Contemporary Clinical Trials 72 (2018) 43–52

Contents lists available at ScienceDirect

Contemporary Clinical Trials

journal homepage: [Link]/locate/conclintrial

An intervention to optimise the delivery of integrated tuberculosis and HIV T

services at primary care clinics: results of the MERGE cluster randomised
trial
⁎
Kufa T.a,b,c, , Fielding K.L.d, Hippner P.a, Kielmann K.e, Vassall A.f, Churchyard G.J.a,b,d,
Grant A.D.b,g,h, Charalambous S.a,b
a
The Aurum Institute, Johannesburg, South Africa
b
The School of Public Health, University of the Witwatersrand, Johannesburg, South Africa
c
Centre for HIV and STIs, National Institute for Communicable Diseases, Johannesburg, South Africa
d
Department of Infectious Diseases Epidemiology, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, United Kingdom
e
Institute for Global Health and Development, Queen Margaret University, Edinburgh, United Kingdom
f
Department of Global Health and Development, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, United Kingdom
g
Department of Clinical Research, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, United Kingdom
h
Africa Health Research Institute, School of Nursing and Public Health, University of KwaZulu-Natal, South Africa

A R T I C LE I N FO A B S T R A C T

Keywords: Objectives: To evaluate the effect of an intervention to optimize TB/HIV integration on patient outcomes.
Tuberculosis Methods: Cluster randomised control trial at 18 primary care clinics in South Africa. The intervention was
HIV placement of a nurse (TB/HIV integration officer) to facilitate provision of integrated TB/HIV services, and a lay
Integration health worker (TB screening officer) to facilitate TB screening for 24 months. Primary outcomes were i) in-
Cluster randomised trial
cidence of hospitalisation/death among individuals newly diagnosed with HIV, ii) incidence of hospitalisation/
death among individuals newly diagnosed with TB and iii) proportion of HIV-positive individuals newly diag-
nosed with TB who were retained in HIV care 12 months after enrolment.
Results: Of 3328 individuals enrolled, 3024 were in the HIV cohort, 731 in TB cohort and 427 in TB-HIV cohort.
For the HIV cohort, the hospitalisation/death rate was 12.5 per 100 person-years (py) (182/1459py) in the
intervention arm vs. 10.4/100py (147/1408 py) in the control arms respectively (Relative Risk (RR) 1.17 [95%
CI 0.92–1.49]).For the TB cohort, hospitalisation/ death rate was 17.1/100 py (67/ 392py) vs. 11.1 /100py (32/
289py) in intervention and control arms respectively (RR 1.37 [95% CI 0.78–2.43]). For the TB-HIV cohort,
retention in care at 12 months was 63.0% (213/338) and 55.9% (143/256) in intervention and control arms (RR
1.11 [95% 0.89–1.38]).
Conclusions: The intervention as implemented failed to improve patient outcomes beyond levels at control
clinics. Effective strategies are needed to achieve better TB/HIV service integration and improve TB and HIV
outcomes in primary care clinics.
Trial registration: South African Register of Clinical Trials (registration number DOH-27-1011-3846).

1. Introduction HIV services refers to co-location and joint delivery of TB and HIV-
related services. TB and HIV integration comprise activities to reduce
South Africa faces a large burden of the dual epidemics of tu- morbidity and mortality from HIV among individuals with TB - HIV
berculosis (TB) and HIV. With an HIV prevalence of 12.8% in the counselling and testing, initiation of cotrimoxazole preventive therapy
general population [1], an estimated TB incidence rate of 781 per (CPT) and earlier initiation of antiretroviral therapy (ART) as well ac-
100,000 of the population [2], and HIV positivity of 59% among in- tivities to reduce morbidity and mortality from TB among individuals
dividuals diagnosed TB in 2016 [2], the country could benefit from living with HIV - intensified case finding, isoniazid preventive therapy
improved integration of TB and HIV services. The integration of TB and (IPT), TB infection control and the early initiation of ART [3, 4]

⁎
Corresponding author at: Centre for HIV and STIs, National Institutes of Communicable Diseases, 1 Modderfontein Road, Sandringham, United Kingdom.
E-mail addresses: tendesayikc@[Link], tendesayik@[Link] (T. Kufa).

[Link]
Received 30 November 2017; Received in revised form 4 July 2018; Accepted 23 July 2018
Available online 25 July 2018
1551-7144/ © 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
([Link]
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

