Abstract
Objective
To assess differences in access to antiretroviral treatment (ART) and patient outcomes across public-sector treatment facilities in the Free State Province, South Africa.
Design
Prospective cohort study with retrospective database linkage. We analysed data on patients enrolled in the treatment programme across 36 facilities between May 2004 and December 2007. We assessed percentage initiating ART and percentage dead, at one year after enrolment. Multivariable logistic regression was used to estimate associations of clinic-level and patient-level characteristics with both mortality and treatment status.
Results
Of 44,866 patients enrolled, 15,219 initiated treatment within one year. 8,778 died within one year, of whom 7,286 died before accessing ART. Outcomes at one year varied greatly across facilities and more variability was explained by clinic-level factors than by patient-level factors. The odds of starting treatment within one year improved over calendar time. Patients enrolled in facilities with treatment initiation available on site had higher odds of starting treatment and lower odds of death at one year compared to those enrolled in facilities that do not offer treatment initiation. Patients were less likely to start treatment if they were male, severely immunosuppressed (CD4 count <50 cells/ mm3), or underweight (< 50kg). Men were also more likely to die in the first year after enrolment.
Conclusions
Although increasing numbers of patients started ART between 2004 and 2007, many patients died before accessing ART. Patient outcomes could be improved by decentralisation of treatment services, fast-tracking the most immunodeficient patients and improving access especially for men.
Introduction
In South Africa, HIV/AIDS is the top single cause of mortality (1–4) and the Free State has the third highest HIV prevalence of the nine provinces, estimated to be 12.6% in 2008 (5). There are few data on outcomes of patients from first contact with ART programmes as most studies only report mortality from initiation of ART and ignore pre-treatment deaths (6). However, one study reported that 87% of deaths occurred in patients not on treatment (7). The Comprehensive HIV and AIDS Management (CHAM) Programme in the Free State started in May 2004 as part of the scale up of access to antiretroviral treatment (ART) in resource-limited settings (8).
Information on factors that influence whether an enrolled patient starts ART or dies before treatment, and the time spent waiting for ART, should inform policy aimed at reducing overall programme mortality. Although treatment programmes aim to start as many eligible patients on treatment as quickly as possible (9), factors such as shortage of staff and facilities, time taken for adherence training and diagnosis and treatment of opportunistic infections can contribute to delays in initiating ART (10–12).
In this paper, we report on patient outcomes at one year after enrolment in the Free State public treatment programme to assess management of patients through the system. Outcomes include the proportion of patients that start treatment and deaths before and on treatment. We examined differences in outcomes according to both facility and patient level characteristics.
Methods
Setting and patients
All patients enrolled in the Free State CHAM programme between May 2004 and December 2007 were included in the analysis if a CD4 value was available. After testing HIV-positive in any Free State clinic, patients were referred to nurse-run assessment facilities for initial assessment of eligibility for treatment which includes a CD4 cell count measurement. According to clinical protocol (13) patients with CD4 cell counts > 200 cells/mm3 were asked to return for another assessment after six to twelve months, or before if they experienced problems. Patients with CD4 cell count of ≤ 200 cells/mm3, or with WHO disease stage IV (AIDS) were referred for treatment. Patients who start ART obtain monthly supplies of medication from assessment facilities and return to treatment facilities for review of their viral loads and repeat prescriptions six monthly. By the end of 2007, 30 clinics and eight hospitals distributed over the five health districts of the province were participating in the programme (Figure 1). Two clinics were excluded from this analysis because of delays in implementation of the electronic data collection system. Of the remaining 28 clinics, 17 were assessment-only facilities and 11 were combined assessment and treatment facilities. Seven of the hospitals provide treatment for patients from assessment facilities, the eighth hospital functions as a specialist referral facility for the province.
Figure 1.
Map of the Free State Province ART facilities
Data Collection
Routine clinical information on patients from clinician notes on standardised forms was entered into the province’s electronic medical record (EMR) system by trained data capturers. Data were downloaded and imported into a data warehouse weekly. During this time period, Meditech SA™ was gradually implemented as the EMR system across the facilities (14). Each patient has a unique identifier within Meditech so that patients can be tracked across facilities within the province. Mortality is ascertained by monthly linkage with the national death register using national identity numbers. The death register is estimated to capture more than 90% of adult deaths in South Africa (15). Patient demographic and clinical characteristics extracted from Meditech included sex, age, weight and CD4 count nearest to enrolment and year of enrolment into the programme. The Meditech database was linked with the National Health Laboratory Services (NHLS) database to obtain CD4 measurements to ensure data were as complete as possible. The Free State Department of Health gave permission for the data to be analysed for this study and the Human Research Ethics Committee of the University of Cape Town approved the protocol. No individual patient consent was deemed necessary as data were collected routinely for the CHAM programme, and no patient identifiers were included in the extracts used for analyses.
Information was obtained from programme staff on the type of facility (assessment only clinic, treatment hospital or combined assessment and treatment clinic), location (rural/peri-urban/urban), health district and distance between the facility and the point of care for initiation of ART (in kilometres). Data on staffing levels were obtained from the Free State Province Human Resources database and included patient load, number of staff per 1000 patients and number of staff vacancies per 1000 patients. Staff per patient was calculated as the mean number of staff in each clinic over the period April 2005 to September 2008 divided by total number of patients enrolled in that clinic.
Outcomes
Patients were followed to December 31, 2008 and therefore all patients potentially had at least one year of follow up time. The outcomes of interest were whether the patient had initiated treatment or died in the first year after enrolment.
Statistical analysis
We tabulated the characteristics of each facility and examined completeness of recording and variation in patient characteristics across facilities. We estimated median waiting times for treatment and mortality and assessed outcomes at one year after enrolment, by district, facility location, distance to ART initiation site and year of enrolment.
We used a mixed-effects logistic regression with random intercepts for facilities to estimate the odds of the patient starting treatment or dying within one year of enrolment, allowing for between-facility variability. Patient level characteristics included in the model were: sex, age, weight, CD4 count and year of enrolment. Year of enrolment was categorised as 2004/05, 2006 and 2007, as 2004 and 2005 were similar in terms of patient outcomes. Facility level characteristics included in the model were: staffing per 1000 patients, location and distance to ART initiation site. Location was categorised as rural or urban/peri-urban and included in the model as a binary variable. We fitted a sequence of four models to the data to assess the variation explained by facility and patient level characteristics: (i) the null model with no covariates, (ii) including patient level characteristics only, (iii) including facility level characteristics only, (iv) including both facility and patient level characteristics. We examined the reduction in the standard deviation of the random effects distribution in each of these models.
