Abstract
Objectives
A prospective cohort study was conducted among HIV-infected inpatients with tuberculosis or other opportunistic infection (OI) in South Africa to estimate subsequent antiretroviral therapy (ART) uptake and survival.
Methods
Logistic regression modeling explored associations between baseline characteristics and starting ART, and ART exposure adjusted incidence of death was estimated over 6 months of follow-up.
Results
Among 49 participants enrolled, median CD4 cell count at hospital discharge was 42 cells/mm3, and the most common presenting OI were tuberculosis (76 %), Pneumocystis pneumonia (8%), chronic diarrhea (8%), cryptococcal meningitis (6%), and Toxoplasmosis gondii (4%). By 6 months, only 20 (45%)had initiated ART, and 4 (8%) were lost to follow-up. ART uptake was independently associated with previous use of traditional medicine (OR 7.2; 95% CI 1.4 – 55.1) and with less advanced HIV infection (baseline CD4 count (per 50 cells/mm3increase OR 1.4; 95% CI 0.9–2.2). A total of 14 (31%)died before initiating ART; the monthly incidence of death did not decrease over the 6 month interval.
Conclusion
High mortality observed within the 6 months following hospitalization with tuberculosis or other acute OI argue that mechanisms are needed to expedite ART for patients after AIDS-defining illness.
BACKGROUND
Significant pre-ART mortality has been documented among HIV-infected outpatients awaiting initiation of antiretroviral therapy in resource-limited settings (ART) (1, 2). In South Africa, the delay of ART until completion of tuberculosis (TB) therapy was associated with an increased risk of death among ART-eligible patients (3). A growing evidence base suggests that patients with recent OI may benefit from rapid ART initiation (4). However, following discharge from the hospital HIV-infected patients with recent OI, patients in resource-limited settings with recent OI face several potential barriers to timely initiation of ART potentially placing them at increased risk for poor outcome. These barriers include the loss of functional capacity associated with hospitalization, as well as structural hurdles such as pre-ART readiness assessment processes. A significant delay in initiating ART would be expected to adversely impact survival, estimated to be only 1.3 years after AIDS-defining illness (5).
In order to improve ART uptake and reduce pre-ART mortality among ART-eligible inpatients with TB or other OI, a pilot program was initiated at McCord Hospital in Durban, South Africa to expedite access to ART after hospitalization. In the pilot program, HIV-infected patients with tuberculosis or other OI transitioned rapidly from the inpatient ward to a step-down center for initiation of ART. Eligible patients who did not enter this pilot program – and instead sought ART as outpatients – were also prospectively followed and their outcomes are presented here.
METHODS
McCord Hospital is a 166 bed general hospital in South Africa that provides inpatient care to patients from Durban townships. HIV-infected patients with tuberculosis or other acute opportunistic infection comprise approximately half of admissions to the McCord Hospital medical wards. As part of routine inpatient care, HIV and CD4 count testing are performed, and at discharge, HIV-infected patients with CD4 count <200 cells/mm3are directed to enroll as out patients for ART at the Sinikithemba Clinic at McCord Hospital (large vertical site) or at one of several other local clinics.
During the period of this study, from December 2006 until February 2007, HIV-infected, ART-naïve patients hospitalized at McCord Hospital with tuberculosis or other acute OI were approached to enter a pilot project which consisted of immediate initiation of ART in a step-down center with D4T-3TC-EFV. Patients were required to pay for the cost of additional hospitalization days (median 14 days [IQR 12–19]) associated with the immediate ART program. Patients who did not participate in this pilot project were prospectively enrolled and followed.
Enrollment criteria
HIV-infected, ART-naïve patients of 18 years of age or greater with a CD4 cell count of <200 cells/mm3 (or with WHO Stage IV disease) were prospectively enrolled. Patients were excluded if they did not verbalize both a readiness to initiate ART after discharge and a willingness to be contacted after 6 months for follow-up.
Data collection
Clinical, demographic and laboratory data were collected during admission and clinical outcomes were collected at month 6. Data collected at admission included age, gender, education, prior OI history, date of HIV diagnosis, use of traditional medicines, body mass index, presence of thrush, CD4 cell count, albumin, alanine aminotransferase (ALT), creatinine and hemoglobin. Six months following hospital discharge, data collected included current medications, date and location of current ART clinic (if enrolled), new clinical events (hospitalizations and OI), and vital status (obtained from emergency contact if necessary). At that time patients were also offered a clinic visit at McCord Hospital if they had not seen a physician in the last month.
