Does HIV status influence the outcome of patients admitted to a surgical intensive care unit? A prospective double blind study

BMJ 1997; 314 doi: (Published 12 April 1997) Cite this as: BMJ 1997;314:1077
  1. Satish Bhagwanjee, lecturer in anaestheticsa,
  2. David J J Muckart, senior lecturer in surgerya,
  3. Prakash M Jeena, lecturer in paediatricsa,
  4. Prushini Moodley, honorary lecturer in haematologya
  1. Faculty of Medicine University of Natal Private Bag 7 Congella 4013 South Africa
  1. Correspondence to: Dr Bhagwanjee
  • Accepted 12 December 1996


Objectives: (a) To assess the impact of HIV status (HIV negative, HIV positive, AIDS) on the outcome of patients admitted to intensive care units for diseases unrelated to HIV; (b) to decide whether a positive test result for HIV should be a criterion for excluding patients from intensive care for diseases unrelated to HIV.

Design: A prospective double blind study of all admissions over six months. HIV status was determined in all patients by enzyme linked immunosorbent assay (ELISA), immunofluorescence assay, western blotting, and flow cytometry. The ethics committee considered the clinical implications of the study important enough to waive patients' right to informed consent. Staff and patients were blinded to HIV results. On discharge patients could be advised of their HIV status if they wished.

Setting: A 16 bed surgical intensive care unit.

Subjects: All 267 men and 135 women admitted to the unit during the study period.

Interventions: None.

Main outcome measures: APACHE II score (acute physiological, age, and chronic health evaluation), organ failure, septic shock, durations of intensive care unit and hospital stay, and intensive care unit and hospital mortality.

Results: No patient had AIDS. 52 patients were tested positive for HIV and 350 patients were tested negative. The two groups were similar in sex distribution but differed significantly in age, incidence of organ failure (37 (71%) v 171 (49%) patients), and incidence of septic shock (20 (38%) v 54 (15%)). After adjustment for age there were no differences in intensive care unit or hospital mortality or in the durations of stay in the intensive care unit or hospital.

Conclusions: Morbidity was higher in HIV positive patients but there was no difference in mortality. In this patient population a positive HIV test result should not be a criterion for excluding a patient from intensive care.

Key messages

  • HIV positive patients admitted to intensive care for diseases unrelated to their HIV status have a similar mortality and duration of stay when compared with HIV seronegative patients

  • The incidence of septic shock and multiple organ dysfunction is higher in HIV seropositive patients and needs further investigation

  • HIV status cannot be used to deny critically ill patients admission to intensive care

  • The HIV and AIDS epidemic raises unique ethical considerations that must be carefully addressed during clinical studies


Extensive data are available on the outcome of patients with AIDS admitted to intensive care units.1 2 The survival rate of these patients has greatly improved over the past 15 years, and refusal to provide intensive care to these patients on the basis of medical futility is therefore deemed unjust.1 In Africa the pattern of HIV disease is different from that in the developed world, more patients manifesting early HIV disease and fewer progressing to AIDS.3 This pattern is reflected in our intensive care unit, where most patients are admitted with diseases unrelated to their HIV status. To our knowledge the outcome of patients with HIV infection admitted to intensive care for other reasons has not been described. Our clinical impression was that their outcome was poor.

Limited resources and the high cost of intensive care have compelled clinicians to rationalise the allocation of resources.1 For example, in our unit it is policy not to admit patients with incurable malignant disease, end stage liver disease, and patients with multiple organ failure who are deemed non-salvageable. The lack of objective data made it unclear whether patients with HIV infection should be treated similarly. To allow rationalisation of the admissions policy with respect to these patients we conducted a prospective study to determine the prevalence of HIV infection among patients admitted to the unit and assess the impact of HIV status (HIV positive, HIV negative, AIDS) on outcome.

The study embraced a major ethical dilemma. On the one hand, the clinician has an obligation of non-maleficence–that is, patients must not be harmed by the actions of the doctor. On the other hand, the doctor has an obligation to society to ensure that available resources are appropriated fairly, based on objective evidence. Though the basic ethical tenets of patient autonomy, justice, beneficence, and non-maleficence4 are useful, they are only the starting points for ethical decision making.

Subjects and methods

The study was conducted in the 16 bed surgical intensive care unit at King Edward VIII Hospital, a large teaching hospital in Durban. All patients admitted to the unit over six months (September 1993 to February 1994) were included. There were no exclusions. Informed consent was not sought. The study protocol was approved by the ethics committee of the University of Natal.

A screening enzyme immunoassay for HIV (Abbott HIV-1/HIV-2 third generation plus kit; Abbott Laboratories, Chicago) was performed on all patients at admission. Positive results were confirmed by the department of virology using an immunofluorescence assay (SEROFLUOR; Virion, Switzerland) and western blotting (HIV western blot 1/2; Diagnostic Biotechnology, Singapore). Patients with positive results in the confirmatory tests were considered HIV positive. The department of haematology was informed of these results and requested a specimen of blood for flow cytometry from three patients, one of whom was HIV positive. Staff were thereby blinded to which patients were HIV positive. Flow cytometry was performed on whole blood samples from all HIV positive patients with Coulter's Q-prep method (commercially produced antibodies from the Coulter Corporation, Miami). Samples were analysed on an Epics Profile II Coulter flow cytometer.

