Our findings suggest that health workers did alter their behaviour in response to the intervention. However, to shed further light on how the intervention worked, two questions warrant further discussion: Why did health workers focus on coercive methods to increase deliveries at facilities? And why didn’t the local authorities pay bonuses according to the set targets as outlined in the directives from the Ministry of Health and Social Welfare?
Change in health worker behaviour
Although the payments ended up not being performance based, health workers were told that they would be, and they responded to the intervention much as the theory predicts that they would: They increased effort in areas where they would receive payments (with a focus on quantity), while strategies to improve the quality of care (for which they were not rewarded) appear to have been largely ignored. Similar results were also reported in a P4P impact evaluation study on maternal and child health in Rwanda where the highest improvements in indicators were observed for those indicators with the highest expected payment [
12].
Health workers reported numerous strategies to attract women to the clinics. One of the strategies they reported was to offer better services, educating women about the benefits of delivering at a clinic and positive attitudes toward women attending the clinic. The use of positive strategies to persuade women to give birth at clinics may certainly increase the number of facility births as some studies in Tanzania have found negative staff attitudes to be among the barriers against facility deliveries [
36,
37]. However, attracting more women to deliver at facilities does not necessarily ensure that the clinical quality of care is high. Due to the difficult circumstances under which many health workers work - lack of equipment, medication, and skills [
38] - it may be difficult to deliver high-quality care even if they want to. In fact, prior to the intervention, health workers were afraid that they would not be able to deliver a high quality of care due to a lack of equipment.
Health workers not only engaged in positive strategies to induce women, coercive strategies were also employed, and these strategies were copied between clinics. While other studies have noted that incentive structures may entail coercive strategies, [
39,
40], to our knowledge, our study is the first to identify coercive practices against home birth in connection with a P4P scheme in Tanzania. Health workers threatened to fine women who did not give birth at a facility or to withhold vaccinations and/or clinical cards for babies who had been born at home. While health workers in Mvomero claim that the threats were meant to “scare” people and that they have not been carried out in actual practice, the district health administration argue that these threats may have kept women from coming to the clinic for postnatal care and vaccinations of their babies after having delivered at home, because they were afraid of being fined.
The effectiveness of both positive and coercive strategies on health outcomes crucially depends on the quality of care offered, which was not directly targeted by the intervention. During our fieldwork in 2011, a woman died of untreated eclampsia because the medical attendant at the dispensary where the woman came to give birth did not recognize the signs of preeclampsia. Since poor quality may be an explanation for why women choose not to deliver at the clinic, health workers could potentially improve the quality to attract more women. However, if health workers offer a low quality care because they are unable to deliver a high quality, coercive strategies forcing women to deliver at the clinic may be perceived as the only strategy they have at hand. Such strategies need not improve health outcomes.
Observe that even if health workers reported that they had implemented various strategies to attract women to the clinics, the strategy that would involve the least amount of effort would be to forge the numbers. Before health workers received payments, they were worried that a convenient strategy to fulfill the delivery indicator would be to forge the numbers. After having received the P4P payments, however, health workers were adamant that this did not take place. Certainly, health workers had a vested interest in making this argument. An assessment of the donor-funded pilot undertaken in the Pwani Region suggests that verification and data validation has been a problem and that forging has indeed taken place [
41].
Why wasn’t P4P implemented in the way that it was designed?
Contrary to the directives from the Ministry of Health, the district administration ended up paying bonuses to all facilities at a flat rate, which means of course that the payments were not performance based. The district administration’s explanation for giving a flat rate was that all health facilities did well in at least one indicator, and that they would be “stricter” if P4P was to be continued. In our view, two factors were behind this decision.
First, as documented in an appraisal study commissioned by donors in 2009, the Health Management Information System (HMIS) in Tanzania is too weak to handle a proper monitoring of a P4P programme [
2]. A weak health infrastructure has been identified as one of the major threats to the effectiveness of P4P [
40]. We witnessed that the health administrators in Mvomero were too busy with other work to take on the additional burden of monitoring P4P. Hence, for an under-resourced district such as Mvomero, monitoring P4P with the help of the district’s own resources probably proved impossible.
