Document(s) 4 of 10

When TEHIP began its operations in Tanzania, its management team set up shop in a large room in the newly renovated offices of Tanzania's National Institute of Medical Research in Dar es Salaam. This made TEHIP a neighbour of the Ministry of Health, the Tanzanian offices of the World Health Organization, and the Tanzania Public Health Association -- appropriate circumstances for a project that sought to influence the formation of health policy at both the national and international levels. This cluster of offices -- a nerve centre for the health sector in Tanzania -- is a short distance away from the shoreline where the turquoise waters of the Indian Ocean meet the city of Dar es Salaam.

More noteworthy than the natural setting, however, was the scene inside the project's office. In contrast to the established norms in Tanzania (and in many other countries around the world), TEHIP opted not for individual offices but for an "open concept" plan. This was much more a reflection of management philosophy than an indication of a particular taste in interior design. With no internal walls and no doors (just chest-high partitions separating desks), the physical form of the office made it easier for a multiplicity of contributors to work together, shoulder-to-shoulder, toward a common goal, using diverse but complementary means. This open concept design, in short, facilitated the practice of "comanagement." Within that fluid office space, Tanzanians and international staff, financial and administrative workers, big-picture planners and "details people," researchers and development workers could trade ideas and enlist each other's expertise and support. Meetings did not have to be scheduled -- conferences would spontaneously occur as managers turned to face (and raise questions with) their coworkers. This constant exchange of information kept managers "in the loop" about the project in general and about what was happening outside their immediate area of expertise. The open office space also made it impossible for individual agendas to undermine common goals; impossible that "factions" should emerge and compete against each other for influence or dominance.

TEHIP also hoped that its management style -- as embodied by the architectural innovations at its project office -- would provide an example of what should happen at the district and community levels. The unique organization of the office demonstrated to other contributors within the health system that the formation of cohesive, functional teams could lead to better results, as a wide range of viewpoints and specialties were brought to bear on individual problems. The management team also tried to show through its organizational preferences that minimal resources could be deployed in a way that produces results cost-effectively. For example, the managers who shared a common office space also shared a vehicle. As well, having a program of proper maintenance of office equipment -- rather than simply replacing computers, for example -- delivered a strong message that the right systemic support could help extend the life and impact of precious investments.

In short, far from being a typical development project, TEHIP sought to be a uniquely collaborative joint venture. In many specific respects, its modus operandi was that combining different skills, talents, and perspectives could provide greater strength to the overall enterprise, leading to better results. In the pages that follow, we examine two critical ways in which a diverse range of participants were brought together within new and innovative frameworks.

Integrating research and development

By integrating its research and development aspects into a cohesive, functional whole, the project sought to strengthen and increase the effectiveness of both of those individual facets of the program. The development side -- which assigned funds to, and encouraged the implementation of, interventions designed to make district health systems more effective -- was strong because its decisions were based on current research, including continuous feedback on how well innovations were working and how they might be improved. Meanwhile, the research side had an immediate policy impact and enjoyed a higher than normal level of credibility within communities because researchers were directly involved in a tangible effort to improve health in the districts. Therefore, researchers were seen not as pursuing an abstract agenda, but as direct contributors to better health.

This practical orientation of TEHIP's research component was exemplified by its use of a "plausibility design." This is distinct from the classical scientific model that painstakingly sets up experiments that control for confounding to prove definitively a relationship between a specific cause and a specific effect. Functioning within a living, dynamic health system -- where the process of change was already underway, and where a multitude of uncontrollable real-life influences could affect health indicators -- TEHIP would never be able to ascribe any positive changes in health outcomes exclusively to the policy changes and interventions it championed. However, by retrospectively comparing health outcomes in the districts with those in neighbouring districts -- and by assessing what role other factors such as rainfall disparities, localized disasters, and resource inequality may have played in changes in those outcomes -- researchers were able to test the "plausibility" that specific policies contributed to the greatly improved health pictures in Rufiji and Morogoro.

