Abstract
The risk for healthcare workers (HCWs) of tuberculosis (TB) attributable to occupational exposure is difficult to determine, as are the conditions contributing to this risk. The objective of the present study was to determine which TB cases among HCWs in the Netherlands were infected during work and to analyse factors which contributed to infection and subsequent disease.
The total study population consisted of 101 cases over a 5-yr period. In 67 (66%) subjects the route of infection could be determined by epidemiological and microbiological information. Of these cases, 28 out of 67 (42%) were due to infection at work, 19 (28%) were community acquired, and 20 (30%) were infected abroad.
The 28 cases infected at work were subject to an in-depth analysis. Delayed diagnosis of the index case, especially in the elderly patient, was the main cause of patient-to-HCW transmission. In some circumstances, inadequate infection-control measures also contributed to transmission.
In conclusion, a high suspicion of tuberculosis by the clinician, adequate infection control measures by hospital authorities, and early identification of latent tuberculosis infection by occupational and public-health specialists are necessary to prevent tuberculosis among healthcare workers.
Healthcare workers (HCWs) are at risk of nosocomial infection with Mycobacterium tuberculosis 1. This risk was high in the pre-chemotherapy era, but declined rapidly after the introduction of effective treatment, which reduced the infectious period of tuberculosis (TB) patients, as well as the absolute number of TB patients in many countries 2. Along with the declining risk, the attention for infection-control practices in hospitals has lessened 2. Recognition of nosocomial transmission as a public-health issue was renewed when extensive HIV-related transmission of multidrug-resistant TB occurred in New York City hospitals (NY, USA), which also affected HCWs, some with fatal outcomes 3–5.
Several research methods, such as cohort studies and case–control studies, have been applied to study and estimate the extent of work-related TB among HCWs 2, 6–8. These studies have methodological limitations in determining the excess risk for HCWs and the importance of conditions of exposure leading to this risk 7, 8. Cohort studies and case–control studies are often unable to differentiate between occupational and nonoccupational risk, while outbreak reports may describe extraordinary situations with a high number of infecting inocula, particular virulent strains or HIV comorbidity, making it difficult to extrapolate such results to nonoutbreak conditions 4, 6, 7, 9.
In the Netherlands, all TB cases are reported to the Netherlands Tuberculosis Register (NTR), which also includes information on risk-group status 10. One of the variables within the register is “working in the healthcare/social-welfare sector”. Every year, on average 20–30 HCWs out of 1,500 cases are reported with TB. The objective of the present study was to determine which cases were really infected during healthcare work in the Netherlands and to analyse factors which contributed to infection and subsequent disease.
METHODS
The current study cohort comprised of all consecutive TB patients registered in the Netherlands during the period of January 1, 1995 to December 31, 1999, who were classified as “working in the healthcare/social-welfare sector”. After approval by the ethical committee of the NTR, all 37 Depts of Tuberculosis Control of the Municipal Health Services (MHSs) responsible for notification of the cases were identified and asked for their collaboration. This essentially meant that nurses who were involved in patient management and contact investigations related to these patients were interviewed to answer three basic questions as follows. 1) Was the patient really a HCW before or at the time of diagnosis? 2) What kind of work was the HCW involved in at the time of diagnosis or before diagnosis? 3) Does epidemiological or molecular information exist to prove or exclude infection during work in the Netherlands? Cases were excluded if the patient was not a HCW or the diagnosis of TB was withdrawn after notification.
Since 1993, all M. tuberculosis isolates in the Netherlands are subject to standardised insertion sequence (IS) 6110-based restriction fragment length polymorphism (RFLP) typing, so-called DNA fingerprinting 11. Clusters are defined as groups of patients having isolates with fully identical RFLP patterns or, if strains harbour less than five IS6110 copies, with identical sub-typing by use of the polymorphic GC-rich sequence probe 12. The fingerprints of all culture-confirmed cases in the cohort were cross-checked with the National DNA fingerprint database.
