Healthcare Analytics: Predicting patient experience with narrative data: a healthcare goldmine
By Sagar Anisingaraju and Mo Kaushal (l-r)
The healthcare industry in the United States is going through a major transformation and is becoming much more consumer friendly. This process of “consumerization” will drive healthcare executives to pay close attention to what patients feel and think about healthcare organizations and services – something the industry has ignored for decades. This trend should continue and be accelerated as we all start paying more out of pocket for healthcare services.
Going forward, the success of healthcare organizations will depend on internalizing patient sentiment insights into everyday decision-making, just as it is for nearly every other industry. By its recent introduction of the “Health Star Rating” system  for hospitals, the Centers for Medicare and Medicaid Services (CMS) sent a strong and clear signal about the huge transformation of the patient experience and associated ratings, and its impact on the healthcare industry – an impact the industry can no longer ignore.
Patient Experience: An Overview
Patients today provide valuable information about their personal experiences with hospitals, clinics, doctors and nursing staff. Patients share their experiences in a variety of mediums, e.g., social media, patient discussion forums and surveys. In addition, key insights are gathered when patients engage with care delivery organizations via call centers and interactive in-room TV apps. When we bring all of these different data sets together along with key hospital objective metrics and CMS data, the insights derived can have a huge impact on the decisions that today’s healthcare executives have to make. These “narrative data assets” provide new perspectives and complete story lines never before available to the healthcare industry.
This article explains new methods on how providers can implement programs using big data and predictive analytics to improve patient experiences while improving providers’ survey scores. Based on work done with a number of leading healthcare providers over the last three years, we will present examples on how insights can be generated and used to predict specific improvements around reduction of re-admission rates, improving Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) scores and reducing patient churn. When patient experiences are heard, understood and acted upon in a holistic way, healthcare organizations can dramatically improve their top and bottom lines while addressing quality of care.
Healthcare Data Landscape
The data landscape – a key to understanding patient behaviors – has dramatically changed over the last few years. As shown in Figure 1, narrative data – ready for processing – is now available from patients before, during and after an encounter.
Figure 1: Narrative data is available from patients before, during and after an encounter.
The medium through which patients are communicating during various touch points has to be understood holistically in order to generate insights to act upon. A holistic program gathers patient data from hospital systems as well as third-party surveys and social websites. Using the new data landscape that is available, a set of patient experience indicators can be generated as shown in Figure 2. Figure 3 shows an example of some of the broader data sources, in terms of both variety and volume, that would enable comprehensive insight into patients when looked at holistically.
Figure 2: A set of patient experience indicators can be generated from today’s data landscape.
Figure 3: Examples of broader data sources that could enable comprehensive insight into patients.
Insights from Data
Once the new data landscape explained above is accessed and understood, we have a wealth of information available to analyze. For example, the narratives from social media sites and surveys can be analyzed using Bayesian interference  to find out emotional attachment of patients to a hospital’s brand and services. A typical patient encounter may leave the patient in an overall satisfied state, but they might have felt “anger” at the billing errors. Advanced data science algorithms can analyze this range of emotions.
Similarly, the key topics that patients are discussing about the hospital and its services can be honed in using data science methods. A standard rating algorithm can be built and applied to rate the overall satisfaction of patients across the mediums used to communicate.
It is important for the providers to understand the key topics of interest among their patient segments. Are the patients thinking about ease of appointments, facilities at the hospital, parking lots or about billing errors and nurses’ communication? Which of these topics have negative connotation and which ones have positive inclination?
Advanced analytics, driven by data science, can now answer these questions with reasonable certainty. Identifying the trending of “hot” topics over time, and associating a rating to these topics, will enable executives to make key decisions within their care-delivery organizations. Hot “green” (positive sentiment) topics could be a key tool for marketing folks, while a hot “red” (negative sentiment) topic could trigger a possible intervention program. Understanding these topics in terms of the same Health Star Rating system being promoted by the CMS makes it easier to analyze and leverage such insights.
Hospitals are constantly looking at continuous improvement to enhance quality of care. Patient sentiment input is crucial to understanding the overall key performance indicators (KPI) that a hospital is tracking. Mapping from key topics of interest for patient segments to the KPIs is needed to understand the value of patient sentiment in proper context.
