Epidemiology of Human Immunodeficiency Virus Testing among patients with Tuberculosis in North Carolina.

ABSTRACT

Background. Human immunodeficiency virus (HIV) testing is recommended for all patients with tuberculosis (TB).

Methods. Surveillance data for all reported cases of TB in North Carolina from 1993 to 1999 were examined to assess HIV testing practices.

Results. Of 3,680 TB patients, 3,119 (85%) had HIV testing data reported. Of these, 604 (19%) were not offered HIV testing, 465 (18%) refused testing, 379 (15%) were HIV seropositive, 29 (0.8%) were tested but results were not reported, and 1 (0.03%) had an indeterminate result. Older patients were significantly less likely to be offered HIV testing and more likely to refuse testing. Males and African Americans were more likely to be offered and to accept testing.

Conclusions. At least 34% of TB patients in North Carolina from 1993 to 1999 did not receive HIV testing. Patients in higher-risk groups were more likely to be tested, but even within the highest-risk groups, testing was not universal. Health care providers should offer HIV testing to all individuals with TB.

Infection with the human immunodeficiency virus (HIV) is a common and serious comorbidity among people with tuberculosis (TB). (1) The Centers for Disease Control and Prevention (CDC) recommended in 1989 that all patients with TB should be offered testing for HIV. (2) Despite these recommendations, recent studies (3,4) have shown that only 35% to 63% of patients with TB are being tested for HIV. These data suggest that health care providers are attempting to selectively test TB patients for HIV according to the providers' perceptions of patients risk of HIV infection. Because physicians often miss important risk factors for HIV, (5) this strategy is considerably less sensitive and efficient than universal testing of TB patients, who are at high risk for coinfection with HIV.

We examined TB surveillance data from North Carolina to determine how well providers are adhering to the recommendations for HIV testing among patients with TB. To further understand patterns of HIV testing in TB patients, we examined demographic characteristics that were associated with HIV testing and HIV seropositivity in this cohort.

METHODS

We examined surveillance data for all cases of TB reported to the state health department from January 1, 1993, to December 31, 1999 (N = 3,680). The Tuberculosis Information Management System (TIMS) database consists of data abstracted from the Report of Verified Case of Tuberculosis, which is the standard data form for reporting information about a case of TB to the CDC in Atlanta, Georgia. The data include demographics, site of disease, diagnostic information, and treatment. The information is collected by review of medical records and patient interviews by county health department employees and is forwarded to the state health department. The data are then forwarded to the CDC on a weekly basis. The HIV status is recorded on the form as one of seven values: negative, positive, indeterminate, refused testing, not offered testing, test done with results unknown, or unknown.

The HIV seroprevalence data for North Carolina was extracted from the North Carolina HIV/Sexually Transmitted Diseases (HIV/STD) Prevention and Care HIV/AIDS Reporting System database. Health care providers report all persons diagnosed with HIV infection or AIDS to the HIV/STD Prevention and Care Branch. Demographic, clinical, laboratory, and risk factor data are collected as part of the reporting process. Morbidity is assigned by county of residence at the time of diagnosis, regardless of where the patient receives medical care. A patient potentially may be reported to the database once as a case of HIV infection and once as a case of AIDS. Patients are counted only once in the database, even if they are reported multiple times. Total HIV prevalence is calculated using the sum of all living persons with HIV infection (non-AIDS) and AIDS.

We sought to determine which groups of TB patients, if any, were more likely to participate in HIV testing and were likely to have HIV infection. Because the TIMS database does not collect information on many HIV risk factors (apart from illicit drug use), we elected to use demographic variables that could serve as partial surrogates for these risk factors. We hypothesized that the demographic information most likely to correlate with HIV risk factors included age (linear and quadratic terms), sex, race/ethnicity (separated into five categories: non-Hispanic white, non-Hispanic black, Hispanic, Native American, or Asian/Pacific Islander), whether a patient had received any directly observed therapy, and foreign birth. A patient's use of illicit drugs within the past year was categorized as either injection or noninjection drug use, with possible responses "Yes," "No," or "Unknown." In addition, the average yearly number of cases of HIV/AIDS in a patient s county of residence from January 1, 1993, to Septembe r 30, 1999, was divided by the estimated population of the county in 1998 (6) (the latest year for which estimates were available) to provide a measure of the average HIV incidence within each county during the study period. To assess secular trends in HIV testing, the year in which a patient was reported as a TB case was also included as a predictive variable.