Study Design
We conducted an observational, non-randomized prospective cohort study. The primary outcomes were ART initiation within 6 months of hospital discharge, and death within 6 months of hospital discharge.
Statistical Analysis
Multivariable logistic regression modeling explored the associations between baseline (i.e. pre-hospital discharge) characteristics and ART initiation at 6 months. Covariates included in the modeling process were age, gender, education, use of traditional medicines, use of cotrimoxazole, date of HIV diagnosis (concurrent with current OI versus prior), body mass index, presence of thrush, tuberculosis (versus other OI), CD4 cell count, lymphocytes percentage, alanine aminotransferase (ALT), hemoglobin, creatinine and albumin. All covariates were initially fit alone (univariate models), and only those suggestive of an association were fit in a multivariable model. Logistic regression analyses were performed using SAS software, version 9.1. All tests of statistical significance were 2-sided, without adjustment for multiple testing.
Summaries of survival used ART exposure stratified and interval specific incidence rate estimation. Both the incidence point estimate and a 95% confidence interval on the estimate are presented. To estimate survival by ART exposure, if and when individuals started ART, their subsequent follow-up survival time contributed to the post-ART initiation stratum. Incidence estimation was performed using STATA software, version 8.2.
Ethical approval
The study was approved by the McCord Hospital Research Ethics Committee in Durban, South Africa.
RESULTS
Forty-nine patients with HIV and tuberculosis or other acute OI requiring hospitalization were prospectively enrolled between December 2006 and February 2007 (Figure 1). The reasons cited for declining participation in the immediate ART program were: cost of hospitalization (74%), family or employment obligation (19%) or other (7%).
Figure 1.
Inclusion of patients in prospective study of ART uptake and mortality after hospitalization with acute OI at McCord Hospital in Durban, South Africa (12/06 – 2/07)
Baseline characteristics at admission with tuberculosis or other acute OI
At admission the median age was 34 (interquartile range (IQR) 28–40), 49% were female, and the median CD4 cell count was 42 cells/mm3 (IQR 14–106)(Table 1). The median time since initial HIV diagnosis was 2.9 months ((IQR) 1.1–3.9 months) and for 67% of patients, the current admission was the first OI. The admission opportunistic infections were: pulmonary tuberculosis; 28%; extra pulmonary tuberculosis (including tuberculosis meningitis), 38%; Pneumocystis pneumonia, 8%; chronic diarrhea (>30 days), 8%; cryptococcal meningitis, 6%; Toxoplasmosis gondii, 4%; and unknown, 8%. More than one third (37%) of patients had a low body mass index of <18.5 kg/m2and the median admission albumin, creatinine, and hemoglobin were 2.65 g/dl, 1.05 mg/dl, and 10.5 g/dl, respectively.
Table 1.
