers were injured before the study year had twice the risk of injury of those whose fathers were not injured (OR, 2.1; 95% CI, 1.5-3.0). Children had increased risk of injury if their mothers were injured before the study (OR, 2.5; 95% CI, 1.7-3.8) or during the study (OR, 2.0; 95% CI, 0.9-4.2). Children whose parents both reported agricultural injuries before the study had a 4-fold increase in injury risk over those with neither parent injured (OR, 4.2; 95% CI, 2.6-6.9).

Conclusions  Positive associations between parents' and children's injuries were observed, with a potential additive effect if both parents were injured. These results indicate a need for further research into the social and/or physical environments driving these associations so that appropriate interventions for pediatric injury can be determined.

Injury, the leading cause of death among children,1 results in substantial costs in terms of disability, rehabilitation, and lost productivity.2 Children working on farming and ranching operations are at a particularly high risk of fatal and nonfatal injury.3,4 Injury-related mortality and morbidity rates for the agricultural trade are among the highest of all US industries.5,6 For farming and ranching families, the proximity between the household and the operation makes hazardous exposures nearly unavoidable, even for children not actively engaged in operation-related tasks.7 In response to this issue, a national initiative for the reduction of agricultural injury among children was established.8

Injuries are not distributed randomly among populations at risk. Positive associations between one's prior injuries and risk of subsequent injury have been observed in studies of sports,9,10 occupational,11 and agricultural12,13 injuries and for medically attended injuries among adults and children.1416 Risk also appears to be associated within families; sibling injury was shown to be associated with at least a short-term increase in children's risk of injury.17,18 Individuals' past injuries may increase their vulnerability to future injuries,12 and the association between sibling injuries could be related to periods of heightened familial stress after initial injuries.17,18 Another likely explanation is that certain environments are inherently more hazardous than others and continue to be so even after an injury occurs.

An area that remains to be examined is the association between parents' and children's injuries. A positive association might be expected, given the theories that increased familial stress or sustained hazardous environments or both lead to multiple injuries within households. However, parents' injuries might lead to implementation of preventive measures for their children. Improved understanding of the effect of parents' injury experiences on children's risk may help practitioners identify high-risk populations and exposures as well as effective interventions. The objective of this study was to determine whether an association exists between parents' agricultural injuries and children's risk of injury among farm and ranch households.


This study was based on data from the 1999 Regional Rural Injury Study–II,19 a prospective, population-based cohort study of agricultural households in Minnesota, North Dakota, South Dakota, Nebraska, and Wisconsin. We conducted a nested case-control study of agricultural injuries involving children younger than 20 years. The University of Minnesota Institutional Review Board Human Subjects Committee approved the protocol for this study.


A random sample of 16 000 agricultural operations was generated (n = 3200 from each state) by using the US Department of Agriculture's National Agricultural Statistics Service Master ListFrame of Farming Operations.20 Eligible operations had to have produced $1000 or more of agricultural goods during 1998 and/or have land enrolled in a Conservation Reserve Program, and they had to be actively engaged in farming or ranching as of January 1, 1999. In addition, the operator's household had to be associated with the farm or ranch and include children younger than 20 years in residence as of January 1, 1999.

Of the 16 000 sampled operations, 12 677 were screened successfully, of which 8288 (51.8%) were found ineligible. Of the 4402 (27.5%) households identified as eligible, 3765 (85.5%) participated in the first 6-month interview, and 3677 (83.5%) participated in both full interviews. Demographic and injury data were collected for 16 538 household members, including 8488 children. An agricultural injury was defined as any event resulting from activities or objects associated with the family agricultural operation leading to restriction of normal activities for 4 hours or more, loss of consciousness or awareness or amnesia for any length of time, and/or treatment by a health professional.


Subsequent to introductory mailings, computer-assisted telephone interviews were conducted to obtain informed consent, establish eligibility, and enroll households. As an incentive to participate, all sampled operations were eligible for a drawing that provided at least a 1 in 32 probability of receiving a $100 US savings bond.

