copy the linklink copied!1. Bringing health into the 21^st century

1.2.1. Digital transformation entails much more than digitising existing processes

A key part a digital transformation relates to how new attitudes and thinking are taken on and adopted. This has been evident in some sectors where disruptive innovators have improved certain aspects of production, which has stimulated adaptation among bigger players. For instance, not too long ago electronic banking was deemed revolutionary. Now the word ‘electronic’ is redundant.

Typically, small-scale disruption is evaluated, adopted and scaled to improve performance across a sector. A related feature, decoupling, is another hallmark of a digital transformation. Smaller operators are taking on specific parts of the value chain and improving these from the consumer perspective. There are some signs of decoupling in health. For example, pre-diabetic care, mental health or radiology services are beginning to be outsourced by established actors such as hospital chains, insurers and health systems to smaller organisations dedicated to more narrow function (Shah, Farkas and Kocher, 2019[14]). GP at Hand, a digital service in the United Kingdom providing health advice, triage and remote GP consultations, was launched in 2017 in partnership with a west London primary care practice. The number of patients enrolled with this service rose from 4 700 to over 50 000 within two years. However, the disruption caused by the digital service is challenging the prevailing order and its institutions (Burki, 2019[15]).

Decoupling may seem ill-suited to health. After all, it conflicts with the rhetoric of service integration (which is often implied to suggest vertical integration). However, disruption need not mean fragmentation. It may actually have the opposite effect if innovation is underpinned by a coherent information infrastructure and policy framework. The urgent need for these is a key finding of this report.

The health sector has unique characteristics. For example, some health data are very privacy-sensitive. They can also be complex. But this does not fully explain the lack of progress. Other types of data, financial or social security, are also considered sensitive. Data mining techniques to analyse complex, unstructured data are available. Complexity is also not inherent. A hallmark of digital transformation is efficient data exchange. This is made possible with adoption of common standards and protocols. The global internet protocol (IP), which enables the seamless exchange of data across the internet, is perhaps the pre-eminent example of this. Yet harmonisation is still largely lacking in health, making health data more complex than it needs to be.

The lack of a common approach is partly responsible for a health data ‘gold rush’. Commercial organisations are profiting from acquiring, harmonising and selling large volumes of personal health data for a variety of commercial purposes (IQVIA, 2019[16]; Forbes, 2018[17]; Fortune, 2016[18]; Computer World, 2019[19]; Healthcare Weekly, 2019[20]). While the commodification of these data is not inherently negative, it does raise a number of questions regarding privacy, consent and efficiency that require a policy response.4 Regulators and policy makers need to prevent the misuse of personal health data, as well as ensuring that the benefits of their use are distributed equitably. After all, most gold rushes did not end well for the majority of those involved.

Despite the uniqueness of health as a sector, the key lesson from other sectors resonates: that digital transformations are not achieved by simply digitising existing practices. In fact, 45% of companies report that ‘digitisation’ fails to deliver the expected returns (McKinsey, 2019[21]). Success depends on deeper institutional change. And while the transformation has certainly entailed disturbance to the established order, has made some actors in the global digital economy wealthy, and has had certain negative social consequences, there is little doubt that it has also generated sizable consumer surpluses and welfare (Brynjolfsson, Eggers and Gannamaneni, 2018[22]). In net terms, the principal beneficiaries have been the public.

1.2.2. Lessons from other sectors are relevant in health

Why has health – a sector where information and knowledge are critical to performance and success – been so slow deploy available technology for better information and knowledge?

Inputs are not the reason. The data exist, ready to be exploited. The technologies to extract, manage, transmit and analyse them are also available and are continuously improving. Investment in AI is accelerating worldwide as is the number of registered AI patents (OECD, 2019[1]). The number and proportion of scientific papers using data mining and big data5 analytics is rising (Galetsi, Katsaliaki and Kumar, 2019[23]). The expertise to put data to work exists – albeit residing mostly in other sectors.

The reasons relate to the very things other sectors have been able to address: the organisational and institutional capacity to harness these inputs. The key enablers – flexibility, re-structuring and working across silos – are anathema to the fundamental way in which health systems are organised. Their institutions, forged in the pre-digital era when it was not possible to combine, share and analyse large amounts of data, are known for being static and resistant to change. In many ways, the arrival of digital technology has simply highlighted these long-standing barriers to achieving health policy objectives.

A digital transformation must begin with an explicit recognition that data are a valuable resource, but have no intrinsic value unless put to work within an enabling institutional environment. This recognition, which in many cases finds expression in strategic orientation (an important element of policy making), must be pan-sectoral in order to begin bridging the silos that impede digital transformation. Digital strategies abound in health, but whole-of-government approaches on which enabling policy frameworks can be built are rare.

Flexibility is another defining feature of any transformation. This means allowing existing actors and new entrants to disrupt thinking and practice. It also means allowing smaller parts of larger organisations to trial new approaches, which are then evaluated and, if viable, scaled up for broader adoption and until the new practice becomes the new normal across entire sectors – as was the case with e-banking.

Health, of course, is not an area where the fail early, fail often mantra holds the same appeal as in Silicon Valley. Nevertheless, a number of examples exist of how innovation is not just possible but can advance health policy objectives. The challenge is to permit some flexibility and controlled disruption, while managing risks. This is eminently possible even in complex, risk laden endeavours. Aviation, for example, has already been mentioned. The British Army has overhauled its operational model, enabling its forces to become more modular and agile (McKinsey, 2019[24]). But in order to work, innovation must be underpinned by common protocols, standards and an integrated information infrastructure. This hinges on appropriate governance designed specifically to maximise the opportunities and minimise the new, unique risks of the digital era.

Capacity to use digital technology, manage data and extract knowledge from them must exist. This requires investment in not just hardware but the expertise to make the most from it. It must target the supply side and the demand side. Providers and patients need to be engaged. Surprisingly, OECD countries typically invest only under 5% of health budgets on managing information. In other sectors investment is four times higher (OECD/WHO/World Bank Group, 2018[25]). Closer inspection reveals that the health spends similar amounts to comparable sectors on tangible products such as ICT hardware, computers and network infrastructure (Calvino et al., 2018[26]). However, spending on intangible products such as software and databases, and the purchases of ICT services is comparatively modest (Figure 1.1).

copy the linklink copied!

Figure 1.1. Investment in software, databases and ICT services by the health sector

Investment in software and databases as a % of non-residential GFCF; purchases of intermediate ICT services as a % of output

Note: Gross fixed capital formation (GFCF) is a measure of spending on fixed assets. Countries covered: Australia, Austria, Denmark, Finland, France, Italy, Japan, the Netherlands, Norway, Sweden, the United Kingdom, and the United States.

Source: Calvino et al. (2018[26]),:”A taxonomy of digital intensive sectors”, https://dx.doi.org/10.1787/f404736a-en.

