Physicians and Breastfeeding: Beliefs, Knowledge, Self-efficacy and Counselling Practices

A B S T R A C T A pilot-tested questionnaire was mailed to 325 obstetricians, pediatricians, family practitioners and general practitioners of a British Columbian maternity hospital to measure aspects relating to physicians attitudes toward breastfeeding counselling. Response rate was 67.3%. The measures of selfefficacy, knowledge and beliefs were added to a regression model containing measures of gender, specialty, years in practice and personal or spousal breastfeeding experience to determine whether additional variance in counselling behaviour could be accounted for. Physicians attempted to convince women to breastfeed if: 1) they believed in the immune properties of breastmilk (OR = 1.23, SE = 0.07) and 2) they were confident in their own breastfeeding counselling (OR = 1.88, SE = 0.36). Likewise, encouraging women to continue breastfeeding in the face of breastfeeding problems was related to confidence in breastfeeding counselling (OR = 1.22, SE = 0.10) and belief in the immune properties of breastmilk (OR = 2.83, SE = 0.45). A B R É G É Différents aspects relatifs à l attitude des médecins par rapport à l allaitement maternel ont été déterminés par un questionnaire préalablement vérifié qui a été expédié à 325 obstétriciens, pédiatres, médecins de famille et généralistes affiliés à une maternité de la Colombie Brittanique. Le taux de réponse était de 67,3 %. Les paramètres de la confiance en soi, des connaissances et des croyances ont été ajoutés au modèle de régression en tenant compte des paramètres du genre, de la specialité, du nombre d années d expérience et de l expérience de l allaitement, personnelle ou de l épouse, pour déterminer si certaines variances pourraient être attribuées au comportement de counselling. Les médecins ont tenté de convaincre les femmes d allaiter lorsqu ils avaient confiance dans les qualités immunologiques du lait maternel (RRE = 1,23, ET = 0,07) ainsi que dans leur aptitude en counselling de l allaitement (RRE = 1,88, ET = 0,36). Pour les mêmes raisons, les médecins ont encouragé les femmes à poursuivre l allaitement si des problèmes surgissaient (respectivement RRE = 2,83, ET = 0,45 et RRE = 1,22, ET = 0,10). Physicians and Breastfeeding: Beliefs, Knowledge, Self-efficacy and Counselling Practices Our understanding of the importance of breastfeeding has increased immensely, 1-9 however, breastfeeding rates remain far from ideal. Fewer than one in four Canadian infants are exclusively breastfed in hospital, and the high rates of premature termination of breastfeeding have not changed over the past 20 years. 10 Reversing the problems of breastfeeding failure and bottle-feeding choice depends on physicians who are skilled in breastfeeding support and management. 11-13 The following study aimed to assist an urban British Columbian hospital intending to develop breastfeeding education for its physicians. Studies have investigated whether physicians attitudes and specialties affect their breastfeeding counselling behaviours. 14-17 More recent studies have added physicians beliefs and selfconfidence to the list of factors affecting counselling behaviours; 18,19 however, most studies have looked at only one specialty or residency program at a time. Green and Kreuter 20 recommend that continuing education to physicians be preceded by an analysis of motivators for and obstacles to the desired change in behaviour. Their PRECEDE model 20 allows conceptualization of interactive factors that influence physicians breastfeeding counselling and training of childbearing Department of Health Care and Epidemiology, Institute of Health Promotion Research, and Department of Family Practice, University of British Columbia Financial support was provided by the Department of Family Practice, University of British Columbia and by the British Columbia Medical Services Foundation. Correspondence: Maria Burglehaus, 2202 London St., New Westminster, BC, V3M 3G2, Tel: 604-540-7279. Reprint requests: Dr. Lorie Smith, 305-2730 Commercial Dr., Vancouver, BC, V5N 5P4 Maria J. Burglehaus, MSc, RDN, Lorie A. Smith, MD, CCFP, Samuel B. Sheps, MD, MSc, FRCPC, Lawrence W. Green, DrPH women, namely predisposing, enabling, and reinforcing factors. These categories are convenient in that they group the more specific behavioural influences such as knowledge, attitudes and beliefs, skills, incentives, and rewards under broader rubrics according to the measures that might be used to change behaviour. Green et al. 21 offer this classification as a useful conceptual framework for analyzing physician behaviour and planning interventions to change it. They specify that the three domains are not mutually independent. Predisposing factors refer to antecedents to behaviour that provide the rationale or motivation for a behaviour. These factors are within the realm of psychology and represent the cognitive and affective dimensions of knowing, feeling, believing, valuing, and having self-efficacy or confidence. 