Background
Annual funding for the control of HIV/AIDS in resource poor countries rose from $US 1.6 billion in 2001 to $US 10 billion in 2008 [
1]. By 2006, an estimated 49% of all external funding disbursed for HIV/AIDS came from two global health initiatives (GHIs) [
2]: The Global Fund to Fight AIDS, Tuberculosis and Malaria and the United States President's Emergency Plan for AIDS Relief (PEPFAR). Between 2002 and 2007, the numbers of people on antiretroviral therapy (ART) in developing countries rose from 300,000 to 3 million, leading to a decline in annual AIDS deaths from 2.2 to 2 million [
3] and an estimated 550,000 life years saved across 14 African countries [
4]. Prevention of Mother to Child Transmission (PMTCT) coverage increased from 9% in 2004 to 33% in 2007 [
3]. In some African countries, external HIV/AIDS funding (mainly from GHIs) has exceeded countries' total spend on their health sectors [
2], accounting for between 67% and 98% of all AIDS funding in five of the poorest countries [
4]. This has fuelled debates about the effects of GHIs on health systems [
5]. However, peer-reviewed [
6] and other multi-country studies [
7,
8], until now, have reported mainly national level perspectives, which report contrasting views and expectations of largely positive or negative effects.
The effects of GHIs on countries' health systems is being researched across 16 countries under the umbrella of the Global HIV/AIDS Initiatives Network (GHIN), which supports independent country research teams that have agreed network aims and principles by which they are researching common themes:
http://www.ghinet.org. The principal GHIN themes include the effects of GHIs on human resources for health (HRH), on other priority services, on the capacity of countries to coordinate GHIs alongside traditional aid mechanisms, and effects on equitable access to services. Research teams from Malawi and Zambia were among four African country teams and GHIN coordinators who agreed on common research questions, approaches and methods at a research planning workshop in Malawi in September 2006.
Between 2004 and 2008, both countries received large grants from GHIs (see Table
1); and national data illustrate the rapid scale-up in the delivery of HIV/AIDS services (see Table
2). Malawi received large levels of funding from only one GHI (the Global Fund) whereas Zambia received funding from both the Global Fund and PEPFAR. We hypothesised, in conducting the comparative analysis, that it might be easier to roll out a coordinated national human resource for health strategy in a less complex GHI arena. PMTCT services have been rolled out to all 28 districts in Malawi and all 72 districts in Zambia; and nationally reported ART coverage was close to 50% in both countries by 2008 [
3]. The World Bank Multi Country AIDS Program (MAP) has also been an external player in funding for HIV in both countries. However, their programme focus was mainly not on health facility scale-up, and therefore was not considered in this paper. This paper presents comparable findings from Malawi and Zambia on the scale-up in service delivery and workload at health facilities, and in numbers and distribution of health workers. The aim is to report trends in health worker numbers, distribution and workload, and to explore and compare the effects of different GHI inputs - Global Fund alone in Malawi and Global Fund and PEPFAR in Zambia - on human resources for health (HRH) strategies and responses, in the light of greatly increased resources for HIV/AIDS.
Table 1
Summary of Global Fund and PEPFAR HIV funding to Malawi and Zambia (in million US$)
Malawi
| | | |
Round 1 | $342.6 m | $229.6 m (Dec 09) | $14.5 m (2004) |
Round 5 | $17.6 m | $13.0 m (Oct 09) | $15.2 m (2005) |
Round 5 (HSS)* | $ 52.0 m | $21.3 m (Aug 09) | $16.4 m (2006) |
Round 8 | $15.1 m | | $18.9 m (2007) |
| | | $23.9 m (2008) |
Zambia
| | | |
Round 1 | $90.3 m | $81.9 m | $82 m (2004) |
Round 4 | $236.3 m | $128.0 m | $126 m (2005) |
Round 8 | $129.4 m | | $147 m (2006) |
| | | $216 m (2007) |
| | | $269.2 m (2008) |
Table 2
Core HIV Indicators in Malawi and Zambia
Indicator
| | | | | | | |
|
2004
|
2005
|
2006
|
2007
|
2005
|
2006
|
2007
|
Population (in millions) | 11.9 | 12.3 | 12.8 | 13.2 | 11.4^ | 11.8^ | 12.2^ |
Adult HIV prevalence (15-49)% Epidemiological indicators | 14.4 (2003) | 14.2 | No data | 12.0 | 13.9 | 13.5 | 13.1+
|
HIV prevalence in pregnant women (%) | 19.8 | 16.9 | No data | 12.0 | 19.1 | 19.1 | 19.3 |
Number (%) of adults and children with advanced HIV infection receiving ART | 13 183 (6%) | 37 840 (14%) | 85 200 (33%) | 130 488 (43%) | 39 351 | 80 030 (32.9%) | 149 199 (50.5%) |
Number (%) of pregnant women needing and receiving ART to reduce the risk of mother to child HIV transmission (PMTCT) | 2719 (3%) | 5076 (7%) | 9231 (22%) | 23158 (35.4%) | No Data | 25,578 29.7% | 35,314 39.1% |
