SEARCH

SEARCH BY CITATION

Keywords:

  • HIV/AIDS;
  • antiretroviral;
  • task shifting;
  • cost-effectiveness;
  • economic evaluation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Objective

To estimate the cost-effectiveness of nurse-led versus doctor-led antiretroviral treatment (ART) for HIV-infected people.

Design

Cost-effectiveness analysis alongside a pragmatic cluster randomised controlled trial in 31 primary care clinics (16 intervention, 15 controls) in Free State Province, South Africa. Participants were HIV-infected patients, aged ≥16 years. Cohort 1 (CD4 count ≤350 cells/μl, not yet receiving ART at enrolment): consisted of 5 390 intervention patients and 3 862 controls; Cohort 2 (already received ART for ≥6 months at enrolment) of 3 029 intervention patients and 3 202 controls. Nurses were authorised and trained to initiate and represcribe ART. Management and ART provision were decentralised to primary care clinics. In control clinics, doctors initiated and re-prescribed ART, nurses monitored ART. Main outcome measure(s) were health service costs, death (cohort 1) and undetectable viral load (<400 copies/ml) (cohort 2) during the 12 months after enrolment.

Results

For Cohort 1, the intervention had an estimated incremental cost of US$102.52, an incremental effect of 0.42% fewer deaths and an incremental cost-effectiveness ratio (ICER) of US$24 500 per death averted. For Cohort 2, the intervention had an estimated incremental cost of US$59.48, an incremental effect of 0.47% more undetectable viral loads and an ICER of US$12 584 per undetectable viral load.

Conclusions

Nurse-led ART was associated with higher mean health service costs than doctor-led care, with small effects on primary outcomes, and a high associated level of uncertainty. Given this, and the shortage of doctors, further implementation of nurse-led ART should be considered, although this may increase health service costs.

Objectif

Estimer le rapport coût-efficacité du traitement antirétroviral (ART) dirigé par un(e) infirmier(e) ou par un médecin pour les personnes infectées par le VIH.

Conception

Analyse coût-efficacité dans le cadre d'un essai contrôlé randomisé pragmatique dans 31 cliniques de soins primaires (16 interventions, 15 témoins) dans la province Free State, en Afrique du Sud. Les participants étaient des patients infectés par le VIH, âgés de ≥16 ans. Cohorte 1 (taux de CD4 ≤350 cellules/μl, ne recevant pas encore l’ART à l'inscription): composée de 5390 patients avec intervention et 3862 témoins; Cohorte 2 (ayant déjà reçu l’ART pour ≥6 mois lors de l'inscription), composée de 3029 patients avec intervention et 3202 témoins. Les infirmier(e)s ont été autorisés et formés pour initier et re-prescrire l’ART. La gestion et la fourniture de l’ART ont été décentralisées vers les cliniques de soins primaires. Dans les cliniques témoins, des médecins ont initié et re-prescrits l’ART, les infirmier(e)s ont surveillé le traitement. Les principaux critères mesurés étaient les coûts des services de santé, les décès (cohorte 1) et la charge virale indétectable (<400 copies/ml) (cohorte 2) au cours des 12 mois suivant l'inscription.

Résultats

Pour la cohorte 1, l'intervention a eu un coût estimatif supplémentaire de 102,52 $US, un effet supplémentaire de 0,42% de décès en moins et un rapport coût-efficacité différentiel (RCED) de 24500 $US par décès évité. Pour la cohorte 2, l'intervention a eu un coût estimatif supplémentaire de 59,48 $US, un effet supplémentaire de 0,47% de plus de charge virale indétectable et un RCED de 12584 $US par charge virale indétectable.

Conclusions

L’ART dirigé par les infirmier(e)s a été associé à une moyenne plus élevée des coûts des services de santé que lorsque dirigé par un médecin, avec de faibles effets sur les résultats primaires et un niveau élevé d'incertitudes. Compte tenu de cela et de la pénurie en médecins, la mise en œuvre ultérieure d’ART dirigé par des infirmier(e)s devrait être prise en considération, même si cela pourrait élever les coûts des services de santé.

Objetivo

Hacer un análisis de coste-efectividad para el tratamiento antirretroviral (TAR) dirigido por una enfermera versus el TAR dirigido por un médico con pacientes infectados con VIH.

