Causes of Death in Persons With Human Immunodeficiency Virus Infection MICHAEL STEIN, M.D., PATRICIAO’SULLIVAN, Ed& TOM WACHTEL, M.D., ALVAN FISHER, M.D., DENNISMIKOLICH, M.D., STEVEN SEPE, M.D., GLENN FORT, M.D., CHARLESCARPENTER, M.D., GAILSKOWRON, M.D., KENNETH MAYER, M.D., Providence, Rhode/s/and

PURPOSE: Pneumocysti carinii pneumonia (PCP) was reported to be the predominant cause of human immunodeficiency virus (HIV)-related deaths prior to 1966, the year that effective pro- phylaxis against PCP entered routine use. Our study was performed to study the causes of HIV- related death since January 1988 in a region where patient tracking is virtually complete.

PATIENTS AND MErHoDs: We surveyed physi- cians associated with the Brown University Ac- quired Immunodeficiency Syndrome (AIDS) Program who cared for greater than 95% of known HIV-positive patients in Rhode Island. These physicians identified alI those HIV-infect- ed persons who had died under their care be- tween January 1963 and July 1990, and deter- mined these patients’ causes of death by chart review. For comparison, death certificates of identified persons were also reviewed at the Rhode Island Department of Vital Statistics.

RESULTS: Among 126 deaths since January 1988, bacterial infections were the most common cause of death (30%), whereas PCP was respon- sible for only 16% of deaths. Persons not receiv- ing any form of PCP prophylaxis were more likely to die from PCP than were those who re- ceived prophylaxis (26% versus 11% [p = 0.041). Cause of death as recorded on actual death cer- tificates was imprecise, although bacterial infec- tions were again the most common cause indi- cated. Only one death occurred in a patient with a CD4 count greater than 200/n& and this was not EIIV-related.

CONCLUSION: PCP has not been the leading cause of death in our region since January 1988. Bacterial infections contribute substantially to mortality, and this may influence future prophy- lactic regimens. HIV-related deaths in patients

From the Department of Medicine, Brown University, Providence, Rhode Island.

Requests for reprints should be addressed to Michael D. Stein, M.D., Division of General Internal Medicine, Rhode Island Hospital, 593 Eddy Street, Providence, Rhode Island 02903.

Manuscript submitted October 30,1991, and accepted in revised form April 6. 1992.

with CD4 counts greater than 2OO/mL are ullusual.

C ause of death statistics allow physicians and health policy planners to understand how par-

ticular diseases are affecting specific populations. Within the epidemic of human immunodeficiency virus (HIV), cause of death analyses might also en- able researchers to study changes in the natural history of a new and complex disease. From 1981 to 1990, 101,000 deaths among persons with the ac- quired immunodeficiency syndrome (AIDS) were reported to the Centers for Disease Control (CDC) M.

Pneumocystis carinii pneumonia (PCP) was re- ported to be the predominant cause of AIDS-relat- ed deaths prior to 1988, the year that effective pro- phylaxis against PCP entered routine use [2]. Both necropsy and clinical studies identifying HIV-relat- ed disease report PCP as the predominant cause of death prior to 1988 [3,4]. In Rhode Island, 55% of all HIV-related deaths prior to 1988 were attributed to PCP (Department of Health, personal communica- tion). There is, however, little information available about specific causes of death related to HIV infec- tion since that time. Cause of death is usually ascer- tained from death certificates, which are universal- ly used and uniform in format. However, death certificates are often imprecise or incorrect [5]. Di- rect questioning of physicians regarding cause of death has been shown to improve the quality of reporting [6].

Rhode Island, a small state with only 10 acute care hospitals and a limited number of physicians caring for HIV-infected persons, offers a unique opportunity to study causes of death in AIDS. The purpose of this study was twofold: to determine causes of death since 1988 in our state and to evalu- ate how accurately these causes are reflected on death certificates.

PATIENTS AND METHODS We surveyed all physicians associated with the

Brown University AIDS Program who cared for pa- tients at each of the state’s 10 acute care hospitals.

October 1992 The American Journal of Medicine Volume 93 387

TABLE I TABLE II Study Population (n = 126) Cause of Death (n = 126)

Age (mean + SD) y&y

IVDU ;Gd Death at home (%) 14 Autopsy (%) 7 CD&defined AIDS diagnosis (%) CD4 count < 200/mm3 (%) El; Survival from notification of HIV positivity (mean + SD) 13.3 + 12.3 mo

I IVDU = mtravenous drug user; CDC = Centers for Disease Control; AIDS = acquired immunodefl crency syndrome; HIV = human lmmunodeficiency virus.

