HIV-related mortality at a district hospital in Botswana

Abstract

We reviewed mortality data among medical inpatients at a tertiary hospital in Botswana to identify risk factors for adverse inpatient outcomes. This review was a prospective analysis of inpatient admissions. All medical admissions to male and female medical wards were recorded over a six-month period between 1 November 2011 and 30 April 2012. Data collected included patient demographics, HIV status (positive, negative, unknown), HIV testing history, HIV related treatment and serological history, admission and discharge diagnoses, and mortality status at final discharge or transfer. Of 972 patients admitted during the surveillance period, 427 (43.9%) were known to be HIV-positive on admission, 144 (14.8%) were known to be HIV-negative, and 401 (41.3%) had an unknown HIV status. Of those with unknown status, 131 (32.7%) were tested for HIV during admission and among these 35 (27.5%) were HIV-positive. Including patients with known mortality status following transfer, 172 (17.9%) patients died during the hospitalization. Death occurred in 105 (23%) of known HIV-positive patients, compared with 31 (13%) of known HIV-negative patients (p = 0.002, HR = 1.56 in adjusted analyses). Among HIV-positive patients who died, a low CD4 cell count (<200 cells/mm³) was associated with death. Overall, patients who died had significantly more neurological and respiratory-related presenting complaints than patients who survived. In conclusion, we identified higher overall mortality among HIV-positive patients at a tertiary hospital in Botswana, and low rates of in-hospital HIV testing and antiretroviral therapy initiation. These data demonstrate that despite available antiretroviral therapy in the population for over a decade, HIV continues to add excess burden to the hospital system and adds to inpatient mortality in Botswana.

Keywords

HIV mortality inpatient Botswana Africa

Introduction

Antiretroviral therapy (ART) is known to have improved survival over the past decade in Africa,¹ and in sub-Saharan Africa, the number of AIDS-related deaths fell by 39% between 2005 and 2013. Botswana recorded a 58% decline in AIDS-related deaths over this time. However, there remains limited data evaluating mortality by HIV status at inpatient facilities.

Recent studies from inpatient settings suggest morbidity and mortality remain high among HIV-infected individuals.^2–4 Outcomes may differ by the available level of care at each facility and by the rate of uptake of ART. Additional data are needed to identify preventable risk factors for mortality in the ART era that may be addressed by targeted interventions. Botswana has long been recognized for its leadership among African nations in HIV testing and provision of ART.⁵ By September 2012 a total of 198,553 Botswana living with HIV were receiving ART.⁶

Our primary objectives were to review mortality data among medical inpatients by HIV status and to identify risk factors for adverse inpatient outcomes. At the time of the review the guideline threshold for commencing ART was CD4 ≤ 250 µl or diagnosis of an opportunistic infection.

Methods

Data collection

We conducted a prospective review of inpatient admissions at Scottish Livingstone Hospital (SLH). All medical admissions adult medical wards were recorded over a six-month period between 1 November 2011 and 30 April 2012. Data were collected as part of a quality assurance program at the hospital and were presented at SLH in June 2012; approval was granted by the hospital and by Botswana Human Research and Development Committee to report the outcome of this surveillance.

SLH is a public district general hospital with 350 beds and serves a population of around 63,248 (estimate 2010). The hospital provides inpatient and outpatient care, including a large Infectious Disease Care Clinic that provides outpatient ART. Surgical interventions are mostly limited to obstetric surgery; there was no intensive care facility on site at the time of this surveillance. Available laboratory tests include hematology (full blood profile), biochemistry (renal profile, liver profile), microbiology (cerebrospinal fluid, sputum smear for acid fast bacilli, urine, and stool cultures). There are radiography facilities but no radiologist. Any higher level of management requires patient transfer to Princess Marina Hospital, Gaborone, which is 37 miles away. SLH is not located in a malaria-endemic area of Botswana.

