Disentangling the impact of alternative payment models and associated service delivery models on quality of chronic care: A scoping review

Payment reforms are frequently implemented alongside service delivery reforms, thus rendering it difficult to disentangle their impact. This scoping review aims to link alternative payment arrangements within their context of service delivery, to assess their impact on quality of chronic care, and to disentangle, where possible, the impact of payment reforms from changes to service delivery. A search of literature published between 2013 and 2022 resulted in 34 relevant articles across five types of payment models: capitation/global budget ( n = 13), pay-for-coordination ( n = 10), shared savings/shared risk ( n = 6), blended capitation ( n = 3), and bundled payments ( n = 1). The certainty of evidence was generally low due to biases associated with voluntary participation in reforms. This scoping review finds that population-based payment reforms are better suited for collaborative, person-centred approaches of service delivery spanning settings and providers, but also highlights the need for a wider evidence base of studies disentangling the impact of financing from service delivery reforms. Limited evidence disentangling the two suggests that transforming service delivery to a team-based model of care alongside a purchasing reform shifting to blended capitation was more impactful in improving quality of chronic care, than the individual components of payment and service delivery. Further comparative studies employing causal inference methods, accounting for biases and quantifying aspects of service delivery, are needed to better disentangle the mechanisms impacting quality of care.


Introduction
Rising prevalence of chronic diseases is challenging health systems to respond to changing and dynamic needs [1].Governments often bear the majority of the fiscal burden, and are thus challenged with delivering high-quality services given budgetary constraints and mounting fiscal sustainability concerns [2].Improving quality is another challenge that, in fragmented healthcare systems, has been related to better coordination, person-centered and 'value-based' care [3].Purchasing has thus been one mechanism at the disposal of policymakers to incentivize improvements in quality of chronic care, given the shortcomings of classical fee-for-service (FFS) arrangements to reach this aim [3][4][5][6][7].
A substantial number of prior reviews have investigated the impact of purchasing reforms on quality, albeit with varying scopes in terms of the type of models studied (e.g.network-level payment models [8], global budgets with quality incentives [9], or mixed payment methods [10]), the type of setting (e.g.primary care [11] or outpatient care [12]) and type of chronic disease (e.g.diabetes [13], strokes [14] and cancer [15]).Given that the success of purchasing is likely to be highly contingent on the context, the contract details, and the organization of service delivery [15], some reviews have focused on particular types of service delivery models, their associated payment models, and their impact on quality, such as Accountable Care Organizations (ACOs) [16], Patient-Centered Medical Homes (PCMHs) [17] or more broadly integrated care models [18].
Past reviews have acknowledged the challenge of understanding the relationship between context, mechanisms and outcomes when it comes to financial and service delivery reforms, yet highlight the importance of doing so to better inform policymaking and the design of payment reforms [8,13,19].While payment reforms often aim to stimulate changes in service delivery, payment reforms can be (and often are) implemented alongside service delivery reforms, thus rendering it difficult to disentangle the impact and effectiveness of the two.While some prior reviews have focused on detailing both the financing and service delivery aspects of payment reforms [8,9,19], none have aimed to disentangle the impact of payment on quality of chronic care from changes to service delivery, likely due to methodological difficulties.
To better understand the mechanisms and contexts in which payment models are successful in improving the quality of chronic care, we aim to gather overarching evidence on a range of population-based alternative payment models (APMs) across settings, by considering both the financing and service delivery model.By APMs, we refer to payment models that reward healthcare providers for delivering high quality care and that deviate from FFS.We focus on population-based models, rather than disease-specific ones, given their potential for improving population health more broadly through better coordination and integration of services.However, we also exclude studies on pay-forperformance (P4P), given the extensive coverage of this model in the literature thus far.A broader focus on different alternative populationbased payment models allows for conclusions to be drawn about what works in which settings.
Furthermore, we aim to disentangle the impact of financing from changes to service delivery where possible, to better inform and understand what precise aspects of purchasing and service delivery lead to improvements in quality.We define service delivery models as different approaches to providing healthcare.These models are used to assess, plan and implement healthcare services.Based on the World Health Organization's (WHO) definition of quality [20], we take a broad view of quality, thus including process and health outcome measures, but also utilization where it pertained to populations with chronic diseases.This decision stems from the view that changes in utilization can be informative towards efficiency of care, one element of quality.We specifically address the following questions: What population-based APMs and associated service delivery models have been used to improve the quality of chronic care?What evidence is there disentangling the impact of payment models from service delivery on quality of care?