1.3.1. The right information to the right people at the right time

The failure to extract and use information contained in health data, which exist already, is a significant missed opportunity to improve services and care. For example, 10% of patients are unnecessarily harmed during care. The health burden of this in OECD countries is on par with diseases such as multiple sclerosis and some cancers. The direct financial impact is as high as 15% of hospital expenditure, and the broader economic drag estimated to be in the trillions of dollars (Slawomirski, Auraaen and Klazinga, 2017[31]). The most common root cause is a failure of communication – information and knowledge not reaching the right person at the right time. Shared electronic information platforms have been shown to improve safety by addressing the communication problem (Banger and Graber, 2015[32]).

Better information exchange makes care not only safer but also more effective and efficient. Care can be better coordinated by different providers and integrated with other services, with better results and less duplication and waste (OECD, 2017[29]). This is especially important for the growing number of people who have multiple chronic conditions, currently estimated to represent at least 20% to 30% of the adult population in OECD countries. For these individuals, accessing care can be frustrating, inefficient and costly in a system fragmented across sectors and disease groupings.

Enabling access to the electronic health or medical record (EHR or EMR)8 by all actors involved in a patient’s care is a key structural component of a high-quality health system (Auraaen, Slawomirski and Klazinga, 2018[33]). While most OECD countries are well on the way to digitalising their health services through the introduction of electronic records in physician practices, hospitals and other settings (Figure 1.2), only 64% of countries report that these form part of an integrated network where secure but unencumbered exchange of information is possible (Oderkirk, 2017[34]).

This can mean either a ‘one patient one record’ approach or one made up of disparate platforms that are set up to exchange data and information. A patchwork of electronic records that cannot efficiently link or exchange information with one another is not effective in improving safety or other aspects of care quality such as co-ordination, efficiency and a positive patient experience. From a systems viewpoint it is not much better that the pre-digital equivalent of paper records stored in individual health care organisations.

copy the linklink copied!

Figure 1.2. Countries are digitalising their health records

Percentage of primary care physician offices and acute care hospitals using electronic medical records, 2016

Note: United Kingdom: England, Scotland and Northern Ireland (excludes Wales).

Source: OECD Survey of Electronic Health Record System Development and Use, 2016; Oderkirk (2017[34]) “Readiness of electronic health record systems to contribute to national health information and research”, https://dx.doi.org/10.1787/9e296bf3-en.

Enabling people to access their health records and interact with their own medical information is a driver of high quality people-centred care. Digital technology provides the ideal platform to enable this access easily and efficiently. It is encouraging that approximately 70% of responding countries to the 2016 OECD survey reported that people can access their record. However, only 43% reported that individuals could interact with their own record (e.g. enter information, send requests, communicate with providers). A study by Barbabella et al. (2017[35]) found that the majority of hospitals in Europe (90%) do not permit patients to access their own health data. Elsewhere, access restrictions can sometimes result from legislation. For example, United States federal and state laws allow patients to request amendment to their health record but not to directly interact with their record.

People’s rights to access their health records may also be underused. In the United Kingdom, fewer than 8% of patients who were able to access their medical records actually did so (NHS Digital, 2019[36]). In programmes to improve care for multi-morbid patients, tools for sharing of information mainly focus on interactions between professionals and provider organisations, not on making information available to patients (Melchiorre et al., 2018[37]). Access is likely even lower among populations with complex needs. In the Netherlands, for example, just 4% of the chronically ill population reported using a personal health record.

Examples of progress can be found. Estonia has a unified EHR, which enables residents to view all of their medical data in one place – including diagnoses, test results, medications. Residents can also interact with their data. For example, they can update their details, supplement existing information, and carry out administrative processes such as obtaining a medical certificate for a driver’s license without needing a specific appointment. Lithuania (which did not respond to the survey on which Figure 1.2 is based) has implemented a centralised ‘one resident – one record’ EHR system that covers 95% of the population. It carries all relevant medical information in integrated electronic workflows covering appointments, referrals and e-prescribing. It also enables provider interaction and patients have secure access to their record through a patient portal.9

1.3.2. Data can help identify health need and target care more accurately

Analysis of existing health data can enable health service planners and policy makers to identify and target individuals who have complex health needs. These people may not always be in contact with the most appropriate health care providers, making data a valuable resource for creating information on where they reside, their level of need, the type of services they will require and the level of resourcing required.

In Spain, a locally developed risk-stratification tool is used in 14 of the country’s 17 autonomous regions. EMR data from primary care providers and hospitals across the regions are linked to create a common data pool. The model uses this pool to identify complex patients based on a classification termed morbidity adjusted groupings (Grupos de Morbilidad Ajustados – GMAs). The system supports targeting of care and helps allocate resources according to health need. One limitation is that the GMAs do not capture variables that are not coded in clinical data such as social and economic circumstances, which could be included with linkage of additional data sets.

Many of the Spanish regions have built on this tool to develop integrated care models for complex patients identified. These often use shared electronic care plans to plan and monitor the care patients receive from various providers, relying on a shared EHR to exchange information with a single provider taking on a coordination role. Such models harness shared information systems while alleviating overburdened providers, for instance by reducing unnecessary hospitalisations and the administrative workload of primary care doctors.10

The United States Office of the National Coordinator for Health IT (ONC) released Draft 2 of the Trusted Exchange Framework and Common Agreement (TEFCA) in April 2019. This outlines a common set of principles, terms, and conditions to support nationwide exchange of electronic health data across disparate networks. The TEFCA is designed to scale health data exchange nationwide and help ensure that Health Insurance Networks, health care providers, health plans, individuals, and other stakeholders have secure access to their electronic health information when and where it is needed to promote care co-ordination and orientate care around patients’ needs.

While some promising examples such as these can be found, there remains a lot more that can be done in this important area of using existing data for the important purpose of identifying and addressing health need.

1.3.3. Most OECD countries are slow to harness data for safer and better treatments

The potential of using data to generate information and knowledge extends beyond improving patient care. Another important area is the assessment and regulation of medical technologies.

Real-world data (RWD) refers to data created during routine health system activity, as opposed to data generated during clinical trials. RWD can be used to generate evidence on how drugs and other medical products perform in routine clinical use. This can complement evidence derived from clinical trials to better inform decisions and actions of clinicians, patients and policy makers.

It also represents value for money. Clinical trials are complex, lengthy and expensive, yet typically include a relatively small number of carefully selected subjects, who may not represent the patients that the product will eventually be used on. Modern analytical techniques drawing on large sets of RWD can include millions of subjects. The statistical power is comparably large, and advanced methods to limit confounding variables continue to evolve. Real-world trials can now reproduce trial results much faster and at a fraction of the cost (Fralick et al., 2018[38])].

This can deliver faster, safer and better treatments – particularly for emerging health challenges such as dementia and chronic conditions, where pooling relevant data from a range of sources is the only way of devising timely detection and effective therapies.

It can also inform better policy. For example, as evidence of a product’s performance accumulates, payers can determine if prices reflect therapeutic benefit – and adjust them accordingly. This can not only improve efficiency and value derived from medical technologies, but also sets incentive signals for future product development.