With the information from the interviews, the NTR and DNA fingerprints, patients were classified into four categories as follows. Category 1: HCW infected during healthcare work in the Netherlands. Cases were included if an epidemiological nosocomial link was confirmed by matching DNA fingerprints, or TB was diagnosed during a contact investigation carried out at the workplace, or a well-documented epidemiological link was present without bacteriological confirmation of the diagnosis in the HCW. Category 2: HCW infected in the community. These cases had either matching DNA fingerprints with a close contact in the community or, if culture negative, were considered infected in the community based on convincing epidemiological evidence, such as contact investigation among household contacts and friends. Category 3: HCW infected abroad. This category includes Dutch HCWs who worked for a long period in a hospital within a TB-endemic country, as well as foreign-born HCWs. Classification was also based on tuberculin skin test (TST) conversion after leaving the Netherlands and on DNA fingerprints (e.g. a unique fingerprint). Category 4: HCW place of infection unknown, including the remaining cases which could not be classified into one of the above categories.
Cases belonging to category one were investigated in more detail by contacting the TB departments of the MHSs and other health institutions involved. Patient and disease characteristics of both the HCW and the index case were obtained from patient records. In addition, information was collected about the circumstances under which transmission occurred. The results were discussed with involved professionals working in these settings. The study design did not allow for collection of standardised information about the airflow (ventilation) in relevant rooms.
RESULTS
In the 5-yr study period, 123 patients were recorded as working in the healthcare/social-welfare sector. Eight cases were misclassified as having TB, seven had a latent TB infection and one case had disease caused by M. avium. Another 21 cases were workers/volunteers involved in the social services in asylum seekers centres, penitentiary institutions and homeless centres who did not have a HCW status. In total, these 29 (24%) cases were excluded from the analysis, leaving a total of 94 eligible HCWs with active TB. However, during the field research, seven additional cases were identified, which were not (yet) included in the present study. The box on the registration form was either not ticked or information was wrongly copied into the national TB register. These cases were included in the current study, giving a total study population of 101 HCWs.
The incidence of TB among HCWs was calculated for hospital workers involved in patient care, since they are at greatest risk for nosocomial infection and their denominator can be more accurately determined. In 1997, 126,500 persons were involved in patient care in Dutch hospitals 13. During the 5-yr study period, 50 cases were hospital-employed HCWs involved in patient care, resulting in a TB incidence of 7.9 per 100,000 per annum. The other 51 cases were employed in nursing homes, home-care organisations or were general practitioners, physiotherapists, student doctors or not employed in the healthcare setting at the time of diagnosis.
Table 1⇓ shows the classification of the cases in the four categories. Of the 47 cases infected in the Netherlands, 28 were work-related (category one) and 19 community acquired (category two). The attributable risk (AR) of healthcare work in the Netherlands can be derived directly by dividing the number of cases in category one by all cases infected in the Netherlands (categories one and two) and is 0.6 (28 out of 47). The relative risk (RR) for healthcare work in the Netherlands can be calculated by AR = (1–1/RR) and is 2.47.
Of the 11 HCWs infected abroad (category three) and born in the Netherlands, all but one case was involved in patient care in developing countries, often for many years. One Dutch HCW had a negative TST when leaving the Netherlands to work in a refugee camp in Kenya and on returning the TST was 6 mm. No treatment for latent TB infection was prescribed in accordance with Dutch guidelines. The patient developed pleural TB within 1 yr of returning, with a M. tuberculosis strain identical to a strain prevalent among Somalian TB patients in the Netherlands.
In-depth investigation focused on the 28 cases infected during work in the Netherlands. Among them, 16 cases were classified based on both an epidemiological link and a matching DNA fingerprint, two cases had an epidemiological link and a unique fingerprint which could be explained (see below), eight cases had culture-negative TB diagnosed in a contact investigation at work, and two HCWs with culture-negative TB had a well-documented contact in the hospital, but were diagnosed after they reported with symptoms. However, five cases were excluded from analysis below since they had no identified patient contact: two HCWs were infected by a regular visitor of the hospital; two HCWs by another HCW; and one laboratory assistant developed cutaneous TB with a unique fingerprint after injuring herself with a needle in a laboratory.
