Prevalence and determinants of tuberculin reactivity among physicians in Edmonton, Canada:

Transcription

1 International Epidemiological Association 2001 Printed in Great Britain International Journal of Epidemiology 2001;30: Prevalence and determinants of tuberculin reactivity among physicians in Edmonton, Canada: Sabrina S Plitt, a,b Colin L Soskolne, a E Anne Fanning c and Stephen C Newman a,d Background Health care workers (HCW) have historically borne a heavy burden of tuberculosis (TB) infection and disease. Unfortunately, physicians are rarely included in HCW surveys of tuberculin exposure and infection. Methods Results Conclusion The prevalence and risk factors for tuberculin reactivity were determined for a sample of the 1732 licensed physicians in Edmonton. Stratified random sampling was used to select 554 specialists and 219 general practitioners. These physicians were contacted by means of an introductory letter and a follow-up telephone call to solicit participation. All eligible physicians were asked to complete a questionnaire and those with either no recorded positive tuberculin test or a previously negative result were two-step tuberculin skin tested. In total, 560 physicians (72.4 %) participated in the study. The overall tuberculin reactivity for this population was 45.9%. Using logistic regression analysis, we determined that risk factors for reactivity were aged over 45 years, of foreignbirth, previous Bacillus Calmette-Guérin (BCG) vaccination, foreign practice experience, and being a respiratory medicine specialist. The prevalence of tuberculin reactivity among physicians is considerably higher than estimates for the general Canadian population. This observed excess risk may be associated with factors linked to their medical practice. The high participation rate suggests physician willingness to participate in this type of research, and emphasizes the need to include them in routine HCW surveillance. Keywords BCG, tuberculosis, health care workers, physicians, epidemiology, tuberculin skin tests Accepted 16 August 2000 Historically, health care workers (HCW) have carried a higher burden of tuberculosis (TB) infection than that of the general public. A recent multicentre study of Canadian HCW showed that 38.7% of the 4651 participants were tuberculin reactors; 5% of these reactions were recent conversions (i.e. from negative to positive tuberculin reactivity within 2 years). 1 Physicians were not included as participants in this study. A survey of Canadian hospitals estimated the annual HCW conversion rate to be 1.7%, 2 a rate considerably higher than the annual infection a Department of Public Health Sciences, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada. b Department of Epidemiology, School of Hygiene and Public Health, Johns Hopkins University, 615 N. Wolfe Street, Baltimore, Maryland, USA. c Division of Infectious Diseases, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada. d Department of Psychiatry, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada. rate of the general public estimated to range from 0.02% to 0.08%. 3 Holton et al. s 1997 paper examining TB infection control programmes in Canadian hospitals estimated that physicians had a significantly higher tuberculin skin test (TST) conversion rate than that of any other occupational group. 2 The risk of physician infection may be higher than that of other HCW because in Canada they are often the point of first contact with the health care system. The large majority of patients with active TB are diagnosed upon presentation to their family doctor with symptoms. For example, in Alberta, in 1994, this was the case for 69% of the 178 TB cases. 4 Health care workers, including physicians, are at high risk of exposure to patients with active TB because of delayed diagnosis and appropriate isolation of infectious patients, and aerosol generating procedures such as bronchoscopy, gastroscopy, or post-mortem examination performed on unsuspected cases. 5 Despite this potentially high exposure risk, physicians have been complacent about their own risk of TB exposure and disease. 6,7 1022

2 TUBERCULIN REACTIVITY AMONG PHYSICIANS 1023 Although there have been many studies of TB infection in HCW, as measured by the prevalence of tuberculin reactivity and active disease, physicians are often either not included, or they comprise only a small percentage of the total population examined. Thus, the representativeness and generalizability of these studies to physicians is poor. Four studies were identified which were directed specifically at physicians and at physiciansin-training. 6 9 Barrett-Connor in 1979, gathered data on 4000 Californian physicians by mailed questionnaire. In this study, she reported that age-specific TB infection among physicians was at least twice that of the general population of the United States. 6 Geiseler et al. in 1982, contacted 4575 medical graduates of the Illinois College of Medicine from and found 47% of the physicians who had never had active TB considered themselves tuberculin positive. 7 More recently in 1994, Fraser et al. reported 24.5% of 351 physicians from an American university-associated tertiary care hospital were skin test positive currently or by history. 