A data abstraction form was devised, most likely diagnoses were ascertained from medical records and from written logbooks that were collated and kept by the nursing staff on the ward. Two members of the medical staff collected the data on a weekly basis. The initial data for each admission were collected from the medical records at the time of admission and upon discharge; if the patients who had short admissions were discharged prior to the data being captured, the charts were retrieved from medical records. Laboratory data were collected using the electronic health system, the Integrated Patient Management System. The data collected included patient demographics, HIV status (positive, negative, unknown), whether HIV testing had been carried out, and, if applicable, ART status and current viral and immunological markers, admission diagnosis, and discharge diagnosis. The final outcome was recorded as ‘discharged from hospital,’ ‘died in hospital,’ ‘transferred to another hospital or ward,’ or ‘absconded.’ For the purpose of the overall mortality analysis, transfers were analyzed as per their final discharge outcome from the location to which they were discharged, i.e. alive on final discharge or dead on final discharge. Absconded patients were included as ‘alive on final discharge.’ Admission duration was calculated inclusive of the day of admission and the day of discharge. Due to the wide variety of presenting complaints, admission diagnoses, and discharge diagnoses, the ten most common presenting complaints were considered. Admission and discharge diagnoses were compared to identify the consistency between these parameters. Sensitivity analyses were performed that included and excluded transferred patients from primary analyses.

Statistical methods

All data were entered into a Microsoft Access database, and analyses were performed using SAS 9.3 (SAS Institute, Cary, NC). The primary outcome analyzed was inpatient mortality. Risk factors for unknown sero-status and, separately, for being HIV-positive were identified by the two-sample T test for continuous variables or Fisher’s exact test for categorical variables. The 5% level was considered statistically significant. Logistic regression model was used to calculate odds ratios and associated 95% confidence intervals for both unadjusted and adjusted analyses. Potential risk factors were considered for inclusion in the final model if they were significant in univariate analysis (p < 0.05).

Results

Demographics

Over the six-month period, a total of 972 patients were admitted to the male and female medical wards. Of the 972 patients, 518 were women (53.3%) and 454 men (46.7%). The mean (standard deviation [SD]) age was 47.8 (SD 20.4) years. The mean duration of admission was 10.1 days (SD 13.2), and the median duration was seven days (interquartile range 4, 12). Table 1 summarizes the patient demographics.

Table 1.

Summary of demographics and outcome stratified by HIV status.

	HIV status
	Positive	(%)	Negative	(%)	Unknown	(%)	Total	(%)
Number on admission	427	–	144	–	401	–	972	–
Number on discharge	463	–	239	–	270	–	972	–
Age on admission (mean (SD))	41(12)	–	51(22)	–	55(24)	–		–
Gender
Male	216	46.7	125	52.3	113	41.9	454	46.7
Female	247	53.3	114	47.7	157	58.1	518	53.3
Outcome
Died	94	20.3	29	12.1	33	12.2	156	16.0
Discharged	321	69.3	183	76.6	222	82.2	726	74.7
Transferred	44	9.5	26	10.9	10	3.7	80	8.2
Absconded	4	0.9	1	0.4	1	0.4	6	0.6
Unknown	0	0	0	0	4	1.5	4	0.4
Outcome following transfer							Total final outcome
Died	11	25.0	2	7.7	3	30.0	172	17.7
Discharged	30	68.2	21	80.8	5	50.0	782	80.5
Absconded	0	–	0	–	0	–	6	–
Unknown	3	6.8	3	11.5	2	20.0	12	1.2

HIV status

Overall, 427 (43.9%) were known to be HIV-positive on admission, and 144 (14.8%) had a documented negative HIV test in the past; 77 (53.5%) of which were within the last year. There were 401 (41.3%) patients with unknown sero-status at the time of admission. Of these, 270 (67.3%) were not tested for HIV during the admission, and 131 (32.7%) were tested for HIV; among those tested during admission, 36 (27.5%) were HIV-positive, and 95 (72.5%) were HIV-negative. Although outside the direct scope of this paper, it was felt important to note that risk factors for having an unknown sero-status at the time of admission included older age (mean [SD] age was 55 [24] versus 43 [15] in those with unknown sero-status versus known sero-status, respectively, p < 0.001) and shorter duration of admission (mean [SD] duration of 9 [11] days versus 11 [14] days in those with unknown sero-status versus known, respectively, p-value = 0.001). There was no difference by gender for those with known and unknown status. Of those tested during the admission, risk factors for a positive test were younger age (mean [SD] of 41 [13] versus 54 [24] in HIV-positive versus HIV-negative, respectively, p-value < 0.001) and longer duration of admission (mean [SD] duration of 12 [14] versus 8 [12] days in HIV-positive versus HIV- or unknown, respectively, p-value <0.001).