Participating households received packets containing study details and materials to facilitate collection of information during each of 2 full-length interviews. The interviews occurred in July 1999 and January 2000, and information was collected about each 6-month recall period of 1999 (January through June and July through December). Nonrespondents for whom eligibility status was not established were sent a 1-page survey to ascertain eligibility of the household. Copies of the materials used, including interview instruments, are available on a Web site.21

The female head of household was the primary respondent for demographic and injury data, whereas the male head of household was preferred for operation exposure information. With parents' permission, children 12 years and older were requested to be their own respondents. The proxy for those who refused, and children younger than 12 years, was the female head of household.


Cases were all agricultural injury events sustained by children younger than 20 years on their own operation during the reporting periods. The control population came from households randomly sampled to achieve a minimum ratio of 3 controls per case on the basis of injury rates from a previous study.3 An incidence-density sampling scheme was used to sample from the agricultural injury-free person-time experience (months) of children younger than 20 years in residence on the operation. One child with at least 1 agricultural injury–free month was selected randomly from each control household. The exposure experience was collected for the month before the month of injury for cases and the month before an agricultural injury–free month for controls, sampled to reflect expected monthly incidence rates.3 This method allowed cases to be sampled as controls for months in which they did not sustain an agricultural injury. The final population included 203 (91.0%) of 223 eligible case events and 755 (97.0%) of 778 sampled eligible controls.


Agricultural injury data also were collected for respondents identified as the children's mothers and fathers. Parents were asked whether they had incurred an agricultural injury before or during the study year. Use of the 6-month recall periods is key for validity of injury reporting;22,23 in this study, this method also was expected to decrease the potential for exposure (parents' injuries) misclassification.

Children were categorized as having a father, a mother, or both parents injured any time before or during the study period. For the analyses to address the temporal nature of the study question, parents' injuries had to have occurred before the case injury month. For controls, parents' injuries had to have occurred before a randomly selected target month. In addition, each parent's injury experiences were combined to reflect their overall injury history through both periods.


Multivariable logistic regression was used to estimate risk of agricultural injury to children, subsequent to parental agricultural injury, while controlling for potentially confounding covariates.24 To determine potential confounders, we used a causal model21 to guide the design and use of directed acyclic graphs, according to the methods in Greenland et al25 and Hernan et al.26 These methods facilitated identification of covariates that may introduce bias if included in the statistical models and reduced potential of overfitting the models. For parent-related covariates, data specific to fathers or mothers were used for the respective analyses of fathers' and mothers' injuries. For analyses of combined parental injury history, mothers' data were included, assuming these factors have a greater effect on children's risk.

Results were adjusted for within-household correlation by using generalized estimating equations.27 To account partially for selection bias, we addressed nonresponse at the household level by weighting observed responses with inverse probabilities of response28 on the basis of characteristics from the Master ListFrame (operation state, operation type, annual revenue by quintile). To account for unknown eligibility for nonrespondents, we estimated probability of eligibility for these same characteristics and used them to adjust weights.29 Parents' injury status was unknown for 4% to 6% of children in each analysis; a separate exposure level was evaluated for these missing data in each model.


Among the 203 cases and 755 controls, agricultural injuries before the study period were reported for 118 (58.1%) and 306 (40.5%) fathers and 63 (31.0%) and 112 (14.8%) mothers, respectively. During the study period, injuries were identified for 18 (8.9%) and 59 (7.8%) fathers and 15 (7.4%) and 29 (3.8%) mothers, respectively.


Case vs control children more often resided in households affiliated with dairy operations (23.2% and 12.8%); controls were more often associated with field, forage, or specialty crop operations or Conservation Reserve Programs (50.9% and 35.0%, respectively) (Table 1). Compared with control households, case households worked more hours per week on their operations: 25.6% vs 11.1% worked 160 hours or more. Nearly 23% of case households had 4 or more children in residence, compared with 12.7% of control households. Control households more often had only 1 child in residence (27.6% vs 16.8%). There were no differences between cases and controls in parents' age or educational status.

