Tropical Medicine and International Health volume 15 no 3 pp 277286 march 2010

doi:10.1111/j.1365-3156.2009.02455.x

Cost-effectiveness of educational outreach to primary care nurses to increase tuberculosis case detection and improve respiratory care: economic evaluation alongside a randomised trial
Lara Fairall1, Max O. Bachmann2, Merrick Zwarenstein3, Eric D. Bateman4, Louis W. Niessen5, Carl Lombard6, Bosielo Majara7, Rene English1, Angeni Bheekie8, Dingie van Rensburg9, Pat Mayers10, Annatjie Peters11 and Ronald Chapman11
1 Knowledge Translation Unit, University of Cape Town Lung Institute, Cape Town, South Africa 2 School of Medicine, Health Policy and Practice, University of East Anglia, Norwich, UK 3 Centre for Health Services Sciences, Sunnybrook Research Institute, Toronto, Canada 4 Department of Medicine, University of Cape Town, Cape Town, South Africa 5 Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA 6 Biostatistics Unit, Medical Research Council, Cape Town, South Africa 7 Department of Community Health, University of the Free State, Bloemfontein, South Africa 8 School of Pharmacy, University of the Western Cape, Cape Town, South Africa 9 Centre for Health Systems Research and Development, University of the Free State, Bloemfontein, South Africa 10 Division of Nursing and Midwifery, School of Health and Rehabilitation Sciences, University of Cape Town, Cape Town, South Africa 11 Free State Department of Health, Bloemfontein, South Africa

Summary

objective To evaluate the cost-effectiveness of an educational outreach intervention to improve primary respiratory care by South African nurses. methods Cost-effectiveness analysis alongside a pragmatic cluster randomised controlled trial, with individual patient data. The intervention, the Practical Approach to Lung Health in South Africa (PALSA), comprised educational outreach based on syndromic clinical practice guidelines for tuberculosis, asthma, chronic obstructive pulmonary disease, pneumonia and other respiratory diseases. The study included 1999 patients aged 15 or over with cough or difcult breathing, attending 40 primary care clinics staffed by nurses in the Free State province. They were interviewed at rst presentation, and 1856 (93%) were interviewed 3 months later. results The intervention increased the tuberculosis case detection rate by 2.2% and increased the proportion of patients appropriately managed (that is, diagnosed with tuberculosis or prescribed an inhaled corticosteroid for asthma or referred with indicators of severe disease) by 10%. It costs the health service $68 more for each extra patient diagnosed with tuberculosis and $15 more for every extra patient appropriately managed. Analyses were most sensitive to assumptions about how long training was effective for and to inclusion of household and tuberculosis treatment costs. conclusion This educational outreach method was more effective and more costly than usual training in improving tuberculosis, asthma and urgent respiratory care. The extra cost of increasing tuberculosis case detection was comparable to current costs of passive case detection. The syndromic approach increased cost-effectiveness by also improving care of other conditions. This educational intervention was sustainable, reaching thousands of health workers and hundreds of clinics since the trial. keywords tuberculosis, lung disease, primary care, education, health economics, randomised trial

Introduction The World Health Organization (WHO) has estimated that a fth of primary care consultations in developing countries are because of respiratory symptoms and that 2% of primary care patients have tuberculosis (WHO 2004). WHO and

developing country health services have increased efforts to improve tuberculosis case detection because of increasing tuberculosis incidence, aggravated by HIV and antibiotic resistance (Gandhi et al. 2006, WHO 2008a,b). They are also increasingly trying to integrate tuberculosis care with primary care for other chronic diseases.