Women and men 15-49 who received a test in the last 12 months and knew their results. | 283 467 | 482 364 | 661 400 | 461 038* | 15.6% | 234 430 (15.4%) | 254 585 (15.4%) |
Numbers of sites providing ART | 20 | 60 | 104 | 109 | 107 | 156 | 322 |
Numbers of sites providing PMTCT | 36 | 40 | 60 | 84 | 67 | 307 | 678 |
Numbers of sites providing HIV Counselling and Testing (VCT) | 146 | 239 | 351 | 370 | No data | 883 | 1028 |
An analysis of Global Fund proposals [
9] and disbursement levels [
9], recorded on the Global Fund website, shows that staff training and supplies for Voluntary Counselling and Testing (VCT) and PMTCT were an important component of Zambia's successful 2003 Round 1 US$90 million HIV/AIDS grant. Zambia's late 2005 Round 4 US$236 million HIV/AIDS allocation included a major component of in-service training for 5,264 health professionals and 32,868 non-health agents. US PEPFAR organisations based in Zambia, where US$ 571 million had been allocated by the end of 2007, reported a range of health systems strengthening, infrastructural development and training components. This included the training in 2006 of 'more than 15,000 Zambian health care workers' in the delivery of a range of HIV services [
10]. In 2003 Malawi was awarded a large (US$342.6 million) Round 1 grant from the Global Fund to HIV/AIDS control. By 2005 it had re-allocated its grant to support its national Emergency Human Resource Programme [
11‐
13]. The significance of this is considered in the Discussion.
Discussion
These findings add to the 'thin and contested ... knowledge base' around the effects of GHIs on countries' health systems [
19]. Data collected directly from facilities and district offices corresponded with nationally reported data [
17,
20], confirming that population-wide scale-up of ART, PMTCT and VCT services has been happening in Malawi (2006-08) and Zambia (2004-07). More importantly, it provides facility level data that demonstrate large increases in HIV service client loads, including an almost threefold increase in ART clients over 30 months in Malawi, and a fourfold increase in ART clients over 48 months in Zambia. The type of intra-facility analysis conducted in this study has been able to demonstrate the correlations in trends between ART scale-up, routine workload and the availability of clinical staff at the facility level. While OPD visits provide only one measure of clinical staff workload, they represent an indicator that was routinely reported by facilities to District Health Management Teams. Such evidence therefore does not rely on special data collection exercises.
In Malawi, there was a modest (10%) increase in clinical staff numbers (doctors, nurses and midwives, and clinical officers and medical assistants) at district hospitals and urban health centres, but not in rural health centres where the increase in staff was principally through non-clinical HSAs. The increase in routine workload in facilities providing ART, notably at the district hospitals but also at rural health centres, suggests a steady increase in client utilisation of these facilities. Whether Malawi's decision to allocate most (91%) of the increases in clinical staff to ART facilities was in response to the increased workload, and/or the greater availability of staff helped to attract more patients, it suggests a coherent approach to health worker distribution when faced with the challenge of delivering ART on top of routine care. The increase in clinical staff in Malawi resulted in a decrease in OPD workload in rural and urban facilities, with a slight increase in semi-urban (district hospital) facilities.
ART scale-up in these three districts of Zambia between 2004 and 2007, was set against a static urban routine outpatient workload, a 24% increase in workload in rural facilities and a 35% rise in smaller rural facilities. A recent study [
21] reported workload as the most important cause of health worker burnout in urban health facilities. These facilities experienced a net decrease in clinical staff numbers, which was proportionately greater in the rural district, and only a modest increase in support staff (technicians and dedicated HIV counsellors). In 2004, rural Mumbwa facility staff were coping with four times as many OPD visits as Lusaka (the capital city) facilities and twice as many as facilities in urban Kabwe. By the end of 2007, dedicated HIV counsellors in Zambia still only accounted for 11% of staff directly delivering a service to clients/patients in surveyed facilities, compared to counsellors and HSAs in Malawi who accounted for 43% of such staff. Unlike Malawi, these district facilities in Zambia did not appear to be using task shifting to non-clinical staff to manage the increased HIV workload during this period. While there was an upward trend in non-HIV workload in ART providing facilities, which may mean they were attracting more patients, the urban-rural disparity was stronger.