However, most health systems are not harnessing data to their full potential in this way. The majority of twenty-six OECD and European Union countries surveyed in 2018 on the use of routinely collected data reported that the principal purpose was to monitor medicine consumption and expenditure (22 countries). Eighteen used these data to monitor provider compliance and 15 to monitor quality of prescribing. Meanwhile, 14 countries reported using routine data to evaluate the safety of medicines (Figure 1.3).

copy the linklink copied!

Figure 1.3. Routine data are mostly used for monitoring medicine use, expenditure and compliance

Note: In most cases, the routine data described only cover medicines dispensed in the community setting and not medicines dispensed in hospitals.

Source: OECD (2019[39]), “Using routine data to inform pharmaceutical policies”, https://www.oecd.org/health/health-systems/Using-Routinely-Collected-Data-to-Inform-Pharmaceutical-Policies-Analytical-Report-2019.pdf.

The positive finding is that at least routine data are beginning to be used. However, in most cases the data excluded drugs dispensed in hospitals, which represents a considerable proportion of medicine consumption. This suggests problems with data linkage and information infrastructure. Moreover, only nine countries reported that evidence generated from routine data was considered by health technology assessment agencies (HTA) in their decision-making (OECD, 2019[39]).

Overall, the majority (19) of responding countries answered ‘no’ to the question of whether routine data were used to their full potential. The main barriers cited were prohibitive privacy legislation and poor data quality (governance issues) and lack of capacity in terms of infrastructure as well as analytical expertise (OECD, 2019[39]).

Perhaps the best example of systematic use of routine data at a national level is the United States Food and Drug Administration’s (FDA) ‘Sentinel’ Initiative. The Sentinel programme monitors the safety of medical products in routine use by accessing personal health data – scattered across a large number of health care organisations, payers, dispensers and other agencies – of over 200 million people. To ensure security and privacy, data never leave their location. Instead, single electronic queries are sent across the entire network. This feature is enabled by common data standards and protocols. Identifying safety problems the old way used to take years. Sentinel does this in weeks or months, reducing harm while saving money and time. Since its establishment in 2014, the initiative has enabled important regulatory decisions. It is also eliminating the need for many expensive post-marketing studies, which can cost millions of dollars to run.11

1.3.4. Big data are not used to their full potential in public health

Public health is arguably the area where data – especially ‘big data’ flowing from non-traditional sources outside of the health system such as social media, web searches and environmental data – can be put to work to advance health and prevent disease. These data can complement traditional information sources to answer many questions and steer public health policy in three key ways:

1. 1. Allowing a more precise identification of at-risk population groups (‘precision public health’);

2. 2. Enabling better surveillance of both communicable and non-communicable diseases;

3. 3. Facilitating better targeted strategies and interventions to improve health promotion and disease prevention.12

These new methods can be used to tackle a range of public health priorities, such as monitoring and responding to food-borne outbreaks, identifying behavioural risk patterns across populations, or monitoring signs of mental ill-health such as suicidal ideation.

Much promising activity is already happening at the municipal level. “Smart Cities” demonstrate how integrating transportation, mobility, food safety and environmental data can contribute to more effective and efficient health promotion and disease prevention. For example, the city of Chicago used publicly available data from a variety of sources (ZIP codes, business licenses, building code violations, and a phone hotline for complaints) to predict restaurants most likely to be in violation of health codes. More recently, researchers used anonymised Google search terms and geolocation data of individual users to identify food-borne disease outbreaks across the city. Problems were identified considerably faster and more accurately than with traditional surveillance methods.

At the national level, Korea is in the process of integrating the national health insurance database (NHID) with clinical records, health care activities as well as data from outside of the health system including climate, pollution and geolocation data. To date, analytical uses of the NHID have included, for example, identifying causality and predicting risk by linking health-screening data with medical history and socioeconomic status, and a surveillance system to target chronic diseases, based on information of individuals’ use of health services. The amalgamation will further equip decision makers, public health experts and citizens with more detailed information to pursue a range of efforts to improve public health and prevent disease.

Some countries have also started exploring the possibility of using genetic data to inform public health prevention. In Estonia, two large-scale clinical trials are working to better target preventive health care services at patients at higher risk of cardiovascular disease or breast cancer, based on genomic algorithms using data from the Estonian Genomic Center. In the United States, a USD 215 million Precision Medicine Initiative was launched in 2016. This will include, among other projects, the All of Us research programme, a 1-million participant study whose mission is “to accelerate health research and medical breakthroughs, enabling individualized prevention, treatment, and care” by studying “individual differences in lifestyle, environment, and biology”.

Big data present a number of opportunities but also have some inherent limitations as well as challenges relating to inter alia ensuring their quality and managing their security and privacy implications. For example, the temptation to over-rely on big data without robust methods for interpreting it can lead to apophenia: “seeing patterns where none actually exist, simply because enormous quantities of data can offer connections that radiate in all directions” (Boyd and Crawford, 2012[40]). Approaches using big data should therefore complement – not replace – traditional public health surveillance methods. Nevertheless, big data have the potential to facilitate the more effective translation of knowledge into effective public health policy, by enabling a better understanding of the interaction between behaviour, genetics, and the physical and social environment.

Transparency represents a key element that can facilitate the success of public health initiatives based on big data. Sharing data and algorithms with other stakeholders (e.g. collaborations between academia, public health departments, industry, and citizens) enables a more effective use of data and facilitates the early detection of any problems, as well as allows other public health authorities to implement similar successful interventions. But transparency needs to be underpinned by good data governance, which ensures that security and privacy are maintained and maximises the utility and use of available data to generate valuable knowledge.

1.3.5. Harnessing data to improve efficiency and value across the health system is rare

Using available data can drive considerable improvements in how complex health systems are managed. The knowledge extracted from data can help decision makers identify and act on problems such as waste and inefficiency, and to make the decisions, and judgements needed for a health system to function. This includes setting priorities, and allocating resources across populations and competing programs.

In some health systems the proportion of unnecessary treatments, procedures and therapies exceeds 40% (O’Neill and Scheinker, 2018[41]). Identifying wasteful spending through better use of available data is a prime opportunity for improving efficiency and outcomes (OECD, 2017[29]).

For example, the Washington State Health Care Authority (WA HCA) has linked claims, billing and EHR data of 2.4 million insured individuals to examine inappropriate and wasteful practices. The algorithm classifies services as necessary, likely to be wasteful, or wasteful. The benefit of linkage across the three data sources is that it allows more nuanced analysis, incorporating the individual context of each case, as opposed to simply tallying the number of procedures on a ‘low value’ list.

Of the 1.52 million services analysed 44% were classified as wasteful (i.e. not adding any health benefit for the recipient). Spending on these services amounted to USD 282 million, or 36% of WA HCA expenditure (WHA, 2018[42]). A similar study in Saint Louis, using data of 1.6 million individuals, identified 46% of services as unnecessary (National Alliance Of Healthcare Purchaser Coalitions, 2018[43]).