We examined the associations between these variables and three separate outcomes: the likelihood of being offered an HIV test, the likelihood of refusing HIV testing once offered, and the likelihood of testing positive for HIV if testing was accepted. Associations between categorical variables and the outcome of interest were assessed with the chisquare or Fisher's exact test, as appropriate. Continuous variables were analyzed using Student's t test or Wilcoxon signed rank sum test, as appropriate. Multivariable analysis was done using unconditional logistic regression. In all of the models, age (both linear and quadratic terms) and the year in which a patient was reported with TB were used as continuous variables. Race was coded by the use of dummy variables for each racial/ethnic group, with white as the baseline case. Illicit drug use was coded by two variables: injection drug use and noninjection drug use. If a patient's use of illicit drugs was recorded as "unknown," then this fact was coded by use of a separate dummy variable for injection and noninjection drug use. Incidence of HIV in the county of residence was separated into quartiles and represented by dummy variables, with the first quartile defined as the baseline. All P values are 2-sided, and 95% confidence intervals (CIs) are reported for all odds ratios. A P value of .05 or less was considered statistically significant. All statistical analysis was done with SAS version 6.12 (SAS Institute, Cary, NC).

RESULTS

A total of 3,680 patients were reported as confirmed cases of TB between January 1, 1993, and September 30, 1999. The cohort was predominantly male and African American, with a median age of 51 years (interquartile range 38 to 72, range 1 month to 106 years) (Table 1). Of the 3,119 patients for whom data were available, 2,515 (81%) were offered HIV testing (Fig 1). Of this group, 465 (18%) refused testing, and another 30 (1.2%) had an indeterminate or unknown result. This left 2,020 patients with known HIV status.

The number of patients reported with HIV among counties in North Carolina varied widely, ranging from a total of 0 cases reported from January 1, 1993, to September 30, 1999, to 1,474 cases (median 31 cases, interquartile range 8 to 72). This corresponded to an estimated yearly incidence of between 0 cases! 100,000 person-years to 34.8 cases/100,000 person-years (median 10.3, interquartile range 4.1 to 17.5). Of all patients reported with HIV/AIDS in North Carolina up to September 30, 1999, 21% were white non-Hispanic, 76% black non-Hispanic, 1% Hispanic, 1% Native American, and the remainder (<1%) were of Asian/Pacific Islander descent. One percent of patients were less than 5 years old at the time of reporting, 3% were 5 to 19 years old, 70% were 20 to 39 years old, 19% were 40 to 49 years old, and 6% were 50 or older.

The proportion of TB patients offered an HIV test by 10-year age group is shown in Figure 2. In logistic regression analysis, both age and age squared significantly predicted whether patients were likely to be offered HIV testing (P < .0001 for both terms). This produced an inverted U-shaped curve with a predicted maximum likelihood to be offered testing at age 40. The proportion of TB patients who were known to have been offered an HIV test increased each year from 48% in 1993 to 82% for 1999 (chi-square for trend = 208, P < .0001). Even after adjusting for changes in demographic risk factors and excluding patients whose HIV testing status was classified as "unknown," a greater proportion of patients were offered HIV testing during each year of the study (odds ratio = 2.7 for TB patients reported in 1999 vs 1993, 95% CI 1.9-3.8). Other factors that were associated in multivariable analysis with being offered an HIV test included black race, receiving any directly observed therapy, noninjection drug use, and being in a county in the third or fourth quartile of average HIV incidence (Table 2).