Diseño

Análisis de coste-efectividad dentro de un ensayo clínico pragmático, controlado y aleatorizado por conglomerados, en 31 centros de atención primaria (16 intervención, 15 controles) de la Provincia del Estado Libre, Sudáfrica. Los participantes eran pacientes infectados con VIH, con edades ≥16 años. La Cohorte 1 (conteo CD4 ≤350 células/μl, aún no recibían TAR en el momento de comenzar el estudio) consistía en 5,390 pacientes recibiendo intervención y 3,862 controles; La Cohorte 2 (ya recibía TAR ≥6 meses al momento de comenzar el estudio) estaba conformada por 3,029 pacientes recibiendo intervención y 3,202 controles. Las enfermeras estaba autorizadas y entrenadas para comenzar y prescribir de nuevo el TAR. La gestión y la provisión del TAR estaban descentralizadas en los centros de atención primaria. En las clínicas control, los médicos iniciaban y prescribían de nuevo el TAR, las enfermeras monitorizaban el TAR. Los principales resultados a evaluar eran los costes de los servicios sanitarios, la muerte (cohorte 1) y la carga viral no detectable (<400 copias/ml) (cohorte 2) durante los 12 meses siguientes a la inclusión en el estudio.

Resultados

Para la Cohorte 1 la intervención tenía un coste incremental estimado de US$102.52, un efecto incremental de 0.42% menos muertes, y la relación de coste incremental (RCI) de US$24,500 por muerte evitada. Para la Cohorte 2 la intervención tenía un coste incremental calculado de US$59.48, un efecto incremental de 0.47% más cargas virales no detectables y un RCI de US$12,584 por carga viral no detectable.

Conclusiones

El TAR dirigido por enfermeras estaba asociado a unos costes promedio de los servicios sanitarios más altos que los cuidados ofrecidos por un médico, con pocos efectos sobre los resultados primarios y un nivel alto de incertidumbre asociado. Por ello, y dada la falta de médicos, debería considerarse el implementar más servicios de TAR dirigidos por enfermeras, aunque ello implique un incremento en el coste de los servicios sanitarios.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Large-scale implementation of antiretroviral treatment (ART) in sub-Saharan Africa has been hampered by scarcity of finance and appropriately trained health workers (Joint United Nations Programme on HIV/AIDS (UNAIDS) 2010). Task shifting of ART from doctors to other health workers is increasingly used to expand ART provision, but requires appropriate training, evaluation and support (World Health Organisation 2007). South Africa has the world's largest HIV burden (Joint United Nations Programme on HIV/AIDS (UNAIDS) 2010), and Free State is the province with the third highest prevalence (13% in 2008 (Lawn et al. 2005)). Although in 2008, an estimated 539 000 people in South Africa were receiving ART, another 809 000 eligible people were still not receiving it, and only 26% of the 109 000 eligible people in Free State were receiving it (Adam & Johnson 2009). Delays in initiating ART among eligible patients have led to high mortality rates among ART-eligible patients awaiting ART initiation (Cleary et al. 2006; Fairall et al. 2008a).

The Free State Department of Health (FSDoH) has worked with the authors since 2000 to develop and evaluate a method of educational outreach to primary care nurses to improve the quality of diagnosis of treatment for respiratory illnesses and HIV (Fairall et al. 2005; Zwarenstein et al. 2011). Since 2007, we have expanded the guidelines, training and managerial support to enable primary care nurses to initiate, monitor and represcribe ART, via a complex intervention called STRETCH (Streamlining Tasks and Roles to Expand Treatment and Care for HIV). As nurse prescribing of ART was not yet national policy, due to concerns about nurses' clinical competence and safety in providing ART, this intervention was implemented and evaluated as a randomised trial (see protocol (Fairall et al. 2008b) and clinical analyses (Fairall et al. 2012)). Here, we use data from this trial to compare the cost-effectiveness of doctor-led care for patients with HIV/AIDS to that of nurse-led care packaged as STRETCH.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Overview

The study was a cost-effectiveness evaluation alongside a cluster randomised controlled trial (see protocol (Fairall et al. 2008b) and clinical analyses (Fairall et al. 2012)), with individual level data on healthcare costs and outcomes, conducted from a public sector health service perspective. Trial registration: ISRCTN46836853.

Participants

All 31 primary care clinics providing ART in the Free State in 2007 were randomly allocated to be either intervention (16) or control (15) clinics. Two cohorts of HIV-infected patients aged ≥16 years and registered with the province's HIV/AIDS programme were enrolled between January 2008 and June 2009; follow-up continued until June 2010. Cohort 1 comprised patients with CD4 count ≤350 cells/μl and not yet receiving ART at enrolment (those with CD4 ≤ 200 cells/μl were eligible for ART according to national guidelines at the time of the trial (South African National Department of Health 2004a), and those with CD4 between 200 and 350 cells/μl were expected to become eligible during the trial). Our hypothesis was that STRETCH would expedite ART initiation, reducing mortality. Cohort 2 comprised patients who at enrolment had already received ART for >6 months. Our hypothesis was that STRETCH would not increase virological failure among cohort 2 patients on ART despite increased workloads due to ART. As more patients were eligible for Cohort 2 than the sample size estimate (Fairall et al. 2012) required, a random sample was enrolled.