Different models have been used to deliver integrated TB/HIV ser- were tailored to the physical layout, human resources availability and
vices at primary care level [5, 6]. These models range from separate needs and priorities of the clinics with respect to TB/HIV integration. In
services with referral between them to complete integration where care order to support the implementation of such activities, the trial in-
for both diseases is provided by the same provider in the same con- troduced two new staff cadres – a TB/HIV integration officer (IO) and a
sultation [6]. Evidence of the benefits of TB/HIV integration is limited TB screening officer (SO). The IOs who were professional nurses with
to evidence of improved TB treatment outcomes, increased ART uptake prior TB/HIV experience. Following study specific training, the IOs
and shorter time to ART initiation among HIV positive individuals worked with clinic managers to identify gaps and barriers to effective
treated for TB [7–10]. Information on the effectiveness of interventions TB and HIV service integration, to support the clinic with better de-
to improve integration of TB and HIV services and the effect of such livery of TB and HIV collaborative services and to transition towards
integration on distal outcomes such as mortality, hospitalisations and the single provider model of TB/HIV integration as recommended by
retention in care is however limited [6, 11–13]. We report on the the national TB/HIV integration guidelines [19]. At placement, both the
outcomes of a cluster randomised trial to evaluate the effect of an in- IOs and clinic staff were made aware that the IOs were meant to be
tervention to optimise integration of TB and HIV-related services on catalysts for change, that the IOspresence in the clinics was limited to
morbidity (hospitalizations), mortality and retention in care among the intervention period (1st September 2011–31st August 2013), and
individuals newly diagnosed with TB and individuals newly diagnosed that the clinics were expected to sustain any TB/HIV integration ac-
with HIV at primary care clinics in South Africa. tivities initiated by the IOs beyond this intervention period.
The broad function of the SOs was to support the screening of HIV
2. Methods positive individuals for TB, the demand for which was expected to in-
crease with better TB/HIV integration. The TB screening officers were
2.1. Study design and setting lay workers with previous experience of working in health care settings
and their placement was considered task shifting for TB screening- that
The setting and methods of this cluster-randomised trial have been is use of lay workers in place of nurses to conduct TB screening. The SOs
previously reported. [14] Clusters comprised primary care clinics in a were meant to work with the IOs, clinic staff and managers to identify
sub-district of Ekurhuleni District, Gauteng province, South Africa. At bottlenecks to TB screening and determine where they would be best
randomization in August 2011, all the participating clinics provided TB located in the clinic in order to maximise the number of HIV positive
and HIV services under the same roof, but these services were run se- patients screened for TB. The location of the SOs in the clinics was also
parately with varying levels of cross-referral between them – that is allowed to vary depending on the physical layout and the flow of pa-
they had co-location of services. This level of integration- co-location of tients in the [Link] were expected to identify a staff member
services- was assumed sub-optimal with the one-stop-shop model- a who would continue with TB screening beyond the intervention period.
model in which both TB and HIV services are provided by the same
provider in the same consultation- considered ideal. [15] In addition, in 2.4. Description of the control clinics
the sub-district where the trial was implemented, there was an on-going
TB/HIV health systems strengthening programme whose main activities In control clinics, TB and HIV services were provided as re-
were mentoring clinic nurses on ART and training on recording and commended by the South African National Department of Health
reporting for TB and HIV. At the time of clinic randomization, South guidelines for TB/HIV integration. At randomization, these guidelines
African national HIV treatment guidelines recommended ART for HIV recommended provision of all TB/HIV collaborative services including
positives with WHO Stage 3 or 4 disease, CD4 count < 200 cells/μl, TB on-site TB diagnostic services and ART initiations and encouraged the
disease or pregnancy with at CD4 counts < 350 cells/μl. [16] These physical integration of the services. [19] However implementation of
guidelines changed in March 2012 making patients who were HIV po- these guidelines in clinics was variable and sub-optimal with no clinic
sitive and had CD4 counts < 350 cells/μl, TB or were pregnant eligible providing the one-stop-shop model at baseline.
for ART regardless of CD4 count [17]. In 2016, CD4 criteria for ART
eligibility were removed, making all positive individuals eligible for 2.5. Baseline assessments and monitoring fidelity to the intervention
ART [18].
At baseline, assessments of participating clinics were conducted to
2.2. Randomization establish clinic characteristics and the extent of co-location of TB and
HIV services. Fidelity to the intervention was assessed through bi-
Of 32 available clinics, 18 were purposively selected, i) ensuring monthly monitoring and supervision visits to intervention clinics by the
availability of TB diagnostic and treatment services, ii) at least 40 TB study coordinators and monthly group meetings for IOs and SOs. In
patients registered at the clinic in the preceding year, iii) good geo- addition, the IOs and SOs also completed time sheets in order to
graphic spread the sub-district (clinics not too close to one another) and document how they spent their time in the clinics. At the end of the
iv) absence of other competing research studies. The selected clinics intervention period, facility assessments were conducted in order to
were stratified into high or low case-fatality clinics (i.e. TB CFR measure the extent of TB/HIV integration at both intervention and
rates > 3.5% OR ≤3.5% among smear-positive TB patients diagnosed control clinics. During the assessments, quantitative data on clinic level
in 2010), and then randomly assigned within strata to either the in- indicators of TB and HIV integration were collected from routine clinic
tervention or the control arm. Randomization was done by pulling data for the preceding month (May 2013). In addition, assessments of
numbers out of bowls. This was done at a meeting with clinic managers the location of key integrated TB/HIV services (i) TB screening for HIV
and sub-district health management staff and facilitated by the trial positive individuals, ii) HIV testing for individuals diagnosed with TB
statistician. patients and iii) ART treatment and monitoring for individuals diag-
nosed with both TB and HIV- were conducted. Fig. S1 shows the timing
2.3. Description of the intervention of these activities during the trial.

The intervention was a set of activities to optimise the integration of 2.6. Enrolment and follow up of evaluation cohorts
TB and HIV services beyond the level prevailing at the control clinics
and has been described previously. [14] Briefly, the intervention were Research assistants placed at each clinic and with assistance from
any activities meant to address clinic level barriers to better delivery clinic staff, identified and enrolled a consecutive sample of clinic at-
and co-location of TB/HIV services at each intervention clinic. These tendees aged ≥18 years and i) newly diagnosed with TB, defined as

