Efficacy of Control Measures in Preventing Nosocomial Transmission of Multidrug-Resistant Tuberculosis to Patients and Health Care Workers

Methods

In May 1991, we investigated an outbreak of multidrug-resistant tuberculosis at Cabrini Medical Center, a teaching hospital in New York City. In that investigation, epidemiologic and laboratory data showed the occurrence of nosocomial transmission of multidrug-resistant tuberculosis to patients and health care workers [5]. Risk factors for acquisition of multidrug-resistant tuberculosis by patients included age, HIV seropositivity, and hospital admission to Cabrini Medical Center within 7 months before onset of multidrug-resistant tuberculosis. Additionally, health care workers assigned to the medical or HIV wards (primary wards housing patients with tuberculosis) were identified as being at increased risk for tuberculin skin test conversion. Conditions facilitating tuberculosis transmission included delayed identification, isolation, or treatment of patients with infectious tuberculosis; lapses in tuberculosis isolation practices; and inadequate isolation facilities for patients with tuberculosis. In response to the outbreak, control measures similar to those recommended in the CDC's 1990 tuberculosis guidelines were instituted to prevent further nosocomial transmission of tuberculosis [16].

Control Measures

Numerous measures were introduced to decrease transmission of multidrug-resistant tuberculosis, including source controls, environmental controls, and respiratory protection for health care workers (Table 1). Source control measures included improved isolation precautions and expanded treatment regimens for patients with suspected or confirmed tuberculosis and included routine use of acid-fast bacilli smears to determine length of isolation and to establish treatment efficacy. To expedite the identification of patients with infectious tuberculosis and the determination of the antimicrobial susceptibility of M. tuberculosis isolates, laboratory procedures were also modified. Thus, personnel and workday hours devoted to processing of mycobacterial isolates were expanded; a hospital-wide computerized database of patients with tuberculosis was developed; a gene probe was introduced; and the reporting of acid-fast bacilli smear results to requesting physicians was standardized.

Environmental modifications included installing exhaust fans in isolation rooms to increase the number of acid-fast bacilli isolation rooms with negative pressure and included using portable Aeroguard (HR Incorporated, Bellevue, Washington) chambers to do cough-inducing procedures, such as sputum induction or aerosolized pentamidine administration. Finally, to further decrease the risk for occupational acquisition of tuberculosis, the respiratory protection worn by health care workers was changed from nonmolded to molded surgical masks (3M 1800+ Aseptex, 3M Incorporated, St. Paul, Minnesota).

To assess compliance with these control measures, we examined the hospital's isolation facilities, reviewed patient charts for documentation of initiation and discontinuation of acid-fast bacilli isolation precautions, interviewed infection control and laboratory personnel, observed isolation practices on hospital wards containing patients with tuberculosis, and evaluated the ventilation system in acid-fast bacilli isolation rooms by testing air-flow direction using smoke tubes.

Epidemiologic Studies

A case-patient was defined as any Cabrini Medical Center patient whose M. tuberculosis isolate was resistant to at least isoniazid and rifampin and as any patient whose clinical course was consistent with active tuberculosis during the study period (1 January 1990 to 11 August 1992). Case-patients and all other patients with tuberculosis were identified by reviewing Cabrini Medical Center laboratory and infection control records during the study period. Medical records of all potential case-patients were reviewed to verify a clinical course consistent with tuberculosis.

To determine whether a decrease had occurred in the incidence of multidrug-resistant tuberculosis after the institution of control measures, we compared the proportion of patients with tuberculosis who had multidrug-resistant strains before (preintervention period) and after (intervention period) the implementation of control measures. The preintervention period (January 1990 to June 1991) was defined as the time from the onset of the outbreak until the formal institution of recommended control measures; the intervention period (July 1991 to 11 August 1992) was defined as the time during and after the institution of recommended control measures until our follow-up study.

Patient-to-Patient Transmission

To investigate possible patient-to-patient transmission of multidrug-resistant tuberculosis, we evaluated case-patient medical records to determine if case-patients had been hospitalized at Cabrini Medical Center within 7 months before onset of tuberculosis. For those case-patients who had been previously hospitalized, we determined whether they had been hospitalized on the same ward at the same time or whether they had a documented nosocomial exposure to another patient with culture-confirmed multidrug-resistant tuberculosis during a previous hospitalization. Data collected included age, sex, race, HIV serologic status, previous opportunistic infections, dates and results of acid-fast bacilli smears and mycobacterial cultures, M. tuberculosis antimicrobial susceptibilities, dates of admission and discharge, hospital room assignments, dates of isolation for tuberculosis, nursing documentation of proper application of isolation precautions, antituberculous medications prescribed, clinical course, and outcome.

