The amount of demand for a service is partially dependent on the degree to which the existing population has already been served. In the ranking exercises, particularly in Gaza, the baseline performance or early performance of specific indicators were mentioned as explanations of difficulty. Improving performance over time, for FP consultation with HIV+ women and facility deliveries, was difficult because many facilities achieved near full coverage (the last mile effect). The average difference in facility deliveries between 2011 (baseline) and 2014 (endline), adjusted for population growth in the seven facilities in Gaza included in our sample was only 9.4%. Health workers in Nampula rarely mentioned previous high performance as a barrier, although as the indicators improved over time they discussed how they were not able to achieve such high increases in number of patients or the resulting overall payment. Health workers reported that some services have reached ‘saturation’, whereby they felt they could not increase the number of patients eligible for a particular service.

We always advise pregnant women to give birth in the maternity ward. At our meetings, the Activistas [community-based health advocates] and traditional birth attendants always advise women to come to give birth in the maternity facilities. And PBF has helped our Actvistas focus their efforts on getting patients into the health facility when appropriate, although there are only so many women in the community- almost all are coming to the facility for delivery already. (I05, Chokwe, Gaza; Medical chief (managerial) discussing the indicator on facility deliveries).

It’s difficult to improve this indicator because... here in the district, HIV-positive women come to the facility only during pregnancy and postpartum. Then they disappear, which makes it difficult to continue to [family planning counseling] with them. (I22, Mabalane, Gaza; health provider discussing the indicator on HIV-infected women who received an FP consultation and a modern contraceptive method).

Health workers reported that for many of the maternal and child health services, the patients are perceived to be knowledgeable as to which services they need to access at the appropriate time. Many of the health workers reported that maternal patients will show up for these services without much additional advocacy or outreach efforts, meaning norms help to influence patient behavior. PBF propagated this further by influencing health worker behavior to counsel patients on when to return and the importance.

On the other hand, health workers were clear that HIV services and ART initiation and treatment remain stigmatized and often burdensome for the patient to maintain over the long-term, thus these indicators were ranked as most difficult. According to respondents, adult HIV and pediatric retention on ART was considered difficult because ‘they don’t show up’. Health workers reported that they used PBF facility funds to pay for car repairs and petrol to increase their outreach capacity and frequency to improve the HIV retention indicators, but few facilities have achieved any improvement. In Gaza, health workers ranked initiation of treatment for children 2-14 years (pediatric patients) as relatively easy because, ‘mothers were good at bringing in their children for the initial consultation and to obtain ART’. Health workers felt that parents prioritized their children’s health over their own for this particular indicator. However, retention on treatment overall remains a difficult indicator given the frequency of visits to the health facility.

Mothers may be sick, but if the child is not sick they do not come back [just for themselves]. The problem is knowledge, it is knowledge. I still think maybe [we need to present] lectures daily to mothers to explain the purpose of postpartum consultations and why they should return to the health unit. (I36, Murrupula, Nampula; health provider discussing the indicator on postpartum women who received at least one postnatal consultation 3-28 days after birth).

"At the prenatal consultation, we educate mothers about the risks facing women during labor and invite them to deliver at the health facility. The number of institutional deliveries appears to be increasing just based on the number of women we have coming and the increasing in midwives assisting deliveries at the health unit. So we are working with all the traditional midwives to build greater awareness.” (I34, Melema, Nampula; health provider discussing the indicator on facility deliveries).

For several health facilities, there were very low patient numbers for some indicators, and thus increasing the ‘achievement’ even by one or two patients is difficult. For Nampula, the improvement of ART initiation is due to the low number of eligible patients-less than 10- so increasing the number of patients per quarter by one is a significant improvement (greater than 10%). Moreover, in Nampula the HIV prevalence is significantly lower compared to Gaza (4.6% and 25.1% respectively) (Ministry of Health, 2010), hence the number of eligible patients for HIV testing and treatment are much lower and harder to locate through outreach.

Now the other issue that may also contribute to this indicators difficulty is that most children under five are already on ART, so the number of children to start on the regimen is very, very small in our communities. It is clear that this child coming in at a higher age than five years it will no longer come through our HIV pediatric program because they have already entered our program before five years, and is in our books. So by this indicator of children 2-14 years, probably will to be reduced because children are already on ART. All children that are eligible for this indicator have already been reached between us and the Activistas. (I15, Chicumbane, Gaza; health provider discussing the indicator on HIV-infected children 2-14 years of age who initiated ART treatment for the first time).

Health workers described population catchment (rural/urban) and geographic spread as factors that make acquiring a patient easier or more difficult. Poor infrastructure and lengthy distances between communities and the facility act as barriers to access; patients may not be able to afford the transport necessary, or the time away from income generating activities. This particular factor was mentioned by health workers, but not discussed in detail as health workers are not generally concerned with the length or degree of difficulty of a patient’s journey.

