Patient Care Intervention Center

Business Hours

Mon - Fri 8AM to 5PM

Connect with us

info@pcictx.org

Help our cause

Contribute | Volunteer

Contact Us

Leave a note

Social

  • Home
  • Contact
  • Social
  • The role of machine learning in effective and efficient care coordination
The role of machine learning in effective and efficient care coordination

The role of machine learning in effective and efficient care coordination

Print Email
(4 votes)

We see machine learning applied in a lot of business these days, from optimal search engine results to shopping recommendations. When IBM Watson won “Jeopardy!” against humans, it marked the beginning of a new era in artificial intelligence, where machines don’t just pull answers from a vast set of available resources but can analyze and process the information within a really short period of time. The healthcare industry is no stranger to machine learning. For example, Google’s machine learning algorithm is now detecting breast cancer with an 89% accuracy, when compared to 73% for a pathologist.

The key lies in training and re-training the model, to make predictions more accurate.

Organizations performing care coordination service are also implementing machine learning and integrating high-tech models to provide better care to patients. For example, tracking data from wearables or cell phones and setting up automated alerts for care coordinators.

So, what is Machine Learning?

In simple words, it is building a system that learns from data.

How can machine learning be applied in the world of care coordination?

Helping organizations understand the connection between factors like demographics, neighborhood, social life, medical diagnoses, chronic conditions, criminal offense and childhood experiences, and the role these factors play, in an individual becoming a High-Need High-Cost (HNHC) patient.
Predicting the effectiveness of a specific care plan for a HNHC patient based on some of the factors mentioned.
Help understand the root causes to determine patterns that result in an individual becoming an HNHC patient, to asses if intervention can be provided earlier in the cycle.
Avoiding a potential ER visit through improved care and help connect the needs of an HNHC patient to the resources available in a more automated manner.

How do we setup and configure machine learning environments?

There are various providers in the market who offer solutions to build predictive analytics systems. Some of the providers are Microsoft, Amazon, and Google. A more extensive list of providers can be found here.

The main steps involved in building a machine learning workflow is shown below:

Machine learning workflow

The key lies in training and re-training the model, to make predictions more accurate.

Why and how are we adopting Machine Learning at PCIC?

Our core mission is to improve healthcare quality and costs through data integration and care coordination. Being a non-profit organization, we do this with limited resources and need to apply lean principles in developing new and more efficient models to the way we provide healthcare. Machine learning helps us understand and detect problems faster, find patterns in the data that help us apply solutions quicker and in a more uniform manner. Manually developed custom solutions for every individual makes any care coordination program expensive and difficult to scale. At PCIC we are trying to solve that by integrating technology like machine learning into our overall workflow of patient care.

Our core mission is to improve healthcare quality and costs through data integration and care coordination.

PCIC is working towards implementing machine learning in its day-to-day operations by integrating trained models into our health record system - StreetEMR. We are updating and re-designing StreetEMR for easier and improved collection of data during care coordination. The Unified Care Continuum Platform that we are building into the system will use machine learning models to identify the best care manager across multiple organizations to assign to a patient as well as to match up the needs of the patient to resources available in the most optimal way.

What are some of the challenges?

Both the quantity and quality of data are equally important to building highly efficient models. It can be challenging at times to get the right data in a timely manner.

We analyze data from a lot of different organizations and systems, including hospitals, health insurance companies, social service agencies, police and fire departments. The solutions to providing effective care to the patient requires bringing together different stakeholders not just through a data sharing platform, but also an effective communication platform. This relationship building step with different agencies requires time, effort and a collaborative mindset by all members. It requires us balancing the needs of an individual agency to that of the whole (community).



Authors


Kallol Mahata

Vice President

Swetha Kiledar

Software Engineer

Last modified on Thursday, 05 October 2017 19:27

Recent Posts

Our Vision

At PCIC we envision a coordinated health safety-net where all stakeholders share data to make better decisions.

Our Mission

To improve healthcare quality and costs for the vulnerable in our community through data integration and care coordination.

Copyright © Healthcare for Special Populations. All rights reserved.
Team Site