Dr Peter Kilima, former Director of Preventive Health Services at the Ministry of Health, uses an evocative metaphor to describe the integration of research and development. "A butterfly cannot fly with one wing," he says. "Likewise, development cannot be achieved without research."

How did the "two wings of the butterfly" assist one another? First, the research wing identified major health issues and budgeting priorities. It also provided the foundation for the development of new management tools designed to help the DHMTs base their planning more upon evidence. Then, the development wing applied specific expertise to the development of new information tools and strategic interventions, and provided the budget so that district teams could implement these innovations. Following this, the research wing helped to identify any weaknesses in the tools and new strategies, thereby assisting in their refinement. The development wing adjusted its work to accommodate the new information, and so on. In this way, the butterfly -- with its mutually reinforcing and mutually dependent wings -- moved forward more quickly and effectively than might otherwise have been the case.

There are indications that this approach of integrating research and development has led to positive changes in the districts. For example, in the post-TEHIP era, the DHMTs have commissioned their own research. This suggests that rather than viewing research either as a threat or as an unhelpful distraction from important daily work, the district teams now acknowledge that research can inform their development initiatives in ways that lead to better results and improved health outcomes.

The consortium approach

Enlisting the expertise of the best local researchers -- skilled in a wide range of specialties -- was also a key characteristic of the project's approach. In keeping with IDRC's philosophy, it was always understood that Tanzanians, who were familiar with the country and the realities of working there, would conduct the research. In addition, this collaborative approach suggested that those researchers should be organized cross-institutionally as "consortia." The formation of those consortia began following the call for proposals, when researchers representing the full range of academic institutions in Tanzania were invited to a briefing at a hotel in Dar es Salaam. The assembled researchers organized themselves into several consortia, with each one subsequently being granted small amounts of funding to prepare proposals addressing various health care issues.

During the peer review process that followed, TEHIP's international scientific advisory committee judged the completed submissions according to the applicants' tactical approach to answering the questions -- did they suggest innovative, cost-effective ways to tackle the research problems? The committee then selected two successful consortia that included researchers drawn from various institutes in the University of Dar es Salaam, the Muhimbili University College of Health Sciences, and the Ifakara Health Research and Development Centre.

Beyond the advantage to the project of being able to recruit the best people, the consortium approach offered several benefits for the Tanzanian research community. For one, it was most Tanzanian researchers first experience of working inter-institutionally and programatically, where individuals had control over a particular piece of a larger program's research agenda and depended on colleagues in other centres for part of their data. This approach -- with people assembled from different departments and different institutions -- helps develop trust, professional rapport, and an ability to bring different perspectives to bear on a common problem. It also provides good experience with the "centres of excellence" style of research support that has become common for broader, large-budget research programs in the developed world in recent years.

The research begins

The Tanzanian research consortia moved into the field armed with three core questions (as we observed in Chapter 1) that focused their efforts upon the practicability of taking a more evidence-based approach to health care planning.

• How and to what extent can Tanzanian district health plans be more evidence-based?
• How and to what extent can evidence-based plans be implemented by decentralized district systems?
• How, to what extent, and at what cost can such evidence-based plans have an impact on population health?

At the level of everyday practice, however, there was the need for more specific inquiry about how Tanzania's districts planned, how communities related to the health system, what mortality rate existed in the districts, and what could be done to reduce levels of mortality. To address the need for this information, work was divided into three thematic areas, each supported by a consortium of researchers:

• Health systems -- how the district health planning process worked;
• Health behaviours -- how the population sought health services; and
• Health impacts -- what happened to patients as a result of ill health.

Proceeding simultaneously, the research areas (described below) fed into a broader understanding of how evidence on mortality could lead to more effective health system planning in the districts. Research in these areas also continued during and after the introduction of innovations into the health system. Following the development of the planning tools, the deployment and use of those tools by DHMTs, and the introduction of other innovations within the health system, the research consortia would ask to what extent these innovations were achieving their goals and how they could be modified or improved. With the district
health systems in a continual state of transformation, it was the researchers' job to assess the impact of changes on the development side and to continually ask "where do we go from here?"

