Please note: The algorithm descriptions in English have been automatically translated. Errors may have been introduced in this process. For the original descriptions, go to the Dutch version of the Algorithm Register.
Care demand typing
- Publication category
- Impactful algorithms
- Impact assessment
- Privacy Quickscan, DPIA
- Status
- In use
General information
Theme
Begin date
Contact information
Link to publication website
Responsible use
Goal and impact
Goal: Better allocation of resources between relatively simple and complex patient groups.
Impact: Providers and insurers agree on groups of patients divided by type of care demand. This leads to more resources for more complex groups. It also reduces the incentive for healthcare providers to give preference to patients with relatively simple care demands. The algorithm reflects the expected group into which a patient is likely to fall based on the completed care demand. This ensures shared trust in the tool between insurer and provider.
Considerations
Insurers and providers have no way to discuss the complexity of target groups in procurement without care demand typing. This led to procurement with tight budget caps and cost per unique client (KPUC). This caused cherry picking of relatively simple patients. The care demand type label allows differentiation in procurement, for example on average cost per patient in a group, or maximum waiting time for a group. This tool is trusted only if it is based on objective measures. The algorithm ensures that the practitioner knows what this objective measure is and ensures that insurers can investigate patterns of deviation in case of distrust.
Human intervention
The practitioner is not bound by the label/care demand type predicted by the algorithm. The practitioner can choose the care demand type that he/she thinks best fits his/her patient's care demand.
Risk management
The risks of using the care demand typing algorithm were looked at in detail. For instance, together with representation from the mental health sector, the outcome of the algorithm was looked at, among other things to gain insight into any bias. This bias turned out not to be present. In addition, there is always a human being who makes the final choice for a care demand type, informed only by the results of the algorithm. Finally, it was agreed with all parties from the mental health sector that the categorisation, informed by the algorithm, will not have consequences at the individual level, but only at the level of groups.
Legal basis
With the new care demand typing, care progression can be better predicted and tracked. It therefore provides more insight into what kind of care and how much care is needed for patients.
The algorithm thus has an important place in the contracting and purchasing discussions and thus falls within the statutory task of market development and performance and rate regulation that the NZa has pursuant to Article 16, opening words and under a, Wmg.
Links to legal bases
Elaboration on impact assessments
On the processes in which this algorithm plays a role, a legitimacy assessment and a GEB were performed according to our standard procedure. Prior to deployment, advice from the Personal Data Authority was requested. Many sector organisations also took part, with them so-called field agreements were made on the healthcare performance model.
Impact assessment
- Privacy Quickscan
- Data Protection Impact Assessment (DPIA)
Operations
Data
Initial input is the HoNOS+ score list that the practitioner scores of the patient's complaints, problems and symptoms, along with a choice of a diagnosis main group. This input is processed by the algorithm, from which follows a prediction of one or a few care demand types. Finally, the practitioner chooses the care demand type that best fits his patient's care demand.
Technical design
The algorithm has a few layers. First, the practitioner chooses whether she wants to do the full HoNOS+ or the 'dynamic' care demand typing (more on the latter later). Next, the practitioner chooses into which main group she thinks the patient's complaints fall and scores the HoNOS+, which are 19 items scored with (integer) values from 0 (no complaints) to 4 (very severe). During completion, the practitioner has access to the text of the questions/items and a more in-depth description of the question, as well as per score a description of the clinical picture associated with the selected score. Depending on the main group, the algorithm is loaded with main group-specific coefficients and decision rules. In the case of full care demand typing (basically Linear Discriminant Analysis), the patient's care demand is linked to the coefficients and constant belonging to the completed scores on the HoNOS+ and the care demand types in the main group. Next, it is checked whether certain 'red lines' have been activated, these are combinations of scores and care demand types that are not expected (expert input), these care demand types are excluded. These coefficients and constants can be summed to produce a probability distribution across the care demand types in the main group. This distribution is shown to the practitioner, she chooses the care demand type that best fits within the main group. In the case of dynamic care demand typing, the practitioner starts a decision tree (structure depending on main group), which is constructed as a means of arriving at a reliable care demand type with as few items as possible (reality modelled on LDA under full care demand typing). The practitioner scores a first item and depending on the score on that first item, she is given a follow-up question, just until a most likely care demand type can be shown to the practitioner with at least 95% certainty. Again, the practitioner is free to choose a care question type within the main group, independent of the prediction.
External provider
Similar algorithm descriptions
- The algorithm helps the CIZ employee choose the care profile that best suits the client's care needs. A care profile states what care a person needs.Last change on 31st of October 2024, at 16:40 (CET) | Publication Standard 1.0
- Publication category
- Impactful algorithms
- Impact assessment
- Field not filled in.
- Status
- In use
- The risk model (the algorithm) helps choose which NOW applications to investigate further. It gives clues as to whether the information given in the NOW application is correct. With these indications from the risk model, SZW reviews the application.Last change on 10th of June 2024, at 13:42 (CET) | Publication Standard 1.0
- Publication category
- Impactful algorithms
- Impact assessment
- Field not filled in.
- Status
- In use
- The Operational Readiness Dashboard provides insight into the current operational readiness and deployability of the armed forces. Specifically, the dashboard provides insight into personnel readiness, materiel readiness and exercise readiness per unit. This dashboard runs in BI/VEST, which stands for Business Intelligence/Improved Steering Information.Last change on 28th of November 2024, at 12:30 (CET) | Publication Standard 1.0
- Publication category
- Other algorithms
- Impact assessment
- DPIA
- Status
- In use
- The Care Module is an LIJ tool and part of the Ritax. With it, a professional identifies concerns about the home situation. The instrument identifies the number of care factors. Afterwards, the Child Protection Board and/or the Youth Probation Service can weigh up whether professional help is needed, in addition to any punishment or the criminal justice process.Last change on 3rd of December 2025, at 13:14 (CET) | Publication Standard 1.0
- Publication category
- Impactful algorithms
- Impact assessment
- DPIA
- Status
- In use
- Rule-based algorithms are used to perform some of the administrative work in the informal care valuation process. The technique used is Robotic Process Automation (henceforth RPA)Last change on 17th of September 2025, at 13:02 (CET) | Publication Standard 1.0
- Publication category
- Other algorithms
- Impact assessment
- Field not filled in.
- Status
- In use