The GHIs, notably Global Fund in both countries and PEPFAR in Zambia, were clearly providing the significant proportion of the external funding which was achieving this impressive scale-up in life-saving HIV/AIDS service coverage. An increase from US$3 (2003) to US$5 (2006) per capita expenditure on HIV in Malawi and from US$10 to US$14 per capita in Zambia was due to external resources [
4]. The perception at the national level in Zambia was that in 2008-09 PEPFAR would account for half and the Global Fund for one third of all funding for ART roll-out [
22]. Several reports and other studies have pointed to a large and longstanding degree of rural-urban inequity in Zambia. Only 52% of all health workers and 24% of doctors live and work in rural areas where two thirds of Zambians reside [
23], and there are high vacancy rates and a rapid turnover of staff in rural areas [
24]. Zambia's Public Expenditure Review national HRH survey [
25] reported much higher vacancy rates in rural compared to urban health centres for the following health worker categories: doctors (91%:38%), clinical officers (58%:43%), midwives (50%:32%), nurses (43%:23%). Attribution of findings on health workforce distribution, trends and incentives to the inputs and influence of the Global Fund and PEPFAR - and to government responses to GHIs - is more difficult. However, the findings from this study show a divergence and a deterioration in rural-urban equity in Zambia, during the period when PEPFAR and the Global Fund were likely to be having a major impact.
WHO specifies a minimum workforce threshold estimate of 2.28 clinical staff (doctors, nurses, midwives) per 1,000 people [
26] (23 per 10,000). Clinical staff densities in our study (between 2.9 and 2.1 in the rural facilities and between 6 and 7 in urban facilities) were lower than the 7.9 per 10,000 that have been reported nationally in Zambia in 2004 which had risen to 9.8 per 10,000 in 2007 [
23]. This could partly be attributed to lack of designated catchment populations for the large district and central hospitals. The University Teaching Hospital did not provide data on staff numbers. Rural Mumbwa district (at 2.9 in 2004 falling to 2.1 in 2007), however, was typical of health worker densities in three of six rural districts cited in an early draft of the Global Fund's Five Year Evaluation [
4], which were categorised as 'poor infrastructure rural' (mean 2.6, range 1.7-3.5). More weight can be given to the Zambian than to the Malawi staff density findings, as in the former all public and private fixed facilities were mapped and were included in the study if they were providing ART. In Malawi, only public sector and faith-based facilities were included, which meant that clinical staff in NGO facilities, likely to be common in urban areas, were not included in the study.
The slightly larger rural-urban difference in nationally reported health worker density in Zambia (4.5:16.0) [
23], compared to Malawi (3.5:11.7) [
27], may reflect contextual differences: an estimated 35% of Zambia's population live in urban areas [
28], compared to 18% in Malawi [
29]. The population density in rural areas of Malawi is six times that of Zambia and is among the highest rural densities in the world [
30]. However, whatever the underlying factors, the evidence (based on one rural district) suggests that some rural areas have been falling behind urban areas in Zambia in terms of clinical staff allocations, during the period that GHI funded scale-up accelerated. While this study did not aim to measure rural-urban ART coverage levels, the high proportion of Zambia's nationally reported ART client estimates that were attending facilities in Lusaka suggests that ART service scale-up was heavily skewed towards the capital city, at least during the 2004-07 period.