8 A 1996 paper examined 284 physicians (resident, attending and associate) at an American teaching hospital where they found 13% of resident and attending (i.e. full-time) physicians and 20% of associate (i.e. part-time) physicians to be tuberculin reactors. 9 Recent studies demonstrating high tuberculin reactivity among HCW in Canada prompted an examination of Canadian physicians. 1,10,11 In order to determine a baseline prevalence of tuberculin reactivity (defined as an induration size of 10 mm or greater) and the risk factors associated with this reactivity among physicians, we conducted a cross-sectional study of tuberculin reactors among practicing physicians of the health catchment area located in and around the city of Edmonton, Alberta, Canada. The average rate of TB in the province of Alberta from was 6.9 cases per The city of Edmonton has a rate twice that of the provincial rate. Tuberculosis rates are four times higher among Aboriginal people, two times higher among foreign-born, and increase with increasing age. 12 Materials and Methods Data collection The study population was selected from the July 1996 Alberta College of Physicians and Surgeons physician directory, a roster of all registered physicians for the province of Alberta. There were 1782 licensed physicians in the Edmonton region, the focus of this study. In order to ensure adequate representation of specialties in the sample, the physicians in the Edmonton region were initially stratified by specialty. Stratified random sampling then was employed to select 1:2 specialists (n = 554) and 1:4 general practitioners (n = 219) for study. Thus, 773 physicians (i.e. 43.4% of all physicians in the region) were contacted by letter and two follow-up telephone calls to solicit their participation. Physician eligibility was based on completion of post-graduate medical training and not having been retired for more than one year. Those physicians considered ineligible were replaced by sequential random sampling by an eligible counterpart from the same specialty stratum. Of the 773 physicians selected as potential participants, 560 (i.e. 72.4%) participated. A questionnaire was administered to all participants. The questionnaire included information on personal demographics, country of birth, history of tuberculin skin testing, past tuberculosis infection, known tuberculosis exposures, BCG vaccination history, place of training, year of graduation from medical school, travel history, and the physician s current patient population, i.e. percentage of Aboriginal patients, percentage of elderly patients (over 70 years of age), and percentage of foreign-born patients. In addition, non-reactors by history who did not report annual skin testing underwent two-step testing. A total of 189 physicians claimed to have had a previously positive test result. However, documentation was available for only 34 physicians (i.e. 18%). Therefore, physicians with no documentation were contacted by the study clinician (AF) and offered repeat skin testing in order to determine the validity of their reactor status. Of these, 35 (i.e. 18.5%) agreed to re-testing and 26 (i.e. 74.3%) were positive. Tuberculin skin testing All skin testing was done by the Mantoux method. The dose of five tuberculin units of purified protein derivative (PPD) was injected intracutaneously into the volar surface of the forearm. Purified protein derivative (Tubersol) from Connaught Laboratories (Toronto, Ontario, Canada) was used. Skin tests were read hours later; the transverse diameter of the induration was measured using calipers and recorded in millimetres. An induration of 10 mm or larger was considered to be a positive test. 5 Four trained personnel administered and read the tuberculin skin tests. For consistency, one member of the study team did the majority of the testing (LZ) and another the reading (SP). In order to ensure reliable readings, physician self-reading of PPD test results was discouraged and only accepted in circumstances when the physician could not be met by one of the study personnel, which occurred in 17.3% of readings. However, only 8.8% of the physicians who had a two-step test did not have either of their tests read by study personnel. Any physician with a previous negative skin test who was a reactor during the course of this study was advised in writing to have a chest X-ray, and submit sputum for culture. The physician was advised to consider preventative drug therapy after TB was ruled out by inquiry, chest X-ray and sputum culture. Statistical analyses All statistical analyses of the data were performed using standard statistical software Associations between the tuberculin skin test status and potential risk factors were assessed using χ 2 significance testing and the calculation of odds ratios (OR) with 95% confidence intervals (CI). The χ 2 for linear trend statistic was used to assess associations between the outcome and ordinal variables. Logistic regression methods were used to determine the risk factors associated with tuberculin reactivity, and to adjust for confounding factors. Interactions between the variables were examined. The goodness-of-fit of this final regression model was checked using the Hosmer-Lemeshow test statistic. 16 Ethical considerations Ethics approval for this research was obtained from the University of Alberta Ethics Review Board. In accordance with

