Machine learning in Medicine

Machine Learning and Evidence-Based Medicine

While waiting for the new book by Eric Topol: Deep Medicine, let's have a look at this article, and at this summary table:

 Xavier Miserachs

Health spending in late life

Predictive modeling of U.S. health care spending in late life

In US, it is said that a quarter of public expenditure for the elderly (Medicare) is spent in the last 12 months of life. Really what happens is that the last year is only close to 10% of the whole lifetime health spending. Anyway, a new article in Science highlights commmon misunderstandings on such figure and disentangles the fundamentals.

These common interpretations of end-of-life spending flirt with a statistical fallacy: Those who endup dying are not the same as those who were sure to die. Ex post, spending could appear concentrated on the dead, simply because we spend more on sicker individuals who have higher mortality—even if we never spent money on those certain to die within the year. Empirically, this suggests using predicted mortality, rather than ex post mortality, to assess end of-life spending.

Less than 5% of spending is accounted for by individuals with predicted mortality above 50%. The simple fact that we spend more on the sick—both on those who recover and those who die—accounts for 30 to 50% of the concentration of spending on the dead. 

Crucial conclusion:

In sum, although spending on the ex post dead is very high, we find there are only a few individuals for whom, ex ante, death is near certain. Moreover, a substantial component of the concentration of spending at the end of life is mechanically driven by the fact that those who end up dying are sicker, and spending, naturally, is higher for sicker individuals. Of course, we do not— and cannot—rule out individual cases where treatment is performed on an individal for whom death is near certain. But our findings indicate that such individuals are not a meaningful share of decedents. These findings suggest that a focus on end-of life spending is not, by itself, a useful way to identify wasteful spending. Instead, researchers must focus on quality of care for very sick patients.

Good article.

PS Eight years ago I made this presentation on estimates of costs of late life. The summary in this post (in catalan)

Club des Belugas - Never think twice

Regulating alcohol marketing

Policy Approaches for Regulating Alcohol Marketing in a Global Context: A Public Health Perspective

The article says:

The range of policy options for alcohol marketing restrictions includes four main categories: no restrictions, voluntary regulation or self-regulation, partial restrictions (e.g., on content, time and place, or particular audiences), and complete bans.

Unfortunately you'll not find a clear assessment of the impact of these policies. Only anecdotical facts. Therefore, no prescriptions can be made with sounding evidence. My impression is that somebody should care about the current advertising strategies that are very far from what WHO considered as comercials some years ago.

Conceptual framework on the growth of alcohol corporations, exposure to alcohol marketing, and alcohol-related public health problems.

Enabling Patients to Stick to their Medication

Investing in medication adherence improves health outcomes and health system efficiency

OECD provides some key figures on medication adherence:

Poor adherence is estimated to contribute to nearly 200 000 premature deaths in Europe per year. Patients with chronic diseases are particularly vulnerable to poor health outcomes if they do not adhere to their medications. Mortality rates for patients with diabetes and heart disease who don’t adhere are nearly twice as high as for those who do adhere.

It is estimated to cost EUR 125 billion in Europe and USD 105 billion in the United States per year in avoidable hospitalisations, emergency care, and outpatient visits.

The three most prevalent chronic conditions – diabetes, hypertension, and hyperlipidaemia – stand out as the diseases with the highest avoidable costs, for
which every extra USD spent on medications for patients who do adhere can generate between USD 3 to 13 in savings on avoidable emergency department visits and inpatient hospitalisations alone.

I'm dubious about the exact figures, anyway if you imagine that it is half ow what the say it would be a lot. Systematic reviews say that non-adherence is 15%. This is a hot topic and the ways to tackle are known.

Acknowledge: Medication non-adherence harms health and increases healthcare costs. The first step for the relevant stakeholders is to acknowledge that this problem exists and to adequately recognise its main drivers. Medication adherence needs to move up the policy agenda in order to raise awareness of the problem and mobilise adequate responses.

Inform: Few countries systematically monitor adherence. Routine adherence measures as well as adherence-related quality and performance indicators should be encouraged in order to improve health system effectiveness and efficiency.