Of patients offered HIV testing, 18.5% refused to have the test. The proportion of patients who refused testing in each 10-year age group is shown in Figure 2. In logistic regression modeling, both age and age squared significantly predicted whether a patient would refuse testing (P < .0001 for both terms). In this model, a 37-year-old patient was most likely to accept testing, while younger and older patients were significantly less likely to accept testing when offered. Other factors that predicted patients would accept HIV testing when offered were male sex, black race, and noninjection drug use (Table 2). Patients whose injection drug use status was unknown were also more likely to accept HIV testing than patients who had not used injection drugs in the past year. There was no consistent relationship between quartile of HIV incidence in the patient's county of residence and likelihood to accept HIV testing. Interestingly, patients were slightly less likely over time to accept HIV testing when offered; 90 % accepted testing in 1993 compared with 83% in 1999 (chi-square = 40.3, P < .001). This trend disappeared when adjustment was made for other demographic and risk factors; the odds of accepting HIV testing in 1999 compared with 1993 in the multivariable model were 0.8 (95% CI, 0.5-1.2).

Among the TB patients for whom HIV status was known, 379 (18.8%) were HIV seropositive. The proportion of patients who were HIV seropositive by age group is shown in Figure 3. Again, both age and age squared were significant predictors of a positive HIV test (P < .0001 for both terms). The age at which patients had the highest likelihood of a positive HIV test was 37. Male sex, black race, and living in a county in the fourth quartile were all significantly associated with testing positive for HIV, as were both injection and noninjection drug use. Asians were significantly less likely to test positive for HIV than whites (Table 2). Among TB patients who were tested in this cohort, the likelihood of testing positive for HIV decreased significantly over time, from 21% in 1993 to 17% in 1999 (chi-square = 17.9, P=.01). This trend was primarily due to changes in the demographic and risk factor composition of the cohort over time; after adjustment for demographic factors, directly observed therapy, drug use, and average HIV incidence in the patient's county of residence, this trend lost statistical significance (P= .52).

DISCUSSION

Ten years after the Advisory Council for the Elimination of Tuberculosis recommended universal HIV testing for patients with TB, we found that we are still far from reaching this goal. In our cohort of TB patients, only 67% were confirmed to have been offered HIV testing. Considering only patients for whom data on HIV testing were available brings that figure up to 80%, which is still far short of universal testing. It appears that providers are attempting to offer testing only to patients perceived as "at risk" for HIV infection. Unfortunately, it has been shown in at least one study that health care providers frequently do not detect important risk factors for HIV infection when they are interviewing patients. Similarly, there was a bias against testing elderly patients despite the fact that the prevalence of HIV infection in this group is likely not negligible. (7) In fact, examination of the number of AIDS cases reported to the CDC between 1990 and 1997 (the most recent year for which full data are publi cly available) reveals that the proportion of AIDS cases in patients more than 50 years old at the time of diagnosis has increased from 9.7% in 1990 to 12.1% in 1997, while the proportion in patients older than 60 has also increased from 2.8% in 1990 to 3.2% in 1997. (8)

Many counties in North Carolina have few patients with TB and a low prevalence of HIV. We hypothesized that in such counties the clinicians and public health department personnel would be less likely to suspect, undiagnosed HIV infection and therefore less likely to test for it. There was a statistically significant relationship. between the average incidence of HIV in each county during the study period and the likelihood of being offered an HIV test, but the effect was small. The odds ratio of being offered testing in a county in the highest quartile of HIV incidence versus in the lowest quartile was only 1.5 after adjusting for other demographic factors. Unfortunately, it appears that more experience with these diseases is not translating into improved HIV testing.