These 15 physicians reviewed charts of all HIV- infected persons they had seen who died between January 1988 and July 1990. Rhode Island statutes mandate that all HIV-seropositive cases be report- ed to the Department of Health with appropriate demographic data, but without individual identifi- ers. Based on HIV/AIDS case reporting, these phy- sicians cared for greater than 95% of known HIV- positive persons in Rhode Island.

Our study sample included all HIV-positive per- sons identified by retrieval systems available to the 15 participating physicians. For hospital deaths, data on subjects were retrieved from hospital com- puter files. As part of the Brown AIDS Program, physicians maintained readily retrievable office files on patients seen. All subjects from these two retrieval systems composed the sample. A question- naire for each subject was completed based on rec- ord review by the subject’s attending physician or one of the authors.

We collected the following data: patient demo- graphics, transmission risk, date of HIV seroposi tivity, date and place of death, AIDS-defining diag- nosis, other AIDS diagnoses, use of zidovudine and PCP prophylaxis for at least the 2 months prior to death, last CD4 count prior to death, resuscitation preferences, and autopsy findings when available. A CD4 count cutoff of 200 cells/mm3 was used to iden- tify persons with advanced HIV disease; at this lev- el, patients are candidates for both PCP prophylax- is and antiretroviral therapy [7].

From chart review, the principal condition being treated at the time of death was specified as the immediate cause of death. For example, if a patient receiving long-term therapy for toxoplasmosis de- veloped bacterial pneumonia and died despite the institution of antibiotic therapy, this death was classified as a bacterial death. The death certifi- cates of the identified persons were then reviewed for cause of death at the Rhode Island Department of Vital Statistics. All causes of death listed on stan- dard death certificates were recorded, including the “immediate” cause, conditions leading to the “im-

Physician Responses (%)

On Death Certificates 1%)

Bacterial infections 0;;’ opportunistic infections

Wastinglencephalopathy Liver failure Unknown Lymphoma Non-HIV-related Kaposi’s sarcoma AIDS Cardiopulmonary arrest

25 ;t 18 16 16

8 ;

;

2” ! 1

2: 11

Total 100 100

;P = Pneumocystis canmi pneumonia; HIV = human lmmunodeficiencyvirus; AIDS = acquire, munodeficlency syndrome.

mediate” cause, and “other significant conditions” contributing to death.

The CDC surveillance case definition for AIDS was used for diagnostic classification [a]. The only bacterial infections identified by participating phy- sicians were pneumonia (presumed or microbiologi- tally documented), sepsis, and endocarditis.

RESULTS Surveyed physicians reported 126 deaths be-

tween January 1988 and July 1990 in known HIV- seropositive persons (Table I). The mean age of the sample was 39.9 f 9.6 years; 71% were white; 86% were male; and 30% had used intravenous drugs. Although 87% had CDC-defined AIDS diagnosis, 98% had a CD4 count of less than 200 cells/mm3 prior to death. Fourteen percent of the patients died at home, the remainder in a hospital. Seven percent had autopsies. The mean survival from notification of HIV positivity was 13.3 f 12.3 months.

The leading AIDS-defining diagnosis was PCP (36%). An opportunistic infection other than PCP was the AIDS-defining disease for 28%, Kaposi’s sarcoma for 15%, wasting or encephalopathy for 5%, and lymphoma for 3%. The remaining 13% of per- sons did not have CDC-defined AIDS at the time of death.

Causes of death as reported by physicians and as recorded on death certificates are shown in Table II. Bacterial infections were the leading cause of death (30%) in both. The 29 patients with pneumo- nias represented 74% of all deaths due to bacterial infections; pathogens found in sputum or broncho- alveolar lavage cultures included Staphylococcus aureus (6) and Huemophilus influenzae (5), Strep- tococcus pneumoniae (4), Legionella species (2), and Pseudomonas aeroginosa (1). More than three

399 October 1992 The American Journal of Medicine Volume 93

quarters (79%) of the patients with the cause of death listed as pneumonia had had bronchoscopy within 2 weeks of death without finding Pneumo- cystis on silver stain, thereby excluding PCP as the cause of death. Ten additional persons had sepsis or endocarditis (diagnosed by positive blood cultures), and the most common pathogen was again S. aur- eus. Of the 38 persona who died of bacterial infec- tions, 32 (82%) had met the CDC surveillance AIDS definition.

