Because it illustrates genetic relatedness among strains of M. tuberculosis, DNA fingerprinting has been used by epidemiologists and clinicians to confirm suspected outbreaks of disease and episodes of laboratory cross-contamination [3, 13, 14]. We hypothesized that routine DNA fingerprinting, done by using restriction fragment length polymorphism analysis, could be used to enhance hospital infection-control surveillance for patient-to-patient transmission of Mycobacterium tuberculosis. To test this hypothesis, we performed DNA fingerprinting on all M. tuberculosis isolates obtained from Cook County Hospital, Chicago, Illinois, over a 1-year period. During this period, no patient-to-patient transmission had been detected by routine hospital infection-control practices.

Methods

Top Methods Results Discussion Author & Article Info References

Setting

Cook County Hospital is a 700-bed public hospital that serves a primarily indigent urban population. The prevalences of tuberculosis and HIV infection are high; the latter accounts for 9% of inpatient days. Administrative controls, including aggressive triage of patients with possible tuberculosis in the emergency department, have led to the widespread use of respiratory isolation for patients with known or suspected pulmonary tuberculosis. Patients with pulmonary tuberculosis are kept in respiratory isolation until they have clinical improvement while receiving therapy and have negative results on acid-fast smears on 3 separate days. Engineering controls consist primarily of isolation rooms that have been retrofitted with window exhaust fans to comply with recommended air-handling guidelines [6]. Health care workers wear personal respirators, approved by the National Institute for Occupational Health and Safety, while in respiratory isolation rooms; isolated patients wear surgical masks when not in negative-pressure rooms. Despite these measures, a few patients who are not isolated promptly on admission receive a diagnosis of pulmonary tuberculosis each year. In addition, the hospital has outpatient areas (including an HIV clinic) in which patients spend many hours in close proximity and where unrecognized tuberculosis can spread.

Laboratory Studies

During the 12-month study period (1 April 1995 to 31 March 1996), one M. tuberculosis isolate from every patient with tuberculosis at Cook County Hospital was sent to the Michigan Department of Community Health for DNA fingerprinting. All isolates were analyzed by standard methods [15] by using a 246-base pair probe representing the IS6110 sequences to the right of the PvuII site. The probe was labeled with horseradish peroxidase for detection by enhanced chemiluminescence (ECL Direct Labeling and Detection System, Amersham, Arlington Heights, Illinois) and was hybridized to DNA from M. tuberculosis isolates restricted with PvuII. PvuII-restricted DNA from M. tuberculosis strain MT14323, containing 14 copies of IS6110, was used as a DNA size marker. Autoradiographs were produced by exposing the hybridized blots to Hyperfilm ECL (Amersham).

Secondary fingerprinting done by using a probe for the polymorphic guanine cytosine-rich repetitive sequence (PGRS) has proven useful for confirming the genetic relatedness of M. tuberculosis isolates [16-20]. This second assay was performed, with the recombinant plasmid pTBN12 and the methods described above, on isolates with identical IS6110 fingerprints of 5 or fewer hybridizing bands and isolates with more than 5 hybridizing bands with fragment patterns that differed by 1 to 2 bands. A 1-kilobase ladder was included as a size marker (Gibco, Gaithersburg, Maryland). The pTBN12 was donated by Don Cave (John L. McClellan Memorial Veterans Hospital, Little Rock, Arkansas).

We compared IS6110 and PGRS fingerprints by using Whole Band Analyzer software, version 3.2.2 (BioImage, Inc., Ann Arbor, Michigan). The average linkage clustering method was used to match patterns with an SD of 2.5%. Matching patterns were compared visually to ensure similarity. For PGRS fingerprinting, only hybridizing bands greater than 1.6 kilobases were analyzed.

Isolates were considered to be genetically related (that is, clustered) if they had identical IS6110 patterns of more than 5 bands, had IS6110 patterns of more than 5 bands that differed by 1 to 2 bands and had identical PGRS patterns, or had IS6110 patterns of 5 or fewer bands and had identical PGRS patterns.

Susceptibility testing was done at the Cook County Hospital laboratory. Susceptibilities on any isolate found to be resistant to one or more first-line drugs (isoniazid, rifampin, ethambutol, or streptomycin) were confirmed by the Illinois Department of Public Health laboratory.