The demographic factors associated with being offered an HIV test were similar to those associated with testing positive. One could argue that the clinicians caring for this cohort of TB patients are conserving resources by testing only .the patients who are at highest risk. It has been estimated (9) that screening a population for HIV infection is cost effective if the population in question has a prevalence of 0.5%. The prevalence of HIV infection among TB patients has been reported to be anywhere from 1.6% to 74.6 %. (10-12) In our cohort, 18.9% of patients with known HIV status were seropositive. Even if we assume that all patients whose HIV status was unknown were seronegative, 10.2% of TB patients in North Carolina would still be HIV positive. Given the availability of highly effective antiretroviral therapy and evidence that TB may accelerate the course of HIV disease (13) and increase HIV-associated mortality, universal HIV testing among TB patients is clearly cost effective. Because many cases of HI V infection are not diagnosed until late in the disease, (14) universal testing offers an excellent opportunity for early intervention.

Even if HIV testing were offered to all of our TB patients, a significant number refuse testing. This study was not designed to explore the reasons for this phenomenon, but there was a significant inverse association between groups that were offered testing and groups that refused testing. Patients less than 20 years old, elderly patients, non-African American patients, and females were less likely to be offered testing and were more likely to refuse testing when offered. This may represent the perception among both patients and health care providers that these groups are not at risk for HIV infection. The attitudes of health care providers who are offering testing may also have a significant effect. Up to 30% more patients may accept testing when the provider strongly encourages the test as opposed to when a more neutral approach is used. (15) Interestingly, among two cohorts of STD clinic patients, people who refused HIV testing were shown to be 2 to 5 times more likely to be infected than individuals who accepted the HIV test. (16 17) Of the patients who refused testing in one of these studies, 42% would not permit HIV testing because they believed they were not at risk for acquiring the virus. Universal testing has been effective at decreasing the rate of refusal of HIV testing. One study showed more than 90% acceptance of HIV testing among a cohort of 19,594 TB patients in the Ivory Coast. (18) This success is even more astounding when it is considered that no antiretroviral therapy was available to HIV-infected patients in this cohort; only supportive care and education were offered to individuals who had a positive test.

Another issue raised by this study is HIV testing among pediatric TB patients. Children are a uniquely vulnerable group, and untreated pediatric HIV infection has a generally rapid course and poor prognosis. (19) Public health nurses in North Carolina reported that the decision to test a child for HIV infection was made based on the perceived parental risk (Dee Foster, RN, oral communication, April 2000). As a result, parents of less than half of children younger than 10 with TB in this cohort were offered HIV testing. Furthermore, parents of more than 40% of children in this group offered testing refused it. In short, only 30% of young children with TB in North Carolina actually received HIV testing. Given the personal and societal consequences of failing to diagnose HIV in even a single child, the low rate of HP/ testing among children with TB must be improved.

This study has several limitations. The information available in our database does not include important risk factors for HIV infection such as sexual behavior and exposure to blood products. We used surrogate factors such as age, race, and sex to predict whether patients would undergo HIV testing and which patients would be infected with the virus. While there are some correlations between known HIV risk factors and the demographic surrogates used in this study, these associations are of varying strength and predictive utility. It is possible that every patient with any risk factor for HIV in our population was offered HIV testing, but this scenario is unlikely given the poor sensitivity of clinicians at identifying persons at risk for HIV disease. Another limitation is the lack of data on HIV testing for 15% (561/3,680) of our cohort. Most of these patients were likely not offered HIV testing since local health department personnel often would record "unknown" if patients were not offered HIV testing durin g the early years of the period included in our analysis (Dee Foster, RN, oral communication, April 2000).

Despite these limitations, the data presented here have important implications for health departments throughout the country. Our study reaffirms that patients with TB are at high risk for coinfection with HIV. Despite good compliance with HIV testing among traditionally high risk demographic groups, many patients did not receive an HIV test, either because it was not offered or because testing was refused. This problem is not unique to North Carolina; similar findings have been reported from other sites within the United States. Offering an HIV test to all TB patients is a simple and inexpensive measure that could significantly reduce the morbidity attributable to both diseases.

Acknowledgements. We thank Del Williams, PhD, who provided the data from the North Carolina HIV/STD Prevention and Care HIV/AIDS Reporting database, as well as insight regarding how these data are collected.

