The future of health care is analytics, but...
The output of predictive analytic models should be the inputs to informed conversations.
We cannot just look at data in hindsight
First, we cannot simply look backwards. Analytics can be an incredibly powerful tool in health care, in no small part because there is a massive amount of data to analyze. Clinical, utilization, socio-demographic, economic, and behavioral data elements all play a role in understanding and treating patients more effectively. It is not, however, sufficient to leverage these data sets for retrospective analysis alone. While data can uncover trends and patterns of what has happened in the past, it is both insufficient and inadequate to stop our analysis there, especially when the goals are increased quality and reduced costs.
Health care is now focused on the right treatment, for the right patient, at the right time, by the right provider, in the right location, yet older studies have highlighted the fact that patients receive the wrong diagnosis 10 percent to 15 percent of the time. Furthermore, when they do get the “right” care, they may be receiving treatment in a location or setting that is more expensive, but no better. Within my organization, we have identified five critical drivers of health care quality, which together represent the best path to success for both the health care system and the patients they treat. Diagnosis, doctor, treatment, hospital, and coping are interdependent drivers of health care quality, and each must be addressed to achieve the best outcome, regardless of the condition.
An abnormal mammogram, while an isolated piece of clinical data, can allow for connection with a patient before she receives an incorrect breast cancer diagnosis, sees a poor-quality provider, follows a poor treatment protocol, has surgery in a poor-quality hospital, or gets lost in the system without appropriate follow up. It is a moment in time, at the outset of a long health care journey, where key treatment decisions can be made in support of the highest quality care, at a reduced cost.
Similarly, lumbar disc disease and low back pain can lead to a reduced quality of life for many patients. Discectomies, laminectomies, and fusions are surgical options in some cases, but often there are alternative treatment pathways for patients with low back pain. Predictive analytic models, using diagnosis and procedure codes along with utilization data, can help brokers, employers and providers target high-risk patients among their populations. Early identification of patients who are progressing towards back surgery can allow for intervention with less-invasive measures such as weight loss, physical therapy, and epidural steroid injections. In many cases, these non-surgical approaches yield significant pain relief, and improvements in the patient’s quality of life, without surgery. In short, predictive analytics give us the chance to save lives and dollars. No wonder everyone in health care is so excited about them.
We are not quite there yet. We still have data issues. What are the appropriate data elements and sets? Where are they stored, and how are they shared? Standardized data sets, collection methods, and use are foundational to comparative analysis and impact studies. We are making progress in these areas, but still have a long way to go.
We can’t look at data alone – we have to look at the human
Second, and I would argue most importantly, we still wrestle with the balance between high tech data analysis and the old-school high-touch model of patient care. In our view, data should power analytic models, and these models should deliver insights. But above all else, the output of predictive analytic models should be the inputs to informed conversations and thoughtful treatment decisions.
Take, for example, social determinate data, the information which helps establish the context within which patients live and are treated. Clearly, this insight has become a focus of population health management initiatives in recent years, as it has become evident that socio-demographic, economic, and behavioral data elements weigh heavily upon an individual’s ability to get and stay healthy. The key, however, is understanding how this information is best communicated between patients and providers. Are predictive analytic models the best approach? Should we rely upon data and proxy variables to get to the details of our patient’s challenges, or should we speak with them, directly, about their circumstances?
I contend that patient context cannot be understood through automation or analytics. Predicting who might have transportation challenges, and thus be at increased risk for appointment “no-show,” may seem to make sense, until you compare it to simply asking the question. Furthermore, a conversation can uncover critical details about a patient’s challenge or risk. Why is transportation a problem for you? Is it due to the lack of a car and no accessible public transportation? Or is it due to a suspended license for multiple parking violations? The first challenge will persist, over significant time, while the second may be limited to a more finite number of days. The risk mitigation strategy is certainly different, in each example, so the detail here is critical. This is not a data problem, nor is it one to be solved via analytics. This insight is generated via a human connection.
In the end, we still need to speak with employees, and talk with our patients, to better understand what they face in their day-to-day, and to determine how those factors contribute to health and well-being. That, I believe, is what’s next in health care: a combination of predictive analytics and old fashioned, high touch, person-to-person support, which together assist patients in becoming more effective consumers of health care. Technology can help identify patients ripe for intervention, but the technology itself cannot solve for the needed improvements in care. Similarly, analytics inform, but they are neither compassionate, nor caring. Applied to their most logical extreme, they help providers help patients; they do not replace them. Patient preferences, anxieties, and fears come directly from people, and they must be understood by health care professionals.
Data is so powerful, and we have an amazing opportunity in health care to leverage it for good. Those who strike a balance between data insight and care delivery will benefit the most. Let’s identify the right data, share it consistently across health care, and most importantly, let’s drive improvements in how we engage with our patients along their health care journey.