Table 1. 
Characteristics of Case and Control Subjects Younger Than 20 Years: 1999 Regional Rural Injury Study–II
Characteristics of Case and Control Subjects Younger Than 20 Years: 1999 Regional Rural Injury Study–II

When controlling for potential confounders, children of fathers who sustained an agricultural injury before the study period had twice the risk of injury as those whose fathers did not report an injury (odds ratio [OR], 2.1; 95% confidence interval [CI], 1.5-3.0) (Table 2). Children whose fathers were injured during the study were no more or less likely to be injured (OR, 1.0; 95% CI, 0.5-1.9) than children whose fathers were not injured. Children whose fathers were injured during both periods had a greater, albeit imprecise, risk of injury than children whose fathers were never injured (OR, 1.8; 95% CI, 0.8-4.1).

Table 2. 
Risk of Children's Agricultural Injury According to Fathers' and Mothers' Injury Histories: 1999 Regional Rural Injury Study–II
Risk of Children's Agricultural Injury According to Fathers' and Mothers' Injury Histories: 1999 Regional Rural Injury Study–II

The risk of injury among children whose mothers were injured before the study period was more than twice that for children whose mothers did not sustain an injury (OR, 2.5; 95% CI, 1.7-3.8) (Table 2). A more modest association was observed for mothers' agricultural injuries during the study (OR, 2.0; 95% CI, 0.9-4.2). The highest risk was observed among children whose mothers reported injuries both before and during the study year (OR, 8.0; 95% CI, 2.5-25.4).

Compared with those with neither parent injured before the study period, increased risks of agricultural injury were observed for children with one (OR, 1.5; 95% CI, 1.0-2.2) or both (OR, 4.2; 95% CI, 2.6-6.9) parents injured (Table 3). Similarly, if both parents were injured during either time period analyzed, the children's risk was also increased (OR, 3.5; 95% CI, 2.2-5.7) compared with children with neither parent ever injured.

Table 3. 
Risk of Children's Agricultural Injury According to Parents' Combined Injury History: 1999 Regional Rural Injury Study–II
Risk of Children's Agricultural Injury According to Parents' Combined Injury History: 1999 Regional Rural Injury Study–II

Although data were too sparse to evaluate all exposures of interest, the effect of parents' injuries on children varied slightly according to sex. For example, the relative risks of injury for boys and girls whose fathers reported an injury before the study period were 1.9 (95% CI, 1.2-3.0) and 2.5 (95% CI, 1.3-4.7), respectively. For those whose mothers reported injuries before the study period, relative risks were 2.4 (95% CI, 1.3-4.2) and 3.2 (95% CI, 1.7-6.0).


Results of this study showed a positive association between parents' history of agricultural injury and children's risk of subsequent injury. Although the underlying reasons for this association cannot be identified from the data, 2 explanations are plausible: (1) certain agricultural environments may be inherently more hazardous than others, regardless of operation type, and (2) behavioral characteristics of some families may lead to increased high-risk exposure. No matter the cause, this association indicates that interventions to prevent childhood injury can be targeted specifically to families with injured parents (ie, parents treated by health professionals can be sentinel events for prevention efforts).

Although it is postulated that prior injury is associated with risk perception30 and that risk perception may lead to improved preventive practices and reduced likelihood of injuries,31 results of this study indicate that parents' experiences with agricultural injuries, alone, do not convey a lower risk to children. Results of studies conducted with other populations support these findings. A recent analysis of self-reported hospital emergency department visits indicated that increased parental emergency department use was associated with increased child emergency department use.32 Further studies showed that parents rarely alter the environment or change safety behaviors to prevent reinjury among children,33,34 and relatively few parents who perceive their agricultural operations as high risk make safety-related changes to mediate this risk.35 These findings may be partly due to parents' beliefs that children's injuries are unpreventable or a normative part of child development.34,3639

