Digital health: panacea or chimera?


An interesting article on Lancet Digital Health, by Kazem Rahimi, reflects on the elusive search for the savings that digital health should allow.

Digital health advocates argue that the digital future will be one of more precise interventions, improved health outcomes, increased efficiency, and ultimately reduced health-care expenditure. But how realistic is the promise of reduced costs, while also seeing improved health, or at least no diminishing of it?

The author admits that the field is at too early a stage to reach a definitive conclusion on this issue, not least because of the need for more empirical evidence. However, in the interim, given the importance of this issue to current national and international policies, the argument of digital cost reduction seems worthy of scrutiny.

The author states that “despite the substantial contributions of technological progress to improvement in health outcomes, examples of cost-cutting effects are a rarity. On the contrary, technological progress is widely seen as the most important driver of the rise in health-care spending. For instance, magnetic resonance imaging will inevitably be more expensive than its alternative, which is usually either no test at all or a cheaper, but less accurate, diagnostic technique.”

However, the author highlights the differences in digital health. Digital technologies often include innovative software solutions and algorithms that could be substantially cheaper than devices or drugs. In addition, these technologies tend to focus on solutions to the notoriously inefficient delivery systems of health care globally, as opposed to the development of new treatments. Given that the alternative to digital technologies would potentially be a more labour-intensive model of care, one might expect their adoption to replace costly health-care professional time or hospital services.

Rahimi does not question the fact that the use of well designed and tested technological solutions will eventually lead to improved matching of resources with the complexity of tasks and, thus, achieve increased productivity or (technical) efficiency. The author makes for example the case of a machine learning algorithm that is able to make diagnoses faster or better than most doctors could be expected to lead to substantial reductions in the price of that particular service. Provided that there is sufficient empirical evidence, one could then directly compare the prevailing approach (doctor diagnosis) with the new digital approach (algorithm plus or minus doctor diagnosis) and conclude that the new intervention will get the same job done at a much lower cost.

But why is such a cost-saving intervention still likely to increase health-care expenditure? This apparent paradox can be explained by the common confusion between microeconomic effects of individual health-care interventions or programmes, and effects on the whole health-care market. Although a microeconomic study might conclude that substituting old with new might lead to net savings, the typical models in such studies assume that health-care utilisation of the service or treatment under investigation remains unchanged and that the two approaches differ only in their costs and health consequences. Thus, an intervention that is of lower price, even without causing a change in health outcomes compared with the alternative, would be expected to result in cost savings.

However, as the author points out, health-care markets tend to be in disequilibrium when demand continues to exceed supply and use. In such a setting, reducing the price of a particular service will invariably lead to an increase in the quantity demanded. Given that total expenditure equals the quantity demanded multiplied by its price, introduction of low-price technologies might lead to an overall rise in expenditure. In other words, medical uses of the new treatment are increased through addressing an unmet demand, and this expansion in use leads to a net rise in expenditure.

The article makes further reflections that I invite you to read here. They are an interesting point of view on a subject that is too often addressed in a dogmatic and simplistic way.

Artificial intelligence in healthcare: where and how to invest

AI Investments

There are four main areas in which using artificial intelligence (AI) technologies can improve the effectiveness and value of ICT solutions in healthcare.

The applications that manage clinical, administrative and logistical processes, focused on data management, have so far been marginally affected by the new AI-based technologies. The functional perimeter of these applications has been constant for many years and their evolution is mainly linked to technological aspects (software architecture, infrastructure, user interface).

AI-based technologies can significantly improve the ability to analyze data related to health processes (Process Data Analytics). Compared to the business intelligence carried out with traditional technologies, AI allows to better estimate and understand current trends, perform predictive analysis, and find correlations between different phenomena.

This increase in analysis capacity can be very useful in resource planning, for example in contact centre, booking, bed and operating room management. More information and awareness on the dynamics of demand for resources allows to articulate and segment the offer of services in order to maximize the results. By using some IA-based technologies it is possible, for example, to create tools to optimize the resources planning and scheduling, as to suggest to system administrators the setting of agendas for a booking system or the management of queues in a contact centre. A big step forward compared to current software where the configuration burden is entirely on the user.

Another area where AI can allow a great leap forward is the support to the decision-making process (Decision Making). Neural networks and deep learning mechanisms can greatly increase the effectiveness of systems for differential diagnosis and decision support (CDSS), overcoming the current limitations of the Evidence Based Medicine approach. The ability to find correlations and to learn from clinical practice, typical of AI-based technologies, will be a determining factor for the expansion of this sector.

Even in the measurement of outcomes (Outcome measurement), AI can allow an important evolution, thanks to the mining, analysis and correlation capabilities that these technologies possess. The measurement of outcomes will be increasingly important with the spread of policies and models based on Value Based Healthcare.

Finally, there is a final consideration to be made: in order to achieve the above-mentioned developments, it is necessary that also at process data management level there is an evolution consistent with the overall picture. In other words, in designing new applications it is necessary not only to focus on the mere management of data and basic processes, as it happens today, but also to consider the needs and implications that Resource Planning, Decision Making, Process Data Analytics and Outcome Measurement imply, with a holistic approach to all areas of digital health care.