4 Towards Integrated Health Systems: Scenarios and Strategies

4.1 Introduction: System-Level Consequences

Improving healthcare delivery requires more than refining payment models in isolation. As outlined in the previous chapters, many of the systemic problems we face (fragmentation, misaligned incentives and poor coordination), stem not from payment alone, but from deeper organizational and institutional structures. A reform agenda must therefore address three interconnected layers of the healthcare system:

Culture: stimulate mutual trust, shared responsibility, professional identity, and leadership. Coordination begins with a culture that values collaboration across boundaries. Leadership at all levels is needed to bridge professional silos, set shared goals, and support change. A data-driven approach to learning is crucial to strengthen professional autonomy and continuous improvement.
Organization: enable trust-based cooperation through governance, shared rules, and institutional support. Cross-sector collaboration requires more than goodwill. It demands clear rules for joint decision-making (e.g., on shared investments or workforce planning), mechanisms for resolving conflicts, and governance arrangements that lower transaction costs and align roles and responsibilities across institutions.
Financing: align payment models with patient needs and long-term system goals. Payment is not neutral, it influences what care is delivered, by whom, and how. Financial mechanisms must reward coordination, continuity, and prevention. They must also be adaptable to local contexts, account for complexity, and distribute resources in ways that enable integrated, person-centered care rather than reinforcing fragmentation.

Yet greater coordination might bring unintended consequences. On the provider side, regional collaborations (e.g., Beter Samen in Noord) may become so essential to care delivery that they acquire market power, so they can act independently of insurers. On the payer side, closer coordination between insurers -while necessary to overcome the public goods problem in payment innovation- reduces the remaining mechanism of competition. Since Dutch insurers cannot engage in premium differentiation or risk selection, their only remaining margin for competition lies in better contracting than their competitors. Follow-up contracting (as described in Chapter 3) for successful innovations, while socially desirable, undermines this mechanism.

These dynamics suggest that more vertical integration beteen providers and between payers and providers may be inevitable. Indeed, this is the direction proposed by Ven and Schut (2024), who advocate for competing regional HMO-style systems. My view is that while more vertical collaboration and integration may be the right path, true competition between regional HMOs is unlikely to emerge. Even in dense areas like Amsterdam-Noord, the integrated network of Beter Samen in Noord leads to a (near) regional monopoly.

Thus, any realistic scenario for the future must assess these effects and weigh trade-offs. This chapter does not offer predictions, but explores three scenarios to illustrate possible trajectories. In each, key questions emerge:

How large should these regional systems be?
Who should define their governance?
What happens to specialized care organized nationally?

These key questions are not yet answered.

4.2 Different Scenario’s

4.2.1 Scenario 1: Controlled Integration

In this scenario, regional provider collaborations grow organically, supported by light regulation. Providers integrate services across care domains (hospital, primary, social, mental health), and insurers form voluntary coalitions to coordinate follow-up contracts. Governance structures remain mostly private, with minimal state involvement.

Pros: Maintains flexibility, encourages local innovation, minimizes bureaucracy.
Cons: Without strong oversight, coordination may fail.

Providers may gain market power without accountability. Payers may lose incentives to contract for value.

The U.S. DSRIP programs suggest the feasibility of voluntary regional coordination (Janus 2019). However, these initiatives often lack structural follow-up, and their population coverage may remain partial. The central risk is that while the system formally allows for coordination, incentives and market structures discourage sustainable implementation.

4.2.2 Scenario 2: Hard Integration

In this scenario, the system embraces full regionalization. Regional budgets are assigned to integrated care organizations, which assume responsibility for a defined population. The government sets the rules for governance, price and budget regulation, data-sharing, and quality oversight.

Pros: Strong alignment of incentives, public accountability, potential for cost control and improved outcomes.
Cons: Bureaucratic rigidity, potential for innovation stifling, difficulty in managing cross-regional or specialized care.