Incentivise: Changes in financial incentives for providers and patients are essential. Shifting to payment systems that reward providers for the quality of patient outcomes would provide strong motivation to improve adherence. Medication adherence could also be considered as a measure for performance based contracts with pharmaceutical companies. Where patients’ co-payments for chronic medications exist, their reduction or removal should be considered to reduce financial barriers.

Steer and Support: The adherence process begins with a patient and a prescribing clinician and a dispensing pharmacist who should all be supported by other health system stakeholders. Payers/system designers can develop IT systems that facilitate optimal prescribing and patient-clinician communication or renewing prescriptions by patients. Educators have a role in equipping health professionals with skills in managing adherence such as person centred communication, shared decision-making, and socio-cultural competencies.

Clinical utility of genomic sequencing

The Path to Routine Genomic Screening in Health Care

Now that whole genome sequencing is knocking at the door of the clinician, it is the time to ask for clinical utility. The understanding of how such information will change diagnostic and therapy is paramount. There is still no need for cost-effectiveness, clinical utility comes first.
And the editorial at Annals explains exactly this issue, highly recommended:

There should be little doubt that individually tailored health care management plans based on DNA analysis are coming, but the timing of their introduction into routine clinical care is contingent on further demonstrations of clinical utility and proven implementation models.

My impression: let's wait for epigenetic biomarkers, beyond whole genome sequencing that provides less than 100 actionable genes out of 20.000. Though,

 The fact that only a small percentage of people would benefit from GS today is counterbalanced by growing evidence that the benefit could be significant, and perhaps even life saving

 

Pepe Castellanos at Barnadas Gallery

Who should get treatment?

Who should receive treatment? An empirical enquiry into the relationship between societal views and preferences concerning healthcare priority setting

The concern for an equitable and fair allocation of healthcare resources requires a prioritisation approach. Otherwise we are going to live in an arbitrary and opaque world.
An article from the Netherlands explains what people think about three perspectives:

The view “Equal right to healthcare” comprises an egalitarian view on health and healthcare. People with this view consider access to healthcare a basic human right. Everyone is equal, hence has an equal right to healthcare. According to people with this view, prioritisation should solely be based on the need for care and prioritisation based on patient, disease, and intervention characteristics, such as the effect of treatment, is opposed. What is considered to be “the right care” is a matter of personal concern for patients and, according to people with this view, patients should be supported in their treatment choices regardless of the costs.

The view “Limits to healthcare” comprises a view with a strong concern for providing “the right care” for patients. People with this view consider health-related quality of life to be an important outcome of treatment. According to people with this view, providing the right care may imply refraining from (life prolonging) treatment. People with this view do not consider cost-effectiveness to be an important criterion for priority setting, although they do consider it important to make good use of money. Hence, providing treatments that generate minimal benefits should be avoided. Priority setting based on patient characteristics is rejected, with an exception made for lifestyle. According to people with this view, patients who are culpable of their own disease should receive lower priority and prevention should receive higher priority in allocation decisions.

The view “Effective and efficient healthcare” comprises a utilitarian view on health and healthcare. People with this view consider it important to generate as much health for society as possible given the budget constraint, and consider a patient’s capacity to benefit from treatment important when setting priorities. Although people with this view focus on the cost-effectiveness of treatments, they do believe it is not possible to “put a [fixed] price on life”. The value of health benefits depends on circumstances and patient characteristics, such as age and culpability, and hence these should be taken into account in priority setting.

 And the result is:

 The majority of respondents was matched to the view “Equal right to healthcare” (64.5%), followed by “Limits to healthcare” (22.5%), and “Effective and efficient healthcare” (7.1%). A minority of respondents (5.9%) could not be matched

My impression is that we change such criteria according to the exact setting we are in a precise moment. That's why beyond societal criteria we do need professional criteria. Sounds too easy to solve the prioritisation exercise according to three principles.

PS. Still waiting for the book:Rationing and Resource Allocation in Healthcare: Essential Readings

 Juan Genovés exhibition at Marlborough gallery

Research and results

The Biomedical Bubble: Why UK research and innovation needs a greater diversity of priorities, politics, places and people

More resources for research are needed. This is the usual mantra. However, what about outcomes?. Since this is not so easy to measure it really lies in an uncertain land. A new report tries to put things clearer, at least for UK. It explains the mismatch about research funding and what is needed to improve health. This is exactly what I consider the right approach. It is useless to ask for more money unless we explain and focus on the priorities for achieveing better health.

