Medicine as a data science (2)

The Evolution of Patient Diagnosis: From Art to Digital Data-Driven Science

Currently medical diagnosis is driven by a standard way to proceed. We could say that the pattern of the decision flow has not changed for years.

A physician takes a history, performs an examination, and matches each patient to the traditional taxonomy of medical conditions. Symptoms, signs, family history, and laboratory reports are interpreted in light of clinical experience and scholarly interpretation of the medical literature.

Data availability, and specifically genetic data could change completely diagnostic process.

Initiatives to develop genetic reference data at the population level could be grouped into 3 categories.First are well-known databases of genotype-phenotype relationships
as observed and submitted by researchers (eg, Online Mendelian Inheritance in Man, ClinVar, and the National Human Genome Research Institute’s Genome-Wide Association Study [GWAS] Catalog). Second are databases, such as the Genome Aggregation Database (gnomAD), the next iteration of the ExomeAggregation Consortium (ExAC) database, and the 1000 Genomes Project, that aggregate sequences
collected from other studies for secondary use. Third, patients and other study participants are invited to donate data to registries like GenomeConnect or enroll
in cohorts like the National Institutes of Health All of Us initiative, which is recruiting 1 million patients to contribute biological samples and EHR data for research.

The reference to these databases is crucial to understand what's going on in US medicine, and how european medicine stands behind.
JAMA article develops the concept of Clinical Information Commons:

There should be a new compact between patients and the health system, such that captured data and biospecimen by- products of the care deliverysystem should be aggregated and linked to build a clinical information commons (CIC) to aid diagnosis

I agree. Saluscoop started as an alternative focused in this approach. As usual, the big question is: who is going to invest in a digital commons?. Unless governments take this initiative as a whole, the future of a data driven medicine is uncertain.

Adrian Piper: A Synthesis of Intuitions, 1965–2016

MoMA, New York, New York

Sat 31 Mar 2018 to Sun 22 Jul 2018

Laboratory medicine as a data science

Data science, artificial intelligence, and machine learning: Opportunities for laboratory medicine and the value of positive regulation

Artificial intelligence (AI) and data science are rapidly developing in healthcare, as is their translation into laboratory medicine.

These are the four areas that the authors consider that AI will have impact:

• Processes and care pathways
• Laboratory test ordering and interpretation
• Data mining, early diagnosis, and proactive disease monitoring
• Personalized treatment and clinical trials

Meanwhile there is a long way ahead.

Jacob Lawrence, This is Harlem, 1943. Gouache and pencil on paper. Hirshhorn Museum and Sculpture Garden, Smithsonian Institution, Gift of Joseph H. Hirshhorn, 1966. Artwork © The Jacob and Gwendolyn Knight Lawrence Foundation, Seattle / Artists Rights Society (ARS), New York; photograph by Cathy Carver

Medicine as a data science

THE PATIENT WILL SEE YOU NOW
The Future of Medicine Is in Your Hands

Maybe the title is the most confounding factor of the new great book written by Eric Topol.  Once you have finished reading it, you'll be convinced that he set the expectations to high, ordinary people should develop certain skills beyond their capabilities to apply such concept. I would say that a greater part of the medicine is in your hands, not medicine at all. The rationale behind the book is that medicine digitization allows patients to know more about their disease and how to "manage" it in certain cases. The most important thesis is that future medicine has to be considered a data science. And this is exactly the impact of the digitization of diagnostic and treatment: pervasive application of Bayes theorem in clinical practice, using big data and analytics.(Remember my archimedes posts, surprisingly Topol forgot it).
The book includes many topics that those that follow this blog it would sound familiar, i.e. ch. 4 about Angelina Jolie and BRCA genetic tests, a must read. And chapter 5 is a journey on the new omics of the medicine, a topic that I have also covered in the blog.
Nowadays, Eric Topol is the writer that is able to capture what's going on in medicine and its impact on society. That's why this book is a key reference of our time and I strongly recommend it.

PS. If you don't believe me, check Forbes, NYT, WP, WSJ.
PS. The book is also an invitation to change the current academic programmes for life sciences universities. Better now than later.

Medicine as a data science (3)

High-performance medicine: the convergence of human and artificial intelligence

If you want to know the current state of artificial intelligence in medicine, then Eric Topol review in Nature is the article you have to read. A highlighted statement:

There are differences between the prediction metric for a cohort and an individual prediction metric. If a model’s AUC is 0.95, which most would qualify as very accurate,
this reflects how good the model is for predicting an outcome, such as death, for the overall cohort. But most models are essentially classifiers and are not capable of precise prediction at the individual level, so there is still an important dimension of uncertainty.

And this is good summary:

Despite all the promises of AI technology, there are formidable obstacles and pitfalls. The state of AI hype has far exceeded the state of AI science, especially when it pertains to validation and readiness for implementation in patient care. A recent example is IBM Watson Health’s cancer AI algorithm (known as Watson for Oncology). Used by hundreds of hospitals around the world for recommending treatments for patients with cancer, the algorithm was based on a small number of synthetic, nonreal cases with very limited input (real data) of oncologists. Many of the actual output recommendations for treatment were shown to be erroneous, such as suggesting the use of bevacizumab in a patient with severe bleeding, which represents an explicit contraindication and ‘black box’ warning for the drug. This example also highlights the potential for major harm to patients, and thus for medical malpractice, by a flawed algorithm. Instead of a single doctor’s mistake hurting a patient, the potential for a machine algorithm inducing iatrogenic risk is vast. This is all the more reason that systematic debugging, audit, extensive simulation, and validation, along with prospective scrutiny, are required when an AI algorithm is unleashed in clinical practice. It also underscores the need to require more evidence and robust validation to exceed the recent downgrading of FDA regulatory requirements for medical algorithm approval

Therefore, take care when you look at tables like this one:

Prediction	n	AUC	Publication (Reference number)
In-hospital mortality, unplanned readmission, prolonged LOS, final discharge diagnosis	216,221	0.93* 0.75+0.85#	Rajkomar et al.96
All-cause 3–12 month mortality	221,284	0.93^	Avati et al.91
Readmission	1,068	0.78	Shameer et al.106
Sepsis	230,936	0.67	Horng et al.102
Septic shock	16,234	0.83	Henry et al.103
Severe sepsis	203,000	0.85@	Culliton et al.104
Clostridium difficile infection	256,732	0.82++	Oh et al.93
Developing diseases	704,587	range	Miotto et al.97
Diagnosis	18,590	0.96	Yang et al.90
Dementia	76,367	0.91	Cleret de Langavant et al.92
Alzheimer’s Disease ( + amyloid imaging)	273	0.91	Mathotaarachchi et al.98
Mortality after cancer chemotherapy	26,946	0.94	Elfiky et al.95
Disease onset for 133 conditions	298,000	range	Razavian et al.105
Suicide	5,543	0.84	Walsh et al.86
Delirium	18,223	0.68	Wong et al.100

1. LOS, length of stay; n, number of patients (training+ validation datasets). For AUC values:
2. *, in-hospital mortality; + , unplanned readmission; #, prolonged LOS; ^, all patients; @, structured + unstructured data; + +, for University of Michigan site.