Integrated systems also require high levels of trust, data capacity, and adaptive governance. Without these, hard integration risks becoming a rigid administrative structure that fails to deliver patient-centered care.

4.2.3 Failure to adapt to better coordination

In this scenario, the system remains fragmented. Provider collaborations remain isolated pilots. Insurers continue to contract independently. There is no structural investment in coordination, and innovation depends on temporary grants.

The potential consequences are:

Growing inequality in access and outcomes.
Workforce shortages may lead to waiting lists and care delivery challenges.
Public distrust of the healthcare system.

This is the “default path” if action is not taken. It reflects the trajectory of health systems that acknowledge systemic problems but fail to implement coherent solutions. Ultimately, this scenario leads to crisis-driven reform rather than proactive change.

4.3 Final Reflection: Experimentation and Learning

The scenarios outlined above are not predictions, but illustrations of different pathways, each with different advantages, risks, and trade-offs. What they make clear, however, is that there are no blueprints for system transformation. Structural change in healthcare is complex, uncertain, and politically sensitive. But uncertainty should not be an excuse for inaction.

In such a context, the best strategy is not to wait for perfect designs, but to start small, experiment, monitor, and adapt. Regional pilots, such as Beter Samen in Noord, demonstrate how trust-based collaboration can emerge and deliver results. But their success depends on more than goodwill—it requires institutional support for learning.

This calls for a different way of thinking: one that acknowledges complexity, embraces uncertainty, prioritizes institutional learning over static design and a good analysis of the process of change.

Experimentation in healthcare should not be reduced to isolated, fixed pilots. Real-world interventions often evolve during implementation as stakeholders learn and adjust. This makes experimentation a dynamic process. It requires structured, but flexible efforts to explore what works in practice. Recognizing that outcomes may differ by context, population, and professional culture.

Adaptive governance is essential to make this work. When no blueprint exists and when outcomes are uncertain, governance must be flexible and resilient. Here, contract theory offers a valuable framework. In particular, the idea of incomplete contracts helps us understand why fixed rules and rigid accountability structures often fail in complex environments. Not all contingencies can be foreseen and not all outcomes can be pre-specified. This means governance should not rely solely on detailed ex-ante contracting but must also stimulate longer-term relationships, mutual trust, and the ability to renegotiate and adapt. In this sense, adaptive governance is the institutional form of contract theory under uncertainty: it is about designing frameworks that evolve as information emerges, as consequences and incentives are better understood through experience.
Bayesian causal learning offers a good framework for this type of real-time learning and decision making. Because experiments are often adapted along the way, involve small sample sizes, and operate in complex environments, Bayesian multilevel models are a good way to estimate effects under uncertainty. Unlike classical methods that rely on large-sample inference and fixed designs, Bayesian approaches allow us to incorporate prior knowledge, continuously update our beliefs, and quantify the probability of specific effects. This is especially helpful under continous monitoring of the effects. However, in real-world policy settings randomized controlled trials are often infeasible due to ethical, logistical, or political constraints. This makes robust causal modeling essential. Well-designed observational studies using (staggered) difference-in-differences, matching, instrumental variables or regression discontinuity designs can provide causal identification strategies. Bayesian methods can strengthen these approaches by modeling uncertainty explicitly, integrating multiple sources of data, and yielding interpretable probability statements. For example, we can estimate the probability that a particular intervention reduces the incidence of chronic disease by 10% within five years. This information is far more relevant to policymakers than conventional p-values. Bayesian learning also accommodates non-linear and dynamic processes, where average effects may hide important heterogeneity in outcomes, risks, and benefits.

For me, this way of thinking, combining systems thinking, contract theory and Bayesian reasoning, is fundamental to what Health Systems Engineering should be. Health Systems Engineering is not just about technical optimization, but about designing systems that are capable of learning, adapting, and improving under real-world conditions. It is a practical and principled response to complexity.