In addition to the intensive care unit staff, patients also were blinded to the results of the HIV tests. The protocol permitted disclosure of HIV status to staff in two instances: (a) if a staff member sustained a needlestick injury–when the injured staff member, the consultant in charge of the patient, and the matron in charge would be informed of the result; (b) if a patient required haemodialysis–when the nurse undertaking haemodialysis and the consultant in charge would be informed of the result.

On discharge all patients were advised that they had been tested for HIV and of the reason for testing and given the option of knowing the result. Post-test counselling was offered to patients when HIV results were disclosed. Results of HIV testing were made available to the research team only after the patient had been discharged and all other data had been collated. On conclusion of the study results of HIV testing were permanently removed from laboratory records.

Three groups of patients were defined. HIV positive and HIV negative patients were identified by HIV testing; patients with AIDS were identified by Centers for Disease Control criteria.5 The following data were recorded in all patients: demographic details; admission diagnosis and referring discipline; APACHE II score (acute physiological, age, and chronic health evaluation) in the first 24 hours after admission6; incidence of organ failure as defined by Knaus et al7; incidence of sepsis and septic shock as defined by the American College of Chest Physicians and the Society of Critical Care Medicine8; incidence of nosocomial sepsis as defined by our intensive care unit protocol; durations of intensive care unit and hospital stay (duration of hospital stay did not include intensive care unit stay); intensive care unit and hospital mortality (hospital mortality did not include intensive care unit mortality).

Admission to the unit and treatment offered were not influenced by HIV status. All patients were treated according to standard intensive care unit protocols.


HIV positive and HIV negative patients were compared by Student's t test and the χ2 test for continuous and discrete variables respectively. A probability value of less than 0.05 was considered significant. Associations between HIV status and outcome variables were adjusted for differences in age and analysed by logistic regression. Kaplan-Meier estimates were used to compute the survival distribution function estimates within the HIV positive and HIV negative groups (non-survivors) and the equality of the distributions tested by the Wilcoxon rank sum test. Age was added as a covariate.


No patient had AIDS. The rest of the data therefore refer to HIV positive and HIV negative patients only. Of the 402 patients admitted to the unit during the six months, 52 (13%) tested positive for HIV. Though the male to female distribution in the two groups was similar, they differed significantly in age (P<0.002; table 1).

Table 1

Age and sex distribution of HIV negative and HIV positive patients

View this table:

Most patients in both groups were admitted after trauma (table 2). HIV infection was more common in patients referred from orthopaedic surgery and obstetrics and gynaecology. There was no significant difference in intensive care unit or hospital mortality or in the duration of intensive care unit or hospital stay (table 3 and 4). Intensive care unit mortality in HIV negative patients was 24% (84/350) compared with 29% (15/52) in HIV positive patients (odds ratio 1.45; 95% confidence interval 0.75 to 2.80); hospital mortality was 6% (16/247) and 3% (1/37) in the two groups respectively. There was no significant difference in survival distribution between the groups (mean survival time 7.3 (SE 2.6) days in the HIV positive group, 6.9 (0.78) days in the HIV negative group; P=0.88) (fig 1). Information regarding hospital mortality and duration of hospital stay could not be retrieved for 19 HIV negative patients. There was no significant difference in mean APACHE II score between HIV negative and HIV positive patients (scores 9 and 8 respectively).

Fig 1
Fig 1

Survival distribution function versus duration of stay in intensive care unit

Table 2

Interdisciplinary distribution of patients

View this table:
Table 3

Comparison of mortality between HIV negative and HIV positive patients

View this table:
Table 4

Mean and median (range) number of days' stay in intensive care unit and hospital for HIV positive and HIV negative patients

View this table:

Organ failure was more prevalent in HIV positive patients. Significant differences were found when cardiac, respiratory, and haematological system failures were compared (table 5). Though there was no difference in the incidence of severe sepsis and nosocomial sepsis, septic shock was significantly more common in HIV positive patients (table 6).

Table 5

Comparison of total and individual organ failures between HIV negative and HIV positive patients

View this table:
Table 6

Incidence of sepsis in HIV negative and HIV positive patients

View this table:

It was not possible to perform flow cytometry on all HIV positive patients. This was because either the patient died soon after admission or the request for testing came after the patient was discharged from the unit and could not be reached. Compared with normal values there were significant differences in T4 count, T4:T8 ratio, and B4 count (table 7). The T4:T8 ratio was reversed and B4 count reduced in HIV positive patients.