Material and methods
This article is based on a scoping review of the academic and grey literature that assessed the impact of APMs and their associated service delivery models, originally commissioned by the WHO and published as a final report [21].Given the broad focus on APMs, a scoping review rather than a systematic review was carried out in order to gather a broad and diverse body of evidence on purchasing and service delivery together.To ensure adherence to quality standards we developed a protocol based on the Joanna Briggs Institute (JBI) guidelines [22] and following the AMSTAR (A Measurement Tool to Assess Systematic Reviews) checklist, as far as criteria for scoping and systematic reviews align [23].The protocol is accessible on OSF Registries [24].

Search strategy
We identified key search terms through existing reviews on payment models, adapted these to our research questions, and refined them through trial searches.The final list of search terms is presented in Table 1.Further details, including particular MesH terms used and search strings per repository, can be found in the Appendix A1.We adopted a three-stage strategy to identify relevant literature: First, we conducted a systematic search of the five academic repositories PubMed, Web of Science, Scopus, Cochrane Database of Systematic Reviews, and Google Scholar.Second, we hand-searched the reference lists of all relevant systematic reviews identified in the first step for relevant singlestudy papers.Third, we performed a targeted search of the grey literature using Google incognito and by searching relevant organizations' repositories (OECD, World Bank, WHO).Despite this broad search strategy, ultimately, only academic articles were included in the review, as the grey literature identified did not meet the inclusion criteria.All searches were conducted in November 2022.

Screening process
The search for literature (after the removal of duplicates) resulted in 3480 resources (Fig. 1).Before starting the full title and abstract screening, two researchers trialed 20 titles, using and clarifying the inclusion and exclusion criteria (Table 2), with more than 75 % of agreement [22].Subsequently, two researchers separately screened all titles and abstracts, documented their decisions in an Excel file, compared these, and solved disagreements by consensus.In a second step, the full texts of 318 resources meeting the inclusion criteria were retrieved and a second pilot test (n = 5) was undertaken.For the full-text screening, we proceeded as in the first screening phase, resulting in the exclusion of 284 resources.The remaining 34 articles were included in the analysis.
Articles were included only if they measured quality indicators related to the management or prevention of various chronic conditions, or if the study population was narrowed down to individuals with chronic conditions.Based on the WHO's definition of quality [20] and Donabedian's model [26], we considered structure, process and health outcome measures that relate to the effectiveness, safety, person-centeredness, timeliness, equity, integration and efficiency of care.In cases of service utilization, only admissions/hospitalizations due to the chronic disease were considered as measures of quality, on the basis that they indicate avoidable admissions and are thus a measure of effective delivery of appropriate care.A common reason for exclusion was insufficient detail on the payment and service delivery model.In terms of the payment model, authors must have clearly outlined the base payment of the reform, additional financial incentives and their design, and whether participation was mandatory.Additionally, information on the setting of service delivery and the type of providers affected were needed.These criteria led to the exclusion of some articles on ACOs and PCMHs, for example, that did not highlight the payment models behind them.

Extraction and analysis
We prepared an extraction sheet based on the Cochrane Effective Practice and Organisation of Care (EPOC) resources [27] with a priori defined categories deemed relevant for our review, guided by relevant factors highlighted by Stokes et al [28].To ensure consistency, we trialed the data extraction with two researchers charting information for two selected research papers and subsequently compared the results.

Table 1
Search terms used to identify relevant literature.
Based on this trial the extraction sheet was slightly amended and categories were clarified; the extraction of the remaining articles was divided between two researchers.We extracted information as outlined in Table 3 according to the authors' findings and refrained from providing our own interpretation of the studies' results individually, including where findings were contradicting.
Given there are no quality appraisal tools for scoping reviews, we ensured the quality of articles in two ways.First, we assessed the risk of bias of each article according to the Critical Appraisal Checklists developed by the JBI for different types of research.Second, an amended version of the GRADE approach based on the type of study, risk of bias,

Table 2
Overview of inclusion and exclusion criteria.

