Haddad MB, Wilson TW, Ijaz K, Marks SM, Moore M. Tuberculosis and Homelessness in the United States, 1994-2003. JAMA. 2005;293(22):2762-2766. doi:10.1001/jama.293.22.2762
Author Affiliations: Division of Tuberculosis
Elimination, Centers for Disease Control and Prevention, Atlanta, Ga.
Context Tuberculosis (TB) rates among US homeless persons cannot be calculated
because they are not included in the US Census. However, homelessness is often
associated with TB.
Objectives To describe homeless persons with TB and to compare risk factors and
disease characteristics between homeless and nonhomeless persons with TB.
Design and Setting Cross-sectional analysis of all verified TB cases reported into the
National TB Surveillance System from the 50 states and the District of Columbia
from 1994 through 2003.
Main Outcome Measures Number and proportion of TB cases associated with homelessness, demographic
characteristics, risk factors, disease characteristics, treatment, and outcomes.
Results Of 185 870 cases of TB disease reported between 1994 and 2003,
11 369 were among persons classified as homeless during the 12 months
before diagnosis. The annual proportion of cases associated with homelessness
was stable (6.1%-6.7%). Regional differences occurred with a higher proportion
of TB cases associated with homelessness in western and some southern states.
Most homeless persons with TB were male (87%) and aged 30 to 59 years. Black
individuals represented the highest proportion of TB cases among the homeless
and nonhomeless. The proportion of homeless persons with TB who were born
outside the United States (18%) was lower than that for nonhomeless persons
with TB (44%). At the time of TB diagnosis, 9% of homeless persons were incarcerated,
usually in a local jail; 3% of nonhomeless persons with TB were incarcerated.
Compared with nonhomeless persons, homeless persons with TB had a higher prevalence
of substance use (54% alcohol abuse, 29.5% noninjected drug use, and 14% injected
drug use), and 34% of those tested had coinfection with human immunodeficiency
virus. Compared with nonhomeless persons, TB disease in homeless persons was
more likely to be infectious but not more likely to be drug resistant. Health
departments managed 81% of TB cases in homeless persons. Directly observed
therapy, used for 86% of homeless patients, was associated with timely completion
of therapy. A similar proportion in both groups (9%) died from any cause during
Conclusions Individual TB risk factors often overlap with risk factors for homelessness,
and the social contexts in which TB occurs are often complex and important
to consider in planning TB treatment. Nevertheless, given good case management,
homeless persons with TB can achieve excellent treatment outcomes.
Numerous tuberculosis (TB) outbreaks have originated among homeless
persons in the United States.1- 4 Homelessness
is associated with an increased risk of exposure to Mycobacterium
tuberculosis,5 undetected and untreated
infection, and subsequent progression to TB disease.6 Homeless
persons with TB are more likely to be hospitalized, and to be hospitalized
longer, than nonhomeless persons with TB; the public sector pays most of those
Better understanding of the characteristics of homeless persons with
TB disease is important for creating strategies to reduce TB incidence in
this high-risk population in the United States. In 1993, the Centers for Disease
Control and Prevention (CDC) standardized national monitoring of TB disease
among homeless persons by asking health departments to indicate whether annually
reported TB cases occurred in homeless persons. Homeless status was reported
incompletely (ie, 70%) that first year but increased to at least 90% each
subsequent year. Thus, 1994 through 2003 represents the first full decade
of national TB surveillance that includes an assessment of homelessness. This
article describes homeless TB patients and compares risk factors and disease
characteristics between homeless and nonhomeless TB patients from 1994-2003.
The US Census Bureau does not enumerate the national homeless population,
so we were unable to use TB surveillance data to determine disease rates among
homeless persons. Instead, we calculated the proportion of all reported TB
cases that occurred in homeless persons.
Health departments have been reporting all TB cases into the US National
TB Surveillance System at the CDC since 1953. The system expanded in 1993
to permit collection of additional information (eg, homeless status, human
immunodeficiency virus [HIV] status, anti-TB drug resistance) and has not
changed since then. Cases must meet the surveillance case definition (ie,
laboratory or clinical evidence of active disease due to M tuberculosis complex) to be included in the national count.8 Standard case reports to the surveillance system also
indicate whether patients have a previous diagnosis of TB disease (ie, >12
months before current diagnosis) but not history of vaccinations, latent TB
infection, or any TB treatment. Molecular epidemiology results are not currently
Our analysis included all cases reported from the 50 states and the
District of Columbia from January 1, 1994, through December 31, 2003, based
on reports submitted to the CDC as of March 30, 2004. In case reports, homelessness
is defined by any of the criteria specified by the McKinney Act (ie, the lack
of regular access to a conventional, nontemporary nighttime residence) during
the 12 months before diagnosis.8,9 Case
reports also note whether persons were residents of a correctional facility
on the day of their TB diagnosis. Therefore, no one incarcerated continuously
for more than 12 months could also be reported as homeless, but more recently
incarcerated persons might be reported as both homeless and incarcerated.