3 1024 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY their approval, all participants signed an informed consent form before participating in the study. Results Descriptive Selected characteristics of the study population are presented in Table 1. The mean age of the population was 47.5 years, 76.4% were male, 65.1% were born in Canada, and 44% had previously been BCG vaccinated. Approximately a quarter of the population had either travelled (11.5%) or practised medicine (11.1%) in a country with mid- to high-tb burden (defined as any country, not including North American and Western European Countries, Australia and New Zealand). Almost twothirds of the physicians recalled being in contact with an active TB case. Of the 560 participating physicians, 56 (10%) did not complete a questionnaire and 12 (1.6%) received only the first test of the required two-step skin test. There was no difference based on gender, location of medical school (i.e. foreign and Canadian), years since medical school graduation, nor on the proportion of specialists to non-specialists between those who Table 1 Selected characteristics of the study population (n = 560): physicians, , Edmonton, Alberta, Canada a Gender, no. (%) Male 428 (76.4) Female 132 (23.6) Age (years), mean (SD) 47.5 (9.8) BCG, b no. (%) Yes 216 (44.0) No 275 (56.0) Time since BCG (years), mean (SD) 33.7 (8.8) Age at BCG (years), mean (SD) 16.8 (8.5) Country of birth, no. (%) Canada 324 (65.1) Other 174 (34.9) Location of medical school, no. (%) Canada 427 (76.2) Other 133 (23.8) Years since graduation, mean (SD) 22.5 (9.9) Known exposure with an active TB c case, no. (%) Yes 321 (65.2) No 171 (34.8) Foreign travel for longer than a month, no. (%) d None 388 (77.4) Travel only 57 (11.5) Travel and medical practice 55 (11.1) NB: a The variables of age, BCG, country of birth, known exposure, and foreign travel/practice are not based on the total study population size (n = 560) owing to incomplete questionnaire information, or questionnaires that were never completed (n = 56). b Bacillus Calmette-Guérin. c Tuberculosis. d i.e. In a country other than North American and Western European countries, Australia and New Zealand. did and did not complete the questionnaire, or between those who had a single-step test and a two-step skin test. In addition, only 5.1% of all of those who had two-step skin testing were negative on the first step and positive on the second step (i.e. boosting). Therefore, the assumption was made that none of the participants without a second skin test would have boosted and they were considered non-reactors. From the Alberta College of Physicians and Surgeons directory, information on the non-participants was obtained including specialty, school of medicine, and year of graduation. This allowed us to determine several significant differences between the participants and the non-participants (i.e = 213). The non-participants were more likely to have attended medical school in a foreign country (P = 0.03), to have graduated from medical school in an earlier year (P = 0.002), and they were more likely to be male (P = 0.04). No difference was found in the numbers of specialists and non-specialists between the two groups (P = 0.4). Just under half of the physicians examined were tuberculin reactors (45.9% [n = 257]) (Table 2). Among tuberculin positive physicians, 70% had previously tested positive. Of those physicians who had been BCG-vaccinated, the prevalence of tuberculin reactivity was 69.9% (Figure 1). Among non- Canadian-born physicians, the prevalence of tuberculin positive results was 70.7% (n = 123). Prior BCG was reported for 63.5% of the non-canadian-born physicians. For those physicians who were both BCG-vaccinated and born outside of Canada, the percentage of tuberculin reactors was 78.3%. The reactivity rate for foreign-born, non-bcg vaccinated individuals was 57.4%. For Canadian-born physicians who never received the BCG vaccine, the prevalence of reactivity was 13.1%, compared to 61.5% of Canadian-born physicians who had been BCG vaccinated. Six (1.2%) of the physicians reported having active TB in the past. Four of these cases were confirmed via the provincial TB registry, and four of the six cases were foreign-born. Of the latter four, two were diagnosed in their country of birth, one in childhood, and one at completion of medical training. The two others were diagnosed in Canada and developed TB after 2 years and 7 years, respectively, of Canadian postgraduate medical training. Of the two Canadian born cases, one had Table 2 Tuberculin reactivity of the study population (n = 560): physicians, , Edmonton, Alberta, Canada No. (%) Negative result 303 (54.1) Positive result 257 (45.9) Of the total positive results (n = 257) New positive a 77 (30.0) Previous positive 180 (70.0) Confirmed by documentation 34/180 (18.8) Confirmed by re-test for study 26/180 (14.4) Not re-tested/no documentation 120/180 (66.6) Those skin tested for study with positive results (n = 103) Positive on first step 88 (85.4) Negative on first step and positive on second step 15 (14.6) a No previous test result or previously negative 1 year.