Although fathers and mothers may perceive injuries as a normal part of childhood, mothers relay more conservative perceptions of injury,39 which may explain the greater awareness and focus on safety-related behaviors among mothers.40,41 In the current study, mothers' injuries appeared to be associated more strongly with children's risk than fathers' injuries were. It is unknown whether this is influenced by the divergence between mothers' and fathers' perceptions of injury or by distinct gender roles among parents on agricultural operations. The latter could explain the strength of association between mothers' and children's injuries. For example, mothers may have been working on their operation while also caring for children. Supervision by a caregiver engaged in farmwork, compared with that in the home, is associated with increased risk of injury among children.42 The associations observed in the current study could result from the tendency of mothers and children to share environments that potentiate injury risk.

A unique finding was the apparent additive effect of parents' agricultural injuries on children's risk of injury. Children whose parents were both injured before the study were 4 times more likely to be injured, and children whose parents both had a lifetime history of injury (either before or during the study) were 3.5 times more likely to be injured. This finding may be attributable to inherent hazard levels of certain operations. The type of operation, however, was controlled in the multivariable analyses to address this problem, although residual confounding cannot be ruled out. A second rationale for this finding is the possible modeling of parental behaviors by children, particularly pertaining to risk awareness. For example, children are more likely to practice safety precautions, such as using personal protective equipment, when such precautions are modeled by their parents.43 The current finding emphasizes the potential benefit of clinic-based injury prevention measures targeted to parents treated for agricultural injuries.

Limitations of this study should be considered when interpreting the results. For example, some families may be better reporters of injuries than others, which could lead to misinterpretation of the observed associations. Furthermore, because of the self-reported nature of the data, the results are susceptible to information bias. However, the use of 6-month recall periods, which improves injury reporting validity,22,23 was expected to minimize reporting bias for the injuries occurring in the study year. Also, a particular strength of this study was the large number of participating households (3765 [85.5%] of 4402 eligible). Although the high participation rate should reduce the potential for selection bias to affect study results, an attempt to control such bias also was made through adjustments for nonresponse and within-household correlation.

The complexity of agricultural operations limits our ability to fully account for environmental characteristics in multivariable analyses. Agricultural injuries result from multiple and complex causal pathways. Although this study is one of the largest of its kind and contains a substantial amount of detailed data, we could not identify single pathways or circumstances that may explain intrafamilial risk of injury. Despite this limitation, these findings are useful in the context of identifying families at higher risk to whom interventions might be targeted, particularly in a clinical setting. Future studies of the shared physical and social environments contributing to such risk may lead to more effective population-based injury prevention measures.

Back to top
Article Information

Correspondence: Susan G. Gerberich, PhD, Regional Injury Prevention Research Center, Division of Environmental Health Sciences, School of Public Health, University of Minnesota, MMC 807, 420 Delaware St SE, Minneapolis, MN 55455 (gerbe001@umn.edu).

Accepted for Publication: June 15, 2006.

Author Contributions: Dr Gerberich had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of these data analyses. Study concept and design: Carlson, Langner, Alexander, Gurney, Gerberich, Ryan, and Renier. Acquisition of data: Ryan, Renier, and Mongin. Analysis and interpretation of data: Carlson, Langner, Alexander, Gurney, Gerberich, Ryan, Renier, and Mongin. Drafting of the manuscript: Carlson, Langner, Gerberich, and Ryan. Critical revision of the manuscript for important intellectual content: Carlson, Alexander, Gurney, Gerberich, Renier, and Mongin. Statistical analysis: Carlson, Alexander, Gerberich, Ryan, Renier, and Mongin. Obtained funding: Gerberich. Administrative, technical, and material support: Carlson, Gerberich, Ryan, and Renier. Study supervision: Alexander and Gurney.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant R01 CCR514375 from the National Institute for Occupational Safety and Health, Centers for Disease Control and Prevention, US Department of Health and Human Services; a National Institute for Occupational Safety and Health grant T42/CCT510-422 for the Occupational Injury Prevention Research Training Program; the National Occupational Research Agenda Program, Midwest Center for Occupational Health and Safety, University of Minnesota; and the Regional Injury Prevention Research Center and Center for Violence Prevention and Control, Division of Environmental Health Sciences, School of Public Health, University of Minnesota.

