Improving the efficiency of monitoring adherence to antiretroviral therapy at primary health care level: a case study of the introduction of electronic technologies in Guguletu, South Africa

Abstract

This paper presents a case study of the efficiency gains resulting from the introduction of electronic technologies to monitor and support adherence to highly active antiretroviral therapy (HAART) in Guguletu, South Africa. It suggests that the rollout of HAART to such resource-poor communities can be assisted significantly by the introduction of modified cellphones (to provide home based support to people on HAART and improve the management of adherence data) and simple bar-coding and scanning equipment (to manage drug supplies). The cellphones have improved the management of information, and simplified the working lives of therapeutic counsellors, thereby enabling them to spend less time on administration and to devote a constant amount of time per patient even though their case loads have risen threefold. It has helped integrate the local-level primary health service provision of HAART with the kind of centralised data capture and analysis that could potentially support a national HAART rollout.

¹Respectively, Researcher, AIDS and Society Research Unit, University of Cape Town; Professor, Health Economics and AIDS Research Division at the University of KwaZulu-Natal; and Senior Lecturer, Department of Civil Engineering, University of Cape Town. The research for this paper was conducted by Xanthe Wessels (see also Wessels, 2005) under the supervision of Nicoli Nattrass, and with the support of Ulrike Rivett. The final version of this article was written by Nattrass.

1. Introduction

This paper explores the efficiency gains resulting from the introduction of electronic technologies to monitor and support adherence among AIDS patients receiving highly active antiretroviral therapy (HAART) in Guguletu, South Africa. It describes and analyses the impact of introducing modified cellphones (to improve the collection and management of adherence data and provide home based support to people on HAART) and simple bar-coding and scanning equipment (to manage drug supplies). It provides evidence that these changes have simplified the working lives of therapeutic counsellors, enabling them to spend less time on administration and to devote a constant amount of time to each patient even though their patient loads have risen threefold. This shift to the use of electronic systems has improved efficiency at clinic level and helped integrate local-level primary health service provision of HAART with the kind of centralised data capture and analysis that could potentially support a national HAART rollout. Note, however, that this is a case study of an urban based clinic, and hence the findings cannot simply be generalised to the national level. The case study should thus be regarded as indicative of possibilities only.

This introductory section describes the South African AIDS epidemic and the progress of the HAART rollout. It concludes with an overview of the Guguletu case study. Section 2 describes the introduction of cellphone based technology and Section 3 the electronic drug supply management system. Section 4 argues that these changes have improved efficiency, and Section 5 provides a concluding discussion.

AIDS is a serious problem in South Africa. According to estimates from the ASSA 2003 demographic model, by 2006 18.3 per cent of South African adults and 11.2 per cent of the total population were HIV-positive. As can be seen in Figure 1, estimates from the ASSA 2003 model concur with national sero-prevalence data (Shisana et al., 2005) and with the government's Antenatal Clinic Survey. ¹

Figure 1: HIV prevalence in South Africa (ASSA 2003 Model)

This has implications for South African development, although the precise economic impact of AIDS is difficult to model and remains controversial (see e.g. Barnett & Whiteside, 2002; Bell et al., 2004; Nattrass, 2004; Young, 2004). Given the scale of human suffering the epidemic entails, policy makers are confronted with the challenge of ‘rolling-out’ HAART to treatment sites throughout the country – many of which will be catering to people living in resource-constrained environments.

There are, of course, many barriers to rolling out HAART. These include disincentives arising from poor coordination between health and welfare policy – especially with regard to the disability grant, which could undermine demand for HAART and undermine adherence (Nattrass, 2006a) – and the socio-economic challenges facing people in informal settlements and rural areas who may find it difficult to access the rollout sites. Challenges specific to the health system include drug procurement, managing drug supplies, training new personnel, the construction of additional infrastructure and the development of testing and monitoring systems (see e.g. Abdullah, 2003, 2005, on the challenge for the South African health sector; Kober & van Damme, 2004; Over, 2004; Gilks et al., 2006; Simon et al., 2006: 496, for developing countries).