Patient-to-Health Care Worker Transmission

To evaluate whether the risk for transmission of multidrug-resistant tuberculosis from patients to health care workers had been reduced, we compared conversion rates of tuberculin skin tests in health care workers during the preintervention and intervention periods. During both periods, the tuberculin skin testing program at Cabrini Medical Center required annual testing of all hospital employees and additional testing of health care workers after a known tuberculosis exposure. A tuberculin skin test conversion was defined as induration of 10 mm or more to purified protein derivative in a Cabrini Medical Center employee with a documented tuberculin skin test result that was negative within the previous 24 months. Employees without negative skin test results at baseline were excluded from our analyses. Conversion rates were compared by period, by job category (frequent direct patient contact compared with infrequent or no direct patient contact), and by ward assignment (primary wards housing patients with tuberculosis [medical and HIV wards] compared with wards infrequently housing patients with tuberculosis).

Health care workers with direct patient contact included physicians, nurses, nursing aides, respiratory therapists, and social workers; workers without direct patient contact included administrative staff, clerks, dieticians, housekeepers, engineers, and laboratory personnel. Health care workers with tuberculin skin-test conversions or active tuberculosis or both were identified by reviewing employee health and infection control records during the study period. Data collected included age, sex, race, history of bacille Calmette–Guérin (BCG) vaccination or tuberculosis exposure or both, country of origin, job title, duration of employment, and work location at time of tuberculin skin-test conversion.

Statistical Analysis

Data were collected on standardized forms, entered into Epi Info Version 5.01b software, and analyzed [17]. Categorical variables were compared by the Fisher exact or chi-square test, and relative risks (RRs) were calculated. Continuous variables were compared by the Student t-test or the Kruskal-Wallis rank-sum test.

Results

Case-Patients

Forty patients met the case definition (Figure 1). Case-patients in the preintervention and intervention periods were similar in age, race, sex, HIV serologic status, and diagnosis of AIDS (data not shown). However, case-patient mortality decreased in the intervention compared with the preintervention period (4 of 10 [40%] compared with 25 of 30 [83%] deaths; P = 0.01). Further, in the preintervention period, 20 of 25 (80%) case-patient deaths were attributed to tuberculosis, whereas in the intervention period, only 1 of 4 deaths was attributed to tuberculosis (the 3 remaining case-patient deaths occurring in the intervention period were attributed to complications of AIDS [n = 2] and to suicide [n = 1]).

The proportion of patients with multidrug-resistant M. tuberculosis infection decreased in the intervention compared with the preintervention period (10 of 70 [14%] compared with 30 of 95 [32%] patients; RR, 0.5 [95% CI, 0.2 to 0.9; P = 0.02]). Because hospitalization at Cabrini Medical Center within 7 months before onset of tuberculosis, particularly on the HIV ward, was identified as a risk factor for development of multidrug-resistant tuberculosis, we next examined the location and duration of hospitalization for case-patients before onset of tuberculosis. A similar number of case-patients in the intervention and preintervention periods had a history of previous hospitalization at Cabrini Medical Center (6 of 10 compared with 24 of 30 patients; RR, 0.8; P =0.2) (Figure 1). Further, case-patients in the intervention and preintervention periods were equally likely to have been admitted to the HIV ward (4 of 10 compared with 17 of 30 patients; RR, 0.7; P =0.5) and were hospitalized for a similar number of days (median, 5 compared with 15 days; P =0.2) before onset of tuberculosis. However, case-patients diagnosed during the intervention period were less likely than those diagnosed during the preintervention period to have had an identified nosocomial exposure to another case-patient during a previous hospitalization (1 of 10 compared with 20 of 30 patients; RR, 0.2; P = 0.003).