44
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

having initiated TB treatment in the 60 days preceding enrolment (TB up time for the hospitalisation/mortality outcome. We assumed a
cohort) or ii) newly diagnosed with HIV, defined as having tested HIV sample size of 60 patients per clinic for the TB cohort and 165 patients
positive in the 60 days preceding enrolment (HIV cohort) respectively. for the HIV cohort per clinic across the 18 clinics taking into account
Individuals who met inclusion criteria for both TB and HIV cohorts refusals and loss to follow up [14].For the TB cohort, assuming an in-
were identified as the TBHIV cohort. At enrolment, baseline ques- cidence of hospitalisation or mortality of 20 per 100 person-years (py)
tionnaires were administered collecting information on participant in the control clinics, the target sample size cited above gave 92%
demographic and socioeconomic characteristics, TB screening, diag- power to assess a 50% reduction in either mortality or hospitalisation at
nosis and treatment, HIV testing, care and treatment and on factors the intervention clinics [14]. For the HIV cohort, we assumed an in-
related to trial outcomes, as well as contact information for follow-up. cidence of hospitalisation or mortality of 15 per 100 py in the control
Enrolment of participants started at intervention clinics at least three clinics which gave 86% power to assess a 40% reduction in either
months after the start of the intervention, to allow time for the inter- mortality or hospitalisation at the intervention clinics. For the retention
vention to have an effect on clinic activity. in care endpoint, we assumed that 70% of TB cohort would be HIV
Participants were followed up through telephone interviews at six positive and that 30% would not be retained in care over a 12-month
monthly intervals for at least 12 months and up to 18 months for some period in the control clinics. [14] This gave 90% power to assess a 50%
participants recruited earlier on. During the calls, data on clinic at- reduction in the proportion not retained over a 12-month period in the
tendance, TB and/or HIV care received were collected. Telephonic in- intervention clinics.
terviews were considered necessary to measure outcomes of partici- For the primary outcome person time at risk was measured from
pants who transferred from or stopped attending care at the clinic of date of HIV test or starting TB treatment to the earlier of first hospi-
[Link] assistants also conducted medical record review talisation, death, or date last known to be alive, for the HIV and TB
and abstractions. If participants could not be reached through the tel- cohorts, respectively. Date last known to be alive was defined using
ephone, their next-of-kin were contacted and information on whether multiple data sources: date of most recent interview; date of most re-
participants had moved, been hospitalised or died obtained. For parti- cent routine clinic visit; date last seen alive as reported by the partici-
cipants who had a valid South African national identification number, pant's next of kin (if participant was lost to follow-up); and the vital
the national vital statistics register was used to determine vital status statistics register (those with a South African identity number).
and date of death if the participant had died. Fig. S1 shows the timing of The analysis of the intervention effect on primary and secondary
participant enrolment in relation to the implementation of the inter- outcomes was based on comparisons of two arms (intervention vs
vention. control) for in the three cohorts (TB, HIV and TBHIV cohorts).The
analyses generated cluster-level summaries and took into account the
2.7. Outcomes stratified randomization. Briefly, the logs of the cluster-level summaries
of the outcomes were used to calculate geometric means in each of the
The trial had three primary outcomes: i) the incidence of death or two arms. An approximate standard error for the log (risk or rate) ratio
hospitalisation in the HIV cohort; ii) the incidence of death or hospi- based on geometric means of cluster risks or rates were calculated by
talisation in the TB cohort; and iii) the 12-month retention in care in the two-way analysis of variance on randomization stratum, arm, and the
TBHIV [Link] was defined as any death occurring during the interaction between stratum and arm. The 95% confidence interval (CI)
follow-up period as reported by any of next of kin OR detected through was calculated from this standard error, using a t-statistic with 14 de-
the clinic records OR the vital statistics register. Hospitalisations were grees of freedom. An adjusted analysis, taking into account baseline
either self-reported or reported by the next of kin during the follow-up imbalances, was also conducted using a two-stage approach re-
period (excluding those associated with pregnancy). Retention in care commended for studies with a small number of clusters. [14, 21, 22]
among HIV positive individuals newly diagnosed with TB was defined Pre-specified subgroup analyses for sex, CD4 count strata (< 350 cells/
as clinic attendance for routine HIV care with receipt of IPT, CPT, and μl versus 350 or over cells/μl), social –economic status and enrolment
ART, CD4 or viral load testing which was self - reported or documented period (last year of enrolment versus earlier) were planned for all pri-
on medical record review during the period 305 to 425 days from the mary outcomes. Socioeconomic status was measured using the socio-
date of starting TB treatment. This corresponded to a 12 month visit economic position index (defined as the number of assets that the-
post ART initiation with a 60 day window around the visit. participant ownedoutof a list of 16 listed (working electric/gas stove,
The study also had several secondary outcomes. These outcomes working vacuum cleaner,working washing machine,working satellite
were the comparison among intervention and control clinics ofi) television,working digital videodisc[DVD] player, working motor/
Proportion of the TB cohort whoknew their HIV status by the end of TB car,working mail/ post box/bag,working mail delivery at home,-
treatment;ii) median CD4 count at enrolment in the HIV cohort;iii) working radio,working television [TV],working computer,working re-
proportion of HIV cohort who started IPT within 12 months of the HIV frigerator,working landline telephone,working cell phone,working bi-
test;iv) the proportion of the TB cohort who successfully completed TB cycle,working motorcycle/ scooter). The scores were determined for
treatment during 12 months of follow up;and v) the proportion of the each participants and distribution of the index determined using prin-
ART-eligible HIV cohortwho initiated ART by 10 weeks after date of cipal component analysis [23]. Participants who belonged to lower,
CD4 count (based on the current guidelines). These secondary outcomes middle and upper thirds of this distribution were identifi[Link] items
were selected because they represented steps in the pathway from di- included in the index were adapted from a questionnaire used by Sta-
agnosis of TB or HIV to primary outcomes and aimed to measure the tistics South Africa household surveys [24].
extent of delivery of integrated TB/HIV services. For both primary and
secondary outcomes individuals were considered to enter the cohort on 2.9. Ethical considerations
the date of the HIV test and date of starting TB treatment for the HIV
and TB cohorts respectively. Ethical clearance was obtained from the research ethics committees
of University of the Witwatersrand and the London School of Hygiene &
2.8. Sample size and statistical analysis Tropical Medicine as well as from the Centres for Disease Control (CDC)
office of the Associate Director for [Link] to conduct the
The sample size calculations took into account the clustered design study was also obtained from the Ekurhuleni District Department of
through the coefficient of variation [20–22] and have been outlined Family Health prior to randomization. The trial was registered on the
previously [14]. Briefly, we assumed nine clinics per arm, a type I error South African Register of Clinical Trials (registration number DOH-27-
of 5%, a coefficient of variation of 0.25 and 15 months average follow- 1011-3846). Participants enrolled in the evaluation cohorts were

45
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

18 clusters (clinics)
Intervention Total participants 4003 Control

9 clusters allocated to intervention 9 clusters allocated to control

Screened for study eligibility 1999 Screened for study eligibility 2004
Median participants per cluster; 242 Median participants per cluster; 252
(Range; 101-348) (Range; 109 -274)

Non-eligible: 379 in 9 clusters Non-eligible: 440 in 9 clusters

(Range; 10-97) (Range; 18-86)
Declined enrolment: 53 in 8 clusters Declined enrolment: 67 in 9 clusters
(range; 1-25) (range; 1-21)

Enrolled: Enrolled:
1567 participants (from 9 clusters) 1497 participants (from 9 clusters)
Median participants per cluster; 183 Median participants per cluster; 193
(Range; 85-248) (Range; 77-224)

Withdrawn: 27 in 8 clusters Withdrawn: 13 in 7 clusters

6 duplicates 8 duplicates
3 found to be under-age 2 missing consent forms
4 missing consent forms 2 illiterate participants had no witnesses
8 illiterate participants had no witnesses when giving consent
when giving consent 1 participant did not sign the consent
6 participants did not sign the consent form
form

Analysed: Analysed:
1540 participants (from 9 clusters) 1484 participants (from 9 clusters)
Median participants per cluster; 180 Median participants per cluster; 191
(Range; 81-247) (Range; 76-224)
1054 (68%) with valid SA ID numbers 975 (66%) with valid SA ID numbers
1125 (73%) with 12-month follow-up 1138 (72%) with 12-month follow-up

Fig. 1. CONSORT diagram for the HIV cohort.

requested to provide written informed consent before enrolment into 3.2. Fidelity to the intervention
the study. The study was conducted according to ICH/GCP guidelines
and had oversight from a Data Monitoring Committee. The intervention was implemented over a 24-month period as
planned. IOs were present at all intervention clinics for 197 person-
months (91% of the intervention period), while SOs were present for
3. Findings 191 person-months (88% of the intervention period). From analyses of
timesheets and in-depth interviews, the main activities of the TB/HIV
3.1. Description of clinics integration officers were providing direct care to patients (mean
monthly proportion 39.6%), administration duties (mean 32.8%), re-
At randomization, intervention clinics were comparable to control porting and recording TB and HIV activities (mean 10.6%), mentoring
clinics with respect to opening hours, adult head count per month and counsellors and data clerks (mean 3.6%) (See Fig. S2), in contrast to the
the number of TB and HIV services provided (Supplementary Table leadership and coordination role envisaged at intervention design. The
S1).During the enrolment and follow up periods, no clinics were main activities of the SOs were screening patients for TB, completing
dropped from the trial (Fig. 1). registers and following up individuals with confirmed TB who failed to
return to initiate TB treatment as intended at trial conception.