Providing treatment to people already infected with HIV should be regarded as part of a public health AIDS prevention strategy, as well as a therapeutic intervention for the individuals concerned (Cohen & Hosseinipour, 2005; Salomon et al., 2005; Montaner et al., 2006). However, for these preventative and therapeutic benefits to be realised, it is essential that people on HAART adhere closely to their treatment regimens (e.g. Hogg et al., 2002). HAART patients are required to take their medication twice a day for the rest of their lives. Missed medication results in an increase in viral load and in more opportunities for the virus to mutate against the drug regimen (Mugavero & Hicks, 2004; Steain et al. 2004; Simon et al. 2006: 496). Not only does this have adverse implications for the health of the individual, it also threatens to undermine the public health benefits of the HAART rollout itself. Promoting good adherence is thus a key objective of all HAART sites.

In May 2001, the first HAART pilot project based in a South African township was launched in Khayelitsha by Médecins Sans Frontières (MSF) in collaboration with the Western Cape Provincial Government (Naimak, 2006). One of the objectives was to explore the challenges of providing HAART to people living with AIDS in a resource-poor context. The project, which included the provision of counselling and support groups, was successful in promoting good adherence to medication regimens, achieving good clinical outcomes and improving patients' lives (see e.g. MSF et al., 2003; Coetzee & Nattrass, 2004; Coetzee et al., 2004; Naimak, 2006). Although it is always difficult to extrapolate from the success of such pilot projects, it is worth noting that similar good adherence was found at other resource-constrained sites in the Western Cape (Orrell, 2005) and in other developing countries (Ferradini et al., 2006; Gilks et al., 2006).

Since 2001, and especially since 2004 when the government's national rollout finally began, there has been a steady increase in the number of people accessing HAART (see Figure 2). By March 2006, there were 141 774 people on HAART. While this is a sizeable achievement in global terms, it is important to note that this was less than a third of the numbers initially envisaged in the 2003 Operational Plan (DoH, 2003) and that a substantial portion of the numbers on HAART can be attributed to external assistance from NGOs, the Global Fund and PEPFAR (the United States President's Emergency Plan for AIDS Relief) (Nattrass, 2006b). Comparative international econometric analysis indicates that South Africa's HAART coverage should have been a lot higher given the country's level of development and institutional characteristics (Nattrass, 2006c). This suggests that a lack of political will on the part of the Health Minister to roll out HAART may have been a significant constraint (see also Hassan, 2005).

Figure 2: Actual versus planned rollout of antiretroviral (ARV) therapy in South Africa

Recent changes to the management of AIDS policy (which has seen the Deputy President and the Deputy Health Minister assume greater responsibility for AIDS policy) suggest that this problem of political will may now finally have been overcome. However, there remains a set of institutional and economic obstacles – particularly with regard to human resources (Nattrass, 2006b) – which will hamper the rollout for some time to come. In other words, the issue of efficiency (i.e. using available resources as effectively as possible) is likely to become of central concern for existing and future treatment sites.

Unfortunately, there are as yet no economic studies of the efficiency gains that could result from the introduction of new technologies and management systems. Up to now, economic analyses of the provision of antiretroviral therapy have concentrated on the total and marginal costs of expanding the existing health service (see the reviews of economic studies in Boulle et al., 2003; Wessels, 2005). ² While these studies differ in their assumptions about rationing and the cost-saving implications of HAART (see e.g. Boulle et al., 2003; Nattrass & Geffen, 2005), they all assume that improved health outcomes are possible only through the expanded allocation of existing inputs, and that the relationship between these inputs and health outcomes is fixed. None of them investigates whether new ways of managing a HAART rollout at primary health care clinic level can achieve benefits in terms of improved efficiency (i.e. getting better outcomes for a given level of resource inputs).

This case study explores the issue of efficiency gains from the introduction of new technologies to monitor and support adherence to HAART at primary health care level. As noted above, promoting good adherence is a central challenge for those providing HAART. For this reason, all treatment sites include clinical and various forms of psycho-social support for patients. The challenge they will all face in the future as the rollout continues is how to provide access to more people without compromising the quality of care – and at a lower average cost per patient on treatment. We explore this issue by means of a case study of the Hannan Crusade Treatment Centre (HCTC) in Guguletu – a primary health care facility and research site operated by the Desmond Tutu HIV Centre at the University of Cape Town (UCT). Quantitative and qualitative information was obtained from interviews with staff at all levels working at the HCTC, the Desmond Tutu HIV Centre, and at Cell-Life. The research took the form of in-depth interviews and requests for statistical information where possible. ³