Beyond precision medicine: high definition medicine

High-Definition Medicine

Some months ago I was posting on medicine as a data science. Now:

The foundation for a new era of data-driven medicine has been set by recent  technological advances that enable the assessment and management of human health at an unprecedented level of resolution—what we refer to as high-definition medicine. Our ability to assess human health in high definition is enabled, in part, by advances in DNA sequencing, physiological and environmental monitoring, advanced imaging, and behavioral tracking. Our ability to understand and act upon these observations at equally high precision is driven by advances in genome editing, celular reprogramming, tissue engineering, and information technologies, especially artificial intelligence.

This is what high definition medicine is about:

the dynamic assessment, management, and understanding of an individual’s health measured at (or near) its most basic units. It is the data-driven practice of medicine through the utilization of these highly detailed, longitudinal, and multi-parametric measures of the determinants of health to modify disease risk factors, detect disease processes early, drive precise and dynamically adjusted interventions, and determine preventative and therapeutic intervention efficacy from real-world outcomes

In this framework, precision medicine is only a small piece of the engine.

The article published in Cell by scholars from Scripps Translational Science Institute sheds light on the new perspectives of the practice of medicine, a milestone on the current knowledge of life sciences and its application.

***

Catalunya, 1 d'octubre de 2017 · .

Medicine as a data science (5)

A guide to deep learning in healthcare

Some months ago, Mckinsey released a guide to AI for executives. It says:

Deep learning is a type of machine learning that can process a wider range of data resources, requires less data preprocessing by humans, and can often produce more accurate results than traditional machine-learning approaches (although it requires a larger amount of data to do so). In deep learning, interconnected layers of software-based calculators known as “neurons” form a neural network. The network can ingest vast amounts of input data and process them through multiple layers that learn increasingly complex features of the data at each layer. The network can then make a determination about the data, learn if its determination is correct, and use what it has learned to make determinations about new data. For example, once it learns what an object looks like, it can recognize the object in a new image.

Now Nature publishes a helpful review article on deep learning in healthcare.

Some of the greatest successes of deep learning have been in the field of computer vision (CV). CV focuses on image and video understanding, and deals with tasks such as object classification, detection, and segmentation—which are useful in determining whether a patient’s radiograph contains malignant tumors

The next step is speech and text. Some advances are already available. Basically, Tensorflow by Google is feeding the beast.

PS. WHO and the classification of digital health interventions 1.0

PS. And the book to read:

Laboratory medicine as a data science (2)

 Recent evolutions of machine learning applications in clinical laboratory medicine

You'll find an interesting review about laboratory medicina and machine learning in this article, with applicatons to chemical chemistry, hematology and microbiology.

There has been a recent rise in various ML applications in the field of clinical laboratory medicine. Despite the potential of ML to ameliorate the efficiency of laboratory processes and optimize diagnostic workflows, translation into routine practice is still slow-going. There is a need to raise more awareness about the vast ML landscape among laboratory professionals. Educational programs dealing with theoretical ML concepts as well as their associated challenges and opportunities could stimulate wider acceptance and exploitation in the clinical laboratory. It is important to realize that ML will not  immediately function as a surrogate of the laboratory professional’s neural networks, but will rather act as a valuable supportive tool with the capability of increasing the odds on optimal outcomes for patients accessing health care.

 Margaret Huntington Boehner

Medicine as a data science (4)

The practical implementation of artificial intelligence technologies in medicine

One of the critical issues for AI implementation in clinical practice is about privacy. In this article you'll find a clear statement on the impact of EU regulation:

The GDPR will affect AI implementation in healthcare in several ways. First, it requires explicit and informed consent before any collection of personal data. Informed consent has been a long-standing component of medical practice (unlike in social media or onlinebased marketing), but having to obtain informed consent for an  collection of data still represents a higher bar than obtaining consent for specific items, such as procedures or surgical interventions. Second, the new regulation essentially lends power to the person providing the data to track what data is being collected and to be able to request removal of their data. In the healthcare context, this will shift some of the power balance toward the patient and highlights the importance of ongoing work needed to protect patient privacy and to determine appropriate governance regarding data ownership. 

More details inside.

Potential roles of AI-based technologies in healthcare.

 Integration of patient health information at multiple interfaces.

A bit worse before it gets better

Toward Precision Medicine: Building a Knowledge Network for Biomedical Research and a New Taxonomy of Disease

A new mental frame was created some weeks ago when President Obama gave a speach on the creation of the initiative on Precision Medicine. To be honest, the term was in the title of a 2011 report by IOM.
In my opinion, it is a bundle: stratified medicine+big data+regulatory science+... This is the bundle of the new buzzword, and unless new details arise, nothing specially new.
Now the New Yorker speaks abouts the problems of precision medicine, and focuses on the risks. The final paragraph illustrates the issue:

For Solomon, genetics is simply a new tool with a learning curve, the same as any other. “When the electrocardiogram was first developed, about a hundred years ago, most physicians thought it was voodoo,” Solomon said. “Now, if you don’t understand it, then you shouldn’t be practicing medicine.” But Mary Norton sees that analogy as too simplistic. The pace of genetics research, the variability of test methods and results, and the aura of infallibility with which the tests are marketed, she told me, make this advance a more complicated one than the EKG. Norton believes that, as genetics becomes increasingly integrated into medical care, “over time everyone will come to have a better understanding of genetics.” But, as the demand for DNA testing increases, she says, “it will probably be a bit worse before it gets better.”

Could we avoid the initial bit worse of  "imprecision of stratified medicine"? . I'm full convinced that appropriate regulatory efforts could mitigate such impact. Unfortunately, governments are on vacation.

Validesa i utilitat de l'òmica

Evolution of Translational Omics: Lessons Learned and the Path Forward

 L'"Òmica" és un terme que abasta múltiples disciplines moleculars, que impliquen la caracterització dels conjunts globals de molècules biològiques, com ara ADN, ARN, proteïnes, i metabòlits. Per exemple, la genòmica investiga milers de seqüències d'ADN, la  transcriptòmica investiga totes o moltes transcripcions de gens, la proteòmica investiga un gran nombre de proteïnes, i metabolòmica investiga grans conjunts de metabòlits.