2010 Blackwell Publishing Ltd

277

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

In South Africa, most patients with respiratory symptoms are treated in public sector primary care clinics by nurses who have little training in diagnosis and treatment of respiratory disease. Among such patients, tuberculosis is often not diagnosed (WHO 2008b), asthma is undertreated (Mash & Whittaker 1997; Stempel et al. 2004) and antibiotics are over-prescribed (Louwagie et al. 2002, WHO 2004). South Africas tuberculosis notication rates more than doubled between 2000 and 2007 (WHO 2009). Thus, an effective and sustainable method of training primary care nurses to diagnose and treat respiratory disease is needed. World Health Organizations Practical Approach to Lung Health (PAL) strategy aims to increase detection of active tuberculosis among adults with cough or difculty breathing, and also to improve the diagnosis, treatment and referral of other respiratory conditions, through integrated clinical management of patients presenting to rst-level facilities with respiratory symptoms (WHO 2005, 2008b). Similar syndromic approaches have been used for other diseases in developing countries, with some evidence of cost-effectiveness (Gilson et al. 1997; Adam et al. 2005; Shresta et al. 2006). The Practical Approach to Lung Health in South Africa (PALSA) combines locally adapted syndromic guidelines with evidence-based methods of educational outreach delivered by usual health department trainers (Bheekie et al. 2006; English et al. 2008). An important advantage over conventional classroom-based methods of educating professionals is that it is brief and does not remove scarce primary care staff from their workplaces. A pragmatic cluster randomised controlled trial conducted in 2003 showed that PALSA signicantly improved the quality of respiratory and tuberculosis care, increasing tuberculosis case detection, asthma treatment and appropriate urgent referral among patients attending clinics staffed by nurses in a resource-poor rural province of South Africa (Fairall et al. 2005). We have shown elsewhere that the guideline, when used by a nurse, was accurate in identifying patients at high risk of pulmonary tuberculosis (English et al. 2006). The present study evaluates PALSAs costeffectiveness. Methods We analysed cost-effectiveness alongside a pragmatic cluster randomised trial, using cost and outcome data about individual patients. The study had a societal perspective, considering the costs of illness to the health service and to the society for the period from the rst consultation until 3 months later. We included societal as well as health service perspectives, because costs of illness 278

and health care use can be considerable for patients, their households and society, especially for chronic disabling diseases such as tuberculosis and obstructive lung disease, and these costs could be inuenced by better care. Intervention and trial The Free State provinces 40 largest primary care clinics outside the capital were randomised to receive the PALSA intervention or to continue with usual training and support. The intervention was based on the PALSA syndromic guideline for management of adult respiratory diseases (Fairall et al. 2005; English et al. 2008), which classied patients into diagnostic and treatment categories according to their symptoms and signs. It covered tuberculosis, obstructive lung disease, acute upper and lower respiratory tract infections and HIV opportunistic infections. It was adapted from WHOs PAL guideline (WHO 2005, 2008b), after consultation with South African primary care physicians, nurses and managers, and harmonised with local guidelines such as the national essential drug list, HIV and tuberculosis programmes. The guideline was published in booklet and desk blotter formats (English et al. 2006). Primary care nurses were trained to use the PALSA guideline at their clinics, to minimise disruption of services. The programme trained eight nurse managers already employed by the health department, who then trained the nurses. Each trainer delivered three or four educational outreach sessions, each lasting 13 h, to all clinical staff in each intervention clinic over a 3-month period. Key messages as, With cough or difculty breathing for two weeks or more, think of tuberculosis were emphasised. The study recruited 1999 patients with cough or difcult breathing attending the trial clinics. After randomisation, the trial arms had equal numbers of clinics in each district, equal proportions of urban and rural clinics and similar numbers of tuberculosis patients per clinic treated during the previous year. Numbers of patients recruited per clinic ranged from 47 to 51 in intervention clinics and from 46 to 51 in control clinics. Patients in intervention and control clinics had similar baseline characteristics (Fairall et al. 2005). They were interviewed after their rst consultation and again 3 months later, so the studys follow-up period was 3 months. At follow-up, 1856 (93%) patients were interviewed. The trial took place between May and November 2003, implemented incrementally as training was rolled to each district. All 20 clinics randomised to receive the intervention did receive it, and therefore all 1000 trial patients in intervention clinics were potentially exposed to the intervention.

2010 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

Outcome measurements We dened two primary trial outcomes at the initial consultation or within 3 months: tuberculosis case detection and appropriate care. Tuberculosis case detection was dened as a new diagnosis of tuberculosis, conrmed by examination of patient-held tuberculosis treatment records. In keeping with national policy, tuberculosis could only be diagnosed if sputum smear tests identied tubercle bacilli or, for smear negative cases, if a medical doctor had made a clinical diagnosis, using all available clinical evidence including radiography and sputum culture results. Thus, all new tuberculosis diagnoses recorded were by denition true positive cases. Appropriate care was dened as (i) new tuberculosis diagnosis or (ii) treatment with inhaled corticosteroids for asthma or (iii) referral for a higher level of care in the presence of dened indicators of severe illness (respiratory rate 30 breaths per min, temperature 38 C, prominent use of accessory muscles,

or confusion). Asthma treatment and urgent referral were established by patient interview and examination of patients medical records. The interventions effects were the differences between arms in these outcomes. Cost of the educational intervention Costs of developing the guideline and training materials were estimated from project expenditure records, diary reviews and interviews with the developers (Table 1). Clinic staff training costs were obtained from attendance registers completed by trainers, and records of session duration, transport and other costs associated with training. The hours each person spent developing, delivering or receiving the intervention were multiplied by their respective hourly costs to their employers. The effects of training were assumed to last only for the 3 months of follow-up during the trial, so all training costs were allocated to this period.