Clinical Data Collection

Demographic information and disease characteristics were gathered by review of the medical record for the first admission during which tuberculosis was diagnosed. This record was available for 166 of 173 patients whose isolates were fingerprinted (96%), and data were available from other sources [such as the outpatient record or the Department of Health tuberculosis database] for most of the other 7. Screening for laboratory cross-contamination was done for all clustered isolates. Laboratory cross-contamination was considered likely when 1) only one culture was positive for tuberculosis, 2) the specimen was negative on an acid-fast smear, 3) the clinical syndrome was not consistent with tuberculosis, and 4) the specimen was processed on the same day as the specimen of a patient whose isolate had a matching DNA fingerprint [13, 14].

To determine whether nosocomial transmission of M. tuberculosis had occurred between patients with clustered isolates, the times and locations of all outpatient visits and inpatient admissions from 1 April 1993 to 31 March 1996 were reviewed. If two or more patients who had isolates in a cluster had been in any inpatient or outpatient area of the hospital on the same day before they had both received a diagnosis of tuberculosis, the medical records were reviewed to determine the likelihood of M. tuberculosis transmission. Data were gathered on hospital or clinic location, use of and compliance with respiratory isolation, acid-fast smear status, HIV infection status, tuberculin skin-test status, and susceptibility patterns. Records were also reviewed to determine whether the patients had concurrently undergone laboratory or radiographic procedures or had obtained medication at the same pharmacy. Data on hospital location, time of registration, and time and location of laboratory and radiographic procedures were obtained from a computerized database that is complete and accurate; data on drug susceptibility were available for all patients with clustered isolates who were on hospital grounds concurrently with another patient with clustered isolates. If patients were registered in the emergency department on sequential days, the records were reviewed to determine whether temporal overlap existed.

Contact Tracing

The results of hospital contact tracing were reviewed. Tracing is done when a patient who has not been appropriately isolated receives a diagnosis of acid-fast smear-positive pulmonary tuberculosis. Employees who may have been exposed to this patient are identified by their supervisors and undergo tuberculin skin testing with chest radiography, if appropriate. All employees receive routine tuberculin skin testing annually. Compliance with this testing is a condition of continued employment.

The results of community contact tracing performed by the Chicago Department of Public Health were reviewed to identify out-of-hospital links between patients whose isolates were clustered. Contact tracing is done for all cases of pulmonary tuberculosis and consists of interview with the patient to identify close contacts. Contacts undergo tuberculin skin testing and, if conversions are detected, the circle of contact tracing is widened by repeated interviews to include more casual contacts, who are also tested. Community contact tracing beyond that performed by the Chicago Department of Public Health was not done in our study.

Statistical Analysis

Chi-square tests (or the Fisher exact test, when expected cell sizes were <5) were used to test the association of clustering with patient demographic and clinical characteristics. Two-tailed P values are presented with relative risks (RRs) and 95% CIs. All variables found to be significantly associated with clustering on univariate analysis were entered into a logistic model by using the statistical software package SPS, version 8.0 (SPS, Inc., Chicago, Illinois). The output of the model, which was the ln(odds) of clustering, was converted to the probability of clustering.

Results

Top Methods Results Discussion Author & Article Info References

Mycobacterium tuberculosis was recovered from 183 patients in Cook County Hospital during the study period. One hundred seventy-three isolates (95%) underwent DNA fingerprinting (3 isolates were not available, and 7 were overgrown). Of the isolates that underwent fingerprinting, 96 (55%) were in clusters. Five of these 96 isolates were found to represent cases of laboratory cross-contamination [3] or specimen mislabeling [2]. Three of the 5 false-positive cultures had already been identified clinically; the other 2 were from patients who were lost to follow-up and did not receive antituberculous therapy. After elimination of the 5 isolates that were false-positive cultures, 91 of 168 isolates (54%) in 15 clusters remained. The clusters ranged in size from 2 to 29 isolates: Six had 2 isolates each, 7 had 3 to 8 isolates each, 1 had 16 isolates, and 1 had 29 isolates.