Weight, age and CD4 count had missing values, which were imputed using the multivariate imputation by chained equations method in Stata™ assuming data were missing at random (16;17). The imputation model consisted of all variables in the analysis model, including outcomes of interest, time to outcomes and other recorded values of pre-treatment weight and CD4. Distributions of imputed data were comparable with observed data. Data were more likely to be missing in urban facilities and facilities where treatment was available on the same site. Given that we identified predictors of missingness, it is not plausible that data were missing completely at random (MCAR). It is not possible to distinguish whether data were missing at random (MAR) or missing not at random (MNAR), but the fact that predictors of missingness tended to be clinic rather than patient characteristics makes the MAR assumption underlying the multiple imputation analyses more plausible. We analysed 25 datasets in which missing values were replaced by imputed values. Point estimates and standard errors were calculated using Rubin’s rules (18). Analyses using imputed data are presented in the main paper and analyses restricted to complete case data are presented in Appendix 2 (19). All analyses were performed using Stata statistical software (20).
Appendix 2.
Adjusted associations of facility and patient level characteristics with patient outcomes at 1-year post enrolment: complete case analysis (N=24510)
| Outcome at 1 year | Patient started treatment | Patient died | |||
|---|---|---|---|---|---|
| N | OR(95% CI) | P>|z| | OR(95% CI) | P>|z| | |
| Facility level characteristics | |||||
| Average staff (per 1000 patients) | |||||
| <1.5 | 8068 | 0.41(0.17,0.97) | 0.044 | 0.84(0.55,1.29) | 0.429 |
| 1.5–2.5 | 9211 | 1 (baseline) | 1 (baseline) | ||
| >2.5 | 7231 | 1.24(0.57,2.71) | 0.589 | 1.04(0.69,1.56) | 0.853 |
| Location | |||||
| Rural | 6245 | 0.57(0.30,1.08) | 0.084 | 1.30(0.90,1.86) | 0.158 |
| Urban/peri-urban | 18265 | 1 (baseline) | 1 (baseline) | ||
| Distance to treatment initiation site | 0.001* | 0.007* | |||
| Same site | 5513 | 1 (baseline) | 1 (baseline) | ||
| <=8km | 7660 | 0.54(0.25,1.16) | 0.113 | 1.63(1.07,2.49) | 0.023 |
| 9km–15km | 7803 | 0.42(0.15,1.14) | 0.088 | 2.10(1.26,3.52) | 0.005 |
| >15km | 3534 | 0.35(0.18,0.71) | 0.004 | 1.50(1.03,2.18) | 0.035 |
| Patient level characteristics | |||||
| Sex | |||||
| Male | 7869 | 0.70(0.66,0.75) | <0.001 | 1.36(1.26,1.47) | <0.001 |
| Female | 16641 | 1 (baseline) | 1 (baseline) | ||
| Age (years) | |||||
| 15–29 | 6789 | 0.88(0.81,0.95) | 0.001 | 0.75(0.68,0.83) | <0.001 |
| 30–39 | 10045 | 1 (baseline) | 1 (baseline) | ||
| 40–49 | 5706 | 1.05(0.97,1.14) | 0.236 | 1.11(1.01,1.21) | 0.023 |
| >=50 | 1970 | 1.02(0.91,1.15) | 0.718 | 1.42(1.25,1.61) | <0.001 |
| Weight (kg.) | |||||
| <40 | 1116 | 0.58(0.50,0.67) | <0.001 | 3.18(2.75,3.68) | <0.001 |
| 40–49 | 5484 | 0.80(0.74,0.87) | <0.001 | 1.61(1.47,1.76) | <0.001 |
| 50–59 | 8533 | 1 (baseline) | 1 (baseline) | ||
| 60–79 | 7737 | 1.11(1.02,1.20) | 0.010 | 0.68(0.61,0.74) | <0.001 |
| >=80 | 1640 | 1.11(0.96,1.28) | 0.173 | 0.53(0.43,0.66) | <0.001 |
| First pre-treatment CD4 count(cells/µL) | |||||
| <=25 | 2279 | 0.57(0.51,0.63) | <0.001 | 4.29(3.84,4.80) | <0.001 |
| 25–50 | 1913 | 0.71(0.64,0.80) | <0.001 | 2.34(2.07,2.64) | <0.001 |
| 50–100 | 3451 | 0.88(0.81,0.97) | 0.007 | 1.75(1.58,1.94) | <0.001 |
| 100–200 | 6255 | 1 (baseline) | 1 (baseline) | ||
| 200–350 | 5356 | 0.15(0.14,0.16) | <0.001 | 0.52(0.46,0.59) | <0.001 |
| >350 | 5256 | 0.02(0.02,0.03) | <0.001 | 0.26(0.22,0.30) | <0.001 |
| Year of enrolment | |||||
| 2004/5 | 12026 | 1 (baseline) | 1 (baseline) | ||
| 2006 | 7131 | 1.20(1.11,1.30) | <0.001 | 0.95(0.87,1.04) | 0.250 |
| 2007 | 5353 | 1.76(1.61,1.92) | <0.001 | 0.85(0.76,0.94) | 0.002 |
p for trend
OR: Odds Ratio from logistic regression analyses adjusted for all variables in the table and including facility as a random effect.
Results
Table 1 shows the characteristics of each health district in the programme, and characteristics of each facility are given in Appendix 1. The Motheo district includes mostly urban facilities and had the largest number of patients enrolled, while Xhariep had only rural facilities and had the smallest number of patients. Figure 1 shows the location of each district within the province, together with the major towns in each district and locations of facilities.
Table 1.
