In this study, we estimated for the first time the number of additional HHWs needed to achieve universal access to HIV treatment in South Africa, using data from a time and motion study. We found that universal access to ART at CD4 cell counts of ≤350 cells/μl will require South Africa to commit a further 2,200 nurses, 3,800 counselors, and 300 doctors to HIV treatment, at a cost of ZAR 929 million (US$ 141 million) in salaries. We found an average HHW:patient ratio of 0.2 per 1,000 patients for doctors (for performing all ART initiations), and 1.2 per 1,000 patients for nurses and 2.1 per 1,000 patients for counselors (for performing initiations of eligible people and maintaining them on treatment).
It is interesting to compare our empirical findings from this time and motion study with estimates based on other sources of information. Based on reports of the total numbers of health workers and patients in HIV treatment programs, Hirschhorn et al. estimated that in 2004 the number of doctors and nurses required to treat 1,000 HIV-infected people were 1–2 and 2–7, respectively, in different developing countries . Based on recommendations ‘by experienced practitioners based on experience at sites in Kenya’, a 2004 WHO report estimated that 1 doctor and 2 nurses were needed to treat 1,000 HIV-infected people . The lower HHW requirements estimate by our time and motion study may be due to the fact that we are now more than five years into the public-sector ART scale-up program in South Africa , and the productivity of HIV treatment delivery has increased because of scale effects and learning over time. Of course, it is also possible that the estimates differ because of quality-of-care differentials. The expert opinions elicited in the WHO report may have reflected a higher standard of quality of care than currently found in the real-life, public-sector HIV treatment program in rural KwaZulu-Natal, where this study took place. Similar to other treatment programs in the region, levels of non-retention and non-adherence in this program are substantial ; nevertheless, the program has been very effective in reducing HIV-related mortality  and increasing life expectancy  in this community, and can therefore be considered to have been successful, delivering relatively high quality of care.
Data on the current health workforce specifically devoted to HIV care in South Africa are not available, but WHO estimates show that in 2004 there were around 35,000 doctors and 180,000 nurses in total in the country ; consequently, the HW:population ratio in South Africa exceeded the threshold of 2.3 doctors and nurses per 1,000 population that has been proposed by WHO as a critical minimum [35, 36]. Although our study determined the additional numbers of health workers required for future ART scale-up under different scenarios, it did not enable us to evaluate this number directly in relation to the supply of health workers in order to establish whether there is a shortage in health workers in South Africa, either nationwide or in regions. The number of doctors required to provide universal access to ART is currently one of the most important capacity constraints in SSA, because only a few thousand doctors graduate from medical schools on the entire subcontinent each year [37, 38], and the rates of doctor migration to countries outside the region remains high . However, scaling up treatment to universal access for initiation at CD4 cell counts of ≤350 cells/μl will require only about 300 additional doctors committed to performing ART initiations, and may thus be feasible without major changes to national health worker production and retention. However, recruiting the required additional 2,200 nurses fully devoted to HIV care may prove to be a greater challenge, given that the total of all professional nurses who graduated from nursing schools in South Africa in 2011 was only about 5,600 .
It is therefore vital to increase efforts to expand the health worker pool for HIV in South Africa by increasing training and retention, considering reinstatement of retired health workers, or increasing HHW productivity, in particular to achieve universal coverage at more relaxed eligibility criteria . Currently, public-sector HHWs are paid a salary on a monthly basis and, additionally, receive contributions to health and retirement pension insurance, as well as a rural allowance for service in under-served areas. Alternative models of contracting and incentivizing HHWs, such as performance-based payment, might improve productivity. However, such new models might also lead to inefficiencies (for example, transaction costs for monitoring performance), and unintended behavioral consequences (for example, decreased quantity and quality of care for services not included in the performance-based payment scheme) . It is possible that transferring public-sector patients on ART to the private sector for routine follow-up and monitoring, as has been done effectively in Botswana and Mexico [41, 42], might increase the pool of available HHWs. At the same time, of course, this strategy might increase the human resources costs per individual on ART, because health worker salaries in the private sector in South Africa are higher than those in the public sector. Health worker interventions, such as shifting tasks from more to less skilled health workers  and integration of ART delivery into the general primary care services  should also be considered as means to free up human resources for HIV treatment. Integration might improve productivity, if it either increases capacity utilization (for example, reducing idle time) or leads to economies of scope, as different health services are combined. Redistribution of HHWs from overstaffed to understaffed clinics could further improve HHW productivity by reducing idle time. Finally, a number of interventions could increase the supply of health workers in South Africa, nationally or in regions, including interventions to decrease health worker migration out of the country and to increase health worker production [45, 46]. One recent example is the 2012 agreement between Cuba and South Africa, which is intended to ensure continued placement of Cuban doctors in South African medical schools and increased training of South African nationals in Cuban medical schools .