Clearly some people receive unnecessary treatments while others, in greater need, miss out. In addition, unnecessary duplication and wasteful administrative processes abound in all health systems. The challenge is using this information and knowledge to develop effective policy responses to reduce these inefficiencies and redeploy resources where they are most needed.

Data linkage, by shedding light on what is really going in a complex health system, is a critical step to assess performance, identify problems such as unwarranted variation, and enable smarter resource allocation. However, a survey of OECD countries conducted in 2013 and repeated in 2019 found considerable variation in both the availability and readiness of key data sets, and their regular linkage for secondary purposes such as system management (Figure 1.4). While the 2019 information is preliminary and the participation of a greater number of countries in the 2019 data collection is necessary for understanding change over time, the results suggest that key health datasets may be less likely to be regularly linked than they were in 2013.

copy the linklink copied!

Figure 1.4. Availability of data is growing but their linkage appears to be stagnant

Percentage of key data sets available and regularly linked, 2013 and 2019

Note: These are preliminary data still missing several countries; only countries that responded to both the 2013 and 2019 survey are shown; *Ireland 2013 data used for 2019 (relevant survey section not completed in 2019).

Sources: OECD (2019[44]) “Survey on health data governance: preliminary results”; OECD (2015[45]) “Health Data Governance: Privacy, Monitoring and Research”, https://dx.doi.org/10.1787/9789264244566-en.

One possible reason for the reported reduction in linkage is that the European General Data Protection Regulation (GDPR) came into force in 2018. The GDPR affords personal health the highest level of protection, and has brought security into sharper focus within the European Union and beyond, with many countries still adapting their systems to respond to the new regulation. Protecting data and putting them to work are not mutually exclusive – both can be achieved with strong data governance frameworks.

Nevertheless, linking disparate datasets to create new knowledge for policy and other purposes is eminently possible. In New Zealand, personal health and other public sector data are linked to create ‘virtual registries’ for diseases such as diabetes (Jo and Drury, 2015[46]). These registries track the health care activity and outcomes of the relevant patient population enabling better assessment, decision-making and resource allocation. Traditionally, such registries are constructed in parallel to existing data sources. This can cost millions of dollars to establish and maintain. Harvesting the data from existing data sets is much cheaper and more efficient. Similar linkage in New Zealand also enabled detailed cost-of-illness study of multi-morbidity, generating hitherto unknown knowledge on the expenditure profiles of patients with various combinations of chronic diseases (Blakely et al., 2019[47]).

Detecting fraud, a significant burden on many health systems (OECD, 2017[29]), can be enhanced with modern data analytics. The United States Department of Health and Human Services-Office of Inspector General (HHS-OIG) mines available data and applies predictive analytics and modelling to enhance oversight of the Medicare and Medicaid programs for fraud. Multi-disciplinary teams use near real time data to examine claims for known fraud patterns, identify suspected fraud trends, and to calculate ratios of allowed services as compared with national averages, as well as other assessments.13

1.3.6. Information systems can enable better health funding models

A major contributor to the problems faced by health systems is the way services are paid for. The rise of chronic diseases and longer life spans means that service delivery should be approached in a more longitudinal and coordinated way that crosses boundaries between sectors (including social care) and settings. Remunerating individual service parcels is becoming less and less conducive to ensuring effectiveness and efficiency in the majority of health care activities.

Payment systems that encourage integration and that cover entire care pathways, better outcomes and efficiency have been discussed for some time. Alternative models to meet the challenges of rising patient complexity and achieve policy objectives include: additional payments made before during or after service delivery for specified outputs or outcomes; bundling – a combined, single payment for entire care cycles across settings and including primary and tertiary care, imaging and pathology, rehabilitation and follow-up care; and population-based payment, in which groups of health providers receive payments on the basis of the population covered, in order to provide most appropriate health care services for that population (OECD, 2016[48]).

While these non-traditional payment models vary in their design, incentives and structure, they have one thing in common: their success relies heavily on an information system with the capacity to integrate data on inputs, outputs, processes and outcomes. This was difficult in the analogue era – part of the reason that fragmented payments such as fee-for-service became the dominant model. However, electronic linkage of clinical, administrative, financial and other data makes these new approaches possible. Payments can be bundled across a set of providers and activities, with data systems ensuring that each component is remunerated appropriately. The clinical and budgetary consequences of a medical error at any point in the pathway become the responsibility of the entire team of providers, as opposed to the ones downstream to where the problem occurred. This encourages better coordination and communication (the lack of which is the most common cause of adverse events). Good information systems can ensure that payment for achieving agreed performance metrics is based on reliable data from several sources that can be more accurately adjusted for patient complexity and other confounders. Technologies such as blockchain, which can enable deployment of ‘smart contracts’, can further enhance the reliability of these payment models.

Similarly, population-based remuneration can also be adjusted to reflect health need, making it possible to transform care. The possibilities for these innovative approaches to payment expand when health data can be linked with social care data. Enabling payment models that encompass a broader range of health determinants could yield better health outcomes than the current fragmented approach. Integrating data in this way increases the accountability of each provider who contributes to a patient’s care pathway.

While the possibilities are many, examples are few, especially at the national level. Partners HealthCare in the United States has implemented a bundled payment framework that encompasses acute and post-acute care providers. The framework comprises four parts: risk stratification of patients, care team convening and planning, monitoring and communication, and patient engagement. The cornerstone of this model is an integrated EMR that enables communication between providers and patients, and that supplies the data for risk stratification and eventually payment. But implementation was heavily reliant on organisational leadership and policy (Safavi, Bates and Chaguturu, 2019[49]).

Some countries are laying the foundations for new payment and resourcing mechanisms. Finland, for example, is establishing a one-stop shop for all secondary use of health and social care data, enabling a wider set of information to be integrated for use by the health system. The integration of data from all providers of health and social services, as well as socio-economic data, is intended to inform the needs-based allocation of resources across the country’s regions.14

However, data infrastructure and digital technology merely provides the possibility for changing the way providers are paid. Institutional and policy changes are needed to adapt remuneration towards a model more suited to modern demands (See Section 1.4.5).

1.3.7. Preparing the health workforce to make the most from a digital transformation is a challenge

The health workforce is pivotal in a successful digital transformation. OECD analysis suggest that automation will create significant workforce disruption in modern economies. But the perception that health workers will be replaced by machines is not supported by the evidence, with their roles and tasks among the least likely to be affected (Figure 1.5). Nevertheless, majority of health workers are likely to see changes in their task composition towards tasks that are difficult to automate, such as those related to creative and social intelligence, teamwork and other ‘soft’ skills.

As discussed, digital technology offers opportunities to improve care delivery. Data-driven tools can support health workers in performing tasks that are repetitive, time-consuming, and heavy on data processing, such as selecting irregular results from large volumes of preventive or routine chronic care tests, synthesising information relevant for a given patient’s condition from numerous sources (patient records, archives, guidelines, specialist recommendations), or analysing patterns in patient outcomes for regular improvements in practice.