Health district characteristics
| District | Number of patients enrolled by end 2007 |
Large towns/ cities |
Facilities | Median (range) of staff per 1000 patients enrolled |
Median (range) of vacancies per 1000 patients enrolled |
Median CD4 at enrolment and range over facilities |
|---|---|---|---|---|---|---|
| Fezile Dabi | 5644 | Sasolburg | 2 treatment facilities (both urban) | 10.89 (1.73–23.70) | 5.00 (0.44–22.59) | 154 (98.5–232) |
| Kroonstad | 5 assessment facilities (2 peri-urban, 3 rural) | |||||
| 1 combined facility (rural) | ||||||
| Lejweleputswa | 8112 | Welkom | 2 treatment facilities (1 urban, 1 rural) | 2.04 (1.60–27.04) | 0.51 (0.41–13.64) | 166 (115–192) |
| 3 assessment facilities (2 urban, 1 rural) | ||||||
| Motheo | 15477 | Bloemfontein | 1 specialist referral facility (urban) | 6.59 (1.28–29.27) | 1.46 (0.16–16.76) | 188 (99–223) |
| Botshabelo | 2 treatment facilities (both urban) | |||||
| Thaba Nchu | 4 assessment facilities (all urban) | |||||
| 1 combined facility (urban) | ||||||
| Thabo Mofutsanyane | 12108 | Phuthaditjhaba | 2 treatment facilities (both urban) | 6.37 (1.25–48.68) | 1.95 (0.09–18.86) | 151 (100–176) |
| Harrismith | 4 assessment facilities (all urban) | |||||
| Bethlehem | 2 combined facilities (both rural) | |||||
| Xhariep | 3525 | Predominantly | 6 rural combined facilities | 8.27 (4.16–22.56) | 9.96 (1.67–38.75) | 211 (164.5–254.5) |
| rural area with small towns | 1 rural assessment facility | |||||
Assessment facility – Nurse-run facility providing assessment of eligibility for ART, preparation for treatment and monthly collection of medication.
Treatment facility – Doctor-run facility providing initiation of treatment, 6 monthly review of viral load and CD4 once on treatment and repeat prescriptions.
Combined facility - Nurse-run assessment facility with visiting doctor support for ART initiation and repeat prescriptions.
Appendix 1.
Facility characteristics
| Clinic | N | Location* | Distance to treatment site (km) |
Hospital or clinic^ |
Staff per 1000 patients |
Vacancies per 1000 patients |
Median CD4 at enrolment |
|---|---|---|---|---|---|---|---|
| District: Fezile Dabi | |||||||
| Boitumelo Hospital | 531 | U | 0 | H | 12.17 | 8.69 | 120 |
| Lesedi CHC | 781 | P | 4 | C | 5.32 | 1.08 | 146 |
| Metsimaholo Clinic | 342 | R | 70 | C | 9.61 | 5.64 | 173 |
| Metsimaholo Hospital | 568 | U | 0 | H | 16.22 | 4.40 | 98.5 |
| PAX CHC | 185 | R | 0 | C | 23.70 | 22.59 | 180 |
| Refengkgotso Clinic | 656 | R | 40 | C | 4.46 | 3.34 | 226 |
| Steynsrus Clinic | 217 | R | 70 | C | 17.69 | 5.60 | 232 |
| Zamdela CHC | 2364 | P | 8 | C | 1.73 | 0.44 | 146 |
| District: Lejweleputswa | |||||||
| Bongani Hospital | 448 | U | 0 | H | 23.97 | 11.43 | 115 |
| Matjhabeng Clinic | 2693 | P | 7 | C | 1.80 | 0.41 | 167 |
| Mohau Hospital | 328 | R | 0 | H | 27.04 | 13.64 | 118 |
| Phomolong Clinic | 2035 | R | 30 | C | 2.04 | 0.46 | 192 |
| Welkom Clinic | 2608 | P | 15 | C | 1.60 | 0.51 | 165 |
| District: Motheo | |||||||
| Batho Clinic | 3383 | P | 8 | C | 1.38 | 0.16 | 211 |
| Botshabelo Hospital | 267 | P | 0 | H | 29.27 | 16.76 | 123 |
| Heidedal CHC | 4722 | P | 0 | C | 1.34 | 0.24 | 223 |
| Jazzman Clinic | 394 | P | 8 | C | 10.82 | 1.87 | 121.5 |
| MUCPP CHC | 4593 | P | 12 | C | 1.28 | 0.20 | 217 |
| National Hospital | 794 | U | 0 | H | 11.57 | 5.31 | 99 |
| Pelonomi Hospital | 277 | P | 0 | C | 9.46 | 10.74 | 173 |
| Winnie Mandela Clinic | 1047 | P | 5 | C | 3.72 | 1.06 | 129 |
| District: Thabo Mofutsanyane | |||||||
| Bohlokong Clinic | 1991 | P | 4 | C | 1.94 | 0.58 | 143 |
| Bophelong Clinic | 339 | R | 0 | C | 13.82 | 17.08 | 176 |
| Mamello CHC | 392 | R | 0 | C | 12.29 | 16.65 | 114.5 |
| Manapo Hospital | 873 | U | 0 | H | 10.15 | 3.33 | 104 |
| Namahadi Clinic | 3774 | P | 15 | C | 1.25 | 0.09 | 157 |
| PhekolongHospital | 180 | U | 0 | H | 48.68 | 18.86 | 100 |
| Tseki Clinic | 2583 | P | 12 | C | 1.98 | 0.12 | 165 |
| Tsiame B Clinic | 1976 | R | 25 | C | 2.60 | 0.16 | 163 |
| District: Xhariep | |||||||
| BophelongP’burg CHC | 682 | R | 0 | C | 8.27 | 3.18 | 254.5 |
| Ethembeni Clinic | 905 | R | 0 | C | 4.16 | 4.55 | 209 |
| Itumeleng Clinic | 703 | R | 0 | C | 7.04 | 9.96 | 181 |
| Lephoi Clinic | 236 | R | 0 | C | 15.61 | 22.64 | 238.5 |
| Matlakeng Clinic | 182 | R | 0 | C | 22.56 | 38.75 | 164.5 |
| Nelson Mandela Clinic | 470 | R | 80 | C | 7.09 | 1.67 | 184 |
| Thembalethu Clinic | 347 | R | 0 | C | 12.89 | 18.81 | 214.5 |
| Total | 44866 | 4.27 | 2.28 | 170 | |||
R=rural, P=peri-urban, U=urban
H=Hospital, C=Clinic
Table 2 shows the variability in completeness of data according to district, location and distance from treatment site. Completeness of weight records was the most variable, ranging from 16.2% to 94.1% recorded; these records were most complete in urban facilities. Rural facilities had the most complete recording of CD4 counts and peri-urban facilities had the most complete ID recording. Facilities with ART initiation on site had the least complete CD4 records.