The one identified exposure of a case-patient to a patient with multidrug-resistant tuberculosis during the intervention period occurred approximately 6 months after the implementation of control measures. The putative source-patient was not isolated at the time of admission, and when finally isolated, the patient was intermittently noncompliant with isolation precautions (walking in hallways without a mask). Of the remaining 5 case-patients in the intervention period with previous hospitalizations, 1 was the partner and housemate of a case-patient, 1 had a history of tuberculosis and poor compliance with medications, 1 had a prolonged hospitalization that included preintervention and intervention periods and had been hospitalized on the same ward as a case-patient during the preintervention segment of the hospitalization, and 2 had no identified source of exposure.

Health Care Workers

Of 1567 tuberculin skin tests administered to health care workers with previously negative skin test results, we identified 48 (3.0%) skin-test conversions. Overall, tuberculin skin-test conversion rates during the preintervention and intervention periods were similar (26 of 840 [3.1%] compared with 22 of 727 [3.0%] conversions; P =0.9).

Health care workers who had tuberculin skin-test conversions had a median age of 34 years (range, 23 to 58 years), 34 (71%) were women, 12 (25%) were black, and approximately half (56%) were born in the United States. Only 4 (8%) were known to have received BCG vaccinations; none had a known history of tuberculosis. Before tuberculin skin test conversion, these health care workers had been in their current positions a median of 5.0 years (range, 0 to 21 years), and most workers (36 of 48 [75%]) had direct patient care responsibilities.

Health care workers with tuberculin skin-test conversions during the preintervention or intervention periods were similar in age, sex, race, and BCG vaccination status. However, health care workers with tuberculin skin test conversions during the preintervention period were more likely than those with conversions during the intervention period to be assigned to the medical or HIV wards (primary wards housing patients with tuberculosis) (Table 2).

Because overall conversion rates for tuberculin skin tests were similar for the two time periods, we next examined conversion rates by job category and ward location. During the preintervention period, health care workers with direct patient contact had higher skin-test conversion rates than did health care workers without direct patient contact (22 of 342 [6.4%] compared with 4 of 409 [1.0%] conversions; RR, 6.6; P < 0.001). During the intervention period, skin test conversion rates among health care workers with and without direct patient contact were similar (14 of 296 [4.7%] compared with 8 of 354 [2.3%]; RR, 2.1; P =0.08) (Figure 2).

When we compared tuberculin skin-test conversion rates for specific wards, a change in the risk for conversion also occurred between the preintervention and intervention periods (Table 3). Before the institution of control measures, conversion rates in health care workers assigned to primary wards housing patients with tuberculosis were higher than those in workers assigned to other wards; however, in the intervention period, conversion rates among workers assigned to wards housing patients with tuberculosis decreased to levels seen in workers assigned to other hospital wards. In contrast, conversion rates for tuberculin skin tests among health care workers assigned to other wards infrequently housing patients with tuberculosis were similar during the two time periods (Table 3).

Active tuberculosis developed in two health care workers. The first had cutaneous anergy and contracted active tuberculosis after an exposure during the preintervention period. This health care worker (a housekeeper) was HIV seropositive, had an M. tuberculosis isolate that matched the restriction fragment length polymorphism pattern for the hospital's epidemic strain of tuberculosis, was treated with a multidrug regimen for tuberculosis, and died of tuberculosis. The second health care worker had a documented skin-test conversion and contracted active tuberculosis during the intervention period. This health care worker, a physician with unknown HIV serologic status, had an upper lobe infiltrate on chest radiograph; no M. tuberculosis isolate was available. This health care worker was treated with isoniazid, rifampin, and pyrazinamide and survived.

Discussion

Recent reports of outbreaks of multidrug-resistant tuberculosis have documented the risk for nosocomial transmission of multidrug-resistant tuberculosis to patients and workers in medical facilities [1-10]. At one hospital that had had an outbreak, Cabrini Medical Center, transmission of multidrug-resistant tuberculosis occurred at a time when the CDC guidelines were not being fully implemented [5]. In response to the outbreak, various control measures were instituted to decrease transmission of tuberculosis.

Our data suggest that after the introduction of these control measures, the risk for nosocomial transmission of multidrug-resistant tuberculosis from patient to patient decreased. Decreases occurred in the proportion of patients with tuberculosis who had multidrug-resistant M. tuberculosis isolates and in the number of case-patients who might have acquired their disease nosocomially despite the similar placement of case-patients on wards housing patients with tuberculosis and despite a similar duration of hospitalization before onset of tuberculosis. At Cabrini Medical Center, the implementation of these control measures was supported by expanded staffing of the mycobacteriology laboratory and by the introduction of a standardized reporting system. Moreover, modifications such as rapid identification and treatment of case-patients with an expanded drug regimen were probably responsible for the reduction in mortality associated with multidrug-resistant tuberculosis seen in the intervention period, despite the similar demographic characteristics of case-patients and the similar severity of underlying disease.