Demographic and disease characteristics and their association with clustering for the 168 study patients are shown in Table 1. In univariate analysis, risk factors for clustering were birth in the United States (RR, 7.3 [95% CI, 2.8 to 18.6]), male sex (RR, 2.1 [CI, 1.3 to 3.6]), African-American ethnicity (RR, 3.5 [CI, 2.0 to 6.3]), current alcohol or illicit drug abuse (RR, 1.8 [CI, 1.3 to 2.6]), and homelessness (RR, 1.4 [CI, 1.1 to 1.9]). Logistic regression analysis showed that only birth in the United States (multivariate RR, 4.6) and male sex (multivariate RR, 2.6) were independently associated with clustering. Positivity for HIV and having a tuberculosis isolate resistant to one or more first-line drugs were not associated with clustering.

Epidemiologic Investigation of Patients with Clustered Isolates

In 13 of 15 clusters, which included isolates for 46 patients (2 to 8 in each cluster), we identified no instances in which patients with clustered isolates were in an inpatient or outpatient area of the hospital concurrently. Possible overlaps occurred among patients whose isolates were in the 2 largest clusters (one had 16 isolates and one had 29 isolates); most were in the largest cluster. In 148 instances, 2 patients whose isolates were from the same cluster were on hospital grounds concurrently. The results of epidemiologic investigation of these 148 instances are shown in Table 2. Nosocomial transmission of M. tuberculosis between patients in these pairs was thought to be unlikely in 144 of the 148 instances because the isolates had different susceptibility patterns (n = 32) or because no evidence of geographic overlap between the paired patients was found (n = 112).

In the remaining 4 of the 148 instances, the patient pairs had pansusceptible M. tuberculosis isolates and had been in a hospital location concurrently (the emergency department in three instances and the HIV clinic in one instance). In one instance, nosocomial transmission was considered unlikely because the possible source patient had exclusively extrapulmonary disease. In a second instance, the time from possible exposure to development of clinical tuberculosis-5 weeks to fibrotic pulmonary disease in an immunocompetent patient-was implausible.

The other two overlaps involved the same possible source patient. Patient A was an HIV-infected woman who presented to the emergency department with a 2-week history of cough, fever, and weight loss. Chest radiography showed a pattern consistent with miliary tuberculosis. Patient A was masked and placed in respiratory isolation within 1.25 hours of admission and was subsequently found to have 1+ acid-fast bacilli in her sputum. The first potentially infected patient (patient B) was also HIV infected and had a CD4⁺ lymphocyte count of 423 cells/mm³. She was in the emergency department concurrently with patient A but was admitted by ambulance after patient A had been placed in respiratory isolation. Because patient B's history included a positive result on a tuberculin skin test, patient B was masked and placed in a respiratory isolation room more than 50 yards from patient A's room. Eight months later, patient B developed pulmonary tuberculosis with a fingerprint matching that of patient A. Droplet nuclei can remain suspended for a considerable period, so nosocomial transmission of M. tuberculosis from patient A to patient B cannot be ruled out. However, because of the prompt masking and respiratory isolation of both patients, in addition to patient B's previously positive tuberculin skin-test result, transmission was considered unlikely.

The second potentially infected patient (patient C) was an HIV-infected woman who was in the HIV clinic concurrently with patient A 5 weeks before patient A's admission to the emergency department. At the time, patient A had a low-grade fever and weight loss, but lack of cough and pulmonary signs was specifically documented. Four months later, patient C developed pleural tuberculosis with a fingerprint matching that of patient A. Although patient A had no pulmonary symptoms during the temporal overlap, nosocomial transmission of M. tuberculosis from patient A to patient C was considered possible.

Tuberculin Skin Testing of Employees

During the study period, the hospital's infection-control program identified eight study patients who received a diagnosis of acid-fast smear-positive pulmonary tuberculosis before being placed in respiratory isolation. Of these patients, two had isolates in clusters and six did not. Testing of 186 employees who were exposed to these patients showed no tuberculin skin-test conversions or secondary cases of tuberculosis. Surveillance tuberculin skin testing was done on 2954 employees during the study period; the overall rate of conversion among employees decreased from 2.87% (22 of 766 employees) in the second quarter of 1995 to 1.75% (14 of 800 employees) in the first quarter of 1996 (RR, 1.6 [95% CI, 0.9 to 3.2; P = 0.14]). This is in keeping with the pooled conversion rate among health care workers of 1.2% to 1.9% that has been reported by hospitals caring for large numbers of patients with tuberculosis [8]. Many of these conversions are thought to represent community-acquired rather than nosocomial tuberculosis because 28 of the 70 employees (40%) who had conversions during the study period had no responsibilities for adult patient care: Nine were office or supply room workers, 9 were health care workers who worked exclusively in pediatrics, 7 were buildings-and-grounds workers, 1 was an animal handler, 1 was an employee trainer, and 1 was a hospital operator.