There are several options for improving HHW productivity in PHC HIV clinics. We found that the average duration of a work day varied widely between clinics. The number of HHWs required for the treatment of 1,000 ART patients could be reduced by about one-third if all clinics achieved the productivity of the most efficient clinic. In addition, the average duration of a work day in our data was 6.3 hours. If we assume an HHW work day to last 8 hours, we find that the total number of nurses, counselors, and doctors required would be reduced by 23%, 24%, and 37% compared with the baseline in clinics A, B, and C, respectively. Another way to improve productivity might be to reduce opening hours in selected clinics. We found that in the least busy (and most rural) clinic in our sample, only 20% of HHW time on one particular day was spent on direct patient contact and 36% was spent on breaks and idle time, because only a few patients visited the clinic on that day. Limiting the opening hours of selected PHC HIV clinics might reduce wastage due to idle time. At the same time, however, such a strategy carries the potential danger of reducing access for particular populations, such as employed people or people living in rural areas. Finally, productivity gains might be achieved by cutting the time spent on breaks and idle time between patients. However, we found that on average, only 9% of the work day was spent in this category, which translates to 43 minutes of break time in a work day of 8 hours, which is very short considering the demands of the job and the need for health staff to eat and take care of themselves during the work day, thus there is little room for considerable productivity gains through this approach.
Universal access for all HIV-infected people (’treatment as prevention’) would require an additional 800 doctors, 6,000 nurses, and 11,000 counselors fully dedicated to HIV treatment. Salary costs to cover all HHWs under such a strategy would be about ZAR 2.6 billion (US$ 394 million) per year, which is an increase of 20% compared to the current total HIV sector budget in South Africa. Of course, salaries are only one part of the running costs of added HHWs and expansion of HIV treatment. Other expenditures necessary for HIV treatment include running costs for drugs, medical supplies, water and heating, and investment costs for equipment, facilities, and continuing HHW education [48, 49]. In addition, the need for support staff not involved in direct patient contact, such as laboratory technicians, supply-chain workers, general management staff, and trainers, will also grow with increasing treatment coverage. At present, a strategy of providing ART for all HIV-infected people seems unrealistic for South Africa without large additional financial commitments and substantial increases in HHW training capacity. However, the future effects of a `treatment as prevention’ strategy through reduced transmission has been suggested to be considerable [15, 50, 51], possibly outweighing the initial investments. Furthermore, the workload of initiating patients with advanced disease is higher than for patients with relatively high CD4 cell counts, resulting in a lower HHW:patient ratio. Further research is needed into the epidemiological and clinical benefits of a ‘treatment as prevention’ strategy, for example, through randomized controlled trials in general populations in SSA .
Alternative models of ART delivery through task shifting or simplifications might provide a solution for the shortage of HHWs. Preliminary studies have shown that the quality of nurse-initiated treatment is comparable with that of doctor-initiated treatment , and this strategy has recently been implemented in South Africa . Our results suggest that this strategy could reduce overall salary costs by 7%. Reducing the required frequency of routine clinic visits could also be a useful tool to decrease the required number of HHWs required to maintain patients on ART. Currently, patients are required to visit the clinic every month until treatment has stabilized, and once every 2 months thereafter. Up to 50 to 60% of total salary costs might be saved if patients visited a clinic only once every 3 or 4 months, or if the number of nurse-attended visits were decreased. Although these figures suggest substantial potential for resource savings for universal HIV treatment coverage, it is currently not understood how the implementation of these alternative treatment delivery models will affect the quality of care. Levels of non-adherence and non-retention are already worryingly high in South Africa’s expanding ART program , and these problems might worsen if routine-visit frequencies were decreased.