Research on health systems planning processes

A health systems research consortium was contracted to conduct both quantitative and qualitative studies of the systems and services operating in Rufiji and Morogoro districts and document changes over time. The essential objective of this research was to determine how DHMTs could use locally generated information on mortality, cost-effectiveness, health system capacity, and community preferences in their planning and distribution of health resources. The consortium began by observing and describing which district officials were responsible for planning and how they went about that task. Later, as the project offered DHMTs a series of planning tools to help them plan more on the basis of evidence, the research consortium shifted its focus to examine what impact those tools actually had on planning. Did the availability of the tools lead DHMTs to alter their funding priorities, or did it remain business as usual? Did the tools dispel previous biases in health care planning? Were the tools easy to use? How could the tools be modified to become more effective? What new tools might be developed to deal with unaddressed concerns and unmet needs?

Research on household health-seeking behaviours

This consortium -- with team members drawn from a wide range of health disciplines such as anthropology, demography, sociology, systems analysis, economics, and epidemiology -- examined a range of issues related to the extent and ways in which citizens used the district health systems. This area of research was based on the idea that even the most dramatic improvements to the health care system would not lead to better population health if clients stayed away from health facilities. The team sought to answer a number of questions. How frequently do people go to health facilities? What makes them decide to go to a facility and what keeps them away? Is there a preference for traditional or modern health care? What do clients think about the quality of care at their local health facility? As innovations were developed and put into use -- such as the Integrated Management Cascade (see Chapter 3), designed to increase efficiency and boost local facilities' access to drugs, lab tests, etc. -- researchers sought to determine whether levels of use and satisfaction increased.

Some research into when and why clients go to a health facility produced startling results. For example, it had been known that although most people are likely to seek modern care for most forms of fever and malaria, they also generally associate late-stage, life-threatening malarial fever with convulsions -- known as degedege -- with evil spirits and changes in weather rather than with malaria. Therefore, at the point at which malaria is most deadly, many Tanzanians were believed to seek out traditional healers rather than modern care. The research showed that this understanding was exaggerated and that most patients did seek modern care as their first resort. The problem now becomes one of making sure this care is of high quality and obtained quickly, rather than wasting efforts on changing beliefs about traditional care.

Research on health impacts

The objective of this research component was to quantify changes in the burden of disease, as measured by mortality, in the two districts under study. Central to the research was the use of Demographic Surveillance Systems (DSSs), which had been set up to gather vital data and track the health of large segments of the populations of Rufiji and Morogoro. Evidence generated by the DSS (which is described below) allowed district officials, firstly,
to identify the major contributors to mortality that should be addressed proportionally in the health budgeting process. At the latter stages, the continuing flow of evidence allowed DHMTs to gauge what impact health reforms had on population health.

The Demographic Surveillance System

The "evidence engine" powering health reforms

Functioning as the "evidence engine" that provides the raw materials -- the raw data -- for the tools, the DSS is a system that has come to play a key role in health planning, not just in Tanzania but increasingly throughout the developing world. In countries where no vital registration system exists, the DSS can serve as a highly reliable mechanism to supply data on burden of disease, mortality, population, household size, and many other key topics.
A DSS is rooted in a sentinel area where the entire population is monitored for changes in health status. But the impact of DSS activity is felt much more widely, with information from these sentinel areas used to create representative profiles for other districts. Urban and rural areas should ideally have separate sentinels, and a sentinel should represent only other districts that are assessed to have similar ecological, geographic, demographic, epidemiological, and socioeconomic characteristics.

Collecting, compiling, and updating population data is a massive task. This is clearly the case at the Rufiji DSS station at Ikwiriri, where station manager Dr E.A. Mwageni recently noted that he hasn't had a "computer-free day" since February 1999. The station is a beehive of activity throughout the week, often including weekends. As the nerve centre of health impact research taking place locally, the station houses the administrators and data processors who input the continuously updated health information that flows in from the field. Station personnel are fed this information from bicycle-equipped enumerators who travel to villages and households to update recent household characteristics and events. These local enumerators are responsible for two facets of data gathering: first, the collection of baseline data on precoded forms and, second, returning to every household every 4 months to update this information. "Key informants" -- leaders in the community -- preinform the enumerators when major changes (such as deaths) have occurred in the community (see box: "The growth of DSS monitoring").