Així comença el llibre de l'IOM sobre una qüestió fonamental de la medicina dels nostres dies.  I el més interessant és com explica la diferència entre l'òmica translacional i els biomarcadors. Malgrat la dificultat que presenta l'avaluació d'un biomarcador, els reptes al que s'enfronta l'òmica són molt superiors. Diu clarament a l'inici:

The complexity of omics research also makes data provenance more challenging and makes sharing of the complex data sets and computational models difficult, which limits the ability of other scientists to replicate and verify the findings and conclusions of omics research studies. Database repositories for genomic data sets are available, but data sharing is not routine, and  without access to the data sets or a precisely defined computational model, replication and  verification are more difficult than for single biomarker tests. While independent confirmation studies are expensive, the need for replication is beneficial in the omics field given the data  complexities that can lead to errors, from simple data management errors to incorrectly  designed computational models. This level of complexity does not exist for single-biomarker  test research, development, and validation.

Massa sovint es vol fer passar aquesta complexitat com inadvertida. I afegeix:

Many hope that the promise that omics science holds for medicine and public health will be realized. With the creation of high-throughput measurement technologies, it is now feasible to take a snapshot of a patient’s molecular profile at specific stages in the progression of disease pathology or at a given location in the body. However, the complexity of these technologies and of the resulting high-dimensional data introduces major challenges for the scientific community, as rigorous statistical, bioinformatics, laboratory, and clinical procedures are required to develop and validate these tests and evaluate their clinical usefulness.

Sobre el tipus de dades òmiques heu d'anar a la pàgina 40 i llegir-ho amb deteniment. Quan un acaba de comprendre el que s'explica de forma planera, aleshores s'adona que els que venen genoma i prou s'han quedat curts, la complexitat és notòria. I en especial la referent a l'epigenoma, del que ja n'he parlat repetidament en aquest blog. El capítol sobre avaluació de les proves esdevé clau. Només fa referència a validesa analítica i clínica, però és el principi sense el qual tots aquells que es plantegin fer cost-efectivitat no podran treballar. I cap al final trobo aquesta conclusió:

 A well-designed test development plan addresses a clinically meaningful question and employs rigorous test discovery, development, and validation procedures. This includes locking down all aspects of an omicsbased test prior to evaluation for clinical utility and use and avoiding overlap between discovery and validation specimens. Choosing an appropriate clinical/biological validation strategy and interacting with FDA prior to initiation of validation studies also reflect a well-designed test development plan. Making data and code available are critical aspects of test development because it enables external verification of the results and generation of additional insights that can advance science and patient care.

El rigor s'imposa i traduir la recerca en aplicacions obliga a comprendre el valor que aporten a la societat. El camí és llarg malgrat sovint apareix als diaris com que és bufar i fer ampolles.

PS. Es poden patentar les proves genòmiques? Avui un tribunal decideix, ho trobareu a WSJ.

PS. Ekaizer a 8TV, fonamental. I també a RAC1

PS. Al Diccionario RAE queda més clar encara: macarra. 1. adj. Dicho de una persona: Agresiva, achulada.

Eliseu Meifren, podeu veure'l a Sant Feliu de Guixols, paga la pena.

Medicine as a data science (7)

Artificial Intelligence in Health Care: The Hope, the Hype, the Promise, the Peril

The National Academy of Medicine’s Special Publication: Artificial Intelligence in Health Care: The Hope, The Hype, The Promise, The Peril synthesizes current knowledge to offer a reference document for relevant health care stakeholders such as: AI model developers, clinical implementers, clinicians and patients, regulators, and policy makers, to name a few. It outlines the current and near-term AI solutions; highlights the challenges, limitations, and best practices for AI development, adoption, and maintenance; offers an overview of the legal and regulatory landscape for AI tools designed for health care application; prioritizes the need for equity, inclusion, and a human rights lens for this work; and outlines key considerations for moving forward.

A must read

Precision medicine: a deep breakthrough in life sciences paradigm

Bioscience - Lost in Translation? How precision medicine closes the innovation gap

It is not so easy to translate knowledge into practice, and this is the case of biosciences into clinical applications. However, recently this trend is accelerating and precision medicine is emerging. A new book gives us the highlights to understand precisely what's going on: Bioscience - Lost in Translation? How precision medicine closes the innovation gap.

Richard Barker (the author of 2030 - The future of medicine) says:

The classic definition of diseases has been in terms of the symptoms they cause and/ or where in the body they appear. This was the best that medicine could do when external observation of the patient was the only or primary means of diagnosing disease. The  powerful new tools of molecular biology are reinterpreting disease in terms of aberrant,
defective, or unbalanced molecular mechanisms at the cellular, organ, or organism level. Molecular level diagnosis becomes a real possibility. Such an approach brings effective therapy immediately closer. Molecular diagnostics can separate diseases with similar symptoms but different underlying causes— and often suggest a different starting point for intervention.

If this is so, what should we do?

The seven changes of mindset and of practice are:
1. Advance the molecular definition of disease and the application of systems biology. We need a more decisive move from a classic definition of diseases— in terms of the symptoms they cause and/ or where in the body they appear— to a definition in terms of aberrant, defective, or unbalanced molecular mechanisms at the cellular level. And we need to marry this with a recognition that singular target- based innovation rarely works: we need a systems biology approach.

2. Partner academia and industry in more collaborative, impact- oriented research. We need to extend the ‘open innovation’ approach in which academia and companies invest together and share IP. We need to define new pre- or non- competitive spaces, especially in work on disease mechanisms and disease models. And we need to provide for new types of links and incentives to break down the barriers between these two worlds. 

3. Move decisively to a more adaptive approach to development, trial and approval design. We need to build on successful experiments in more flexible trial design, development pathways, and regulatory appraisal to a globally accepted adaptive approach. This involves collaborative design of the evidence package needed to secure approval and reimbursement, and greater teamwork through the process. 

4. Create new reward and financing vehicles for leading edge innovation. We need to move from reward systems based purely on unit sales of products, irrespective of outcome, to rewarding innovators for positive outcomes, patient by patient. We also need to design financing mechanisms that bridge between cost- effectiveness and affordability. We must be able to accommodate high- cost precision therapies that offer cures and so generate long- term returns for the system.

5. Engineer tools and systems for faster and better innovation adoption and adherence. We need to move from reliance solely on promotion to doctors and passive patient participation to a disciplined approach to establishing new pathways of care. These will be based on modern behavioural science, clinical decision support, and other digital technologies.

6. Develop an infrastructure for real- world data- driven learning. We now have the opportunity to study in large populations how lifestyle and treatment choices lead
to outcomes, learning from every patient as if in a clinical trial. New analytical tools will empower this.

7. Bring patients into the mainstream of decision- making and engage them  hole heartedly throughout the process. It is time to move from a process and mindset in which patients are regarded as passive subjects for clinical trials and recipients of products and procedures. Their input and engagement needs to be sought along the whole innovation chain: on treatment benefits, acceptable risks, optimal clinical trial design, adherence support, and outcomes.

Highly recommended.

