Unlocking scal space of expanding the health workforce in district health systems in Uganda in the context of PEPFAR transition

Although expanding scal space for health worker recruitments could reduce workforce shortages in Sub-Saharan Africa, effective strategies for achieving this are still unclear. We aimed to understand the process of transitioning health workers (HWs) from PEPFAR to Government of Uganda (GoU) payrolls and to explore the facilitators and barriers encountered in increasing domestic nancial responsibility for absorbing this transitioned workforce. We conducted a multiple case-study of 10 (out of 87) districts in Uganda which received PEPFAR support between 2013 and 2015 to expand their health workforce. We purposively selected eight districts with the highest absorption rates (‘High absorbers’) and two with the lowest absorption rates (‘Low absorbers’). A total of 66 interviews were conducted with high-level ocials in three Ministries of Finance, Health and Public Service (n = 14), representatives of PEPFAR implementing organizations (n = 16), District Health Teams (n = 15) and facility managers (n = 22). Twelve focus groups were conducted with 87 HWs absorbed on GoU payrolls. We utilized the Consolidated Framework for Implementation Research (CFIR) to guide thematic analysis. At sub-national level, facilitators of transition in ‘high absorber’ districts were identied as the presence of transition ‘champions’, prioritizing HWs in district wage bill commitments, host facilities providing ‘bridge nancing’ to transition workforce during salary delays and receiving donor technical support in district wage bill analysis- attributes which were absent in ‘low absorber’ districts. At national-level, multi-sectoral engagements (incorporating the inuential Ministry of Finance), developing a joint transition road map, aligning with GoU salary scales and recruitment processes emerged as facilitators of the transition process. Overall, PEPFAR support acted as a catalyst for increasing GoU and facility-level budget allocations towards expanding the health workforce in focus districts in Uganda. Our case-studies offer implementation research lessons on effective donor transition and insights into pragmatic strategies for expanding scal space for health in a low-income setting.

Decentralized health worker recruitments in SSA is however beset by a myriad of institutional constraints. These include; delays in recruitment owing to cumbersome and lengthy administrative procedures, the ubiquitous 'vacancies but no wage bill' constraint, shortage of pay roll analysis expertise and increasingly dysfunctional district recruitment committees (Sumah and  In Uganda, decentralization was a part of governance reforms that date as far back as 1992 (Tashobya et al. 2016;Awortiwi, 2010). Health sector decentralization was formally provided for under the 1995 national constitution and further operationalized in the Local Government Act of 1997 (Tashobya et al. 2016). Uganda's district health system comprises of a district hospital and lower-level primary care health facilities (Alonso-Garbayo et al. 2017). In 2014, management of the public sector payroll and salary processing system was further decentralized to districts (Lwanga et al. 2018). District recruitment bodies known as 'district service commissions' conduct interviews and selections of health workers. However, devolved recruitment, in practice, is a shared responsibility between the districts (which declare vacancies and make recruitment decisions) and relevant central government line ministries such as Finance and Public Service which commit funds to the public sector wage bill. There exists a heavy dependence by districts on central government grants for both capital development and basic operational funds (Alonso-Garbayo et al. 2017).
Health worker shortages are pervasive at all levels of the health system in Uganda (Zakumumpa et al. 2017). In 2012, the Uganda government implemented an aggressive health workforce recruitment program known as 'the surge' in which 7,112 health workers were recruited to plug severe sta ng gaps at the primary care level (Jaskiewicz et al. 2016). Speci cally, health workers were recruited at the level of Health Centre IVs (sub-district) and Health Centre IIIs (sub-county). However, severe sta ng gaps remained at the level of district hospitals which have relatively high HIV client loads (Zakumumpa et al. 2016). In response to these sta ng gaps, and in order to accelerate progress towards HIV epidemic control in Uganda, PEPFAR developed a 3-year 'Human Resources for Health Support Program' in 2013 that was implemented in 87 focus-districts (USAID,2019). According to the implementation plan, the health workers would be initially recruited on contract by PEPFAR for a period of two years but would subsequently be absorbed into the mainstream public service as soon as scal space allowed. A total of 3,154 health workers were recruited by PEPFAR in Uganda between 2012 and 2015. Of these, 694 were enrolled onto the Government of Uganda (GoU) payroll between 2013 and 2017 (USAID,2019). An additional 1,965 health workers were expected to be transitioned to GoU in a phased manner between 2017 and 2020.
There has been increasing international assistance in addressing the human resources for health crisis in SSA in the quest to sustain public health gains registered during periods of donor support and to ensure the long-term sustainability of these outcomes during situations of declining donor aid (Micah et al. 2018). This has motivated scal space analyses by external donors and recipient governments with a view to increase reliance on domestic nancing . Fiscal space has been de ned as 'the capacity of government to provide additional budgetary resources for a desired purpose without any prejudice to the sustainability of its nancial position' (Heller, 2006). As donors like PEPFAR reduce their nancial support to workforce costs, little is known about which factors hinder or facilitate the expansion of scal space for health worker recruitments within the Government. These data are critical to understanding the dynamics involved, and strategies needed for increasing domestic nancial , there is little empirical attention to the prospect of creating scal space for expanding the health workforce in decentralized settings in low-income countries. We aimed to understand the process of health worker transition from PEPFAR to Uganda Government payrolls and to explore the facilitators and barriers encountered in increasing domestic nancial responsibility for this transition.