The growth of DSS monitoring Since the first Demographic Surveillance System (DSS) was established in Bangladesh in the 1960s, the concept has grown as an international force to the point where DSS data is now gathered from large populations in approximately 30 locations in Africa. In 1998, the INDEPTH Network (an umbrella organization embracing 40 field sites in Africa, Southeast Asia, and Oceania) was constituted in a conference hosted by TEHIP and other Tanzanian DSS supporters in Dar es Salaam in 1998 to facilitate standardization and to promote evidence-based planning and the sharing of demographic information across borders. Increasingly, DSS is seen as essential not just in gathering and compiling household-level information on health, but also for monitoring trends related to poverty-reduction strategies, education, food security, and the environment.

Verbal autopsies

In cases where there has been a death in a household, specially trained interviewers are sent to conduct "verbal autopsies." These workers visit a household 2 to 4 weeks after a death has occurred. Deaths are also recorded in the system from biweekly visits to about 150 community key informants. Because of the sensitivity required when collecting information from bereaved families, conducting verbal autopsies is the sole task of these specialists. The autopsies follow a standardized international format and
take several hours to complete. The advantage of a thorough interview -- rich in context and detail -- is that it minimizes the chance of misdiagnosis. In cases where symptoms, such as fever, could be caused by a number of underlying conditions, the contextual information could help point to the cause of death more precisely. Regardless, rates of accuracy will vary. However, statistical analysis confirms that, despite individual variations, in aggregate, verbal autopsies provide an accurate reflection of rates of disease within the general population. Causes of death are then assigned by a panel of three independent physicians.

The evolution of the tools

The DHMTs needed practical tools allowing them to easily understand information on local causes of mortality -- the information that was collected through the DSS -- and to incorporate that information into the health budgeting process. Exactly what tools would be needed, however, was not clear at the outset. Rather, new tools were developed intermittently over a 4- to 5-year period, as research pointed to gaps in the health planning process and as DHMTs identified areas where technical assistance was required.

What was known from the beginning, however, was that managers' practical requirements were different from the demands of the academic demographers and epidemiologists who had the time and training to sift through stacks of statistical charts and tables and to interpret the numbers. District planners -- already burdened with excessive amounts of paper, statistics, and responsibilities -- needed tools that could be integrated into their existing schedules and that would simplify and streamline their ongoing tasks, rather than add complexity to their routines.

The tools were therefore designed to give planners a quick overview of what bearing local health indicators had on budgeting and planning priorities. Information presented by the tools is invariably expressed in graphical form. Colourful charts and graphs allow managers to compare information easily and quickly and are much more readily understood than tables of numbers. The idea behind these tools was to empower district officials to deal with local needs. Said Morogoro District Executive Director John Gille: "The knowledge and skills we have acquired as a team have propelled us to a level whereby we can identify our problems, analyze our priorities, and construct an evidence-based plan." In addition, since local managers were integrally involved in developing the tools -- by identifying gaps in the budgeting process and by helping to perfect those tools through use and experience -- there was less risk than might otherwise exist that the tools would be rejected as "foreign" impositions and therefore not used.

Another test of how fully the tools could be integrated into planners' routines was whether they could perform more than one function. District planners are more likely to use a tool if it can help them to solve a variety of routine problems. By contrast, having individual, discreet tools that are brought out to solve individual problems represents extra work and distraction for busy managers. In this respect, the tools can be more aptly compared to a Swiss army knife than to a hammer or a saw. Both the burden of disease profile tool and the budget-mapping tool (1 and 2 below) serve about 10 routine, minor functions in addition to the main task they were designed to accomplish. They continue to evolve with time, experience, and use.

1. District burden of disease profile tool

This tool's primary job is to repackage population health information from the DSS in a way that district officials can easily understand. It provides an annual feedback at the start of the annual planning cycle on the population's health status and, consequently, its needs in terms of health interventions. This has led to planning revolutions in the two districts. District managers now have the means to set priorities during their planning that directly address the local burden of disease.