4 TUBERCULIN REACTIVITY AMONG PHYSICIANS 1025 Table 3 Tuberculin reactivity according to demographic variables: physicians, , Edmonton, Alberta, Canada a Figure 1 Tuberculin reactivity of physicians according to country of birth and Bacillus Calmette-Guérin (BCG) vaccination status, , Edmonton, Alberta, Canada primary TB as an infant. The other was a very weak reactor after BCG vaccination at medical school entry and developed pulmonary TB after more than 20 years of practice; this case is the only one that is almost certainly linked to physician practice. Of the 183 previously positive physicians without active disease, only 13.3% had reported taking Isoniazid (INH) prophylaxis. Of the reactor physicians who had never been on TB prophylaxis, 90.1% had never been offered medication, while the remaining 9.9% had been offered, but refused. The variables foreign-birth, BCG, older age, previous contact, and practice in a mid or high burden country demonstrated significant associations with tuberculin reactivity (Table 3). Those physicians whose current practice included higher numbers of high-risk patients (i.e. Aboriginal, elderly, or foreign-born) did not demonstrate an increased risk of tuberculin reactivity. We hypothesized that we would find a higher proportion of reactors among physicians working in specialties with high-risk procedures such as bronchoscopy (respiratory medicine), intubation (anaesthesia), post-mortem (pathology) or gastroscopy (gastroenterology) in addition to those working in the emergency room or in infectious diseases. Respiratory medicine, internal medicine and general surgery showed the highest proportions of tuberculin reactivity of all the specialties. Respiratory physicians had an extremely high prevalence of reactivity compared to the other specialty groups: 83% for the 18 physicians examined. Table 4 indicates that compared to family medicine, none of the other hypothesized high-exposure specialties (emergency medicine, gastroenterology, infectious diseases, anaesthesia and pathology) had significantly high proportions of tuberculin reactivity (ranging from 20% for emergency to 57.1% for gastroenterology specialists). However, general surgeons had a 70% prevalence and a statistically significant odds ratio (OR) of 3.48 (95% CI : ), and 68% of the internal medicine physicians examined were reactors with a statistically significant OR of 3.17 (95% CI : ). Logistic regression analysis of tuberculin skin test status demonstrated that age 45 years, previous BCG vaccination, non-canadian country of birth, practising medicine in a midhigh burden country for longer than one month, and pulmonary specialty were independent risk factors for tuberculin Mantoux positive Total 257 (45.9) 560 Gender Age No. (%) Total OR (95% CI) b Male 204 (47.7) ( ) Female 53 (40.2) years 149 (59.4) ( ) 45 years 68 (27.9) 244 Country of birth Foreign-born 123 (70.7) ( ) Canadian-born 95 (29.3) 324 BCG c Yes 151 (69.9) ( ) No 64 (23.3) 275 Known exposure to an active TB d case Yes 153 (47.7) ( ) No 63 (36.8) 171 Foreign travel and practice (for longer than one month) No travel 151 (38.9) Travel only 26 (45.6) ( ) Travel and medical practice 40 (72.7) ( ) a The information for the variables of age, BCG, country of birth, known exposure and foreign travel/practice are not based on the total study population size (n = 560) owing to incomplete questionnaire information, or questionnaires that were never completed (n = 56). b OR = odds ratio; CI = confidence interval. c Bacillus Calmette-Guérin. d Tuberculosis. Table 4 Tuberculin reactivity for those medical specialties deemed at high risk for tuberculosis exposure: physicians, , Edmonton, Alberta, Canada a Mantoux positive No. (%) Total OR (95% CI) b Family practice 61 (40.1) 152 Referent Anaesthaesia 15 (45.5) ( ) Emergency medicine 1 (20.0) ( ) Gastroenterology 8 (57.1) ( ) Geriatric medicine ( ) General surgery 14 (70.0) ( ) Infectious diseases 6 (42.9) ( ) Internal medicine 17 (68.0) ( ) Pathology 7 (41.2) ( ) Respiratory medicine 15 (83.3) ( ) a This table includes only those specialties that have historically been deemed as high risk for TB exposure or those which were identified in this study as having a high prevalence of tuberculin reactivity. Family practice has been included as the referent category. b OR = odds ratio; CI = confidence interval. reactivity (Table 5). Age 45 years was used as a cut-point because of a natural break in the log odds of reactivity at