Inclusion criteria Exclusion criteria
Themes covered • Studies examining alternative payment models (beyond pay-for-performance) with the aim of improving quality for chronic diseases  Theme Details and effect size, was used to discern the certainty of evidence across articles (Appendix A2) [29].We clustered the articles included in this review into five groups based on similarity of the type of payment models they analyzed, informed by a classification by Tsiachristas [30] based on level of financial and care integration: 1) capitation/global budget arrangements, 2) bundled payments, 3) shared savings and shared risk arrangements, 4) pay-for-coordination arrangements, and 5) blended capitation.Furthermore, a final cluster included articles that compared different variations of APMs.

Description of the literature
The scoping review identified 34 articles that met the inclusion criteria, highlighted in Table 4.In terms of country representation, most articles hailed from high-income countries.A majority focused on the United States (74 %, n = 25), followed by Canada (n = 3), China (n = 2), Germany (n = 2) and the Netherlands (n = 2).
Five types of APMs were identified in the scoping review.The largest number of articles analyzed capitation or global budget arrangements (36 %, n = 13), from which six focused on the Alternative Quality Contract (AQC) in Massachusetts.The second-largest cluster focused on pay-for-coordination arrangements (26 %, n = 9), followed by shared savings/shared risk arrangements (17 %, n = 6).Three articles analyzed a blended capitation model based in Ontario, Canada.One article assessed a bundled payment in the Netherlands.The two remaining articles compared across variations or types of models.
Most articles studied single-payer arrangements (n = 21), with public (n = 8) or private purchasers (n = 13), while 11 analyzed multi-payer APMs.The remaining articles did not provide information on the purchaser(s).As the focus of the scoping review was on population-based models rather than disease-specific payment models, the target population of the APMs was primarily the general population and narrower, such as to children (n = 1) or frail older adults (n = 1), in only a few cases.
Given the strict inclusion criteria of causal inference methods to discern the causal impact on quality of care, most articles used a difference-in-differences methodology (67 %, n = 23).Fewer used regression models with time effects (n = 8).Two used interrupted timeseries and one employed a randomized-controlled trial (RCT).
The certainty of evidence was generally assessed as quite low, despite the use of causal mechanisms and methods for overcoming biases, with only 4 articles identified as having a high or moderate certainty of evidence (Table 4).One of the most prevalent issues was selection bias, both in terms of the providers where participation was voluntary (i.e.those most likely to benefit and with necessary capacity participating), but also the expectation that patients with worse health would be more likely to enroll with these providers.
The most common type of quality indicator used were process measures related to chronic disease prevention and screening, (63%, n = 22), such as management of diabetes and cardiovascular diseases and cancer screenings.Fewer articles (n = 14) used health outcomes (e.g.self-reported health, mortality, survival rates, patient-perceptions of quality, readmissions, hospital admissions due to chronic disease).From these nine, only three included patient-perceptions of quality.

APMs, associated service delivery models and their impact on quality of chronic care
The scoping review identified five clusters of APMs.In the following we provide a short description for each of the APMs, the service delivery models in which they were embedded and their impact on quality of chronic care (Table 5).