Medicine as a data science (6)

Adapting to Artificial Intelligence: Radiologists and Pathologists as Information Specialists

While some physicians are lobbying for creating more specialties, Jah and Topol argue exactly the opposite. Radiology, pathology and in vitro diagnostics should be under the same umbrella: "the information specialists":

Because pathology and radiology have a similar past and a common destiny,
perhaps these specialties should be mergedinto a single entity, the “information specialist,” whose responsibility will not be so much to extract information from images and histology but to manage the information extracted by artificial intelligence in the clinical context of the patient.

 There may be resistance to merging 2 distinct medical specialties, each of which has unique pedagogy, tradition, accreditation,and reimbursement.However, artificial intelligence will change these diagnostic fields. The merger is a natural fusion of human talent and artificial intelligence. United, radiologists and pathologists can thrive with the rise of artificial intelligence. 

The history of automation in the broader economy has a reassuring message. Jobs are not lost; rather, roles are redefined; humans are displaced to tasks needing a human element. Radiologists and pathologists need not fear artificial intelligence but rather must adapt incrementally to artificial intelligence, retaining their own services for cognitively challenging tasks.A unified discipline, information specialists would best be able to captain artificial intelligence and guide medical information to improve patient care.

You may agree or not. Technology is breaking barriers and creating bridges. Food for thought.

Josep Segú - Brooklyn Bridge

Exhibition at Sala Parés

Understanding Ethics of AI

 The Oxford Handbook of ETHICS OF AI

The approach to the ethics of AI that runs through this handbook is contextual in four senses:

• it locates ethical analysis of artificial intelligence in the context of other modes of normative analysis, including legal, regulatory, philosophical, and policy approaches,

• it interrogates artificial intelligence within the context of related modes of technological innovation, including machine learning, Big Data, and robotics,

• it is interdisciplinary from the ground up, broadening the conversation about the ethics of artificial intelligence beyond computer science and related fields to include other fields of scholarly endeavor, including the social sciences, humanities,and the professions (law, medicine, engineering, etc.), and

• it invites critical analysis of all aspects of—and participants in—the wide and continuously expanding artificial intelligence complex, from production to commercialization to consumption, from technical experts to venture capitalists to self-regulating professionals to government officials to the general public.

Outline

Part I. Introduction & Overview

1. The Artificial Intelligence of Ethics of AI: An Introductory Overview

2. The Ethics of Ethics of AI: Mapping the Field

3. Ethics of AI in Context: Society & Culture

Part II. Frameworks & Modes

4. Why Industry Self-regulation Will Not Deliver 'Ethical AI': A Call for Legally Mandated Techniques of 'Human Rights by Design'

5. Private Sector AI: Ethics and Incentives

6. Normative Modes: Codes & Standards

7. Normative Modes: Professional Ethics

Part III. Concepts & Issues

8. Fairness and the Concept of 'Bias'

9. Accountability in Computer Systems

10. Transparency

11. Responsibility

12. The Concept of Handoff as a Model for Ethical Analysis and Design

13. Race and Gender

14. The Future of Work in the Age of AI: Displacement, Augmentation, or Control?

15. The Rights of Artificial Intelligences

16. The Singularity: Sobering up About Merging with AI

17. Do Sentient AIs Have Rights? If So, What Kind?

18. Autonomy

19. Troubleshooting AI and Consent

20. Is Human Judgment Necessary?

21. Sexuality

IV. Perspectives & Approaches

22. Computer Science

23. Engineering

24. Designing Robots Ethically Without Designing Ethical Robots: A Perspective from Cognitive Science

25. Economics

26. Statistics

27. Automating Origination: Perspectives from the Humanities

28. Philosophy

29. The Complexity of Otherness: Anthropological contributions to robots and AI

30. Calculative Composition: The Ethics of Automating Design

31. Global South

32. East Asia

33. Artificial Intelligence and Inequality in the Middle East: The Political Economy of Inclusion

34. Europe's struggle to set global AI standards

Part V. Cases & Applications

35. The Ethics of Artificial Intelligence in Transportation

36. Military

37. The Ethics of AI in Biomedical Research, Medicine and Public Health

38. Law: Basic Questions

39. Law: Criminal Law

40. Law: Public Law & Policy: Notice, Predictability, and Due Process

41. Law: Immigration & Refugee Law

42. Education

43. Algorithms and the Social Organization of Work

44. Smart City Ethics

Fundamental misconceptions about health economists and economics

Economics: the biggest fraud ever perpetrated on the world?

Twitter is a risky tool. Your short messages are seen worldwide, be careful. Richard Horton, editor of The Lancet, sent 10 tweets about economics and economists. Ten misconceptions, one behind the other. Certain people consider that such sentences doesn't deserve an answer, they only reflect the personality and knowledge of the author. Others, like David Parkin, John Appleby and Alan Maynard think the opposite and they decided to write a comment this week in The Lancet. This is an article for the files. And it fits perfectly as a recommended reading for those that share these controversial views of Richard Horton:

Panel: Tweets from @richardhorton: “Economics, second only to ‘management’, may just be the biggest fraud ever perpetrated on the world.”
The case against economics:
1 The promise economics offers is seductive: how to allocate scarce resources in society.
It’s a false promise.
2 Economists write as if the economy=society, and societal problems=economic problems. The confl ation is false too.
3 Once there was political economy = economics, ethics, politics. Economists have stripped morality from economics, leaving an arid science.
4 The high points of economic thinking are theories, not data. Reliable experimentally derived data are anathema for most economists.
5 Economists see health as an economic good. It is an opportunity cost, with zero intrinsic value.
6 Rationality, for the economist, means subjecting every thought/decision to a cost-benefit analysis. A wholly narrow view of humanity.
7 The big idea in economics is the market. The assumption is that human beings make cost-benefit decisions based only on self-interest.
8 The essence of economics is price. For those in health who argue for access free at point of delivery, we kill the soul of the economist.
9 Economists deny the existence of citizens. They see only consumers.
10 Finally, it’s acceptable to worsen the lives of some provided the gains of others compensate. Economists institutionalise inequality.A sum of nonsense sentences, one behind the other. 

After reading the comment to each of these tweets, it will be difficult to maintain the same position.
And the authors' conclusion:

What motivated Horton’s critical outburst about economics and economists is not clear. More than 40 years ago, an essay by Alan Williams to defend economic evaluation admitted its imperfections, but concluded with Maurice Chevalier’s view on old age: “Well, there is quite a lot I don’t like about it, but it’s not so bad when you consider the alternative!”Economics, like medicine, is imperfect. The challenge for practitioners of each is to ensure that the perfect does not drive out the good. Our practices may at times be imperfect, but that should not inhibit our drive to improve clinical practice and economic activity for the benefit of all our patients and citizens. We all must strive to avoid confused analysis in displays of modest understanding of each other’s work.