A biomedical bubble has developed, which threatens to unbalance the UK’s research and
innovation system, by crowding out the space and funding for alternative priorities. This
is not a speculative bubble, as developed for tulips in the 1630s, or dotcoms in the early
2000s; there is far too much substance in the biomedical sciences for this. But it is a social, political and epistemic bubble (similar to the ‘Westminster bubble’, or the ‘filter bubble’), in which supporters of biomedical science create reinforcing networks, feedback loops and commitments beyond anything that can be rationalised through cost-benefit analysis.

The biomedical bubble represents a risky bet on the continued success of the pharmaceutical industry, despite mounting evidence that this sector faces a deepening
crisis of R&D productivity, and is cutting its own investment. And it favours a particular approach to the commercialisation of science, based on protectable intellectual property and venture capital based spinouts – despite the evidence that this model rarely works. Our health and social care system is under growing strain, and as the NHS marks its 70th birthday this month, there is renewed debate about its long-term affordability. Too often, the biomedical bubble distracts attention and draws resources away from alternative ways of improving health outcomes. Only 5 per cent of health research funding is spent on researching ways of preventing poor health. And more than half is spent in three cities - London, Oxford and Cambridge - despite variations in life expectancies of up to eight years across the country. This paper argues for a more balanced distribution, aligned to what the evidence clearly shows are crucial social, economic, environmental and behavioural determinants of better health outcomes.

 Food for thought.

Shomei Tomatsu in Barcelona

Fake lab tests

Bad blood

We all know that we live in a post-truth society. But this may have strong effect in your health. The case of Theranos, a US lab is explained in an excellent book by John Carreryrou is a precise example. You create an expectation that with a blood drop all tests can be done, you develop the narrative, the social and entrepreneurial support, and...while the regulator is on vacation (as usual) you perform fake test that can endanger your health. Imagine that your coefficient of variation is 34-48%, while it should be less than 10%!. The results may hurt you directly.
The book provides a lot of details:

As for the lab itself, it was a mess: the company had allowed unqualified personnel to handle patient samples, it had stored blood at the wrong temperatures, it had let reagents expire, and it had failed to inform patients of flawed test results, among many other lapses

And strong messages for start-up investors:

By positioning Theranos as a tech company in the heart of the Valley, Holmes channeled this fake-it-until-you-make-it culture, and she went to extreme lengths to hide the fakery. Many companies in Silicon Valley make their employees sign nondisclosure agreements, but at Theranos the obsession with secrecy reached a whole different level.

FT recommends this book for this summer, and I agree that it will help you to understand and avoid similar fake business that we can detect around us. It should never happen again.

Cost-effectiveness of new (genomic) benefits, it's just the begining

HERC database of health economics and genomics studies 
Cost-effectiveness of cell-free DNA in maternal blood testing for prenatal detection of trisomy 21, 18 and 13: a systematic review

Just yesterday our government suddenly decided to introduce a new benefit in public insurance coverage: contingent DNA based non-invasive prenatal screening. And the question is: does someone know if this new benefit is cost-effective?
You can get the answer after reading this review article, and the summary is:

 In total, 12 studies were included, four of them performed in Europe. Three studies evaluated NIPT as a contingent test, three studies evaluated a universal NIPT, and six studies evaluated both. The results are heterogeneous, especially for the contingent NIPT where the results range from NIPT being dominant to a dominated strategy. Universal NIPT was found to be more effective but also costlier than the usual screening, with very high incremental cost-effectiveness ratios. One advantage of screening with NIPT is lower invasive procedure-related foetal losses than with usual screening. In conclusion, the cost-effectiveness of contingent NIPT is uncertain according to several studies, while the universal NIPT is not cost-effective currently.

If this is so, since uncertainty is the word that better reflects its current cost-effectiveness, why do the have introduced? Because they don't care about it. These are not the best days for a health economist (and for the society as a whole). Maybe it's just the begining of a new world without scarcity, and I can't figure out.

Manuel Anoro