Research Design
We utilized a qualitative case-study research design. Case-studies are recommended for an in-depth understanding of complex phenomena within organizations (Gilson 2012;Yin, 1993). We conducted a multiple case-study of 10 districts in Uganda categorized into two a) Eight districts with the highest rates of absorption of health workers recruited with PEPFAR support dubbed 'high absorbers' cases b) Two districts with the lowest absorption rates or the 'low absorbers' cases. We then conducted a comparative analysis across the two categories of cases with regard to facilitators and barriers to health worker transition.

Case-studies selection
The ten case-study districts were purposively selected from 87 districts in Uganda which received PEPFAR support in recruiting health workers between 2013 to 2017.
Study districts were purposively selected based on secondary analyses of databases in the Human Resources Information System (HRIS) and a locally-based international PEPFAR implementing organization's databases of health workers recruited between 2015 and 2017. From these databases, we selected districts with the highest number of health workers transitioned from PEPFAR to Government of Uganda payrolls. Table 1 shows we selected the districts with the highest number of health workers transitioned from each of eight geographic sub-regions as de ned by the Uganda Bureau of Statistics (Iganga, Sheema, Apac, Kasese, Napak, Nwoya, Tororo and Kampala) and based on HIV burden (which was the key focus of PEPFAR support). We then selected two districts with the lowest number of health workers absorbed onto their pay roll (Nakaseke and Bushenyi). Each of the two districts had absorbed only one health worker since 2013 when PEPFAR's health workforce transition program commenced in Uganda.

Conceptual framework
We utilized an implementation research lens (Proctor et al. 2009) to better understand the process of transitioning health workers from PEPFAR to Uganda Government payrolls and in order to explore the facilitators and barriers involved in this process. More speci cally, we adopted the Consolidated Framework for Implementation Research (CFIR) as the analytical framework underpinning this study (Damschroder et al. 2009). The CFIR is a 'meta-theoretical' framework that was informed by earlier implementation research frameworks and is derived from a robust systematic review of factors in uencing effective or successful implementation of interventions (Means et al. 2020). The CFIR framework provides a multi-level analysis lens that entails 39 constructs categorized under ve 'domains' (Intervention characteristics, outer setting, inner setting, characteristics of individuals, and process of implementation). The CFIR framework guided this study in three ways. It informed the diverse range of study participants selected for this study especially those involved in the transition of the contract workforce onto the public sector payroll. It helped in constructing our qualitative interview guides during data collection and provided an overarching deductive thematic framework for our synthesis and interpretation of study ndings and in their presentation (Means et al. 2020).

