We identified 21 005 San Francisco patients with AIDS with ≥1 AIDS-OI diagnosed during the observation period. We excluded 147 patients: 109 were reported with a death date >1 month before the last reported laboratory test, 33 were aged <13 years, and 5 had no date associated with their first AIDS-OI. This left data from 20 858 San Francisco patients with a first diagnosed AIDS-OI available for analyses, including 3002 (14.4%) with diagnosis in the pre-ART era, 14 097 (67.6%) in the mono/dual ART era, and 3759 (18.0%) in the cART era.

The demographics of AIDS-OI case patients shifted over time toward a distribution consisting of fewer white persons, fewer gay, bisexual or other men who have sex with men (MSM), and more black and Hispanic men and women who acquired HIV through injection drug use or heterosexual transmission. Across the 3 treatment eras, the median ages at first AIDS-OI diagnosis were 37, 39, and 43 years, respectively, and the majority of patients were white non-Hispanic (85.3%, 75.7%, and 57.8%, respectively), male (99.4%, 97.1%, and 91.9%, respectively), and MSM (82.8%, 76.9%, and 59.6%, respectively) (Table 1). Owing to differential mortality, patients with a first AIDS-OI diagnosed in the cART era had a longer median follow-up (5.16 years; range, <1 to 15.6 years) than those diagnosed in pre-ART and mono/dual ART eras (1.00 years [range, <1 to 29.3 years] and 1.52 years [<1 to 25.5 years], respectively). Of the 20 858 patients with AIDS-OIs, death occurred in 17 099 (82.0%) by the end of observation: 2957 (98.4%), 12 569 (89.2%), and 1573(41.9%) in each treatment era, respectively. Within each era, the majority of decedents with death certificate information available had an AIDS-OI mentioned on the death certificate.

The annual number of AIDS-OIs diagnoses peaked in the mono/dual-ART era, followed by a pronounced decline in the cART era (Table 2). Across the entire study period, the 10 most frequently diagnosed AIDS-OIs in 1981–2012 were PCP (n = 8163), Kaposi sarcoma (KS; n = 4195), HIV wasting syndrome (n = 1864), esophageal candidiasis (n = 1381), MAC (n = 1069), HIV encephalopathy (n = 932), extrapulmonary cryptococcosis (n = 883), chronic cryptosporidiosis (n = 694), immunoblastic lymphoma (n = 676), and cytomegalovirus disease other than retinitis (n = 643). The 5 least frequently diagnosed AIDS-OIs were invasive cervical cancer (n = 12), recurrent Salmonella septicemia (n = 19), disseminated coccidioidomycosis (n = 32), chronic isosporiasis (n = 36), and disseminated histoplasmosis (n = 78).

In Kaplan–Meier analyses of survival after diagnosis by AIDS-OI type, survival probabilities were significantly higher in the cART era than in previous treatment eras (P < .05; logrank test). The overall 5-year survival probability after first AIDS-OI was 7% in the pre-ART era but almost doubled to 18% in the mono/dual-ART era and showed the greatest improvement in the cART era, increasing to 65% (Figure 1).

When we analyzed survival after selected types of AIDS-OIs, we found that in the cART era, the median survival time after the diagnosis of PCP, KS, and cryptosporidiosis was >15 years, compared with <5 years in the previous treatment eras (Figure 2). Although there have been improvements in survival in the cART era for MAC and immunoblastic lymphoma, the median survival times for these AIDS-OIs remained <5 years.

For each of the 10 most common AIDS-OIs during 1981–2012, the 5-year survival probabilities after diagnosis increased from the pre-ART era to the cART era, with the greatest survival gain for patients with PCP: from 1% in the pre-ART era to 69% in the cART era (Figure 2). In the cART era, patients with a diagnosis of chronic cryptosporidiosis had the highest 5-year survival probability (81%), followed by those with KS (75%); patients with immunoblastic lymphoma had the lowest 5-year survival probability (47%).

