Succeeding with predictive analytics in healthcare: 10 steps to get started

Listen
Text
  • Small
  • Medium
  • Large

A recent survey by CDW Healthcare Analytics indicated more than two-thirds of healthcare decision-makers consider analytics to be among their top three priorities1.

Given its immense potential to help increase patient safety and care quality while reducing healthcare costs, analysts estimate the healthcare analytics market to be over $20 billion by the year 2020 (Market and Markets, iQ4i).

Another study by Health Information and Management Systems Society (HIMSS) and Jvion surrounding the use of predictive models in healthcare showed that while a small fraction of providers today (about 15 percent) actually use predictive modeling, more than 90 percent of these organizations use the outputs to make better decisions on patient care and management2. The study identified specific areas – readmissions, patient deterioration, sepsis and general patient health – where providers are focusing their predictive analytics initiatives.

While the outlook towards adoption of predictive tools and technologies looks positive, the truth is a majority of organizations, even larger ones, are still in the planning phase. Smaller providers, often times, are waiting to learn from the larger ones. The challenge stems from organizations' readiness to adopt, as well as technology and data constraints in using predictive analytics. Based on our work in data analysis and predictive analytics, here are the 10 important steps that organizations need to take:

Step 1: Be ready to go the distance
Broad-based predictive analytics initiative is a long-term investment in technology and skilled resources. Considering the immense variability in format, structure, source system, nature and type of data, all within a single organization, data scientists don't have a complete picture of use cases they have to develop or analytical problems they need to solve.

Different types of use cases need to aggregate data from a variety of sources. Technology and data teams need to ensure they invest in a robust data aggregation framework that is compatible with industry standards such as QDM and V3 RIM as well as leverage existing investments in analytics technology (Business Intelligence (BI) / Data warehouse (DWH)). It is also imperative to build a data sandbox where data analysts can build, discard and rebuild multiple datasets, predictive models and reports/dashboards. Organizations shouldn't expect immediate returns, but stay invested over time to reap the benefits.

Step 2: Hire the right team
The best predictive models are developed by people who understand health data, IT systems like electronic health records (EHR) and practice management systems (PM), and can adapt to constantly changing data exchange standards such as HL7, SNOMED, etc. It is key to look for personnel or vendors who are qualified in data analysis and predictive modelling, as well as understand data safeguards required to work with health data. These vendors should have the necessary certifications to work with Personal Health Information (PHI) positive data. Aggregating disparate data sources can be a PHI violation if the data is not scrubbed and de-identified.

Step 3: Start small
For predictive analytics, start small with proof of concept projects. These initial projects allow clinicians and administrators to observe the theory getting converted into practice and see the value, which they can then use as a launching point to convince skeptics that analytics work.

Step 4: Keep it as simple as possible
Simplicity is key. Concentrate on models that allow you to identify the factors responsible for predictions. Keep it simple enough for a healthcare domain expert to ensure that the predictions are clinically accurate and robust.

Step 5: Avoid a one-size-fits-all approach
One solution or predictive model does not work for all scenarios, use cases and populations. Every organization is unique in the type of patient population it caters to, therapeutic focus areas, quality of care and work culture. Develop or test predictive models on your own data to understand its effectiveness and accuracy. It is likely that you will need to customize these tools for seamless integration and success within your own IT system.

Step 6: Think of the end-user
Predictive analytics can be a complex subject for people to understand. Organizations should make sure the results of a successful predictive model are easy to use for all stakeholders within and outside the organization. Decide the scope of the predictive model based on the end-user. While conceptualizing the development of a specific predictive model, think beyond data procurement and type of predictive model to operationalize it.

Step 7: Educate your staff
Administrators and data scientists have to spend time educating the staff about the merits and use of analytical tools as well as the role of good quality data to develop these tools. Listening, engaging and incorporating the feedback of end-users (i.e. physicians, nurses, aides, therapists, case manager, etc.) is crucial to the success and application of real-time analytics within a healthcare ecosystem.

Step 8: Collaborate
It is important to collaborate with key stakeholders and potential end-users to identify relevant use cases for predictive analytics. While developing predictive analytics for popular use cases (e.g., readmission management, claims denials, etc.), listen to your staff to understand the immediate analytics needs of your organization. Nurture a collaborative environment where clinicians and administrative staff can come up with novel use cases or problem statements for data analyst to solve. For example, case managers wanting to predict which patients will need home health services in the future, clinicians who want to study disease progression models that could incorporate socio-economic and environmental factors or administrators interested in simulating different scenarios for Case Mix Index (CMI) to compare provider performance and reimbursement.

Step 9: Build strong executive sponsorship
It almost goes without saying that hospital executives have to be involved and show support for analytics-based projects for the staff to follow suit. In such cases, it is better to adopt a top-down approach, even for small projects.

Step 10: Partner with organizations that have strong experience in healthcare analytics
A majority of healthcare experts, especially physicians and caregivers, are skeptical about the potential of analytics. The best approach right now is to partner with professionals or organizations that have credibility in developing and deploying analytical solutions within a healthcare ecosystem. Collaborations with a unique combination of healthcare domain expertise and advanced technology helps minimize risks, improve operational efficiency, clinical and patient outcomes and treatment efficacy.

Emma Mendonca is a Healthcare Consultant for CitiusTech Inc.

The views, opinions and positions expressed within these guest posts are those of the author alone and do not represent those of Becker's Hospital Review/Becker's Healthcare. The accuracy, completeness and validity of any statements made within this article are not guaranteed. We accept no liability for any errors, omissions or representations. The copyright of this content belongs to the author and any liability with regards to infringement of intellectual property rights remains with them.

Copyright © 2021 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.

 

Featured Whitepapers

Featured Webinars