Global budget/capitation
The largest share of articles discussed capitation-based/global budget models (n = 13), where providers received a global budget to cover all their patients or a per-member fee for those covered by the predetermined services [31][32][33][34][35][36][37][38][39][40][41][42][43].Variations within this type of arrangement exist in terms of the financial responsibility/risk taken on by providers, the services covered and the setting in which they were provided.In nine of these articles, financial incentives based on quality performance were used in addition to global budgets/capitation.Within this cluster, most articles found positive impacts on a range of quality measures.
Four articles examined capitation in primary care, including comprehensive care and chronic disease management programs provided by General Practitioners (GPs) [31,34,36], or team-based comprehensive care provided in Patient-Centred Medical Homes (PCMHs) [33].While in some of these primary-care based models GPs acted as gatekeepers to other services [34,36], others provided team-based multidisciplinary care [31,33].All of these arrangements uncovered mostly positive impacts on different measures of quality of chronic care.Additionally, a capitation model with financial incentives for monitoring chronic care patients in Germany improved survival rates [36], reduced hospitalizations due to ambulatory-care sensitive conditions (ACSH) and reduced chronic-care related hospitalizations for some conditions [34].However, there were no changes to re-admissions for patients with chronic conditions in a risk-adjusted partial capitation PCMH model [33].The use of health information technologies (HIT) and electronic health records (EHR) were also highlighted as initiatives used to improve information transfer and decision-making across each of these primary care models.
In two articles, global budgets were used to reimburse service delivery in hospitals, albeit with mixed findings.A hospital-based global budget model with P4P based on quality measures in China showed improved treatment indicators for acute myocardial infarction, but reduced oxygenation index assessments for chronic asthma [37].The Maryland All-Payer model, a global budget with P4P characteristics and shared savings/risk, mandatory monthly data reporting and eligibility for lump sum investment money to reorganize service delivery, had no impact on chronic-disease specific case-mix adjusted readmission rates or risk-standardized mortality [35].
A single article examined the impact of a managed care organisation (MCO) with full-risk, risk-adjusted capitation covering all services for Medicaid beneficiaries, finding that patients with severe and chronic conditions in the MCO received more ancillary services and post-acute care treatment (home health services) as compared to FFS arrangements [32].
Within this cluster, six articles [38][39][40][41][42][43] analyzed the Alternative Quality Contract (AQC), a risk-adjusted population-based global budget model with quality bonuses and shared savings/risk based on 64 quality measures, covering primary and specialty care under the same private payer ACO.A few of the six articles evaluated the model as positively affecting quality of care.Aggregate chronic disease management scores had improved after 4 and 8 years of implementation [38,39].Health outcomes of glycated hemoglobin and blood pressure control increased [39,42].One article found minor increases in psychotropic medication management visits for individuals with behavioural risks among AQC organisations that took on some financial risk for behavioural health [42].Other articles found no/limited spillover effects for Medicare beneficiaries [43], no improvements in process measures for managing asthma among children [40], and limited improvement in equality of chronic disease management between lower and higher socioeconomic groups [41].
Pay-for-coordination was often introduced as part of the implementation of PCMHs or similar primary-care based service delivery models.PCMHs usually encompass multi-professional care teams, aiming at the provision of comprehensive care and the coordination of care across healthcare settings.Furthermore, they focus on chronic disease management, population health, evidence-based guidelines, and programs for specific patient groups such as smoking cessation programs.In many cases, this form of service delivery was supported by the use of health data and HIT.To increase chances of successful implementation, some of the PCMHs received additional initial investment funding or technical support and engaged in learning collaboratives.In some cases, providers were requirements to obtain official PCMH recognition and to participate in quality improvement initiatives.In one case [50], patient engagement in the form of focus groups and surveys was used to monitor and inform quality initiatives.
The majority of articles in the pay-for-coordination cluster found some positive impact on the quality of chronic care (6 of 9).Most of the articles assessed changes to diabetes care (n = 9) and cancer screenings (n = 7).The most common measures used in connection with diabetes care were hemoglobin testing, eye exams, nephropathy screening, and LDL cholesterol screening.Three of the nine studies that assessed LDL cholesterol screening found a positive effect [44,49,50], with the remaining articles finding no effect [45,47,48,51,52].Two studies found a positive effect on hemoglobin testing [45,49] and eye exams [44,45].A positive effect on nephropathy screening was found in one article [51], while another found a negative impact [45].Three studies found no changes to diabetes care [47,48,52].Of the seven articles that assessed changes to different types of cancer screening, three found a positive impact on breast cancer screening [44,48,51] one for cervical cancer screenings [44], and none for colorectal cancer screenings.
Individual articles also assessed other chronic care measures.Fifield et al [48] reported positive outcomes for hypertensive blood pressure control with high certainty of evidence, but no changes for cardiovascular lipid testing [47].In the two articles that found no impact on quality of care, the authors speculate this may have been due to a lack of a facilitators to help support the implementation of the reform [47], limited time frame of the evaluation, and weak financial incentives to improve performance [52].