Table 1 Cost of developing, delivering and receiving the training intervention Total cost of the PALSA intervention Category Guideline development Cost (2009 $US) 96 499 Resources used 11 months of junior and 2 months of senior researcher time, 1 month of health care worker time, 6 days WHO consultants time; cost of workshops including 10 local and 2 international ights, accommodation and venues; graphic layout 7 months of junior and 2 months of senior researcher time, 0.5 month of health care worker time; cost of pilot sessions including four local ights, accommodation and venues; illustrations and graphic layout 0.5 month junior and 0.5 month senior researcher time, 3 months health care worker time; 11 local ights, accommodation and road travel; support materials for trainers 60 educational outreach visits to 148 nurses in 20 clinics including 49 months of health care worker time, travel (6569 km) and materials

Training programme and materials development

54 901

Training the trainers

31 774

Educational outreach visits

15 940

Total cost of PALSA Cost of intervention per clinic attendance Category Guideline and material development Training trainers and outreach visits Total cost of intervention per attendance

199 113

Cost in 2009 $US 0.12 1.04 1.15

Comments Development costs of guideline and training programme were treated as capital costs, allocated over 6 years 100% allocated to intervention clinics assuming training was not used elsewhere or after the trial endpoint

PALSA, Practical Approach to Lung Health in South Africa. Development costs incurred before 2003 were inated to 2009 values using the average South African consumer price index. Costs were allocated to patient attendances using routine data on total clinic attendances and assuming 30% (51 797) of all 172 636 patients who attended trial clinics during the trial had cough or difcult breathing and could have beneted from the intervention.

2010 Blackwell Publishing Ltd

279

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

Resources used in developing the guideline and training materials were considered as a capital cost. Assuming that the guideline and training materials would remain valid for 6 years, based on other guideline research (Shekelle et al. 2001), we allocated 4.2% (3 months 6 12 months) of these capital costs to the 3 months of follow-up in each clinic during the trial. These guidelines were still in use in 2009 but had been expanded to cover other diseases. We assumed that 30% of all clinic patients during the 3-month trial period had respiratory symptoms and could have beneted from the intervention, based on our previous study in a similar setting (Fairall et al. 2001). Development and training costs per clinic attendance by a PALSAeligible patient were then estimated by dividing these total costs by 30% of the total number of clinic attendances recorded during the trial period (Table 1). Patients resource use, unit costs and total costs of resources Information on patients health care use was collected through an interviewer-administered questionnaire at baseline and 3 months later (Table 2). It elicited informa-

tion on numbers of clinic visits, investigations performed, drugs prescribed, numbers and duration of hospital admissions, ambulance transport, private sector health practitioners fees, patients transport costs and other episode-related out-of-pocket expenditure. Unit costs of resources were obtained from pharmaceutical services, laboratory tariffs, department of health tariffs and a literature review (Table 2). We conducted a work study of 433 consultations by adult patients with cough or difculty breathing with nurses in 10 intervention and 9 control clinics. This showed no difference in duration of consultations. To calculate the total cost of resources used, for each resource the unit cost was multiplied by the respective number of resources used. As the trial was conducted in 2003, earlier development costs were inated to 2003 values (Statistics South Africa 2009). Discounting of costs and outcomes after 2003 was unnecessary because the study was conned to 2003. Costs were converted from South African rands to United States dollars using the average exchange rate for the trial period ($1 = ZAR 7.35), then adjusted for ination to 2009 United States dollars (United States Department of Labor 2009).