20 Enabling factors are the antecedents to behaviour that allow a motivation to be realized. They include skills, resources or barriers. Reinforcing factors are factors subsequent to a behaviour that provide the continuing reward or incentive for the behaviour and contribute to its persistence or repetition. Reinforcing factors include rewards, feedback and colleague support. 20 Our study sought specifically to determine which predisposing factors to target in order to improve physicians willingness, motivation and ability in breastfeeding counselling. Predisposing factors include knowledge, attitudes, beliefs, values and perceived needs and abilities that might motivate physicians counselling and supportive practices. Demographic factors including age, physician specialty and gender were included in our study since specific subgroups may demonstrate needs for different interventions. NOVEMBER DECEMBER 1997 CANADIAN JOURNAL OF PUBLIC HEALTH 383

The aims of the investigation were: 1) to provide baseline data on attitudes, beliefs, knowledge and self-efficacy concerning breastfeeding counselling in a generalizable population of physicians; and 2) to estimate the relative amount of variance in counselling practices accounted for by measures including cognitive factors, gender, specialty, years in practice and personal or spousal breastfeeding experience. This study aimed not only to further investigate the variety of factors affecting physicians breastfeeding counselling behaviours but also to expand the literature by including four specialties and several years in practice. METHODS An urban hospital s mailing list was used to identify the 325 general practitioners, family practitioners, obstetricians and pediatricians providing care to the maternity patients of the hospital. Family practitioners are physicians who have completed the required training to become registered with the College of Family Physicians. We developed an eight-page questionnaire which consisted of 40 closed-ended questions. It was based on factors cited in the literature concerning breastfeeding counselling and management and on items from questionnaires by Lawrence 15 and by Michelman et al. 18 Our questionnaire items elicited beliefs, attitudes, self-efficacy and knowledge and the extent to which physicians provide counselling concerning breastfeeding. Knowledge questions concerned the effect of domperidone on breastmilk supply, correct positioning of the infant at the breast and advice about starting solids. Additional questions included asking physicians about their practices surrounding supplementation with water or formula and whether they had read the WHO-UNICEF Ten Steps to Successful Breastfeeding. In order to pilot-test the questionnaire, it was mailed to 14 physicians providing obstetrical care at a local community hospital. The physicians were asked to complete the questionnaire and include any comments they had regarding the questionnaire s clarity or the time involved to complete it. Eleven questionnaires were returned. The only change resulting from the pilot test involved moving one question to the end of the questionnaire. The amended questionnaire was mailed to the physicians offices (N = 325), followed by a reminder postcard, a second mailing, follow-up phone calls and a third mailing. Thirteen questionnaires were returned in cases where physicians no longer provided maternity care or had moved out of the area. Data collection was carried out from June to November 1994. Reliability testing was limited to testing for internal consistency using Cronbach s alpha, the results of which are presented in the results section. Correlational analyses were performed to detect effects of specialty, years in practice, gender, ethnicity and personal breastfeeding experience on 1) counselling behaviours and 2) predisposing factors (beliefs, knowledge and self-efficacy). The questionnaire items for counselling behaviours, beliefs, and self-efficacy were treated as interval measures. The Pearson correlation was used as this test statistic is reasonably robust to departures from normality. The point biserial correlation was used in testing for association between any interval variables against a dichotomous variable. Knowledge questions were dichotomous and non-parametric tests were used for correlation analyses. Both correlational analyses and backwards stepwise logistic regression were performed to investigate whether beliefs, knowledge and self-efficacy were associated with physician counselling behaviour. The regression model controlled for physician factors such as specialty and years in practice. The five-point Likert scales of the individual counselling behaviour items limit their variability. Logistic regression was