Models of care

Intervention clinics implemented STRETCH, a complex educational and organizational health systems intervention, described in detail elsewhere (Uebel et al. 2011). Selected nurses in each clinic were authorised to prescribe first line ART drugs, and received additional training and guidelines covering ART prescribing, drug effects and side-effects, and criteria for referral to a doctor. These nurses prioritised assessment of eligibility and initiation of ART and monitored treatment once initiated, including represcriptions. Other intervention clinic nurses, who had received the same training as control clinic nurses, provided routine HIV care to patients not yet eligible for ART. Managerial decisions about ART were delegated to clinic managers who received support from the STRETCH coordinator (a doctor [KU]).

Control clinics followed usual practice. Nurses assessed and monitored HIV positive patients and referred them to doctors once they became eligible for ART (CD4 < 200 cells/μl). Doctors initiated ART, reviewed patients 3–6 monthly and repeated or modified ART prescriptions. Between doctor visits, patients attended monthly clinics, where nurses provided monitoring. Nurses continued to receive periodic training covering the management of HIV/AIDS, sexually transmitted infections and tuberculosis (Uebel et al. 2011), but stating that nurses should not prescribe ART.

Costs

For each patient, we estimated the health service cost incurred in the first 12 months post-enrolment, as described below. Costs are reported in 2009 US Dollars (US$), a conversion rate of US$1 = 8.42 South African rands (ZAR) was applied (the average exchange rate for 2009 (http://www.x-rates.com/d/ZAR/USD/hist2009.html, accessed 23 July 2012). No discounting (Drummond et al. 2005) was undertaken as costs were estimated over 12 months. The same methods were used for both Cohorts 1 and 2, except for ART costs (see below).

Set-up & implementation costs

Before nurse-led ART started, guideline material was developed, trainers of healthcare staff were trained, and programme roll-out was planned (set-up). Ongoing intervention support was also provided (implementation). Associated levels of resource use and unit costs were recorded. Total set-up and implementation costs were estimated and apportioned across all eligible patients (for either cohort) who attended the intervention clinics during the trial period.

Clinic visit costs

For each patient, the number of nurse and doctor visits was recorded. To estimate unit costs, fieldworkers timed doctor and nurse consultations 3 clinics per arm: one small (20 visits), one medium (30 visits) and one large (40 visits). Mean consultation times, salary and patient contact data were used to estimate the staff cost per consultation. Overhead costs, which include utilities (water and electricity) and non-clinical staff (administrative, cleaning and security personnel), were also included and based on previous data (Cleary et al. 2006) (here when clinical staff costs were added to overhead costs, overheads constituted 58.6% of the total).

ART and tuberculosis (TB) treatment costs

For ART drugs for Cohort 1, the date of ART initiation and regimen type was recorded. We assumed a prescription was subsequently provided every 28 days until the earliest of: (i) date of death (if applicable), (ii) 28 days after their last health service contact or (iii) 12 months post-enrolment. Unit costs of first (1a: stavudine, lamuvidine, efavirenz and 1b: stavudine, lamuvidine, nevirapine) and second (zidovudine, didanosine and lopinavir/ ritonavir) line ART regimens were obtained from FSDoH and inflated by 8% to reflect pharmacy handling costs (a standard fee applied throughout the country). All other ART regimens were assigned the weighted average cost of regimens 1a, 1b and 2. Patients were assumed not to switch ART regimens (switches within a year of initiation are very rare). For ART drugs for Cohort 2, we applied the same methods as for Cohort 1, although any switching of ART prescriptions was also monitored. The numbers of CD4 and viral load tests were recorded and multiplied by FSDoH unit costs.

TB treatment

We recorded TB treatment dates. Regime types and content were estimated from guidelines (World Health Organisation 2010) and combined with drug costs (from FSDoH). Informed by Sinanovic et al. (2003) each TB treatment clinic visit was assumed to last 5 min with a nurse.

Hospital admissions

Data on admissions and in-patient days were extracted from the FSDoH hospitalisation database. Missing lengths of stay were assumed to equate to the average for admissions with complete data (Irvine et al. 2010), in the same cohort. Costing was based on previously estimated (Cleary et al. 2006) unit costs inflated to 2009 levels using the South African consumer price index (Statistics South Africa 2010).

Total costs per patient, over the 12 months post-enrolment, were estimated by summing the costs of set-up and implementation (intervention patients only), clinic visits, ART and TB treatment and hospital admissions.

Effects

In Cohort 1, the primary outcome was death within 1 year of enrolment (Fairall et al. 2008b), compared between trial arms. Date of death was obtained either by linkage with the national register of death certification, using patients' national identity numbers or from patients' medical records.

In Cohort 2, the primary outcome was undetectable viral load (<400 copies/ml) one-year post-enrolment (Fairall et al. 2008b). Detectable viral load indicates treatment failure. Viral load tests should have been repeated 6 monthly during ART (South African National Department of Health 2004b) but in practice intervals between tests varied. We used the viral load closest to one year after enrolment and within 6 months of that date.

