Opportunistic Infections in the United States: Focusing Health Care and Needs for People with AIDS

Citation: Eckholdt, H., Chin, J., Harris, C., and Kim, D. (1998). Opportunistic infections in the United States: Focusing health care and needs for people with AIDS. Poster #13236 presented at the 12th World  AIDS Conference. Geneva, Switzerland.

Abstract.
Past efforts to measure health care access, delays in diagnosis, and health cost have focused on individual opportunistic infections excluding some groups due to low frequency (Asians and Pacific Islanders and Native Americans). Measuring the total number of AIDS indicator infections could help health care providers target groups in greatest need of outreach efforts for earlier HIV/AIDS diagnosis and treatment. Data on the total number of AIDS indicator infections among adolescents and adults diagnosed with AIDS in the US from 1981 through December 1994 (N=441,528) were analyzed using the Centers for Disease Control and Prevention's AIDS Public Information Data Set. The number of presenting opportunistic infections was related to survival time, as well as probability of death. The groups at greatest risk for increased infection were Asian/Pacific Islanders and Native American, Intravenous Drug Use + Men who have sex with Men and MSM’s, Central U.S. region of residence, earlier years, and younger ages. The identified groups need to be targeted for early HIV treatment interventions.

Background.
Past efforts to measure health care access, delays in diagnosis, and health cost have focused on individual opportunistic infections excluding some groups due to low frequency (Asians and Pacific Islanders and Native Americans). We measured the total number of AIDS indicator infections to further our understanding of health care needs especially for lower frequency groups. If meaningful, this measure would help health care providers target groups in greatest need of outreach efforts for earlier HIV/AIDS diagnosis and treatment.

Changes in the epidemiology of Acquired Immune Deficiency Syndrome (AIDS), the health care system, and AIDS treatment regimens in the United States (US) have brought renewed attention to issues of access to health care and health care needs for people who test positive for the Human Immuno-deficiency Virus (HIV+) and people diagnosed with AIDS. Past efforts to measure health care access focused on trends in Pneumocystis carinii pneumonia (PCP) as a presenting AIDS defining infection (Eckholdt & Chin, 1997). PCP was chosen for such modeling because PCP is preventable with the use of commonly available prophylactic agents. The appearance of PCP has been explained as indicative of barriers to receiving or maintaining appropriate care (Chien, Rawji, et al, 1992; Graham, Zeger, et al, 1991; Piette, Stein, et al, 1991) or group differences in microbe strain or exposure history (Smulian, Sullivan, et al, 1993; Walzer, Kim, et al, 1989). Past analyses using data on the frequency and proportion of confirmed PCP diagnoses as the presenting AIDS defining infection among adolescents and adults diagnosed with AIDS in the US from January 1984 through December 1994 obtained directly from the Centers for Disease Control and Prevention’s Public Information Data Set (CDCP, 1997) showed that Asians and Pacific Islanders were at increased risk for PCP compared with all racial groups when controlling for relevant temporal, demographic, and HIV transmission/exposure factors.

Recent efforts to measure the clinical impact, cost, and cost effectiveness of treatment for prophylaxis of various AIDS related infections showed that increased costs and reduced life expectancy were associated with failure to treat diseases prophylacticly (Freedberg, Scharfstein, et al, 1998). In an effort to further our understanding of access to health care, we developed another such proxy using the total number of AIDS indicator infections from the same data.

AIDS indicator infections are reported to the CDCP which maintains a national surveillance of AIDS through the receipt of AIDS case reports submitted by individual state and local health departments. The criteria for a diagnosis of AIDS has changed over the years but generally consists of the probable or definitive diagnosis of one or more infections, in the absence of other non-AIDS related etiology, designated by the CDCP. Due to the regional and temporal differences in the rigors and resources applied to the surveillance system, the CDCP acknowledges that these data are to be considered minimal estimates for AIDS diagnoses and AIDS indicator infections.

Methods.

Measure. As we set out to further our investigation of measures that can serve as proxies for "barriers to health services" in these data, we devised a sequence of logical steps and hypotheses that would reduce our chances of committing a Type II error – running all possibilities and finding significance in the absence of any real world phenomena – as well the burdens on our own team resources. Our choice for such a measure was the sum of presenting AIDS indicator infections.