"Before TEHIP we did not identify and prioritize our interventions," recalls Dr Harun Machibya, Morogoro District Medical Officer. "Rather, we implemented plans worked out centrally. Even in budgeting, the tendency was to add some percentages to previous years' planned and budgeted activities." Peter Nkulila, a clinical officer on the DHMT concurs: "We did things blindly."
Not knowing how allocated resources corresponded to the local causes of mortality led to waste: for example, nonmalarious highland areas might receive (through a centrally planned process) a full allotment of antimalarial drugs more appropriate for an endemic area.

The burden of disease tool has allowed local officials to correct such problems by supplying them with information in a form that makes practical sense. Although a DSS had been operating and collecting data on the burden of disease in Morogoro since 1992, this information was not fully used by district health planners until the tools were introduced in 1997. Instead of unwieldy books of charts and numbers, the tool offers simple, computer-generated graphical representations of key indicators. These are known collectively as the "burden of disease profile." This profile is updated using the same sentinel DSS and distributed to all the district planners each year at a time corresponding to the district health planning period.

This tool is highly practical because it expresses burden of disease information in terms of "intervention addressable shares." Instead of presenting the burden of disease by specific disease categories, the profile shows the proportional burden addressed by various health interventions and strategies. These cost-effective interventions form a "package" of choices available to DHMTs. This is really a "district health intervention profile" and is more relevant to district planners -- making it possible to appreciate the burden of disease in the context of health intervention priorities and the proportional use of resources to support such interventions. The tool also provides graphical representations of data on the age and seasonal distribution of deaths in the districts; place of birth or death; and health-seeking behaviour in the events preceding death. And since Tanzania has no vital statistics registry and national censuses are infrequent, the tool's provision of updated projections of population structure such as age and sex, current fertility, and age-specific mortality rates can help planners predict the number of births, deaths, infants, under-fives, pregnancies, and so on in the district in the next planning year. In turn, this can help guide the allocation of resources.

2. District health accounts tool

Also known as the "district health expenditures mapping tool," this software tool analyses budgets in a standard way to generate graphics that show how plans for spending -- or current spending commitments -- coalesce as a complete plan. Planners consulting these graphs are able to see how individual spending options translate as a proportion of their overall budget, where budget funding is coming from, and which interventions and activities are being funded.

The tool is issued annually on diskette at the start of the district planning cycle. Planners input budget figures -- representing the funding they want to allot to particular spending areas -- into a matrix of standard line items and funding sources. The computer generates graphs illustrating the overall budgetary picture that would emerge from these inputs. One of the major advantages of this tool is that it reduces the complexity and extraneous details that often make it difficult for planners to keep track of where, proportionally, their funding is coming from and where it is going. Typically, district budgets provide multiple pages of detail and contain over a thousand budgeted items and activities, with hundreds of subtotals and dozens of major line items. Not only is it difficult to discern trends within this over-abundance of information, it is also difficult to spot errors and to make changes that will correct problems without creating new ones. The tool allows for changes to be made and then instantly generates the graphics showing how those changes will affect other aspects of the budget.

The tool also addresses a number of weaknesses that have emerged in the government­donor approach to health basket funding. As figures are entered, the tool checks the inputs against the ministry's expectations. If key issues have been left out or ceilings have been exceeded, for instance, the tool will bring this to the attention of planners. In several ways, this tool helps to ensure that budgets are not going to be rejected because they are noncompliant -- any errors become readily apparent when the graphs are generated. The graphics also serve as important communication focal points in discussing the plan with stakeholders. DHMTs can also compare and contrast each other's annual plans from year to year.

3. District health service mapping tool

This tool uses a statistical database produced by the Health Management Information System in combination with an extensive catalogue of maps. The maps and the database are cross-referenced through a public domain tool developed by
WHO called HealthMapper, a free software application, which
was installed on all health planners' computers in Rufiji and Morogoro districts in June 2001.