The HCTC is a particularly interesting case because it has introduced an electronic system based on cell-phone technology to improve the management of drug supplies and the monitoring of adherence and side-effects in patients. We argue here that although this has imposed additional up-front costs, the greater efficiency this project has made possible has helped the HCTC increase the number of people on treatment without compromising the quality of care. Indeed, there is reason to believe that the service provided to each patient has probably improved as a consequence. The case study suggests that electronic systems can be used to good effect to link primary health care service provision of this kind to a central database – which in turn can feed necessary information back to the HAART clinics. In the case of the HCTC, the central database was located at UCT. However, it is a model that could be transferred to other public sector HAART clinics. ⁴

2. Cellphone Based Electronic Technology at the HCTC

The HCTC began operations in September 2002 as part of a research initiative into the provision and monitoring of HAART in Guguletu. This township, which comprises an estimated 60 000 households, dates back to 1958 when it was known as Nyanga West. Assuming that HIV prevalence in the population is similar to that among black South Africans in general, ⁵ it is likely that about a fifth of adults in the area are HIV-positive – although clinic staff at the HCTC believe that HIV prevalence is closer to 25–27 per cent (Catherine Orrell, HCTC site manager, 30 June 2005). The HCTC is the only site providing HAART in Guguletu and is currently providing treatment to over 1000 patients. It is located at the Guguletu Community Health Care Centre, but operates independently of it – being linked instead to the Desmond Tutu HIV Centre at UCT. Although this is problematic on one level (because the provision of HAART is not integrated into the public sector in the same way that the MSF clinic in Khayelitsha is), it has the advantage of providing a stand-alone case study of a monitoring system which could be adopted by the public sector at a later date.

As was the case for the MSF clinic in Khayelitsha, treatment outcomes have been very positive at the HCTC, with average adherence rates in the range of 94–98 per cent (Orrell et al., 2003). Adherence support is provided to patients by ‘therapeutic counsellors’ – i.e. members of the community who are themselves HIV-positive, and who have been hired and specially trained by the HCTC to provide such services. Each patient is assigned a counsellor who helps with the initiation and continuation of treatment. One of the most important functions of these counsellors is to monitor adherence and look out for adverse side effects.

The main tool for monitoring adherence at the HCTC is to count the number of pills in a patient's possession (and then compare this to the number expected to be remaining if the patient has adhered completely to the treatment regimen). Every time patients come to the clinic for their monthly visit, they are required to bring their remaining tablets with them. These are then counted by the doctor during the appointment. This is the main adherence measure used by doctors on site. The therapeutic counsellors supplement this data by making surprise visits to the homes of HAART patients and conducting pill counts in ‘the field’. This and other information about side effects and health problems is noted down by the counsellors and then relayed to a central secure database by means of a cellphone.

This innovative use of cellphone technology was pioneered by the ‘Cell-Life’ project which was initially conceived in the Department of Civil Engineering at the University of Cape Town (see http://www.cell-life.org). The project transformed cellphones into data acquisition and transmission tools by installing menu applications directly on the SIM card of the handset. The data, once entered on the handset, are sent via SMS (Short Messaging Service) to a central secure database. The system is designed so that accredited administrative, managerial and clinical staff can access the database to retrieve patient information via a secure internet connection. Reports of relevant cohort data can also be downloaded from the website in encrypted format.

Prior to the introduction of this cellphone based monitoring system, the process of data capture and entry was a great deal more labour-intensive. The counsellors would note down the pill counts and other relevant information in their ‘patient diaries’, then they would transcribe the information onto data capture forms and deliver these to the clinic when they had the time. These forms, in turn, were then entered manually into the database by a third party at the Desmond Tutu HIV Centre at UCT. The cellphone system was designed to cut out all these stages, enabling the counsellor to key in data directly from the patient's home to the database. However, because of the problem of cellphone theft, Sister Lulu Mtwishe (the counsellors' supervisor) advised the counsellors not to take the cellphones with them on home visits, but rather to continue recording patient information into their diaries (30 June 2005). They then enter the data into their cellphones in the evening. Despite this unanticipated crime-related glitch, the cellphone system appears to have made the counsellors' lives considerably easier, as the following statements reveal:

I would not want to work without my phone as I would have to keep my own paper records and every week I would have to submit the pieces of paper to the clinic and this is a lot of work. (Flora, 26 June 2005)

With the paper system it is a lot of work and it is not professional. With the cellphones you don't spend a lot of time writing down the pill counts and filling in forms. (Thembekha, 26 June 2005)

Not only is the electronic system easier for the counsellors, it has also improved the database immensely by overcoming the problem of mislaid forms that plagued the previous paper based system:

Paper is something that you or your supervisor can lose and at the end of the day it appears as though you have not done your visit, as the record is lost. With the cellphones it's much better than working on paper. (Thembekha, 26 June 2005)

The cellphones are much quicker and easier to use. You don't lose patient records as often and most of the information is already on the database so it is not too much of a problem if some patient data is lost. (Fundi, 28 June 2005)

The introduction of the cellphone system also appears to have reduced the administrative burden of data capture and analysis, and to have improved the flow of information, to the benefit of doctors and patients alike. According to Sister Lulu Mtwisha, the electronic system is vastly superior to the old paper based system:

It is incomparable, the paperwork was good for that particular period, but with the advent of Cell-Life it is something that I don't wish to revisit. With the old system I would get the information a week after everything had happened. I would make the counsellors hand in the paperwork of their visits for the past week on a Monday or Tuesday. I would then bring the pile of forms back to my office at Desmond [the Desmond Tutu HIV Centre at UCT] and only when I had time would I peruse and check. I may not have noticed anything going wrong for an entire month and only at the end of the month, when I could look at the data for that whole month, would I notice that a certain patient had not been visited. Now with the Cell-Life system I can pick this up quickly and there is really nothing that can fall off the plate without me noticing. The system allows me to prevent situations from happening rather than having to fix a situation once it has already happened. It's a form of damage control. (26 June 2005)

The new system has not only improved the monitoring of adherence, it has also saved lives. For example, a database administrator noticed that one of the patients was experiencing erratic and unusual symptoms. He alerted Sister Mtwisha, who in turn notified the doctor, who diagnosed lactic acidosis – a potentially fatal side effect of antiretroviral therapy requiring immediate cessation of the medication. This patient owes his life to the new system which was able to collect the information needed to save him (Sister Mtwishe, 30 June 2005).

It would thus appear incontrovertible that the electronic system is an enormous improvement on the old paper based system. It has resulted in greater efficiency in terms of time saving for the counsellors (and the clinic staff who were previously involved in shuffliging piles of paper and searching for lost forms) and the quality of care provided to patients. But all this, of course, comes at a cost. Cell-Life estimates the cost of implementing the system at the HCTC (including all training, hardware, software, cellphones and airtime) as R37.48 per patient per month (for more detail see Wessels, 2005). The economic question that needs to be confronted is whether the improved system justifies the additional resources required to implement and maintain it. This is the task undertaken in Section 4 below. However, before we approach this subject, it is necessary to describe the second component of the electronic adherence monitoring system at the HCTC, i.e. the antiretroviral drug dispensary and stock recording system.

3. Electronic Antiretroviral Drug Dispensary and Stock Recording

Prior to the introduction of the electronically based ‘pharmacy plan’ at the HCTC, all dispensary and stock recording functions were performed physically by the pharmacy staff and recorded using a paper system. Paper records indicated which patients were due for renewed dosages, drug pick-ups by patients were recorded manually, and at the end of the day the remaining drug stock was counted and recorded manually onto paper. This system was time consuming and plagued by stock misplacement, theft and unnoticed drug expirations.

The automated pharmacy plan, which was implemented at the beginning of 2003, comprises bar-code labelling and scanning, a basic stock control system and a cellular and internet based monitoring of patient antiretroviral drug collections. This has greatly simplified supply chain management. The patient receives a prescription of drug type and dosage from the doctor, which is handed to the pharmacist at the HCTC on collection of the initial treatment dose of antiretroviral therapy. The pharmacist enters the prescribed triple therapy for a particular patient into the system via the computer. Each prescription lasts six months, but drugs are dispensed in one-month dosages to ensure that the patient visits the clinic at least once a month. This periodic collection of drugs forms part of the adherence monitoring process by allowing the pharmacist to keep a record of which patients have, or have not, collected their regimens for the month.