Our study has several limitations. First, the time and motion data used in this study were from three different PHC HIV clinics in a rural area, and covered only 13 days of observation. However, the time spent per patient did not differ significantly between the clinics. Although the duration of the work day did differ significantly between two of the clinics, this could have been caused by the location of the clinics (rural versus more urban), which is associated with patient load, rather than with differences in work routine or the full potential productivity of the HHW. In the clinic with the shortest effective work day, an average of 14% of the time was spent on breaks and idle time between patients, whereas in the other clinics this percentage was only 5% and 7%.
Second, we extrapolated our data from this specific rural setting to South Africa as a whole. Similar studies in other parts of South Africa are needed to confirm whether this extrapolation is justified.
Third, HHWs were aware that they were being observed, which might have induced them to increase their productivity during observed visits compared with unobserved visits. However, such a Hawthorne effect  is likely to be limited, for a number of reasons. The observer was not involved in the HIV treatment program and was not known to the HHW. Moreover, the observer completely abstained from providing performance feedback to the observed HHW, even if asked for such feedback, so as to eliminate this underlying cause of the Hawthorne effect . Finally, we did not detect substantial changes in time allocation patterns within individual clinics between the different observation days, suggesting that observer bias, which is likely to wane (owing to increasing HHW tolerance to being observed) or grow (owing to feedback and learning) over time, may not have played an important role in this study.
Fourth, the observed activities might not be all-encompassing, as some infrequent activities such as continued training and attending workshops or conferences were not observed. However, these activities will take up only a limited proportion of the total number of work hours over a year  and any bias resulting from excluding these activities will thus be limited.
Finally, there are several assumptions in our analyses that might hold precisely true. We assumed that the measured doctor time was only for initiations, while the counselor and nurse time was for both initiation and follow-up. This simplification could have biased our estimates, especially for doctor time, as doctors also perform tasks other than initiations, such as completing disability grant forms or looking after more complex cases (for example, suspected treatment failure, side-effects, or adverse events).
Task time allocation and HHW productivity is likely to depend on the model of HIV treatment delivery. The current delivery model may change, because South Africa has embarked on a major reform of PHC, with the goal of providing universal access to a comprehensive package of healthcare services in the public sector through a national health insurance scheme . Several initiatives included in the reform may improve the availability of HHWs. So-called district specialist teams will be recruited and deployed in all of the 52 districts in the country to improve the standard of care delivered in community-based healthcare facilities . These teams will consist of four medical specialists and three advanced professional nurses. In addition, each PHC ward will have at least one PHC outreach team consisting of a professional nurse, environmental health and health promotion practitioners, and four to five community health workers. The role of the outreach team will include health promotion and prevention campaigns, early detection and interventions for selected health problems, and support for treatment retention and adherence . These initiatives, which are currently tested in pilot studies, could help to further devolve HIV treatment to communities and homes, freeing up HHWs in healthcare facilities.
We provide a point estimate for the HHWs required for one year of treatment at the present time. In the longer run, the current HHW and ART coverage levels may themselves affect HHW requirements, for a number of reasons . On the one hand, because ART is effective in reducing mortality, the more patients receive ART in the current period, the more patients will require treatment in future periods, assuming that HIV incidence remains unchanged [58, 59]. On the other hand, ART can also effectively prevent HIV transmission [15, 50, 51, 60–62], which would lead to a reduction in the number of future cases, and thus a reduction in the number of HHWs required in future periods. Dynamic models are needed in future studies to examine the effects of the mortality-reduction and transmission-prevention effects of ART on long-term HHW requirements. Those studies should also take into account that ART may change the type of patient needing ART. For example, ART may shift the age composition of patients on ART towards older ages [63, 64], increasing the average morbidity among ART patients and the average health worker time required for providing appropriate treatment for each patient. Moreover, the case mix of patients on ART might change over time due to increasing rates of ART failure and long-term ART toxicities. These changes could increase the average health worker time required per patient because they necessitate time-consuming counseling, ART switches, and complex treatments for the side-effects of ART .