Twenty-five percent of patients died of other op- portunistic infections, with PCP (diagnosed with bronchoalveolar lavage) the third leading cause of death (16%) according to physician responses. Along with bacterial infections, these were the most common immediate causes listed on death certifi- cates as well. However, 33% of deaths on death cer- tificates were ascribed to “AIDS” and “cardiopul- monary arrest,” which are nonspecific. No death certificates listed cause of death as “unknown,” al- though physicians listed 6% of deaths as unknown after chart review. Non-HIV-related causes of death included gastrointestinal bleeding and myo- cardial infarction. All persons who met the CDC AIDS definition had been reported to the State Health Department.

Table III presents all causes of death listed on death certificates. Although 32 certificates (25%) listed bacterial infections as the immediate cause of death, bacterial infections were listed on 4 addition- al death certificates. PCP was the immediate cause of death on 20 certificates and was listed as an addi- tional cause on 6 others. “AIDS” was listed alone as the immediate cause of death on 28 certificates.

Seventy-two percent of persons had used zidovu- dine for at least 2 months prior to death. No patient had granulocytopenia (fewer than 1.0 X log cells/L) at the time of death. Sixty-one percent of patients used PCP prophylaxis for at least 2 months prior to death; 7% of these persons had used trimethoprim- sulfamethoxazole (TMP/SMX) prophylaxis.

Persons not receiving any form of PCP prophy- laxis were more likely to die of PCP than were those who had received prophylaxis for at least 2 months prior to death (26% versus 11%; p = 0.04), based on physician review.

COMMENTS After 1987, PCP was not the leading cause of

HIV-related death in our state. Earlier diagnosis, better treatment, and increased use of prophylaxis for PCP all probably influenced the decline of PCP as a cause of death. One previous report documents a similar decline in fatal PCP infections [3] but without specifically identifying how cause of death was defined. This decline in fatal episodes of PCP

F

TABLE III All Causes of Death on Death Certificates

Cause No.

Bacterial infections* Alone With opportunistic infections With PCP With AIDS With lymphoma

32 (total)* 16

;

0t~he;~portunistic infections

With bacterial infection With opportunistic infections With PCP With AIDS

23 (total) 13

:

i

PCP Alone With bacterial infection With opportunistic infections With Kaposi’s sarcoma With AIDS

20 (total) 13

!

.:

3 (total) 2 1

Liver failure With opportunistic infections With PCP

3 (total) 2 1

Lvmphoma 1 (total)

Non-HIV-related 1 (total)

Kaposi’s sarcoma 1 (total)

AIDS 28 (total)

Cardiopulmonary arrest Alone With bacterial infection With PCP With lymphoma With AIDS

lbrevlatlons as in Table II.

14 (total)

z

! 5

*Totals represent number hstec as the rmmed~ate cause of death.

points out that access to PCP prophylaxis needs to continue for all populations when clinically indicat- ed [9].

Bacterial infections are emerging as an important cause of death in the AIDS epidemic. These infec- tions are not included in the CDC AIDS surveil- lance definition but are a rising source of mortality in patients with advanced disease as indicated by CD4 counts lower than 200 cells/mm3. Bacterial in- fections were also commonly noted early in the course of HIV infection (often due to S. aureu.s and H. influenzae), and these infections were also noted here [lo]. Our data, from a predominantly male homosexual sample, confirm the effect of bacterial infections on mortality reported among intrave- nous drug users in New York City [ll].

Appropriate antibiotic treatment in patients with CD4 counts less than 200 cells/mm3 may pre- vent premature death. In cases of terminal HIV

October 1992 The American Journal of Medicine Volume 93 389

CAUSES OF DEATH IN HIV INFECTION / STEIN ET AL

disease, treating bacterial processes may, of course, be futile; in this series, reporting physicians clearly believed that antibiotic therapy was indicated and had a reasonable chance of success as indicated by their responses to the cause of death query. Indeed, terminal bacterial infections may have been under- reported given the probability that there were in- stances when cultures were not done and physicians decided treatment was not indicated. Only 7% of our sample used TMP/SMX prophylaxis for PCP, in part because of patient preference in our region for aerosol pentamidine. It is possible that with more widespread use of TMP/SMX, bacterial in- fections will be less common. Our data may provide an additional reason for choosing TMP/SMX as a first-line PCP prophylactic regimen [7].