It’s difficult to improve this indicator because... here in the district, HIV-positive women come to the facility only during pregnancy and postpartum. Then they disappear, which makes it difficult to continue to [family planning counseling] with them. (I22, Mabalane, Gaza; health manager discussing the indicator on HIV-infected women who received an FP consultation and a modern contraceptive method).

It’s been quite difficult for these caregivers to bring their children to appointments routinely, initially because of financial constraints, distance, and so on. But when they come to the clinic, we show them we care for them, we try to explain how their child will benefit from antiretroviral treatment, that the child could, for example, attend school, enjoy childhood, be healthier with fewer opportunistic infections... to be sure the child does not abandon treatment. (I15_Chicumbane_Gaza; health providers discussing the indicator on HIV-infected children 0-14 years of age alive 12 months after initiating ART).

The complexity, including the clinical protocols, number of steps or tasks for treatment, and intensity of effort to complete treatment or services were all described as ways that indicators can increase in difficulty. Typically, maternal and child services were less complex than HIV services and pediatric HIV. PMTCT, although clinically more intensive than MCH services, was perceived as easier than adult HIV services.

We always deliver BCG after childbirth, then follow up with the other vaccines when mothers bring their children in for the checkups, and the vaccination takes very little time that most parents agree and remember to bring the vaccination booklet. We also have community campaigns for vaccination which makes it even easier. But there are a few [parents] who do not return with their children to meet the vaccination schedule. (I06, Chokwe, Gaza; health provider discussing the indicator on children who received full vaccination for BCG, DPT, polio, and measles in the first 9 months).

Initiating and retaining patients on ART is difficult from a clinical perspective. For infants in particular there are many confirmatory testing, counseling for parents and family members, ensuring access to ART medications, and the infant has to return more frequently than adult patients for checkups. An adult can choose to initiate and continue ART, but a child requires the mother’s [agreement and participation]. She must take charge of the child’s treatment. While the initiation of treatment is easy… after nine months, the initial treatment is finished, [and] the mother disappears. (I33_Mutuali_Nampula; health provider discussing the indicator on HIV-infected children 0-14 years of age alive 12 months after initiating ART).

PBF has influenced HIV recordkeeping very positively … we test all women who seek our maternity services and record the results and contacts of the tested women. If the test result is positive, [the women] starts treatment immediately. So the system work much better now. It is not complex for us to test every maternity patient, and we have increased our prenatal consultation numbers so these patients count for multiple indicators. (I01, Macia, Gaza; health provider discussing the indicator on HIV-infected pregnant women who initiated ART).

The duration of consultation is considered a driver of difficulty because health workers felt that their time was limited and workload high. Patient consultations or services that require a high level of time commitment were considered difficult to improve due to the limited time of human resources to dedicate to more patients. Health workers stated that their workload had increased due to PBF without an increase in clinical human resources, thus the time to serve a patient was a limiting factor, and services that take prolonged time to complete like HIV treatment initiation are difficult during periods of peak volume (generally in the morning hours). Facility deliveries were the one exception to the duration of consultation. Facility deliveries have long duration, but due to the demand side factors of mothers showing up accompanied by families, the low complexity (for normal deliveries) and availability of staff allow for facility deliveries to be considered easier than some of the other time intensive services.

When a new patient arrives, it is true that our mission is to start treatment, but a little counseling is necessary, which is sometimes a challenge to deliver and the patient often has to wait. Once the treatment begins, it is important to follow up, but that requires resources to do it, the Activista or a nurse who can call on the patient. Retaining patients on ART requires working hard to look for them-we know they’re out there. (I23, Chicualcuala, Gaza; health provider discussing the indicator on HIV-infected children 0-23 months of age who initiated ART treatment for the first time).

Woman who complete at least one postnatal visit three to twenty-eight days after birth is easy because it is can happen at the same time as the newborn follow-up. So the mother will bring in her baby and we send her to the maternity department for their postpartum consultation while she is here and it is less time for the MCH nurse to do both at the same time. (I26, Mapai, Nampula; health provider discussing the indicator on postpartum women who received at least one postnatal consultation 3-28 days after birth).

Adults were noted to be more difficult to maintain on treatment because of factors such as employment, migration, and stigma. Children on the other hand were more likely to be brought in by mothers. Providers noted that mothers cared more about their child’s health than their own, making child HIV indicators easier to achieve.

Woman who complete at least one postnatal visit three to twenty-eight days after birth is easy because it is can happen at the same time as the newborn follow-up. So the mother will bring in her baby and we send her to the maternity department for their postpartum consultation while she is here. (I26, Mapai, Nampula; health provider discussing the indicator on postpartum women who received at least one postnatal consultation 3-28 days after birth).

