Tuberculosis (TB) results from rapid progression of a recently acquired Mycobacterium tuberculosis infection or from reactivation of a remote infection (1). It is critical that recent M. tuberculosis infections be identified because TB is more likely to develop in persons with recent infections (2). Furthermore, the finding of recently infected persons suggests ongoing transmission of TB, which can be interrupted by prompt identification and treatment of undiagnosed cases (3). However, finding undiagnosed cases remains a challenge (4). Name-based contact investigations have traditionally been used for this purpose, but such investigations are resource-intensive (5), making them less practical in countries to which TB is endemic. Genotyping of M. tuberculosis has emerged as a complementary method to detect ongoing transmission because persons who have the same TB genotype may be involved in the same chain of transmission (6). Although this assumption is relatively reliable in low-incidence countries, it is yet to be determined whether genotyping in TB-endemic settings can similarly detect ongoing transmission.

We investigated a TB cluster of 24 patients with matching M. tuberculosis genotypes in Gaborone, Botswana, a city with a high number of TB cases (7). Because almost half of these patients received care at the same hospital, nosocomial transmission was suspected. We conducted an investigation to determine if TB transmission occurred among these patients within the hospital and to identify possible alternate sites of ongoing transmission of this TB strain.

The Study

During August 2012–January 2015, all consenting persons with TB at 26 facilities in Gaborone provided sputum samples for culture as part of the Kopanyo study (8). M. tuberculosis isolates were genotyped by 24-locus mycobacterial interspersed repetitive units–variable number tandem repeats (9).

We assessed nosocomial transmission by reviewing dates of hospital visits for overlap among the 24 TB cluster-associated patients. The hospital’s electronic billing system was used to obtain all previous dates of admission, discharge, and visits to the accidents and emergency department that had occurred for these patients since 2004.

We interviewed each patient, using an investigation form (Technical Appendix), to learn their primary residence; contacts; places of work and worship; and other frequented locations, including bars and combi (minibus) routes used in the 6 months before diagnosis. Primary residences of patients were mapped by using global positioning system coordinates (8). Ethical approval was obtained from the University of Pennsylvania, US Centers for Disease Control and Prevention, Botswana Ministry of Health, and University of Botswana.

We looked for epidemiologic links that might suggest ongoing transmission. An epidemiologic link was defined for patients having at least 1 of the following associations: overlapping visits at the hospital, living within 1 km of another cluster-associated patient (spatial link), frequenting the same locations as another cluster-associated patient, and naming another cluster-associated patient as a contact.

During the study, ≈60% of M. tuberculosis isolates from reported TB patients in Gaborone were genotyped. The cluster discussed in this report includes 24 (2.3%) of 1,033 total genotyped cases from Gaborone.

Figure 1

Figure 1. Timing of hospital visits and treatment for 10 tuberculosis (TB) cluster–associated patients, Gaborone, Botswana, 2013–2015. Patients were hospitalized or seen in the accident and emergency ward, and all had a history...

All patients had pulmonary disease involvement (Table 1). Ten (42%) had received care at the hospital since 2004; most visits occurred after October 2013 (Figure 1). Except for visits by 2 patients, no patients’ visits overlapped at the hospital. Patients V and X overlapped in the hospital for 3 days, albeit in separate buildings. Patient V was admitted to the hospital with a known diagnosis of TB and had started TB therapy the day before admission. Patient X was in the hospital for a week but did not start TB therapy until 13 days after patient V was admitted. No members of this cluster were healthcare workers.

Figure 2

Figure 2. Residence-associated data for patients in a tuberculosis cluster, Gaborone, Botswana, 2012–2015. A) Primary residences of 20 patients are indicated by red dots. Inset map shows location of Gaborone in Botswana. Black...

Twenty (83%) patients had a primary residence in Gaborone; 14 resided in 4 distinct neighborhoods (Figure 2, panel A). The spatial link definition (residing <1 km from another cluster-associated patient) was met by 13 (54%) patients. One patient (B) lived slightly >1 km from the nearest patient (Figure 2, panel B). Two of 4 spatially linked patients (C, D, I, and W) (Figure 2, panel B) did not name each other during enrollment when asked about contacts; when interviewed again as part of this investigation, these patients confirmed spending time together around the time of their diagnosis.

One patient died and 2 were lost to follow-up. Thus, 21 patients contributed to the remaining epidemiologic links obtained from interviews (Table 2). The median number of specific locations reported by patients was 6 (range 1–13). Of the 21 patients, 11 (52%) reported attending the same bar as another patient; 1 of these patients worked as a bartender. Site visits to 2 of the bars revealed an environment conducive to TB transmission (closed, poorly ventilated space crowded with patrons and employees). Similar crowded conditions were observed in combis, and 16 (76%) patients reported using the same combi route as another patient. Eight (38%) patients attended the same church as another patient, and 6 (29%) patients named each other as a contact, suggesting transmission could have occurred among them.

Top