Data collection
In keeping with the CFIR framework's multi-level analysis lens, we selected study participants involved in the transition process at the policy & planning, programmatic and implementation levels; a) national-level policy & planning actors e.g. sector ministry o cials ( Ministry of Health, Finance and Public Service) and PEPFAR as well as its 'implementing partner (IP)' playing the overall national coordination function of overseeing the transition process b) Sub-national operational-level actors (e.g. District Health O cers, District personnel o cers and PEPFAR implementing organizations at the sub-national operational level) c) Facility-level actors (Hospital administrators and the Principal Nursing O cers (Head Nurse) and transitioned health workers across diverse cadres. The category of participants is shown in Table 2. We conducted 15 face-to-face Key Informant Interviews (KIIs) with national-level actors who had 'insider' insights into the health worker transition process right from inception (e.g. during signing of MoUs) such as the overall coordinating PEPFAR implementing organization, program o cers in the United States embassy in Uganda and high-level technocrats in the line Ministries of Health, Finance and Public Service who were directly involved in the inception meetings and in the consensus building between PEPFAR and the Uganda Government around absorption of health workers after their two-year contract period.
We then conducted 24 in-depth interviews (IDIs) with district-level actors in 10 case-study districts who were directly involved in the implementation of health workforce transition at the sub-national level or with in their decentralized settings. These included the District Health O cers and District Human Resources O cers. The interview guide used in our interviews was constructed around the ve major CFIR-derived domains (Process of implementation, Intervention characteristics, outer setting, inner setting, Characteristics of Individuals). This overarching framework helped in eliciting the facilitators and barriers to health worker transition. Data were collected over two rounds. For the 'high absorber' cases, data were collected between June and September 2018 (round 1) and January to March 2020 (round 2) among the 'low absorber' case districts.
Twelve focus group discussions (FGDs) were conducted with 87 health workers who were transitioned onto the public pay roll in the case-study districts to better understand transition enablers and barriers from their perspective. The FGDs were conducted on-site at 10 district hospitals. The interviews were conducted in English by HZ and JR with the assistance of four research assistants who operated the recorder and took notes.
To augment respondent data, we conducted a desk review of relevant documents such as 'PEPFAR's Human Resources for Health(HRH) Support for Recruitment -Implementation plan of April 2013'. We reviewed written memos from two central government line Ministries of Public Service and Health addressed to district local government leaders urging them to absorb the health workforce recruited with PEPFAR support.

Data analysis
Qualitative data were analyzed in line with the procedures recommended by Miles and Huberman (1994). Interviews were recorded in English, transcribed verbatim into text transcripts by four research assistants. Data were analyzed, in an iterative process, involving four major steps; a) Data familiarization: HZ, JR read the interview transcripts multiple times) b) Developing a coding framework: We adopted the ve CFIR-derived domains (Intervention characteristics, outer setting, inner setting, characteristics of individuals, and process of implementation) as an overarching deductive thematic framework as well as inductively, from the data (Fereday and Muir-Cochrane, 2006) c) Data abstraction of the coded data into thematic categories while engaging in a constant comparative analysis across the two categories of cases of 'High absorber' and 'lower absorber' districts (Glaser and Straus, 1999) d) Overall interpretation and synthesis: The nal analyses were reached by consensus in a process involving at least four of the authors.
In addition, as shown in Table 3, we adopted the recommended procedures for ensuring rigour in casestudy analysis (Gilson et al., 2011). We spent at least two weeks in each of the 10 case-study districts. Multiple on-site visits were made to facilities hosting transitioned health workers and spent engaging in informal discussions with facility in-charges, conducting formal, face-to-face interviews with multiple informants, reviewing documents and formal reports relating to the PEPFAR transition process. . Use of theory We adopted the meta-theoretical Consolidated Framework for Implementation Research (CFIR) as a guiding conceptual framework for this study. Case selection We selected 10 (out of 87) districts in Uganda that received PEPFAR Human Resources for Health support. We selected eight districts with the highest absorption rates of PEPFAR health workforce and two with the lowest absorption rates. Sampling In line with the multi-level analysis lens of the CFIR framework, we selected participants from national-level actors (e.g. high-level central government ministry actors) sub-national actors (e.g. District Health Teams and Regionalbased PEPFAR agencies) and facility-level participants (Hospital Administrators, Head Nurses). Selected participants span across the policy & planning, programmatic and implementation spectrums.