Of the 427 known to be HIV-positive at admission, 240 (56.2%) were taking ART at the time of admission, and 332 (77.8%) had a known CD4 count, 151 (45.5%) had a CD4 count ≥200 cells/mm³, and the remaining 181 (54.5%) had a CD4 count <200 cells/mm³. Of these 181, 90 were receiving ART at the time of admission. HIV RNA (viral load) was known for 157 (36.8%) and of these, 125 (79.6%) had a viral load result of ≤400 copies/ml documented within the past year. At discharge a total of 463 (47.6%) patients were known to be HIV-positive, 239 (24.6%) were known to be HIV-negative, and 270 (27.8%) remained unknown.

Mortality

Among all admissions, 156 (16.1%) died before discharge or transfer, 726 (74.7%) were discharged to home alive, 80 (8.2%) were transferred to a tertiary care centre or another ward (e.g. surgical or tuberculosis specialist ward), four had unknown discharge status, and six patients absconded (considered as alive for final outcome). Of the 80 (8.2%) patients transferred, 58 (72.5%) were transferred to a tertiary care facility (Princess Marina Hospital in Gaborone) and the remaining 22 (27.5%) to a specialist ward within SLH. A final mortality status was available for 72 (90%) of these: 16 (22.2%) died after transfer, and 56 (77.8%) were discharged alive from the place of transfer. Thus, during the six months of surveillance, outcome/mortality status was known for 960 (99%) of 972 patients. Including the transfers with known mortality status, a total of 172 (17.9%) of patients died. HIV-positive status among transferred patients did not differ significantly from non-transferred patients (66% versus 63%, respectively, p = 0.60).

Among known HIV-positive individuals, 105 (23%) of 456 died, and being HIV-positive was significantly associated with a higher risk of death in univariate analyses when compared with either known HIV-negatives (p = 0.003) or with all other patients, including those with unknown HIV status (p = < 0.001). Table 1 shows univariate predictors of mortality among HIV-positive patients. Among HIV-positive patients who died, a low CD4 cell count (<200 cells/mm³) was associated with greater odds of death. Among HIV-positive patients, neither being on ART on admission nor having a viral load <400 was significantly protective (Table 2).

Table 2.

Univariate predictors of mortality among patients HIV-positive at discharge.

	Alive at final discharge (n = 355) n (%)	Dead at final discharge (n = 105) n (%)	OR (95% CI)	p-value
CD4 (cells/mm³)
<200	142 (52.3%)	53 (70.7%)
>200	135 (48.7%)	22 (29.3%)	0.44 (0.25,0.76)	0.003
ART status
No	172 (48.5%)	49 (46.7%)
Yes	183 (51.5%)	56 (53.3%)	1.07 (0.69,1.66)	0.75
Viral load
<400	101 (80.8%)	25 (78.1%)
>400	24 (19.2%)	7 (21.9%)	1.18 (0.46,3.04)	0.73

ART: antiretroviral therapy.

Table 3 shows univariate and multivariate predictors for mortality among patients with a known HIV status at outcome. Univariate analyses showed that patients who died were predominantly male and HIV-positive in models adjusted for both gender and HIV status at outcome together, both remain statistically significant.

Table 3.

Univariate and adjusted logistic regression models predicting probability of death; excluding patients with unknown HIV status at discharge or death.

	Univariate models			Adjusted model
	N	OR (95% CI)	p-value	N	OR (95% CI)	p-value
Older age	696	1.01 (0.997,1.02)	0.18	–	–	–
Male gender	696	1.54 (1.06,2.25)	0.03	696	1.60 (1.09,2.34)	0.02
Known HIV-positive	696	1.96 (1.27,3.03)	0.003	696	2.01 (1.30,3.12)	0.002

CI: confidence interval; OR: odds ratio.

Table 4 shows univariate and multivariate predictors for mortality among all patients (those with unknown HIV status are included in the category for HIV-negative patients). As before, univariate analyses showed that patients who were male and HIV-positive were at higher risk for death. When all patients were included, older age was also a significant predictor of death (though of low magnitude). In the adjusted model older age, being male and HIV-positive all remained statistically significant risk factors for death (Table 3). In both adjusted models presented in Tables 3 and 4, the risk of death for men remained higher, even after adjusting for HIV status.