The database provides and stores information on the exact combination of drugs that need to be packaged for each patient every month. Drugs are packaged each month at the Desmond Tutu HIV Centre and bar-coded (under the previous system, pill packaging was outsourced). The bar-code system is used for both stock recording and patient collection record purposes. When drugs are scanned out, the bar-code system records a set of information indicating that the drugs have left the Desmond Tutu HIV Centre. This system notifies management of the specific drug type, the quantity, and the patient label showing for whom the drugs are intended. When the drugs arrive at the HCTC the details are scanned into the system for stock recording and, simultaneously, the Desmond Tutu HIV Centre is notified over the network that the drugs have arrived safely at the clinic. When the patient arrives to collect a month's supply of drugs, the system identifies the triple therapy required for that patient and the correct stock is taken off the shelves and scanned out of the clinic. The bar-code scanning process concurrently records the stock supply reduction and the fact that a certain patient has collected his or her prescription. This simultaneously notifies the pharmacist at the Desmond Tutu Centre that the patient's pill pack for the next month can be packaged. Barcode scanning thus ensures a continuous feedback process, allowing the drugs to be tracked down at any location at any time. As Jalal Ghiassi-Razavi, the Systems Support Manager at Cell-Life, puts it, ‘a great benefit of the system is that, just like a DHL [courier] package, you can track it and know exactly where it is’ (2 July 2005).

Since the implementation of the pharmacy plan, the various functions of the pharmacist have been greatly simplified and organised into a single system whereby stock recording, collection recording and adherence monitoring are all performed simultaneously. Jalal Ghiassi-Razavi explains:

The system has made her [the pharmacist's] life a lot easier. She currently has over 1000 patients on the system and I wonder how she would have managed this on paper. When we started there were about 500 patients, even less, on the pharmacy system and now we have a massive database. (2 July 2005)

While the pharmacy plan has contributed significantly to solving the previous difficulties experienced in the paper based stock management and drug dispensary processes, it has the disadvantage of implementation and maintenance costs. Cell-Life estimates the total (appropriately discounted) cost of the system to be R32.00 per patient per month (assuming a patient load of 500). These costs include design and development, implementation, training and testing, and maintenance per annum. However, to offset this, recall that the Desmond Tutu HIV Centre is now responsible for packaging the drugs (which had previously been outsourced). For the purposes of this analysis, we assume that the additional costs for the Centre are equivalent to the previous costs of the outsourcing (see Wessels, 2005, for more details).

The pharmacy and cellphone systems monitor adherence by functioning interactively. The pills issued to the patient by the pharmacy are recorded on collection onto the database, and the surprise pill counts performed in the field are sent over the network onto the same database. The two measures are directly comparable and complement each other. If a patient who is due to collect his or her treatment has not arrived within a week, this is identified on the system and a counsellor will be sent out to locate the client. Discontinuation of treatment is thus prevented.

Furthermore, the possibility of issuing the incorrect drugs to clients is significantly reduced – as is the danger of issuing drugs that have passed their expiry date. Drug expiry dates are recorded and pharmacy staff is automatically notified when the expiry date of a certain batch is imminent. More importantly, the system enables management to make decisions about drug type and manufacturers. For example, the database will allow management to identify whether certain drug batches are resulting in similar side effects (such as Nevirapine rashes) or whether clients who have been issued with a specific drug experience similar symptoms. The system operates to single out these drugs, enabling management to make more reliable decisions about drug quality.

In short, it appears that there are clear benefits for managing drug supplies and for monitoring adherence, side effects and other health related events among patients. As with the introduction of cellphone technology to help therapeutic counsellors monitor adherence, the economic question that needs to be posed is whether these benefits are sufficient to justify the increased costs. Section 4 below considers this issue with regard to the introduction of both the electronic supply management system and the cellphone monitoring system.