Multiple methods
Multiple methods were used including face-to-face interviews, focus group discussions with transitioned health workers, secondary analyses of (HRiS) human resources information system data bases, document review and informal engagements with District Health Teams. Triangulation Case studies were constructed based on multiple data sources (e.g. interviewee data , document review and respondent in-put such as with transitioned health workers themselves). Negative case analysis Emergent themes that contradicted initial assumptions and the theory underpinning the study were identified. Peer debriefing and support Data analysis at each of the four stages involved a team-based process involving at least three authors. Across-case analyses were agreed upon by consensus involving all authors.

Respondent validation
The initial synthesis report by the investigators was presented before the broad-based Human Resources for Health Technical Working Group (TWG) of Uganda's Ministry of Health for their in-put based on their combined experience in the focus area.

Results
The identi ed facilitators and barriers to health worker transition emerging from this study are presented based on the ve CFIR-derived domains; a) Process of Implementation b) Characteristics of the intervention b) Inner setting c) Outer setting and d) Characteristics of Individuals.
Process of Implementation Table 5 shows the milestones in the process of implementation of the health worker transition process. Implementation happened at three major levels; a) national-level policy planning and coordination b) Subnational level programmatic supervision c) Facility-level implementation.

National-level stakeholder engagement and transition planning
At the national-level the process involved consensus building meetings between PEPFAR and Government of Uganda (GoU) high-level actors around absorption of the recruited workforce after phasing out of support. This culminated in a formal Memorandum of Understanding (MoU) between the two parties. The MoU stipulated that PEPFAR would provide funds for the recruitment process and salary support for the initial 2 years and GoU would subsequently enroll the recruited health workers on the public sector payroll as soon as scal space allowed. Inter-sector meetings were convened incorporating the relevant line Ministries of Health, Finance and Public Service. PEPFAR was represented by its overall national coordinating agency-an international NGO, which consulted with relevant program o cers at the United States embassy in Uganda. In 2013, a health worker transition implementation plan and road map was jointly agreed through a consultative process involving the two parties. Salary harmonization was key point whereby PEPFAR would pay the recruited workforce (during their two-year contract phase) salaries that were equivalent to public sector salary scales. With the exception of payment of a housing (accommodation) allowance to the contract workforce, PEPFAR's pay structure was well aligned with that of the Uganda Government.
The PEPFAR national coordinating agency continually monitored the transition process and regularly shared insights and progress reports with Ministry of Health's Human Resources for Health Technical Working Group (TWG).
At the district-level, regionally-based PEPFAR implementing partner (IPs) organizations in the 87 focusdistricts in Uganda held transition planning meetings that engaged district-level actors such as District Health Teams, Chief Administrative O cers (CAOs) and District Human Resource O cers as well as District Service Commissions (DSCs) which make personnel selection decisions. During such meetings, a transition road map at the district-level and the roles of the varied stakeholders were agreed upon. District health teams in conjunction with District Human Resources O ces determined the health worker cadres to be recruited based on the needs of individual districts. Table 4 shows that the bulk of health workers recruited across case-study districts were midwives, nurses and clinical o cers. These vacancies were advertised in national newspapers and through district and health facility notice boards. The processes of initial formal recruitment were led by the districts with the nancial support of PEPFAR provided through its regionally-based IPs. Across all districts, contract staff were vetted by the District Service Commissions to ensure that they met the Uganda public service standards for recruitment. The IPs managed contracts and payrolls during the two-year contract phase for the transition workforce. In most of the 87 focus-districts an independent PEPFAR contractor was mentioned as the personnel contracts and payroll management agency. The District Health Teams together with IPs monitored performance of contract health workers through instruments such as time sheets which were a basis for approving salaries and later on absorption. At the facility-level, contract staff were oriented in public service structures and processes by their immediate supervisors. The district health teams and facility service managers were instrumental in providing supervision and appraisal of contract staff. This formed the basis of selection of health workers on contract who were to be absorbed onto the public pay roll.