used as this method is compatible when the dependent variable displays limited variability. In logistic regression the dependent variable must be dichotomized. Thus, the counselling behaviour variables were dichotomized with usually and always grouped together and sometimes, infrequently and never grouped together. This classification scheme reflects the goal of an education intervention: to target the factors that are associated with lower rates of discussing and advocating breastfeeding to patients, implied by the terms sometimes, infrequently and never. The regression analyses were carried out using stepwise forward logistic regression. In each regression the following control variables were entered first: years in practice, gender, specialty, personal or spousal breastfeeding experience, and caucasian ethnicity. Years in practice was a continuous variable, while gender, breastfeeding experience and caucasian ethnicity were dichotomous. Dummy variables were used for specialty. The analyses involved three sets of regressions. The first set, consisting of nine regressions, regressed each of the three components (knowledge, then beliefs, and finally self-efficacy) on each of the three counselling variables. Next, three regressions were run, one for each of the counselling variables in which each of the components were entered stepwise; knowledge, then beliefs and finally self-efficacy. The final set of three equations entered in any components that had been determined significant in the preceding sets of regressions. RESULTS Response rate was 67.3% (210/325). Response rate for physicians tends to be lower than rates for other populations; our response rate was similar to those of recent physician studies. 14-17 Response rate varied between the various groups of the study. Female physicians were much more likely to respond than were male physicians; response rates of 86% and 57% respectively. Family practitioners were more likely to respond than were physicians in other specialties. Finally, physicians with fewer years in practice were more likely to respond. The mean years in practice was 13 (Table I). There were no significant differences among the specialties regarding gender distribution nor with respect to spousal or personal breastfeeding experience. Reliability analyses using Cronbach s alpha were performed on questionnaire items used to elicit information about counselling behaviour, beliefs, knowledge and self-efficacy. We explored the possibility of making composite scores of the questions within categories. In most cases, relia- 384 REVUE CANADIENNE DE SANTÉ PUBLIQUE VOLUME 88, NO. 6

bility was low (i.e., alpha values did not approach 0.8) indicating the questions were measuring different aspects of their TABLE I Response Rates Responded Sampled Percent Gender Male 115 202 56.9 Female 95 110 86.4 Specialty Obstetricians 24 37 64.9 Pediatricians 20 31 64.5 Family practitioners* 69 91 75.8 General practitioners 97 153 63.4 Years in Practice Less than 13 years 96 114 84.2 13 or more years 114 198 57.6 TOTAL 210 312 67.3 * Family practitioners refer to those physicians who have completed the required training to become registered with the College of Family Physicians. TABLE II Occurrence of Counselling by Counselling Behaviours Categories Discuss Attempt to Encourage Breastfeeding Convince Mothers Continued Prenatally to Breastfeed Breastfeeding n=197* n=208 n =208 Score Category N Percent N Percent N Percent 5 Always 129 65 86 41 95 46 4 Usually 51 26 96 46 100 48 3 Sometimes 12 6 22 11 10 5 2 Infrequently 5 3 4 2 3 1 Mean 4.5 4.3 4.4 Standard dev. 0.72 0.73 0.65 Note: None of the responses were in the Never category. * Pediatricians tended to indicate this question was not applicable TABLE III Final Regression Models Counselling Behaviour Variable β p Partial R I. Discuss breastfeeding Diagram 1.16 0.074 0.12 in the prenatal period (n=178) Model Chi-Square = 2.9 Significance = 0.086 II. Convince women to Children -0.92 0.094-0.08 breastfeed when they Pediatrician -1.38 0.055-0.11 intend to bottle-feed (n=188) Immune 0.63 0.080 0.09 Self-efficacy 0.21 0.004 0.22 Model Chi-Square = 8.9 Significance = 0.003 III. Encourage women to Immune 1.04 0.021 0.21 continue breastfeeding Self-efficacy 0.20 0.038 0.17 in the face of breastfeeding problems (n=189) Model Chi-Square = 4.6 Significance = 0.032 1. Each of the models controlled for ethnicity, personal or spousal breastfeeding, gender, years in practice and specialty. 2. Diagram is the knowledge question requiring physicians to select the diagram that showed correct positioning of the infant at the breast: 1 = correct choice; 0 = incorrect choice 3. Children: 1 = physician or spouse breastfed a child; 0 = neither physician nor spouse breastfed a child 4. Pediatrician : 1 = specialty is pediatrics; 0 = specialty is not pediatrics. 