El disseny de sistemes de pagament

 A Framework for the Design of Risk-Adjustment Models in Health Care Provider Payment Systems

A partir d'avui aquest blog es trasllada a Substack. Durant unes setmanes serà accessible simultàniament per blogger i per substack. Anoteu l'adreça: econsalut.substack.com

Article resumit amb IA.

Aquest article presenta un marc conceptual integral per al disseny de models d'ajust de risc (RA) en el context de models de pagament prospectiu a proveïdors d'assistència sanitària. L'objectiu és desenvolupar un marc que expliciti les opcions de disseny i les compensacions associades per tal de personalitzar el disseny de l'RA als sistemes de pagament a proveïdors, tenint en compte els objectius i les característiques del context d'interès.

Introducció (1-3): Durant les últimes dècades, els reguladors i els responsables polítics de la salut han fet esforços per millorar l'eficiència de la prestació d'assistència sanitària mitjançant la reforma dels sistemes de pagament a proveïdors. Específicament, l'eficiència s'ha perseguit mitjançant la introducció d'elements prospectius en els models de pagament, donant lloc a diversos Models de Pagament Alternatius (MPA) com els acords de qualitat alternatius i els pagaments agrupats. Aquests MPA tenen com a objectiu incentivar l'eficiència traslladant (part de) la responsabilitat financera dels pagadors als proveïdors. Una característica típica dels pagaments prospectius a proveïdors és que es basen en un "nivell de despesa normatiu" per a la prestació d'un conjunt predefinit de serveis a una determinada població de pacients. El nivell de despesa normatiu es refereix al nivell de despesa que "hauria de ser" depenent de la població de pacients d'un proveïdor, en lloc de la despesa observada. Un element clau en la determinació dels nivells de despesa normatius és la correcció de les diferències sistemàtiques en les necessitats d'assistència sanitària de les poblacions de pacients dels proveïdors, comunament coneguda com a ajust de risc (RA). L'RA és crucial per garantir un terreny de joc igualitari per als proveïdors i per evitar incentius per a comportaments no desitjats, com la selecció de riscos.

Nova Contribució (8-10): Tot i les contribucions conceptuals existents sobre el disseny de l'RA, actualment no hi ha un marc integral per adaptar el disseny de l'RA al pagament de proveïdors i a les característiques essencials del context. Aquest article desenvolupa aquest marc sintetitzant, ampliant i aplicant coneixements de la literatura existent. La metodologia va incloure una revisió de la literatura combinada amb consultes a experts en el camp de l'RA i els sistemes de pagament. La informació recopilada es va sintetitzar per desenvolupar el marc, del qual van sorgir tres criteris per al disseny de models d'RA i es van agrupar les opcions i les compensacions en dues dimensions principals: (a) la tria dels ajustadors de risc i (b) la tria de les ponderacions de pagament.

Definicions de Conceptes Clau (11-13): Els models de pagament prospectiu i els MPA traslladen la responsabilitat financera dels pagadors als proveïdors per tal d'incentivar el control de costos i l'eficiència. Qualsevol trasllat de responsabilitat financera requereix que el pagador determini el nivell de despesa normatiu, que reflecteix el nivell de despesa apropiat donades les necessitats d'assistència sanitària d'una població i els objectius dels MPA. El nivell de despesa normatiu no fa referència necessàriament al nivell de despesa absolut o òptim, sinó al nivell considerat apropiat donat el nivell/objectius d'eficiència perseguits pel MPA.

Fonts de Variació de la Despesa i el Paper de l'RA i la Mancomunació de Riscos (14-19): Quan s'estableixen nivells de despesa normatius, és important considerar tres fonts de variació de la despesa: (a) variació sistemàtica impulsada per factors fora del control dels proveïdors (variables C o "factors de compensació"), (b) variació sistemàtica impulsada per factors que els proveïdors poden influir (variables R o "factors de responsabilitat"), i (c) variació aleatòria. Per evitar que els proveïdors assumeixin riscos excessius que no poden influir, els MPA solen aplicar alguna forma de mancomunació de riscos. L'RA prospectiu s'utilitza per compensar la variació de la despesa deguda a les variables C. La naturalesa i el grau en què s'ha de compensar la variació de la despesa resultant de les variables C forma el punt de partida d'un model d'RA.

Tres Criteris per al Disseny de Models d'RA (19-26): L'objectiu general de l'RA en els MPA és compensar els proveïdors per la variació de la despesa deguda a les variables C, alhora que els manté responsables de la variació de la despesa deguda a les variables R. Això implica dos criteris clau: (a) compensació adequada per a les variables C i (b) cap compensació per a les variables R. Un tercer criteri important és la viabilitat.

• Criteri 1: Compensació Adequada per a les Variables C (20-26): Per evitar problemes de selecció, l'RA hauria de compensar adequadament les variables C que són rellevants a la llum de les possibles accions de selecció de riscos per part dels proveïdors (atraure/dissuadir pacients sans/no sans). També hauria de compensar les variables C que varien entre les poblacions de proveïdors per evitar la participació selectiva en el MPA.
• Criteri 2: Cap Compensació per a les Variables R (26-29): Per evitar ineficiències, l'RA no hauria de compensar les variables R. La compensació per la variació de la despesa de les variables R pot donar lloc a problemes d'eficiència, com la perpetuació de les ineficiències existents ("biaix d'status quo") i la creació d'incentius per a noves ineficiències (reducció dels incentius per al control de volum i preu, codificació ascendent).
• Criteri 3: Viabilitat (29-30): Un tercer criteri crucial és la viabilitat, que inclou la disponibilitat de dades i l'acceptació per part de totes les parts interessades (pacients, proveïdors, pagadors, reguladors).

Un Marc per al Disseny de Models d'RA (30-31): Aquest marc distingeix entre preguntes de disseny, opcions associades i consideracions i compensacions clau pel que fa a (a) la tria dels ajustadors de risc i (b) la tria de les ponderacions de pagament.

La Tria dels Ajustadors de Risc (31-47): Aquesta secció aborda tres preguntes principals de disseny:

• Quin tipus d'informació es basa els ajustadors de risc? (32-38): Les opcions inclouen informació demogràfica, socioeconòmica, subjectiva (de salut), diagnòstica, d'utilització, clínica, de despesa (retardada) i del costat de l'oferta. L'ús d'informació endògena (diagnòstics, utilització, despesa) és altament predictiu de la despesa de tipus C, però pot perpetuar ineficiències i introduir nous incentius perversos per al volum i el preu. L'ús d'informació exògena (demogràfica, socioeconòmica) no manté ni introdueix incentius perversos relacionats amb el volum o el preu, però el seu poder predictiu és generalment baix.
• A quin període de temps (període base) pertany la informació? (38-45): Es pot distingir entre ajustadors concurrrents i prospectius. Els efectes d'incentiu relatius d'aquestes opcions no estan clars a priori.
• Com dissenyar els ajustadors de risc? (46-47): Això inclou l'especificació de l'escala de mesura, l'operacionalització dels ajustadors (considerant condicions, jerarquies, restriccions) i les interaccions entre ajustadors.