Characteristics of the intervention District wage bill budget analysis support
Technical support for district wage bill analyses was extended by the coordinating PEPFAR implementing organization to districts. This was reported as a facilitator for HW transition in 'high absorber' districts.
Although there was a widely held perception, among actors within the district administrations that their budgets could not accommodate any new personnel recruitments, technical support in scrutinizing district wage bills revealed unutilized funds in the wage bill that were subsequently deployed to absorbing the contract workforce in the transition MOU with PEPFAR.
'PEPFAR helped us analyze the wage bill budget. There was some confusion with the Ministry of Public service and Ministry of Health and here at the district. We were in the dark. So, PEPFAR came and analyzed and found that we had a balance (funds for salaries) which we were not using. At least we would nd there was something ( scal space) for recruitments' [KII, District O cial Iganga].
Conversely, in 'low absorber' districts participants reported that they did not receive technical support in wage bill analyses. Hence, donor support in wage bill analyses emerged as a distinguishing feature between the two categories of 'high absorber' and 'low absorber' districts.
However, Fig. 1 shows that even across the 'high absorber' districts there were still a signi cant number of contract staff who were not enrolled onto the public sector payroll. Wage bill ceilings limited the ability of districts to absorb a higher number of contract staff.
We observed that 'low absorber' districts had a higher number of their contract staff seconded to PNFPs (Private Not-For-Pro ts) such as mission hospitals which had an even weaker absorption capacity (at 30%) compared to district local governments (at 55%).
'The absorption has been very slow in PNFPs because these did not have money to absorb them. Most PNFPs were comfortable offering services with low cadre staff and do not have a budget to hire high cadre staff. But PEPEFAR hired these staff for the HIV response but health facilities do not have income to maintain them. So transition in PNFP is very challenging' (KII, National-Level o cial) Figure 2 shows the number of health workers absorbed between 2012 and 2017. Secondary analyses of HRIS and PEPFAR data bases revealed that over 500 of the recruited workforce were not absorbed in GoU service after transition. Across case-study districts, a number of health workers left government service before they were formally absorbed. In the focus group discussions, health workers indicated that variable delays in accessing the public payroll after their 2-year PEPFAR contracts had run out, a lack of private accommodation (especially in rural Northern Uganda) and challenging work environments such as chronic stock-outs of supplies contributed to the reduction of the pool of workforce available for absorption. Many opted for alternative employment -mostly private sector providers.

Support in convening district personnel recruitment committees
PEPFAR support helped in unlocking the inherent and long-standing organizational barriers to expansion of the existing district health workforce in 'high absorber' cases. A lack of basic operational funds for supporting the lengthy procedures required for hiring new staff was a constraint raised across all casestudy districts. District Service commission (DSCs) are standing committees that make personnel selection decisions and are meant to sit every three months. However, the DSCs were widely reported to be dysfunctional owing to a chronic inability to raise monetary allowances for paying the non-full time DSCs which are comprised of retired senior public servants. Running district job adverts in national newspapers was said to be prohibitively expensive which further impeded personnel recruitments. PEPFAR provided the necessary funding to kick-start recruitment processes in form of paying for newspaper job adverts, providing monetary allowances to DSCs and sent observers to meetings where job interviews were conducted which enhanced transparency and objectivity in the selection processes. Transparency in recruitment of the transitioned health workforce A number of health workers reported that before the PEPFAR intervention, DSCs had a reputation of questionable objectivity in the selection of personnel due to a widely-held perception that nepotism and bribery were common in district personnel recruitment decisions. Given this context, the selection of PEPFAR-supported health workers through transparent and merit-based processes lent special legitimacy to transition workforce which enhanced their absorption prospects into public service. District and facilitylevel managers perceived PEPFAR-supported personnel as having been recruited through rigorous and objective procedures.
'I look at it as a good strategy for recruiting staff. This issue of our local politics of you are going to recruit this one's daughter (nepotism), you are going to solicit bribes… those ones didn't surface anywhere. It was a puri ed process that government didn't have any reason whatsoever to object to their absorption. Someone recruited by an NGO interested in health you can't doubt their quali cations, you can't doubt their capabilities and then I think it also eliminated this issue of tribalism (ethnic biases) in recruitments [KII, District O cial, Tororo].
The transitioned workforce was perceived as competent and suitable for absorption into government service. The two-year contract phase funded by PEPFAR allowed facility-level managers to identify resilient and dependable HWs for absorption. In addition, this phase also provided HWs with an opportunity to be inducted and initiated into government systems and work environments.
'When they came the health workers on contract exhibited professionalism in their work. They were good people and immediately, they started working. The quality of service, was realized by the community. I think there is a visible change in the hospital since they came in [KII, District O cial, Apac].
Outer setting