5. Immune: believe that exclusively breastfed infants have fewer gastrointestinal infections, respiratory illnesses, eczema and/or allergic reactions than formula-fed infants. Five-point Likert scale from strongly disagree (score of 1) to strongly agree (score of 5). constructs. The self-efficacy questions taken from the study by Michelman et al., 18 however, had an acceptable reliability in this study (alpha = 0.79) and were combined to make a composite score. Counselling behaviour Ninety percent of the respondents reported always or usually discussing breastfeeding with their patients prenatally (65% always do) and 88% reported always or usually attempting to convince mothers to breastfeed if they intend to bottle-feed (41% always do). In the face of breastfeeding problems, 94% of physicians reportedly always or usually encourage patients to continue breastfeeding (46% always do). The measures of self-confidence, knowledge and beliefs were added to a regression model containing measures of gender, specialty, years in practice and personal or spousal breastfeeding experience to determine whether additional variance in the counselling behaviour could be accounted for. Physicians attempted to convince women to breastfeed if they believed in the immune properties of breastmilk (OR = 1.23, SE = 0.07) and were confident in their own breastfeeding counselling (OR = 1.88, SE = 0.36). Likewise, encouraging women to continue breastfeeding in the face of breastfeeding problems was related to confidence in breastfeeding counselling (OR = 1.22, SE = 0.10) and belief in the immune properties of breastmilk (OR = 2.83, SE = 0.45). Correlational data showed weak to moderate associations between the reported cognitive factors and counselling behaviours. Knowledge of how to position the infant at the breast was associated with whether physicians reported discussing breastfeeding with patients (p<0.05). Physicians who disagreed with the statement that cereals help the baby sleep through the night were more likely to encourage women to continue breastfeeding in the face of breastfeeding problems. Self-efficacy, beliefs and knowledge Female physicians expressed greater selfefficacy in counselling the mother about breastfeeding problems and in positioning the infant at the breast (p < 0.001) (Table III). Both male and female physicians whose children were breastfed expressed NOVEMBER DECEMBER 1997 CANADIAN JOURNAL OF PUBLIC HEALTH 385

greater self-efficacy in counselling about breastfeeding problems and in positioning the baby at the breast (females r = 0.44, p < 0.001; males r= 0.30, p< 0.01). There were no significant differences across specialty regarding self-efficacy. However, there were significant differences across specialty regarding beliefs in the benefits of breastfeeding. While family practitioners more strongly believed that breastfeeding provides adequate nutrition than did physicians in other specialties (r=0.29, p<0.01), general practitioners were more likely to be neutral or disagree (r=-0.20, p<0.01) with this belief. Pediatricians were more likely to disagree or take a neutral stand concerning the belief that breastfeeding protects the infant from illnesses (r=-0.17, p<0.01). Knowledge deficits were highlighted in this study. Twenty-four percent of physicians felt that adding cereals to the infant s diet helps the infant sleep through the night. Pediatricians were less likely than were other physicians to answer that cereals help the infant sleep through the night (p<0.01), while general practitioners were more likely to answer that cereals help (p<0.01). Family practitioners were more aware than were other physicians of domperidone as a means to increase breastmilk supply (p<0.05), while obstetricians were less aware of the use of domperidone for this purpose (p<0.05). When asked to select from two pictures the one showing correct positioning of an infant at the breast, 12% of physicians chose the wrong picture. There were no significant differences across specialty for this question. However, female physicians scored higher on each of these three knowledge questions (p<0.01). Physicians felt that supplementation is indicated for twins, infants with a difficult latch (suckling technique), and in cases requiring an emergency C-section. These are not, however, medical indications for supplementation (Table IV). Only 34% of the physicians had read the WHO - UNICEF Ten Steps for Successful Breastfeeding, a policy that hospitals are to promote to their health professionals in efforts to improve practices and support for breastfeeding. DISCUSSION The following limitations of the current study should be noted. First, the higher response rates from female physicians, Caucasian and/or family physicians reduce the generalizability of the findings. It is likely that the nonrespondents represent physicians who are less interested and/or less skilled in breastfeeding counselling. Specifically, the counselling