Table 2.
Completeness of recording of data in the Meditech database of the Free State ART Programme by a) district b) facility type and c) location
| % of patients in each facility with the variable recorded | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Variable of interest | CD4 at enrolment | Weight at enrolment | South African national identity number | |||||||||
| N | Median | IQR* | Range | Median | IQR* | Range | Median | IQR* | Range | |||
| Overall | 44866 | 72.1 | (62.5,87.5) | (36.7,98.2) | 71.1 | (58.4,81.8) | (16.2,94.1) | 81.1 | (69.6,87.9) | (39.5,94.2) | ||
| District (number of facilities) | ||||||||||||
| Fezile Dabi (8) | 5644 | 69.2 | (53.2,85.8) | (37.1,98.2) | 66.4 | (59.4,73.7) | (49.5,88.9) | 86.8 | (78.0,89.7) | (67.7,92.4) | ||
| Lejweleputswa (5) | 8112 | 87.5 | (67.4,87.8) | (65.0,88.0) | 77.0 | (76.3,81.4) | (72.9,89.0) | 76.8 | (69.8,87.7) | (69.3,90.3) | ||
| Motheo (8) | 15477 | 70.5 | (54.7,74.5) | (48.7,79.1) | 42.0 | (23.5,78.0) | (16.2,85.9) | 82.8 | (65.1,91.1) | (39.5,92.2) | ||
| Thabo Mofutsanyane (8) | 12108 | 66.9 | (57.7,80.1) | (46.1,90.0) | 61.5 | (57.8,79.1) | (40.7,85.3) | 81.9 | (70.8,84.7) | (39.8,94.2) | ||
| Xhariep (7) | 3525 | 86.5 | (80.7,92.8) | (75.8,97.7) | 82.2 | (68.1,90.2) | (58.6,94.1) | 72.3 | (50.0,84.3) | (43.8,85.5) | ||
| Location | ||||||||||||
| Urban | 3394 | 55.0 | (46.1,65.0) | (37.1,71.4) | 79.6 | (76.3,85.3) | (58.2,85.9) | 81.9 | (76.8,88.6) | (76.3,92.2) | ||
| Peri-urban | 31477 | 71.7 | (67.0,79.1) | (48.7,90.0) | 59.6 | (31.3,77.0) | (16.2,89.0) | 86.2 | (79.0,90.5) | (39.5,94.2) | ||
| Rural | 9995 | 85.0 | (71.6,89.3) | (53.1,98.2) | 70.6 | (60.5,84.9) | (40.7,94.1) | 71.0 | (66.4,84.5) | (39.8,92.4) | ||
| Distance from treatment site | ||||||||||||
| Same site | 12959 | 65.0 | (51.3,82.2) | (37.1,97.7) | 76.3 | (58.2,85.2) | (23.2,94.1) | 76.8 | (66.3,84.7) | (39.5,92.4) | ||
| <=8km | 12653 | 76.9 | (69.5,87.8) | (67.0,90.0) | 69.3 | (23.6,78.5) | (16.2,89.0) | 88.2 | (79.0,90.5) | (75.2,90.9) | ||
| 8–15km | 13558 | 72.1 | (71.6,79.9) | (71.2,87.5) | 69.7 | (59.6,77.2) | (56.7,77.5) | 89.7 | (86.2,93.0) | (84.7,94.2) | ||
| >15km | 5696 | 88.0 | (83.5,92.8) | (62.7,98.2) | 70.6 | (64.6,82.1) | (58.8,88.9) | 68.5 | (66.4,79.7) | (43.8,85.4) | ||
IQR=interquartile range
44,866 treatment-naïve patients aged 15 years or over were enrolled across the 36 facilities between 2004 and 2007. Median CD4 cell count at enrolment was 170 cells/mm3 and ranged across facilities from 98 to 255 cells/mm3. 67% of patients were female, with a range of 54–76% across facilities. Mean age at enrolment was 36 years (range 33–38 years) and mean weight 57.6kg (range 54.6–59.9kg). Data were missing for age in 22 (0.05%) patients, for weight in 15,550 (34.66%) patients and for CD4 count in 11,650 (25.97%) patients.
Table 3 shows patient outcomes by district, distance to ART initiation site, location and year of enrolment. Xhariep had the lowest pre-ART mortality (13.6%) and the highest median CD4 cell count at enrolment (211 cells/mm3). In contrast, Thabo Mofutsanyane had the highest pre-ART mortality (19.2%) and the lowest median CD4 count (151 cells/mm3). Fezile Dabi had the highest percentage initiated on ART (48.4%), and for those who initiated, had the shortest median waiting time for treatment (1.65 months). Despite the presence of major urban centres, Motheo had the longest waiting times for treatment (3.70 months). Over all clinics, percentage initiated on ART ranged from 7.9% to 82.2%, and mortality ranged from 8.1% to 32.8%. For those who were treated, median waiting times ranged from 1 day (in a hospital setting) to 5.4 months. 83% of all deaths occur in patients not receiving ART.
Table 3.