Baseline characteristics of patients at hospital admission with acute OI in South Africa
| Characteristics | |
|---|---|
| N=49 | |
| Median age – years (intraquartile range) | 34 (28–40) |
| Women – no. (%) | 24 (49) |
| Education less than secondary school – no. (%) | 26 (53) |
| Median time since initial HIV diagnosis – months (IQR) | 2.9 (1.1–3.9) |
| On an ART waiting list at admission – no. (%) | 10 (20) |
| Alternative medicine use in the last three months | |
| Product sold as “immune booster” – no. (%) | 14 (29) |
| Traditional medicine – no. (%) | 10 (20) |
| History of prior OI – no. (%) | 16 (33) |
| Opportunistic infection at admission – no. (%) | |
| Pulmonary tuberculosis | 14 (28) |
| Extrapulmonary tuberculosis (including TB meningitis) | 18 (38) |
| Pneumocystis pneumonia | 4 (8) |
| Chronic diarrhea(≥30 days) | 4 (8) |
| Cryptococcal meningitis | 3 (6) |
| Toxoplasmosis gondii | 2 (4) |
| Other | 4 (8) |
| Median weight – kg (IQR) | 57.3 (50–62) |
| Median body mass index – kg/m2 (IQR) | 20.3 (17.6–25.4) |
| Body mass index < 18.5 kg/m2 (underweight) — no. (%) | 18 (37) |
| Median CD4 count – cells/μl (IQR) | 42 (14–106) |
| CD4 cell count category – no. (%) | |
| < 50 cells/μl | 27 (55) |
| 51 – 199 cells/μl | 18 (36) |
| > 200 cells/μl | 3 (6) |
| Biochemical parameters: | |
| Albumin — median 1 | 2.65 g/dl |
| Creatinine — median | 1.05 mg/dl |
| Hemoglobin — median | 10.5 g/dl |
| Alanine aminotransferase (ALT) – median 1 | 25.5 IU/l |
Fifteen patients did not have a baseline albumin and ALT measured
ART initiation within 6 months of hospital discharge
Follow-up at 6 months following hospital discharge was available for 45 of 49 (92%) patients (Table 2). By this time, 20 (45%) of 45 patients had initiated ART. Among those who initiated ART, the observed time from discharge until beginning ART therapy was a median of 83 days (interquartile range 34.5–119.5). Four participants started ART within 1 month of discharge, 2 between 1 and 2 months, 6 between 2 and 3 months, 3 between 3 and 4 months, and the remaining 5 between 4 and 5 months. Pre-discharge covariate associations with initiating ART were estimated (Table 3). Use of traditional medicine at baseline was associated with higher odds of 6-month ART initiation; 78% of patients using traditional medicine at baseline initiated ART by 6 months compared to 36% among patients who denied recent use of traditional medicines (P=0.04). Furthermore, patients with a higher CD4 cell count at baseline of >50 cells/ul had an increased odds of initiating ART by 6 months compared to those with aCD4 cell count of ≤50 cells/ul (OR, 3.6; 95% CI 1.04–12.8). This observation was partially explained by an increased mortality among patients with baseline CD4 count of <50 cells/ul (10 of 25 participants with CD4 < 50 cells/ul later died compared to 4 of 19 participants with CD4 ≥ 50 cells/ul, Fisher’s exact test p=0.2). In univariate analysis, there was not a significant association identified between ART initiation and other baseline factors including education, date of HIV diagnosis (concurrent with hospitalization versus prior), BMI, baseline tuberculosis (versus other OI), ALT, hemoglobin, creatinine and albumin.
Table 2.
Outcomes through 24-weeks following discharge with acute OI in KwaZulu-Natal, South Africa
| Patients N=45† |
|
|---|---|
| Overall incidence of death after discharge, per 100 person/years (95% CI) | 0.19 (0.11–0.32) |
| Incidence of death prior to ART initiation | 0.24 (0.14–0.42) |
| Incidence of death following ART initiation | 0.05 (0.0–0.37) |
| Initiation of ART by 24 weeks | |
| Initiated ART, no. (%) | 20 (46) |
| Median observed days from admission to ART initiation [IQR] (n=20) | 83 [34.5–119.5] |
| Mortality at 24 weeks | |
| Died before initiating ART – no. (%) | 13 (29) |
| Died after initiating ART – no. (%) | 1 (2) |
| Median observed days from discharge to death [IQR] (n=14) | 98.5 [51–149] |
Four patients were not traceable with contact information provided among 49 patients enrolled.
Table 3.