Characteristics of HCWs and index patients
The median (range) age of the HCWs was 28 (21–65) yrs with the majority being female (18 out of 23). The following professions were involved: nurses (n = 14); doctors (n = 4); ward assistants (n = 2); bronchoscopy assistants (n = 2); and an assistant of an outpatient department (n = 1). In total, 21 were infected in the hospital and two outside the hospital. The hospital workers were deployed at the following departments: a pulmonology ward (n = 9); an internal medicine (including AIDS) department (n = 5); a bronchoscopy unit (n = 2); an outpatient department (n = 2); and at other wards (n = 3). In total, 10 cases had pulmonary TB (among them two cases were smear positive), 12 had pleural TB and one had TB of the skin. None of the cases reported a previous history of TB. There were no HCWs with HIV co-infection. However, as the HCWs were not systematically tested for HIV, this information is incomplete. All isolates were drug susceptible and all 23 HCWs completed treatment.
In 21 out of the 23 patients, the infection could be attributed to a specific index case, while in two cases only incomplete information of the presumed index patient could be obtained. Of the known sources, 16 index patients each caused one secondary case among HCWs, one index case was the source of two secondary cases and one index case caused three secondary cases among HCWs. Almost all index patients had smear-positive pulmonary TB, except for two cases: one with disseminated TB with chest radiography abnormalities, a negative sputum culture and a positive stool culture for M. tuberculosis; and one case with a TB abscess of the knee joint that was surgically managed and drained. The median (range) age of the index patients was 45 (25–87) yrs, with 44% of the index patients (eight out of 18) aged >60 yrs. Two index cases were HIV infected.
Interval between infection and diagnosis of TB in HCW
The date of infection was determined as the date of admission of the index case to the hospital or, for nonhospitalised patients, the date of first contact with the HCW. The date of TB diagnosis of the HCW was the date of admission to the hospital, or for nonhospitalised HCWs, the date of specimen collection for TB examination or the first date of presenting with symptoms at a health post. For two HCWs, both infected by the same index patient, the interval could not be determined because the index patient was undiagnosed for a long time, probably >1 yr. One of these HCWs even developed pleural TB 4 months before the index patient was diagnosed.
The interval between infection and disease of TB cases among HCWs is presented in figure 1⇓. The median interval was 32 weeks for all 21 secondary cases and 34 weeks for the 13 culture-confirmed secondary cases. The factors that contributed to infection are summarised in table 2⇓. In 10 index cases, mostly elderly patients with comorbidity, TB was initially not suspected and, thus, adequate isolation was delayed. This diagnostic delay was the main cause of patient-to-HCW transmission, while in some circumstances inadequate isolation measures contributed to infection.
Screening for TB or latent TB infection
Eight HCWs were diagnosed with TB in a contact investigation or pre-employment screening. Three HCWs developed TB despite treatment for latent TB infection (LTBI) with 6 months isoniazid (one was diagnosed later in a pre-employment screening). In three HCWs, LTBI was missed due to a false-negative TST. Results are shown in table 3⇓.
Seven HCWs with TB were not enrolled in a contact investigation or periodical screening. Two of them should have been included in a periodical screening, both employed at a pulmonology ward. Another two should have been included in a contact investigation, but in one situation the index patient refused to mention contacts with health professionals. In the remaining three cases, contact investigation or periodical screening for TB was not indicated and the link with the index patient was only determined by matching DNA fingerprints retrospectively.
DISCUSSION
In the present study, a case series of HCWs with TB, information from a comprehensive national DNA fingerprinting database and detailed epidemiological information from TB departments was used to distinguish nosocomial and non-nosocomial routes of transmission. In 67 out of 101 (66%) HCWs with TB, the route of infection could be determined. Among them, 28 out of 67 (42%) were infected during work in the Netherlands, 19 (28%) were community acquired and 20 (30%) were infected abroad. In 34% of cases, the route of transmission could not be determined, mainly due to lack of bacteriological confirmation.
The TB incidence rate for hospital-employed HCWs with patient contacts (7.9 per 100,000) was approximately two times higher than the incidence rate for Dutch citizens during the study period (average 4.4 per 100,000), but still lower than the rates for all citizens in the Netherlands (average 9.8 per 100,000 with immigrants accounting for >50% of all cases) 10. The present authors also calculated the relative risk for healthcare work in the Netherlands by classifying cases into categories. The RR of 2.47 compares with findings in England and Wales where TB rates in HCWs were two to three times higher than those in similar occupational groups 14. The fact that in one-third of cases the route of transmission could not be determined (category four) probably does not influence the distribution of cases among the three categories and the RR in a significant manner.
