INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The risk of occupationally acquired tuberculosis remains a concern (1, 2), even as case-rates of tuberculosis have declined nationally (3). The cornerstone of determining the effectiveness of tuberculosis control efforts in hospitals is accurate monitoring of employee tuberculin conversion rates. A recent survey of 210 U.S. hosptials conducted by the Society for Healthcare Epidemiology of America (SHEA) and the Centers for Disease Control and Prevention (CDC) found 561 (0.64%) tuberculin conversions among 87,156 tests performed in health care workers in 1992 (4). Rates in certain urban settings may be higher and exceed 3% annually, even in non-outbreak settings (5, 6).

Despite acceptance that care of tuberculosis patients confers occupational risk, many studies have suggested that community transmission, rather than occupational exposure, constitutes the dominant mode of transmission to employees (7). In addition, job-specific rates of transmission have not predictably identified groups at risk, complicating development of targeted infection control strategies. These reports, however, have come from hospitals that treat relatively few patients with tuberculosis.

In contrast, tuberculosis case-rates at St. Clare's Hospital and Health Center in New York City are far in excess of national rates. To determine the relative contribution of occupational versus community exposure to Mycobacterium tuberculosis, we determined rates of tuberculin conversion among health care workers at St. Clare's from 1991 to 1994, the interval when CDC guidelines to control tuberculosis were systematically implemented (10). Occupational exposure was examined by stratification according to job category, while community transmission was assessed by analyzing rates of tuberculosis in an employee's postal zone of residence.

Unlike other studies (7), we found that job category was strongly associated with risk of tuberculin conversion, while postal zone of residence was not. In addition, the group that included housekeeping, laundry, and security personnel continued at significant risk even after implementation of CDC guidelines. Educational efforts targeting health care worker groups at risk are needed to continue to control the intra-institutional spread of tuberculosis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

St. Clare's Hospital and Health Center is a 250-bed, acute-care hospital, including a 92-bed AIDS unit and a 25-bed secured prison unit. There are no pediatric or obstetrical services. The hospital is located on the Upper West Side of Manhattan and has 850 full-time employees and a medical staff of approximately 230 voluntary physicians. Over 100 new full-time employees are hired annually (11).

Employee health records are maintained by employee health service staff. For voluntary physician staff who had not completed evaluation through the employee health service, additional demographic and tuberculin status data were obtained through the medical staff office. At time of initial employment, all individuals complete a health questionnaire that includes information on country of birth, BCG vaccination status, age, gender, occupation, and postal zone of residence.

A positive tuberculin was defined as 10-mm induration 48 to 72 h following 5-TU tuberculin administration. A tuberculin conversion was defined as an increase of at least 10-mm induration on serial tests performed within 2 yr and read at employee health service. Tuberculin tests formally interpreted in writing by outside physicians were also allowed. At St. Clare's, in 1991 and 1992, Aplisol (Parke-Davis, Morris Plains, NJ) was the tuberculin preparation used, while in 1993 and 1994, Tubersol (Squibb-Connaught, Swiftwater, PA) was used.

In February 1993, two-step testing of new employees was introduced. Fifteen health care workers were identified with a boosted response to serial testing (12). These employees do not fit meaningfully into the categories of tuberculin-positive, tuberculin-negative, or tuberculin converter, and so are excluded from the current analysis. Results of analysis of the data, including categorization of the 15 employees with a boosted response on two-step testing as "converters," are provided in the footnotes of the tables.

Because there are more than 100 unique job descriptions, jobs were categorized into one of five groups, similar to other studies (7- 9). The "laboratory" group included employees from the laboratory and blood bank and technicians (radiology, respiratory, pathology, cytology). The "physician-nurse" group included physicians (staff, physician's assistant, medical students), nurses (RN, LPN, nurse practitioners), and emergency medical service personnel. The "social service" group included employees from pastoral care, occupational and physical therapy, social service and utilization review, case managers, and patient advocates. The "housekeeping" group included personnel from housekeeping, security, engineering and maintenance, laundry, dietary, and transportation. The "finance" group included workers from finance, data processing, accounting, medical records, human resources, the mail room, telecommunications, pharmacy, purchasing, and central supply.

Differences between means were tested either by the student test or the rank sum test. Differences in categorical variables were tested by exact chi-square (13). Differences in tuberculin conversion rates between 1991-92 and 1993-94 were compared using exact mid-p probabilities (14). The relative risk of tuberculin conversion among job categories was determined by Cox's proportional hazard model (15). Two models are presented, including one adjusted for age, BCG vaccination status, U.S. or foreign birth, gender, and rate of tuberculosis in the employee's postal zone of residence (n = 500). Since the rate of tuberculosis in the area of residence was not a significant predictor of tuberculin conversion and this information was not available for 182 employees living outside of New York City, a second model was used excluding the tuberculosis rates per postal zone (n = 682). Results from the models were similar. Statistical significance was defined as p  0.05.