Shared savings (and shared risk)
The third cluster included articles discussing shared savings arrangements (n = 4) [53][54][55]58] or arrangements combining shared savings with shared risk (n = 2) [56,57], with both usually being based on a FFS model.In a shared savings arrangement, providers are eligible to receive a portion of the savings, typically determined by quality performance, if their total expenditure is less than the previously agreed-upon benchmark.With shared risk, providers are also held accountable for overspending if costs exceed the benchmark.
Regarding healthcare setting and service delivery models, shared savings/shared risk were implemented in primary care settings in connection with PCMHs [53,54], or in primary care more generally [56,58].Two studies assessed shared savings/shared risk in an ACO, which also included hospitals [55,57].PCMHs and ACOs put an emphasis on team-based care through collaboration and care coordination across professions, providers, and settings.Furthermore, in these models, a focus was also on health promotion and prevention, care management, and improving care transitions and follow-up care, supported by digital health tools such as EHR, web-based disease registries, automated alert systems, data-driven approaches to decision-making, and e-prescribing.Additional incentives consisted primarily of performance feedback, and in cases of PCMH models, the requirement to receive PCMH recognition by a certain deadline.
All six articles in this cluster reported some positive outcomes on the quality of chronic care, although with some conflicting findings in some areas of chronic disease management.Most found a positive effect on process measures of diabetes control, including hemoglobin testing and nephropathy screening [53][54][55]58], eye exams [54,55,57,58] and diabetic blood pressure control as a health outcome [56].Conversely, some reductions were seen for eye examinations for diabetes [53,59].
Nearly all articles also assessed changes in cancer screenings, albeit with mixed findings.Three articles found a positive impact on breast cancer screening [54,55,57], two on cervical cancer screening, and one on HPV vaccination uptake [57].Other studies found a negative impact on cervical [56] and colorectal cancer screenings [53,57].
Overall, those studies in which shared savings/shared risk arrangements were implemented alongside a patient-centered, integrated service delivery model, namely PCMHs [53,54] or ACOs [55,57], seemingly reported slightly better outcomes on the quality of chronic care than the studies where payment reforms took place in a general primary care setting [56,58].For example, a shared saving arrangement in a PCMH with additional bonus payments linked to prevention and chronic disease management measures found improvements in preventive care processes and quality of care, though suggested "choice overload" contributed to no changes or decreases in other measures [53].An early ACO model comprising a network of services, including PGPs, integrated delivery systems and hospitals, with shared savings, found positive impacts on four process measures of diabetes control, breast cancer screening and lipid profile tests for coronary artery disease [55].
In contrast, some authors [58] found that a shared savings model with P4P and bundled payments for patients with chronic diseases in primary care in the Netherlands led to increases in 8 process measures for diabetes control, yet a decline in four process measures of COPD control, suggesting prioritization of certain measures.Similarly, Navathe et al [56] concluded a capitation-based shared savings model for primary care had only limited improvements in quality overall in the first year with increases in diabetes blood pressure control, but reductions in cervical cancer screenings.

Blended capitation
Three articles examined the impact of physicians voluntarily shifting from a blended FFS model to blended capitation from 2003 onward in Ontario, Canada [59][60][61].The former model entails most earnings arising through risk-adjusted capitation based on a defined bundle of services provided to enrolled patients with limited earnings from FFS, while the latter comprises the reverse.Additionally, both arrangements included P4P elements, relating to disease prevention and chronic disease management, aftercare following hospital discharge and after-hours care.These APMs were implemented in primary care settings as a type of patient-centered medical home that provided comprehensive care, requiring changes relating to organization of service delivery, the introduction of formal patient enrollment and after-hours care provision.The blended capitation arrangement additionally included financial incentives for physicians to form team-based practices and provide multi-disciplinary care services.
Literature from this cluster found that switching from blended FFS to blended capitation in primary care settings in Ontario, Canada, had a positive impact on process quality of diabetes care [59] and on outcomes related to mental health, including ER visits due to mental health reasons [60], suggesting a more efficient use of resources [61].

Bundled payments
The fifth type of APM represented bundled payments but included only one article [62].In general, in bundled payment arrangements providers are reimbursed based on expected costs for a clinically defined episode of care for certain diseases, covering all the services provided during this episode.The bundled payment of Care Chain Frail Elderly (CCFE) was part of a Dutch program that aimed to support frail elderly patients with complex care needs to live in their homes as long as possible.The program included the integration of primary and community care, multi-professional healthcare teams, individual case management, and person-centered care plans, all supported by HIT and the sharing of information among different professionals.The authors assessed changes to the quality of chronic care through patient-reported outcome measures and found a positive impact on person-centeredness, but also a decrease in autonomy and an increase in the burden of medication [62].