Machines are also beginning to match human performance and accuracy in, for example, analysing radiological images and retinal scans. Workers aided by these tools could become more productive. This includes having greater opportunities for face time with patients or colleagues, and to employ human skills to address their professional obligations more effectively and with requisite empathy.

copy the linklink copied!

Figure 1.5. Health sector jobs are at a comparatively low risk of automation

10 sectors with the highest, and 10 with the lowest risk of automation for average job within the sector

Notes: High mean probability of job automation means that the mean job in a given sector is highly automatable based on tasks it involves. Low mean probability of job automation indicates that the mean job in a given sector might change with regards to how some of its tasks are carried out. Not all tasks related to caring for and assisting patients that cannot be automated could be included in the calculation; hence, estimates for the health sector are biased upwards.

Source: Nedelkoska and Quintini (2018[50]), “Automation, skills use and training”, https://doi.org/10.1787/2e2f4eea-en.

Leveraging data to improve clinical as well as other activities of health care systems also means that new occupations – and roles within existing occupations – are likely to emerge. For example, oncologists, radiologists and anaesthesiologists could work with coders and data scientists to design, evaluate and refine the algorithms that enhance their practice.

However, things are not as simple as feeding reams of data into a machine – as the aforementioned problems with of autonomous vehicles suggest (Section 1.2.1). In the health space, difficulties with initiatives such as IBM ‘Watson’ suggest that the automation of complex and nuanced clinical decision-making is more complex than initially thought. Watson was trained to give treatment advice for cancer patients, but frequently made unsafe recommendations. A key problem was that Watson was trained on hypothetical, not real-world data, highlighting the importance of strong data governance to enable putting real-world data to work for productive purposes.

Nevertheless, the fact remains that health workforces are ill-prepared for a digital transformation, which makes change harder and slower. For example:

• Requirements to undertake continuous professional development have not kept up with technological and labour markets changes affecting the economy as a whole. Up to 70% of health professionals report not being accustomed to using digital solutions due to gaps in knowledge and skills in data analytics (Hegney et al., 2007[51]; Foster and Bryce, 2009[52]; Skills for Health, 2012[53]; European Commission, 2013[54]; European Health Parliament, 2016[55]; Quaglio et al., 2016[56]; Melchiorre et al., 2018[37]). A health professional who does not understand how an algorithm behind a clinical decision support tool was developed, or the data used for its conclusion, will see it as a “black box” and may be reluctant to use it (Galetsi, Katsaliaki and Kumar, 2019[23]).

• A fragmented, disease-centred approach to health care delivery means that many digital solutions are not fully deployed because of limited teamwork among health workers belonging to different professional categories, specialties, or provider organisations (Rudin et al., 2016[57]).

• Tasks in health care are defined based on the type of employment. For example, only doctors can perform certain tasks. This limits the opportunity to leverage digitalisation to improve efficiency and address shortages by facilitating shifting of tasks from doctors to nurses and community health workers.

• In-house expertise to put data to work is also missing. Making the most of digital opportunities and ensuring that the changes result in favourable outcomes requires ICT specialists and data scientists. Sector-specific specialisation for these occupations is becoming increasingly common and valued. Yet, recent OECD analysis suggests that the health sector has relatively few specialists (less than 1% of employees) in these fields compared to education, public administration and finance (Figure 1.6).

Some health systems are addressing these issues, particularly in terms of improving digital literacy among their health workers. Australia, Canada, Norway, Switzerland, New Zealand, and the United Kingdom completed a review or established a regular process to assess how technological and other developments (including ICT, AI, genomics, or demographics) are likely to change the skill requirements and the roles and functions of health workers over the next two decades. These reviews include the consequences for the education of future, and the training of current health workers.

Based on such a review, the Norwegian government, for example, is currently restructuring national curriculum regulations in health to make these more future-oriented. In Switzerland, eHealth Swiss has published guidelines for educators on how to integrate digital-health topics into the education and professional training of health workers. Canada Health Infoway – an independent, not-for-profit organisation, fully funded by the federal government – supports the Digital Health FACTS programme, which engages inter-professional faculty and students of medicine, nursing, and pharmacy to promote and scale up development of digital skills that match the demands posed by the emerging digital technologies.

copy the linklink copied!

Figure 1.6. The health sector has a relatively low proportion of ICT specialists

ICT specialists as a % of total employment, by sector (2013-15)

Note: Calculated as the number of workers employed in ICT specialist occupations, over total employment in the sector. Sector-specific values are weighted averages over the considered countries, where weights are equivalent to the country’s total employment in the sector. The values are then averages 2013-15. Countries covered are: Australia, Austria, Denmark, Finland, France, Italy, Japan, the Netherlands, Norway, Sweden, the United Kingdom, and the United States.

Source: Calvino et al. (2018[26]), “A taxonomy of digital intensive sectors”, https://dx.doi.org/10.1787/f404736a-en.

Many countries have developed new programmes and accreditation standards in Clinical Informatics, with some – the United Kingdom and the United States, for example – creating hybrid degrees that closely tie clinical leadership with informatics and digital transformation. These hybrid programmes allow clinical leaders to obtain new competencies in ICT-based quality improvement and change management to provide the needed leadership and ensure buy-in from health workers in general. In the United States, the HITECH Act funded two distinct health IT workforce training programs, which have trained 21 437 students and working professionals.15

However, more is required to ensure that the skills health workers (and the health workforce) needed for effective and safe use of emerging digital tools, technologies and data analytics are taught routinely. The same goes for human capital in administration, management and policy making, where digital, data and statistical literacy are equally crucial to drive positive transformation. With the current accelerated pace of technological development, it will become more necessary for health workers to regularly update their skills. As such, it is important to reconsider the skills delivery pathway, and encourage a shift towards a lifelong learning model. But concrete strategies for achieving this are still to be formulated in most countries.

1.4.1. Transforming skills and attitudes across the sector

Using data (and to a lesser extent digital technology) is still too often seen through a lens of risk as opposed to opportunity. This is an outdated, 20th century view, but is reflected in existing policies that make it very difficult to put data to work. Too often use of data is seen principally as a risk to privacy, cybersecurity threats and the costs of security breaches (including liability).

While data security and privacy are crucial, forward-thinking countries and health systems equally recognise and communicate the opportunities of using data. In contrast to previous decades, the 21^st century approach is to maximise the benefits while minimising the risks. Both are possible but only with strong, fit-for-purpose governance and policy frameworks.

Paternalistic attitudes that tend to prevail in health sector are also not conducive to promoting the use of data and information for purposes ranging from empowering people to take greater control of their health or effective sharing of data across specialties and settings of care. Beyond care delivery, this presents a barrier to secondary uses of data for research, public health and performance monitoring.

The way in which digital tools are developed serves to illustrate why this is a problem. For example, few health apps are designed with involvement of the end user – the individual with a particular health need or the health professional. Often, there is little input from clinical experts. More typically, digital tools and gadgets are developed by IT experts. Another example are EHRs, which are often developed and designed with a purpose other than making clinical care more effective and efficient. The technology ends up ‘getting in the way’ of not only delivering good services but also of generating high quality data that can then be used for other important purposes.