Cost-effectiveness analysis

As data were from a cluster randomised trial, with patients sampled and the intervention implemented at clinic level, the cost-effectiveness analyses took account of intraclinic correlation of costs and outcomes, and correlation between costs and effects (Nixon & Thompson 2005). We accounted for this intraclinic correlation of costs and outcomes using a two-stage bootstrap procedure: first, randomly sampling clusters with replacement and also, within each sampled cluster, resampling individuals with replacement (Bachmann et al. 2007; Fairall et al. 2010). Resampling was repeated 10 000 times, and each iteration provided an estimate difference in incremental costs and outcomes between trial arms. The percentiles of these incremental costs and outcomes' differences provided 95% confidence interval (95% CI) estimates. For each cohort, we estimated the incremental cost, incremental effect and incremental cost-effectiveness ratio (ICER). In cohort 1, the ICER was in terms of the incremental cost per death averted, and for cohort 2, we estimated the incremental cost per extra patient with an undetectable viral load. It should, however, be acknowledged that we are not aware of specific thresholds, on either of these scales, below which an intervention would be argued to constitute value for money in the South African healthcare system. Additionally, the cost-effectiveness acceptability curve (CEAC) (Barton et al. 2008; van Hout et al. 1994), which estimates the probability of each option being cost-effective at different levels of the cost-effectiveness threshold, was estimated for each cohort.

Sensitivity analysis

To test whether results were robust to assumptions about set-up and implementation costs, we conducted sensitivity analyses (Drummond et al. 2005) to reflect what might happen if the intervention was (i) rolled out to another province (reduced set-up costs) or (ii) continued in current intervention clinics (reduced implementation costs). Scenario (i) included all training and implementation costs, along with 50% of planning costs, but excluded guideline development costs. Scenario (ii) included 50% of implementation costs, set-up costs were excluded.

Ethical issues

The study was approved by the Research Ethics Committees of the Medical Schools of the University of Cape Town and the University of the Free State. Data confidentiality was maintained, and the research did not influence individual patient care. Informed consent was obtained from participating doctors, nurses and managers, but not from patients as the intervention was targeted at entire clinics and staff (not at individual patients), because data were obtained from anonymised medical records, and it was not feasible to obtain such large-scale consent without impairing access to care (Haines et al. 2000; Medical Research Council 2002). Blinding was not possible, but outcomes were assessed objectively using electronic databases.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Participants

For cohort 1, 5 390 patients were enrolled in the intervention arm and 3 862 in the control arm. For cohort 2, 3 029 of 8 202 eligible patients were enrolled in the intervention arm and 3 202 of 7 705 in the control arm. Details of exclusions and adverse events are reported elsewhere (Fairall et al. 2012). In each cohort, the intervention and control arms had similar baseline characteristics (see Table 1). Complete data were obtained for all cost variables and, in cohort 1, for the measure of effect (deaths). In cohort 2, complete 12-month follow-up viral load data were available for 85.3% of intervention patients and 83.0% of control patients.

Table 1. Patient baseline characteristics at enrolment
 Intervention groupControl group
Cohort 1
No. of patients53903862
Women3604 (67%)2681 (69%)
Median age (years) [IQR]36 [30–43]35 [29–42]
Median CD4 (cells per μL) [IQR]141 [70–201]137 [70–197]
Mean weight (kg) (SD)59 (14)58 (14)
Current tuberculosis301 (6%)200 (5%)
Cohort 2
No. of patients30293202
Women2113 (70%)2332 (73)
Median age (years) [IQR]38 [32–44]38 [32–45]
Median duration on ART (months) [IQR]13.9 [6.8–21.7]13.7 [7.3–22.3]
Viral load <400 copies per ml2378 (79%)2507 (78%)
Mean weight (kg) (SD)61 (13)62 (13)
Current tuberculosis241 (8%)186 (6%)

Costs

Total set-up and implementation costs amounted to US$330 986 (Table 2). As 14 872 eligible patients attended intervention clinics between January 2008 and June 2010, the mean set-up and implementation cost was US$22.26 per patient.

Table 2. Set-up and implementation costs
Set-up costs – component partsResources costedCost (US$)
  1. a

    15% of these costs were included as the ARV-related elements were undertaken in conjunction with the national roll-out of the PALSA (Fairall et al. 2005) programme.