Comparison across arrangements
The final cluster comprised two articles comparing variations of APMs across different locations or types.One article compared an Oregon-based Medicaid ACO (risk-adjusted global budget with full financial risk) against a Colorado Medicaid ACO (PMPM and bonus payments on top of FFS) [63].Both models consisted of primary care medical homes that coordinated care, had automatic enrollment of patients, and additional financial incentives.However, the Colorado model focused more on enhanced payment for coordination and case management, while the Oregon model was more comprehensive in the services covered and included funding for administrative staff, data infrastructure, implementation resources and training for its implementation.The results indicate that while the Oregon model was superior in a number of measures, neither model proved superior to the other in terms of chronic care related measures of quality, including appropriate medication for individuals with asthma.
The second article compared care for children enrolled in various Medicaid Managed Care models (i.e.capitated arrangements between the state and managed care plan).Hall et al [63].found that Pediatric Only Plans (POPs) within Provider Service Networks (PSNs) were generally rated better across a number of subjective process measures by parents than in HMOs.However, no evidence was found of improved ratings of care in POPs by parents of children with chronic conditions.

Disentangling the impac tof payment models from service delivery
Reforms to payment models often go hand-in-hand with changes to service delivery models.Disentangling the two to determine what aspects of financial or service delivery reforms impact quality of chronic care specifically when both are present, is therefore difficult to do.In a limited number of articles, authors were able, or attempted in some way, to disentangle the two to draw conclusions on their contributions to improvements in quality of chronic care.None of the articles however, quantified and tested aspects of service delivery or reimbursement as variables in their analyses.
In exploring the impact of a blended capitation model in Ontario, Canada, versus a blended FFS model, Kiran et al [59] additionally distinguished between blended capitation with a multidisciplinary team-based care model (incentivized by additional funding) versus blended capitation or blended FFS in settings without team-based care.The authors found that blended capitation alongside forming team-based practices had the most positive impact on process quality of diabetes care, relative to blended capitation alone and blended FFS.This suggests a superiority of financing and service delivery reforms together in bringing about positive change to process quality for some chronic diseases, versus changes to financing alone.
In another example of trying to disentangle the impact of service delivery from purchasing reforms, Hall et al [64] explored differences in parents' perceptions of quality of care for their children with chronic conditions, in a HMOs versus a Pediatric-Only Plan (POP).As a sort of Provider-Sponsored Network, the authors hypothesized that these POPs would be rated better by parents with children with chronic conditions, as the result of differences in organizational structure (i.e.size), not-for-profit ownership, mission to serve only Medicaid patients, and their general focus on pediatric care.The authors found that while POPs were generally rated better by parents over a number of access and quality ratings, parents of children with chronic conditions did not provide higher ratings among children-focused POPs versus HMOs, despite their specialized plans geared towards children with chronic diseases.
McConnell et al [63].compared two different ACO models based on different payment models across two U.S. states, Colorado and Oregon, to better understand what approaches are effective for improving the quality of care.While both ACOs were comparable in terms of enrollment of beneficiaries and service delivery model (i.e.Primary care medical home), minor additional initiatives were implemented by each.The Oregon model implemented a hospital-to-home transition program, while the Colorado model implemented a centralized data repository to track and report clinical performance.The authors found relative performance improvements of the Oregon model compared to the Colorado model, although not for chronic care.
In three articles, payment reforms were seemingly taken in a topdown approach with no explicit changes to the service delivery model, but where the intention was to incentivize different behaviour among providers through the new reimbursement method, with mandatory participation [32,35,37].In these cases, it can be inferred that changes to service delivery resulted from changes in the payment model, thus suggesting the directionality of impact and allowing us to attribute impact to the payment model.For example, Munnich and colleagues [32] detailed a legislative intervention in Florida that mandated a shift from FFS to full-risk capitation (per-enrollee-per-month payment) via managed care contracting for Medicaid beneficiaries, with the expectation that managed care plans would coordinate all care and manage chronic diseases of patients.For some chronic diseases, this mandatory reform to capitation via managed care contracting resulted in an increase in ancillary services and post-acute care treatment, likely as means for better management of chronic patients and improvement in post-acute care transitions to reduce future costs.Similarly, a local government in China chose 16 hospitals to switch from FFS to a global budget as means for reducing expenditure, with positive impacts on some chronic disease-specific process measures of quality [37].The mandatory shift to global budgets for all hospitals in the Maryland All Payer model however, found no changes to quality measures for three cardiovascular conditions [35].
In the remainder of the articles, authors were unable to or did not attempt to disentangle the impacts of financing from those of service delivery.In some of the clusters, such as the pay-for-coordination cluster and most of the shared savings/shared risk group, payment and service delivery reforms were inherently intertwined as they were implemented concurrently in bottom-up approaches, rendering it difficult to disentangle the impact of one from the other.As the PCMH model emphasizes comprehensive, patient-centered, coordinated care led by physicians, and financial reimbursement that supports this, then reforms to reimbursement go hand-in-hand with service delivery changes.Similarly, of the articles on shared savings/risk, payment reforms often accompanied explicit and pre-determined changes to service delivery, primarily through agreements between payers and providers, such as in ACO models.