Homeless and nonhomeless persons with TB were compared on the basis
of demographic characteristics, including sex, age, country of birth, and
self-reported ethnicity/race. Stratifying TB incidence by ethnicity/race is
valuable because disparities in TB case rates have persisted.10
We examined self-reported excessive alcohol use and any unprescribed
drug use, either injected (ie, intravenous, subcutaneous, or intramuscular)
or noninjected (ie, ingested, inhaled, or smoked), during the 12 months before
TB diagnosis. We calculated prevalence ratios (PRs) with 95% confidence intervals
(CIs) to compare substance use between homeless and nonhomeless persons with
TB.11 Because the age distribution of homeless
patients was centered, we adjusted for the effect of age by recalculating
each PR with the subset of persons aged 30 to 59 years.
All states reported positive HIV test results, and every state except
California also reported negative results to the CDC. After excluding California
and including only reports with HIV test results (ie, positive, negative,
or indeterminate) in the denominator, we calculated the prevalence of documented
We examined the infectiousness of TB disease by ascertaining the frequency
of pulmonary disease that was also culture-confirmed or had a positive sputum
smear for acid-fast bacilli.12 In addition,
the frequency of cavitary pulmonary disease, a sign of advanced TB, was reviewed.
We noted differences in the susceptibility of initial cultures to first-line
anti-TB medications (ie, isoniazid, rifamycins, pyrazinamide, and ethambutol).
Primary isoniazid resistance was defined as resistance to isoniazid, and primary
multidrug-resistant TB was defined as resistance to at least isoniazid and
rifampin in patients without a previous history of TB.8 We
also examined drug resistance patterns among patients with a previous TB diagnosis.
Because drug-resistant TB strains can be associated with overseas TB treatment,
we stratified patients with previous TB into those born inside or outside
the United States and recalculated for each stratum the prevalence of drug
resistance by homeless status.
Treatment guidelines for TB did not substantially change during the
10-year period of this study. For evaluating treatment and outcomes, analysis
was restricted to 1994-2001 because health departments continue to collect
and submit those data for up to 2 years after reporting a case. We noted who
provided care (ie, health department, private clinician, or both) and type
of treatment (ie, whether patient received a recommended initial 4-drug anti-TB
regimen and/or directly observed therapy [DOT]).13 For
all patients who began an anti-TB drug regimen, we examined whether treatment
was eventually completed (regardless of length of time). The Healthy People
2010 initiative of the US Department of Health and Human Services sets the
goal that 90% of TB patients will complete therapy within 1 year.14 Using previously described criteria for evaluating
the timeliness of treatment completion,8 we
excluded all patients who died or whose therapy might be reasonably expected
to take longer (eg, rifampin resistance).
Analysis of the deidentified data from the National TB Surveillance
System was performed with SAS 8.02 (SAS Institute, Cary, NC) and Microsoft
Excel 2002 (Microsoft, Redmond, Wash). There was greater than 85% completeness
of reporting on each analyzed variable except for HIV status. The χ2 test for comparisons between the homeless and nonhomeless groups consistently
yielded P values that were ≤.01 (as would be expected
given the data set size), so those are not individually listed.
Patients were categorized as either homeless or nonhomeless in 178 517
(96%) of the 185 870 total TB case reports sent to the CDC from the 50
states and the District of Columbia for 1994 through 2003. The 7353 reports
missing information on homeless status were excluded from all subsequent analyses.
The annual number of cases of TB reported to occur in homeless persons
decreased from 1392 in 1994 to 913 in 2003, while the total number of TB cases
decreased from 24 205 to 14 874. The proportion of TB cases associated
with homelessness was stable, ranging annually from 6.1% to 6.7% (overall
mean: 6.4%). Regional differences were apparent, with western and some southern
states reporting a higher proportion of cases associated with homelessness
Characteristics of patients with TB according to homeless status during
the 12 months before diagnosis are summarized in the Table. Males accounted for 87% of TB cases in homeless persons during
the 1994-2003 period. This proportion was stable over time. Among nonhomeless
TB patients, male predominance was less marked (61%). The age distribution
of homeless persons at the time of TB diagnosis was centered (ie, 82% of persons
were aged 30-59 years), with a slight aging trend from a median age of 41
years in 1994 to 46 years in 2003. Unsurprisingly, the age distribution of
nonhomeless TB patients was wider (ie, 47% were aged 30-59 years), with the
median essentially unchanged (44-45 years) during the 10-year period.