The core function of the tool is to allow health administrators to use only a few keystrokes to get a quick, visual representation of the availability of specific health services or the attendance at health facilities for various diseases across the district. Maps provide detail down to the village level, showing roads, rivers, villages, administrative boundaries, health facilities, and schools. Those maps can in turn be overlaid by information depicting, for example, the immunization coverage in that particular area, the degree of malaria risk, and so on. Since human beings are inherently spatial and territorial, this mapped-out representation of the distribution of health indicators has a more visceral and direct impact on the reader than tables containing numbers and listings.

4. Community voice tool

"The lifeblood of the health sector reform is community participation," said Rufiji District Executive Director Faustin Fissoo. "Using their own resources, communities have actively involved themselves in initiating, planning, and managing their development projects." Noting some villages' initial skepticism about
the idea that communities could make strides without leadership from upper levels of government, Mr Fissoo relates that the community participation tool has unleashed a new civic enthusiasm, which has had many practical benefits. These range from the rehabilitation of village clinics to the realization of numerous goals generated by community members themselves.

The community voice tool was introduced in recognition of the value of communities expressing their own needs and wishes and participating in bringing these goals to fruition. Sometimes these communities' wishes are directly related to health care. Other times, the communities identify their primary concerns as areas indirectly connected with health (water supply, the condition of roads, schooling, etc.).

Using an approach known as participatory action research, animators have offered communities the opportunity to reflect on their development preferences, promoted dialogue on ways of achieving local participation in planning, and identified individuals and groups who can advance community participation. When communities are directly involved in identifying and solving their own problems, community members become a powerful force in programs of social improvement.

For example, in Rufiji's Kilimani village, elders decided to raise money through a tax on local products such as cashews, rice, fish, and timber. The new funds have allowed Kilimani to provide piped water to the community. This has had multiple benefits: not only do many residents now have showers in their homes, but there has also been a reduction in the number of women injured by crocodiles since they no longer bring their laundry to the river bank. In Bungu, meanwhile, the community has built a new dispensary to replace a dilapidated one and has formed a women's group that has taken on projects such as making bricks to build a new school.

There is no proviso that strictly limits the use of the community voice tool to the health sphere. Indeed, one unexpected consequence of these activities has been that, once communities have been mobilized to express themselves on health care planning, the focus of community discussion often shifts to broader governance issues and to other matters such as the state of schools and roads. When community voice does focus directly on health issues, it sometimes facilitates an expression of community concerns (such as the provision of dental services and of first aid for skin conditions) that are more related to "quality of life" than to the mortality rate. Community voice also sometimes promotes discussion of how the health system can provide better service and what community members can do to improve their own health.

5. Cost-effectiveness and district cost information system tool

This is the "tool that got away." Understanding the incremental cost-effectiveness of health interventions was originally conceived of as one of the most important tools in the planning toolbox. It became apparent that two vital ingredients for cost-effectiveness analysis -- costs and coverage -- were not available to district planners. So the first step was to develop a cost-tracking tool. A very practical district-level computer database was developed to capture facility-specific intervention costs from a bottom-up approach. However, it found very little variance in costs between facilities or over seasons so it did not appear practical to run it continuously. It therefore stands as a useful tool to apply periodically to estimate the actual costs of interventions. As a research tool it was very useful to understand general technical inefficiencies and noncompliances. Districts still lack practical tools to estimate actual coverage of essential health interventions (with the exception of immunization and antenatal care where denominators are known). Without such developments, incremental cost-effectiveness considerations for decentralized planners are not possible, and planners are restricted to general information regarding generic cost-effectiveness from other settings.

Moving to the next stage

The tools have provided significant support to districts attempting to reform the way they plan to meet health challenges. The introduction of these tools coincided with two other developments that greatly strengthened the districts' abilities to plan health care delivery in line with the local disease burden: a budgetary top-up that brought health spending closer to the level recommended by WDR93 and a series of "supportive interventions and strategies"' designed to improve the efficiency of health care administration in Rufiji and Morogoro. The next chapter explores in greater detail how districts have used these tools and interventions to redesign and improve health care delivery.

Document(s) 4 of 10