Acknowledgment: We thank the US Department of Agriculture's National Agricultural Statistics Service for collaboration on this project and are indebted to the farm and ranch families who shared their time and personal information during participation in the Regional Rural Injury Study–II.

Anderson  RNSmith  BL Deaths: leading causes for 2001 Natl Vital Stat Rep 2003;52 ((9)) 1- 85
Baker  SP The Injury Fact Book.  New York, NY Oxford University Press1992;11- 15
Gerberich  SGGibson  RWFrench  LR  et al.  The Regional Rural Injury Study–I (RRIS-I): A Population-Based Effort. A Report to the Centers for Disease Control and Prevention.  Minneapolis Regional Injury Prevention Center, University of Minnesota1993;
Hard  DMyers  JSnyder  K  et al.  Young workers at risk when working in agricultural production Am J Ind Med 1999;(suppl 1)31- 33
National Safety Council, Injury Facts, 2005-2006 Edition  Itasca, Ill National Safety Council2006;68
 2004 Census of Fatal Occupational Injuries: fatal occupational injuries by industry and event or exposure, all United States, 2004 US Department of Labor, Bureau of Labor Statistics Web site http://www.bls.gov/iif/oshwc/cfoi/cftb0196.pdfAccessed April 11, 2006
Pickett  WBrison  RJBerg  RLZentner  JLinneman  JMarlenga  B Pediatric farm injuries involving non-working children injured by a farm work hazard: five priorities for primary prevention Inj Prev 2005;116- 11
Castillo  DHard  DMyers  JPizatella  TStout  N A national childhood agricultural injury prevention initiative J Agric Saf Health 1998;1 ((special issue)) 183- 191.http://asae.frymulti.com/abstract.asp?aid=15368&t=2Accessed November 15, 2005Article
Emery  CAMeeuwisse  WH Risk factors for groin injuries in hockey Med Sci Sports Exerc 2001;331423- 1433
Hagel  BEFick  GHMeeuwisse  WH Injury risk in men's Canada West University football Am J Epidemiol 2003;157825- 833
Schneider  GABigelow  CAmoroso  PJ Evaluating risk of re-injury among 1214 army airborne soldiers using a stratified survival model Am J Prev Med 2000;18 ((3)(suppl)) 156- 163
Browning  SRTruszczynska  HReed  DMcKnight  RH Agricultural injuries among older Kentucky farmers: the Farm Family Health and Hazard Surveillance Study Am J Ind Med 1998;33341- 353
Low  JMGriffith  GRAlston  CL Australian farm work injuries: incidence, diversity and personal risk factors Aust J Rural Health 1996;4179- 189
Kendrick  DMarsh  P Injury prevention programmes in primary care: a high risk group or a whole population approach? Inj Prev 1997;3170- 175
Madden  CGarrett  JMCole  TBRunge  JWPorter  CQ The urban epidemiology of recurrent injury: beyond age, race, and gender stereotypes Acad Emerg Med 1997;4772- 775
Spady  DWSaunders  DLSchopflocher  DPSvenson  LW Patterns of injury in children: a population-based approach Pediatrics 2004;113522- 529
Johnston  BDGrossman  DCConnell  FAKoepsell  TD High-risk periods for childhood injury among siblings Pediatrics 2000;105562- 568
Johnston  BDGrossman  DCThompson  RS Transient elevation in risk of injury in siblings following injury encounters J Pediatr 2003;14279- 83
Gerberich  SGGibson  RWFrench  LR  et al.  Etiology and Consequences of Injuries Among Children in Farm Households: A Regional Rural Injury Study: Technical Report to the Centers for Disease Control and Prevention.  Minneapolis Regional Injury Prevention Center, University of Minnesota2003;
National Agricultural Statistics Service, US Department of Agriculture Web site http://www.nass.usda.gov/Data_and_Statistics/Quick_Stats/index.aspAccessed August 17, 2006