Table 2 Use of health care resources and mean costs (in 2009 US$) per patient during 3 months of follow-up Intervention Resources Health service costs Public health service utilisation Clinic visits Outpatient visits Inpatient days Ambulance trips Medication Beta-agonists Inhaled corticosteroids Antibiotics (excluding TB drugs) Other drugs Investigations Sputum tests for TB Chest X-ray HIV rapid test kits PALSA intervention costs Total health service costs Household costs Fees paid to private providers Travel to the clinic Travel to other HCP Total household costs Total societal costs Unit cost ($) Mean no. units Mean cost ($) Control Mean no. units Mean cost ($) Source of unit costs

14.16 11.29 42.52 95.84

1.51 0.05 0.13 0.04 0.40 0.15 0.57 0.82

21.46 0.62 5.45 3.97 1.54 1.04 0.70 1.76 3.66 0.65 0.28 1.73 42.83 5.15 1.41 1.64 8.20 51.03

1.42 0.04 0.21 0.05 0.26 0.08 0.54 0.78 0.55 0.06 0.08

20.13 0.50 8.87 5.21 0.98 0.52 0.65 1.12 2.74 0.40 0.22 0.00 41.34 4.10 0.89 1.11 6.07 47.40

Daviaud (1999) Haile (2000) Cleary et al. (2004) FS DOH FS FS FS FS DOH DOH DOH DOH

5.02 7.25 2.84 1.15

0.73 0.09 0.10 1.51

FS DOH FS DOH FS DOH Table 1

FS DOH, Free State Department of Health; PALSA, Practical Approach to Lung Health in South Africa.

280

2010 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

Cost-effectiveness analysis This analysis used patient level data on costs and outcomes from the trial, described previously. We calculated incremental cost-effectiveness ratios (ICERs) for both primary outcomes (tuberculosis case detection and appropriate care) in an intention-to-treat analysis. The ICER represents the average extra cost for each additional patient who attains the respective, benecial, outcome. ICER point estimates were calculated by dividing differences in mean costs by differences in outcome probabilities. Statistical analysis of individual patient data was conducted as follows. Regression models were used to compare trial arms while accounting statistically for cluster randomisation within strata. The mean effects and cost differences were obtained from regression models, and ICER condence intervals and probabilities that the intervention was cost-effective were obtained from bootstrap analyses. Differences in mean costs between the two trial arms were estimated with linear regression. Differences in outcomes were estimated with binomial logistic regression. In all regression models, district as well as trial arm were included as explanatory variables, because randomisation was stratied by district. Patients characteristics were not included as model covariates because, owing to randomisation, they were not confounders (Fairall et al. 2005). Clinics were not included as covariates because they were completely collinear with arm. Regres-

sion analyses used HuberWhite robust adjustment for intra-clinic correlation. We used non-parametric bootstrap sampling to quantify statistical uncertainty and to account for intra-clinic clustering of costs and outcomes (Carpenter & Bithell 2000). We drew 2000 iterations per analysis, rst sampling clinics within arms and then sampling individuals within clinics, both with replacement (Bachmann et al. 2007). We used the bias-corrected accelerated percentile method to estimate ICER condence intervals (Carpenter & Bithell 2000). Bootstrap sampling resulted in scatter plots of differences in cost and outcomes, showing the uncertainty about effects and differences in mean cost per patient (Figures 1 and 2). We disregarded ICER condence intervals from analyses of tuberculosis case detection, because they included some samples worse with outcomes and higher costs as well as others with better outcomes and lower costs (Figure 1). In such situations, condence limits cannot be interpreted (Briggs 2001). We plotted cost-effectiveness acceptability curves, showing the probability (y-axis) that the intervention is costeffective for various acceptability thresholds (x-axis) (Briggs 2001). Briey, this analysis is based on the equation: nb = ek ) c where nb is the net benet of an intervention, e is the effect of the intervention, k is the amount society is willing to pay

15

Difference in mean cost ($US)

10

Figure 1 Cost-effectiveness plane with additional cases of tuberculosis diagnosed as effect.

10 0.02 0.01 0 0.01 0.02 0.03 0.04 0.05 Effect (difference in proportion diagnosed with tuberculosis) 0.06 0.07

2010 Blackwell Publishing Ltd

281

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

20

15

Difference in mean cost ($US)

10

10

15 0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

Effect (difference in proportion with appropriate care)

Figure 2 Cost-effectiveness plane with additional patients appropriately managed as effect.

for a unit of effect and c is the extra cost of the intervention. As k is not known in this context, a range of values are used. The intervention is cost-effective if ek > c. This is determined for each bootstrap sample, for a range of k values, and the probability that the intervention is cost-effective at each k values is the proportion of bootstrap samples for which ek > c. Statistical analyses were conducted with stata 9.0 (STATA Corp.,). We conducted several sensitivity analyses to test the robustness of our ndings. First, we halved training costs, reecting a scenario where the training was effective for 6 instead of 3 months. Secondly, for newly diagnosed tuberculosis cases, we included costs of tuberculosis treatment during the whole period of follow-up. Thirdly, we changed other key parameters to 50% and 150% of their assumed values in the base case analysis. These parameters included the proportions of all clinic patients assumed to have cough or difculty breathing, unit costs of health service resources and depreciation of the guideline and training materials. Ethics approval The study was approved by the research ethics committee of the Faculty of Health Sciences, University of the Free State.