Training of nurses as outreach trainers, training programme development and subsequent educational outreach trainingTrainer and trainee time, travel/material and venue costsa11 139
Guideline and toolkit developmentResearcher and clinician time, travel and material costs42 474
Implementation planning with health department and clinic staffResearcher, health department and clinic staff time, travel and venue costs121 354
Ongoing input by trial coordinatorResearcher time117 541
Follow-up outreach support visitsOut-reach trainer and clinic staff time, travel and venue costs8 232
Trainer's quarterly meetingsTrainer and out-reach trainer time and travel costsa3 910
Ongoing meetings with clinical staff/mangersOut-reach trainer and clinic staff time and travel costs26 336
 Total cost (base-case) (per eligible patient)330 986 (22.26)
Sensitivity analysis (i) reduced set-up and all implementation costs (total) (per eligible patient) 227 835 (15.32)
Sensitivity analysis (ii) reduced implementation costs (total) (per eligible patient)76 055 (5.11)
Clinic costs

Nurse and doctor visit rates are shown in Table 3. It can be seen that, on average, in both cohorts, intervention patients received more nurse visits than those in the control group. Similarly, in Cohort 1, intervention patients also received more doctor visits, whereas in Cohort 2, control patients tended to receive more doctor visits than those in the intervention group (Table 3). Nurse visits were estimated to be longer in the intervention arm (mean 10.52 min (based on 80 consultations)) than in the control arm (mean 5.62 min (58 consultations)), this also applied to doctor consultations (mean 14.07 min (10 consultations) versus 9.02 min (32 consultations)). Estimated unit costs per visit are shown in Table 4, and it can be seen that total nurse and doctor costs were higher for intervention patients in both cohorts (Table 3).

Table 3. : Estimated levels of resource use and health sector costs (in the first 12 months post-enrolment)
 Cohort 1Cohort 2
Mean number per patientMean total cost (US$)% of total cost difference Mean number per patientMean total cost (US$)% of total cost difference
Intervention clinicsControl clinicsIntervention clinicsControl clinicsIntervention clinicsControl clinicsIntervention clinicsControl clinics
Set-up and implementation$22.2621.9$22.2636.9
Nurse visits6.744.55$43.71$15.7827.57.515.50$48.70$19.0449.1
Doctor visits1.901.60$37.88$20.3917.21.231.45$24.38$18.539.7
All clinic visits8.646.15$81.58$36.1744.68.746.95$73.09$37.5858.8
ART prescriptions6.185.56$119.97$105.1214.613.7113.75$280.43$279.74-1.2
TB treatment (days)44.2429.15$32.20$21.2210.86.075.39$4.41$3.920.8
In-patient days0.950.90$110.45$104.376.00.380.38$44.46$44.50-0.1
CD4 test1.431.40$10.79$10.610.21.191.08$9.00$8.181.4
Viral load test0.610.56$23.06$21.072.01.511.41$48.46$46.453.3
Total (unadjusted) cost  $400.30$298.56100.0  $481.42$421.05100.0
Table 4. Unit costs for both cohorts (US$, 2009 prices)
 SourceUnit cost (US$)
  1. FSDoH = Free State Department of Health.

  2. a

    Taken from Cleary et al. (Cleary et al. 2006), inflated to 2009 cost levels using the using the average South African consumer price index (Statistics South Africa 2010).

Nurse visit (intervention clinic)Estimated within study6.49
Nurse visit (control clinic)Estimated within study3.46
Doctor visit (intervention clinic)Estimated within study19.89
Doctor visit (control clinic)Estimated within study12.76
ART prescription regimen 1aFSDoH23.58
regimen 1bFSDoH11.78
regimen 2FSDoH89.07
(weighted average)FSDoH19.52

TB treatment (weighted average drug cost per day)

(weighted average visit cost per day)

FSDoH0.33
Assumption informed by Sinanovic et al. (Sinanovic et al. 2003) and above nurse visit costs0.40
Cost per in-patient day of a hospital admissionExternal dataa115.96
CD4 testFSDoH7.55
Viral load testFSDoH37.76
Treatment costs

In Cohort 1, 65.5% of intervention patients were initiated on ART (65.3% received regimen 1a, 33.0% regimen 1b), compared with 59.5% of control patients (61.0% regimen 1a, 36.0% regimen 1b). Mean ART prescription costs were consequently higher in the intervention arm (Table 3). In Cohort 2, ART prescription numbers and costs were similar in the intervention arm (at enrolment, 61.5% received regimen 1a, 33.7% regimen 1b) and control arm (64.2% regimen 1a, 31.8% regimen 1b) (Table 3). In Cohort 1, 18.3% of the intervention arm and 12.5% of the control arm started TB treatment. In Cohort 2, the rates were 3.2% and 2.5%, respectively. Mean TB treatment costs were consequently higher for intervention patients (Table 3).

In Cohort 1, 13.0% of both intervention and control patients were admitted to hospital, mean lengths of stay were 7.4 and 7.0 days, respectively. In Cohort 2, 5.2% of intervention patients were admitted, compared with 4.3% in the control arm, mean lengths of stay were 7.4 and 9.2 days, respectively. In-patient costs were similar in both arms (Table 3).

