rates on HIV and medicine wards became essentially the same as the rates on other wards (5.1 percent versus 4.0 percent; p = 0.5).

Again, the impact of the individual control measures could not be determined, but clearly, the overall impact was significant. The authors note that the overall PPD conversion rate was unchanged. They highlight the importance of determining job-specific rates.

Blumberg and colleagues (16) reported the efforts made at Grady Memorial Hospital in Atlanta. These were in response to nosocomial transmissions of drug-sensitive tuberculosis in 1991 and early 1992 (5). Control measures implemented included the following:

March 1

• Expanded isolation policy—all patients with known or suspected tuberculosis (including all patients for whom AFB smear and culture were ordered), also any patient with HIV infection (or risk for HIV infection with unknown serology) with abnormal CXR. Increased surveillance by infection control to ensure that patients for whom smears ordered were in isolation.

• Isolation stopped only after three negative AFB smears (previously stopped after 2 weeks of therapy)

• Increased physician education

• Window fans added to 90 rooms to provide negative pressure

June 1

• Submicron masks used for personal respiratory protection

July 1

• PPD testing done every 6 months now included nonemployee health care workers (e.g., attendings, house staff, medical students)

• tuberculosis nurse epidemiologist hired

To determine the effectiveness of these measures the authors reviewed tuberculosis exposure episodes (from July 1, 1991, to June 30, 1994) and PPD conversions (from January 1, 1992 to June 30, 1994). Over the 3-year period there were 752 admissions (673 patients) with tuberculosis; for 461 admissions (61 percent) the patients had positive AFB smears and were considered infectious. The results for these patients are shown in Table D-1.

Employee PPD conversion rates fell steadily from 3.3 percent to 0.4 percent during the postintervention period (for trend, p < 0.001). For the January-June 1994 PPD conversions (23/5,153 [0.4 percent]) no clustering by work area was noted. In fact only 10 health care workers had direct patient contact on wards where tuberculosis patients were housed, 4 had patient contact on low-risk tuberculosis areas (e.g., neonatal intensive care unit [ICU]), and 9 had no patient contact, suggesting that more than half of the conversions may have been community acquired.