Ready for the AI, Blockchain and IoT Wave in Healthcare? Here are Some Practical Considerations.

Digital innovations using connectivity and technologies like blockchain as well as artificial intelligence (AI) offer promises to transform healthcare. Some claim that these technologies are trending towards the peak of the Gartner’s hype curve, if not already there. In fact, about $2.7B worth of venture investment has gone into AI in healthcare over last 6 years according to Rock Health – just over 10% of all venture dollars invested in digital health during that period.

One must wonder if these investments are going to pay off. What is real beyond just the hype? Where are the meaningful opportunities? The answers to these questions require gaining a better understanding of the underlying technologies, their current maturity levels, their targeted applications, and the requisite eco-system. Some of the key terms are defined below to better understand the topic:

Internet of Medical Things (IoMT): It is a network that connects various applications, devices and “things” with clinical relevance. The network enables collection, sharing and analysis of data.

Blockchain: It is a chain of databases, called blocks, which are linked to share data in a secured manner, assuming the presence of 3rd party adversaries, without changing any attributes to the sources and keeping record of transaction data at each level with timestamp.

Artificial Intelligence (AI): These are adaptive (or learning) algorithms (or systems) that use available input-output data to model a system as well as it can. The model of the system automatically updates itself, as well and as fast as it can, based on new data of changing operating conditions. Hence, the model can be used to predict outcomes assuming the environment doesn’t materially change. The algorithms typically require trade-offs among requisite data size, accuracy of the prediction, and speed of response to changing operating conditions.

The value from IoMT, blockchain and analytics based advanced decision support is conservatively estimated to be $93B in the US. The value reflects productivity gains in administrative and operational areas as well as in the clinical side, including reduction in hospital acquired infection and medical errors. Various studies suggest that simply connecting medical devices (i.e. IV pumps, vital sign monitors, ventilator, etc.) to EMR/EHR could release up to 8% of nurses’ capacity from administrative tasks such as charting data; it also helps prevent about 800 human data entry centric errors daily in a typical hospital. Simultaneously, some of the advanced decision support systems allow pre-problem alerts in critical care environment to better serve patients. As shown in the figure below, these advanced decision support systems are currently offered by large market presence players who have ample amount of data in a given domain.

Advanced Decision Support in Healthcare IoMT

In general, the digital trend in healthcare is looking to move from the left with connecting devices to the right where vision of personalized medicine can be realized as shown in the digital eco-system figure below. The progression from the left to the right includes moving from having comprehensive set of data in a single domain (i.e. radiology, hematology, etc.) for many patients to limited cross-domain data at the enterprise level for only those patients who are treated at this enterprise, and eventually to having comprehensive cross-domain data from all enterprises for a given individual. This progression would benefit from both IoMT and blockchain technologies. Blockchain offers the promise of having various entities share patient data in a more secure manner once patient consent is obtained. Traditional deterministic optimization or newer AI based analytics performed using the available data enables advance decision support in form of software applications.

Digital eco-system in healthcare using IoMT, blockchain and AI

The key initial challenges to progress on this roadmap include interoperability and ability to connect various disparate systems from many entities. About $200M worth of venture investment has gone into this area over last 6 years, excluding transaction values of medical device interoperability companies like Capsule and iSirona. Today, many if not most of the devices have proprietary interfaces. They often connect through proxy devices such as Capsule, iSirona, etc. Similar to EHR vendors, various IS vendors are also hesitant to open their interfaces to share data. Hence, the challenges are of both technical and commercial natures.

Beyond the key initial challenges, there are 5 operational considerations that shouldn’t be underestimated:

1. Data integrity

The primary reason for building the digital eco-system is to obtain necessary data in a timely manner to enable effective decision making. It is hard to generate high quality decisions if data going into the system is flawed. While AI algorithms offer some resiliency to bad data over time, they still require an initial set of good data to train the system. Hence, it is important to understand if there is sufficient amount of data for the mission, if the data is validated to be accurate, and if the validity holds true over time with changing operating conditions.

Data integrity shouldn’t be taken for granted. Many examples have recently surfaced that call for special attention to data integrity.

  • A recent CNBC article outlined how a patient’s medical record indicated that she’d given birth twice, when in reality, she’d never been pregnant. A different record for the same patient wrongly included a diagnosis of diabetes.

  • Recently, a $32M malpractice award was upheld in Massachusetts Appeals Court for essentially not having complete information in the medical record. In this case, patient’s primary care physician failed to note a brain aneurism in the medical history. Consequently, the patient suffered a brain hemorrhage after giving birth because the obstetricians was not made aware about her aneurism.

Some of the experts estimate that about 70% of medical records have wrong information. A Joint Commission study indicates the data error via human-computer interface to be the leading cause of adverse patient events in Health IT. Underestimating the data integrity challenge could have serious patient safety and legal liability implications.

2. Clinical network integrity

Clinical/Health IT (HIT) networks are mostly built to collect and share data. As shown in the figures above, they connect medical devices, databases, and applications that take information manually or in automated manners. Hence, it is important that these networks are designed, commissioned, and managed over their lifecycle to operate with high reliability, offering the highest availability. The error of commission, error of data omission or transmission, and/or incompatibility between multi-vendor products could lead to either malfunction of systems or data availability/integrity challenges.