For both cohorts estimated CD4 and viral load test costs were slightly higher in the intervention arm. In both cohorts, total costs were higher in the intervention arm, with clinic visit and set-up and implementation costs accounting for most of the difference (Table 3).

Effects

In Cohort 1, 16.5% of intervention patients and 17.1% of control patients died within 12 months of enrolment. In Cohort 2, of those who completed the viral load test at 12 months, 83.5% (N = 2156) had an undetectable viral load in the intervention arm, whereas 83.9% (N = 2230) in the control arm did.

Cost-effectiveness

In Cohort 1, after taking account of clustering, the estimated incremental cost of the intervention was US$102.52 (95% CI –US$6.19 to US$162.05), with an incremental effect of 0.42% fewer deaths (95% CI 10.38% to –11.83%). The consequent ICER was US$24 499.77 per death averted. The CEAC showed that the estimated probability that the intervention was cost-effectiveness was <50% at all cost-effectiveness thresholds (Figure 1).

image

Figure 1. Plots of cost-effectiveness acceptability curve (CEAC) and incremental cost-effectiveness ratio (ICER) for the intervention group for both cohorts 1 (incremental cost per death averted) and 2 (incremental cost per undetectable viral load).

Download figure to PowerPoint

In Cohort 2, the estimated incremental cost was US$59.48 (95% CI US$12.84 to US$106.11), with an incremental effect of 0.47% more undetectable viral loads (95% CI –10.61% to 11.55%), giving an ICER of US$12 583.90 per undetectable viral load. According to the CEAC, the probability the intervention was cost-effective was <5% at all cost-effectiveness thresholds (Figure 1).

Scenario (i) (reduced set-up costs) (Table 2) resulted in an incremental cost of US$95.58 and an ICER of US$22 842.17 per death averted in Cohort 1 compared with US$52.54 and US$11 116.45 per undetectable viral load, respectively, in Cohort 2. Scenario (ii) (reduced implementation costs) (Table 2) resulted in an incremental cost of US$85.37 and an ICER of US$20 403.13 per death averted in Cohort 1 compared with US$41.36 and US$8 261.21 per undetectable viral load, respectively, in Cohort 2. The CEACs remained largely unchanged (not shown).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

The STRETCH intervention of nurse-led ART was associated with higher health service costs and a small statistically non-significant effect on mortality among patients awaiting ART and equivalent viral load suppression rates for patients already on treatment, compared with the current policy of doctor-led ART. Levels of uncertainty were high, as indicated by wide confidence intervals around the incremental costs and effects. Large-scale expansion of ART based on doctor-led care may, however, not be a practical option, given the scarcity of doctors in South Africa's public health system and Nurse Initiation and Monitoring of ART has become national policy since the trial was conducted.

The increased cost of nurse-led care could, however, be accompanied by increased quality of ART, as shown by other evidence from the randomised trial (Uebel et al. 2011). In Cohort 1, intervention patients were more likely to be diagnosed with TB, remain in the ART programme and to increase their CD4 counts (Fairall et al. 2012). In Cohort 2, intervention patients were more likely to switch ART regimens and had larger increases in CD4 count and weight (Fairall et al. 2012).

The higher mean costs for intervention patients are largely explained by more frequent clinic visits with longer consultations (Tables 3 and 4). Visit frequencies suggest that nurses partly substituted for doctors in cohort 2 patients. However, in Cohort 1, it seemed that nurses obtained more doctor visits for their patients, possibly due to better adherence to referral criteria. Finally, it should also be noted that in South Africa, there is also a shortage of spare capacity with regard to nurses, which meant that sometimes the implementation of STRETCH was hampered and/or resulted in increased workloads for other team members (Georgeu et al. 2012).

The main strength of this study is the detailed individual level data, with complete cost data on over 15 000 patients in a province-wide pragmatic randomised trial, providing original evidence that is generalisable to other public sector ART programmes in South Africa and other countries with severe HIV/AIDS epidemics and scarcity of doctors.

The main limitation is that the primary outcomes considered, and the 12-month follow-up period potentially underestimates the benefits of the intervention. Indeed, the main effectiveness analysis completed after a further six months of follow-up showed that intervention patients in Cohort 1 had fewer and shorter hospital admissions, which would have helped offset higher primary care visit costs (Fairall et al. 2012). Some of the assumptions we made could also be questioned. Although guideline development costs have previously been allocated over 6 years (Fairall et al. 2010), we apportioned them over the 2.42 years of the trial, because of more rapid changes in HIV/AIDS care. As such, the base-case ICER may be considered a conservative estimate.