A fundamental part of addressing this revolves around equipping the entire health workforce with the skills and attitudes to make the most of data-driven technologies. For example by reforming both initial training and continued professional education and supporting health workers in acquiring new digital and teamwork skills. It concerns how health workers are socialised and taught to interact with colleagues in their as well as other professions and their patients. It also requires ensuring that individuals and patients feel empowered to work in partnership with their care providers (see Section 1.4.3).

As discussed, most countries are still at the preparatory stage of restructuring of health workers’ curricula to respond to the demand for updating of skills. Similarly, courses and teaching methods such as inter-professional education that aim at reinforcing skills related to teamwork and inter-professional collaboration are not yet routinely included in the core curricula. Only some countries – Canada, for example – have built in these skills explicitly into their core competency frameworks informing health professional education. The need for strengthening the skills of health workers to work with (as opposed to around) patients, such as skills in effective person-centred communication and shared decision-making, is also usually marginalised in education and training (OECD, 2018[58]).

Other opportunities include facilitating the integration of new professions and roles in health systems and relaxing rules around the division of work between health professionals, in light of the reorganisation of tasks that digitalisation can offer at all levels of the health system. More than half of the OECD countries have made progress in this direction by expanding the roles of nurses and allowing task shifting from physicians (and nurse practitioners) to nurses (and nurse assistants) (Maier, Aiken and Busse, 2017[59]) and by creating new medical specialists such as Clinical Information Officers.

1.4.2. New ethical frameworks are needed

At the new digital frontier, a growing number of health care activities will be performed by humans together with machines. How do health workers answer questions about collaborating with AI? Even relatively simple machine-learning models already used – such as those automatically stratifying patients into at-risk and intervention groups – give rise to questions regarding health workers’ and machines’ respective roles, accountability, and about how to ensure that digital systems do not crowd out the personal touch between patients and providers.

Other questions concern how to inform a patient when a risk-prediction model did not recommend treatment, what mechanism exists to override the model’s recommendation if necessary or, again, what happens if the model’s recommendation leads to a suboptimal outcome.

No progress will be made unless professional and ethical frameworks are updated. The 2019 OECD Recommendation on Artificial Intelligence can guide countries in this regard. The recommendation comprises two sections. The first outlines the principles for responsible stewardship of AI (inclusivity, sustainability, human-centric values, transparency, security, safety and accountability). The second lays out the policy requirements for trustworthy AI (investment, a digital ecosystem, an enabling policy environment, building human capital and preparing for labour market transformation, and international co-operation) (OECD, 2019[60]). Further delay will make health professionals hesitant to use data-enabled tools and other technologies that enhance cooperation among providers across settings.

The novelty and complexity of big data analytics raises a new set of ethics-related challenges for researchers, regulators and policy makers (Ienca et al., 2018[61]). Growing possibility to use non-traditional data sources such as social media are also challenging the existing ethical frameworks, especially when these data are used for purposes that benefit public health, such as disease surveillance and outbreak management. Work is needed by researchers, ethicists and policy makers to develop acceptable decision making frameworks. For example, Vayena et al. (2015[62]) established three categories for the ethical use of personal data in public health: 1. Context sensitivity (e.g. commercial versus public health use); 2. The nexus of ethics and methods (e.g. robustness of algorithms and data provenance); and 3. Legitimacy requirements (e.g. pest practice, regulatory bodies, public communication).

Data ownership is another related area requiring ongoing discussion (Koskinen and Kimppa, 2016[63]). Here, two challenges are emerging. First, the commodification of health data, with commercial entities acquiring and on-selling personal data for commercial purposes. While these data are typically, but not always, de-identified this still raises questions about consent, ownership and about who should profit from the knowledge extracted from them. Evidence suggests that the public are reticent about their data being used for commercial purposes. Yet a considerable part of health system activity could be deemed commercial and certainly has a commercial component. Excluding companies with the expertise and resources from using health data to, for example, develop improved diagnostic tools or precision therapies that maximise clinical effectiveness and minimise risk, may inhibit potential advances to human health and welfare.

Second is the opinion that individuals are the sole ‘owners’ of their data. In a world where these data could theoretically be re-deployed in an infinite number of ways, this becomes an important issue and a technical as well as policy challenge. Ownership can be debated based on who paid for the activities that generate the data (Koskinen, Kainu and Kimppa, 2016[64]). If paid for by a third party (typically the taxpayer or insurance plan enrolees) data may be considered a public good with the payers as well as the data subjects potentially entitled to a say in how the data are used (Rodwin, 2009[65]). These and other questions become even more challenging when genomic data are considered.

Even where personal health data are voluntarily sold by the data subject, this emerging market needs to be examined through the lens of ethics and policy. People should have the opportunity to sell their data. But the likelihood that these people would overwhelmingly be from the poorer and more vulnerable social strata, and the consequent risk of exploitation needs to be considered. For similar reasons, the World Health Organisation advises against paid organ donation (World Health Assembly, 2004[66]). In addition, the potential bias in the data could affect the findings of research and other outputs they are used for.

Clues on where societies stand on these issues can be found in regulatory frameworks such as the GDPR, which recognises the societal benefits that can be derived from health data (and puts these data in a special category). But this is a fast-moving field and these questions require deeper, ongoing discussion and the development of dedicated, clearly articulated ethical frameworks and charters.

1.4.3. Engaging patients while ensuring that no one is left behind

People appear to be quite willing to engage with data and information. In 2017, 3.7 billion health related smartphone apps were downloaded globally, up from 1.7 billion in 2013. The proportion of adults seeking health information on line more than doubled between 2008 and 2017 (Figure 1.7).

copy the linklink copied!

Figure 1.7. More people are seeking health information on line

Percentage of adults who sought health-related information online, 2008 and 2017

Source: OECD (2019[67]), “ICT Access and Usage by Households and Individuals”, https://dx.doi.org/10.1787/b9823565-en.

Digital transformation holds some promise to ‘democratise’ health and health care, and certainly to helping people better manager their own health. But for this to happen in practice, without widening health disparities, more needs to be done. For example:

• Helping patients access and use effectively their own medical records. Section 1.3.1 outlined the growing digitalisation of health records and the rising proportion of countries that enable people to access their data electronically. However, few appear to permit individuals to interact with their own data and evidence suggests that a small proportion of people actually do so (especially those with high levels of health need – see below). In some cases legal restrictions are in place preventing interaction with data. These may need to be re-examined with a view to what is best for health outcomes and care quality.

• Improving health and digital literacy is key. In 18 OECD countries recently surveyed, at least 30% of the population has poor health literacy levels and in 12 of those countries the proportion is over 50% (Moreira, 2018[68]). Ways to address this gap include raising digital and health literacy, but also better design of web portals or smartphone apps as well as better information about their availability.

  Canada Health Infoway recently launched a country-wide campaign, Access 2022, to promote the use of electronic health records by patients and health professionals. In Norway, the government has allocated NOK 27.5 million since 2014 to increase participation in the digital health portal and developed training programs for digital skills at the municipal level.