Modeling and Assessment. It was important that our measure of barriers, the sum of presenting AIDS indicator infections, was (a) clinically meaningful. Our second step involved the construction of hypotheses regarding the (b) empirical meaning of the measure through the statistical / epidemiologic behavior of the constructed measure. We reasoned that the measure would exhibit a systematic process where people with (1) etiologically relevant histories of exposure, or (2) a history of neglect of past and immediate health needs would rate highest on this sum of infections. Where a random processes would show uniform distributions across demographic, geographic, and transmission groups (once samples reach a large enough size). Data in support of sum of infections as the outcome of (1) prior exposures would show systematic processes like increased risk among specific groups likely to possess relevant exposure histories (i) country of birth or (ii) HIV exposure group. Data in support of a (2) barrier interpretation would show systematic group processes favoring (i) race, or (ii) region of residence. Our third step in the modeling process involved a kind of outcome (c) validation whereby our constructed measure of barriers to health care would be related to (1) likelihood of death, and (2) survival time among those cases data which contain life/death status and date of death.

Analysis. In order to test these hypotheses, data on the total number of AIDS indicator infections among adolescents and adults diagnosed with AIDS in the US from 1981 through December 1994 were analyzed using the Centers for Disease Control and Prevention’s AIDS Public Information Data Set (CDCP, 1997) were analyzed. The outcome consisted of the sum of presumptive and definitive diagnoses of the following presenting opportunistic infections: Severe Immunosuppression, Infections: bacterial, Lymphoma: Burkitts, Candidiasis: esoph, Candidiasis: pulmonary, Cytomegalovirus: other, Cytomegalovirus: retinitis, Coccidioidomycosis, Cryprococcosis: extrapulmonary, Cryprococcosis: intestinal, HIV encephalopathy, Histoplasmosis, Herpes Simples, Lympnoma: immunoblastic, Isosporiasis, Kaposis Sarcoma, Lymphiod, Mycobacterium: avium, Mycobacterium: other, Pneumocystit carinii pneum, Lymphoma: brain, Leukoencephalopathy, Salmonella septicemia, Tuberculosis: extrapulmonary, Toxoplasmosis, Wasting syndrome, Cervical Cancer, Tuberculosis: pulmonary, Pneumonia).

Race group was dummy coded as contrasts with each race group against "White", covariates were: year of diagnosis (entered as a class variable 81 through 94), transmission exposure group (using the AIDSPIDS exposure categories: injecting drug use, heterosexual contact with a person with or at increased risk for HIV infection, other exposures including hemophilia and blood transfusion, contrasted against men who have sex with men), age at diagnosis (using the AIDSPIDS age category entered as a class variable: 13 to 19 years, 20 to 24 years, 25 to 29 years, 30 to 34 years, 35 to 39 years, 40 to 44 years, 45 to 49 years, 50 to 54 years, 55 to 59 years, 60 to 64 years, 65 years or older), geographic region of residence in the United States (using the AIDSPIDS categories: Central, Western, Southern, Mid-Atlantic, and smaller msa (50,000 to 1,000,000), contrasted against North), gender (female, male), and birthplace (born in U.S, born outside U.S.).

The a priori hypotheses were assessed using simple odds ratios and Pearson correlations. Confirmation of our simpler results was conducted with Piosson regression models. Post hoc analyses were conducted using linear regression models. All data were analyzed with SAS (1998). Poisson distribution and the Log link function were conducted

Results.

Initial data analyses showed that the distribution of opportunistic infections at diagnoses is a highly skewed distribution whereby 56.4% of all AIDS diagnoses through December 1994 (total n=441,528) report only one opportunistic infection (see Figure 1).

In an effort to simplify this discussion, much of the following analyses were conducted with a binary version of the number of infections (0=1 infection, 1=2 or more infections). This allowed us to investigate risks associated with the increased infections. First and foremost was an increased risk for death. People with 2 or more infections at diagnosis with AIDS were 1.9 (CI=1.88, 1.93) times more likely to be reported as dead in the database in comparison to people reporting only 1 infection at diagnosis. This finding was interpreted as a form of predictive validity for the total number of infections.

Our initial hypotheses concerned the relationship between the number of infections and race group. Table 1 shows the odds ratios for the risk of 2 or more infections among Asians and Pacific Islanders, and Native Americans in contrasts to Whites, Blacks, and Hispanics.