Multi-sectoral engagements in transition process
At an institutional level, multi-stakeholder engagements involving actors at the national, sub-national and facility-levels were identi ed as a major transition facilitator by participants in 'high absorber' districts.
At national-level, PEPFAR was involved in multi-sectoral engagements of high-level actors with authority for approving health worker recruitments in relevant central government sector ministries such as 'We had several interactions. Ministry of Health invited us. As a district, we are supposed to implement Ministry of Health policies. The policy was such that PEPFAR would recruit those health workers on contract and with time, the districts, with help of Ministry of Health and Finance would avail a wage bill to absorb them. So that was the understanding. First between Ministry of Health and PEPFAR, then we as implementing partners as districts and local government. That's how we came on board' (KII District O cial Sheema).
Crucially, PEPFAR worked within established Uganda Government recruitment process and structures.
Districts determined the cadres that would be hired based on their needs. District Service Commissions (DSCs) made the ultimate hiring decisions. This lent legitimacy to the cohort of health workers recruited with PEPFAR support.
'All recruitment of contract staff was done by district service commissions (DSCs). So when it comes to absorption, such health workers are regularized because they were already recognized as legitimate staff hired through competent structures' [KII, National-Level o cial].

Inner setting
Prioritization of health workers in district personnel recruitments The prioritization of health workers in district personnel recruitments was a key distinguishing feature between 'high absorbers' and 'low absorbers' cases. In 'high absorber' districts such as Kasese and Sheema, participants were unequivocal in relaying the notion that their district administrations deliberately prioritized the health workforce in recruitments. In the 'high absorber' cases, whenever some scal space in the district wage bill emerged, slots for health workers were 'ring-fenced' as the overall priority-taking the shape of an informal recruitment policy.
'In fact we had to trade off some cadres, those ones who were not extremely needed or useful we had to keep them off in order to bring in the more useful staff like the midwives and clinical o cers' (KII, District O cial, Apac) 'You may have the wage bill but how are you going to prioritize the cadres of peoples you are going to recruit? You may say my entire Health Centre IIs need a security guard. You may recruit like 20 porters. I know they are needed there but is it a priority? (KII, District O cial, Sheema).
Although we found that 'high absorber' districts prioritized health workers in their wage bill, national-level informants reported that this was further reinforced by formal written memos from central government Ministries of Public Service and Health to the district political and technical leadership asking that they prioritize the absorption of PEPFAR-supported workforce in the available wage bills .These memos were written in March 2013 as a result of the protracted engagements by PEPFAR and high-level actors in sector ministries which were in line with the jointly developed health worker transition road map.