La Tria de les Ponderacions de Pagament (48-60): Per trobar ponderacions de pagament apropiades, els responsables de la presa de decisions s'enfronten a tres decisions principals de disseny:

• Quina mostra d'estimació? (49-52): Es requereix una mostra d'estimació representativa de la població d'interès i dels nivells de despesa normatius. En la pràctica, sovint s'utilitzen dades històriques i poblacions de pacients similars.
• Quines intervencions de dades? (52-58): Quan la mostra d'estimació no és representativa, s'han de considerar intervencions de dades sobre la població de pacients i/o les dades de despesa per millorar la coincidència amb la població d'interès i el nivell de despesa normatiu. Això pot incloure correccions per biaixos i inequitats.
• Com derivar les ponderacions de pagament? (59-60): Això implica decidir quins ajustadors de risc incloure (considerant el biaix de la variable omesa) i quin criteri d'optimització utilitzar per estimar aquestes ponderacions. Les opcions van des de criteris d'optimització estàndard (OLS, GLM) fins a criteris personalitzats (regressió restringida, aprenentatge automàtic).

La Interconnexió Entre les Opcions de Disseny per als Ajustadors de Risc i les Ponderacions de Pagament (61-62): Les decisions de disseny dins i entre aquests dos temes estan altament interrelacionades. Per exemple, la tria de la informació en què es basen els ajustadors de risc afectarà la seva especificació i operacionalització. De la mateixa manera, les decisions sobre com es deriven les ponderacions de pagament depenen tant de la tria dels ajustadors de risc com de la tria de la mostra d'estimació (modificada).

Discussió (63-68): No hi ha un enfocament únic per al disseny de models d'RA, i el disseny adequat pot variar segons la configuració i les evolucions al llarg del temps. És crucial la decisió normativa sobre quines variables es consideren C i quines R. L'abast de la preocupació pels possibles incentius de selecció i control de costos pot variar segons el context. Les consideracions de viabilitat, com la disponibilitat de dades i l'acceptació de les parts interessades, també són importants.

Consideracions Més Amplies per al Disseny de l'RA en el Finançament de l'Assistència Sanitària (69-70): Tot i que aquest article se centra en el pagament a proveïdors, el marc proposat també podria beneficiar altres reformes de finançament, com les iniciatives de participació del consumidor, tot i que es necessita més recerca.

Conclusió (71): El disseny de models d'RA per a sistemes de pagament prospectiu a proveïdors és un exercici complex que requereix una consideració explícita de moltes preguntes, opcions i compensacions difícils. El procés de disseny ha de guiar-se per tres criteris clau: compensació adequada de les variables C, cap compensació de les variables R i viabilitat. Les diverses preguntes i opcions de disseny es poden classificar en la tria dels ajustadors de risc i la tria de les ponderacions de pagament. Es necessita més recerca per donar suport a les decisions normatives sobre les variables C i R, així com per desenvolupar mètriques d'avaluació integrals per a la valoració dels efectes dels incentius.

Referències

A continuació es mostren les referències citades en les fonts:

• Adams Dudley, R., Medlin, C. A., Hammann, L. B., Cisternas, M. G., Brand, R., Rennie, D. J., & Luft, H. S. (2003). The best of both worlds? Potential of hybrid prospective/concurrent risk adjustment. Medical Care, 41(1), 56–69.
• American Medical Association. (2019). Improving Risk adjustment in alternative payment models.
• Andersen, R., & Newman, J. F. (2005). Societal and individual determinants of medical care utilization in the United States. The Milbank Memorial Fund Quarterly. Health and Society, 83(4), 1468-0009.2005.00428.x.
• Anderson, G. F., & Weller, W. E. (1999). Methods of reducing the financial risk of physicians under capitation. Archives of Family Medicine, 8(2), 149–155.
• Anthun, K. S. (2021). Predicting diagnostic coding in hospitals: Individual level effects of price incentives. International Journal of Health Economics and Management, 22(2), 129–146.
• Arrow, K. J. (2004). Uncertainty and the welfare economics of medical care. Bulletin of the World Health Organization, 82(2), 141–149.
• Ash, A. S., & Ellis, R. P. (2012). Risk-adjusted payment and per-formance assessment for primary care. Medical Care, 50(8), 643–653.
• Ash, A. S., Mick, E. O., Ellis, R. P., Kiefe, C. I., Allison, J. J., & Clark, M. A. (2017). Social determinants of health in managed care payment formulas. JAMA Internal Medicine, 177(10), 1424–1430.
• Bäuml, M. (2021). How do hospitals respond to cross price incen-tives inherent in diagnosis-related groups systems? The impor-tance of substitution in the market for sepsis conditions. Health Economics, 30(4), 711–728.
• Bergquist, S. L., Layton, T. J., McGuire, T. G., & Rose, S., & National Bureau of Economic Research. (2018). Intervening on the data to improve the performance of health plan pay-ment methods (Ser. NBER working paper series, no. w24491). National Bureau of Economic Research.
• Brown, J., Duggan, M., Kuziemko, I., & Woolston, W. (2014). How does risk selection respond to risk adjustment? New evi-dence from the Medicare Advantage Program. The American Economic Review, 104(10), 3335–3364.
• Buchner, F., Wasem, J., & Schillo, S. (2017). Regression trees iden-tify relevant interactions: Can this improve the predictive per-formance of risk adjustment? Health Economics, 26(1), 74–85.
• Cattel, D., Eijkenaar, F., & Schut, F. T. (2020). Value-based pro-vider payment: Towards a theoretically preferred design. Health Economics, Policy and Law, 15(1), 94–112.
• Chang, H.-Y., Lee, W.-C., & Weiner, J. P. (2010). Comparison of alternative risk adjustment measures for predictive modeling: High risk patient case finding using Taiwan’s national health insurance claims. BMC Health Services Research, 10, 343– 343. https://doi.org/10.1186/1472-6963-10-343.
• Chernew, M. E., Mechanic, R. E., Landon, B. E., & Safran, D. G. (2011). Private-payer innovation in Massachusetts: The “alter-native quality contract.” Health Affairs, 30(1), 51–61.
• Chien, A. T., Newhouse, J. P., Iezzoni, L. I., Petty, C. R., Normand, S.-L. T., & Schuster, M. A. (2017). Socioeconomic background and commercial health plan spending. Pediatrics, 140(5).
• Constantinou, P., Tuppin, P., Gastaldi-Ménager, C., & Pelletier-Fleury, N. (2022). Defining a risk-adjustment formula for the introduction of population-based payments for primary care in France. Health Policy, 126(9), 915–924.
• Dafny, L. S. (2005). How do hospitals respond to price changes? The American Economic Review, 95(5), 1525–1547.
• Douven, R., McGuire, T. G., & McWilliams, J. M. (2015). Avoiding unintended incentives in ACO payment models. Health Affairs, 34(1), 143–149.
• Douven, R., Remmerswaal, M., & Mosca, I. (2015). Unintended effects of reimbursement schedules in mental health care. Journal of Health Economics, 42, 139–150.
• Dowd, B. E., Huang, T-y, & McDonald, T. (2021). Tiered cost-sharing for primary care gatekeeper clinics. American
• Duan, N., Manning, W. G., Morris, C. N., & Newhouse, J. P. (1983). A comparison of alternative models for the demand for medical care. Journal of Business & Economic Statistics, 1(2), 115–126.
• Durfey, S. N. M., Kind, A. J. H., Gutman, R., Monteiro, K., Buckingham, W. R., DuGoff, E. H., & Trivedi, A. N. (2018). Impact of risk adjustment for socioeconomic status on Medicare advantage plan quality rankings. Health Affairs, 37(7), 1065–1072.
• Eijkenaar, F., van Vliet, R. C. J. A., & van Kleef, R. C. (2018). Diagnosis-based cost groups in the Dutch risk-equalization model: Effects of clustering diagnoses and of allowing patients to be classified into multiple risk-classes. Medical Care, 56(1), 91–96.
• Einav, L., Finkelstein, A., Ji, Y., & Mahoney, N. (2022). Voluntary regulation: Evidence from Medicare payment reform. The Quarterly Journal of Economics, 137(1), 565–618.
• Ellis, R. P. (1998). Creaming, skimping and dumping: Provider competition on the intensive and extensive margins. Journal of Health Economics, 17(5), 537–555.
• Ellis, R. P., Martins, B., & Rose, S. (2018). Risk adjustment for health plan payment. In T. G. McGuire & R. C. van Kleef (Eds.), Risk adjustment, risk sharing and premium regu-lation in health insurance markets: Theory and practice (pp. 55–104). Elsevier.
• Epstein, A. M., & Cumella, E. J. (1988). Capitation payment: Using predictors of medical utilization to adjust rates. Health Care Financing Review, 10(1), 51–69.
• García-Goñi, M., Ibern, P., & Inoriza, J. M. (2009). Hybrid risk adjustment for pharmaceutical benefits. The European Journal of Health Economics, 10(3), 299–308.
• Geruso, M., & Layton, T. (2020). Upcoding: Evidence from Medicare on squishy risk adjustment. Journal of Political Economy, 128(3), 984–1026.
• Geruso, M., & McGuire, T. G. (2016). Tradeoffs in the design of health plan payment systems: Fit, power and balance. Journal of Health Economics, 47, 1–19.
• Gilmer, T., Kronick, R., Fishman, P., & Ganiats, T. G. (2001). The Medicaid Rx model: Pharmacy-based risk adjustment for pub-lic programs. Medical Care, 39(11), 1188–1202.
• Glazer, J., & McGuire, T. G. (2002). Setting health plan premiums to ensure efficient quality in health care: Minimum variance optimal risk adjustment. Journal of Public Economics, 84(2), 153–173.
• Gravelle, H., Sutton, M., Morris, S., Windmeijer, F., Leyland, A., Dibben, C., & Muirhead, M. (2003). Modelling supply and demand influences on the use of health care: Implications for deriving a needs-based capitation formula. Health Economics, 12(12), 985–1004.
• Hayen, A. P., van den Berg, M. J., Meijboom, B. R., Struijs, J. N., & Westert, G. P. (2015). Incorporating shared savings pro-grams into primary care: From theory to practice. BMC Health Services Research, 15, Article 580.
• Heckman, J. J., & Honoré, B. E. (1990). The empirical content of the Roy model. Econometrica, 58(5), 1121–1149.
• Hildebrandt, H., Hermann, C., Knittel, R., Richter-Reichhelm, M., Siegel, A., & Witzenrath, W. S. (2010). Gesundes Kinzigtal Integrated Care: Improving population health by a shared health
• Hollenbeak, C. S. (2005). Functional form and risk adjustment of hospital costs: Bayesian analysis of a Box-Cox random coef-ficients model. Statistics in Medicine, 24(19), 3005–3018. https://doi.org/10.1002/sim.2172.
• Horn, S. D., Horn, R. A., & Sharkey, P. D. (1984). The severity of illness index as a severity adjustment to diagnosis-related groups. Health Care Financing Review, 1984, 33–45.
• Hughes, J. S., Averill, R. F., Eisenhandler, J., Goldfield, N. I., Muldoon, J., Neff, J. M., & Gay, J. C. (2004). Clinical risk groups (CRGs): A classification system for risk-adjusted capitation- based payment and health care management. Medical Care, 42(1), 81–90.
• Humbyrd, C. J., Hutzler, L., & DeCamp, M. (2019). The ethics of success in bundled payments: Respect, beneficence, and social justice concerns. American Journal of Medical Quality, 34(2), 202–204.