behaviours and the belief, self-efficacy, and knowledge scores may be specific to the respondents. Thus any bias will likely overestimate breastfeeding support at the hospital as the factors predicting response to the questionnaire (particularly being female and being a family practitioner) are for the most part also associated with better levels of counselling practice, knowledge, self-efficacy and beliefs. Second, self-reports of breastfeeding counselling may bias the results, most likely in the direction of overestimation. Additionally, the survey was within one geographic area. The results may not be generalizable to other physicians providing maternity and infant care. Medical training in the past has had gaps in the area of general preventive medicine and patient education. 21 Training for breastfeeding counselling was found to be inadequate by 63% of the physicians in the 1985 study by Reames. 16 Furthermore, evidence suggests that breastfeeding education to physicians has not improved since 1985. Years in practice failed to demonstrate a TABLE IV Mean Self-efficacy Scores by Gender Male Female Mean Standard Deviation Mean Standard Deviation Self-efficacy composite score 13.83 3.74 16.69 2.90 (t= -6.2, p < 0.001) Scale for self-efficacy composite score is 0 to 20 TABLE V Reports of Indications for Supplementation* n No Water Formula Water & Formula Twins 185 148 16 20 1 Emergency C-Section 185 141 19 21 4 Difficult latch 188 144 10 31 3 * Select responses to the question Which of the following are indications for supplementation with non breastmilk substances (formula/sterile water)? Please answer independently for each indication. significant influence on counselling behaviour in our study and in previous studies of physicians counselling behaviour concerning breastfeeding. 18,19 Apparently, both older and younger physicians in the current study have similar levels of breastfeeding counselling. Self-reports of refraining to counsel or advocate about breastfeeding are as common among the recently trained physicians as they were among physicians trained in an era less supportive of breastfeeding. In the 1993 study by Goldstein and Freed, 19 67% of family practice residents found their training in breastfeeding counselling inadequate and many reported they have counselled in situations in which they admitted they were not prepared. Goldstein and Freed 19 also found that having breastfed a child was not predictive of counselling about breastfeeding; 24% of the total sample of physicians were males with a spouse who had breastfed and 13% were females who had breastfed. Our study reproduced this finding: 29% of the physicians were males with spousal breastfeeding experience and 32% were females with personal breastfeeding experience. The experience of breastfeeding a child does not appear to provide the confidence necessary to counsel patients; this confidence must be developed through training in breastfeeding counselling or through rewarding experiences in doing it successfully. Low self-efficacy among physicians, a feeling of unpreparedness to counsel 386 REVUE CANADIENNE DE SANTÉ PUBLIQUE VOLUME 88, NO. 6

patients about their lifestyles, may explain most of their diffidence to do so. Like Goldstein and Freed, 19 we found that selfefficacy was important in accounting for variance in breastfeeding counselling behaviours. However, our results suggest that this relationship between lower selfefficacy and lower levels of counselling is consistent across a wide range of years in practice and four areas of specialty. We looked at factors associated with self-efficacy. Our study supports the finding by Goldstein and Freed 19 that while personal or spousal breastfeeding experience was associated with increased selfefficacy ratings (females r=0.44, p=0.000; males r=0.30, p=0.001), women physicians who had breastfed a child reported greater self-efficacy than did male physicians whose spouses had breastfed a child. It is important to remember, however, that personal or spousal breastfeeding experience does not appear to predict physician s breastfeeding counselling behaviours. While research shows evidence of greater self-efficacy in general preventive counselling among more recently trained physicians, 20 our study showed that recent graduates did not have greater self-efficacy in breastfeeding counselling than did physicians with greater years in practice. This supports findings that current medical training continues to be inadequate regarding training in breastfeeding. In summary, our findings suggest low levels of self-efficacy regarding breastfeeding counselling among physicians practicing in four broad specialties. Moreover, we demonstrate the prevalence of erroneous advice, lack of interest and misinformation regarding breastfeeding. Given evidence that physicians can influence both breastfeeding initiation and duration, 11-13 our results