Characteristics of Individuals Presence of transition 'Champions'
The presence or absence of transition 'champions' differentiated between 'high absorber' and 'low absorber' districts.
Whereas 'high absorber' districts reported the presence of internal transition 'champions', their absence in participant discourses in 'low absorber' districts was unmistakable.
Transition 'champions' were individuals who went above and beyond the call of duty of their positions to promote the absorption of health workers onto the public pay roll. These champions were reported at both the district and facility-levels. The presence of champions at multiple levels created synergies in promoting health worker absorption in 'high absorber' districts. The frequently cited champions include in uential actors such as Chief Administrative O cers (CAOs) of host districts, District Health O cers (DHOs) and Hospital administrators who actively pushed for the recruitment and absorption of health workers and enrollment on the government payroll.
'We had a smooth transition because the team in XXX (District) is very proactive. They don't operate like they are in government. The CAO (Chief Administrative O cer) was an experienced man so he was quick to come in and push the recruitment process along. Much more than it normally is. The District Human Resource O cer was very active. They did their work in a timely way and actively pushed to have the HWs absorbed. Issues of health workers were prioritized. Actually, the rst batch of health workers we even gave them appointment letters before their contracts were over' [KII, District O cial, Iganga].
Champions tirelessly worked to expedite processes in the context of the typically lengthy administrative procedures in the Ugandan public sector. They acted as 'persistence enhancers' for health workers and even appropriated district nances to create 'stop-gap' monetary allowances for health workers before they were able to access the public payroll.
Actors at the facility-level were frequently cited as transition champions. Facility in-charges were motivated by a need to avoid losing skilled health workers who had been posted at their health facilities as contract staff. As such, they were instrumental in ensuring timely appraisal of contract health workers but also engaged in active follow-up with the District Service Commissions at the district administration headquarters for absorption of health workers to avoid losing their contract staff. Facility in-charges in 'high absorber' districts actively engaged their transition workforce in activities such as surgical camps and community outreaches to enable them secure some eld monetary allowances to sustain them as they awaited enrolment on the payroll which was characterized by prolonged delays in several of the case-study districts.
'The salaries could delay for two to three months.
We have PHC (primary health care) funds earmarked to this facility. We used some of this to buy them basics such as soap and sugar that could also help them to persist and endure' [KII, Facility in-charge, Nwoya District].

Discussion
We conducted a multiple case-study of 10 districts in Uganda to better understand why they had variations in absorption rates of the health workforce transitioned from PEPFAR payroll support. We found distinguishing features between the two 'low absorber' districts and the eight 'high absorber' districts. We found that in the latter cases, conducting a wage bill analysis of district budgets to discover unutilized funds, the presence of transition 'champions' and prioritizing health workers in the available district wage bill differentiated them from the 'low absorber' districts where these attributes were absent.
At an institutional level, multi-stakeholder and multi-sectoral engagements, agreeing on a joint transition road map and PEPFAR's alignment with Uganda government pay scales and recruitment processes enabled over 694 health workers to be added to the public sector payroll. However, limitations in district wage bills, prolonged delays in enrollment onto the public sector payrolls and a lack of accommodation for transition health workers were common across all districts. In this study we found that Private Not-for-Pro t (PNFP) facilities had lower absorption rates of the workforce transitioned from PEPFAR support at (30% absorption) compared to district local governments with a 55% absorption rate.
Our study adds to accumulating calls for government support to the private sector in bolstering human resources for health including in payroll support (Zakumumpa et al. 2016)  The adopted ve domains of the CFIR framework were helpful in providing a broad deductive framework for our overall synthesis and interpretation of study ndings as well as in their presentation-we note that although the framework categorizes into ve domains, some of our ndings appeared to cut across more than one domain. For instance, we found that the prioritization of health workers in district wage bill commitments derived from 'inner setting' priority-setting but was re-enforced by 'outer setting' factors Our study does however highlight the potential in uence of 'change agents' in driving health system reform and in unlocking scal space for health in a resource-constrained setting. At an institutional level, we found that the presence of transition 'champions' at multiple levels including within district governance systems but also at the facility-level was a key enabler of increasing budgetary allocations for expanding the health workforce in Uganda. We nd that PEPFAR support had a 'trigger effect' that synergized the role of internal 'champions' in promoting health system strengthening. In in uencing health system strengthening in Uganda, PEPFAR can be said to have been acting as a catalyst. The role of 'external change agents' is recognized in implementation research (Means et al. 2020

Conclusion
Overall, PEPFAR's Human Resource for Health support acted as a catalyst for increasing GoU and facilitylevel budget allocations towards expanding the health workforce in focus districts in Uganda. Our casestudies offer implementation research lessons on effective donor transition and insights into pragmatic strategies for expanding scal space for health in a low-income setting. Declarations