Results Effects of the intervention The trial (Fairall et al. 2005) found that tuberculosis was newly diagnosed in 4.5% of intervention arm patients and 2.3% of control arm patients (risk difference 2.2%, 95% condence interval (CI) 0.44.1%; P = 0.03). Appropriate care was recorded in 20.8% of intervention arm patients and 10.8% of control arm patients (risk difference 10.0, 95% CI 4.915.0; P < 0.001). The 10.0% risk difference for appropriate care included risk differences of 2.2% for tuberculosis case detection, 3.1% (7.8% vs. 4.7%) for urgent referrals and or 6.0% (13.7% vs. 7.7%) for inhaled corticosteroid prescriptions. Costs of the intervention and of provision and utilisation of health care The total cost of the training intervention was $199 113 (Table 1). Of this, the cost of developing the guideline and materials was $151 400, and 4.2% of this was allocated to the 3 months of the trial. Training sessions cost $47 713, all of which was allocated to the 3 months of the trial. The training intervention thus cost an average of $1.15 per patient clinic visit (Table 1). Intervention group patients visited clinics on average 1.51 times during the trial, so the

282

2010 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

intervention alone cost on average $1.73 per intervention group patient (Table 2). In the primary analysis, the intervention was associated with a $1.49 (95% CI $6.38$9.37) higher mean health service cost and a $3.64 (95% CI $4.37$11.65) higher mean societal cost, compared with usual care. These differences were largely because of more clinic visits, tuberculosis tests, asthma medication and greater costs of travel and private providers, despite lower inpatient and ambulance costs among intervention arm patients (Table 2). When we included costs of tuberculosis treatment after diagnosis for the period of follow-up, the mean health service cost was $3.00 (95% CI $4.88$10.92) higher in the intervention arm than the control arm. When we allocated only half of training costs to the 3-month trial period, the mean health service cost was $0.73 (95% CI $7.10$8.54) higher in the intervention arm. When unit costs of each resource item were increased or decreased by 50%, the differences in mean cost (and thus ICERs) were most sensitive to changes in unit costs of inpatient days (cost differences increased or decreased by 58%), ambulance trips (by 26%) and sputum smear tests (by 15%), and by <10% for all other items. Cost-effectiveness of the intervention In the primary analysis, the intervention was associated with a $68 higher health service cost, and a $166 higher total societal cost, per extra tuberculosis case diagnosed (Table 3). It was associated with a $15 higher health service cost, and a $36 higher total societal cost, per extra patient appropriately managed. Cost-effectiveness estimates were sensitive to inclusion of tuberculosis treatment costs and assumptions about how long training was effective for (Table 3).

The cost-effectiveness acceptability curves show that, when increases in appropriate care were considered (Figure 3), the intervention appeared much more likely to be cost-effective at any willingness to pay threshold, compared to analyses with tuberculosis case nding as outcome (Figure 4). In the latter analyses, there was wide variation in the acceptability thresholds for which the intervention would be cost-effective (Figure 3). This reects the greater statistical uncertainty about the interventions effect on tuberculosis case nding alone (Figure 1), together with uncertainty about costs. Discussion The study shows that the PALSA educational outreach intervention costs more, and was more effective, than usual training and support. The extra cost of $68 per extra tuberculosis case detected is similar to the cost of passive case nding in South African clinics, previously estimated at $39 per case diagnosed in 1997 (Sinanovic et al. 2003), or $67 at 2009 prices (US Department of Labor 2009). The extra cost would therefore probably be acceptable to South African health services, although costs of enhanced case nding would be additional to current costs of passive case nding. It compares favourably with an earlier South African studys nding that enhanced case nding cost $166 per extra case detected (Hausler et al. 2006). The intervention is more cost-effective if one also considers its additional effects on asthma and severe respiratory illness. The cost-effectiveness of the intervention would be much better than we estimated if training effects lasted longer than the 3 months we conservatively assumed (Table 3) and reached more trainees. Other studies (Sanci et al. 2000; Morgenstern et al. 2003) have shown that educational outreach training can last longer than