4. Exploring the Nature of Improved Efficiency

In economics, the concept of efficiency relates specifically to the relationship between inputs and output. Put simply, if the ratio of inputs to output falls, then greater efficiency has been achieved. Improvements in efficiency occur either when a greater output is achieved for a similar level of inputs – or if the same amount of output is achieved for a lower level of inputs. The tricky issue when thinking about improved efficiency in the provision of health care is how to measure inputs and outputs and conceptualise the relationship between them. For example, if the relevant ‘input’ is staff hours per patient and the relevant output is ‘patient care’, then efficiency can be said to increase if either the same quality of care is provided for a lower number of hours, or an improved quality of care is produced using the same labour effort. But the notion of ‘quality of care’ is extremely difficult to measure as it is a function of the number of hours in the company of doctors, nurses and counsellors, the skills and dedication of such personnel, and the support systems they are provided with. The job of supporting HAART patients (in order to ensure adequate adherence) is complicated further by the specific socio-psychological needs of the HAART patients, the counselling and personal skills of the counsellors, and the specific relationship between counsellor and patient. Understanding efficiency in this context is thus more of an art than a science and requires both qualitative and quantitative analysis, drawing especially on how counsellors perceive their working lives to have changed and the opportunities provided by the electronic technology to enable them to take better personal care of their patients.

With regard to the HCTC case study, interviews with key players suggest very strongly that the introduction of the electronic systems has improved the quality of adherence data without undermining the quality of care for patients (understood as counselling time available for patients). This can be seen most clearly in the pharmacy plan, which appears to have improved managerial efficiency at clinic level, supplied supplementary data on adherence for the patient database, and provided patients with a better and safer service.

The impact of the cellphone system is a little more complex to unpack. There are strong indications that it has improved the quality of adherence data, reduced the administrative burden on the counsellors and improved the management of health-related information for the benefit of patients. However, we also need to take into account that there has been a rising burden (increased patient load) on each counsellor as the rollout has progressed. It may well be that the cellphone technology has important benefits – but if these are overwhelmed by rising patient numbers per counsellor, then the overall analysis of efficiency (at clinic level) needs to take into account a possible overall negative impact on quality of care (that is, less time spent per patient).

Table 1 provides some indicative hourly and costing data on this issue. ⁶ It shows that between November 2002 and November 2004 the numbers of people on HAART rose sharply from 44 to 437. Although there was a significant expansion in all categories of staff at the HCTC, this was not sufficient to prevent a decline in the ratio of clinic staff to HAART patients. This is not necessarily a problem for quality of care because one would expect an antiretroviral treatment site to start operations with significant excess capacity (particularly with regard to doctors and nurses). However, the trend is worrying with regard to counsellors whose employment one would expect to be more closely correlated with the numbers of patients on HAART. As can be seen from the table, the number of patients per counsellor rose from 5.5 to 15.6 over this period. In other words, since the electronic systems were introduced, the number of patients per counsellor has risen threefold. If this meant that counsellors had less time to spend with patients, then we would need to adjust our efficiency calculation on the output side.

Table 1: Key input indicators, November 2002–November 2004

	November 2002	November 2004
Number of patients	44	437
Number of doctors	1	4
Number of nurses	2	3
Number of counsellors	8	28
Administrative staff	2	4
Total staff	13	39
Total staff per ARV patient	0.3	0.1
Number of patients per therapeutic counsellor	5.5	15.6
If each patient is visited once a week for one and a half hours (including travel), % of time spent by therapeutic counsellors visiting patients^1	20%	60%
% of time used by therapeutic counsellors for administration, data entry etc. (calculated as 100% minus the line above)	80%	40%
Monthly cost of the pharmacy plan (R32 per ARV patient) – data from Cell-Life		R13 984
Monthly cost of the cell phone adherence system (R37.48 per ARV patient) – data from Cell-Life		R16 379
Indicative monthly salary bill (in 2005 prices)^2
Therapeutic counsellors	R29 280	R102 480
Total salary bill (i.e. including doctors, nurses, administrative staff and therapeutic counsellors) at HCTC	R120 054	R322 095
Salary bill for therapeutic counsellors plus cost of the cell phone adherence system	R29 280	R118 859
Salary bill for therapeutic counsellors plus the cost of the cell phone adherence system per ARV patient^3	R666	R272
Total salary bill plus costs of electronic system	R120 054	R352 458
Total salary bill plus costs of electronic system per ARV patient	R2 729	R807
^1This is calculated as the number of patients multiplied by 1.5 times 4 (to get the number of hours per month spent on visiting patients) expressed as a percentage of the total number of counsellor hours available for all work (administrative and counselling).
^2These are not actual salaries from HCTC. They are indicative salaries based on the average cost estimates for 2001 provided in Geffen et al. (2003), inflated by 10% per annum to provide a 2005 figure.
^3This is calculated as total costs (the total salary bill for therapeutic counsellors plus the costs of the cell phone system) divided by the number of patients. Although the costs rose over the period, the number of patients rose faster, and hence cost per patient fell.

