4 minute read

Anthony Tugman
E534
9/30/20
Assignment 6

The Covid-19 pandemic has brought a focus to tele-health due to its ability to promote social distancing as well as its ability to handle the sheer number of patients. The focus on tele-health is most often in its use of connecting patients virtually with healthcare providers. While this task is accomplished successfully, largely due to the high quality cameras and internet that are widely available, the increased use of tele-health and implementation of big data can bring further benefits. In fact, it is already making large contributions in the healthcare industry improving care, accuracy, and efficiency for both patients and healthcare providers. Prior to Covid-19 tele-health adoption was at 11%, increasing to 45% during the pandemic 1. The increase in adoption has led to a flood in big data. Through the use of AI analytics, this data can provide countless meaningful insights.

The technology surrounding tele-health has been under development and deployment for many years, however the pandemic increased adoption and provided a catalyst for this industry. The overall goal of tele-health is to provide healthcare to underserved individuals through increased connectivity. With the increase in connectivity comes the increase in big data and connected monitoring devices. The information and insights produced by these monitoring devices is an untapped trove of medical patterns, statistics, and associated information that can inform healthcare and policymakers in their decisions. The most important outcome from this is the newly sourced data that can be used in a variety of ways.

To explore how this technology is being applied to tele-health an example use case of photo recognition for skin conditions will be described. In itself, the creation of an algorithm to detect skin conditions is nothing novel. But when the algorithm is making potentially life or death decisions there are some additional intricacies built in. Google has created an algorithm that is able to identify 26 skin conditions with as much accuracy as trained dermatologists 2. To achieve such accuracy the AI system was designed to mimic human dermatologist behaviors. This means using associated metadata such as region, race, gender, health history, etc. as well as the image itself to make an informed list of possible conditions, just as a dermatologist would. That is, an actual diagnosis is not being made, but rather suggestions as to how to narrow down lab tests for result determination 2. And while the success rate is impressive, this system has drawbacks predicting rare or undertrained conditions. These drawbacks arise from the limited training datasets available. In fact HIPAA and additional privacy protections have made it difficult for tele-health with AI to become more mainstream 3.

It seems that the patient privacy protections laws are standing in the way between tele-health and its true potential. While the technology is ready to go, as it has been developed for other fields and would be easy to transfer, time needs to be taken to stand back and establish trust with the public. To establish public trust, especially concerning sensitive health information, standards, transparency, and accountability must be established. With the increased adoption of tele-health, now is a better time than ever to gain the public’s trust. Once this has been accomplished it will be possible to amass larger, more organic datasets that contain even the rarest ailments making diagnoses more accurate.

Although not a goal of the discussed algorithm, one of the most interesting facets of this form of data collection is the patterns that can be observed. If suddenly all health information was available to researchers it would be possible to see where certain diseases are concentrated, what factors may be contributing, as well as how resources should be distributed. With the addition of an exponential amount of datasets, these AI systems could even become predictive in use. That is, they could be used to monitor conditions against a standard to alert of certain abnormalities in the healthcare system such as an incoming pandemic.

  1. “How intelligent data transforms health in the time of COVID-19”, MobiHealthNews, 2020. [Online]. Available: https://www.mobihealthnews.com/news/how-intelligent-data-transforms-health-time-covid. [Accessed: 15- Oct- 2020]. ↩︎

  2. K. Wiggers, “Google says its AI detects 26 skin conditions as accurately as dermatologists”, VentureBeat, 2020. [Online]. Available: https://venturebeat.com/2019/09/13/googles-ai-detects-26-skin-conditions-as-accurately-as-dermatologists/. [Accessed: 15- Oct- 2020]. ↩︎

  3. “How HIPAA Is Undermining IT and AI’s Potential To Make Healthcare Better - Electronic Health Reporter”, Electronichealthreporter.com, 2020. [Online]. Available: https://electronichealthreporter.com/how-hipaa-is-undermining-it-and-ais-potential-to-make-healthcare-better/. [Accessed: 19- Oct- 2020]. ↩︎