5 1026 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 5 Final model for logistic regression analysis of tuberculin reactivity status: physicians, , Edmonton, Alberta, Canada (n = 470) Covariates this age. Although previous BCG vaccination and foreign-birth are highly correlated (OR = 3.4), both variables contribute independently to the model. The interaction between BCG and foreign-birth, as well as all other interaction terms, were found to be non-significant. A sub-analysis of Canadian-born, non-bcg-vaccinated physicians was performed and only two variables remained significant: having practised in a mid-high burden country (OR = 7.92 [95% CI : ]) and age 45 years (OR = 2.78 [95% CI : ]). Discussion OR (95% CI) a Age 45 years 2.64 ( ) Previous BCG b vaccination 4.76 ( ) Non-Canadian birth 4.47 ( ) Foreign practice c 4.37 ( ) Respiratory medicine specialist 7.58 ( ) a OR = odds ratio; CI = confidence interval. b Bacillus Calmette-Guérin. c Practising in a country other than North American and Western European countries, Australia and New Zealand. The results of this study indicate that almost one-half (45.9%) of a random sample of physicians from the Edmonton area are tuberculin skin test reactors. Although BCG vaccination and foreign-birth account for a large proportion of this reactivity, practice in mid- to high-burden countries and age 45 years increased the risk. The reactivity prevalence of 45.9% is comparable to that of other HCW. 10 This appears to be higher than that of the general Canadian population. However, tuberculin reactivity surveys are rare. A 1988 survey reported that the prevalence of TB infection among Canadian-born, non-bcg vaccinated young workers ( 25 years old) was approximately 4.3%. 17 A compilation of skin test results from large TB screening programmes in British Columbia, Canada, for example from industry workplace screenings, indicate that the tuberculin reactivity in the general population is approximately 7% (Ted Allen, BC TB Control, former director, personal communication, 1997). This is contrasted with the 13% reactivity of Canadian-born, non- BCG-vaccinated physicians in this study. This difference in reactivity suggests that factors associated with medical practice may increase the risk of tuberculin reactivity. The risk factors in this study correlate well with other tuberculin reactivity prevalence studies among physicians and HCW. Positive tuberculin skin test results are associated with foreignbirth, older age, and previous BCG vaccination in the majority of other studies. 6,8,10 Practising medicine in a foreign country is a significant risk factor in this study, but it does not appear to have been examined in other studies and hence cannot be corroborated from the literature. The reactivity prevalence for different specialties is not well defined in previous studies on either HCW or physicians. Although internal medicine and family medicine specialties have historically shown high tuberculin reactivity, 6,7,9 recent HCW studies have produced varying results. The high tuberculin reactivity in pulmonary medicine found in our study was previously discussed by Malasky et al. in their study, which identified that 11% of pulmonary fellows converted their tuberculin reactor status within 3 years. 18 The effect of changing specialties during one s career was not assessed in this study. The prevalence of reactivity detected in this study is only slightly higher than that of 38% for HCW in a recent Montreal study. 10 In addition, the percentage of Canadian-born, non- BCG-vaccinated reactors is 13% in this study, less than the 22.6% reactivity found among Canadian-born, non-bcgvaccinated HCW in the Montreal study. 