Among all reported TB cases, a higher proportion of nonhomeless were
born outside the United States (44% vs 18%). However, persons born outside
the United States accounted for a growing proportion of TB among homeless
persons (from 14% in 1994 to 23% in 2003), with a mean of 11 years since immigration.
Mexico was the country of birth for 47% of the homeless persons with TB who
were born outside the United States.
Among all reported cases of TB disease during the study period, black
individuals represented the highest proportions among the homeless and nonhomeless.
Among homeless TB patients from 1994-2003, the annual number who were white
declined from 415 to 254. Outside the South, the annual number who were black
also decreased from 450 to 155, but numbers in the southern states remained
constant (ie, 207 blacks in 1994 and 2003). During this period, the annual
number of Hispanic homeless patients with TB decreased from 256 to 215, although
their number as a proportion of total homeless TB patients increased from
18% to 24%.
Nine percent of the TB cases (n = 1032) associated with homelessness
were diagnosed in persons who were residents of correctional facilities on
the day of their TB diagnosis. The type of facility was usually a local jail
(83%), with a smaller proportion in a state or federal prison (13%). Three
percent (n = 5228) of nonhomeless persons with TB were incarcerated
on the day of their diagnosis.
The prevalence of excessive alcohol use was 4.6 times greater among
homeless (54%) than among nonhomeless (12%) persons with TB (95% CI, 4.5-4.7).
Homeless TB patients also had a higher prevalence of noninjected (29.5% vs
5%; PR, 5.6; 95% CI, 5.4-5.8) and injected (14% vs 2%; PR, 6.7; 95% CI, 6.4-7.1)
drug use. All PR point estimates remained greater than 3 when this comparison
was restricted to patients aged 30 to 59 years.
Homeless TB patients during 1994-2003 were more likely than nonhomeless
TB patients to obtain an HIV test (76% vs 53%). Among TB patients with HIV
test results reported into the National TB Surveillance System (California
excluded), TB-HIV coinfection was higher among homeless (34%) than nonhomeless
(20%) patients. A difference persisted (35% vs 29%) when analysis was limited
to the 30- to 59-year-old age group.
Compared with nonhomeless patients, homeless patients were more likely
to have pulmonary disease (93% vs 81%), and those with pulmonary disease were
more likely to have culture-confirmed or smear-positive disease. The prevalence
of cavitary pulmonary disease was also slightly higher among homeless persons
(29% vs 25%). However, neither homeless persons with a primary episode of
TB nor homeless persons with a previous diagnosis of TB were more likely than
nonhomeless persons to have a drug-resistant strain of M tuberculosis. Although homeless persons were more likely to have
a previous history of TB, nonhomeless persons with a previous history of TB
were about twice as likely to be infected with a drug-resistant strain 5.3%
vs 2.7%), even after controlling for country of birth.
Health departments managed 81% of the TB cases associated with homelessness
during 1994-2001. Most homeless TB patients received an initial regimen of
4 anti-TB drugs (81%) and were managed with DOT (86%).
Fewer homeless (77%) than nonhomeless (84%) TB patients eventually completed
TB treatment. A similar proportion in both groups died during therapy (9%),
but more homeless prematurely stopped treatment for reasons such as moving
away or being lost to follow-up (12%, in contrast to 7% of nonhomeless). However,
among homeless patients, those who received DOT were 1.4 times (95% CI, 1.3-1.5)
more likely to complete therapy in 1 year or less, compared with those who
did not receive DOT.
Homelessness has persistently cultivated a “pocket of endemicity”
for TB,15 which continues to be associated
with poverty.16,17 Although the
number of TB cases associated with homelessness during the 1994-2003 decade
decreased, certain states had higher proportions of TB cases associated with
homelessness, and the annual number of black homeless patients with TB in
southern states was unchanged.