Table 4.

Univariate and adjusted logistic regression models predicting probability of death; including all available patients (patients with unknown HIV status at discharge or death are included in HIV- category).

	Univariate models			Adjusted model
	N	OR (95% CI)	p-value	N	OR (95% CI)	p-value
Older age	960	1.01 (1.00,1.02)	0.005	960	1.02 (1.01,1.03)	<0.001
Male gender	960	1.51 (1.09,2.11)	0.01	960	1.42 (1.01,1.99)	0.04
Known HIV-positive	960	1.91 (1.37,2.68)	<0.001	960	2.72 (1.84,4.02)	<0.001

CI: confidence interval; OR: odds ratio. Notes: 1. n = 3 HIV-positive patients whose discharge survival status was unknown. 2. HIV-positive with known CD4 included in analysis: n = 352. 3. All n = 460 HIV-positive patients had known ART history and are included in analysis. 4. HIV-positive with known viral load included in analysis: n = 157. 5. P-value from logistic regression, modeling probability of dying. Reference group for each covariate indicated in the first column. 6. Older refers to age as a continuous variable.

Although not a primary aim of this study the data collected with regards to discharge diagnosis are presented below. Of the 156 patients who died prior to transfer, 90 (57.7%) had no known or recorded past medical history, 28 (17.9%) had a history of confirmed TB, 13 (8.3%) had known hypertension, 11 (7.1%) had hypertension and diabetes mellitus, and three (1.9%) had a known history of hypertension and congestive cardiac failure (with not all being mutually exclusive diagnoses). Patients who died had significantly more neurological and respiratory-related presenting complaints than patients who survived, whereas patients who survived had significantly more gastrointestinal, miscellaneous, and endocrine-related presenting complaints than patients who died (Table 5).

Table 5.

The ten most common discharge diagnoses by survival and HIV status at discharge.

	HIV-negative		HIV-positive		Unknown
	Discharged alive	Died prior to discharge	Discharged alive	Died prior to discharge	Discharged alive	Died prior to discharge
Tuberculosis	18	4	70	23	7	3
Diabetes	35	5	61	17	50	6
Pneumonia	19	9	50	26	18	9
Gastroenteritis	16	1	48	12	16	2
Severe anemia	4	1	15	5	4	2
Para suicide	12	0	7	1	27	1
Hypertension	20	0	5	1	34	3
CVE	5	0	8	1	17	6
CCF	11	3	8	2	12	7
CCM	0	2^a	16	6	7	1

CCF: congestive cardiac failure; CCM: cryptococcal meningitis; CVE: cerebrovascular event.

Presumptive diagnosis made prior to HIV test result.

Discussion

This is the first published documentation of mortality rates by HIV status from a medical inpatient ward in the ART era in Botswana. While the mortality rate overall was found to be 18%, the rate was significantly higher among HIV-positive individuals at 23%. Mortality was also significantly higher among men. Although 48% of admissions were among known HIV-positive patients, these individuals accounted for 61% of all inpatient deaths (and 77% of deaths among those with known HIV status). While access to HIV testing and care is free for Botswana citizens, we also found that 41.3% of adults admitted to the inpatient ward did not know their HIV status and that a new diagnosis of HIV was made on 26% of individuals tested during admission.

Few studies have reported on inpatient mortality by HIV status as ART has become available in resource-limited settings. Pre-ART era studies in Africa reported high overall mortality among medical inpatients, likely exacerbated by the HIV epidemic. In Zimbabwe, hospital admission mortality rates increased from 13.3 to 28.6% between 1992 and 2002.^6–8 In the ART era, mortality rates at hospitals in South Africa (2012) and Nigeria (2009) were 17 and 23.9%, respectively,^10,11 but neither considered HIV or ART specifically. Although overall mortality in this study was lower than some settings, the fact that at least 61% of deaths were among HIV-positive patients is concerning. Recent publications have found that overall mortality is falling in sub-Saharan African countries but deaths attributed to HIV continue to rise^12,13 albeit at a very slow rate.