Community Contact Tracing

The results of community contact tracing were available for 73 of 80 clustered cases of pulmonary tuberculosis. Only one outpatient link between patients with clustered isolates was identified: Two patients were named contacts of a third patient with tuberculosis who was not a study patient.

Discussion

Top Methods Results Discussion Author & Article Info References

In this large, public hospital, DNA fingerprinting of all tuberculosis isolates from a 1-year period revealed many related isolates. However, retrospective epidemiologic evaluation showed only one possible case of nosocomial transmission. We believe that this finding reflects a low pretest probability because of the widespread compliance with the CDC guidelines for tuberculosis infection control at Cook County Hospital [21]. The paucity of nosocomial transmission of M. tuberculosis to patients or health care workers in a hospital with a high prevalence of tuberculosis and HIV infection supports the use of the administrative and engineering controls recommended by the CDC [5, 6]. Routine DNA fingerprinting may be useful for identifying cases of nosocomial transmission of M. tuberculosis in an institution in which the rate of delayed respiratory isolation of patients with tuberculosis is higher than that at Cook County Hospital.

Our cohort of patients with clustered isolates was predominantly male (88%) and African American (89%) and had a high prevalence of substance abuse (74%) and homelessness (36%). Characteristics similar to these have been identified as risk factors for clustering in recent population-based fingerprinting studies [22]. We did not find that HIV infection was associated with clustering, as earlier fingerprinting studies have noted [23, 24], perhaps because tuberculosis in HIV-infected persons is diagnosed more promptly and treated more aggressively today. Other than birth in the United States and male sex, no demographic features were independently associated with clustering. We hypothesize that our cohort of patients with clustered isolates may not have been diverse enough to allow us to detect other independent risk factors for clustering.

We did not address the question of where transmission of clustered strains occurred. The epidemiology of M. tuberculosis transmission has not been characterized by population-based DNA fingerprinting in Chicago as it has been in other large cities. In New York and San Francisco, DNA fingerprinting studies have suggested that the clustering of isolates often represents relatively recent transmission, usually community transmission [23, 24]. We assume that community transmission accounts for most of the clustering seen in our study. The high prevalence of homelessness and substance abuse among patients with clustered isolates suggests that transmission may have occurred at shelters or in locations where drugs were obtained or used. The fact that contact tracing identified only one link between patients with clustered isolates is not surprising. Several population-based studies of urban M. tuberculosis transmission have reported similarly poor ascertainment by traditional investigative methods [22, 24]. In these studies, more intensive epidemiologic investigation often showed links between patients with clustered isolates.

Although our cohort was large and the number of clustered isolates was high, our study has several limitations in the assessment of the usefulness of DNA fingerprinting for surveillance for nosocomial transmission. First, our 1-year study period may have been too short to capture all possible cases of nosocomial transmission. Transmission may have occurred during the study period without disease becoming clinically apparent during that period. Only a minority of immunocompetent persons infected with M. tuberculosis would be expected to manifest clinical disease within 1 year. Second, for most patients, only one isolate was fingerprinted; thus, infection with a second tuberculosis strain could have been missed. Third, in-hospital transmission could have occurred during undocumented episodes of geographic overlap (for example, in elevators or other public areas).

Our study provides no evidence to support expansion of the use of DNA fingerprinting as a tool for tuberculosis infection control outside of the settings of suspected outbreaks and episodes of laboratory cross-contamination. Although DNA fingerprinting identified one possible instance of nosocomial transmission and several false-positive tuberculosis cultures, this time-consuming and expensive procedure did not substantially increase ascertainment of nosocomial M. tuberculosis transmission. It also led to no change in clinical care or infection-control procedures. The fingerprinting results do suggest, however, that compliance with the current CDC tuberculosis infection-control guidelines can control patient-to-patient transmission in a high-risk urban hospital with a high prevalence of tuberculosis and HIV infection.