46
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

18 clusters (clinics)
Intervention Total participants 4003 Control

9 clusters allocated to intervention 9 clusters allocated to control

Screened for study eligibility 1999 Screened for study eligibility 2004
Median participants per cluster; 242 Median participants per cluster; 252
(Range; 101-348) (Range; 109 -274)

Non-eligible: 1519 in 9 clusters Non-eligible: 1625 in 9 clusters

(Range; 89-275) (Range; 101-222)
Declined enrolment: 53 in 8 clusters Declined enrolment: 67 in 9 clusters
(Range; 1-25) (Range; 1-21)

Enrolled: Enrolled:
427 participants (from 9 clusters) 312 participants (from 9 clusters)
Median participants per cluster; 61 Median participants per cluster; 35
(Range; 11-77) (Range; 5-61)

Withdrawn: 6 from 5 clusters Withdrawn: 2 from 2 cluster

2 found to be under-age 2 duplicate enrolment
1 missing consent form
2 illiterate participants had no witnesses
when giving consent
1 participant did not sign the consent
form

Analysed: Analysed:
421 participants (from 9 clusters) 310 participants (from 9 clusters)
Median participants per cluster; 59 Median participants per cluster; 35
(Range; 11-77) (Range; 4-60)
312 (74%) with valid SA ID numbers 209 (67%) with valid SA ID numbers
303 (72%) with 12-month follow-up 246 (79%) with 12-month follow-up

Fig. 2. CONSORT diagram for the TB cohort.

3.3. Delivery of integrated TB/HIV service at intervention and control up until the end of July 2014 as planned. A total of 3024 participants
clinics at the end of the intervention period (1540 in intervention and 1484 in control arm) were eligible for in-
clusion in the HIV cohort (Fig. 1) and 731 (421 in intervention and 310
Overall, there was no difference in location and extent of delivery of in control clinics) were eligible for inclusion in the TB cohort (Fig. 2).
TB and HIV collaborative services based on facility assessments and There were 427 participants (173 control and 254 intervention clinics)
analysis of routine data at the end of the intervention period (see Tables included in both cohorts and therefore made up the TBHIV cohort.
S2 and S3). However, at the end of the intervention period, there was a Figs. 1 and 2 show CONSORT diagrams for the TB and HIV cohorts.
higher proportion of newly diagnosed HIV positive clinic attendees Participants in both cohorts were comparable with respect to most
screened for TB (94.3% vs 39.1%) and a lower proportion of newly characteristics at enrolment, by arm (Tables 1 and 2). For the HIV co-
diagnosed HIV positive clinic attendees initiated on IPT initiation hort, however, participants in the intervention arm were more likely to
(29.1% vs 40.2%) at intervention clinics compared to control (see Table be South African born, to be unemployed, to have a lower asset count
S3). and to be in the bottom third of the socio-economic position index, to
be on CPT at enrolment and less likely to be on ART at enrolment
compared to participants in the control arm. For the TB cohort, parti-
3.4. Description of evaluation cohorts cipants in the intervention arm were also more likely to be South
African born and to be in the bottom third of the socio-economic po-
Between January 2012 to July 2013, 4003 individuals were sition index compared to those in the control arm.
screened for eligibility and 3393 participants were enrolled for follow

47
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

Table 1 Table 2
Characteristics of individuals who were newly diagnosed with HIV at enrolment Characteristics of individuals who were newly diagnosed with TB at enrolment
by arm (n = 3024). by arm (N = 731).
Characteristic Intervention Control (1484) Characteristic Intervention(421) Control (310)
(1540)
Age in years, median (IQR) 36.6 (29.3–42.7) 35.9(29.6–42.7)
Age in years, median (IQR) 34.2 (28.5–40.2) 33.7(28.5–40) Males, n (%) 225 (53.4) 173(55.8)
Males, n (%) 608(39.5) 530(35.7) Married or cohabiting, n (%) 151(35.9) 125(40.3)
Married or cohabiting, n (%)a 590(38.3) 678(45.7) South African born, n (%)a 379(90) 249(80.3)
South African born, n (%)a 1285 (83.4) 1167 (78.6) Completed grade 12 or higher, n (%)a 152(36.1) 143(46.1)
Completed grade 12 or higher, n (%)a 528(34.3) 606(40.8) Employed, n (%) 226(53.7) 179(57.7)
Employed, n (%)a 819(53.2) 876(59.0) Total asset count, median (IQR),b 7 (4–8) 7 (5–9)
Asset count, median (IQR)a,b 6 (4–8) 7 (5–9) In bottom one third of socio-economic 164 (39.0) 77 (24.8)
In bottom one third of socio-economic 575 (37.3) 410 (27.6) position index, n (%)c
position index, n (%)c Distance from clinic < 5 km, n (%) 324 (77) 234(75.5)
Distance from clinic < 5 km, n (%) 1154 (74.9) 1111 (74.9) Worked in the mines, n (%)a 31(7.4) 11(3.5)
Lived alone, n (%) 273(17.8) 303(20.4) Lived alone, n (%) 71(16.7) 50(16.1)
Length of time in town (years),median 7 (1–14) 5(1−13) Length of timein town, median (IQR) 7(1–15) 6 (2–14.5)
(IQR), n Ever smoked, n (%) 133(31.6) 93 (30)
Ever smoked, n (%) 375 (24.4) 361 (24.3) Currently drink alcohol, n (%) 74 (17.6) 71 (22.9)
Currently drink alcohol, n (%) 345 (22.4) 345 (23.2) Previous TB treatment (n, %) 65(15.4) 40(12.9)
Hospitalized in the last 12 months, n (%) 264 (17.1) 233 (15.7) Hospitalised in the last 12 months, n (%) 115 (27.3) 82 (26.5)
Know someone who was ill or died from 814 (52.9) 870 (58.6) Know someone who was ill or died from 185 (43.9) 187(60.3)
HIV, n (%) HIV, n (%)a
Believes that traditional healers can treat 99 (6.5) 134 (9.0) Believes that traditional healers can treat 23 (5.5) 14(4.5)
HIV, n (%) HIV, n (%)
Ever treated for TB 120 (7.8) 100 (6.7) Documented HIV test at enrolment, n (%) 396 (94.1) 288(92.9)
Newly diagnosedwith TB 254 (16.5) 173 (11.7) HIV positive at enrolment, n (%) 329(78.1) 254 (81.9)
Duration since HIV test (median, IQR) 12 (7–21) 11 (7–20) Duration since HIV test (median, IQR) 20 (7–44) 21 (8–62)
On CPT at enrolment, n (%)a,d 602(39.1) 472(31.8) Newly diagnosed HIV positive, n (%) 245 (58.2) 172 (55.5)
IPT use at or prior to enrolment, n (%)a,d 168(10.9) 208(14.5) On CPT at enrolment, n (%)a,d,e 186(56.5) 106(41.7)
Attending usual clinic for HIV care, n (%) 1461 (94.9) 1418 (95.6) On ART at enrolment, n (%)d,e 91 (27.7) 79(31.1)
On ART at enrolment, n (%)a,d 331 (21.5) 397(26.8) Pregnant, n (%)f 8 (4.1) 5 (3.7)
Pregnanta,e 157 (16.9) 236 (24.8)
IQR- interquartile range.
IQR - interquartile range. a
Variables for which there appeared to be imbalance between arms.
a b
Variables for which there appeared to be some imbalance between arms. The number of assets the participant owned out the 16 listed (working
b
The number of assets the participant owned out 16 assets listed (working electric/gas stove, working vacuum cleaner, working washing machine,
electric/gas stove, working vacuum cleaner, working washing machine, working satellite television, working digital video disc (DVD) player, working
working satellite television, working digital video disc (DVD) player, working motor/ car, working mail/ post box/bag, working mail delivery at home,
motor/car, working mail/ post box/bag, working mail delivery at home, working radio, working television (TV), working computer, working re-
working radio, working TV, working computer, working refrigerator, working frigerator, working landline telephone, working cell phone, working bicycle,
landline telephone, working cellphone, working bicycle, Working motorcycle/ Working motorcycle/ scooter).
scooter). c
Based on a socio-economic position index with the following elements
c
Based on a socio-economic position index with the following elements based on a socio-economic position index with the following elements (assets,
(assets, water source, type of toilet, type of floors and walls at participants water source, type of toilet, type of floors and walls at participants dwelling).
dwelling). Variables to include in the SEP index determined through principal Variables to include in the SEP index determined through principal component
component analysis. analysis.
d d
Started Isoniazid preventive therapy (IPT), Cotrimoxazole preventive Denominator is HIV positive TB patients (329 in the intervention arm and
therapy (CPT) or antiretroviral therapy (ART) prior to or on the day of enrol- 254 in the control arm).
ment. e
started Cotrimoxazole preventive therapy (CPT) or antiretroviral therapy
e
Out of 931 females at intervention clinics and 953 females at control (ART) prior to or on the day of enrolment.
clinics. f
Out of 196 females at intervention clinics and 136 females at control clinics.