Quantification of inputs and expenditure on specific health systems components, and efforts by us and by the Global Fund [
4] to track funds to the district and facility level, were unsuccessful. Therefore, establishment of a causal chain and reliable attribution of health systems effects to particular GHIs is not possible. However, our district level findings do provide empirical evidence that supports other mainly national level studies and government and Ministries of Health reports of increasing workload for health staff, especially in rural areas. Malawi appears to have been somewhat more successful than Zambia in recruiting clinical staff, and more so in allocating HSAs and counsellors to supporting scale up. Despite Zambia's efforts and donor support to its rural health worker incentive and retention scheme [
18], progress in implementing its human resources strategic plan has been slow and postings have favoured urban areas at the expense of rural areas [
17,
23]. The scheme has had limited success due to accommodation shortages, a short timeframe for retention allowances and eligibility criteria that until 2007 included only doctors, though it has since been extended to include nurses and nurse tutors [
23]. According to the Ministry of Health in 2009, the current staff establishment contained 32,688 approved positions, though not necessarily funded posts, representing 65% of the staffing requirements for the new structure [
31]. Zambia's national Human Resources for Health Strategic Plan [
18] has also lacked concerted GHI-support for hiring new health workers [
31].
Two explanations may account for the overall less effective scale-up in clinical staff in Zambia: the country may have produced additional clinical staff over 2004-07, but was losing them to better funded posts in the NGO and private for profit sectors (and to emigration) [
32], or it was not producing sufficient clinical staff to meet replacement needs. Others have commented on how rural-to-urban staff migration is compounded by public-to-private provider brain drain, as part of a broader phenomenon of rural-urban inequity [
33]. Key informant interviews in our study reported that urban facilities in Zambia had benefited more than rural facilities from large levels of new resources; and they also reported significant migration from government employment to well funded NGOs, which we could not confirm and quantify. Two studies have reported that the higher wages offered by PEPFAR-funded NGOs were attracting staff away from the public sector [
22,
34]. Up to 2007, PEPFAR was paying salary top-ups and overtime payment for ART delivery [
34]. Together, these findings suggest a PEPFAR-effect that was benefiting the facilities it supports at the expense of other facilities. Prior to the GHIs becoming major players, NGOs were reported to be paying between 23% and 46% more than government [
35]. As Dussault and Franchescini have reported, even where countries have comprehensive health worker policies and strategies, funding may not follow and geographical imbalances result: "Highly-skilled professionals and institutions respond more to incentives than to control mechanisms" [
33].
Malawi's health workforce response suggests differences to Zambia in GHI health systems' effects. Support from donors in April 2005 [
11], including the Global Fund which agreed to the re-allocation of Malawi's Round 1 grant, enabled Malawi to start to implement its Emergency Human Resource Programme [
12]. Demand-side differences, whereby Malawi exerted pressure on the Fund, or supply-side differences, whereby Global Fund portfolio managers interpreted the Fund's guidelines differently in Malawi, could have accounted for this decision to re-allocate the Round 1 grant. As a result, Malawi's Programme has focused on funding basic training (doubling the number of nurses and tripling the number of doctors in training), staff recruitment, deployment (including to rural areas), retention (partly through salary top-ups), basic training and retraining of HSAs to deliver HIV services, and incentives for training tutors [
11‐
13]. Malawi, with the support of the Global Fund through a central pooled mechanism, has been able to invest a greater proportion of its resources on basic training: "... a 165% increase in pre-service training and 79% increase in post-basic training" [
12], compared to Zambia.
Conclusions
The importance of these findings is that they represent what the Global Fund Five Year Evaluation was unable to demonstrate - facility level scale up in clients and service episodes, associating these with indicators of health systems capacity - in this case health worker categories and numbers. The data time-periods are not the same - Malawi's baseline data range from the last quarter of 2005 to early 2008, compared with the start of 2004 to the end of 2007 for Zambia - but clear differences as well as similarities in trends are evident.
Getting better evidence for action
Our findings illustrate much of the 'messiness' associated with reliance on the data obtained from routine health facility information systems, which health systems in sub-Saharan African countries generate and on which they rely for evidence for action. Routine data that are based on health facility records are prone to errors at all stages from initial recording in facility registers, through compilation of data at the facility level for returns to district health offices, during compilation at the district level for reporting to national level, and in analysis at the national level. Data analysis in this study enabled outliers and data of questionable plausibility to be identified and checked, using original research tools/proformas where available. However, this could not preclude errors earlier in the health information system chain, at the level of the health facility recording and reporting system. Health information performance and problems can also be programme-specific. For example, routine PMTCT data in Malawi was not considered to be reliable up to 2007.