We are aware of one other study that has previously compared nurse-led to doctor-led care for HIV patients with regard to cost-effectiveness (Long et al. in press). This study was also based in South Africa and enrolled patients already on ART for several months (≥11 months on ART (Long et al. in press) compared with 6 months for our cohort 2). In contrast to our results, it estimated that nurse-led care was both less costly and more effective (in terms of death and loss to follow-up) (Long et al. in press). That said, mean estimated 12-month costs under nurse-led care (US$492) (Long et al. in press) were similar to our estimate of US$481. Explanations for other differences may include a more stringent inclusion criteria, excluding patients with clinical problems or virological failure from referral for nurse-led care, and that the doctor-led care data were taken from one urban site in a wealthier province (Long et al. in press). Our results are more in line with a systematic review comparing primary care for other chronic diseases provided by doctors and nurses, which found that the care provided by nurses was generally as effective as doctors; however, nurses care tended not to be more cost-effective due to lengthier consultations, more tests and the costs of medical supervision (Laurant et al. 2004). Two other randomised trials, in South Africa (Sanne et al. 2010) and Uganda (Jaffar et al. 2009), compared doctor-led care with non-medical care after ART initiation and showed similar outcomes. However, those trials did not evaluate ART initiation and did not include economic evaluations.

Within the cost-effectiveness analysis, costs are considered in relation to per death averted (Cohort 1) and per undetectable viral load (Cohort 2). We are not aware of specific thresholds, on either of these scales, which have been argued to indicate a level of cost increase, in relation to these benefits, that would be acceptable to the South African health service. Thus, it is difficult to make judgments about cost-effectiveness when using such outcome measures. Further research might therefore estimate outcomes in terms of the quality-adjusted life year (Drummond et al. 2005) as guidance, for example (NICE 2008), is available as to what ICER might be considered cost-effective. Similarly, Shilcutt et al. (Shillcutt et al. 2008) have outlined what cost per disability-adjusted life year averted should be considered attractive in low-income countries.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Outcomes were similar in a nurse-led system, implying that appropriately trained nurses have the ability to monitor patients till they become eligible for ART and then initiate ART promptly (Cohort 1) and undertake longer-term ART monitoring (Cohort 2). In view of the lack of doctors, it seems inevitable that more patients will receive nurse-led care. Here, we provide some support for this shift as nurses were able to provide more contacts and spend more time with patients, they also achieved improvements in certain outcomes, for example TB detection, CD4 count and weight (Fairall et al. 2012). Based on our results, implementation of nurse-led ART may, however, increase healthcare costs.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