Discussion
We found that most clusters of APMs indicated some positive impact on quality of chronic care, however, quality indicators varied substantially between articles, limiting direct comparisons.With an overall view across different types of APMs and their associated service delivery models, it appears that population-based purchasing reforms in combination with more collaborative forms of service delivery were more frequently successful in improving the quality of chronic care than in cases where purchasing changes were limited to a single setting.This was seen in the shared savings and risk cluster, where collaborative and team-based forms of service delivery were seemingly more likely to lead to positive impacts of process measures of preventive care and chronic disease management, versus those only including primary care providers.This was also evident with global budget and capitation models which appeared to be more effective in comprehensive or gatekeeping primary-care based models and in ACOs covering primary and specialty care, and less so in hospital-based service delivery models.The more mixed findings in hospital settings may in part be due to the limited scope of services covered by the global budget and limited coordination with outside services, such as primary care, which individuals with chronic conditions are most likely to benefit from in the longer term [65].This speaks to designing payment models in a way that covers a range of providers in collaborative forms of service provision along the continuum of care, to ensure sufficient accountability, responsibility and buy-in.Purchasing reforms isolated to single settings or providers may not provide sufficient incentives to organize and provide care in a way that patients with chronic conditions need.
Payment reforms have the aim to either indirectly incentivize changes in provider behavior and service delivery, or in agreement with providers, to more explicitly reimburse them based on a certain model of service provision.Few articles aimed to, or could, disentangle the impact of financing reforms from changes to service delivery.Where authors were able to do so, this was often by comparing arrangements that differed in only one particular aspect, often through difference-indifference models.The directionality of impact was also possible to discern in cases where payment reforms were mandatory, but did not specify or require explicit changes to service delivery.Given the limited number of articles achieving this and the mixed findings, only one concrete implication arose from this: the combination of support to transform service delivery to a team-based model of care alongside a purchasing reform shifting to blended capitation was more impactful in improving quality of chronic care, than the individual components [59].Although the evidence-based is thin, these results suggest that financing reforms are just one element of initiatives for improving quality of care, and that service delivery is a crucial piece of the puzzle.In fact, purchasing and service delivery reforms were implemented concurrently in most studies.These reforms often occurred within the scope of implementing an ACO or PCMH model, which require collaboration between payers, who set the terms of reference for reimbursement, and providers, who must implement the service delivery changes.It's therefore likely that in these cases, the widespread positive impacts seen are due to sufficient buy in on both sides of the purchaser-provider relationship.
Unsurprisingly, despite efforts to keep a global reach, the review revealed a strong concentration of literature stemming from high income countries, particularly the U.S. Particularities of the U.S. health care system, namely the highly complex and fragmented nature and the multi-payer structure concentrated on private payers, likely limit the transferability of some of the models to other health care systems.The incentives behind different payment models may have different effects on provider behavior in contexts where providers do not have the option to choose their patients based on their type of coverage, as in other healthcare systems [66].In systems where providers can choose their patients based on their coverage, this could lead to inequities in patients taken on and treated, thus biasing upward the impact of interventions on quality of chronic care.