Table 3 Sensitivity analysis: incremental cost-effectiveness ratios (ICERs) when different outcomes and costs are considered Outcome considered Costs considered Base health service cost Societal (health service and household) cost Health service cost allocating training costs over 6 instead of 3 months Health service cost including tuberculosis treatment Tuberculosis diagnosed ICER [% of base estimate] 67.89 [100] 165.11 [243] 32.61 [49] 137.11 [202] Appropriate care ICER (95% CI) [% of base estimate] 14.91 ()97.59, 146.94) [100] 36.44 ()92.41, 141.15) [144] 7.25 ()65.07, 151.68) [49] 30.21 (4.18, 75.13) [203]

CI, condence interval. Negative condence limits not interpretable because they include bootstrap samples having worse outcome with higher costs, and others having better outcomes with lower costs (Figure 1). Tuberculosis diagnosed or preventive asthma treatment or appropriate urgent referral.

2010 Blackwell Publishing Ltd

283

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

1 0.9

Probability intervention is cost effective

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Base health service cost Base societal cost Health service cost including TB treatment Health service cost; training costs allocated over 6 months Equal probability intervention is or is not cost effective

50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Acceptability threshold ($US)

Figure 3 Cost-effectiveness acceptability curves for tuberculosis case detection, with different costs considered.

1 Probability intervention is cost effective 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 50 100
Base health service cost Base societal cost Health service cost including tuberculosis treatment Health service cost; training costs allocated over 6 months Equal probability intervention is or is not cost effective

150 200 250 300 350 Acceptability threshold ($US)

400

450

500

Figure 4 Cost-effectiveness acceptability curves for appropriate care, with different costs considered.

3 months. As this study PALSA has been expanded to cover HIV AIDS and implemented in four provinces of South Africa, reaching over 6000 health workers in about 900 clinics. During province-wide implementation, each trainer was assigned twice as many clinics as in the trial. With large-scale implementation, the marginal costs of the guideline and educational materials would have decreased. Modelling studies have shown that increasing tuberculosis case detection is the most cost-effective way to control tuberculosis in countries with high HIV prevalence (Baltussen et al. 2005, Currie et al. 2005). However, there is minimal evidence identifying effective and cost-effective ways of increasing tuberculosis case detection (Golub et al. 2005). To our knowledge, PALSA is the rst educational intervention with randomised trial evidence of improved

tuberculosis case detection in Africa. To evaluate PALSAs effects on future tuberculosis transmission and death, one needs to consider evidence from other studies. A WHO modelling study estimated that a 1% increase in case detection and treatment in South Africa would prevent 29000 new cases and 22000 deaths between 2000 and 2009 (Currie et al. 2003). That is equivalent to preventing 0.97 new cases and 0.73 deaths for every extra case detected, assuming an average of 300 000 cases detected every year without PALSA (Currie et al. 2003). The incremental health service cost of PALSA would thus be $70 per case prevented or $93 per death prevented, excluding tuberculosis treatment costs. If health services costs of treating extra cases detected and prevented are included, and assumed to average $500 per case (Sinanovic

284

2010 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

et al. 2003), then the incremental health service costs of PALSA would be $92 per new case of active tuberculosis prevented and $269 per death prevented (not considering costs saved by preventing deaths). The study is an advance on other economic evaluations of similar interventions because it is based on a large rigorous randomised controlled trial with detailed data on patients costs and outcomes. However, the study has several limitations. First, the short duration of follow-up meant that long-term costs and health outcomes which are relevant to chronic diseases such as tuberculosis and asthma could not be measured. Second, the three effects grouped together as appropriate care were given equal weights to produce this composite outcome. We could instead have used different, subjective, weights or more complex statistical methods for composite outcomes (Bjorner & Keiding 2004; Negr & Vazquez-Polo 2006). However, interpreting such results would be less straightforward. Finally, the absence of an accepted acceptability or willingness to pay threshold for these particular outcomes means that one cannot make a straightforward decision about whether or not the intervention is costeffective from this study alone. This study shows that, although the intervention increased costs in the short term, it could be cost-effective, depending on how much funders were willing to pay for better care. These results cannot automatically be generalised to other countries. However, they show that carefully developed and implemented clinical practice guidelines, with educational outreach, can improve tuberculosis case detection and other primary respiratory care at affordable cost. This supports global efforts such as PAL to improve and integrate the primary health care of adults in developing countries. Acknowledgements We thank the Free State Department of Health nurse trainers: Leona Smith, Annette Exley, Sandra Korkie, Tebogo Mothibeli, Francis MacKay, Elizabeth Bolofo and the late Mariana Thirtle; Ineke Buskens for training the nurse trainers; eldwork supervisors Mariette van Rens burg and Gloria Gogo, Centre for Health Systems Research and Development, University of the Free State; Michael Wyeth and Val Myburgh for training materials design; Chris Seebregts and Clive Seebregts, Medical Research Council (MRC), for data management; Chrismara Guttler and Amanda Fourie, MRC for data capture; Sonja Botha for data editing; Sonja van der Merwe and Annette Furter from the Free State Department of Health for arranging pilot sessions and providing routine data; and Miranda Mugford, University of East Anglia for methodological