For example, we can look at patient churn as a KPI and analyze the impact of sentiment on potential patient churn. While this is a fairly standard practice across other industries, the healthcare industry has largely ignored it. Data platforms with correct, proactively linked intervention programs have the potential to reduce patient churn while improving quality of care, outcomes and revenue.
The current survey-based approaches used by hospitals provide a certain level of insights but have very long cycle time for actions. They typically take about eight to 12 weeks to send back the insights, and even then they fail to paint the complete picture. The approach explained here is more comprehensive and faster, thus enabling hospitals to implement intervention programs when they are needed. Furthermore, this approach can also be used to monitor and track programs initiated to impact certain KPIs.
Insights to Intervention Programs
We have seen some methods that hospitals can use to track the leading indicators such as patient sentiment from narrative data assets and analyzing their impact on few of the key performance indicators. As noted, the approach presented here is more comprehensive and faster, thus enabling hospitals to generate insights from patient experiences across touch-points.
The real value of the insights from these narrative data assets is when hospitals use them to define intervention programs, thus potentially generating a desired state of outcomes. Using data science methods, a model can be built that can analyze the impact of changing the set of forcing variables on the outcome of a selected target value. For example, a key set of variables that are forcing a patient to switch hospitals can be identified from the test data, and the model can predict the outcome of patient switch based on changing values.
Depending on the hospital, these influencing variables can be “admission process,” “billing errors,” “staff responsiveness,” etc. The model can describe the most probable variables that are causing the churn based on the data. Hospital management can change the values of these variables to see the impact of the potential churn, and a desired state can be reached with a few predictive iterations.
Once a desired state is achieved, the values of influencing variables and the sum of budget requirements to arrive at those values will then become the starting point to define an intervention program that the hospital can plan and execute. Further, the advanced machine-learning capabilities available today also enable intervention programs to be self-learning. For example, it is feasible for a business unit to define an aspirational state with respect to their patient population and let the system predict a best set of operating parameters to achieve those objectives within time and budget constraints .
As healthcare organizations brace themselves for the major consumerization wave, they can feel confident and comforted knowing that technology solutions have evolved and are available to guide and assist them in this incredible transformation of the data landscape.
Sagar Anisingaraju is the chief strategy officer at Saama Technologies, a big data solutions and services company headquartered in the Silicon Valley. A recipient of the Chief Strategy Officer of the Year award from Innovation Enterprise, Anisingaraju founded and ran InfoSTEP Inc. for 11 years as CEO until its acquisition by Saama.
Mo Kaushal is a partner at Aberdare Ventures, a venture capital firm based in San Francisco and focused on innovative healthcare technology start-ups. Previously, he was the director of Connected Health with the Federal Communications Commission, where he established the agency’s first dedicated healthcare team. During his time in the Obama administration, he was also a member of the White House IT task force, a cross-agency team focused on implementing the technology aspects of the Affordable Care Act. He is also a consulting associate professor at Stanford University.
- Provost, F. and Fawcett, T., 2013, “Data Science for Business,” O’Reilly Media.
- Kass, G.V., 1980, “An exploratory technique for investigating large quantities of categorical data,” Applied Statistics, Vol. 29, No. 2, pp. 119-127.
- 55Following the 2016 presidential election, the first two months of the new year were quite tumultuous for the U.S. healthcare system. Interestingly, the Affordable Care Act (ACA) has survived so far, albeit the process to repeal the law has begun.
- 51The healthcare analytics industry is making great strides. As part of my work I talk to many data analytics companies who report they are very busy with implementation projects. One large company told me that its implementation staffs are booked until end of the first quarter of 2017. This is…
- 50The 2016 election is a watershed moment for the U.S. healthcare industry. Any presidential election and change of guards come with changes in policies. It happened in 2008 when President Obama was sworn into the office. That led to the establishment of the Affordable Care Act (ACA) or Obamacare. To…
- 49March/April 2015 Welcome to the ‘We Economy’: Finding the business in your data By Arnab Chakraborty, Michael Svilar and Prith Banerjee (l-r) The Internet of Things (IoT) and the data it is producing are becoming main catalysts for change and transformation for businesses and consumers alike. Today, with every object,…