10 While two recent physician studies have shown a much lower prevalence of reactivity (24.5% and 13% to 20%), 8,9 both of these study populations were considerably younger and consisted mainly of American-born physicians. This study has both a notable strength and some limitations. Unlike most other studies performed on the tuberculin reactivity status of HCW, two-step tuberculin skin testing was the protocol for testing in this study. Two-step testing helps eliminate false negative results by up to 80% in previously BCG vaccinated people, 19 and provides a more accurate baseline prevalence. Although attempts were made by the study personnel to read all tuberculin skin tests, some self-reading could not be avoided. In order to assess a potential bias of self-reading on the results, the analysis was redone after dropping the 11.7% of physicians who either self-read the second test of their two-step skin test, or who self-read their first and only positive skin test. The final logistic regression model using this altered dataset did not differ significantly from that of the model using the entire dataset. The retrospective nature of the data collection for variables such as BCG vaccination, past travel, tuberculosis exposures, and tuberculin reactivity status may have lead to inaccuracies in this study. The issue of recall is a problem inherent in tuberculin prevalence studies. Because of our reluctance to re-test people who recalled a previous strongly positive skin test, we chose to rely on the recollection of the participants, rather than risk an underestimation of reactors who refused re-testing. However, of those individuals claiming to be reactors who were re-tested for the study, 25% were negative. To examine this issue of potential misclassification from recall error, a sensitivity analysis was undertaken to determine the effect if 50% of those physicians who claimed to have had a previously positive skin test and were not re-tested, were actually negative. A logistic regression analysis performed on this altered data set revealed the same risk factors to be significant, with very similar coefficients, when compared with the actual data. A participation bias might also exist in this study because known reactors may not have been as willing to participate. Upon comparison of participating and non-participating physicians, a significantly higher number of foreign-born and older physicians did not participate. With the assumption that older physicians were more likely to have been BCGvaccinated, the prevalence found in this study may, in fact, be an underestimation of the true proportion; the non-participants

6 TUBERCULIN REACTIVITY AMONG PHYSICIANS 1027 may have had a higher percentage of two of the primary risk factors associated with tuberculin reactivity, namely foreignbirth and prior BCG vaccination. When examining tuberculin reactivity, it is important to consider the effect of prior BCG vaccination on present reactivity status. Evidence has shown that after an interval of years, 25% of individuals BCG vaccinated after the age of 5 years are reactors, and only 8% of individuals vaccinated in infancy are reactors. 20 A total of 86% of BCG vaccinated physicians in this study were vaccinated after the age of 5 years and, on average, 33.7 years had elapsed since the vaccination. This would suggest that the reactivity in previously BCG-vaccinated physicians should not be attributed to the BCG vaccination. Interestingly, no significant difference (P = 0.8) in induration size was noted between those tuberculin reactors who were BCG vaccinated (mean: 17.9 mm) and those who were not vaccinated (mean: 18.4 mm). Conclusions and Recommendations This is the first study to examine tuberculin reactivity among Canadian physicians. It demonstrates that physicians in low prevalence regions have significantly elevated reactivity rates and the associated risk factors are identified. These factors are: age 45 years, practice in mid- to high-burden countries, and previous BCG vaccination. These factors do not differ from risk factors found in other HCW studies. Physicians practising respiratory medicine, surgery and internal medicine experience a high prevalence of reactivity whereas physicians practising gastroenterology, emergency medicine and pathology do not. An examination of physician practice populations indicates that high proportions of Aboriginal, foreign-born or elderly patients do not increase the risk of reactivity. The Canadian and American guidelines for preventing nosocomial transmission of TB in health care facilities outline the need for a strong TB infection-control programme, which includes tuberculin skin testing of employees at the time of hiring, annually and after an exposure. 21,22 In our cohort, the average length of time since the last tuberculin skin test for those physicians who were non-reactors was 9.6 years. This is consistent with the fact that even though physicians have the highest conversion rates of any occupational group in Canadian hospitals, they are the least monitored for infection. 