Imagine if the critical network were down and critical patient data didn’t reach clinicians on time or in the right format!

  • Many studies performed by the Joint Commission have uncovered adverse events including patient deaths, permanent loss of function, unexpected additional care, etc. due to HIT issues. An article in HealthData Management suggests that 18% of EHR related patient safety incidents originate from interoperability issues.

  • A study by The Doctor’s Company indicates substantial increase in claims involving EHR. Close to 50% of the claims point to design or integration related errors.

Clearly, underestimating Health IT network integrity challenge could also have serious patient safety and legal liability implications.

3. Cybersecurity

Connecting more devices, applications and/or “things” increases the cyber-attack-surface. At the same time, healthcare is currently the industry most targeted and susceptible to hacking. Cyberattacks in healthcare could lead to malfunctioning of medical devices, clinical systems or applications; it could also potentially lead to omission and/or alteration of data. In other words, cyberattacks could not only directly impact patient safety, but also indirectly through impacting data integrity and clinical network integrity.

In the above figures, medical devices are at base of the pyramid. According to MediTechSafe discoveries, on average 44% of the connectable medical devices in provider settings are hackable.

4. Efficacy of AI models

In the given digital eco-system, a large amount of data would be analyzed for advanced decision support. In many use cases, deterministic optimization by computing a large amount of data could be sufficient to deliver incremental value. The accuracy in such cases would be very high if the data integrity were high. The aspiration in healthcare, however, is to move from diagnostic capability to prognostic capability. Could we accurately predict health issues? This is where the industry looks forward to using AI models.

The important part of using AI is to know what level of accuracy is attainable in a given use case and if that’s sufficient. More importantly, what are the consequences of inaccurate outcome?

  • Recent articles highlight how IBM’s Watson system gave unsafe and incorrect treatment recommendations during trial diagnoses at Memorial Sloan Kettering Cancer Center. False positives could also be concerning. For example, if an AI application falsely indicated probability of a potential health complexity, what would the next steps be? Considering potential legal liabilities, many clinicians may choose to either prescribe more tests or start some therapies. This could not only increase cost of care but also add health related worries to the patient.

  • In a different use case, Arkansas’s AI based Medicaid benefit allocation system inaccurately cut a resident’s benefits, which resulted into a successful lawsuit.

In a complex healthcare environment, it is difficult to achieve very high level of accuracy from the AI systems. To put things into perspective, 3 Sigma level of accuracy (93.3%) would generate inaccurate outcomes in about 30 million US patient visits every year if every patient were exposed an AI based health predictor.

5. Governance

Before a new drug or medical device is introduced in the market, it goes through a rigorous FDA approval process due to of the potential patient safety concerns. Similarly, digital health applications and Health IT system implementations can also benefit from an appropriate regulatory scrutiny because of the above-mentioned patient safety concerns. Moreover, increasing use of AI in patient care raises interesting governance related questions. For example,

  • Who is held accountable for an incorrect harmful outcome (i.e. wrong treatment) per an AI recommendation?

  • How should a resolution be reached in the case of differing opinion between the man (i.e. doctor) and the machine (i.e. an AI application) for a given situation? Who should bear the burden of the resolution process?

  • What % of inaccuracy in outcome should be allowed for a given application? Who is liable for those who are negatively impacted from this decision?

The advanced technologies like AI, blockchain and IoMT can certainly help in improving quality of care and cost. In fact, some of the initial limited scope pilots have demonstrated the potential. To drive any meaningful change, visionaries have to paint the picture of what’s feasible and create the hype; it helps in building the initial momentum and directing investments towards the vision. The industry, however, now needs to address some of the technical, ethical, regulatory/legal, data and business model related challenges to realize the vision.

At a micro-level, health systems could look at their cultures, processes, organizational structures and talent.


  1. Consider having an organization, inclusive of an Engineering VP role, and necessary talent DNA to drive digital led transformation. It may require optimizing and re-purposing headcount from different areas to manage cost.

  2. Strengthen the risk/crisis management program by including clinical network integrity (IoMT) and cybersecurity of medical devices/network types of risks. Board members have to be aware of these risks as they may be held liable if they failed to exercise oversight with respect to their organization’s cybersecurity risks.

  3. Initiate a data integrity program that may start with improving quality of the current data but also include a mechanism to sustain the quality over time.

  4. Institutionalize a rigorous RoI minded innovation adoption process. A program approach is ideal in this case with C-suite participation in steering committee

A MediTechSafe leader can be available to discuss the recommendations further. Some best practices are also included in our article: Digital Led Transformation in Healthcare: 4 Practices to Adopt from the “Product” Companies.

MediTechSafe has developed a proprietary solution to help healthcare providers manage their cybersecurity, medical devices and clinical networks related risks considering both IT and Operations Technology (OT) needs. If you are a healthcare provider (or a biomed services provider) interested in learning more about MediTechSafe’s solution, you could reach us at