We used data from the HIV Research Network

(HIVRN), a consortium of sites (see Appendix) that provide primary and subspecialty care to HIV-infected patients in 14 cities throughout the USA. To participate in the HIVRN, a site had to have a minimum data set including the patients’ age, sex, race, HIV transmission risk factor, AIDS-defining illnesses, CD4 Nutlin-3a order cell count, HIV-1 RNA level and use of antiretroviral medication. Eleven HIVRN sites that treated adult HIV-infected patients, nine with academic affiliations, also collected data on resource utilization and in-patient ICD-9 codes. Data from 10 of these sites, located in the Northeastern (six sites), Western (two), Midwestern (one) and Southern (one) USA, were included in the analysis. One nonacademic site discontinued participation in the HIVRN during this period and was excluded from analyses. All adult HIV-infected patients (≥18 years old) at these 10 sites with at least one out-patient visit between 2000 and 2008 were eligible for inclusion in the study. Each site abstracted the data elements described above from electronic or paper

records. After removal of identifying information, the sites sent the abstracted data Buparlisib chemical structure to a data co-ordinating centre in an electronic format. For this analysis, data collection encompassed the period from 1 January 2000 to 31 December 2008, as recorded by the date of encounter, not the date of billing or claim payment. Development, maintenance and use

of the database are approved by the Institutional Review Board of the Johns Hopkins University School of Medicine, which serves as the data co-ordinating centre, as well as the Institutional Review Boards of each of the participating institutions. Because bacteraemia is nearly always treated in in-patient settings, we focused on hospital admissions data recorded at each study site. Each Celecoxib HIVRN site reported dates of admission and discharge and all ICD-9 codes associated with an in-patient episode. Any text descriptions were translated to ICD-9 codes. All in-patient episodes were reviewed, starting from 1 January 2000 or the patient’s first recorded out-patient visit to the HIV clinic, whichever came later, and ending at date of death or 31 December 2008, whichever came first. ICD-9 codes were examined to identify all in-patient cases of bacteraemia or septicaemia during the study period. The ICD-9 code for septicaemia (038.XX) intrinsically includes the organism of interest whereas the ICD-9 code for bacteraemia (790.7) does not include an organism. Therefore, in these cases we used the second ICD-9 code (041.XX), which indicates the organism causing bacteraemia. Table 1 shows the classification of bacteraemia/septicaemia episodes in terms of types of organisms and their mapping to ICD-9 codes. In addition, analyses also included a small number of episodes associated with Salmonella (003.1) and Listeriosis (027.0).