One objective of this paper has been to illustrate the potential from analysing health facility data and our analysis demonstrated some of the methodological problems and responses: median workloads (staff-client ratios) are better measures than means for taking into account changes in smaller facilities with low client numbers, because a small number of facilities with large client numbers can have a disproportionate effect on an analysis that uses means, but both measures are important. The collection of facility level data on trends in this study, which the Global Fund Five Year Evaluation did not attempt, demonstrated how health facilities in Malawi and Zambia have been managing to deliver HIV and AIDS services to much greater numbers, while coping with routine workload. The key informant interview data corroborated and helped to illustrate the effects - and the potential for burnout among health workers. The findings are also consistent with and reinforce other findings on rural-urban inequities in Zambia, particularly in terms of workload. Considerable effort was invested by researchers in Zambia to obtain complete data-sets directly from facilities at baseline (2006-07) and again at follow-up (2008) using improved tools. The objective was to show trends in facility outputs of interest: numbers of HIV and non-HIV clients and service episodes. Similar data were collected from national programme offices in Malawi.
In mid-2008, data sets recording OPD and non-HIV priority service clients and episodes were obtained in electronic format directly from district health management offices in Zambia. Reasons for greater completeness of district records, where this was found, were that many health facilities kept no copies of the returns they had sent to district offices; and some, over-time, discarded or mislaid original records. District health offices in Zambia were more consistent than facilities in recording catchment populations (numbers of adults, under ones and under five year old children, women of child bearing age), which facilitated calculation of coverage rates, including immunisation and family planning coverage (data not shown).
The value of staff-population density calculations is more limited in areas where there is a mixture of government and non-government (for-profit and non-profit) providers, and where there are tertiary specialist hospitals that attract patients from afar. Both of these features are characteristic of urban areas. Where staff density data are more useful is to demonstrate health worker allocations and policy responses in rural districts, as in the case of rural Mumbwa district in Zambia where staff densities were falling. The data in this study do not definitely show a growing health worker density gap between rural and urban facilities, but they point to such a gap in those facilities providing HIV service that had catchment population data. Even in the absence of data from non-public facilities, as was the case in Malawi, the available data can still be translated into evidence that should be available to government, with respect to staff allocations to public sector facilities, and to assist with implementation of the WHO rural retention guidelines and policy recommendations [
36].
Acting on the evidence
Staff retention is not only about salaries, top-ups and financial incentives and includes motivational factors that stem from having the infrastructure, management systems, drugs and other commodities for delivering services [
37], which the GHIs have supported. The Global Fund was contributing an estimated 23% of its funding to human resources, though mostly (apart from Malawi) on improving the capacity of existing staff rather than on training and hiring new staff [
19]. Malawi's receipt of large levels of resources from only one GHI - the Global Fund, which was aligning itself with government and pooling its funding with other donors and government - may have made it easier for government to roll out a coordinated national health workforce strategy. The training of new clinical staff, which started in 2005-06 in Malawi, would take time; and the training of volunteers and HSAs as HIV counsellors has been a useful quick response [
38]. However, task-shifting and short-term in-service training should not be considered panaceas [
39] and need to be part of comprehensive government-led strategies [
40]. An even greater investment by donors and governments in the basic pre-service training of nurses, clinical officers, medical assistants and doctors is required. It is shortages and lower densities of clinical staff that lead to higher maternal, infant and under-five mortality rates [
41].
Up to 2007, PEPFAR had a limit of $1 million per-country to be spent on pre-service training, which was raised to $6 million (or 3% of country budgets) from 2009 [
34]. A limited pool of health workers provokes an inevitable competitive tension between programmes funded by government and different donors, especially where GHIs can fund higher salaries and incentives. Reports have highlighted to PEPFAR its lack of support for the production of new health workers and its effects on health worker distribution [
31]. The 2008 PEPFAR reauthorisation promised to take the bold step of training 'at least 140,000 new healthcare workers in HIV/AIDS prevention, treatment and care' [
42], by 2013, with an initial phase (2009-2010) of identifying opportunities for joint health worker training with GHIs [
10]. This may form part of the health systems strengthening component of the new US Global Health Initiative [
43]. If overall levels of GHI funding to countries such as Zambia 'flat-line' or decrease [
44,
45], decisions around the use of available funds to produce and retain new clinical staff, as the Global Fund has enabled to happen in Malawi, will become even more important.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
RB led on study design, data analysis, and drafting of the article. JK participated in study design, data analysis (particularly the Malawi data) and drafting of the article. JS participated in data collection, data analysis (particularly the Zambia data) and drafting of the article. PD participated in data analysis and drafting of the article. VM participated in study design, data analysis (particularly the Malawi data) and drafting of the article). AW participated in data collection, data analysis and drafting of the article. All authors read and approved the final manuscript.