Of note, risk factors for TB overlap with many of the risk factors associated
with persistent homelessness (eg, being male, or having a history of incarceration
or substance abuse).18 Excessive alcohol and
illicit drug use may have importance at several points across the infection-to-disease
continuum. For example, social settings in which alcohol or drugs are shared
may facilitate transmission of M tuberculosis.19,20 Alcohol17,21 and
inhaled22,23 or injected23 drugs enhance susceptibility to respiratory tract
infections. Even in the pre-HIV era, drug dependency was associated with more
rapid progression to TB disease.24 Untreated
HIV infection remains an important risk factor for TB.6,12 Testing
for HIV is recommended for all homeless shelter clients25;
new rapid tests, with results available within 20 minutes, can be done in
Our surveillance system could not capture all the variables of interest
for understanding the complex social contexts in which TB often occurs. For
example, we were unable to examine the relationship of socioeconomic indicators
such as education and income variables with homeless status or ethnicity/race.
In addition, psychiatric disorders are frequent among homeless persons,16,18,27 but we had no data
to assess the prevalence of mental health problems among homeless TB patients.
These are important considerations, not only in understanding the epidemiology
of TB, but also in planning its treatment, among homeless persons.
The most urgent priority for controlling TB in the United States is
interrupting new transmission of M tuberculosis.28 Opportunities for transmission arise when homeless
persons with infectious TB frequent homeless shelters,1- 4 emergency
departments,29 and jails.2,30 Locations
in a particular community where homeless persons typically congregate are
appropriate places to focus resources to prevent, identify, and treat TB.
Nine percent of homeless TB patients were residents of a correctional facility
on the day of their TB diagnosis. The burden of TB that is associated with
the incarceration of homeless persons is likely much higher, but no further
incarceration history is recorded for TB surveillance purposes. The 1996 National
Survey of Homeless Assistance Providers and Clients found that 49% of homeless
persons reported at least 1 incarceration of at least 5 days in a jail, and
18% reported incarceration in a prison, at some point during their lifetime.27 Although a history of incarceration is a well-documented
risk factor for TB,6,31 it might
also be regarded as a unique opportunity for more outreach in this hard-to-reach
population. Local public health staff could collaborate with correctional
facility staff (ie, for staff training, information sharing, and discharge
planning) to initiate supervised treatment of latent TB infection and TB disease.32
We had no means to determine the extent to which homeless persons experienced
delays in obtaining an evaluation for TB, but the slightly higher prevalence
of cavitary pulmonary TB suggests more advanced disease at the time of diagnosis.
Once diagnosed, however, homeless TB patients received good case management,
including laboratory diagnostic evaluation, appropriate use of a 4-drug regimen,
and excellent treatment outcomes for persons given DOT (recommended for all
TB patients). Controlling this public health problem demands considerable
resources but is integral to responding to the Institute of Medicine’s
call to eliminate TB in the United States.28
Corresponding Author: Maryam B. Haddad,
MSN, MPH, FNP, Surveillance, Epidemiology, and Outbreak Investigations Branch,
Division of Tuberculosis Elimination, Centers for Disease Control and Prevention,
1600 Clifton Rd, Mailstop E-10, Atlanta, GA 30333 (email@example.com).
Author Contributions: Ms Haddad had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Haddad, Wilson, Ijaz,
Acquisition of data: Wilson, Moore.
Analysis and interpretation of data: Haddad,
Drafting of the manuscript: Haddad, Wilson.
Critical revision of the manuscript for important
intellectual content: Haddad, Wilson, Ijaz, Marks, Moore.
Statistical analysis: Haddad, Wilson, Marks.
Administrative, technical, or material support:
Study supervision: Ijaz.
Financial Disclosures: None reported.
Funding/Support: All financial and material
support for the design and conduct of this study; collection, management,
analysis, and interpretation of the data; and preparation, review, and approval
of the manuscript were provided by the Division of Tuberculosis Elimination
at the Centers for Disease Control and Prevention. All the authors are full-time
employees of the US Department of Health and Human Services. Analysis of data
in the National TB Surveillance System is part of the authors’ usual
Previous Presentation: Presented in part at
the 132nd Annual Meeting of the American Public Health Association, Washington,
DC, November 6-10, 2004.
Acknowledgment: We gratefully acknowledge the
staff from local and state health departments throughout the United States
who have collected and reported data about TB cases for more than 50 years.
We also thank Martha Burt, PhD, for assistance with understanding the National
Survey of Homeless Assistance Providers and Clients; Robert H. Pratt, BSIE,
for support with SAS coding; the National TB Surveillance System Analytic
Steering Committee for guidance with study objectives and design; and Jose
Becerra, MD, MPH, Lauren Lambert, MPH, and Ann Lanner, BA, for review of the