3.5. Effect of the intervention on the co- primary outcomes of hospital randomization strata and variables showing baseline imbalance (see
admissions and or mortality Table 3 and Tables S5 in the supplementary appendix). The sample was
too small to conduct subgroup analyses for this cohort.
The median follow-up for the HIV and TB cohorts was 12.3 and
12.2 months, respectively, and similar by [Link] individuals in the
3.6. Effect of the intervention on the co-primary outcome of retention in
HIV cohort, the incidence of hospitalisation or death in 12 months of
care
follow up was 12.5 per 100 person-years (py) (182/1459) and 10.4 per
100 py (147/1408) in the intervention and control clinics respectively,
Of 338 individuals in the TB-HIV cohort, 213/338 (63.0%) were
giving an incidence rate ratio [IRR] of 1.17 (95% CI 0.92, 1.49), ad-
retained in care by 12 months in the intervention arm, compared to
justing for randomization [Link] adjusting for age, sex, randomi-
143/256 (55.9%) in the control arm, giving an adjusted risk ratio (RR)
zation strata and variables showing baseline imbalance the adjusted
of 1.11 (95% CI 0.89–1.38) (Table 3), adjusting for randomization
IRR was 1.08 (95% CI 0.84–1.38) (Table 3 and Table S4 supplementary
strata. A fully-adjusted analysis gave similar results. The sample was too
appendix). In subgroup analyses, the effects of the intervention were
small to conduct subgroup analyses (see Table 3 and Table S5 in the
similar by sex, socioeconomic position index (low versus high), and
supplementary appendix).
year of enrolment and among those with CD4 counts < 350 cells/μl
(Table 4). In the TB cohort, the incidence of hospitalisation or death
was 17.1 per 100 py (67/392py) and 11.1 per 100py (32/289py) in the 3.7. Effect of the intervention on the secondary outcomes
intervention and control clinics respectively, giving an adjusted IRR
1.22 (95% CI 0.70–2.13) in a model adjusting for age, sex, Table 5 shows the effect of the intervention on the secondary

48
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

Effect of intervention on the primary outcomes of incidence of hospitalizations and mortality among individuals newly diagnosed with TB and individuals newly diagnosed with HIV and proportion retained in HIV care by

Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, employment status, socio-economic position index, on antiretroviral therapy at enrolment, Cotrimoxazole

Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, employment status, socio-economic position index, on antiretroviral therapy at enrolment and Cotrimoxazole
outcomes. Because of high coverage of HIV testing at enrolment in both

p-Valueb
p-Value

0.46d
0.51b
arms (94.1% in the intervention arm vs. 92.9% in the control clinics),

0.28
the outcome of proportion of the TB cohort who knew their HIV status
by the end of TB treatment was not determined. There was no differ-
Adjusted incidence rate

ence in the mean CD4 counts at enrolment among those in the HIV

1.22 (0.70, 2.13)d

1.08 (0.84, 1.38)b

Adjusted risk ratioe

cohort between the arms (232 cells/μl [standard deviation (SD)

1.10 (0.92, 1.31)

ratio (95% CI)

181 cells/μl] in the intervention arm vs 246 cells/μl [SD 197 cells/μl in
the control arms], adjusted mean difference in square root of CD4 count
0.14 (95% CI -0.64, 0.92)]. There wereno significant differencesin the i)
proportions of ART eligible individuals in the HIV cohort who initiated

Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, socio-economic position index, CPT at enrolment and being in the HIV cohort.
ART by 10 weeks after CD4 count testing- a median 10 days after HIV
p-Valuea

testing (67% vs 70.4%, RR 0.90 [95% CI 0.75–1.10]), ii) ART eligible

0.18

0.25

individuals in the TB cohort who started ART by 10 weeks of HIV

p-Valuea

testing (39.3% vs 38.8%, RR 1.05 [95% CI0.67–1.63]), iii) individuals

Adjusted incidence rate

0.33

in the HIV cohort who started IPT by the end of 12 months (39.8% vs
42.6, RR 0.94 [95% CI 0.64–1.33]) and iv) individuals newly diagnosed
1.17 (0.92, 1.49)

1.37 (0.78, 2.43)

ratio1 (95% CI)

with TB who successfully completed TB treatment during 12 months of

follow up (76.7% vs 78.4%, RR 0.94 [95% CI 0.83–1.06]), after ad-
Adjusted risk ratioa

justing for randomization strata (Tables 5 and S7–S11 in the supple-

1.11 (0.89–1.38)

mentary appendix).

4. Discussion
follow up) Incidence per 100 person-years
(Hospitalisation or deaths/person-years of

This cluster-randomised trial evaluated the effect of an intervention

to optimise the delivery of integrated TB and HIV care on patient re-
levant outcomes. The intervention as implemented in the trial did not
55.9% (143)

have an effect on morbidity, mortality and retention in care among

individuals newly diagnosed with TB or HIV. This was most likely be-
% (n)

cause the intervention did not succeed in improving coordinated de-

livery and co-location of TB/HIV services beyond the levels at the
(147/1408) 10.4

(32/289) 11.1

control clinics.
This trial intervention was intended to address lack of coordination
256
N

as a barrier to TB/HIV integration while taking into account variability

Control arm

in other TB/HIV integration-related challenges and clinic-level re-

Two clusters (one intervention and one control) had zero outcomes. For analysis 0.5 events assumed.