Tuberculosis rates per postal zone were determined as follows: the Surveillance Unit of the Bureau of Tuberculosis Control, New York City Department of Health, provided the annual number of cases of tuberculosis diagnosed in each of 180 postal zones in New York City, from 1991 to 1994. For each postal zone, this case-number was divided by the population in that postal zone as estimated by the 1990 national census. The latter information was provided by the United States Census Bureau.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

From 1991 to 1994, 56 to 118 new cases of tuberculosis were diagnosed annually at St. Clare's Hospital, and case-rates ranged from 1,200 to 2,500 per 100,000 hospital discharges. Respiratory isolation days per year ranged from 6,360 to 10,883. Information on tuberculin status was available for 1,303 employees (> 90% of all employees), including 15 with a positive boosted response who are excluded from further analysis. Of the remaining 1,288 employees, 577 (44.8%) had a pre-existing positive tuberculin and 711 were tuberculin-negative at the start of 1991. Features associated with a pre-existing positive test included older age, female gender, foreign country of birth, and previous BCG vaccination (Table 1). Laboratory (56%), physician-nurse (52%), and housekeeping (48%) groups had a higher prevalence of pre-existing tuberculin reactivity than social service (31%) and finance (25%) groups.

During the 4-yr study period, 97 tuberculin conversions occurred among the 711 tuberculin-negative hospital staff who were followed over 1,869 person-years (Table 2). Risks for conversion included occupation, age, BCG vaccination, and country of birth. The housekeeping group had the highest conversion rate (9.2/100 person-yr), and finance the lowest (2.5/ 100 person-yr).

The association between tuberculosis rate in postal zone of residence and risk of tuberculin conversion was analyzed to determine the potential contribution of community versus occupational transmission (Table 3). Complete information, including tuberculosis rate in postal zone of residence, was available for 500 (70%) of 711 tuberculin-negative employees. Pertinent postal zone-related rates of infection were unavailable for 182 employee who live outside of New York City, and no postal zone was available for 29 employees. In multivariate analysis adjusted for age, BCG vaccination status, country of birth, gender, and tuberculosis rate in postal zone of residence, the housekeeping group had a risk 4.3 times greater (p < 0.001) than the finance group. Similarly, the physician/nurse group had an elevated risk (RR 2.4, p < 0.05) compared with the finance group.

The potential association between tuberculosis rate in postal zone of residence, tuberculin conversion, and occupation was examined further. The mean 4-yr case-rate of tuberculosis per 100,000 persons was calculated for each postal zone of residence for individual employees according to job category. The rate for New York City during the 4-yr interval was 47 cases per 100,000 persons. The case-rates of tuberculosis in the postal zones where employees resided did not vary significantly for any of the five job category groups (p = 0.31; range 50 to 73 cases per 100,000). Postal zones where no hospital employees resided had a combined case-rate of 30 cases per 100,000 persons.

Since tuberculosis rate in postal zone of residence was not associated with risk of tuberculin conversion, job-specific rates of tuberculin conversion were calculated in the multivariate model, exclusive of postal zone tuberculosis occurrence rates, among the 682 employees with complete data exclusive of postal zone (Table 3). In this analysis, the housekeeping group maintained a significant risk for tuberculin conversion (RR 3.4, p < 0.001). Rates of tuberculin conversion for other job categories were not statistically significant in this model.

Numerous infection and tuberculosis control measures were implemented during the study period. To determine the effect of these interventions, the tuberculin conversion rate per occupational group was calculated for 1991-92, before most changes were implemented, and for 1993-94. In this analysis (Table 2), improvement in the conversion rate was seen in each occupational category (p for heterogeneity = 0.48) but, because of varying sample sizes, the changes were statistically significant only for the total of all employees (p < 0.001) and the physician/nurse (p < 0.01) and social service groups (p < 0.04). Because of a concern that a bias toward improvement would be introduced by the exclusion of the employees with a boosted response on two-step testing, the analysis was repeated, with the 15 employees with boosted responses regarded as "converters." In this analysis, only the employee groups as a whole had a significant improvement (p = 0.03).

The housekeeping group had the highest observed conversion rate both before and after implementation (Figure 1). The likelihood of remaining tuberculin-negative over the study period by occupational group, after adjusting for all covariables including postal zone of residence in the proportional hazard model, is shown in Figure 2. The analysis adjusting for all covariables except postal zone is similar (data not shown). In all analyses, the housekeeping group remains at significant risk for tuberculin conversion, while the risk decreases for other groups.

View larger version (16K): [in this window] [in a new window]   Figure 1.   Observed probability of no tuberculin conversion in hospital employees by job type 1991-1994.

View larger version (19K): [in this window] [in a new window]   Figure 2.   Expected probability of no tuberculin conversion in hospital employees by job type 1991-1994.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have found a strong association between job category and risk of tuberculin conversion at a hospital in New York City with high annual rates of tuberculosis and have found no association between postal zone of residence and the occurrence of tuberculin conversion. This is in distinction to previous reports from hospitals with lower annual case-rates of tuberculosis (7) and in concordance with a recent study from Montreal (16), which found an unexpectedly high prevalence of tuberculin reactivity, as well as tuberculin conversions, among workers. Establishment of occupational risk in non-outbreak situations is important for future control strategies and demonstrates the importance of an active tuberculosis control program.