From the Department of Internal Medicine, Division of Infectious Diseases and International Health, Duke University Medical Center, Durham, NC (Drs. Stout and Hamilton); the North Carolina Department of Health and Human Services, Raleigh (Drs. Ratard and Southwick); and the Centers for Disease Control and Prevention, Atlanta, Ga (Dr. Southwick).

Supported by NH-I/National Institute of Allergy and Infectious Diseases grant No. A107392 and NIH/National Heart, Lung, and Blood Institute grant No. HL03759. Dr. Hamilton also acknowledges support from the Department of Veterans Affairs and the Centers for Disease Control and Prevention.

Reprint requests to Carol Dukes Hamilton, MD, Duke University Medical Center, Department of Internal Medicine, Division of Infectious Diseases and International Health, Box 3279, Durham, NC 27710.

References

(1.) Havlir DV, Barnes PF: Tuberculosis in patients with human immunodeficiency virus infection. N Engl J Med 1999;340:367-373

(2.) Tuberculosis and human immunodeficiency virus infection: recommendations of the Advisory committee for the Elimination of Tuberculosis (ACET). MMWR Morb Mortal Wkly Rep 1989;38:236-250

(3.) Asch SM, London AS, Barnes PF, et al: Testing for human immunodeficiency virus infection among tuberculosis, patients in Los Angeles. Am I Respir Crit Care Med 1997;155:378-381

(4.) Katz DJ, Hall WN, Keon NB, et al: HIV testing in patients with tuberculosis, physician response to national recommendations. Am Rev Respir Dis 1993;147:1283-1286

(5.) Barnes PF, Silva C, Otaya M: Testing for human immunodeficiency virus infection in patients with tuberculosis. Am J Respir Crit Care Piled 1996; 153:1448-1450

(6.) (CO-98-1) County Population Estimates for July 1, 1998 and Population Change for July 1, 1997 to July 1, 1998. Population Estimates Program, Population Division, US Bureau of the Census, Washington, DC 20233

(7.) el Sadr W, Gettler J: Unrecognized human immunodeficiency virus infection in the elderly. Arch Intern Med 1995;155:184-186

(8.) AIDS Public Information Data Set Software Version 2.7. Centers for Disease Control and Prevention, 1997

(9.) McCarthy BD, Wong JB, Munoz A, et al: Who should be screened for HIV infection? a cost-effectiveness analysis. Arch Intern Med 1993;153:1107-1116

(10.) Snyder DC, Mohle-Boetani JC, Chandler A, et al: A population-based study determining the incidence of tuberculosis attributable to HIV infection. J Acquir Immune Defic Syndr Hum Retrovirol 1997;16:190-194

(11.) Harries AD, Maher D, Mvula B, et al: An audit of HIV testing and HIV serostatus in tuberculosis patients, Blantyre, Malawi. Tuber Lung Dis 1995;76:413-417

(12.) Leitch AG, Rubilar M, Curnow J, et al: Scottish national survey of tuberculosis notifications 1993 with special reference to the prevalence of HIV seropositivity. Thorax 1996;51:78-81

(13.) Whalen C, Horsburgh CR, Hom D, et al: Accelerated course of human immunodeficiency virus infection after tuberculosis. Am J Respir Crit Care Med 1995;151:129-135

(14.) Wortley PM, Chu SY, Diaz T, et al: HIV testing patterns: where, why, and when were persons with AIDS tested for HIV? AIDS 1995;9:487-492

(15.) Irwin KL, Valdiserri RO, Holmberg SD: The acceptability of voluntary HIV antibody testing in the United States: a decade of lessons learned. AIDS 1996;10:1707-1717

(16.) Jones JL, Hutto P, Meyer P, et al: HIV seroprevalence and reasons for refusing and accepting HIV testing. Sex Transm Dis 1993;20:334-337

(17.) Hull HF, Bettinger CJ, Gallaher MM, et al: Comparison of HIV-antibody prevalence in patients consenting to and declining HIV-antibody testing in an STD clinic. JAMA 1988; 260:935-938