HIV-positive individuals remain disproportionately represented on the medical wards compared with the general population, where the HIV prevalence among 15–49-year-olds is 21.9%⁵ in Botswana. These data support that despite available ART in the population, HIV continues to add excess burden to the hospital system. In addition, the very high mortality observed among HIV patients suggests that ART availability alone has not reduced HIV-related mortality to baseline levels expected in the population. Our data also show that the burden of illness in HIV-positive patients was not all due to opportunistic infections. Comorbidities from non-communicable diseases such as hypertension and diabetes seen in the aging HIV-positive populations were apparent. Among HIV-positive patients, our study identified a protective effect of higher CD4 cell count on mortality. However, we did not demonstrate a significant survival benefit from being on ART or having a reported suppressed viral load at admission, and it remains unknown the extent to which incomplete ART access, late ART access, or ART failure may contribute to mortality risk. Acknowledging that this is representative of inpatients only, ART may have prevented admission.

A concerning number of inpatients at SLH with unknown HIV status were not tested during admission. Patients who were not critically ill may have opted to prefer to test at a local clinic or simply have declined testing and at time of this study the age for consent was 21 years old. Lack of documentation for no test being completed was also identified as a concern.

Although some untested patients may have been deemed at low risk because of their age, or not tested because of a non-infectious presenting illness, at least 167 of those untested were aged between 18 and 50 years and presented with an infection. In addition, high HIV positivity has been reported among older age cohorts in Africa and testing based on perceived HIV risk may underestimate infections.¹⁵ Opt out testing strategies have been shown to be effective, admission ‘opt-out’ testing protocols could be beneficial.¹⁶ Implementation could lead to a dramatic reduction in unknown HIV status and could enhance inpatient care. Such a policy might reduce barriers to testing among younger and older age groups.

We identified low rates of inpatient ART initiation. While certain clinical situations might limit the feasibility of this majority of are likely to benefit from a rapid initiation of ART. Early ART may result in improved clinical outcomes,¹⁷ particularly in settings where there may be a delay between discharge and follow up in the outpatient setting. Most studies to date report the time to treatment among those with AIDS-defining illnesses¹⁸ but few compare starting ART as an inpatient versus outpatient in a sub-Saharan African setting. Further study is needed to evaluate this, where a variety of programs and providers may be responsible for ART delivery and loss to follow-up is of great concern.¹⁹ A possible barrier to initiating treatment in the inpatient setting were guidelines that encouraged involvement of an adherence partner at ART initiation. While having an adherence partner and disclosure of HIV status to another person has been found to improve overall ART adherence, this practice must be balanced against the mortality risk associated with failure to initiate ART promptly.

A lack of diagnostics and the nature of this study make it difficult to draw many conclusions regarding diagnoses. Neurological and respiratory illness were statistically more likely to result in a patient dying, this maybe related to AIDS-defining conditions, including but not limited to such pathologies of TB, Pneumocystis jeruvici pneumonia, and cryptococcal meningitis.

Limitations of this study included the use of chart extraction rather than direct questioning of patients. CD4 and viral load values therefore may not have actively reflected the actual count at the exact time of admission. The number of patients with unknown HIV status may be overestimated if some had tested in community voluntary counseling and testing, highlighting the need for accurate documentation. The lack of known HIV status for 270 (28%) may have led to lack of precision in our HIV-specific mortality rate (with potential bias toward a lower HIV-specific mortality rate when including those with unknown status in the denominator). We detected no evidence of bias in the patients transferred to the tertiary facility (e.g. by HIV status) but cannot exclude that possibility. We performed sensitivity analyses excluding transfers, and results did not change significantly in these analyses (data not shown). Transfer rates at SLH were relatively high because no intensive care ward existed at the facility at the time of the study; it is unknown how the lack of intensive level care, or the ability to transfer to a tertiary facility an hour away in the capital city Gaborone, affected overall mortality.

In summary, this study identified a disproportionately high mortality rate among HIV-positive medical inpatients at a district hospital in Botswana. Outcomes may be improved by targeted treatment and/or early use of ART. As in other studies, lower CD4 cell count was associated with increased mortality, and we believe that maximizing HIV testing and ART use among eligible persons both before and after hospitalization could play a role in mortality reduction (although further data are required to determine whether inpatient ART initiation has a direct mortality benefit in a resource-limited setting). Prioritizing HIV testing and ART program uptake among inpatients will also improve ART coverage among individuals facing the highest mortality risk in the population.