Often HIV-positive pregnant women are isolated from the rest of the community. When they learn that they are infected, [their] men abandon them, so it is difficult to come to treatment and follow-up. This is more common in rural communities. (I20, Murrupula, Nampula; health provider discussing the indicator on HIV-infected pregnant women who initiated ART).

Usually when the children become older, especially when the child is already working, coming back to the health center is not easy… so parents will only bring them when they are really sick. It’s hard to search for those not already on the drugs, and it’s difficult to follow them. Advising the parents is difficult because of the long distances between the health center and some communities. (I12, Murrupula, Nampula; health provider discussing the indicator on HIV-infected children 2-14 years of age who initiated ART treatment for the first time).

Our study demonstrates that there are factors that fall within health workers’ control and factors that lie outside it (Figure 1). PBF practitioners must take this into consideration when defining and pricing indicators. When indicator-relevant factors are beyond health workers’ control, health workers cannot improve service performance, regardless of the price. As recounted in one example, if health workers are incentivized to test positive partners but test kit stock-outs are common, health workers are unable to complete this indicator (test partners), regardless of the HIV testing indicator’s price. On the other hand, when the delivery of services for an indicator involves modifying factors within health workers’ control, then they can be incentivized to improve performance (i.e., increase the number of completed services for more patients).

PBF programs attach an incentive to a specific service to encourage greater production, in turn improving a health-related outcome (Fritsche; Soeters; Meessen, 2014). This assumes that health workers overcome constraints within the existing system, without additional inputs or changes. Incentives, however, are not able to influence all factors in the second and third ring. PBF practitioners must understand which factors health workers can control and which they cannot. For factors that cannot be controlled, PBF programs must either deprioritize those indicators (regardless of the price) or provide technical assistance in conjunction as a part of PBF programming to reduce that constraint. Increasing price, regardless of how high, does not solve many of the fundamental challenges that health workers face in their external environment (Lohmann et al., 2018).

Health workers must expend time and effort to acquire patients. The greater the amount of time and the more intense the patient engagement, the smaller the net profit for the health worker. Our study finds that time and engagement intensity depends on patient demand, disease prevalence, and the external environment that patients are embedded within.

PBF incentives alone do not enable a patient to seek or access services, nor do incentives address the potential opportunity cost that health workers face when deciding what services to provide (Mabuchi; Sesan; Bennett, 2018). To address this cost, PBF interventions should include demand side strategies that seek to change care seeking norms and reduce access barriers. Bridging the gap between the community and the facility requires far more intentional interventions than simply incentivizing health workers.

Indicators must account for the context in which they are implemented, including other interventions and ongoing programs (Bhatnagar; George, 2016). If the target population has already received services under existing interventions, or if disease prevalence is low, there may be lower patient service demand and a reduced number of available patients. PBF program designers must consider this to implement PBF in contexts where demand exists in the first place.

As in acquiring patients, when health workers decide how to prioritize services (and which PBF incentivized indicators to pursue), they factor in the associated costs. From our findings, engagement time and intensity, in relation to treatment, is a function of treatment complexity (i.e., number of steps and clinical protocol), consultation duration, and patient characteristics.

Simply increasing the price of complex services relative to easier services fails to account for the potential costs and external factors that may increase the difficulty for health workers to deliver care. As the health worker payment literature from high-income countries shows, there is ample evidence demonstrating that health workers vary both the intensity and volume of their workload based on payment. Under fee-for-service systems, where health workers are paid per service delivered, health workers are willing to deliver a greater volume and higher intensity of services as long as their net profit is maximized. In the context of PBF, health workers behave in a similar way. In many PBF programs however, such as the Mozambique program, services are priced based on what program designers or policy makers deem important based on their own values or what they view as beneficial to the target population, without considering health worker preferences. This approach ignores the basic concept of reservation utility and does not consider whether the ‘contract’ is acceptable to the provider.

Going forward, PBF program designers and implementers must consider health workers’ perception of indicators and clinical level realities that impede indicator achievement. The current guidance on indicator selection in the PBF toolkit follows the SMART goal-setting formula. The recommendation follows a central planning model that fails to understand provider preferences and motivation for service delivery. Moreover, price setting is often based on the program’s budget availability rather than consideration of the costs and difficulties associated with indicator achievement. In other words, indicator selection and pricing are ‘system-centered’ rather than ‘user-centered’.

PBF programs designers and implementers should involve health workers during the indicator selection and pricing process. This engagement should aim to better understand the production factors driving health worker motivation to deliver on incentivized services. With this information, overall reservation utility can be modeled to determine the most effective price for positive results.

Our study demonstrates that health workers are more likely to deliver services for which they can control production factors. When service-related factors are beyond health workers’ control but necessary for adequate delivery of care, PBF programs must include supplementary technical assistance programs to support their improvement.