To investigate an increase in reports of legionnaires' disease by multiple hospitals in San Antonio, Texas, and to study risk factors for nosocomial transmission of legionnaires' disease and determinants for Legionella colonization of hospital hot-water systems.

The 16 largest hospitals in the cities of San Antonio, Temple, and Austin, Texas.

Review of laboratory databases to identify patients with legionnaires' disease in the 3 years prior to the investigation and to determine the number of diagnostic tests for Legionella performed; measurement of hot-water temperature and chlorine concentration and culture of potable water for Legionella. Exact univariate calculations, Poisson regression, and linear regression were used to determine factors associated with water-system colonization and transmission of Legionella.

Twelve cases of nosocomial legionnaires' disease were identified; eight of these occurred in 1996. The rise in cases occurred shortly after physicians started requesting Legionella urinary antigen tests. Hospitals that frequently used Legionella urinary antigen tests tended to detect more cases of legionnaires' disease. Legionella was isolated from the water systems of 11 of 12 hospitals in San Antonio; the 12th had just experienced an outbreak of legionnaires' disease and had implemented control measures. Nosocomial legionellosis cases probably occurred in 5 hospitals. The number of nosocomial legionnaires' disease cases in each hospital correlated better with the proportion of water-system sites that tested positive for Legionella (P=.07) than with the concentration of Legionella bacteria in water samples (P=.23). Hospitals in municipalities where the water treatment plant used monochloramine as a residual disinfectant (n=4) and the hospital that had implemented control measures were Legionella-free. The hot-water systems of all other hospitals (n=11) were colonized with Legionella. These were all supplied with municipal drinking water that contained free chlorine as a residual disinfectant. In these contaminated hospitals, the proportion of sites testing positive was inversely correlated with free residual chlorine concentration (P=.01). In all hospitals, hot-water temperatures were too low to inhibit Legionella growth.

The increase in reporting of nosocomial legionnaires' disease was attributable to increased use of urinary antigen tests; prior cases may have gone unrecognized. Risk of legionnaires' disease in hospital patients was better predicted by the proportion of water-system sites testing positive for Legionella than by the measured concentration of Legionella bacteria. Use of monochloramine by municipalities for residual drinking water disinfection may help prevent legionnaires' disease.

In 1994, a hospital reported an increase in nosocomial legionnaires' disease after implementing use of a rapid urinary antigen test for Legionella pneumophila serogroup 1 (Lp-1). This hospital was the site of a previous nosocomial legionnaires' disease outbreak during 1980 to 1982.

Infection control records were reviewed to compare rates of nosocomial pneumonia and the proportion of cases attributable to legionnaires' disease during the 1994 outbreak period with those during the same period in 1993. Water samples were collected for Legionella culture from the hospital's potable water system and cooling towers, and isolates were subtyped by monoclonal antibody (MAb) testing and arbitrarily primed polymerase chain reaction (AP-PCR).

Nosocomial pneumonia rates were similar from April through October 1993 and April through October 1994: 5.9 and 6.6 per 1,000 admissions, respectively (rate ratio [RR], 1.1; P=.56); however, 3.2% of nosocomial pneumonias were diagnosed as legionnaires' disease in 1993, compared with 23.9% in 1994 (RR, 9.4; P<.001). In 1994, most legionnaires' disease cases were detected by the urinary antigen testing alone. MAb testing and AP-PCR demonstrated identical patterns among Lp-1 isolates recovered from a patient's respiratory secretions, the hospital potable water system, and stored potable water isolates from the 1980 to 1982 outbreak.

There may have been persistent transmission of nosocomial legionnaires' disease at this hospital that went undiscovered for many years because there was no active surveillance for legionnaires' disease. Introduction of a rapid urinary antigen test improved case ascertainment. Legionella species can be established in colonized plumbing systems and may pose a risk for infection over prolonged periods.