are disappointing for the locality (urban British Columbia), but suggest a possibility of increasing rates of counselling by strengthening the self-confidence of physicians in counselling about breastfeeding through highly specific and short-term training sessions. Despite more than a decade of research identifying the need for breastfeeding education to physicians, little if any education has been provided. Revision of medical school curricula and development of continuing education are necessary to improve physicians interest in breastfeeding and to provide specific skills that will increase physicians competence and self-efficacy in counselling patients about breastfeeding. The methods found effective in systematic reviews of over 100 studies of continuing education employing the PRECEDE model as a framework for meta-analysis of studies, must include strategies directed at enabling skills and reinforcing support, in addition to the predisposing selfconfidence factor noted here. 22-24 ACKNOWLEDGEMENTS We would like to express our sincere gratitude for financial support received from the Department of Family Practice, University of British Columbia and from the British Columbia Medical Services Foundation. REFERENCES 1. Adebonojo FO. Artificial vs. breastfeeding: Relation to infant health in a middle class American community. Clin Pediatr 1972;11:25-29. 2. Aloia JF, Cohn SH, Vaswani A, et al. Risk factors for post-menopausal osteoporosis. Am J Med 1985;78:95-100. 3. Byers T, Graham S, Rzepka T. Lactation and breast cancer. Evidence for a negative association in premenopausal women. Am J Epidemiol 1985;121:661-74. 4. Canadian Paediatric Society Nutrition Committee. Meeting the iron needs of infants and young children: An update. CMAJ 1991;144:1451-54. 5. Cunningham AS, Jelliffe DB, Jelliffe EFP. Breastfeeding and health in the 1980s: A global epidemiologic review. J Pediatr 1991;118:659-65. 6. Damus K, Pakter J, Krongard E, et al. Postnatal medical and epidemiological risk factors for the sudden infant death syndrome. In: Harper RM, Soffman HJ (Eds.), Sudden Infant Death Syndrome: Risk Factors and Basic Mechanisms. New York: PMA Publications, 1988;187-201. 7. Davis MK, Savitz D, Graubard BI. Infant feeding and childhood cancer. Lancet 1988;2:365-68. 8. Frank JW, Newman J. Breast-feeding in a polluted world: Uncertain risks, clear benefits. CMAJ 1993;149(1):33-37. 9. Layde PM, Webster LA, Boughman AL, et al. The independent associations of parity, age at first full term pregnancy and duration of breastfeeding with the risk of breast cancer. J Clin Epidemiol 1989;42:963-73. 10. Avard D, Hanvey L. The Health of Canada s Children: A CICH Profile. Ottawa: Canadian Institute of Child Health, 1989. 11. Labbok MH, Simon SR. A community study of a decade of in-hospital breastfeeding: Implications for breastfeeding promotion. Am J Prev Med 1988;4(2):62-67. 12. Winikoff B, Meyers D, Laukaran V, Stone R. Overcoming obstacles to breastfeeding in a large municipal hospital: Applications of lessons learned. Pediatrics 1987;80:423-33. 13. Rajan L. The contribution of professional support, information and consistent correct advice for successful breastfeeding. Midwifery 1993;9(4):197-209. 14. Hollen BK. Attitudes and practices of physicians concerning breastfeeding and its management. Environmental Child Health 1976;22:288-93. 15. Lawrence RA. Practices and attitudes towards breast-feeding among medical professionals. Pediatrics 1982;70:912-20. 16. Reames ES. Opinions of physicians and hospitals of current breast-feeding recommendations. J Am Diet Assoc 1985;85:79-80. 17. Lowe T. Breastfeeding: Attitudes and knowledge of health professionals. Austral Fam Phys 1990;19:392, 395-96, 398. 18. Michelman DF, Faden R, Carlson Gielen A, Buxton KS. Pediatricians and breastfeeding promotion: Attitudes, beliefs, and practices. Am J Health Prom 1990;4:181-86. 19. Goldstein AO, Freed GL. Breast-feeding counseling practices of family practice residents. Fam Med 1993;25:524-29. 20. Green LW, Kreuter MW. Health Promotion Planning: An Educational and Environmental Approach. Toronto: Mayfield Publishing Company, 1991. 21. Green LW, Eriksen MP, Schor EL. Preventive practices by physicians: Behavioral determinants and potential interventions. Am J Prev Med 1988;4(4):S101-7. 22. Davis DA, Thomson MA, Oxman AD, Haynes RB. Evidence for the effectiveness of CME: A review of 50 randomized controlled trials. JAMA 1992;267:1111-17. 23. Davis DA, Thomson MA, Oxman AD, Haynes RB. Changing physician performance: A systematic review of the effect of continuing medical education strategies. JAMA 1995;274:700-5. 24. Oxman AD, Thomson MA, Davis DA, Haynes RB. No magic bullets: A systematic review of 102 trials of interventions to improve professional practice. CMAJ 1995;153:1423-31. Received: November 29, 1996 Accepted: July 16, 1997 NOVEMBER DECEMBER 1997 CANADIAN JOURNAL OF PUBLIC HEALTH 387