• Jha, A. (2014) The Health Care Blog. https://thehealthcareblog. com/blog/2014/09/29/changing-my-mind-on-ses-adjustment/
• Iezzoni, L. I. (2012). Risk adjustment for measuring health care outcomes (3rd ed.). Health Administration Press.
• Iommi, M., Bergquist, S., Fiorentini, G., & Paolucci, F. (2022). Comparing risk adjustment estimation methods under data availability constraints. Health Economics, 31(7), 1368–1380.
• Isaksson, D., Blomqvist, P., Pingel, R., & Winblad, U. (2018). Risk selection in primary care: A cross-sectional fixed effect analysis of Swedish individual data. BMJ Open, 8(10), Article e020402.
• Jones, A. (2010). Models for health care [Working papers]. HEDG, c/o Department of Economics, University of York, Health, Econometrics and Data Group (HEDG).
• Kapur, K., Tseng, C. W., Rastegar, A., Carter, G. M., & Keeler, E. (2003). Medicare calibration of the clinically detailed risk information system for cost. Health Care Financing Review, 25(1), 37–54.
• Lamers, L. M. (1998). Risk-adjusted capitation payments: Developing a diagnostic cost groups classification for the Dutch situation. Health Policy, 45(1), 15–32.
• Lamers, L. M. (1999). Pharmacy costs groups. Medical Care, 37(8), 824–830.
• Lamers, L. M., & van Vliet, R. C. J. A. (2003). Health-based risk adjustment; Improving the pharmacy-based cost group model to reduce gaming possibilities. European Journal of Health Economics, 4(2), 107–114.
• Layton, T. J., McGuire, T. G., & van Kleef, R. C. (2018). Deriving risk adjustment payment weights to maximize efficiency of health insurance markets. Journal of Health Economics, 61, 93–110.
• Lemke, K. W., Pham, K., Ravert, D. M., & Weiner, J. P. (2020). A revised classification algorithm for assessing emergency department visit severity of populations. American Journal of Managed Care, 26(3), 119–125. https://doi.org/10.37765/ ajmc.2020.42636.
• Levy, S., Bagley, N., & Rajkumar, R. (2018). Reform at risk— mandating participation in alternative payment plans. The New England Journal of Medicine, 378(18), 1663–1665.
• Liao, J. M., Pauly, M. V., & Navathe, A. S. (2020). When should Medicare mandate participation in alternative payment mod-els? Health Affairs, 39(2), 305–309.
• Martinussen, P. E., & Hagen, T. P. (2009). Reimbursement sys-tems, organisational forms and patient selection: Evidence from day surgery in Norway. Health Economics, Policy, and Law, 4(Pt. 2), 139–158.
• McGuire, T. G. (2000). Physician agency. In A. J. Culter & J. P. Newhouse (Eds.), Handbook of health economics (1st ed., Vol. Volume 1a, Handbooks in economics, p. 17). Elsevier.
• McGuire, T. G. (2011) Physician agency and payment for primary medical care. In S. Glied & P. C. Smith (Eds), Chapter 9: The Oxford handbook of health economics (online ed., pp. 462– 528). Oxford Academic.
• McWilliams, J. M., Hatfield, L. A., Landon, B. E., Hamed, P., & Chernew, M. E. (2018). Medicare Spending after 3 years of the Medicare shared savings program. The New England Journal of Medicine, 379(12), 1139–1149.
• McWilliams, J. M., Weinreb, G., Ding, L., Ndumele, C. D., & Wallace, J. (2023). Risk adjustment and promoting health equity in population-based payment: Concepts and evidence: Study examines accuracy of risk adjustment and payments in promoting health equity. Health Affairs, 42(1), 105–114.
• Newhouse, J. P. (1994). Patients at risk: Health reform and risk adjustment. Health Affairs, 13(1), 132–146.
• Newhouse, J. P., Price, M., McWilliams, J. M., Hsu, J., & McGuire, T. G. (2015). How much favorable selection is left in Medicare advantage? American Journal of Health Economics, 1(1), 1– 26. https://doi.org/10.1162/ajhe_a_00001.
• Politzer, E. (2024). Utilization thresholds in risk adjustment sys-tems. American Journal of Health Economics, 10(3), 470–503.
• Pope, G. C., Kautter, J., Ellis, R. P., Ash, A. S., Ayanian, J. Z., Lezzoni, L. I., Ingber, M. J., Levy, J. M., & Robst, J. (2004). Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financing Review, 25(4), 119–141.
• Pope, G. C., Kautter, J., Ingber, M. J., Freeman, S., Sekar, R., & Newhart, C. (2011). Evaluation of the CMS-HCC Risk Adjustment Model [Prepared by RTI]. Centers for Medicare & Medicaid Services, Medicare Plan Payment Group, Division of Risk Adjustment and Payment Policy.
• Porter, M. E., & Kaplan, M. S. (2016). How to pay for health care? Harvard Business Review, 94(7/8), 88–102.
• Robinson, J. C., Whaley, C., Brown, T. T., & Dhruva, S. S. (2020). Physician and patient adjustment to reference pricing for drugs. JAMA Network Open, 3(2), Article e1920544.
• Rose, S. (2016). A machine learning framework for plan payment risk adjustment. Health Services Research, 51(6), 2358–2237.
• Rose, S., Zaslavsky, A. M., & McWilliams, J. M. (2016). Variation in accountable care organization spending and sensitivity to risk adjustment: Implications for benchmarking. Health Affairs, 35(3), 440–448.
• Schokkaert, E., & Van de Voorde, C. (2004). Risk selection and the specification of the conventional risk adjustment formula. Journal of Health Economics, 23, 1237–1259.
• Schokkaert, E., & Van de Voorde, C. (2006). Incentives for risk selection and omitted variables in the risk adjustment formula. Annales d’economie et de Statistique, 83-84, 327–352.
• Struijs, J. N., de Vries, E. F., Baan, C. A., van Gils, P. F., & Rosenthal, M. B. (2020). Bundled-payment models around the world: How they work and what their impact has been [Issue brief]. The Commonwealth Fund.
• van Barneveld, E. M., Lamers, L. M., van Vliet René, C. J. A., & van de Ven, W. P. M. M. (2001). Risk sharing as a supplement to imperfect capitation: A tradeoff between selection and effi-ciency. Journal of Health Economics, 20(2), 147–168.
• van de Ven, W. P. M. M., & Ellis, R. P. (2000). Risk adjustment in competitive health plan markets. In A. J. Collyer & J. P. Newhouse (Eds.), Handbook of health economics (pp. 1003– 1092). Elsevier Science.
• van de Ven, W. P. M. M., & van Vliet, R. C. J. A. (1992). How can we prevent cream skimming in a competitive health insur-ance market? The great challenge for the ‘90’s. In P. Zweifel & H. E. French (Eds.), Health economics worldwide (pp. 23–46). Kluwer Academic Publishers.
• van Kleef, R. C., Eijkenaar, F., van Vliet René, C. J. A., & Nielen, M. M. J. (2020). Exploiting incomplete information in risk adjustment using constrained regression. American Journal of Health Economics, 6(4), 477–497.
• van Kleef, R. C., McGuire, T. G., van Vliet René, C. J. A., & van de Ven, W. P. P. M. (2017). Improving risk equalization with constrained regression. The European Journal of Health Economics: Health Economics in Prevention and Care, 18(9), 1137–1156.
• van Kleef, R. C., & Reuser, M. (2021). How the covid-19 pan-demic can distort risk adjustment of health plan payment. The European Journal of Health Economics, 22(7), 1005–1016.
• van Kleef, R. C., & van Vliet René, C. J. A. (2012). Improving risk equalization using multiple-year high cost as a health indicator. Medical Care, 50(2), 140–144.
• Veen, S. H. C. M., Kleef, R. C., Ven, W. P. M. M., & Vliet, R. C. J. A. (2018). Exploring the predictive power of interaction terms in a sophisticated risk equalization model using regres-sion trees. Health Economics, 27(2), 12.
• Vermaas, A. (2006). Agency, managed care and financial-risk sharing in general medical practice [Dissertation]. Erasmus Universiteit.
• Werbeck, A., Wübker, A., & Ziebarth, N. R. (2021). Cream skim-ming by health care providers and inequality in health care access: Evidence from a randomized field experiment. Journal of Economic Behavior and Organization, 188, 1325–1350.
• Withagen-Koster, A. A., van Kleef, R. C., & Eijkenaar, F. (2020). Incorporating self-reported health measures in risk equaliza-tion through constrained regression. The European Journal of Health Economics, 21(4), 513–528.