Only 20% of all HCWs with TB were foreign born. This differs from studies in low-incidence countries with high percentages of foreign-born HCWs with TB 14–17. Almost all HCWs with work-related TB in the present study developed early manifestations of TB, such as primary pulmonary TB, pleural TB or TB of the skin. Although active-case finding activities such as pre-employment screening and contact investigation detect TB in an early stage, often without bacteriological confirmation, the current authors also found a relatively short interval between infection and disease for HCWs who presented with symptoms. In these cases, awareness of the HCW and a high suspicion of the physician might have limited diagnostic delay, as has been observed elsewhere 14, although in one other study the health-seeking behaviour was similar for HCWs and controls 16.
The reported TB cases among HCWs are the tip of the iceberg of nosocomial transmission of M. tuberculosis bacteria. After all, only 10% of TB infections will eventually lead to active TB 18, so many more infections have occurred. Furthermore, through pre-employment screening, contact investigation and periodical screening of HCWs, a number of latent TB infections are identified (250 annually in the Netherlands) and treated in the majority of cases 10. However, as the present data show, a significant proportion of these infections have been acquired outside the hospital, as has been observed in other studies 19–21. Furthermore, other patients (and visitors) might be at even greater risk if TB is not suspected and diagnosis is delayed. This is well illustrated by the transmission of M. tuberculosis from one index case to three HCWs, two other hospitalised patients and one visitor.
Early recognition of TB and adequate isolation of cases remain the most important interventions to prevent transmission 2, 8, 22, 23. In the current study, diagnostic delay was the main cause of patient-to-HCW transmission in 50% of the cases, often in an elderly Dutch patient with comorbidity. The association between initially missed diagnosis and older age has been described by others 23. Greater experience in TB management, with increased TB suspicion and compliance with diagnostic algorithms, has the potential to reduce diagnostic delay and nosocomial transmission 23. In some cases in the present study, failure to apply with infection-control procedures, such as the use of appropriate masks, contributed to infection. Hospital infection-control measures are relevant to prevent transmission of TB to healthcare workers 18, 23, 24. Although the study design did not allow for a standardised assessment of ventilation during the study period, the relevance of airflow is also well known 21. Four cases in the current study were infected during high-risk procedures, i.e. two HCWs while assisting with bronchoscopies, one laboratory attendant due to a needle stick injury and one HCW during irrigation of a TB abscess (syringing). The occupational risk of these procedures, as well as autopsies, has previously been described in case reports and reviews 2, 8, 23, 25–27. Adequate personal protection measures should be taken during these procedures.
In spite of adequate infection-control measures, transmission of TB might still occur in healthcare institutions, and early diagnosis of latent TB infection or active TB in HCWs is needed in high-risk settings. In the present study, 70% of cases (16 out of 23) had been examined at least once because of high-risk activities or unprotected exposure to a patient at work. The reasons for not detecting and/or effectively treating LTBI were related to the inability of screening procedures to identify infections, the limitations of the TST and failure of (unsupervised) preventive treatment. However, the current study focused on a relatively limited number of HCWs in which infection control measures failed. It needs to be emphasised that a much greater number of infected HCWs have benefited from screening and preventive treatment.
In conclusion, DNA fingerprint surveillance can be used to confirm expected and reveal unexpected cases of nosocomial transmission, thus providing the necessary evidence for policy makers and professionals to take appropriate action. High suspicion of tuberculosis by the clinician, adequate infection-control measures by the hospital authorities and early identification of latent tuberculosis infection in healthcare workers by occupational specialists form the essential components of a comprehensive package to prevent tuberculosis in healthcare workers.
Acknowledgments
The authors would like to thank the public-health nurses and physicians of the Depts of Tuberculosis Control of the Municipal Health Services who participated in this study.
- Received March 21, 2006.
- Accepted June 14, 2006.
- © ERS Journals Ltd