advice. Victor Lithlakanyane and Mosiuoa Shuping, Free State Department of Health; European Union-funded AfroImplement and PRACTIHC collaborations; and Reloe Matji, formerly South African National tuberculosis control programme, provided early support and advice. Robert Scherpbier and Salah-Eddine Ottmani, WHO, conceptualised PAL which is the basis of PALSAs screening and syndromic approaches. The WHO-led multi-country PAL evaluation includes studies in Chile, Kyrgyzstan, Morocco, South Africa and Nepal. This research programme was funded by the International Development Research Centre, Canada, the South African Medical Research Council, the Free State Department of Health and the University of Cape Town Lung Institute. The authors declare that they had full control of all primary data and did not enter an agreement with funders that may have limited their ability to complete the research as planned. References
Adam T, Manzi F, Schellenberg JA, Mgalula L, de Savigny D & Evans DB (2005) Does the integrated management of childhood illness cost more than routine care? Results from Tanzania. Bulletin of the World Health Organization 83, 369377. Bachmann MO, Fairall L, Clark A & Mugford M (2007) Methods for analyzing cost effectiveness data from cluster randomised trials. Cost Effectiveness and Resource Allocation 5, 12. Baltussen R, Floyd K & Dye C (2005) Cost effectiveness analysis of strategies for tuberculosis control in developing countries. BMJ 331, 1364. Bheekie A, Buskens I, Allen S et al. (2006) The Practical Approach to Lung Health in South Africa (PALSA) intervention: respiratory guideline implementation for nurse trainers. International Nursing Review 53, 261268. Bjorner J & Keiding H (2004) Cost-effectiveness with multiple outcomes. Health Economics 13, 11811190. Briggs AH (2001) Handling uncertainty in economic evaluation. In: Economic Evaluation in Health Care (eds M Drummond & A McGuire) OUP, Oxford, pp. 172214. Carpenter J & Bithell J (2000) Bootstrap condence intervals: when, which, what? A practical guide for medical statisticians. Statistics in Medicine 19, 11411164. Cleary S, Boulle A, McIntyre D & Coetzee D (2004) Cost-effectiveness of Antiretroviral Treatment for HIV-Positive Adults in a South African Township. University of Cape Town, Cape Town. Currie CSM, Williams BG, Cheng RCH & Dye C (2003) Tuberculosis epidemics driven by HIV: is prevention better than cure? AIDS 17, 25012508. Currie CS, Floyd K, Williams BG & Dye C (2005) Cost, affordability and cost-effectiveness of strategies to control tuberculosis in countries with high HIV prevalence. BMC Public Health 5, 130138.

2010 Blackwell Publishing Ltd

285

Tropical Medicine and International Health L. Fairall et al. Cost-effectiveness of educational outreach