• Focussing on uptake among the patients who stand to benefit the most. Evidence suggests that the very complex or vulnerable people that could benefit the most from better care models do not engage with digital technology like their more healthy counterparts. The structural inequalities of digitalisation, which mirror more long-standing social stratification (Figure 1.8), are very telling in this context. In the Netherlands, for example, just 4% of the chronically ill population reported using a personal health record. Addressing this requires a specific focus on vulnerable population segments. For example, educational courses and tutorials targeted at patients with low digital literacy have been implemented in Estonia and are being trialled in the United States.

copy the linklink copied!

Figure 1.8. Socio-economic disparity is observed in most on line activities – including seeking health information

Diffusion of selected online activities among individuals aged 16-74 in OECD countries, 2018

Source: OECD (2019[67]), “ICT Access and Usage by Households and Individuals”, https://dx.doi.org/10.1787/b9823565-en.

1.4.4. Opening data availability and facilitating their integration

The institutions and habits formed in the pre-digital era have entrenched fragmentation and a silo approach to doing things. Meanwhile, the utility of data for producing actionable knowledge rises exponentially if they can be combined and pooled – a requirement that is blocked by fragmentation.

Making data available to all stakeholders who can turn them into valuable knowledge is also important. In non-health sectors, opening public sector data has been shown to catalyse innovation by the private sector (OECD, 2019[1]). The openness of government data in general varies across OECD countries (Figure 1.9). Opening access to non-rivalrous health data will help transition them towards being seen as a public good. The benefits to societies and humanity of making them available on the broadest possible scale can therefore be maximised (Grossmann et al., 2010[69]; OECD, 2019[1]).

copy the linklink copied!

Figure 1.9. Openness of government data in OECD countries

Open-Useful-Reusable Government Data Index (OURdata), 2017

Note: The OURdata indices take values between zero and one, with one being the most open. Each component can score a maximum of 0.33.

Source: OECD (2019[1]) ”Going Digital: Shaping Policies, Improving Lives”, https://dx.doi.org/10.1787/9789264312012-en.

Addressing the issue of openness, and facilitating comprehensive data linkage require policy to create the right environment and implement the right levers to make progress. For example:

• Improving the preparedness for EHR data to be used for secondary purposes such as research. Most countries are capturing clinical data electronically, but based on a 2016 OECD survey, only the countries in the upper right quadrant of Figure 1.10 report readiness to put them to work. Of note is Estonia, which has implemented a state-of-the-art digital infrastructure. Yet, it reports not being very well prepared to use this information for secondary purposes, highlighting the need for data governance and capacity to put data to work. The required operational, policy and governance levers include:

  • A national plan or strategy that includes the secondary uses of these data.

  • Having a legal framework that enables these data to be securely extracted and used for secondary purposes

  • Building the capacity to analyse the data to generate meaningful information and knowledge

  • Ensuring data completeness (population coverage), minimum data specifications, common data terminology standards, and unique patient and provider identifiers.

  • Financial incentives to not only adopt and maintain high quality electronic records, but deploy data and digital technology to improve services and outcomes (see Section 1.4.5).

• Encouraging common approaches to data terminology and exchange standards. The lack of standards creates barriers to – and inefficiencies in – sharing and diffusing data within countries and across them. Other sectors have developed protocols and standards for computer communications, data exchange and cybersecurity – to the benefit of entire economies and societies. There is no reason why this cannot happen in health care. For example, Internet Protocols (TCP/IP) were adopted across the world; exchange standards for data sharing among banks, clients and financial services; and cybersecurity standards for aviation have enabled globalised control of air transportation.

  The bespoke approach that has been applied to health data development, on the other hand, causes health care organisations, systems and countries to speak ‘different data languages’ – a major barrier to modernising health systems and achieving similar economic and social benefits as have been realised in other sectors. The inconsistency of health data is also a major reason for the commodification of personal health data, with profits to be made from aggregating, cleaning and harmonising large datasets.

  There is some progress on developing common data terminology standards that harmonise data and enable their pooling, with a number of public and private sector initiatives currently under way. The EU has established policy to support the sharing of health data across borders, including work toward fully interoperable electronic records for research, diagnosis, treatment and disease prevention, and policies promoting effective sharing of genomic datasets to advance personalised medicine. The European Health Data and Evidence Network (EHDEN) is a shared public/private investment in developing an approach to standardising a wide range of health data. EHDEN to create a common data model to facilitate health statistics, monitoring and research undertaken by governments, universities and the private sector entities, such as for pharmaceutical research (EHDEN, 2019[70]). Global private sector initiatives such as the DaVinci project and FHIR are also aiming to establish common approaches and standards to enable the sharing of health data across borders.16

  These initiatives should be welcomed. But there is a risk that their multitude exacerbates the very challenge they are trying to solve. The work that has taken place to date needs to be consolidated, and countries should agree and reach consensus on standards for the growing range of data relevant to health that can and should be shared across borders. This effort requires global coordination involving a range of stakeholders in both the public and private sectors. An international organisation such as the WHO or the OECD could facilitate this needed collaborative work.

• Ensuring that data are available broadly, securely and at low cost. In contrast to data that are generated as a pure by-product of health care delivery or other human activities, access to some types of data whose generation entails costs may need to be subject to exclusivity through intellectual property rights. Policy on intellectual property, generally developed for tangible products, may need to evolve to strike the right balance between economic incentives to generate data and the societal benefits of open access to data (OECD, 2019[1]). This relies on solid data governance and policy frameworks (see Section 1.5.2).

copy the linklink copied!

Figure 1.10. Some countries are ready to deploy EHR data for secondary purposes

Source: OECD Survey of Electronic Health Record System Development and Use, 2016; Oderkirk (2017[34]), Readiness of electronic health record systems to contribute to national health information and research””, https://dx.doi.org/10.1787/9e296bf3-en.

1.4.5. The right incentives play a major part, and rely on policies and institutions

Transformation, especially in complex systems, does not occur organically. Transformation only occurs under the right environment, when the right incentives are aligned. The first industrial revolution began in Britain not because of the invention of steam power, which had existed for decades. Rather, it was the high cost of labour and low cost of energy (coal) in Britain at end of the 18th century (Allen, 2011[71]). Steam power was integrated into the industrial system – or rather the industrial system was reconfigured to make the most of steam technology – only when it made economic sense to do so. When it was, it transformed the industrial production methods and entire economies in the process.

Health is a highly regulated sector and the incentives, which may be financial or not, throughout health systems are heavily influenced by policy. Policies and their enabling institutions are therefore pivotal in driving a digital transformation. However, it can be argued that current policy settings create incentives that do not promote the intelligent use of data and digital technology. This dissonance may be a core reason for the slow pace of digital transformation in the health sector. It highlights the importance of re-designing the policy and institutional settings that underpin how a health system functions – and that digital transformation is not possible without institutional reform.