(18.) Abouya L, Coulibaly IM, Wiktor SZ, et al: The Cote d'Ivoire national HIV counseling and testing program for tuberculosis patients: implementation and analysis of epidemiologic data. AIDS 1998;12:505-512

(19.) Quinn TC, Ruff A, Modlin J: HIV infection and AIDS in children. Annu Rev Public Health 1992;13:1-30

[Figure 2 omitted]

[Figure 3 omitted]

TABLE 1

Tuberculosis Patients Offered Human immunodeficiency Virus (HIV)
Testing, Accepting HIV Testing, and HIV Positive

                                  Number       Offered        Accepted
Characteristic                 (% of Total)  (% of Group)   (% of Group)

Sex
  Male                         2,430 (66.0)  1,733 (71.3)   1,456 (59.9)
  Female                       1,250 (34.0)    782 (62.6)     594 (47.5)

Race/ethnicity
  Non-Hispanic white             972 (26.4)    543 (55.9)     397 (40.8)
  Non-Hispanic black           2,213 (60.1)  1,599 (72.3)   1,339 (60.5)
  Hispanic                       273 (7.4)     216 (79.1)     188 (68.9)
  Native American                 32 (0.9)      19 (59.4)      13 (40.6)
  Asian/Pacific Islander         189 (5.1)     137 (72.5)     112 (59.3)

Any directly observed therapy
  Yes                          2,371 (64.4)  1,733 (73.1)   1,410 (59.5)
  No/unknown                   1,309 (35.6)    782 (59.7)     640 (48.9)

Foreign birth
  Yes                            479 (13.0)    380 (79.3)     328 (68.5)
  No                           3,199 (86.9)  2,133 (66.7)   1,721 (53.8)
  Unknown                          2 (0.1)       2 (100.0)      1 (50.0)

Injection drug use
  Yes                             69 (1.9)      65 (94.2)      63 (91.3)
  No                           2,849 (77.4)  2,065 (72.5)   1,625 (57.0)
  Unknown                        762 (20.7)    385 (50.5)     362 (47.5)

Noninjection illicit drug use
  Yes                            294 (8.0)     279 (94.5)     267 (90.8)
  No                           2,617 (71.1)  1,850 (70.7)   1,423 (54.4)
  Unknown                        769 (20.9)    386 (50.2)     360 (46.8)

Year reported
  1993                           594 (16.1)    266 (44.8)     249 (41.9)
  1994                           565 (15.4)    343 (60.7)     283 (50.1)
  1995                           518 (14.1)    340 (65.6)     273 (52.7)
  1996                           554 (15.1)    383 (59.1)     326 (58.8)
  1997                           463 (12.6)    369 (79.7)     281 (60.7)
  1998                           498 (13.5)    397 (79.7)     295 (59.2)
  1999                           488 (13.3)    417 (85.7)     343 (70.3)

                               HIV Positive
Characteristic                 (% of Group)

Sex
  Male                          300 (12.3)
  Female                         79 (6.3)

Race/ethnicity
  Non-Hispanic white             38 (3.9)
  Non-Hispanic black            313 (14.1)
  Hispanic                       25 (9.2)
  Native American                 2 (6.3)
  Asian/Pacific Islander          1 (0.5)

Any directly observed therapy
  Yes                           280 (11.8)
  No/unknown                     99 (7.6)

Foreign birth
  Yes                            32 (6.7)
  No                            347 (10.8)
  Unknown                         0 (0)

Injection drug use
  Yes                            41 (59.4)
  No                            248 (8.7)
  Unknown                        90 (11.8)

Noninjection illicit drug use
  Yes                           112 (38.1)
  No                            177 (6.8)
  Unknown                        90 (11.7)

Year reported
  1993                           56 (9.4)
  1994                           61 (10.8)
  1995                           67 (12.9)
  1996                           54 (9.7)
  1997                           41 (8.9)
  1998                           46 (9.2)
  1999                           54 (11.1)
TABLE 2