Footnotes

Acknowledgements

We would like to acknowledge all the staff and patients of Scottish Livingstone Hospital, Molepolole, Botswana.

Authors’ contributions

Study design (MEOP, LFW, JM, RLS), study implementation (MEOP, LFW, JM, KK, KMP, RLS), data management (MEOP, LFW), data analysis (MEOP, KB, RLS), manuscript preparation (MEOP, LFW, JM, KK, KMP, KB, MM, JM, RLS).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: KMP receives salary support from the National Institute of Child Health and Human Development (1K23HD070774-01A1).

References

Fatti

Meintjes

Shea

. Improved survival and antiretroviral treatment outcomes in adults receiving community-based adherence support: 5-year results from a multicentre cohort study in South Africa. J Acquir Immune Defic Syndr 2012; 61: e50–58–e50–58.

Balogou

Saka

Kombaté

. Causes of mortality associated with HIV/AIDS in health-care facilities in Togo: a six-month prospective study. Trop Doct 2011; 41: 215–217.

Lewden

Drabo

Zannou

. Disease patterns and causes of death of hospitalized HIV positive adults in West Africa: a multicountry survey in the antiretroviral treatment era. J Int AIDS Soc 2014; 17: 18797–18797.

Kra

Aba

Yao

. Clinical, biological, therapeutic and evolving profile of patients with HIV infection hospitalized at Infectious and tropical diseases unit in Abidjan (Ivory Coast). Bull Soc Pathol Exot 2013; 106: 37–42.

http://www.unaids.org/en/regionscountries/countries/botswana/ (accessed 2 May 2016).

Van der Plas

Meintjes

Schutz

. Complications of antiretroviral therapy initiation in hospitalised patients with HIV-associated tuberculosis. PLoS One 2013; 8: e54145–e54145.

Bardgett

Dixon

Beeching

. Increase in hospital mortality from non-communicable disease and HIV-related conditions in Bulawayo, Zimbabwe, between 1992 and 2000. Trop Doct 2006; 36: 129–129.

Pallangyo

Håkanson

Lema

. High HIV seroprevalence and increased HIV-associated mortality among hospitalized patients with deep bacterial infections in Dar es Salaam, Tanzania. AIDS 1992; 6: 971–976.

Ole-Nguyaine

Crump

Kibiki

. HIV-associated morbidity, mortality and diagnostic testing opportunities among inpatients at a referral hospital in northern Tanzania. Ann Trop Med Parasitol 2004; 98: 171–179.

10.

Myer

Smith

Mayosi

. Medical inpatient mortality at Groote Schuur Hospital, Cape Town, South Africa, 2002–2009. S Afr Med J 2012; 103: 28–31.

11.

Chijioke

Kolo

. Mortality pattern at the adult medical wards of a teaching hospital in sub-Saharan Africa. Int J Trop Med 2009; 4: 27–31.

12.

http://www.health-e.org.za/2014/03/21/report-statssa-mortality-causes-death-south-africa-2011/, update http://www.southafrica.info/about/health/mortality-190314.htm#.Vcj-1VKjs_i (accessed 2 May 2016).

13.

14.

Moses

Zimba

Kamanga

. Prevention of mother-to-child transmission: program xchanges and the effect on uptake of the HIVNET 012 regimen in Malawi. AIDS 2008; 22: 83–87–83–87.

15.

Negin

Cumming

. HIV infection in older adults in sub-Saharan Africa: extrapolating prevalence from existing data. Bull WHO 2010; 88: 847–853.

16.

Creek

Ntumy

Seipone

. Successful introduction of routine opt-out HIV testing in antenatal care in Botswana. JAIDS 2007; 45: 102–107.

17.

Grant

Zolopa

. When to start ART in the setting of acute AIDS-related opportunistic infections: the time is now!. Curr HIV/AIDS Rep 2012; 9: 251–258.

18.

Couzigou

Semaille

Le Strat

. Differential improvement in survival among patients with AIDS after the introduction of HAART. AIDS Care 2007; 19: 523–531.

19.

Geng

Bwana

Muyindike

. Failure to initiate antiretroviral therapy, loss to follow-up and mortality among HIV-infected patients during the pre-ART period in Uganda. J Acquir Immune Defic Syndr 2013; 63: e64–71.