Our results also suggest that the introduced control measures were effective in reducing transmission of multidrug-resistant tuberculosis from patients to health care workers. After control measures were introduced, health care workers with and without frequent direct patient contact had similar rates of tuberculin skin-test conversion. More importantly, the skin-test conversion rate among workers assigned to primary wards housing patients with tuberculosis decreased by 70%, returning to a rate similar to that of health care workers assigned to other hospital wards. The observed decreases in skin-test conversion rates could not be attributed to other differences (for example, age, race, or BCG vaccination status) between workers, factors that may confound conversion rates [18, 19]. These data suggest that the combination of source and environmental controls together with the use of molded surgical masks were all effective in reducing tuberculin skin-test conversion rates among health care workers at this hospital. At present, the efficacy of these molded surgical masks is unknown, and the contribution of their use to the reduction in risk for tuberculin skin-test conversion remains unclear. Recently, however, other models of molded surgical masks have been shown to filter mycobacterial organisms with more than 95% efficiency [20].

Tuberculin skin-testing programs serve two important functions in health care settings: 1) monitoring tuberculosis transmission among health care workers and 2) targeting health care workers who need prophylactic therapy or treatment of active tuberculosis. Therefore, all workers who might have had contact with infectious patients who have tuberculosis should be included in an annual tuberculin skin testing program. As our data show, hospital-wide conversion rates for tuberculin skin tests may be insensitive in determining tuberculosis transmission within an institution. Instead, calculation of job- and ward-specific (for example, nursing and bronchoscopy personnel, HIV ward) conversion rates for tuberculin skin tests may be a superior means of identifying high-risk groups for whom more frequent skin testing and adequate respiratory protection should be targeted.

Although our findings strongly suggest that nosocomial transmission was reduced as a result of the introduced control measures, our study had inherent limitations. First, results of acid-fast bacilli smears were not available throughout the entire study period, making it difficult to systematically assess the infectiousness of case-patients on the basis of smear-positivity data. Second, we used the tuberculin skin test, not the person, as the unit of analysis for our evaluation of skin-test data because health care worker-specific data for all tests were not available for analysis. However, because tuberculin skin testing was done annually and because each period (preintervention and intervention) was between 12 and 18 months in duration, the test unit is probably also a good approximation of the number of health care workers tested. We were also unable to assess potential community acquisition of tuberculosis among health care workers and how such transmission may have affected conversion rates of tuberculin skin tests. However, no clustering of conversions was noted by residence zip code and, in the absence of multiple, concurrent community outbreaks of multidrug-resistant tuberculosis, the rate of community acquisition should have been constant. Therefore, it is unlikely that community transmission of tuberculosis substantially affected the differences seen in skin-test conversion rates during the two periods. Third, it could not be determined with absolute accuracy whether some skin-test conversions occurred in the preintervention or the intervention period unless an occupational exposure was identified. However, the substantial reduction in skin test conversion rates among health care workers after the introduction of control measures provides strong evidence for a decreased risk for occupationally acquired tuberculosis infection. Fourth, because the implementation of control measures was associated with a decrease in the number of case-patients, the effectiveness of these control measures in the presence of a high concentration of infectious patients with multidrug-resistant tuberculosis during a prolonged period could not be fully evaluated. Finally, because several interventions were instituted simultaneously or consecutively, we were unable to determine the independent importance of any single control measure in decreasing nosocomial transmission of multidrug-resistant tuberculosis.

Nosocomial outbreaks of multidrug-resistant tuberculosis have been associated with significant morbidity and mortality among patients and health care workers [1-8]. Our data show that full implementation of a multifaceted approach, similar to that described in the CDC 1990 guidelines [16], can effectively reduce the risk for transmission of multidrug-resistant tuberculosis in health care settings and can reduce morbidity and mortality in affected patients. To reduce the risk for nosocomial transmission of tuberculosis to patients and health care workers, similar infection control programs should be implemented at all U.S. hospitals caring for patients with tuberculosis.