Analysis of the Current System

To address systemic issues in healthcare, we must first demonstrate the specific areas where the current model is failing by using verifiable data. While the Netherlands is frequently ranked as having one of the best healthcare systems in Europe, it is currently experiencing a “care gridlock” known as a zorginfarct. This is not a future projection but a present reality where 38% of specialist referrals exceed legal waiting standards and many general practitioner (GP) practices are closed to new patients. Research from 2023 indicates a projected shortfall of 301,000 healthcare workers by 2035; this number represents a structural impossibility because the system cannot function with the workforce that will be available.

The pressure on the system is driven largely by an aging population. Currently, 20% of Dutch citizens are over the age of 65. Within the group aged 75 and older, 96% have at least one chronic condition and nearly half have two or more. The healthcare system was originally designed to provide acute care for a younger population, and the architecture of the system has not evolved to meet these new demographic needs.

There are three primary factors driving the current failure of the system, and each factor serves to worsen the others.

1. The Administrative Burden

A Dutch GP spends 30% of their working time on administrative tasks such as documentation, insurance coding, and quality reporting. This is an architectural fact of the system where every new regulation or reporting requirement adds minutes to a patient’s visit without removing any existing tasks. For example, a GP who could theoretically see 200 patients per week is limited to approximately 124 consultations because the remaining 76 slots are consumed by the system’s own reporting requirements. When combined with an 8% sick leave rate, a GP operates at only 62% of their theoretical capacity.

2. Breaches of the Treeknorm

The Treeknorm represents the legal maximum time a patient should wait for specialist care, which is generally four weeks for diagnostics and seven to twelve weeks for treatment. In practice, these norms are frequently ignored.

When patients wait too long, their conditions deteriorate; this often turns a simple primary care issue into a more complex and expensive emergency room visit. The system generates the complexity it cannot handle.

3. Data Fragmentation

The Dutch healthcare network consists of roughly 12,000 GP practices and 90 hospital organisations, yet only 11% of health data is shared effectively between them. Most providers still use EDIFACT, which is a messaging standard from the 1980s originally created for shipping logistics. This lack of communication means that five different providers treating the same patient may maintain five independent and disconnected records.

The Capacity Formula

Cammelot calculates the effectiveness of a GP practice using a specific equation to show how much time is actually available for patients:

In this formula, L_admin represents the 30% administrative loss and V_ziekte represents the 8% loss due to sick leave. In the current system, there is no AI assistance, so E_ai is set to 1.0. This means a practice meant for 2,400 patients can only effectively manage 1,488 people before the quality of care begins to decline and burnout occurs.

Health Deterioration and Mortality

Mortality in this simulation is a direct result of delays in the system. When a citizen waits longer than the Treeknorm allows, their health points (HP) begin to drain according to this formula:

The “drain rate” increases based on how serious the illness is. For a patient in a critical state, health points drop four times faster than for someone with a mild condition. The simulation also accounts for the fact that administrative overhead slows down every step of the care chain, which further accelerates how quickly a patient’s health declines. The model distinguishes between “natural mortality,” such as a person dying of old age at 92, and “system-failure mortality,” where a person dies specifically because they waited 18 weeks for a cardiologist.

The Disease Engine and Comorbidities

The simulation tracks 10 chronic conditions using official ICD-10 codes. Each condition progresses through a five-state Markov chain where the system checks every simulation tick to see if a patient moves from “healthy” to “mild” or from “severe” to “critical.”

Waiting for treatment does not just delay help; it actively moves patients toward worse health states. We also model how different diseases interact. For example, clinical research shows that having COPD increases the risk of cardiovascular disease by 2.5 times. These “relative risk multipliers” ensure that the simulation reflects the actual danger that multimorbid patients face when the system is backed up.

Observations from 200 Simulation Runs

When we average the results of 200 independent runs for the current system, the following patterns emerge:

Researcher Disclosure and Limitations

It is important to note that I have worked in Big Tech for the last six years. I built Cammelot as a personal research project to test these systems independently. The code is open source and the opinions expressed here are my own.