Patient outcomes at 1 year after enrolment by a) district b) facility type c) location and d) year of enrolment
| N | Median CD4 |
N (%) dead | N (%) dead before initiating treatment |
N (%) treated | Median waiting time to treatment (months) |
Median time to death for those who died (months) |
|
|---|---|---|---|---|---|---|---|
| Overall | 44866 | 170 | 8778 (19.6) | 7286 (16.2) | 15219 (33.9) | 2.7 | 3.0 |
| Patients who were eligible at enrolment | 19120 | 90 | 5130 (26.8) | 4047 (21.2) | 10194 (53.3) | 2.98 | 3.34 |
| Patients who were not eligible at enrolment | 14096 | 350 | 878 (6.2) | 799 (5.7) | 1704 (12.1) | 3.54 | 5.55 |
| Patients with unrecorded CD4 at enrolment | 11650 | NA | 2770 (23.8) | 2440 (20.9) | 3321 (28.5) | 1.49 | 1.79 |
| District | |||||||
| Fezile Dabi | 5644 | 154 | 1072 (19.0) | 829 (14.7) | 2731 (48.4) | 1.65 | 2.74 |
| Lejweleputswa | 8112 | 166 | 1642 (20.2) | 1287 (15.9) | 3210 (39.6) | 2.74 | 3.27 |
| Motheo | 15477 | 188 | 2648 (17.1) | 2368 (15.3) | 4084 (26.4) | 3.70 | 3.17 |
| Thabo Mofutsanyane | 12108 | 151 | 2813 (23.2) | 2322 (19.2) | 4191 (34.6) | 2.38 | 2.94 |
| Xhariep | 3525 | 211 | 603 (17.1) | 480 (13.6) | 1003 (28.5) | 2.84 | 3.17 |
| Distance from treatment site | |||||||
| Same site | 12959 | 167 | 1931 (14.9) | 1445 (11.2) | 5364 (41.4) | 2.05 | 3.04 |
| <=8km | 12653 | 160 | 2675 (21.1) | 2136 (16.9) | 5005 (39.6) | 2.58 | 2.98 |
| 8–15km | 13558 | 176 | 3107 (22.9) | 2794 (20.6) | 3382 (24.9) | 3.90 | 3.31 |
| >15km | 5696 | 186 | 1065 (18.7) | 911 (16.0) | 1468 (25.8) | 2.61 | 2.61 |
| Location | |||||||
| Urban | 3394 | 105 | 714 (21.0) | 474 (14.0) | 2316 (68.2) | 1.42 | 3.04 |
| Peri-urban | 31477 | 173 | 6338 (20.1) | 5431 (17.3) | 9702 (30.8) | 3.07 | 3.17 |
| Rural | 9995 | 188 | 1726 (17.3) | 1381 (13.8) | 3201 (32.0) | 2.28 | 2.74 |
| Year of enrolment | |||||||
| 2004 | 6567 | 168 | 1260 (19.2) | 1117 (17.0) | 1748 (26.6) | 4.50 | 4.23 |
| 2005 | 10872 | 184 | 2273 (20.9) | 2005 (18.4) | 2708 (24.9) | 3.83 | 3.47 |
| 2006 | 12877 | 185 | 2485 (19.3) | 2072 (16.1) | 4259 (33.1) | 2.68 | 2.91 |
| 2007 | 14550 | 147 | 2760 (19.0) | 2092 (14.4) | 6504 (44.7) | 1.95 | 2.31 |
Compared with assessment-only facilities, facilities where ART was available on site performed well in terms of proportion of patients who started treatment, except for those in Xhariep district where there was no doctor for several months during the observation period. Xhariep contains only rural clinics and all but one of these clinics were combined assessment and treatment facilities served by doctors on a sessional basis. Rural facilities had higher median CD4 counts at enrolment than peri-urban or urban facilities. Treatment hospitals had lower median CD4 counts, were mostly in urban areas and served sicker patients.
Table 4 shows mutually adjusted odd ratios for the association of facility and patient level characteristics with firstly starting treatment and secondly death within one year of enrolment. Type of facility was strongly related to distance to ART initiation site and so we were not able to assess the effect of both variables. Similarly, staffing levels, patient load and vacant posts were all strongly related to each other and so we cannot assess the independent effects of all these variables. We chose to include the distance to ART initiation site, and staffing levels, in our models. The odds of initiating treatment declined with increasing distance to ART initiation site (p for trend <0.001). Patients enrolled in facilities where ART was not available on the same site were more likely to die (OR 1.55, 95% CI 1.28 to 1.89) compared to those enrolled in facilities where ART was available on site. Men were less likely to have started treatment (OR 0.67, 95% CI 0.64 to 0.71) and were more likely to have died within one year of enrolment (OR 1.30, 95% CI 1.23 to 1.37). Compared with older patients, those aged 15–29 years had lower odds of starting treatment (OR 0.80, 95% CI 0.75 to 0.85). The odds of starting treatment increased with weight and with CD4 count up to the threshold of 200 cells/mm3. The odds of dying within one year of enrolment increased with age and decreased with weight and CD4 count. Patients enrolled in 2007 were more likely to start treatment than those enrolling in earlier years.
Table 4.
Adjusted associations of facility and patient level characteristics with patient outcomes at 1-year post enrolment, using 25 imputed datasets Odds ratios are from logistic regression analyses adjusted for all variables in the table and including facility as a random effect.