sponses to them. The IO and SO were meant to be catalysts for im-

1484

63.0% (213)
310

proved TB/HIV integration while demonstrating that assigning specific

N

health cadres to focus on TB/HIV integration alone could improve de-

% (n)

livery of the services. Implementation of the intervention was expected

to result in increased proportions of HIV positive individuals tested
follow up) Incidence per 100 person-years
(Hospitalization or deaths/person-years of

earlier in the course of the infection, screened for TB and started on IPT
as well as increased proportions of individuals diagnosed with TB who
338

are tested for HIV, started on CPT or ART thereby reducing morbidity
N

and mortality.
Although the implementation of this trial intervention was sub-op-
timal and failed to improve integration, observational studies con-
Proportion ofHIV positive individuals with TB retained in careby 12 months

preventive therapy (CPT) at enrolment and being in the TB cohort.

ducted elsewhere have not consistently demonstrated reductions in

(182/1459) 12.5

(67/392) 17.1

mortality with better integration. Most studies that have assessed the
effect of improved TB/HIV integration have evaluated the effect of
Intervention arm

improved co-location of TB and HIV services, comparing vertical ser-

vices or with the same services before and after implementation. Some
of these studies reported more rapid ART start and higher median CD4
1540

421

counts at ART initiations among HIV positive individuals diagnosed

N

with TB [7–10] while others reported no effect on ART uptake [5,

12 months among HIV-positive TB patients.

24–26]. In addition, with respect to TB treatment outcomes among HIV

individuals newly diagnosed with HIV

preventive therapy (CPT) at enrolment.

Incidence of morbidityor mortality among

Incidence of morbidityor mortality among

individuals newly diagnosed with TBc

Adjusted for randomization strata.

positive individuals, some studies found reduced mortality and better

retention [9, 10, 27, 28], where others did not [10, 25, 26, 29–31].
Complex health systems interventions to improve TB/HIV integration
have also been evaluated in trials. The PALSA PLUS trial evaluated the
effect of a non-didactic, outreach- and case-based training on HIV,
CI – confidence interval.

sexually transmitted infections, and TB for nurses. This intervention

increased TB case detection and CPT uptake among eligible individuals
Primary outcome

newly diagnosed with HIV but did not reduce mortality [32]. The
STRETCH trial added task shifting of ART initiation, further ART
training, introduction of ART initiation algorithms and management
Table 3

support to the PALSA PLUS training; this did not reduce mortality
d
b
a

e
c

among patients eligible but not started ART at enrolment or improve

49
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

Table 4
Effect of the intervention among subgroups of sex, socio-economic position (SEP) index, enrolment period and baseline CD4 on incidence of hospitalizations and
mortality among individuals newly diagnosed with HIV.
Intervention arm Control arm Adjusted rate p-Valuea Adjusted rate p-Valueb
ratioa ratiob
N (Hospitalisation or deaths/person-years of N (Hospitalisation or deaths/person-years of
follow up) Incidence per 100 person-years follow up) Incidence per 100 person-years

Sex:
Male 608 (82/571) 14.4 530 (56/498)11.2 1.21 (0.81, 0.33 1.1 (0.76, 0.51
1.79) 1.71)
Female 931 (100/887) 11.3 953 (91/909)10.0 1.15 (0.83, 0.37 1.05 (0.76, 0.74
1.58) 1.46)
SEP index:
Low 982 (119/926) 12.8 646 (64/620) 10.3 1.11 (0.69, 0.64 1.02 (0.65, 0.91
1.79) 1.61)
High 558 (63/533) 11.8 838 (83/788) 10.5 1.03 (0.61, 0.91 0.94 (0.53, 0.82
1.73) 1.66)
Enrolment
periodc
Earlyd 358 (48/365) 13.2 281 (41/290) 14.1 N/A N/A
Late 1182 (134/1095) 12.2 1203 (106/1118) 9.5 1.37 (0.92, 0.11 1.26 (0.84, 0.24
2.03) 1.90)
e
Baseline CD4
< 350 1100 (147/1037) 14.2 1036 (110/988) 11.1 1.25 (0.97, 0.09 1.14 (0.88, 0.31
1.61) 1.48)
≥350 f
251 (9/253) 3.6 273 (20/263) 7.6 N/A N/A

N/A not applicable – analysis not conducted as too few outcomes at the clinic level.
a
Adjusted for randomization strata.
b
Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, employment status, Socioeconomic position index, on
antiretroviral therapy (ART) at enrolment, Cotrimoxazole preventive therapy (CPT) at enrolment and in TB cohort.
c
Early enrolment period defined as being enrolled before 1 September 2012 (≥ 12 months of intervention).
d
Four clusters in the control arm and three clusters in the control arm have < 10 participants in the early enrolment period. Analysis is not possible for this
stratum.
e
CD4 count measured within 90 days of enrolment. 12% (364/3014) of participants have a missing CD4 count (175/1484 and 189/1540) in the control and
interventions arms, respectively).
f
Two clusters in the intervention and control arms (4/18) have zero outcomes. Analysis was not undertaken for this stratum.

12 month viral suppression rates among those on ART for at least six reasonable number of clinics. Secondly, because we assumed that the
months at enrolment [32]. delivery of TB/HIV services was already integrated, to a variable de-
Our study had a number of important strengths. Firstly, we used a gree, in most clinics, we designed an intervention, which was pragmatic
clustered randomised study design with a large sample size and a and allowed the activities for optimising integrated TB/HIV care to vary

Table 5
Effect of the intervention on secondary outcomes of CD4 counts at enrolment, time to starting ART among ART eligible individuals, proportion started on Isoniazid
preventive therapy (IPT) by the end of 12 months follow up and proportion who successfully completed TB treatment during 12 months of follow up.
Intervention arm Control arm Adjusted mean difference p-valueb Adjusted mean difference p-value
(Overall mean, N) (Overall mean, N) (control-intervention)a (intervention-control)c

Square root of CD4 counts at enrolment 14.0, 1315 14.4, 1309 −0.28 (−1.14, 0.58)a 0.50 0.14 (−0.64, 0.92) 0.72
among participants newly diagnosed
with HIVa

n/N, (%) n/N, (%) Adjusted RR p-value Adjusted RR p-value

d
Proportion of individuals newly diagnosed with HIV,ART eligiblewho initiated ART 597/891 (67) 626/889 0.90 (0.75–1.10) 0.30 0.91 (0.71–1.10) 0.30
by the end of 10 weeks after CD4 testing (70.4)
Proportion ofHIV positive individuals newly diagnosed with TB who started ART by 88/224 (39.3) 62/160 (38.8) 1.05(0.67–1.63) 0.83 0.99 (0.64–1.54)d 0.93
10 weeks after HIV testing
Proportion of individuals newly diagnosed with HIV who started IPT by 12 months 503/1264 552/1295 0.94(0.64–1.33) 0.71 0.96 (0.68–1.35)e 0.79
after HIV testing (39.8) (42.6)
f
Proportion of individuals newly diagnosed with TBwho successfully completed TB 323/421 243/310 0.94 (0.83–1.06) 0.28 0.93(0.82–1.05) 0.23
treatment during 12 months of follow up (76.7) (78.4)
a
Untransformed: for the intervention arm overall arithmetic mean is 230.8 (and arithmetic mean of cluster means is 231.5); for the control arm overall arithmetic
mean is 245.7 (and arithmetic mean of cluster means is 242.0).
b
Adjusted for randomization strata.
c
Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, employment status, Socioeconomic position (SEP) level, on
antiretroviral therapy (ART)at enrolment, Cotrimoxazole preventive therapy(CPT)at enrolment and in TB cohort.
d
Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, employment status, SEP level, CPT at enrolment.
e
Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, employment status, SEP level, on ART at enrolment, CPT at
enrolment.
f
Adjusted for randomization strata, sex, age group, country of birth, education level, marital status, SEP level, CPT at enrolment and being in the HIV cohort.