In the qualitative interviews with the counsellors, a great deal of attention was paid to probing this issue. Despite acknowledging the increased patient load, the counsellors reported that each patient continues to receive the same amount of attention as had been the case at the start of the rollout (Wessels, 2005). They reported that they had been able to find the extra time to cope with the higher patient load by spending less time on administration and data recording. When asked whether using the cellphone meant that the counsellors spent less time with each patient, one of them (Flora) replied sharply ‘No, because my cellphone does not have wheels’ (26 June 2005).

Table 1 shows that if we assume that each patient continued to enjoy the same amount of time with the counsellors, and if each patient is visited once a week and the visit and round trip take up one and a half hours of counselling time, then the number of hours spent by counsellors on administration has probably dropped from about 80 per cent to 40 per cent as a consequence of using cellphones. However, this drop is probably overstated, since the counsellors were no doubt spending more time in the early period being trained and learning how best to conduct home visits. They would thus no doubt have become more productive over the period anyway, and this is not captured by the simple exercise reported here.

In short, we can reasonably conclude that the quality of care provided to patients on HAART by the counsellors has probably not declined, and hence that the introduction of cellphone technology has improved the efficiency of this service in the sense that more people are accessing quality care for a smaller input of total therapeutic counselling hours. But has this outcome been too expensive in terms of the additional costs required to implement and operate the cellphone system? Table 1 shows that if we use the estimated monthly cost provided by Cell-Life of R37.48 per patient on HAART and add this cost to an estimated salary bill for the counsellors, then the total monthly cost rises almost fourfold over the period. However, as the rate of increase in the number of patients on antiretroviral therapy was double that, the cost of providing counselling services to people on antiretroviral therapy was more than halved from R666 to R272. In other words, the introduction of cellphone technology has not only freed the counsellors to spend more time with their patients but has also probably had the additional benefit of helping to lower the average cost of providing such services to people on HAART.

As discussed in Section 3, the second arm of the electronic system introduced at the HCTC entailed the pharmacy plan for managing drug supplies and stocks. While this, too, entailed extra costs in terms of systems implementation and management, it appears to have at least been consistent with a general shift towards greater efficiency at the HCTC. As shown in Table 1, the average amount spent per HAART patient on the wage bill plus the total (monthly) costs of the electronic systems declined sharply over the period from R2729 to R807. Note that this does not include the total costs of operating the HCTC. Such a calculation would require information about rent, drug costs, equipment, and so on. The analysis reported here was limited specifically to human resources and systems costs in order to explore the more limited question of the efficiency gains of introducing an electronic system for managing the health and adherence of patients on antiretroviral therapy.

5. Concluding Discussion

The introduction of the electronic cellphone and pharmacy systems appears to have improved adherence monitoring and health evaluation. As Sister Mtwishe observes:

Now it's efficient because the counsellors are given the information today and today Catherine [Dr. Catherine Orrell] can get her results. She can quickly see the adherence, and is able to check if it is right. The system can even measure the adherence of a patient for the whole programme in a jiff. Whereas before we had to wait for weeks on end before Catherine could get this information. (26 June 2005)

However, there remain some glitches in the system. While surprise pill counts from the field and data on pills issued from the pharmacy are sent directly through to a single database, clinic based return pill counts that are calculated during the patients' periodic clinic visits continue to be noted on paper records by the doctors. This means that without a direct comparison of clinic pill counts with surprise pill counts, discrepancies between the two values cannot easily be identified at an individual level. While adherence data gathered by the clinic serves well as a cohort comparison, whereby average values and adherence results can collectively be compared to those of other sites, on an individual patient basis the clinic information is not relayed to the source of pill issuance fast enough for the pill count values to be of significant use to doctors.