The question of what an "acceptable" tuberculin conversion rate might be is unclear and is confounded by the problems with reproducibility and interpretation of the tuberculin test. In a survey of 210 U.S. hospitals conducted by SHEA and the CDC (4), the conversion rate in 1992 was 0.64%, with higher rates seen in larger hospitals (> 473 beds), in those hospitals treating  6 cases of tuberculosis annually (17), and among bronchoscopists. At hospitals in New York City, non-outbreak associated conversion rates may be about 3% per year (5, 6, 18). At our hospital, tuberculin conversion occurred in 7.2% of employees annually in the 2 yr before implementation of guidelines versus 3.3% after control measures were fully instituted.

Our finding that housekeeping personnel had the highest rate of conversion has been suggested by others (18, 19), including a rate of 8.6% annually among housekeeping personnel at Bellevue Hospital in New York (18). Others studies that have compared rates by job category have found less difference (7, 9, 20, 21). Many of these studies, however, have not provided information regarding the BCG vaccination status and postal zone of residence of the employees.

We are uncertain exactly what occupational trait might confer the increased risk to this group of employees. However, common to the occupations in this group is brief daily contact with many patients throughout the hospital, usually by entering a patient room. We speculate that because the contacts are frequent and brief, little attention is paid to an intervention such as wearing an appropriate mask. In addition, because members in this occupational group have historically not been considered "at risk" for airborne infections, less educational effort might have been made to instruct these workers on the principles of airborne transmission and the rationale for interventions such as masks. Since identifying the risk for this group, we have increased our educational efforts in the form of frequent small "in-service" presentations.

There are several limitations to our study. First, our method of determining postal zone-based case-rates of tuberculosis was only an estimate. We used the 1990 population determined by the Census Bureau as our denominator for all 4 yr; certainly population varied, possibly considerably, in certain postal zones during this time. Also, some have questioned the accuracy of the 1990 census, particularly in urban settings such as New York City. In addition, we used the postal zone of employees at initial hire. This information is not updated in a routine fashion when an employee moves to another area (and postal zone) during employment.

Our decision to exclude the 15 employees who had a positive boosted response from the analysis certainly affected the results, since many of these 15 persons may have "pseudo-converted" during the second 2 yr of the study period. Thus, the number of conversions in the second 2 yr of the study was a more accurate estimate of true conversions, while inevitably some persons who were considered converters in the first 2 yr of the study did not truly convert their tuberculin test but rather had a boosted response to serial testing. Our re-analysis of the data to include these persons as "convertors" did decrease the magnitude of the 1991-92 versus 1993-94 improvement in tuberculin conversion rate (Table 2)) but had no effect on the relative risk of occupation groups in the multivariate analysis (Table 3).

Another study limitation is potential heterogeneity in the occupational groupings. Because of the many (> 100) occupations at our hospital, we had to make certain presumptions in our groupings, based both on similarities of activity as well as anticipated level of patient contact. Because of these broad groupings, however, we are not certain which particular occupations in each group are at highest risk. The study from Montreal divided health care workers among eight clinical units and examined rates for other workers, such as respiratory therapists, as well (16). In that study, occupation was significantly associated with tuberculin conversion (p < 0.01) (16).

Other study limitations included the fact that the HIV serostatus of health care workers was unknown. Because of the increasing rates of skin test anergy expected as immunity wanes due to progressive HIV infection (22), both the prevalence of tuberculin reactivity and the rate of new latent M. tuberculosis infection may have been underestimated. Also, information was gathered for voluntary physicians through the medical staff office, making data less complete.

Finally, the usual problems that plague all studies that rely on the tuberculin skin testthat is, the relatively poor specificity and sensitivity of this old testcreate the same uncertainties in the current study (23). This is further compounded in this report by our change in early 1993 from Aplisol to Tubersol. The association of false-positive tuberculin tests with Aplisol has been noted by many (24, 25) but not all (26) investigators. We have previously reported on rates of tuberculin positivity among new employees from 1991 to 1993 (11); in that study, we did not find a change in the annual rate of tuberculin reactors, despite the switch during the study from Aplisol to Tubersol. In addition, the change in tuberculin preparations cannot account for the continued high rate of tuberculin conversions we found among the housekeeping group in 1993-94 (6.7/100 person-years) in the current study.

In conclusion, we have found a significant association between job category and risk of tuberculin conversion at a hospital in New York City with high annual rates of tuberculosis and have found no association between postal zone of residence and the risk of tuberculin conversion. We have also identified an occupational group with continued risk despite implementation of recommended tuberculosis control guidelines. These findings have important implications for tuberculosis control programs and demonstrate the need for educational efforts targeted to persons in specific jobs.