50
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

slightly depending on clinic needs. Thirdly, the trial enrolled in- the test and treat strategy which saw all HIV positive individuals being
dividuals newly diagnosed with TB, as well as individuals newly diag- eligible for ART regardless of CD4 count. [16] The impact of this
nosed with HIV, and would have been able to evaluate the potential strategy will be limited by how early HIV positive individuals are
impact of the intervention on outcomes specific to individuals in both identified and initiated on [Link], there is a need to scale up in-
groups. Our trial showed that individuals newly diagnosed with HIV terventions that promote the scale-up of IPT in order to prevent TB
alone did not have worse outcomes where TB/HIV integration was among HIV positive individuals. In the trial, there were low levels of
being actively promoted [12]. IPT initiation in 12 months (42% of eligible) following HIV testing.
A number of factors may explain our intervention's apparent lack of In conclusion, the trial intervention did not show an effect on pa-
effect on both improved TB/HIV integration and patient relevant out- tient relevant outcomes, because it had an insufficient effect on in-
comes. These factors are related to: i) limited implementation of the tegration of [Link] strategies are needed to achieve closer
intervention as intended, ii) differences between arms and, iii) lack of integration of TB and HIV care, taking resource constraints and existing
statistical power. organisational culture into account, along with evaluation of the effect
From the facility assessments at the end of the intervention period on patient-relevant outcomes.
and process measures (secondary outcomes), the extent to which in-
tegrated care was delivered at intervention clinics was not significantly Acknowledgements
different from that at the control clinics. More robust measurements of
TB/HIV integrated service delivery levels were not possible prior to, The authors would like to thank Chetna Kholi, Sandra Torosilva and
during or at the end of the trials as there were no validated tools or Don Mudzengi for assistance development of facility assessment tools
systems in [Link] intervention may also have failed to deliver the and collection of facility assessment data, Sarah Yates for assistance
expected result because the integration officers did not play the co- with data management, Violet Chihota and Dave Clark for operational
ordination and mentorship role envisaged but instead spent significant support.
amounts of time providing direct care to patients. The integration of-
ficers may have focussed on this aspect in order to better ‘fit in’ to the Author contribution
clinic working [Link] outsiders, the IO may have faced some re-
sistance to changing practice given the prevalent organisational culture TK, KLF, GJC, ADG, SC designed the study.
within clinics, and hence spent more time on being accepted by con- KK and AV collected the data on the process evaluations including
tributing to clinic work rather than pushing for change. This can also be time use (fidelity to intervention).
partly explained in the context of staff shortages as well as lack of staff TK, PH, SC collected the data on the primary and secondary out-
training and motivation in both intervention and control clinics that comes.
limited the extent to which integrated care at intervention clinics could TK, KLF and PH analysed the data.
be improved [33, 34]. TK and KLF drafted the manuscript.
As standard of care evolved over time, control clinics may have TK, KLF, PH, KK, AV, GJC, ADG, SC critically reviewed the manu-
advanced in the delivery of integrated TB/HIV services more than ex- script for content.
pected, which may have diluted the effect of the [Link] IOs TK, KLF, PH, KK, AV, GJC, ADG, SC approved the manuscript for
and SOs presence in the clinics may also have improved record keeping submission.
in the intervention clinics leading to better ascertainment of hospitali-
zations and other outcomes, although this was unlikely to be a major Competing interests
contributory factor as there were statistically differences in follow up
by arm. In addition, participants enrolled into the evaluation cohorts at The authors have no competing interests to declare.
control clinics were more educated, more likely to be employed, less
likely to be in the lower third of the socio-economic position index and Funding source
more likely to have been on ART at enrolment compared to those at the
intervention clinics. On the other hand, intervention clinics may have This study was supported by the President's Emergency Plan for
been able to improve linkage into care for the very ill as indicated by AIDS Relief (PEPFAR) through the Centers for Disease Control and
the higher proportion of HIV positive individuals newly diagnosed with Prevention (CDC) under the terms of [Cooperative agreement
TB in that arm. The adjustments for cluster level imbalance at analysis 5U2GPS000811]. Its contents are solely the responsibility of the au-
are likely to have minimised the effects of these imbalances. thors and do not necessarily represent the official views of CDC.
The trial was also unable to meet the desired sample size and
median duration of follow up (expected 15 months follow up) for both Appendix A. Supplementary data
cohorts and therefore may not have had adequate power to detect the
large differences (40% reduction) between the arms assumed at sample Supplementary data to this article can be found online at https://
size calculations. However, the small effect sizes for all the primary [Link]/10.1016/[Link].2018.07.013.
outcomes and their consistency across a range of sensitivity analyses
suggest that that no important differences between the arms would References
have been observed had the trial been adequately powered.
Despite these limitations, this trial contributed some valuable les- [1] O. Shisana, T. Rehle, L.C. Simbayi, K. Zuma, S. Jooste, N. Zungu, D. Labadarios,
sons about the delivery of integrated TB/HIV services and patient D. Onoya, et al., South African National HIV Prevalence, Incidence and Behaviour
Survey, 2012, Cape Town (2014).
outcomes. Firstly, that task shifting for TB screening is feasible at pri- [2] World Health Organisation. Global Tuberculosis Report 2017. Geneva, Switzerland.
mary care facilities. Secondly, there is a need to strengthen HIV testing WHO/HTM/TB/2017.23.
and linkage into HIV care at higher CD4 counts. This is because re- [3] World Health Organisation, WHO policy on collaborative TB/HIV activities,
Guidelines for national programmes and other stakeholders. Geneva, Switzerland,
gardless of arm, there were low CD4 counts at HIV testing and linkage 2012.
into care; 48.5% of individuals in the HIV cohort had CD4 count < [4] World Health Organisation. WHO Three I's meeting: Intensified Case Finding (ICF),
200 cells/ml and 69% had initiated ART 10 weeks after CD4 count Isoniazid Preventive Therapy (IPT) and TB Infection Control (IC) for people living
with HIV. 2–4 April 2008. Geneva, Switzerland 2008.
testing) despite a national HIV testing campaigns and increased [5] G.A. Ansa, J.D. Walley, K. Siddiqi, X. Wei, Delivering TB/HIV services in Ghana: a
thresholds for ART initiation both introduced during the intervention comparative study of service delivery models, Trans. R. Soc. Trop. Med. Hyg. (2014
period. In September 2016, the South African government introduced Jul 24) tru110.