Among the 3759 patients with their first AIDS-OI in the cART era (1997–2012), 1,579 (41.2%) died. The median follow time was 5.0 years, and the median time to death was 1.5 years. In addition, 42.6% of patients had a CD4 cell count of 50–199 cells/mm³, 84.9% were prescribed ART during the follow-up, 78.1% were prescribed PCP prophylaxis, and 36.5% were prescribed MAC prophylaxis (Table 3). Using PCP as the first AIDS-OI for the referent group and adjusting for age, HIV transmission category, CD4 cell count, use of any ART, and prophylaxis against MAC and PCP, mortality rates were the highest for brain lymphoma (hazard ratio [HR], 5.14; 95% confidence interval [CI], 2.98–8.87), progressive multifocal leukoencephalopathy (HR, 4.22; CI, 2.49–7.17), multiple AIDS-OIs (HR, 3.22; CI, 2.75–3.77), and immunoblastic lymphoma (HR, 2.56; CI, 1.90–3.42) (Table 3). Burkitt lymphoma (HR, 2.40; CI, 1.58–3.65), HIV encephalopathy (HR, 2.23; CI, 1.67–2.96), HIV wasting (HR, 1.81; CI, 1.44–2.29), and recurrent pneumonia (HR, 1.56; CI, 1.17–2.08) were also associated with significantly higher mortality rates than PCP. In contrast, chronic cryptosporidiosis was associated with lower mortality rates than PCP (HR, 0.44; 95% CI, .24–.82). Mortality rates after diagnosis of the other AIDS-OIs did not differ significantly compared with those for PCP.

Persons with a first AIDS-OI diagnosed in the cART era were more likely to die during follow-up if they were not prescribed prophylaxis against PCP (HR, 1.33; CI, 1.15–1.53) or MAC (HR, 1.42; CI, 1.26–1.59) or were not prescribed ART (HR, 2.50; CI, 2.15–2.91). Compared with persons who had geometric mean CD4 cell counts ≥350 cells/mm³, those with lower CD4 counts were more likely to die (HR, 4.47 [CI, 3.61–5.52], 2.12 [1.73–2.59], and 1.57 [1.26–1.95] for CD4 cell counts of <50, 50–199, and 200–349 cells/mm³, respectively). Compared with persons in all other HIV transmission categories (combined), those who acquired HIV through injection drug use were also more likely to die (HR, 1.42; CI, 1.29–1.55).

In the large population of persons with AIDS in San Francisco, survival after first AIDS-OIs has increased markedly from the pre-ART to cART treatment era, but it has remained poor for certain AIDS-OIs. Diagnoses of all AIDS-OIs declined in cART era, consistent with results reported in previous studies [9–11]. In earlier analyses (from 1993 to 2008), Schwarcz et al [10] showed that the incidence of AIDS-OIs among San Francisco patients with AIDS declined since the introduction of cART, with the incidence rate reductions for individual AIDS-OI events ranging from 84% to 99% comparing the earliest cART (1996–2000) and the latest cART (2001–2008) period. Although the incidence of AIDS-OIs in San Francisco has been declining since the 1990s [10], the ranking of first AIDS-OIs by frequency has not changed meaningfully since the beginning of HIV epidemic. PCP and KS consistently remained the 2 most frequently reported AIDS-OIs in all eras. The predominance of PCP among patients with AIDS-OIs diagnosed in the cART era has been reported in previous studies [21, 33], although declines in PCP incidence have been noted with the introduction of PCP prophylaxis in the early 1990s [10].

Our findings of improvement in survival after AIDS-OI diagnosis are consistent with those from previous studies in industrialized countries from pre-ART and cART treatment eras [20, 23, 27]. These survival improvements, especially in the most recent treatment era, probably reflect a combination of factors, including use of cART, improved prophylaxis and treatment of AIDS-OIs, earlier diagnosis of HIV infection, and better linkage to effective HIV care [34–36]. Despite these overall improvements, changes were not uniform, and there remained significant disparities in survival across AIDS-OIs in the cART era in our analyses, which were also noted in aforementioned studies, although respective results cannot be directly compared, owing to varying study populations, time frames, and analytic designs [20, 23, 27]. Survival after PCP diagnosis demonstrated the largest improvement. The use of PCP prophylaxis to prevent the acquisition of PCP among HIV-infected persons with a CD4 cell count <200 cells/mm³ has been recommended since 1989 and has improved survival among HIV-infected persons [37, 38]. Prophylaxis has also been used to prevent MAC infection, but in these analyses the improvement in survival after a diagnosis of MAC infection as the first AIDS-OI was not as great as that after PCP, possibly owing to the more profound immunosuppression at which MAC typically occurs.