| Outcome at 1 year | Patient started treatment | Patient died | |||
|---|---|---|---|---|---|
| N† | OR (95% CI) | P>|z| | OR (95% CI) | P>|z| | |
| Facility level characteristics | |||||
| Average staff (per 1000 patients) | |||||
| <1.5 | 16472 | 0.37 (0.17,0.79) | 0.010 | 1.06 (0.80,1.41) | 0.670 |
| 1.5–2.5 | 14274 | 1 (baseline) | 1 (baseline) | ||
| >2.5 | 14120 | 1.22 (0.61,2.43) | 0.580 | 1.07 (0.83,1.39) | 0.611 |
| Location | |||||
| Rural | 9995 | 0.71 (0.42,1.20) | 0.206 | 1.02 (0.81,1.29) | 0.848 |
| Urban/peri-urban | 34871 | 1 (baseline) | 1 (baseline) | ||
| Distance to treatment initiation site | |||||
| Same site | 12959 | 1 (baseline) | 1 (baseline) | ||
| <=8km | 12653 | 0.67 (0.35,1.30) | 0.239 | 1.51 (1.16,1.95) | 0.002 |
| 9km–15km | 13558 | 0.48 (0.20,1.16) | 0.105 | 1.88 (1.35,2.61) | <0.001 |
| >15km | 5696 | 0.35 (0.19,0.64) | 0.001 | 1.53 (1.18,1.98) | 0.001 |
| Patient level characteristics | |||||
| Sex | |||||
| Male | 14953 | 0.67 (0.64,0.71) | <0.001 | 1.30 (1.23,1.37) | <0.001 |
| Female | 29913 | 1 (baseline) | 1 (baseline) | ||
| Age (years) | |||||
| 15–29 | 12739 | 0.80 (0.75,0.85) | <0.001 | 0.72 (0.67,0.77) | <0.001 |
| 30–39 | 18114 | 1 (baseline) | 1 (baseline) | ||
| 40–49 | 10234 | 1.04 (0.98,1.10) | 0.236 | 1.07 (1.00,1.14) | 0.036 |
| >=50 | 3779 | 0.99 (0.90,1.07) | 0.737 | 1.33 (1.22,1.46) | <0.001 |
| Weight (kg.) | |||||
| <40 | 2619 | 0.55 (0.48,0.62) | <0.001 | 2.60 (2.30,2.95) | <0.001 |
| 40–49 | 9500 | 0.82 (0.76,0.88) | <0.001 | 1.55 (1.43,1.67) | <0.001 |
| 50–59 | 14560 | 1 (baseline) | 1 (baseline) | ||
| 60–79 | 15420 | 1.05 (0.99,1.12) | 0.118 | 0.69 (0.64,0.75) | <0.001 |
| >=80 | 2767 | 1.21 (1.07,1.37) | 0.003 | 0.48 (0.40,0.57) | <0.001 |
| First pre-treatment CD4count(cells/µL) | |||||
| <=25 | 3872 | 0.73 (0.67,0.80) | <0.001 | 3.75 (3.41,4.12) | <0.001 |
| 25–50 | 3457 | 0.87 (0.80,0.96) | 0.004 | 2.30 (2.08,2.54) | <0.001 |
| 50–100 | 6472 | 0.95 (0.89,1.03) | 0.208 | 1.71 (1.58,1.85) | <0.001 |
| 100–200 | 11637 | 1 (baseline) | 1 (baseline) | ||
| 200–350 | 10100 | 0.24 (0.22,0.26) | <0.001 | 0.54 (0.50,0.60) | <0.001 |
| >350 | 9328 | 0.05 (0.05,0.06) | <0.001 | 0.25 (0.22,0.28) | <0.001 |
| Year of enrolment | |||||
| 2004/5 | 17439 | 1 (baseline) | 1 (baseline) | ||
| 2006 | 12877 | 1.23 (1.15,1.31) | <0.001 | 0.99 (0.93,1.06) | 0.788 |
| 2007 | 14550 | 1.51 (1.42,1.60) | <0.001 | 0.92 (0.86,0.99) | 0.017 |
Average over imputations
In the complete case analyses (N = 24,510 reported in Appendix 2), the association of staffing levels with the odds of starting treatment was attenuated and the associations of location, distance from ART initiation site and CD4 were strengthened. The association of year of enrolment with the odds of dying were strengthened. In the imputed data, weight had a median of 56.5kg (IQR 49.0–65.4), compared to 56kg (IQR 49.0–64.0) in the complete case data. CD4 had a median of 173 cells/mm3 (IQR 80-316) in imputed data compared to 170 cells/mm3 (IQR 76-318) in complete case data.
Most of the variation across facilities in the odds of starting treatment was explained by facility, rather than patient, level characteristics. In the null model for starting treatment, facility level standard deviation was 1.58 (95% CI 1.33 to 2.08): this reduced to 1.46 (95% CI 1.26 to 1.83) when patient characteristics were added and to 1.22 (95% CI 1.13 to 1.38) when facility level characteristics were added. Including both patient and facility level variables resulted in a standard deviation of 1.19 (95% CI 1.11 to 1.33).
Discussion
Main findings
Based on a large record linkage study in the Free State province, we found a large amount of variation in odds of starting treatment across clinics and that differences in facility-level characteristics, rather than patient population, appear to drive this variation. Overall, by the end of 2008 around 34% of adults in the Free State public-sector programme received ART within one year of enrolment. The odds of starting treatment within one year increased between 2004 and 2007. However, large numbers of patients continue to die, and the majority of deaths (83%) occur in patients not receiving ART. Patients who were eligible at enrolment and started ART waited a median of almost 3 months before starting treatment. Patients were less likely to start treatment within one year if they were male, severely immunosuppressed (CD4 < 50 cells/mm3), underweight (<50kg), or sought care from a facility where ART initiation was more than 15km away. Men and patients enrolled at facilities where treatment could not be initiated were also more likely to die in the first year after enrolment.
Strengths and limitations
This cohort is one of only a small number of studies to report on outcomes after enrolment in HIV care and not just from treatment initiation (9;21;22). A strength of this study is that it is a large cohort, representative of the HIV-infected population in the Free State, based on routinely collected data. There is an established data capture system in place in the Free State with mechanisms to follow-up specific data queries in each clinic. National identity numbers have been documented for a large proportion of patients and regular linkage with the national death register means that ascertainment of death is fairly complete, even in those who were lost to the programme. In an earlier study using the Free State data, only 20% of all deaths were recorded in the medical records (7) which emphasises the importance of linking to the national death register.
A limitation of this study is the large number of patients with missing weight and CD4 measurements. Although CD4 measurements were routinely taken on almost all patients, the database recorded them on only half of the patients in the cohort. This was improved to three quarters of the patients by linkage using deterministic matching with records of blood results obtained from the NHLS.
Comparison with other studies
To the best of our knowledge, no studies have been able to consider these outcomes across a range of facilities as we have done. Three South African studies have shown high mortality rates in patients who had not initiated ART (9;21;23). They concluded that mortality could be reduced by minimizing unnecessary in-programme delays in treatment initiation. Our study found that facilities with shorter delays had lower mortality and therefore supports these findings. In addition to these studies, a simulation exercise (24) also concluded that prioritising those with the lowest CD4 counts should improve outcomes. A cohort based in Durban, which is well resourced by funding from non-governmental organisations (23) reported that 55% of ART eligible patients had started treatment within one year of diagnosis. The Free State CHAM programme appears to perform similarly as 53% of known ART eligible patients started treatment within one year of enrolment. South Africa faces unique challenges in the roll out of public-sector ART (25). Whilst the majority of deaths were in untreated patients, our study does not directly show the effect of ART in reducing mortality. A study by The Africa Centre, KwaZulu-Natal estimates that standardised mortality rates declined with a rate ratio of 0.78 for women and 0.71 for men after scale-up of public-sector ART (26).