Factors Associated With Being Offered Human Immunodeficiency Virus (HIV)
Testing. Accepting Testing, and Testing Positive for HIV Among
Tuberculosis Patients in North Carolina (*)

                                      Adjusted OR to be  Adjusted OR to
                                       Offered Testing   Accept Testing
Factor                                   (95 % CI)         (95 % CI)

Sex
    Female                            1.0 (reference)    1.0 (reference)
    Male                              1.2 (0.9-1.5)      1.3 (1.0-1.7)

Race/ethnicity
    Non-Hispanic white                1.0 (reference)    1.0 (reference)
    Non-Hispanic black                1.5 (1.2-1.9)      1.5 (1.1-2.0)
    Hispanic                          1.0 (0.5-2.0)      1.1 (0.6-2.1)
    Asian/Pacific Islander            0.8 (0.4-1.6)      1.0 (0.5-2.0)
    Native American                   0.8 (0.3-2.4)      1.0 (0.3-3.4)

Directly observed therapy
    None received                     1.0 (reference)    1.0 (reference)
    Any received                      1.6 (1.3-2.0)      1.1 (0.8-1.4)

Place of birth
    United States                     1.0 (reference)    1.0 (reference)
    Foreign                           1.3 (0.7-2.3)      1.3 (0.7-2.3)

HIV incidence in county of residence
    1st quartile                      1.0 (reference)    1.0 (reference)
    2nd quartile                      0.8 (0.6-1.1)      0.9 (0.6-1.2)
    3rd quartile                      1.5 (1.1-2.1)      0.5 (0.4-0.7)
    4th quartile                      1.5 (1.1-2.1)      1.3 (0.9-1.9)

Injection drug use
    No                                1.0 (reference)    1.0 (reference)
    Yes                               1.0 (0.3-3.4)      2.3 (0.5-10.4)
    Unknown                           3.5 (0.9-14.5)     3.9 (1.2-12.8)

Noninjection drug use
    No                                1.0 (reference)    1.0 (reference)
    Yes                               3.1 (1.4-6.5)      2.4 (1.3-4.5)
    Unknown                           2.3 (0.6-9.0)      0.9 (0.3-2.8)

                                      Adjusted OR to Test
                                       Positive for HIV
Factor                                     (95 % CI)

Sex
    Female                            1.0 (reference)
    Male                              1.8 (1.3-2.4)

Race/ethnicity
    Non-Hispanic white                1.0 (reference)
    Non-Hispanic black                2.1 (1.4-3.0)
    Hispanic                          1.5 (0.7-3.2)
    Asian/Pacific Islander            0.1 (0.0-0.8)
    Native American                   1.7 (0.3-8.8)

Directly observed therapy
    None received                     1.0 (reference)
    Any received                      1.1 (0.9-1.5)

Place of birth
    United States                     1.0 (reference)
    Foreign                           0.9 (0.5-1.7)

HIV incidence in county of residence
    1st quartile                      1.0 (reference)
    2nd quartile                      1.0 (0.7-1.6)
    3rd quartile                      1.0 (0.6-1.6)
    4th quartile                      2.0 (1.4-3.0)

Injection drug use
    No                                1.0 (reference)
    Yes                               4.2 (2.3-7.7)
    Unknown                           1.2 (0.6-2.3)

Noninjection drug use
    No                                1.0 (reference)
    Yes                               1.9 (1.4-2.7)
    Unknown                           1.7 (0.9-3.3)

(*)The odds ratios are adjusted for age (linear and quadratic terms),
sex, race, directly observed therapy, place of birth, illicit drug use
(injection and noninjection), HIV average incidence in the patient's
county of residence, and the year in which the patient was reported with
active TB.

OR = Odds ratio, CI = confidence interval.

RELATED ARTICLE: KEY POINTS

* Ten years after the Advisory Council for the Elimination of Tuberculosis recommended universal HIV testing for patients with TB, we found that we are still far from reaching this goal.

* Health care providers frequently do not detect important risk factors for HIV infection when they are interviewing patients.

* Universal HIV testing among TB patients is clearly cost-effective.