Death certificates were incomplete in providing specific information on the development of infec- tions and malignancies in persons with HIV. In- deed, in 33% of cases (42 persons), the recorded immediate cause of death (AIDS, cardiopulmonary arrest) was imprecise. On only 4 of these 42 death certificates were HIV-related diseases (for example, lymphoma or PCP) noted. Death certificates may have been completed by covering or resident physi- cians who did not know the patient as well as the deceased’s personal physician and, therefore, inex- act mechanisms of death such as “cardiorespiratory arrest” may have been used rather than precise causes.

Even careful surveillance cannot identify deaths in unrecognized cases of HIV infection or deaths in which physicians do not mention AIDS or HIV on death certificates. However, improved public health surveillance has identified deaths among persons diagnosed with HIV infection without AIDS since update of the ICD-9 classification scheme for HIV in 1987. Underdiagnosis and un- derreporting were expected to be less of a problem in Rhode Island, given the limited number of physi- cians and hospitals who care for virtually all per- sons with HIV infection. Notably, all persons in this series who met the CDC AIDS definition had been reported to the Health Department.

A review of deaths in New York City noted that 9% of HIV-infected persons did not have illnesses meeting the AIDS case definition [12]; in our series, 13% did not have CDC-defined AIDS. Interestingly, if the CDC extended the AIDS definition to all per-

sons with a CD4 count of less than 200/mm3, all but one person in this series would have AIDS.

Because only 7% of deaths were followed by au- topsies, cause of death as described by physicians here remains inconclusive. Previous surveys have demonstrated that autopsies often reveal AIDS-re- lated diseases that were not suspected clinically [4]. In the autopsies done in our series, however, no unsuspected diseases were discovered that would explain cause of death. In addition, it is possible that some pneumonia deaths classified as bacterial may have been due to PCP; however, the high rate of bronchoscopy close to the time of death makes this possibility less likely.

Only a single person in our series with a CD4 count higher than 200 cells/mm3 died, and this death was not HIV-related. Due to improved diag- noses and treatments, including antiretroviral and PCP prophylaxis, deaths due to PCP, the most common opportunistic infection in the first years of the epidemic, are declining. Early HIV testing and entry into the health care system should permit persons with HIV infection to live longer.

REFERENCES 1. Centers for Disease Control. Mortality attributable to HIV infection/AIDS- United States, 1981-1990. MMWR 1991; 40: 41-4. 2. Kovacs JA, Masur H. Prophylaxis of Pneumocysfis carinii pneumonia: an update. J Infect Dis 1989; 160: 882-6. 3. Peters ES. Beck EJ, Coleman DG. eta/. Changing disease patterns in patients with AIDS in a referral center in the United Kingdom: the changing face of AIDS. BMJ 1991; 302: 203-7. 4. Wickes MS, Fortin AH, Felix JC. et al. Value of necropsy in acquired immuno- deficiency syndrome. Lancet 1988; 2: 85-8. 5. Rosenberg HM. Improving cause-of-death statistics. Am J Public Health 1989; 79: 563-4. 6. Hopkins DD, Grant-Worley JA, Bollinger TL. Survey of cause-of-death query criteria used by state vital statistics programs in the US and the efficacy of criteria used by the Oregon Vital Statistics Program. Am J Public Health 1989; 79: 57D-4. 7. Volberding PA. Recent advances in the medical management of early HIV disease. J Gen Intern Med 1991; 6 Suppl: S7-12. 6. Centers for Disease Control. Revision of the CDC case definition for acquired immunodeficiency syndrome. MMWR 1987; 36: lS-15s. 9. Piette J, Stein M. Mor V, et a/. Patterns of secondary prophylaxis with aerosol

pentamidine among persons with AIDS. J AIDS 1991; 4: 826-8. 10. Polsky B, Gold JWM, Whimbey E, eta/. Bacterial pneumonia in patients with the acquired immunodeficiency syndrome. Ann Intern Med 1986; 104: 38-41. 11. Stoneburner !?A, DeJarlais DC, Benezra D. Alarger spectrum of severe HIV-1 related disease in intravenous drug users in New York City. Science 1989; 242: 916-9. 12. Hindin RH, Thomas P, Nicholas A, et al. Evaluating completeness of New York City’s case registry. Presented at the Fifth International Conference on

AIDS, Montreal, June 4-9. 1989.

390 October 1992 The American Journal of Medicine Volume 93