 Regional Injury Prevention Research Center Web site http://enhs.umn.edu/riprc/riprc.htmlAccessed October 15, 2005
Braun  BLGerberich  SGSidney  S Injury events: utility of self-report in retrospective identification in the USA J Epidemiol Community Health 1994;48604- 605
Gerberich  SGGibson  RWGunderson  PD  et al.  The Olmsted Agricultural Trauma Study (OATS): A Population-Based Effort: A Report to the Centers for Disease Control.  Minneapolis Regional Injury Prevention Center, University of Minnesota1991;
Breslow  NEDay  NE Statistical methods in cancer research: volume II—the design and analysis of cohort studies IARC Sci Publ 1987; ((82)) 1- 406
Greenland  SPearl  JRobins  J Causal diagrams for epidemiologic research Epidemiology 1999;1037- 48
Hernan  MAHernandez-Diaz  SWerler  MMMitchell  AA Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology Am J Epidemiol 2002;155176- 184
Liang  KYZeger  SL Longitudinal data analysis using generalized linear models Biometrika 1986;7313- 22Article
Horvitz  DGThompson  DJ A generalization of sampling without replacement from a finite universe J Am Stat Assoc 1952;47663- 685Article
Mongin  SJ Adjustment for non-response in the Minnesota Nurses Study Division of Environmental and Occupational Health, University of Minnesota Web sitehttp://www1.umn.edu/eoh/NewFiles/researchpdffiles/response.adjustment.1.pdfAccessed October 15, 2005
Glik  DKronenfeld  JJackson  K Predictors of risk perception of childhood injury among parents of preschoolers Health Educ Q 1991;18285- 301
Conrad  PBradshaw  YSLamsudin  RKasniyah  NCostello  C Helmets, injuries, and cultural definitions: motorcycle injury in urban Indonesia Accid Anal Prev 1996;28193- 200
Mistry  RDHoffmann  RGYauck  JSBrousseau  DC Association between parental and childhood emergency department utilization Pediatrics 2005;115e147- e151
Peterson  LBartelstone  JKern  TGillies  R Parents' socialization of children's injury prevention: description and some initial parameters Child Dev 1995;66224- 235
Langley  JSilva  P Childhood accidents: parents' attitudes to prevention Aust Paediatr J 1982;18247- 249
Zentner  JBerg  RLPickett  WMarlenga  B Do parents' perceptions of risks protect children engaged in farm work? Prev Med 2005;40860- 866
Morrongiello  BADayler  L A community-based study of parents' knowledge, attitudes and beliefs related to childhood injuries Can J Public Health 1996;87383- 388
Coffman  SMartin  VPrill  NLangley  B Perceptions, safety behaviors, and learning needs of parents of children brought to an emergency department J Emerg Nurs 1998;24133- 139
Bennett Murphy  LM Adolescent mothers' beliefs about parenting and injury prevention: results of a focus group J Pediatr Health Care 2001;15194- 199
Lewis  TDiLillo  DPeterson  L Parental beliefs regarding developmental benefits of childhood injuries Am J Health Behav 2004;28(suppl 1)S61- S68
Garling  AGarling  T Mothers' anticipation and prevention of unintentional injury to young children in the home J Pediatr Psychol 1995;2023- 36
Garling  AGarling  T Mothers' supervision and perception of young children's risk of unintentional injury in the home J Pediatr Psychol 1993;18105- 114
Pryor  SKCaruth  AKMcCoy  CA Children's injuries in agriculture related events: the effect of supervision on the injury experience Issues Compr Pediatr Nurs 2002;25189- 205
Hu  XWesson  DFParlin  PCChipman  MLSpence  LJ Current bicycle helmet ownership, use, and related factors among school-aged children in metropolitan Toronto Can J Public Health 1994;85121- 124