Generalisability
Our results may be generalisable to patients enrolling in care in similar settings, but cannot be applied to the population who are HIV-infected and not seeking care. We have shown large differences in outcomes between facilities and this has implications on how HIV programmes are monitored. HIV prevalence is commonly monitored through sentinel surveillance and specific research questions are often addressed using sentinel sites (27). However, sentinel sites may over represent urban settings and not be representative of the general population (28). Care should be taken when evaluating programmes based on these data.
Policy implications
Patients who are assessed in one facility, but have to attend another to initiate ART, appeared to be disadvantaged compared to those who attend combined assessment and treatment facilities for all their care. This suggests that more should be done to enable assessment only facilities to provide treatment or to ensure that referrals to treatment facilities are done promptly. The referral process is one example of a barrier to accessing care. Drug Readiness Training may be another barrier as it can be protracted. Distance to appropriate facilities and transport costs may also be a barrier to access particularly in rural locations. It is important to note that it is not always clinic practices that lead to delayed uptake of ART. A study showed that in 2008, the Free State had the lowest ART coverage of all South African provinces, estimated at 25.8% which is likely due to the large proportion of eligible patients who are unaware of their HIV status (29). The 2005 Human Sciences Research Council household survey found that only 36% of HIV-infected adults had ever been tested for HIV (30). This lack of awareness of HIV status contributes to the delayed uptake of ART and campaigns should continue to target people who are unaware of their status.
We have shown that many of the sickest patients in this cohort die before they can initiate treatment, and demonstrated the need to fast-track the most at-risk patients. Fast-tracking would be easier to implement if treatment initiation were offered at assessment facilities, which is increasingly the case. This is being achieved through sessional visits by treatment facility doctors, and the STRETCH, or Streamlining Tasks and Roles to Expand Treatment and Care for HIV, intervention currently under evaluation in a pragmatic randomised controlled trial in the Free State. The aim of this intervention is to optimise the use of scarce clinical skills by allowing designated nurses to re-prescribe and initiate treatment in selected adults, thereby freeing up doctors to see sicker patients. The trial seeks to provide evidence that a triaged system, where nurses are supported to prescribe ART using clearly-defined clinical protocols, will improve outcomes in this setting (31).
Our findings support evidence that men are disadvantaged in terms of treatment access. It has previously been postulated that women may be disadvantaged in access to health care because of familial and child-rearing duties (32) and as a result, many services have focused on women. Provision of services for women are more developed through the need for antenatal and child health clinics. However, health-seeking behaviours between sexes are known to differ, with men less likely to access services they need (33). In addition, stigma, work responsibilities and homophobia might make it more difficult for men to accept an HIV diagnosis and to seek treatment (33). Other studies from similar resource-poor settings emphasise the need to encourage men to seek care earlier in HIV disease and to start ART, and suggest that health services to specifically address the needs of men should be developed (34–36).
Conclusion
Although increasing numbers of patients started ART between 2004 and 2007, many patients died before accessing ART. Patient outcomes could be improved by decentralisation of treatment services, fast-tracking the most immunodeficient patients and improving access especially for men.
Acknowledgements
This study was funded by the National Institute of Allergy and Infectious Diseases (NIAID, grant 1 U01 AI069924-01). Suzanne Ingle was funded by a UK Medical Research Council PhD Studentship and Margaret May was funded by UK Medical Research Council grant GO600599. The Free State ART data warehouse was developed with the help of a research grant from the International Development Research Centre, Canada (IDRC-102411). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We also gratefully acknowledge: the data capturers, nurses and doctors of the Free State programme, MediTech for assisting with data downloads to the warehouse, Terry Marshall and Sue Candy for assisting with the preparation of NHLS data, and Adrian Spoerri who helped with data linkage. We acknowledge all those who died without access to ART.
Footnotes
Conflict of interest
The authors declare that they have no conflict of interest.
Reference List
- 1.Bradshaw D, Groenewald P, Laubscher R, Nannan N, Nojilana B, Norman R, et al. Initial burden of disease estimates for South Africa, 2000. S Afr Med J. 2003 Sep;93(9):682–688. [PubMed] [Google Scholar]
- 2.Bradshaw D, Schneider M, Dorrington R, Bourne DE, Laubscher R. South African cause-of-death profile in transition--1996 and future trends. S Afr Med J. 2002 Aug;92(8):618–623. [PubMed] [Google Scholar]
- 3.South African Medical Research Council. South African National Burden of Disease Study 2000: Estimates of Provincial Mortality. 2005 [Google Scholar]
- 4.Dorrington R, Johnson L, Bradshaw D, Daniel T. The Demographic Impact of HIV/AIDS in South Africa. National and Provincial Indicators for 2006. 2006 [Google Scholar]
- 5.Shisana O, Rehle T, Simbayi LC, Zuma K, Jooste S, Pillay-van-Wyk V, et al. South African national HIV prevalence, incidence, behaviour and communication survey 2008. A turning tide among teenagers? 2009 [Google Scholar]
- 6.Lawn SD, Myer L, Wood R. Efficacy of antiretroviral therapy in resource-poor settings: are outcomes comparable to those in the developed world? Clin Infect Dis. 2005 Dec 1;41(11):1683–1684. doi: 10.1086/498030. [DOI] [PubMed] [Google Scholar]
- 7.Fairall LR, Bachmann MO, Louwagie GM, van Vuuren C, Chikobvu P, Steyn D, et al. Effectiveness of antiretroviral treatment in a South African program: a cohort study. Arch Intern Med. 2008 Jan 14;168(1):86–93. doi: 10.1001/archinternmed.2007.10. [DOI] [PubMed] [Google Scholar]
- 8.World Health Organisation. Scaling up antiretroviral therapy in resource-limited settings: Treatment guidelines for a public health approach. Geneva: WHO; 2004 [cited 2008 Sep 8];Available from: URL: http://www.who.int/hiv/pub/prev_care/en/arvrevision2003en.pdf.