Factors associated with initiation of ART after a cute OI in KwaZulu-Natal, South Africa
| N | ART initiated % | Univariate Odds Ratio 95% CI | Multivariate Odds Ratio 95% CI | |
|---|---|---|---|---|
| All subjects | 45 | 44 | ||
| Gender | ||||
| Male | 23 | 39 | ||
| Female | 22 | 50 | 1.6 (0.5 – 5.1) | |
| Age | ||||
| 0–29 years | 15 | 47 | ||
| ≥30 | 30 | 43 | 0.9 (0.3 – 3.0) | |
| Baseline traditional medicine use in past three months | ||||
| None | 36 | 36 | ||
| Traditional medicine use in prior 2 months | 9 | 78 | 6.2 (1.1 – 34.3) | 7.2 (1.4 – 55.1) |
| Diagnosed with HIV simultaneous with OI admission | ||||
| No | 25 | 48 | ||
| Yes | 20 | 40 | 0.9 (0.8 – 1.0) | |
| Opportunistic infection | ||||
| Tuberculosis (pulmonary or extrapulmonary) 29 | 45 | |||
| OI other than tuberculosis | 16 | 44 | 1.1 (0.3 – 3.6) | |
| Initial CD4 cell count | ||||
| 0–49 cells/ul | 25 | 32 | ||
| >50 cells/ul | 20 | 60 | 3.6 (1.04 – 12.8) | 1.4 (0.9 – 2.2) |
| Baseline BMI | ||||
| ≥18.5 kg/m2 | 27 | 41 | ||
| <18.5 kg/m2 | 18 | 50 | 1.5 (0.4 – 4.8) | |
| Hemoglobin baseline | ||||
| ≥9 | 35 | 43 | ||
| <9 | 10 | 50 | 1.3 (0.3 – 5.5) | |
| Baseline albumin 1 | ||||
| ≥ 2.5 mg/dl | 21 | 38 | ||
| < 2.5 mg/dl | 13 | 46 | 1.4 (0.3 – 6.5) | |
Fifteen patients did not have a baseline albumin measured
In multivariate analysis, only recent use of traditional medicines remained associated with increased odds of initiating ART by 6 months (OR, 7.2; 95%CI 1.4–55.1; P=0.03; Table 3) and CD4 cell count at baseline of > 50 cells/ul showed a borderline association with an increased odds of initiating ART (OR, 1.4; 95% CI 0.9–2.2; P=0.14).
Mortality within 6 months following discharge with tuberculosis or other acute OI
Overall, by 6 months of follow-up, 14 (31%) of 45 patients with follow-up information available had died (Table 2). In total, 13 of the deaths occurred prior to initiation of ART and 1 occurred 5 days following ART initiation. (A total of 2 deaths occurred within the 1st month following discharge, 3 between 2 and 3 months, 5 (including 1 following ART initiation) occurred between 3 and 4 months, and 2 in each of the intervals of 4 to 5 and 5 to 6 months following discharge.)
This cohort had a total of 20 years of person-time follow-up; the overall incidence of death was 0.19 per 100 person/years (95% confidence interval 0.11 to 0.32) (Table 2). The overall incidence of death following initiation of ART (with a total of 5.23 years of follow-up available) was 0.05 (95% CI 0.01 to 0.37); the overall incidence rate of death before starting ART was 0.24 (95% CI 0.14 to 0.42). Before initiating ART, the monthly incidence rate (per 100 person years)of death was as follows: 0.16 ( 0 up to 1 month following discharge), 0.29 (2 up to 3 months), 0.54 (3 up to 4 months), 0.45 (4 up to 5 months), and 0.692 (5 up to 6 months).
DISCUSSION
In summary – after discharge – patients with TB and other OI enter a dangerous clinical trajectory in South Africa that includes poor ART access and high early mortality. Six months after discharge, more than half of HIV-infected patients with tuberculosis (or other OI) had not initiated ART. During the 6 months after discharge, the incidence of death remained high for those not initiating ART. Patients with the most advanced disease (CD4 count <50 cells/ul) appear to have been less likely to initiate ART after admission (32% initiated) compared to patients with less advanced HIV disease (60% initiated). The overall high mortality in this cohort – and the observation that the incidence of mortality prior to starting ART did not fall over time – suggests that, in South Africa, mechanisms are urgently needed to expedite ART access for patients with advanced HIV disease after OI. This subgroup does not appear to have received adequate attention in an ART roll-out, which has been oriented towards the outpatient sector.
Although acute admission with AIDS-defining illness continues to be an important initial presentation of HIV in southern Africa, inpatients have been not been a major focus of the ART roll-out. In South Africa –after inpatient admission – patients with HIV infection qualifying for ART must queue with out patients (often with less advanced disease), and participate in the same pre-ART readiness assessment process. As this study demonstrates, the current model may impose significant barriers for this vulnerable patient group. To improve the continuity of care after inpatient admission, new treatment models are required for HIV-infected patients presenting to inpatient wards. These could include the use of an inpatient ART initiation model or outpatient “fast-track” mechanisms that could triage patients with tuberculosis or other OI to outpatient ART initiation within weeks, not months, with less stringent preparation and training requirements.