It is thus unsurprising that during interviews, Mr Ghiassi-Razavi (from Cell-Life) and Dr Catherine Orrell (from the HCTC) recommended that a complete electronic system be introduced, whereby all patient information collected by doctors at the clinic is entered into and stored on the same database instead of in separate components (Wessels, 2005). This, however, would require additional resources for administration at the clinic level – perhaps even the addition of a post for a data capturer and analyst. There is clearly a need for more resources to be allocated to making the most of the information that is currently being collected. Consider the case referred to earlier of the database administrator noticing the irregular symptoms of the patient with lactic acidosis and reporting it to the clinic. This anecdote is both comforting and alarming: comforting in that the system is capable of identifying problems, but alarming because it was the result of an observant individual perusing the data and not the result of the kind of systematic ongoing analysis of the available data that is necessary to generate the best health outcomes for all patients. There are clearly still challenges ahead to transform data collection from what is essentially an after-the-fact research tool to a tool for ongoing, immediate use by clinicians.

While this will obviously raise costs, it is unlikely to do so to anywhere near the extent that would be required to reverse our overall conclusion of improved efficiency. As can be seen in Table 1, the new system helped facilitate the expansion of HAART to more people at a cost that was about a quarter of the cost per patient two years earlier. Furthermore, by improving the integrity of the database for adherence monitoring, additional benefits in terms of improved health outcomes will result.

In sum, this exploratory case study points to the potential of simple electronic technologies to monitor adherence and drug supplies at primary health care level. It suggests that the rollout of HAART to resource-poor communities can be assisted significantly by the introduction of cellphone based technologies, and simple bar-coding and scanning equipment. Most importantly from the perspective of providing support services to people on HAART, the cellphone technology is easily accessible to peer counsellors who, being HIV-positive and from the community themselves, do not always possess high levels of schooling. The cellphone based technology, through which they key in data directly to a database through a secure internet connection, has made their work simpler, enabling them to spend more time with patients and less on administration. It has, in other words, helped integrate the local-level primary health service provision with the kind of centralised data capture and analysis necessary to monitor a national-level HAART rollout. As such, it is a rare example of good news in a sea of Cassandra-like academic writing on the infrastructural and administrative obstacles facing the national HAART rollout.

Notes

¹Respectively, Researcher, AIDS and Society Research Unit, University of Cape Town; Professor, Health Economics and AIDS Research Division at the University of KwaZulu-Natal; and Senior Lecturer, Department of Civil Engineering, University of Cape Town. The research for this paper was conducted by Xanthe Wessels (see also Wessels, 2005) under the supervision of Nicoli Nattrass, and with the support of Ulrike Rivett. The final version of this article was written by Nattrass.

¹This is to be expected given that the ASSA demographic modellers use data from the Antenatal Clinic Survey along with death statistics, fertility statistics, assumptions about interventions, etc., to come up with the demographic model. The ASSA 2003 model is so named to indicate that the most recent Antenatal Clinic Survey is from 2003. It is available from the Actuarial Society of South Africa at http://www.assa.org.za.

²A copy of Wessels (2005) is available at the CSSR (Centre for Social Science Research), University of Cape Town, and can be accessed through inter-library loan.

³It is possible that interviews with patients would have added to the value of the study. However, while interesting, such research would not have been relevant because the patients were not directly involved in the electronic system or the paper system. The only change they would have experienced would have been changes in the amount of time that the counsellors spent with them – and this would have varied from patient to patient and depended on their specific needs and on their relationship with the counsellor. Rather, in order to address the question at hand, it was more appropriate to interview counsellors.

⁴Note that this study only considers the impact of this highly targeted intervention for managing HAART patients. We do not consider its possible use for other interventions (such as managing TB patients) as this was beyond the scope of the study.

⁵This is supported by the ASSA model. The ASSA provincial model for the Western Cape predicted that HIV prevalence in the African population in the Western Cape was 13.9 per cent, as compared with 13.8 per cent among Africans in the country as a whole. Both models can be downloaded from the ASSA website at http://www.assa.org.za.

⁶The analysis shown in Table 1 differs from Wessels (2005) in that it assumes that all categories of staff are working full-time on the antiretroviral treatment programme. Wessels (2005) allows for small differences in the numbers of hours, which affects the point estimates but not the overall result.