Persons with active TB were considered homeless if they lived in one of the shelters or residential substance-treatment centers or were reported as homeless in clinic records between 1988 and 1998. Data were available for 1995 to 1998 on whether TB diagnoses were detected by the screening program or by staff at a medical facility. We obtained the numbers of homeless persons from studies done in 1995 and 1998 by the Colorado Department of Human Services (7). These studies enumerated the persons living in shelters and residential substance treatment centers and also estimated the number of persons living in other temporary housing and on the streets. Data for homeless populations in 1996 and 1997 were not available, so we estimated this number, assuming a linear increase in the total. We analyzed TB incidence for the 11-year period, from 1988 (the year that isolates were first available for DNA fingerprinting) through 1998, compiling data for 7 years before and 4 years after the screening program was implemented for the homeless in 1995.

We analyzed identical fingerprinting patterns in homeless patients from 1988 through 1998. Initial DNA fingerprinting used the IS6110 technique, and secondary fingerprinting used pTBN12 for isolates having five or fewer hybridizing bands (4,8). To estimate recent TB transmission in the homeless population, we calculated the proportion of cases clustered within a 2-year period of a preceding case with the identical DNA fingerprint (9). We compared the proportion of cases caused by recent transmission during the 7-year period before the program (1988–1994) with the proportion caused during the first 4 years of the program (1995–1998) to assess the effect of the screening program on TB transmission.

Analyses were done by using Epi Info, version 6.0 (Centers for Disease Control and Prevention, Atlanta, GA). The proportion clustered in the two time periods was compared by using the Fisher exact test.

The institution of a TB-screening program for the homeless population in congregate settings in Denver in 1995 was associated with a substantial decrease in the incidence of active TB in this high-risk population. The mechanisms by which the program could have caused this reduction in TB incidence include treating active cases earlier and treating persons with latent TB infection, thus preventing future active TB cases. Treating persons with latent TB infection was unlikely to have played a major role, however, since we found low rates of starting and completing treatment for this type of infection, a finding noted in previous reports (1,10).

The screening program more likely reduced TB transmission by earlier diagnosis and treatment of active TB cases (the top priority activity in TB control). The program was effective in finding TB patients, detecting over half the active TB cases reported in homeless persons from 1996 to 1998. In contrast to the low acceptance and completion of latent TB infection treatment, high rates of treatment completion for active TB cases can be achieved with directly observed therapy supported by public health regulations (11,12). The results of the DNA fingerprint analysis are consistent with the conclusion that early detection of TB through the screening program leads to a reduction in transmission. New, unique strains of M. tuberculosis have continued to appear among the Denver homeless population, but cases from recent transmission (clustered in the homeless population within 2 years) decreased. The estimated number of active TB cases decreased despite an increasing population of homeless persons, leading to a drop in the annual incidence rate from 510 to 121 per 100,000 persons.

Other potential explanations for the reduction in TB deaths in the homeless population include failure to detect subsequent cases and errors in population estimates. The mandatory TB screening program for residents in shelters and residential substance treatment programs may have led some homeless persons to live in other settings or to leave Denver; the latter reason led to the diagnosis of subsequent cases in other jurisdictions. However, the Denver Metro TB Clinic receives reports of cases diagnosed in Metropolitan Denver and conveys them to the TB Control Program at the Colorado Department of Public Health and Environment; cases were not likely missed unless diagnosed outside Colorado. Obtaining an accurate count of homeless persons is difficult so we based our denominator in this study upon estimates obtained by the Colorado Department of Housing (7). Even if the population at risk had remained the same, despite indications otherwise, the incidence rate would still have decreased from 510 to 210 (rather than 121) per 100,000 persons per year from 1995 to 1998. We detected substantial TB transmission in the homeless population for only two of nine homeless clusters, raising the possibility that the increase in cases was the result of one or two outbreaks running their natural course. Further studies with larger numbers of cases are needed to address this issue; however, our data support the concept that early case detection results in lower rates of transmission in homeless populations.

The possibility exists that clusters that occurred over several years were strains that lingered in the community but were not detected in our study; we obtained DNA fingerprints for only 55% of the cases in the community among nonhomeless persons and cannot rule out that possibility. For clustered cases occurring many years apart, we may have missed links in transmission because of patients who were not tested in the community or who moved out of the area without diagnosis. As we increase our understanding of how to interpret the results of DNA fingerprinting, we may find that clusters that are far apart represent a remote transmission that resulted in latent TB infection with later reactivation. The secondary cases would then share the same pattern as the original (source) case, but the connection between them would be more difficult to identify since the contact in these cases occurred many years earlier.

Of the 42 homeless cases with unique patterns, seven (17%) were diagnosed in foreign-born persons. Of those, one patient arrived in United States <2 years before TB diagnosis, three 2–5 years before diagnosis, two 5–10 years before diagnosis, and one for an unknown period of time. Few of the unique patterns found among homeless patients represent cases imported from foreign countries.

Various screening programs have been proposed to decrease TB transmission among homeless persons (13). Chest radiographic screening provides a rapid assessment for potentially transmissible disease, but this method would have been too expensive for us to implement. Sputum examination by smear has been used (14) but does not detect smear-negative cases and is difficult to implement for large numbers of persons. Tuberculin skin testing does not provide an immediate assessment of infectious TB and gives negative test results for up to 25% of persons with active pulmonary TB. Furthermore, skin testing requires a second visit 2–3 days later for a reading.

Despite reaching only two thirds of the homeless population in 1998, the screening program was notably successful. The true incidence of skin test positivity is likely higher than that shown in Table 1 because those already known to be positive were not tested. The symptom survey was an important component of our program. In spite of the challenges of skin testing, our screening program proved to be feasible for a large population staying at a number of shelters and residential programs. Our results suggest that congregate settings, the focus of our program, are the most important loci of TB transmission among homeless persons.

The low completion rate of treatment for latent TB infection provided further evidence that the primary effect of the program was to increase early detection of active cases of TB, not to identify and treat latent TB infection. Despite the decrease in TB incidence following the screening program, the TB rate remained high (121 per 100,000 population) compared to the overall rate for the metropolitan area (8.2 per 100,000 population). Further progress in decreasing TB among homeless persons will require improvements in the acceptance and completion of latent infection treatment (6).