From January to December 1997, 250 (78%) of 319 patients admitted to the ward had a completed questionnaire and physical examination and at least one adequate sputum specimen. Forty patients (16%) had sputum cultures positive for Mycobacterium tuberculosis, and 26 of these had positive sputum smears. One patient had a positive smear but a negative culture. Only three patients had a diagnosis of HIV infection; none of the three had a positive sputum specimen. Of the 69 ward patients who declined to participate or were unable to provide an adequate sputum specimen, 4 (6%) had been admitted with a diagnosis of suspected TB. If we assume all excluded patients to be negative for TB, the minimum estimated TB prevalence on the ward was 13%.

Patients with a cough of any duration, a cough that lasted >2 weeks, reported weight loss, hemoptysis, or a family history of TB were more likely to have sputum cultures positive for TB (Table 1). Anorexia was associated with a lower likelihood of TB. Because of logistic constraints, we were able to place and read a TST at 48 to 72 hours only on a subset of patients. Of the 67 patients with TST results, a positive reading was observed in 11 (55%) of 20 culture-positive patients compared with 10 (21%) of 47 patients without TB (p=0.007). Among culture-positive patients, those with a positive TST response were younger than those with a negative reading (median 23 years of age[range 19-66] vs. 47 years [range 25-88], p=0.02 by Wilcoxon 2-sample test). The socioeconomic status of patients with and without TB was similar.

Of the 181 patients who reported past BCG immunization, 178 (98%) had a scar. No vaccine scars were observed among the 68 persons who reported no history of BCG immunization. However, having a BCG scar was not associated with any apparent protective effect (Table 1). The presence of a BCG scar was not associated with a positive TST, even when TB culture positive patients were excluded (p=0.7).

Of 40 patients with at least one positive sputum culture, 23 (58%) had strains resistant to at least isoniazid, 8 (20%) to rifampin, 4 (10%) to ethambutol, and 1 (3%) to streptomycin. None were resistant to ciprofloxacin or capreomycin. Eight patients (20%) had TB resistant to both isoniazid and rifampin and were classified as having MDRTB. All 8 patients with resistance to rifampin also had resistance to isoniazid, and 15 patients had strains resistant to isoniazid but not to rifampin. Of the eight strains resistant to both isoniazid and rifampin, one was also resistant to ethambutol, one to streptomycin, and one to both ethambutol and streptomycin. Of 8 patients with MDRTB, 3 had a previous history of TB treatment.

Culture-positive patients for whom TB was the admitting diagnosis differed from those in whom TB was not suspected at the time of admission (Table 2). Patients whose TB had not been suspected were older and less likely to have the classic findings of cough, hemoptysis, weight loss, and prior personal or family history of TB. Patients whose TB had not been suspected at the time of admission were less likely to have a positive sputum smear, but this difference did not reach statistical significance (p=0.16 by Fisher's exact test). However, patients whose TB had not been suspected were significantly more likely to have MDRTB. Six (75%) of 8 patients with MDRTB were not suspected to have TB on admission; 3 (50%) of these six were also smear positive. Admitting diagnoses among culture-positive patients whose TB had not been suspected on admission included two patients with diabetes mellitus, one with systemic lupus erythematosus, and one with a lung lesion thought to be a hydatid cyst.

The overall prevalence of TB among our study patients was high: at least 13% of all patients admitted to this general medicine ward had active TB. Two-thirds of TB patients were smear positive and therefore highly infectious, one-fifth had multidrug-resistant strains, and 75% of the patients with MDRTB had not been suspected of having TB when they entered the hospital. As in most Latin American hospitals, no masks or other respiratory devices were used to prevent spread in this hospital, even when the patient was known to be smear positive and highly infectious.

Nosocomial outbreaks of MDRTB in the United States in the 1980s and early 1990s heightened enforcement of stringent hospital control measures (11), leading to measurable decreases in TST conversion rates among hospital staff (12). Although the rate of TB in Peru is approximately 20 times higher than that of New York City (13), no concerted effort has been made to improve TB control measures in Peruvian hospitals.

The spread of MDRTB threatens control efforts (14). The fact that the majority of our patients with MDRTB had no history of past treatment of TB implies that person-to-person transmission of multidrug resistant strains occurs in Peru. Our data suggest that hospital wards may be one of the sites of transmission.

In developing countries where resources are limited, TB control programs focus on identification and treatment of infectious cases (15). Although treatment is clearly an important component of control, person-to-person spread of resistant strains makes isolation a high priority for preventing transmission. TST testing was not useful in identifying the group in need of screening. Anergy, which was common among culture-positive TB cases, was associated statistically with older median age and was perhaps related to concurrent systemic illness and poor nutritional status among hospitalized patients.

Although Peru has implemented an effective community-based TB control program, hospital control has not been a focus. Control measures such as isolation and respiratory precautions, stringently enforced in the past, were relaxed worldwide after the advent of inexpensive, effective anti-TB medications. After 50 years of selective drug pressure, the outbreak of MDRTB in New York City (5) dramatically highlighted the consequences of lapses in infection control.

Our data show that in countries or locales with a known high prevalence of TB, hospitals should screen all patients with respiratory symptoms by sputum smear within 12 hours of admission to hospital. Those found to be smear-positive should be placed in respiratory isolation, apart from TB-negative patients, until the smear becomes negative. Hospital personnel should observe respiratory precautions in caring for these patients. A system of rapid culture diagnosis and susceptibility testing should be implemented, allowing the presumptive diagnosis of MDRTB within 2 weeks (16). In combination, admission screening for TB, re-implementation of effective hospital respiratory control, and rapid TB diagnosis can substantially decrease the transmission of TB, especially MDRTB, in countries like Peru.