In December 2016, a case study in ScienceDirect examined the impact of electronic health record (EHR) information on modern healthcare approaches. They found that EHR data contains highly valuable longitudinal and real-time clinical data with the potential to optimize clinical workflows, improve patient safety and create positive health outcomes. Specifically, their team reduced the time to diagnostic testing for colorectal and prostate cancer using EHR-triggered software, and increased the proportion of tested patients that received an appropriate follow-up examination [1].
Unfortunately, EHR data isn’t so easy to obtain. As the vast majority of modern health systems and clinics operate on EHRs, the probability of attack from malicious hackers has also increased. Cybersecurity breaches in healthcare are more common than expected, with millions of EHRs being hacked per year resulting in annual costs of $6.2B [2]. This sensitive nature of health data, specifically what is considered Protected Health Information (PHI) under HIPAA regulations, has necessitated bureaucracy and regulations into health systems. From fragmented information technology (IT) and internal consulting teams to training courses and cybersecurity reviews, the ecosystem around retrieving health data has become a painfully slow-moving nightmare.
The same research team that published the case study mentioned above analyzed their challenges in working with EHR data. The main problem areas were getting access to the EHR system, interpreting vastly unstructured data, and working with local IT and EHR personnel.
Any researcher who has attempted to gain access to EHR data at universities and hospitals is well-acquainted with these pain points. Restrictions differ from place to place, and specific teams dedicated to handling these restrictions rarely communicate with one another. Training courses that a researcher must take for approval at one institution may be different at another. The worst health systems have far more complicated paths to approval that can take over a year. For example, many universities have processes in place where the researcher submits a request for data in the form of a proposal, which is stuck in review for weeks or months. Upon acceptance, the researcher usually finds out that certain data requires approval from multiple other departments and must repeat the process for each department, resulting in exponential research delays [3].
Furthermore, network security, software compatibility, limited EHR features and services ,and the research team’s lack of resources and time make it difficult to devise efficient software pipelines that integrate with the data they need. The table above mentions that the research team, although granted access to the EHR, was not approved to use the secure messaging feature. Instead, they had to manually call providers. They also had issues setting up a remote network connection for retrieving the data they needed, resulting in slow and unreliable data transfers. For data scientists this is one of the most frustrating aspects of their work. A feeble network connection that is fed through a string of VPNs, firewalls and other gateways throttle API connections that transfer data [3].
These horror stories are not acceptable.
Advances in computational methods, hardware, and cloud services have made it far easier to run large-scale statistical analyses on health data. Thanks to the convergence of these technological developments, what were once ethereal pieces of information that rarely amounted to insight have now turned into solid foundations for accurate and precise approaches in healthcare. The large scale adoption of electronic health records and the growth in self-generated data through mobile and wearable devices have created an abundance of data on patients, especially data that was never available before [4]. Many studies have already been done that have shown the ability of modern software and statistical approaches to extract important patterns and events from data, which translates into interventions, recommendations and actions that save lives, money, and time. These initial studies and software have turned into publications, open-source software packages, health A.I. companies and large scale initiatives and groups such as IBM Watson Health [5].
This new paradigm of healthcare, characterized by concepts such as “personalized health” and “precision medicine,” must be supported in full if we want to turn idealistic goals into achievable objectives. That starts with opening up resources for collaboration, specifically data as no algorithm can run without it.
Sources
[1] Russo, E., Sittig, D., Murphy, D.R., Singh, H. (2016). Challenges in Patient Safety Improvement Research in the era of Electronic Health Records. Science Direct, 4(4), 285–290. https://www.sciencedirect.com/science/article/pii/S2213076415300907
[2] Becker’s Hospital Review (2017). Healthcare Breaches cost $6.2B annually. https://www.beckershospitalreview.com/healthcare-information-technology/healthcare-breaches-cost-6-2b-annually.html
[3] Fleetcroft, R. (2015). Difficulty Accessing Data from Randomised Trials of Drug for Heart Failure: A Call for Action. BMJ, 351. https://www.beckershospitalreview.com/healthcare-information-technology/healthcare-breaches-cost-6-2b-annually.html
[4] — Chang, Y.Y., et. al. (2017) The Asthma Mobile Health Study, a large-scale Clinical Observational Study using ResearchKit. Nature Biotechnology, 35, 354–362. https://www.nature.com/articles/nbt.3826
[5] — Wilbanks, J.T., Topol, E.J. (2016) Stop the Privatization of Health Data. Nature: International Weekly Journal of Science, 7612(535), 345–348. https://www.bmj.com/content/351/bmj.h5002