A significant association was found between the baseline use of traditional medicine and an increased odds of ART initiation by 6 months. The use of traditional medicines may be a marker for health-seeking behavior. In addition, an association was found between CD4 cell count at the time of admission and subsequent likelihood of ART initiation; the most immunologically advanced patients appeared least likely to initiate ART by 6 months.
The study has important limitations. The small sample size makes it difficult to model multiple baseline factors in a single model to predict the primary outcome of ART initiation after OI and therefore, the study had power to detect only large associations with this outcome. The findings presented in this report cannot be generalized to all settings in South Africa. The site of the study – McCord Hospital – includes a hospital and ART clinic with adequate human resources. It is likely that the outcomes described among HIV-infected patients with tuberculosis and other acute OI in this report represent, compared to the public hospital sector in KwaZulu-Natal, a best-case scenario. A potential source of bias relates to the inclusion criteria for this study; participants were patients with acute OI who were eligible but elected not to participate in an ongoing inpatient ART initiation program that required an additional two week hospitalization with associated costs.
Despite an initial intention by all patients to initiate outpatient ART, less than half succeeded by 6 months. To meet the needs of this patient group, creative and pragmatic new care models are needed. Researchers must initiate high quality patient-oriented research to meet this gap and urgently communicate results from ongoing programs focused on this vulnerable patient group.
Acknowledgments
We acknowledge the critical efforts of the staff at McCord Hospital in Durban, South Africa, particularly the supporting work of the counselors, medical records staff, nurses, and medical officers. We received key technical assistance from Sajiv Pertab.
Funding
The project was supported by the Mark Schwartz Fellowship in Global Health, Partners AIDS Research Center, McCord Hospital, and by NIH grants K24-RR016482 and P30 AI060354.
Contributor Information
Richard A. Murphy, Doctors Without Borders USA, [Study design, data collection, analysis and manuscript preparation]
Henry Sunpath, McCord Hospital, Durban, South Africa, [Study design and analysis].
Bushra Taha, Harvard Medical School, Boston, MA, [Data collection and manuscript preparation].
Shanthi Kappagoda, Stanford University, Div. of Infectious Disease and Geographic Medicine, Palo Alto, CA, [Data collection and manuscript preparation].
Khotso Tony M. Maphasa, Zoe-Life, Durban, South Africa, [Data collection and instrument design]
Daniel R. Kuritzkes, Section of Retroviral Therapeutics, Brigham and Women’s Hospital, Boston, MA, [Study design]
Laura Smeaton, Ctr for Biostatistics in AIDS Research and Harvard School of Public Health, Boston, MA, [Data analysis].
References
- 1.Bassett IV, Wang B, Chetty S, Mazibuko M, Bearnot B, Giddy J, et al. Loss to care and death before antiretroviral therapy in Durban, South Africa. J Acquir Immune Defic Syndr. 2009 Jun;Jan;51(2):135–9. doi: 10.1097/qai.0b013e3181a44ef2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Lawn SD, Myer L, Orrell C, Bekker LG, Wood R. Early mortality among adults accessing a community-based antiretroviral service in South Africa: implications for programme design. AIDS. 2005 Dec 2;19(18):2141–8. doi: 10.1097/01.aids.0000194802.89540.e1. [DOI] [PubMed] [Google Scholar]
- 3.Abdool Karim S, Naidoo K, Grobler A, Padayatchi P, Nair G, Bamber S, et al. Initiating ART during TB Treatment Significantly Increases Survival: Results of a Randomized Controlled Clinical Trial in TB/HIV-co-infected Patients in South Africa CROI. Montreal. 2009 [Google Scholar]
- 4.Zolopa A, Andersen J, Powderly W, Sanchez A, Sanne I, Suckow C, et al. Early antiretroviral therapy reduces AIDS progression/death in individuals with acute opportunistic infections: a multicenter randomized strategy trial. PLoS One. 2009;4(5):e5575. doi: 10.1371/journal.pone.0005575. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Moore RD, Chaisson RE. Natural history of opportunistic disease in an HIV-infected urban clinical cohort. Ann Intern Med. 1996 Apr 1;124(7):633–42. doi: 10.7326/0003-4819-124-7-199604010-00003. [DOI] [PubMed] [Google Scholar]