The documented low number of health workers assigned to rural health facilities and absenteeism in this study are comparable to other findings from Tanzania and elsewhere [11, 30, 31]. However, while other studies presented individual problems related to human resources productivity, capacity or incentives packages [32–39], the current study has brought them all together to show multifaceted nature of the human resources problem. These findings have serious implications for health service provision in southern Tanzania given that no more than one-fifth of the number recommended by the Ministry of Health's own guidelines were actually employed; of those employed, about half were absent from their duty station on the day of our survey; over half of the nursing staff followed during routine vaccination days were non-productive for at least three hours of the working day; and that supervision visits by district health staff to peripheral health facilities were infrequent and of variable quality.
The Ministry of Health established recommendations for staffing levels by interviewing key informants, observational studies and consultative meetings with staff in all levels of service provision . The final criteria for staffing levels were based on the type of services provided, the type of health facility and the number of patients anticipated.
The norms might be appropriate for some places (e.g. urban dispensaries with a high utilization rate) but for others not (e.g. rural remote facilities covering a relatively small population). This may explain why the study identifies both time shortages and an inefficient use of available staff time. Accounting for service demand is crucial as utilization is likely to differ between remote facilities with lower population densities and few users compared to urban facilities with high population densities.
We found that only 14% of nurses' and 20% of clinical staff positions had been filled, lower than the national average of 35% . We noted marked variation in staffing levels between the districts in our project area. The particularly marked lack of staff in rural settings has been documented previously  and results in service delivery being predominantly provided by untrained health workers. Mæstad suggested possible incentive schemes to attract trained people to work in rural areas . "Pull incentive packages" could involve provision of hardship allowances, housing, improved management, local recruitment or clear career development plan; "push incentives" could involve implementation of coercive measures such as bonding, in which health workers are obliged to serve in rural areas for a number of years upon completion of internship. Testing how well such incentive schemes work in developing countries needs to be given priority.
Inadequate staffing levels were compounded by a high level of absenteeism which is not acceptable as it reduces access to services. Approaching a third of all employed staff were absent from their work place, resulting in only about 12% of the recommended staff actually being available at the health facility. Improved health services management is required to reduce the health workers in rural facilities being pulled in different ways - to attend seminars, to collect their salaries and sometimes vaccines or other supplies from the district capitals. Such distractions further undermine their ability to provide services. However, despite understaffing, the nurses in primary facilities did not appear to be overworked, suggesting that for preventive care there is a lack of balance between service supply and demand compared to recommendations of the Ministry of Health and Social Welfare and the internationally set requirement to attain the health Millennium Development Goals of 2.5/1000 health workers per population. Where nursing staff had been employed and were available on site in primary facilities, a surprising amount of time was non-productive, with over half the nurses being unproductive for at least three hours on a vaccination clinic day, considered to be the busiest time of the week. As observed and documented by researchers during our study, the variation in productivity was largely a function of patient flow compounded by lack of management: when patients were not present, nurses lacked the initiative to undertake other activities like filling HMIS forms or doing outreach clinics. The possible explanation could be the presence of untrained staff in primary facilities. This has an impact on quality of some services that require trained health workers for example maternal health and major issues related to HIV or non-communicable disease problems . Patients in most instances value and search for services that they perceive to be of better quality. They could by-pass primary level facilities and seek care directly from higher level facilities perceived to have high quality, leading to loss in functionality of referral systems [42, 43]. The consequence could be underutilization of lower level facilities, overload of hospitals as seen here and cases of high out-of-pocket payments for use of private facilities [44, 45]. This is likely to be particularly detrimental for the poorest, increasing poverty through spending more than the limited resources available for basic needs. To increase access and client confidence for health service requires better availability of skilled health workers, improved service management, and support to reduce absenteeism.
In the decentralized Tanzanian health system, the district Council Health Management Team (CHMT) is responsible for the health services provided in its district. Those persons in-charge of primary facilities have a role in overseeing the day to day activities of their facilities and communicating with CHMTs on various requirements related to drugs, supplies and equipments. The CHMT members are supposed to visit each facility on a monthly basis to supply commodities, review HMIS data and support front-line staff. We found that such supervisory visits were infrequent and not always supportive. Adequate supervision could reduce absenteeism and mitigate some of the factors that reduce health workers' productivity . However, CHMTs face genuine challenges in providing supportive supervision to peripheral health facilities. Many CHMTs plan a monthly supervision schedule, often found posted on their notice boards, but find it difficult to keep to it (personal observations and communications with District Medical Officers in rural districts). Competing interests lead CHMT members to attend training seminars, after which they are obliged to train front-line health workers, taking the latter away from their duty stations. Molyneux and others recommended more training in health facilities and fewer seminars in district head quarters in order to increase health workers' time for patient care and to increase the relevance of the training . Another reason for failure to perform supervision and execute other duties on a timely basis is delay in disbursement of basket funds to the districts from the Ministry of Health and Social Welfare [Personal communications with DMOs of Lindi Rural and Nachingwea in November, 2008]. Additional local factors, such as the breakdown of vehicles and unavailability of fuel, compound the situation. In addition these same people are required to manage the HMIS, look after visiting officials and health stakeholders, who often arrive at very short notice, and to contribute directly to service provision in their districts. The distribution of paperwork such as guidelines and checklists is not enough to effect change: these needs to be complemented by agreed set of priorities, budget, follow-up, audit and feedback to lead to changes and influence performance . Integrated supervision has been proposed to improve the efficiency of supervision visits as part of Tanzania Essential Health Intervention Programme (TEHIP), and this is worth taking forward . Improved supervision is likely to require timely disbursement of funds, sufficient staff, prior notification of visits, appropriate training for supervision and improved supervision of CHMTs by regional and national level staff.
Our study may help those formulating polices to alleviate human resource problems. The number of health workers can be increased by promoting the WHO approach to recruit and train local people, residents of respective cultural zones within a country, and also to use mid level providers . This will orientate health worker training and development of career incentives to encourage service in rural and disadvantaged areas to counteract the tendency of health workers to cluster around cities. The application of health worker management strategies through supportive supervision, improved supply of essential goods and integrated on the job training could reduce absenteeism and non productivity .
There were methodological limitations associated with this study. The facilities and health workers included in the time and motion study were purposively sampled. Nevertheless we believe they were representative of health facilities in the area. The time and motion study did not include private providers, where productivity patterns may differ from government providers. Although the time and motion approach is considered a gold standard in measuring health workers time use , it is subject to the so-called Hawthorne effect where what is being observed changes as a result of being observed. However this would likely result in positive bias , meaning that the documented productivity is higher in health workers under observation. We suspect the extent of this bias was reduced by the fact that interviewers carried out the time and motion study after they had spent several days at the facility, so that health workers had got used to their presence, and they used PDA technology which is less conspicuous than clipboards and pens. Another way in which the time and motion study may have over-estimated the productivity of health workers is that the study was done on the busiest day of the week, when vaccination activities were taking place.