Table 7

Flow cytometry results in HIV positive patients. Cell counts are means (SE)

View this table:

Accidental disclosure of HIV status occurred in one instance as a result of a laboratory error. The researcher who became aware of the result did not divulge it to other staff and did not participate in management decisions regarding the patient. Data for this patient were collated by another member of the team, who was unaware of the result. As permitted by the protocol, the HIV status of five other patients became known to relevant staff members before patient discharge. One case involved a needlestick injury to a staff member, and the remaining disclosures were in preparation for haemodialysis. In all instances management decisions and collation of patient data were by other, blinded researchers. Of all 402 patients tested, only three wished to be informed of their HIV status. No patient objected to having been included in the study without prior informed consent.


Mortality is the best measure of outcome of patients treated in an intensive care unit. Markers of morbidity may be subjective and are therefore less reliable end points. In this study there was no difference in intensive care unit or hospital mortality between HIV positive and HIV negative patients when results were adjusted for age (table 3).

HIV positive patients were more prone to septic shock and organ failure, and we were therefore surprised that the duration of hospital stay and mortality were not increased. This finding is unlikely to have been the result of observer error because, except for one inadvertent disclosure, HIV results were not available until all other data were collated. Abnormalities in flow cytometry results may have been an important factor. HIV positive patients had low T4 counts whereas T8 counts were comparatively high. As a consequence the T4:T8 ratio was reduced but the total (T3) was not affected. The B4 count was also low. All these features are consistent with the latent phase of HIV infection.9 Though the behaviour of these cell populations predicts clinical progression of HIV disease to AIDS,10 to our knowledge its impact on intensive care unit patients admitted for non-HIV related disease has not been described. Conceivably the immune response to major trauma and sepsis is altered. The observation by Munoz et al that HIV negative patients with sepsis and impaired macrophage responsiveness are more prone to subsequent sepsis11 lends credence.

Immunological mechanisms have been postulated to play a major part in the pathogenesis of septic shock and multiple organ failure.12 13 The immune response is complex and paradoxical, pro-inflammatory and anti-inflammatory responses occurring simultaneously and both being mediated by cytokines.14 This has prompted the use of new drugs which alter the immune response in sepsis.15 16 We therefore postulate that, though HIV positive patients have disturbances in immune function which make them more susceptible to septic shock and multiple organ failure, the inflammatory response is also altered such that there is no increase in mortality.

The patients in this study were young, predominantly male, and admitted primarily after trauma or surgery. That no patient had AIDS concurs with Gilks's observation that the pattern of HIV infection in Africa differs from that in the developed world.3 Non-HIV disease is far more prevalent in Africa, with rapid progression from seroconversion to HIV to death from an AIDS defining condition.3 Data relating to outcome in patients with AIDS cannot therefore be extrapolated to our patients. This emphasises the importance of describing the outcome in patients admitted to intensive care with non-HIV related disease.

Issue of informed consent

Decisions on initiating and terminating care for critically ill patients are difficult.17 The unique nature of the AIDS epidemic in Africa,3 the tremendous costs associated with advanced life support,1 as well as the particular ethical considerations in patients with HIV infection18 are compelling reasons for these decisions to be based on sound ethical principles and objective evidence of disease outcome. In view of the lack of clinical information in our patient population the acquisition of objective data was imperative. A major ethical dilemma arose when the decision was made not to seek informed consent. This was thought to be essential, as patients who were likely to be at risk for HIV infection would also be inclined to refuse the study, which would seriously limit its value.

There were two consequences of the study. Firstly, patients were denied the option of being excluded and, secondly, they were at risk of having their HIV status disclosed. The first consideration was evaluated in terms of the potential benefit of the study to society as a whole. The consensus of the research team and the ethics committee was that the clinical implications of the study were enough to warrant denying patients the right of refusal. With respect to the second consequence, every effort was made in the design and execution of the study to ensure that indiscriminate disclosure of HIV results did not occur. To our knowledge HIV results were not disclosed except for study purposes and, furthermore, patient care was not influenced by HIV status.

There was no reason to suspect that the racial background of our patients would have any bearing on their outcome. Race as a demographic variable is considered only rarely in South Africa.19 Our main criterion for denying patients admission to intensive care is futility. This study showed no significant difference in mortality between HIV positive and HIV negative patients. Though the incidence rates of septic shock and organ failure were higher, this did not influence mortality or duration of stay. We therefore conclude that in our patient population HIV status cannot be used as a criterion for denying patients admission to the intensive care unit. Our observations regarding septic shock and organ failure require further evaluation.


Part of this study was presented at the 12th annual critical care congress of the South African Critical Care Society (1995) and at the eighth European congress of intensive care medicine. We thank Mrs Q A Karim, Dr S S A Karim, Professor H M Coovadia, Dr E M Barker, Professor D J Pudifin, Professor A N Smith, Professor J Lipman, and the ethics committee for advice and Miss E Gouws for statistical analysis. We also thank Mr T Doorasamy, technicians in the department of virology, Mr H Benimadho, Mr N Bhimsan, and Mr R Loykisoonlal for technical help and Mrs A Pillay for secretarial work.

Funding: The University of Natal's research and travel committee and the Medical Research Council of South Africa.

Conflict of interest: None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
View Abstract