Second, the district administration argued that a flat rate was given since the P4P initiative was to be a financial motivation (motisha) for the health workers, and that all had done well with at least one indicator. Under P4P, health workers are rewarded for a change in indicators. The extent of change a single facility is able to achieve will be a function of factors both inside and outside health worker control. For a variety of reasons, some facilities will find it difficult to increase the utilization of services (i.e. long distances/lack of transport, strong preferences for home birth among the local population, etc.), while others may find it easier. In clinics where the know-do gap is high and low utilization is explained in part by a low motivation, there may be a high potential for achieving an increase in indicators.
The extent to which bonus payments are perceived as fair – even in cases where they depend upon factors that are partly outside the control of the health workers - depends upon the type of fairness principle that prevails in a particular culture. Based on different ideas about what individuals should be held responsible for, notions of fairness can be divided into two broad categories: libertarianism and egalitarianism [
42]. According to libertarianism, performance-based incentives will be viewed as fair since individuals receive payments based on how they perform in relation to a predefined target. On the other hand, according to strict egalitarianism, P4P will be perceived as unfair since people should not be held responsible for factors such as talent or other external aspects outside of their own control. An egalitarian fairness principle may stand strong in socialist societies, where individual behaviour is believed to be shaped by society, and inequality as such is a function of an unfair societal structure rather than being due to any fault of the individual. By contrast, in capitalist societies where ideas of liberalism and libertarianism stand stronger, pay for performance may be viewed as a fair way to differentiate income, which according to libertarianism should be distributed in relation to effort and talent (see Nozick [
43] for a defence).
Due to its sociocultural setup and political history, Tanzania is characterized by an egalitarian mode of thought. The local government in Mvomero did not appear compelled to reward good performance and withhold money if the performance was not optimal; instead, they stated that they wanted to display gratitude for what they saw as a positive effort. In addition to the practical limitations mentioned earlier, it is possible that the social relationships between local policymakers and health workers made it difficult for the health administration to pay some facilities less than others even if they had not met the set targets.
Because principles of fairness vary across cultures and countries, P4P may be viewed as fair in some places, while being perceived as deeply unfair in other settings. The literature reports a number of cases where health workers are demotivated by P4P programmes because the rewards are perceived as being unfairly allocated [
40]. When we told our informants that the bonuses had been given at a flat rate that disregarded actual performance, none of them argued that this was unfair; nor did informants contest the practice of distributing the bonus equally to all staff members at the facility. One reason for this may be that we focused on nurses. There is the possibility that medical doctors at the three hospitals in the district would have expressed more negative attitudes toward a system where they receive the same bonuses as non-clinical staff.
If it is correct that an egalitarian fairness principle shaped the way that the locally funded P4P was implemented, then P4P may work very differently when implemented by local agents compared to when it is implemented by external entities. The donor-sponsored P4P in the Pwani Region is being piloted and led by professional external agents. Based on our experience from the locally funded P4P in Mvomero, we expect that the intervention may work very differently if/when it is taken over by the government in its entirety.
Our study has a number of implications for the design of P4P in Tanzania and other low-income contexts, and for the studies of such schemes. Firstly, contextual factors affect the nature and operation of a P4P scheme (such as a country’s sociocultural context) and it is therefore important for these factors to be taken into account during the design to achieve sustained results. Secondly, in cases where a P4P design is supply-side oriented (such as the Tanzanian case), there is need to accommodate and promote community views and participation in order to safeguard against coercive practices. Thirdly, our study to a certain extent implies that P4P is prone to adverse effects when introduced in a context where systems constraints are substantial. Finally, the above points suggest that studies of P4P impact on beneficiaries should not only measure utilization of services, but include both health outcomes and a qualitative element, as paying for increased utilization does not necessarily improve health outcomes.