2 Despite physicians having less predictable schedules, 24 the need for their surveillance cannot be overstated. The willingness of physicians to participate in this tuberculin reactivity survey suggests they are amenable to inclusion in routine surveillance. Acknowledging potential nosocomial exposure is an important step towards reducing infection among physicians and all other HCW. 21,22 It was the impression of the study personnel that this research raised awareness and corrected misconceptions about the significance of tuberculin reactivity, BCG protection, TB exposure and infection risk. With greater diligence in testing and the use of flexible schedules for physicians, their TB risk could be minimized and their health protected. Acknowledgements This work was supported by a grant from the Alberta Lung Association. The authors sincerely thank Ms Lori Zapernick for her time and extraordinary effort in completing the tuberculin skin testing. This paper derives from the Masters thesis completed by the first author under the title Prevalence and Determinants of Tuberculin Reactivity among Physicians in the Edmonton Capital Health Authority, Autumn This study was presented in part at the Annual Meeting of the International Union Against Tuberculosis and Lung Diseases (IUATLD), Paris, France, October, References 1 Menzies D, Fanning A, Yuan L, Fitzgerald JM, and the Canadian Collaborative Group in Nosocomial Transmission of Tuberculosis. Tuberculosis in Health Care Workers: a multicentre Canadian prevalence survey: preliminary results. Int J Tuberc Lung Dis 1998; 2(9):S98 S Holton D, Paton S, Gibson H, Taylor G, Whyman C, Yang TC. Status of tuberculosis infection control programs in Canadian acute care hospitals, 1989 to 1993 Part 1. Can J Infect Dis 1997;8: Schwartzmann K, Menzies D. Tubersulosis: 11 Nosocomial Disease. CMAJ 1999;161: Alberta Health. Tuberculosis Services Annual Report, Edmonton, Alberta, Canadian Lung Association. Canadian Tubersulosis Standards. 4th edn. Canada, Barrett-Connor E. The epidemiology of tuberculosis in physicians. JAMA 1979;241: Geiseler J, Nelson KE, Crispen RG. Tuberculosis in physicians: compliance with preventive measures. Am Rev Resp Dis 1987;135: Fraser VJ, Kilo CM, Bailey TC, Medoff G, Dunagan WC. Screening of physicians for tuberculosis. Infect Control Hosp Epidemiol 1994;15: Ramphal-Naley L, Kirkhorn S, Lohman WH, Zelterman D. Tuberculosis in physicians: compliance with surveillance and treatment. Am J Infect Control 1996;24: Schwartzman K, Loo V, Pasztor J, Menzies D. Tuberculosis infection among health care workers in Montreal. Am J Respir Crit Care Med 1996;154: Fanning A, Reid R, Hennig B, Zapernick L. Tuberculin reactor status of house staff in a post graduate training centre, Alberta, Canada. Am J Resp Crit Care Med 1994;149:A Alberta Health. Tuberculosis Services Annual Report, Edmonton, Alberta, CDC: Centers for Disease Control. EpiInfo, version 6.0. Atlanta, Georgia, SPSS Inc. SPSS for Windows, Release Chicago, Illinois, StataCorp. Stata Statistical Software: Release 5.0. College Station, Texas Hosmer DW, Lemeshow S. Applied Logistic Regression. Toronto: Wiley, Menzies D, Chan CH, Vissandjee B. Impact of immigration on tuberculosis infection among Canadian-born schoolchildren and young adults in Montreal. Am J Respir Crit Care Med 1997;156: Malasky C, Jordan T, Potulski F, Reichman LB. Occupational tuberculous infections among pulmonary physicians in training. Am Rev Respir Dis 1990;142: Horowitz HW, Luciano BB, Kadel JR, Wormser GP. Tuberculin skin test conversion in hospital employees vaccinated with Bacille Calmette-Guerin: recent infection or booster effect? Am J Infect Control 1995;23:

7 1028 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY 20 Menzies RI, Vissandjee B. Effect of Bacille Calmette-Guérin vaccination on tuberculin reactivity. Am Rev Respir Dis 1992;145: CDC: Centers for Disease Control. Issues and recommendations on the role of BCG vaccine in the prevention and control of tuberculosis. Am Fam Physician 1996;54: Health Canada. Guidelines for preventing the transmission of tuberculosis in Canadian health care facilities and other institutional settings. Can Comm Dis Rep 1996;22(Suppl.1):1 54.