Although the AIDS-OIs of some persons may have resolved before death, and persons may have died of causes other than AIDS-OIs, 71% of decedents for whom death certificates were available had AIDS-OI mentioned on those certificates. After adjusting for clinical and demographic variables, we found that survival after first AIDS-OIs was poorest for persons with AIDS-defining cancers or AIDS-defining OIs associated with very low CD4 counts, such as progressive multifocal leukoencephalopathy, and diagnosis of multiple AIDS-OIs. Our findings support previous studies finding that survival remains poor after diagnosis of lymphomas, notably brain lymphoma, in the cART era [19–21, 27, 29, 39]. In analyses similar to ours, brain lymphoma was associated with the worst survival among persons with AIDS in New York City [21].

In the cART era, AIDS-OIs that had effective treatments generally were associated with survival similar to that after a first AIDS-OI diagnosis of PCP. Persons with chronic cryptosporidiosis as their first AIDS-OI were less likely to die than those with PCP, possibly because the definitive treatment for this condition with comparatively low mortality in the pre-ART and mono/dual-ART era (Figure 2) is effective cART. We found that higher CD4 cell counts within a year of the first AIDS-OI diagnosis, as well as cART and prophylaxis against PCP and MAC, were all associated with better survival, highlighting the continued importance of cART and OI prophylaxis. We also found that the risk of dying after first AIDS-OI diagnosis was higher among persons with a history of injection drug use than among those with a history of likely exposure to HIV through sexual contact but not injection drug use. As reported in other HIV-infected populations, the reasons for this decreased survival may be poorer health, less access, and adherence to cART, initiation of ART at a more advanced stage of HIV infection, or accidents and overdose [39–43].

Our analyses had several limitations. We relied on routinely collected HIV surveillance data with a limited number of predefined variables. For example, our ART data included only the date of treatment initiation but no information on treatment interruption or discontinuation, and the CD4 cell count data were routinely collected only in the cART period. Importantly, HIV diagnoses (vs AIDS diagnoses only) were not subject to mandatory reporting to the San Francisco Department of Public Health until 2002, and universal reporting of CD4 cell counts to HIV surveillance was introduced in 2008; thus, we could not systematically quantify whether and how earlier HIV diagnosis and cART may have contributed to improved survival after first AIDS-OI.

Because we used an intent-to-treat approach in our analyses, our estimates of the effect of ART on mortality are conservative. We did not have information on coexisting non–AIDS-related conditions, which may have contributed to mortality risk. In addition, the population of San Francisco is not representative of the US general population. The population in our analyses was predominately MSM and therefore provided a rich source of data on KS (which is more common among MSM than among persons in other HIV transmission categories), but with so few women in the study population, we could not study cervical cancer. The San Francisco Department of Public Health surveillance staff conducts follow-up chart reviews at clinical sites that serve large numbers of HIV-infected persons, but we may have missed AIDS-OIs among persons who were seen at smaller private practices or had moved outside San Francisco after their HIV diagnosis.

Some AIDS-OIs that require documentation of recurrence (eg, recurrent pneumonia) or fulfillment of other specific clinical criteria (eg, HIV wasting) may have been missed or incorrectly diagnosed by some reporting providers, and the extent of such overreporting or underreporting is not clear. We expect this misclassification to be nondifferential; thus, our overall estimates remain valid. In addition, our adjustment for clinical factors was limited to variables collected as part of routine AIDS surveillance and lacked more complete measures of whether or not persons were receiving or adhering to HIV care. Finally, our data are from a city with widespread and early treatment programs, so the survival estimates we found may not be generalizable to patients with AIDS in other areas of the United States with different levels of care.

In conclusion, using active HIV surveillance data from San Francisco, we have shown that survival after a first AIDS-OI diagnosis, especially for the most frequently diagnosed AIDS-OIs, has improved markedly since 1981. Despite these improvements, 35% of persons with an AIDS-OI diagnosed in the cART era died within 5 years. Some AIDS-OIs remain associated with substantial mortality risk even after adjustment for CD4 cell count. Better prevention and treatment strategies, including earlier HIV diagnosis, are still needed for AIDS-OIs in the cART era.