volume 15 no 3 pp 277286 march 2010

Daviaud E (1999) Assessing Costs of PHC Services in an Urban District. Department of Community Health, University of Cape Town, Cape Town. English RG, Bachmann MO, Bateman ED et al. (2006) Diagnostic accuracy of an integrated respiratory guideline in identifying patients with respiratory symptoms requiring screening for pulmonary tuberculosis: a cross-sectional study. BMC Pulmonary Medicine 6, 22. Guideline available at: http:// www.biomedcentral.com/1471-2466/6/22/additional/ (accessed 15 December 2009). English RG, Bateman ED, Zwarenstein MF et al. (2008) Development of a South African integrated syndromic respiratory disease guideline for primary care. Primary Care Respiratory Journal 17, 156163. Fairall L, Donald K, Zwarenstein M & Scherpbier R (2001) Evaluation of clinical predictors of lung diagnoses in primary care in Cape Town. South African Respiratory Journal 7, 72. Fairall LR, Zwarenstein M, Bateman ED et al. (2005) Educational outreach to nurses improves tuberculosis case detection and primary care of respiratory illness: a pragmatic cluster randomized controlled trial. BMJ 331, 750754. Gandhi NR, Moll A, Sturm AW et al. (2006) Extensively drugresistant tuberculosis as a cause of death in patients co-infected with tuberculosis and HIV in a rural area of South Africa. Lancet 368, 15751580. Gilson L, Mkanje R, Grosskurth H et al. (1997) Cost-effectiveness of improved treatment services for sexually transmitted infections in preventing HIV-1 infection in Mwanza region, Tanzania. Lancet 350, 18051809. Golub JE, Mohan CI, Comstock GW & Chaisson RE (2005) Active case nding of tuberculosis: historical perspective and future prospects. The International Journal of Tuberculosis and Lung Disease 9, 11831203. Haile B (2000) The Costs of Adult Inpatient Care for HIV Disease at GF Jooste Hospital. Health Economics Unit, University of Cape Town, Cape Town. Hausler PH, Sinanovic E, Kumaranayake L et al. (2006) Cost of measures to control tuberculosis HIV in public primary care facilities in Cape Town, South Africa. Bulletin of the World Health Organization 84, 528536. Louwagie GMC, Bachmann MO & Reid M (2002) Formal clinical primary health care training. Does it make a difference? Curationis 25, 3237. Mash B & Whittaker D (1997) An audit of asthma care at Khayelitsha Community Health Centre. South African Medical Journal 87, 475476. Morgenstern LB, Bartholomew LK, Grotta JC, Staub L, King M & Chan W (2003) Sustained benet of a community and professional intervention to increase acute stroke therapy. Archives of Internal Medicine 163, 21982202.

Negr MA & Vazquez-Polo FJ (2006) Bayesian cost-effectiveness analysis with two measures of effectiveness: the cost-effectiveness acceptability plane. Health Economics 15, 363372. Sanci LA, Coffey CM, Veit FC et al. (2000) Evaluation of the effectiveness of an educational intervention for general practitioners in adolescent health care: randomised controlled trial. BMJ 320, 224230. Shekelle PG, Ortiz E, Rhodes S et al. (2001) Validity of the Agency for Healthcare Research and Quality clinical practice guidelines: how quickly do guidelines become outdated? JAMA 286, 1461 1467. Shresta N, Samir KC, Baltussen R, Kae KK, Bishai D & Niessen L (2006) Practical approach to lung health in Nepal: better prescribing and reduction of cost. Tropical Medicine and International Health 11, 765772. Sinanovic E, Floyd K, Dudley L, Azavedo 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. Statistics South Africa (2009) Consumer Price Index excluding interest rates on mortgage bonds (CPIX). Available at: http:// www.statssa.gov.za/keyindicators/cpix.asp (accessed 16 December 2009). Stempel DA, Roberts CS & Stanford RH (2004) Treatment patterns in the months prior to and after asthma-related emergency department visit. Chest 126, 7580. US Department of Labor (2009) Bureau of Labor Statistics. Consumer Price Index. Available at: http://www.bls.gov/CPI/ (accessed 16 December 2009). WHO (2004) Respiratory Care in Primary Care Services A Survey in 9 Countries (WHO HTM TB 2004.333). WHO, Geneva. WHO (2005) PAL: A Primary Health Care Strategy for the Integrated Management of Respiratory Conditions of People of Five Years of Age and Over (WHO HTM TB 2005.351; WHO NMH CHP CPM CRA 05.3). WHO, Geneva. Available at: http://www.who.int/tb/publications/pal/en/index.html (accessed 15 December 2009). WHO (2008a) Stop TB Partnership. 20072008 XDR and MDR Tuberculosis Global Response Plan. WHO, Geneva. Available at: http://www.stoptb.org/resource_center/assets/ factsheets/MDR_FACT_SHEET_2007.pdf (accessed 15 December 2009). WHO (2008b) Practical Approach to Lung Health. Manual on Initiating PAL Implementation (WHO HTM TB 2008.410; WHO NMH CHP CPM 08.02). WHO, Geneva. Available at: http://www.who.int/tb/publications/pal/en/index.html (accessed 16 December 2009). WHO (2009) Global Tuberculosis Control. Epidemiology, Strategy, Financing. WHO, Geneva. Available at: http:// www.who.int/tb/publications/global_report/2009/en/index.html (accessed 16 December 2009).

Corresponding Author Max O. Bachmann, School of Medicine, Health Policy and Practice, University of East Anglia, Norwich NR4 7TJ, UK. E-mail: m.bachmann@uea.ac.uk

286

2010 Blackwell Publishing Ltd