New payment models (see Section 1.3.6) can illustrate this issue. Remuneration plays a key part in dictating individual and institutional behaviour. While data infrastructure and digital technology provide the tools, uptake will be slow without the deeper institutional and policy changes. For example, continuing to pay for individual fragments of care will do little to stimulate the adoption of technologies that can make service delivery more people-centred materials (Murdoch and Detsky, 2013[11]). Alignment of budgetary and expenditure accountability across organisations, regions or care settings is required, and providers and provider organisation would need to move towards a team-based approach of delivering care across settings. Such fundamental changes in practice and behaviour requires tailored incentives.

Payers also need to ensure that integration of new technology does not destabilise existing services as this may compromise clinical safety as well as trust and morale among patients and providers. NHS England, for example, is in the process of updating its contracting and payment rules for primary care to address emerging challenges and embrace opportunities of digital technology because a growing share of primary care activity occurs on digital platforms. The new payment system will account for the ‘mix’ of ICT-based and face-to-face contacts. While this can only be achieved by using various forms of available data, reporting of data on the costs of digital provision in general practice is a critical part of the new model. This will enable to enable close monitoring, and continuous development of payment rules (NHS England, 2018[72]).

Harmonised information systems, appropriate skills and attitudes, ethical frameworks and engaged stakeholders provide the possibility for a positive digital transformation. But progress will be slow without a significant overhaul of existing institutional and policy frameworks that steer health system behaviour.

1.5.1. An overarching strategy is the foundation

All countries that are on track to harness the opportunities of health data and digital technology have one thing in common: an overarching digital strategy. This begins with the acknowledgement that data and digital technologies are a valuable resource to advancing human health, and a vision of how this resource can be used. Clearly articulating a set of principles that align with broader health system objectives can convey a number of cross-cutting advantages to driving a digital transformation at all levels of a health system. This allows better planning and execution of the necessary changes, as well as adaptation of organisational structures, institutions and workflows.

For example, in Sweden successive digital health strategies have been in place since 2006 (Swedish Ministry of Health and Social Affairs; SALAR, 2016[73]). The latest strategy was jointly developed by the national government and the Swedish Association of Local Authorities and Regions (SALAR) and endorsed in 2016 to guide the digital transformation through to 2025. The goal of the strategy is to “make it easier for people to achieve good and equal health and welfare, and to develop and strengthen their own resources for increased independence and participation in the life of society” (ibid, p.3). The Swedish strategy comprises three streams:

1. 1. increasing digital information exchange, both between different public authorities and with citizens, while safeguarding privacy and data security;

2. 2. advancing semantic interoperability of data in the health system; and

3. 3. ensuring technical interoperability of ICT systems.

As shown previously, Sweden reports near-universal use of electronic medical records in primary care and hospitals, the second highest rate of availability and linkage of key health datasets for secondary use and the use of routine data for monitoring medicine use and expenditure. The proportion of people who access health-related information online in Sweden has increased from less than 30% in 2007 to more than 65% in 2017, which is among the highest shares across OECD countries.

1.5.2. Good data governance can develop the key element of trust

Governance, in the context discussed here, means creating a policy environment that enables digital technology to be deployed, and knowledge to be extracted from data while at the same time ensuring that this is done securely, and in a way that respects individuals’ privacy and their preferences. Governance includes technical, legal and policy levers. In most countries, the key elements of governance in the context of data and digital technology such as privacy legislation were developed and instituted last century. The HIPAA Act, for example, which sets out the data privacy and security provisions for safeguarding medical information in the United States, was enacted in 1996.

The Recommendation of the OECD Council on Health Data Governance, welcomed by Health Ministers of OECD countries in 2017, sets out the fundamental elements countries need to manage and use personal health data for purposes that serve the public interest, while protecting privacy and ensuring data security. (OECD, 2019[75]). The Recommendation’s twelve mechanisms can be grouped into categories relating to technical, policy and communication requirements. Implementing the Recommendation will go some way in addressing the barriers of using data (and digital technology) and putting these to work for positive system transformation.

For example, the Recommendation provides clarity for health system leaders to communicate the benefits of a digital transformation. This enables an informed public discourse that encompasses the opportunities as well as the risks, which include, of course, the benefits foregone of not putting data to work. It also dispels the notion of a trade-off between data protection and their use. The Recommendation also contains technical requirements to ensure efficient data exchange can occur, and necessitates the institution of strict cybersecurity policies and processes. Perhaps most importantly, it is designed to promote harmonisation of policies and legal requirements governing access, management and use of personal health data within and across countries. Such ‘policy interoperability’ is critical in pooling of large datasets for public policy purposes.

1.5.3. Building capacity for a digital transformation requires investment

Strategy and governance are very important. But without the institutional, operational and policy capacity to extract knowledge from data and then deploy this knowledge effectively, any potential for positive system transformation will not be realised.

Capacity building covers the barriers and enablers discussed in Section 1.4. It also means ensuring that the underlying infrastructure is in place to enable data to be put to work effectively. This applies to hardware as well as software and other intangible products to generate, store, link and analyse data.

Institutional capacity must not be neglected. The best digital strategies, governance frameworks and skilled workforce will not ensure that knowledge extracted from data is applied to meet policy objectives. Existing policies, processes and workflows need to be reformed in a way that enables data-driven knowledge to be applied most effectively. It means equipping agencies and organisation with access to data and the ability to extract knowledge from them. It also means arming these institutions with the necessary ‘teeth’ to act on the knowledge generated. The roles, responsibilities and powers of regulators, payers, public health authorities and other health system actors must be aligned with what a data-driven approach can do.

This includes re-thinking the prevailing macro-policy settings and institutions that dictate how things are done in health. This concerns not just remuneration, but also how regulation is set and enforced or how performance is evaluated. Policy makers must think about lowering the barriers for smart deployment of digital technologies and intelligent uses of data that can help achieve policy objectives – from better quality care, to using scarce resources more efficiently at all levels of the system.

1.5.4. But the returns can be considerable

Digital technology is general-purpose, meaning that it can penetrate all aspects and activities of an organisation or endeavour (Brynjolfsson, Rock and Syverson, 2019[27]). Rather than creating new things to do (a main driver of spending growth in health), data and digital technology are particularly suited to improve virtually every existing process and activity in an organisation, industry or sector (see Section 1.2.2). The resulting efficiencies and productivity gains of small and incremental as well as large innovations are considerable but also difficult to quantify ex ante. For example, thirty years ago it would have been impossible to predict the myriad ways that the internet would transform the global economy.

Capturing the macroeconomic effects of a fundamental system transformation, the costs and benefits within and outside the health system, is a challenging task. Suitable economic models are yet to be developed (Marino and Lorenzoni, 2019[13]). Nevertheless, by making virtually every aspect of what a health system does more efficient and productive, the expected health and economic dividends of a digital transformation in health are considerable – even when factoring in increased investment The dividend can be broken into two parts:

1. 1. reducing wasteful and inefficient practices, redundant processes and clinical harm, and

2. 2. addressing unmet health need, improving public health and enabling access to safer and better therapies.