51
T. Kufa et al. Contemporary Clinical Trials 72 (2018) 43–52

[6] H. Legido-Quigley, C.M. Montgomery, P. Khan, R. Atun, A. Fakoya, H. Getahun, cluster-randomized trials of interventions against infectious diseases, Stat. Methods
A.D. Grant, Integrating tuberculosis and HIV services in low- and middle-income Med. Res. 9 (2) (2000) 95–116.
countries: a systematic review, Tropical Med. Int. Health 18 (2013) 199–211. [22] R.J. Hayes, S. Bennett, Simple sample size calculation for cluster-randomized trials,
[7] B. Kerschberger, K. Hilderbrand, A.M. Boulle, et al., The effect of complete in- Int. J. Epidemiol. 28 (2) (1999) 319–326.
tegration of HIV and TB services on time to initiation of antiretroviral therapy: a [23] S. Vyas, L. Kumaranayake, Constructing socio-economic status indices: how to use
before-after study, PLoS One 7 (10) (2012) e46988. principal components analysis, Health Policy Plan. 21 (6) (2006) 459–468.
[8] P. Owiti, R. Zachariah, K. Bissell, et al., Integrating tuberculosis and HIV services in [24] Statistics South Africa, Community Survey Household Questionairre, Available
rural Kenya: uptake and outcomes, Pub. Health Action 5 (1) (2015) 36–44. from, 2007. [Link]
[9] S.M. Hermans, B. Castelnuovo, C. Katabira, et al., Integration of HIV and TB services [25] R. Kaplan, J. Caldwell, L.G. Bekker, et al., Integration of TB and ART services fails to
results in improved TB treatment outcomes and earlier prioritized ART initiation in improve TB treatment outcomes: comparison of ART/TB primary healthcare ser-
a large urban HIV clinic in Uganda, J. Acquir. Immune Defic. Syndr. 60 (2) (2012) vices in Cape Town, South Africa, S. Afr. Med. J. 104 (3) (2014) 204–209.
e29–e35. [26] A.B. Schwartz, N. Tamuhla, A.P. Steenhoff, et al., Outcomes in HIV-infected adults
[10] J.M. Ikeda, C.A.L. Tellez, E.S. Hudes, et al., Impact of integrating HIV and TB care with tuberculosis at clinics with and without co-located HIV clinics in Botswana,
and treatment in a regional tuberculosis hospital in rural Guatemala, AIDS Behav. Int. J. Tubercul. Lung Dis. 17 (10) (2013) 1298–1303.
18 (1) (2014) 96–103. [27] T.D. Ledibane, S.C. Motlhanke, A. Rose, W.H. Kruger, N.R. Ledibane,
[11] L.V. Adams, E.A. Talbot, K. Odato, H. Blunt, K.R. Steingart, Interventions to im- M.M. Claassens, Antiretroviral treatment among co-infected tuberculosis patients in
prove delivery of isoniazid preventive therapy: an overview of systematic reviews, integrated and non-integrated facilities, Pub. Health Action 5 (2) (2015) 112–115.
BMC Infect. Dis. 14 (2014) 281. [28] S.A. Schulz, H.R. Draper, P. Naidoo, A comparative study of tuberculosis patients
[12] A.B. Suthar, G.W. Rutherford, T. Horvath, M.C. Doherty, E.K. Negussie, Improving initiated on ART and receiving different models of TB/HIV care, Int. J. Tubercul.
antiretroviral therapy scale-up and effectiveness through service integration and Lung Dis. 17 (12) (2013) 1558–1563.
decentralization, AIDS (London, England) 28 (Suppl. 2) (2014) S175–S185. [29] J. Uyei, D. Coetzee, J. Macinko, S.L. Weinberg, S. Guttmacher, The influence of
[13] S. Sweeney, C.D. Obure, C.B. Maier, R. Greener, K. Dehne, A. Vassall, Costs and integrated tuberculosis and human immunodeficiency virus service delivery on
efficiency of integrating HIV/AIDS services with other health services: a systematic patient outcomes, Int. J. Tubercul. Lung Dis. 18 (3) (2014) 315–321.
review of evidence and experience, Sex. Transm. Infect. 88 (2) (2012) 85–99. [30] A. Ndagijimana, E. Rugigana, C.B. Uwizeye, J. Ntaganira, One-stop TB/HIV services
[14] T. Kufa, P. Hippner, S. Charalambous, et al., A cluster randomised trial to evaluate evaluation in Rwanda: comparison of the 2001–2005 and 2006–2010 cohorts, Pub.
the effect of optimising TB/HIV integration on patient level outcomes: The “merge” Health Action 5 (4) (2015) 209–213.
trial protocol, Contemp. Clin. Trials 39 (2) (2014) 280–287. [31] M.D. Nglazi, R. Kaplan, J. Caldwell, et al., Antiretroviral treatment uptake in pa-
[15] J.C. Chehab, P. Vranken, A. Peters, J.D. Klausner, Current integration of tubercu- tients with HIV-associated TB attending co-located TB and ART services, S. Afr.
losis (TB) and HIV services in South Africa, 2011, PLoS One 8 (3) (2013) e5779116. Med. J. 102 (12) (2012) 936–939.
[16] Department of Health, Republic of South Africa, The South African antiretroviral [32] M. Zwarenstein, L.R. Fairall, C. Lombard, et al., Outreach education for integration
treatment guidelines, Pretoria South Africa, 2010. of HIV/AIDS care, antiretroviral treatment, and tuberculosis care in primary care
[17] Department of Health, Republic of South Africa, Accelerating Access to ART clinics in South Africa: PALSA PLUS pragmatic cluster randomised trial, BMJ 342
Services and Uptake (Update on Guidelines) [Press Release], (14 April 2012). (2011) d2022.
[18] Department of Health, Republic of South Africa, Implementation of the Universal [33] L. Fairall, M.O. Bachmann, C. Lombard, et al., Task shifting of antiretroviral
Test and Treat Strategy for HIV Positive Patients and Differentiated Care for Stable treatment from doctors to primary-care nurses in South Africa (STRETCH): a
Patients (Update on Guidelines), Pretoria, South Africa, 2016. pragmatic, parallel, cluster-randomised trial, Lancet 380 (9845) (2012) (889–9).
[19] Department of Health, Republic of South Africa, A Practical Guide for TB and HIV [34] J. Uwimana, D. Jackson, H. Hausler, C. Zarowsky, Health system barriers to im-
Service Integration at Primary Health Care Facilities, Pretoria, South Africa, 2011. plementation of collaborative TB and HIV activities including prevention of mother
[20] R.J. Hayes, L.H. Moulton, Cluster Randomised Trials, Chapman & Hall/CRC Taylor to child transmission in South Africa, Tropical Med. Int. Health 17 (5) (2012)
Francis Group, 2009. 658–665.
[21] R.J. Hayes, N.D. Alexander, S. Bennett, S.N. Cousens, Design and analysis issues in

52