Table 1. Significant odds ratios for risk of 2 or more infections at AIDS diagnosis for Asians and Pacific Islanders and for Native Americans in contrast to each other race group.

Odds Ratio (95% confidence interval)

Asian & Pacific Islander

Black 1.287 (1.197, 1.383)

Hispanic 1.448 (1.346,1.557)

Native American

White 1.186 (1.052,1.338)

Black 1.467 (1.301,1.654)

Hispanic 1.650 (1.464,1.861)

Our next level of analysis considered the strength of these effects in the context of other meaningful factors in the database. Therefore, we devised regression models using race group to predict the number of opportunistic infections while controlling for Region of Residence, Year of Diagnosis, Transmission Group, Gender, Age at Diagnosis, and Country of Birth. These models confirmed that Asians and Pacific Islanders as well as Native Americans have higher numbers of opportunistic infections at diagnosis with AIDS in comparison to Blacks and Hispanics, and that Native Americans have more infections than Whites, all while controlling for the factors listed above. This a priori model was run using normal, binomial, as well as poisson distributions. Figure 2 illustrates these proportion of 2 or more infections for each race group at each year of diagnosis with AIDS.

The regression models also revealed other significant relationships that we wish to report as post hoc findings that would be of interest. In these analyses we found that for Sex: male > female; Transmission Group: MSM & IDU > MSM > Heterosexual=Blood > Unknown > IDU=Hemophilia; Region: Central > West=Mid Atlantic=Non MSA=Small=South > Northeast; Year: Earlier > Later; Country of Birth: In US > Out of US.

Conclusions.

Relationships with survival time and death suggest that number of infections is meaningful and relevant in HIV/AIDS care research. Differences between: (1) Race groups and Regions of residence may be indicative of systematic barriers to health care, (2) Exposure groups suggest the role of prior exposure, and (3) Year of diagnosis show trends in the successful treatment of some infections. The identified groups need to be targeted for early HIV treatment interventions.

References.

CDCP. (1995). AIDS Public Information Data Set (AIDSPIDS). Anonymous. Anonymous. Atlanta, Georgia (404) 639-2020:Division of HIV/AIDS. (through December 1994). http://www.cdc.gov/hiv/dhap.htm

S. M. Chien, M. Rawji, S. Mintz, A. Rachlis, and C. K. Chan. (1992). Changes in hospital admissions patterns in patients with Human Immunodeficiency Virus infection in the era of Pneumocystis carinii prophylaxis. Chest 102:1035-1039.

H. Eckholdt and J. Chin. (1997). Pneumocystis carinii pneumonia in Asians and Pacific Islanders. Clinical Infectious Disease 24:1265-1266.

Kenneth Freedberg, Julie Scharfstein, George III Seage, and et al. (1998). The Cost-Effectiveness of Preventing AIDS-Related Opportunistic Infections. Journal of the American Medical Association 279(2):130-130.

N. M. Graham, S. L. Zeger, L. P. Park, J. P. Phair, R. Detels, S. H. Vermund, M. Ho, and A. J. Saah. (1991). Effect of zidovudine and Pneumocystis carinii pneumonia prophylazis on progression of HIV-1 infections to AIDS. Lancet 338:265-269.

J. Piette, M. Stein, V. Mor, J. Fleishman, K. Mayer, T. Wachtel, and C. Carpenter. (1991). Patterns of secondary prophylaxis with aerosol Pentamidine among persons with AIDS. Journal of Acquired Immune Deficiency Syndromes 4 (8):826-828.

SAS System for Windows. (1998). Cary, North Carolina:SAS, Inc. (919) 677-8008. (6.12)

G. Smulian, D. W. Sullivan, M. J. Linke, N. A. Halsey, T. C. Quinn, A. P. MacPhail, M. A. Hernandez-Avila, S. T. Hong, and P. D. Walzer. (1993). Geographic variation in the humoral response to Pneumocystis carinii. The Journal of Infectious Diseases 167:1243-1247.

P. D. Walzer, C. K. Kim, and M. T. Cushion. (1989). Pneumocystis carinii, New York:Marcel Dekker, Inc., 1989.

send correspondence to:

Haftan Eckholdt, Ph.D., M.S.

Albert Einstein College of Medicine / KC-923

1410 Pelham Parkway South

Bronx, New York 10461

email: eckholdt@aecom.yu.edu