Recommendations for future research
Based on this scoping review, several key considerations for research can be inferred.First, the certainty of evidence was generally quite low, despite the use of causal mechanisms and methods for overcoming biases associated with voluntary participation.In the future, researchers must continue to consider these biases in their methodology and interpret their results accordingly.
The evidence base is thin on studies explicitly aiming to disentangle components of the purchasing and service delivery reform, likely out of methodological challenges.Future research should aim at disentangling purchasing and service delivery characteristics through the use of comparative methods, such as difference-in-difference models, and by quantifying service delivery aspects, to better understand the relationship between context, mechanisms and outcomes.
Few articles assessed the impact of APMs on the quality of chronic care through patient-reported outcomes [31,62,64].While studies based on clinical indicators can provide useful information about changes to processes and physical outcomes, this provides only a partial picture of quality of chronic care.While not without their criticism, patient-reported measures are important, particularly if care is to be person-centered and aligned with individuals' goals and preferences.Future research could draw on existing patient-centered methodologies [67,68] and thereby contribute to a more comprehensive understanding of the effects of APMs on the quality of care.
Finally, future research would benefit from better understanding which chronic care measures providers prioritize in cases of choice overload.A large number of studies uncovered positive impacts on cancer screening measures and diabetes control measures related to diabetes, suggesting these may be easy measures for providers to prioritize.

Limitations
While this scoping review used a robust methodology to collate evidence, the review was not exhaustive.Only English texts were included, potentially limiting the representation of countries.Heterogeneity of terminology used in the literature may also have limited the studies identified.While we narrowed the scope of this review to chronic conditions and excluded multimorbidity, many studies did not distinguish between individuals in their sample that had one chronic condition versus more.As past literature has suggested that most payment and service delivery models are not well-suited for addressing the needs of individuals with multimorbidity [69], the limited impact on quality in some articles may be due to this oversight.
Although our inclusion criteria required a certain level of information on the payment and service delivery model, we generally found that beyond these criteria, articles did not provide sufficient information on the payment and service delivery models.Details concerning how payment is dispersed to providers in group-based arrangements, the precise measures used to evaluate performance, and the magnitude of financial incentives relative to providers' pay, were frequently missing.Across most articles, vague language centring around coordination and integration of services was often used, without further explanations of how this was operationalized in practice.

Conclusion
This scoping review set out to compile evidence on purchasing reforms and associated service delivery models at the disposal of policymakers for incentivizing improvements in quality and to better disentangle the impact of payment reforms from service delivery reforms on quality of chronic care.The broader evidence suggests that population-based payment models achieve the best outcomes in combination with collaborative, patient-centered forms of service delivery spanning different settings and providers.While this review is useful for policymakers in indicating how purchasing fits into broader service delivery models and how payment models compare across settings, a wider evidence base of articles isolating the impact of particular elements on quality of care is still needed to better understand the mechanisms impacting quality of care, to better inform policymaking.To achieve this, researchers will need to employ comparative causal inference methods through exploiting timing of mandatory payment reforms.Researchers are further encouraged to quantify aspects of service delivery in their models and to compare similar arrangements that vary only in a particular aspect to better isolate particular mechanisms.

Fig. 1 .
Fig. 1.PRISMA 2020 flow diagram for new systematic reviews which included searches of databases and other sources [25].

C
.Simmons et al.

Table 3
Details extracted from articles.

Table 4
Summary of literature included.

Table 4 (continued ) # Year First Author Country APM studied Program (if applicable) Service delivery model (if any) Purchaser Target group Intervention (Year) Outcomes assessed Methods Study Design Certainty of evidence
(continued on next page) C.Simmons et al.

Table 4
(continued ) (continued on next page) C.Simmons et al.

Table 4
(continued ) Low (continued on next page)C.Simmons et al.

Table 5
Summary of quality incentives, service delivery and impact on quality indicators across articles.
(continued on next page) C.Simmons et al.

Table 5
(continued on next page) C.Simmons et al.

Table 5
+(continued on next page) C.Simmons et al.

Table 5
+(continued on next page) C.Simmons et al.