We would like to thank intervention clinic nurses who were willing to take on additional clinical responsibilities despite their tremendous workloads; doctors, clinic managers, pharmacists and pharmacy assistants at ART sites and primary care services; Pat Mayers and our STRETCH trainers; local area and district managers; district ARV coordinators; the Free State Department of Health's Pharmaceutical and Therapeutics Committee; Wendy Adolph, Elsie Pretorius and our team of data clerks; and senior managers from the department. We also wish to acknowledge the National Health Laboratory Service for providing electronic blood result data, members of our trial monitoring and steering committees, and the project team. Lastly, we wish to recognise the patients who participated, especially those who died before they could be started on ART. Funding for this trial was provided by the UK Medical Research Council, the Development Cooperation Ireland and the Canadian International Development Agency.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  • Adam MA & Johnson LF (2009) Estimation of adult antiretroviral treatment coverage in South Africa. South African Medical Journal 99, 661667.
  • Bachmann MO, Fairall L, Clark A & Mugford M (2007) Methods for analyzing cost effectiveness data from cluster randomized trials. Cost Effectiveness and Resource Allocation 5, 12.
  • Barton GR, Briggs AH & Fenwick EA (2008) Optimal cost-effective decisions: the role of the cost-effectiveness acceptability curve (CEAC), cost-effectiveness acceptability frontier (CEAF) and expected value of perfect information (EVPI). Value Health 11, 886897.
  • Cleary SM, McIntyre D & Boulle AM (2006) The cost-effectiveness of Antiretroviral Treatment in Khayelitsha, South Africa – a primary data analysis. Cost Effectiveness and Resource Allocation 4, 20.
  • Drummond MF, Sculpher MJ, Torrance GW, O'Brien BJ & Stoddart GL (2005). Methods for the Economic Evaluation of Health Care Programmes, 3rd edn. Oxford University Press, New York.
  • Fairall LR, Zwarenstein M, Bateman ED, et al. (2005) Effect of educational outreach to nurses on tuberculosis case detection and primary care of respiratory illness: pragmatic cluster randomised controlled trial. BMJ 331, 750754.
  • Fairall LR, Bachmann MO, Louwagie G, et al. (2008a) Effectiveness of antiretroviral treatment in a South African program: cohort study. Archives of Internal Medicine 168, 8693.
  • Fairall LR, Bachmann MO, Zwarenstein MF, et al. (2008b) Streamlining tasks and roles to expand treatment and care for HIV: randomised controlled trial protocol. Trials 9, 21.
  • Fairall L, Bachmann MO, Zwarenstein M, et al. (2010) Cost effectiveness of educational outreach to primary care nurses to improve respiratory care: economic evaluation alongside a randomised trial. Tropical medicine and international health 15, 277286.
  • Fairall L, Bachmann MO, Lombard C, et al. (2012) Task shifting of antiretroviral treatment from doctors to primary-care nurses in South Africa (STRETCH): a pragmatic, parallel, cluster-randomised trial. Lancet 380, 889898.
  • Georgeu D, Colvin CJ, Lewin S, et al. (2012) Implementing nurse-initiated and managed antiretroviral treatment (NIMART) in South Africa: a qualitative process evaluation of the STRETCH trial. Implementation Science 7, 66.
  • Haines A, Ashcroft R, Coggon D, et al. (2000). Personal Information in Medical Research (MRC Ethics Series). MRC, London.
  • van Hout BA, Al MJ, Gordon GS & Rutten FF (1994) Costs, effects and C/E-ratios alongside a clinical trial. Health Economics 3, 309319.
  • Irvine L, Conroy SP, Sach T, et al. (2010) Cost effectiveness of a day hospital falls prevention programme for screened community dwelling older people at high risk of falls. Age Aging 39, 710716.
  • Jaffar S, Amuron B, Foster S, et al. (2009) Rates of virological failure in patients treated in a home-based versus a facility-based HIV-care model in Jinja, southeast Uganda: a cluster-randomised equivalence trial. Lancet 374, 20802089.
  • Joint United Nations Programme on HIV/AIDS (UNAIDS). Global Report: UNAIDS report on the global AIDS epidemic Geneva, 2010. Available at http://www.unaids.org/documents/20101123_GlobalReport_em.pdf (Accessed 23 July 2012).
  • Laurant M, Reeves D, Hermens R, Braspenning J, Grol R & Sibbald B. (2004) Substitution of doctors by nurses in primary care. Cochrane Database Systematic Review 4, CD001271.
  • Lawn SD, Myer L, Orrell C, Bekker LG & Wood R (2005) Early mortality among adults accessing a community-based antiretroviral service in South Africa: implications for programme design. AIDS 19, 21412148.
  • Long L, Brennan A, Fox MP, et al. (in press) Treatment outcomes and cost-effectiveness of shifting management of stable ART patients to nurses in South Africa: An observational cohort. PLoS Med 8, e1001055.
  • Medical Research Council (2002). Cluster Randomised Trials – Methodological and Ethical Considerations (MRC Clinical Trial Series). MRC, London.
  • NICE. Guide to the Methods of Technology Appraisal. National Institute of Health and Clinical Excellence (NICE) publications, London, 2008.
  • Nixon RM & Thompson SG (2005) Methods for incorporating covariate adjustment, subgroup analysis and between-centre differences into cost-effectiveness evaluations. Health Economics 14, 12171229.
  • Sanne I, Orrell C, Fox MP, et al. (2010) Nurse versus doctor management of HIV-infected patients receiving antiretroviral therapy (CIPRA-SA): a randomised non-inferiority trial. Lancet 376, 3340.
  • Shillcutt S, Morel C, Goodman C, et al. (2008) Cost-effectiveness of malaria diagnostic methods in sub-Saharan Africa in an era of combination therapy. Bulletin of the World Health Organization 86, 101110.
  • Sinanovic E, Floyd K, Dudley L, Azevedo V, Grant R & Maher D (2003) Cost and cost-effectiveness of community-based care for tuberculosis in Cape Town, South Africa. The international journal of tuberculosis and lung disease 7, S56S62.
  • South African National Department of Health. National antiretroviral treatment guidelines (1st Edition). Jacana (Minuteman Press) Department of Health, 2004a. http://www.doh.gov.za (Accessed 23 July 2012).
  • South African National Department of Health. National Antiretroviral Treatment Guidelines. Department of Health, 2004b. http://www.kznhealth.gov.za/arv/arv5.pdf (Accessed 2 March 2013).
  • Statistics South Africa. Consumer Price Index excluding interest rates on mortgage bonds (CPIX). 2010. http://www.statssa.gov.za/keyindicators/cpix.asp (Accessed 23 July 2012).
  • Uebel KU, Fairall LR, van Rensburg HCJ, et al. (2011) Task shifting and integration of HIV care into primary care in South Africa: the development and content of the streamlining tasks and roles to expand treatment and care of HIV (STRETCH) intervention. Implementation Science 6, 86.
  • World Health Organisation. Task Shifting to Tackle Health Worker Shortages. WHO, Geneva, 2007
  • World Health Organisation. Treatment of TB guidelines (4th edition). WHO Press, Geneva, 2010. www.who.int/tb (Accessed 23 July 2012).
  • Zwarenstein M, Fairall LR, Lombard C, et al. (2011) Outreach education integrates HIV/AIDS/ART and Tuberculosis care in South African primary care clinics: a pragmatic cluster randomized trial. BMJ 342, d2022.