- 9.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):2141–2148. doi: 10.1097/01.aids.0000194802.89540.e1. [DOI] [PubMed] [Google Scholar]
- 10.Wouters E, Van DW, Van LF, van RD, Meulemans H. Public-sector ART in the Free State Province, South Africa: Community support as an important determinant of outcome. Soc Sci Med. 2009 Aug 17; doi: 10.1016/j.socscimed.2009.07.034. [DOI] [PubMed] [Google Scholar]
- 11.Hirschhorn LR, Skolnik R. Making universal access a reality--what more do we need to know? J Infect Dis. 2008 May 1;197(9):1223–1225. doi: 10.1086/587185. [DOI] [PubMed] [Google Scholar]
- 12.Lawn SD, Harries AD, Anglaret X, Myer L, Wood R. Early mortality among adults accessing antiretroviral treatment programmes in sub-Saharan Africa. AIDS. 2008;22(15):1897–1908. doi: 10.1097/QAD.0b013e32830007cd. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.South African National Department of Health. National Antiretroviral Treatment Guidelines. 2004 [Google Scholar]
- 14.Meditech Information Technology I. Meditech www meditech co za. 2009 Available from: URL: http://www.meditech.co.za/01000000.html. [Google Scholar]
- 15.Statistics South Africa. Mortality and causes of death in South Africa, 2005: Findings from death notification. Statistical Release P0309 3. 2007 [cited 2008 Sep 9];Available from: URL: http://www.statssa.gov.za/publications/P03093/P030932005.pdf. [Google Scholar]
- 16.Royston P. Multiple imputation of missing values. The Stata Journal. 2004;4(3):227–241. [Google Scholar]
- 17.Royston P. Multiple imputation of missing values: update of ice. The Stata Journal. 2005;5(4):527–536. [Google Scholar]
- 18.Rubin D. Multiple imputation for nonresponse in surveys. New York: Wiley; 1987. [Google Scholar]
- 19.Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. doi: 10.1136/bmj.b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.StataCorp LP. STATA Statistical Software: Release 11. College Station, TX: Stata Corp LP; 2009. [Google Scholar]
- 21.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 1;51(2):135–139. doi: 10.1097/qai.0b013e3181a44ef2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Thai S, Koole O, Un P, Ros S, De MP, Van DW, et al. Five-year experience with scaling-up access to antiretroviral treatment in an HIV care programme in Cambodia. Trop Med Int Health. 2009 Sep;14(9):1048–1058. doi: 10.1111/j.1365-3156.2009.02334.x. [DOI] [PubMed] [Google Scholar]
- 23.Bassett IV, Regan S, Chetty S, Giddy J, Uhler LM, Holst H, et al. Who starts ART in Durban, South Africa?…Not everyone who should [Abstract WEAD102]. Presented at the 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention; Cape Town, South Africa. 2009. Jul, [Google Scholar]
- 24.Walensky RP, Wood R, Weinstein MC, Martinson NA, Losina E, Fofana MO, et al. Scaling up antiretroviral therapy in South Africa: the impact of speed on survival. J Infect Dis. 2008 May 1;197(9):1324–1332. doi: 10.1086/587184. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Ojikutu B, Jack C, Ramjee G. Provision of antiretroviral therapy in South Africa: unique challenges and remaining obstacles. J Infect Dis. 2007 Dec 1;196 Suppl 3:S523–S527. doi: 10.1086/521119. [DOI] [PubMed] [Google Scholar]
- 26.Herbst AJ, Cooke GS, Barnighausen T, KanyKany A, Tanser F, Newell ML. Adult mortality and antiretroviral treatment roll-out in rural KwaZulu-Natal, South Africa. Bull World Health Organ. 2009 Aug 14;87(10):754–762. doi: 10.2471/BLT.08.058982. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Boulle A, Bock P, Osler M, Cohen K, Channing L, Hilderbrand K, et al. Antiretroviral therapy and early mortality in South Africa. Bull World Health Organ. 2008 Sep;86(9):678–687. doi: 10.2471/BLT.07.045294. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.World Health Organisation. Towards Universal Access Scaling up priority HIV/AIDS interventions in the health sector. Geneva: WHO; 2008 Available from: URL: http://www.who.int/hiv/pub/Towards_Universal_Access_Report_2008.pdf.
- 29.Adam MA, Johnson LF. Estimation of adult antiretroviral treatment coverage in South Africa. S Afr Med J. 2009 Sep;99(9):661–667. [PubMed] [Google Scholar]
- 30.Shisana O, Rehle T, Simbayi LC, Parker W, Zuma K, Bhana A, et al. South African National HIV Prevalence, HIV Incidence, Behaviours and Communication Survey, 2005. Cape Town: HSRC Press; 2005. [Google Scholar]
- 31.Fairall LR, Bachmann MO, Zwarenstein MF, Lombard CJ, Uebel K, van Vuuren C, et al. Streamlining tasks and roles to expand treatment and care for HIV: randomised controlled trial protocol. Trials. 2008;9:21. doi: 10.1186/1745-6215-9-21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Box TL, Olsen M, Oddone EZ, Keitz SA. Healthcare access and utilization by patients infected with human immunodeficiency virus: does gender matter? J Womens Health (Larchmt) 2003 May;12(4):391–397. doi: 10.1089/154099903765448907. [DOI] [PubMed] [Google Scholar]
- 33.Mane P, Aggleton P. Gender and HIV/AIDS: What Do Men have to Do with it? Current Sociology. 2001;49(6):23–37. [Google Scholar]
- 34.Braitstein P, Boulle A, Nash D, Brinkhof MW, Dabis F, Laurent C, et al. Gender and the use of antiretroviral treatment in resource-constrained settings: findings from a multicenter collaboration. J Womens Health (Larchmt) 2008 Jan;17(1):47–55. doi: 10.1089/jwh.2007.0353. [DOI] [PubMed] [Google Scholar]
- 35.Cornell M, Myer L, Kaplan R, Bekker LG, Wood R. The impact of gender and income on survival and retention in a South African antiretroviral therapy programme. Trop Med Int Health. 2009 Jul;14(7):722–731. doi: 10.1111/j.1365-3156.2009.02290.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Keiser O, Anastos K, Schechter M, Balestre E, Myer L, Boulle A, et al. Antiretroviral therapy in resource-limited settings 1996 to 2006: patient characteristics, treatment regimens and monitoring in sub-Saharan Africa, Asia and Latin America. Trop Med Int Health. 2008 Jul;13(7):870–879. doi: 10.1111/j.1365-3156.2008.02078.x. [DOI] [PMC free article] [PubMed] [Google Scholar]