Healthcare is undergoing a revolutionary transformation driven by data, analytics, artificial intelligence, machine learning and engagement platforms.
Population Health seeks to identify and engage with at-risk patient populations to improve outcomes and reduce cost. Population health initiatives benefit patients, healthcare organizations, and the public.
When overall population health improves, that means a higher quality of life for patients and their families as well as lower costs across the community. Additionally, healthcare organizations benefit by meeting key quality metrics such as preventing readmission or supporting healthier populations.
The input side of population health initiatives involves data collection and aggregation. Integrating, de-duplicating, normalizing and standardizing data from both internal and external data sources is critical to implementing an effective population health solution. This is among the biggest of all big data challenges in healthcare.
The broader the swath of data and information inputs, the more meaningful the risk analytics. However, big data also needs to be meaningful and actionable.
After examining population health inputs, we’ll review the output side and where CRM and marketing technology fits in. We’ll conclude by asking what can be learned from Amazon Prime membership.
Population health input categories
At a macro level, patient outcomes are a result of, in descending statistical order:
- Individual behaviors
- Social determinants of health and environmental factors
- Clinical care
In other words, a fully informed understanding of a patient goes well beyond analyzing clinical data. To the extent that inputs can be gathered from as many disparate sources as possible, risk categories and strata can be more precisely defined.
Population health input details
Within each of the above categories are subcategories of potential data points. Here are several of them.
1. Individual behaviors
This is potentially the largest bucket of inputs.
While there are privacy issues associated with automatically collecting data about many types of individual behaviors, there can be voluntary data provided through self-reporting, including data from non-clinical wearables such as smartwatches.
Self-reported family medical history and genomic testing can help identify latent conditions that have not yet presented any symptoms. This category is evolving rapidly but undoubtedly will play a significant future role in population health.
3. Social determinants of health and environmental data points
In some patient populations, the zip or postal code can be the number one risk factor for a given condition. Identifying social and environmental variables and marrying them with clinical data can yield very valuable insights into patient health. Some examples of critical social and environmental data:
- Income and employment status
- Housing, community and transportation access
- Literacy and language
- Social integration and support
- Personal safety
4. Clinical data sources
While clinical data is generally the most accessible, there remain many collection and consolidation challenges. Also, remember that this data represents only a fraction of the population health picture. Clinical data sources can include:
- Electronic medical records
- Health information exchanges
- Pharmacy data
- Lab data
- Physician interactions
5. Augmented and Inferred Data
Beyond collecting data to use for population health risk stratification, augmenting data with public and private data repositories can be used to greatly enhance the data on hand. For example, using weather data from publicly available data sources can identify key periods when low income or elderly patients may be at risk.
With all of this data available, adding machine learning and artificial intelligence can yield new relationships in the data that otherwise might have been missed with human review.
Population health outputs
As consumers, we’ve come to expect an Amazon-like level of personalized engagement. Individual behavior is, in turn, the area that can be most easily influenced by proper outputs—including relevant, personalized and targeted communications.
A variety of data-driven outreach media efforts can encourage better individual behaviors. These media include:
- Text messages
- Patient portal content and messaging
- Outbound phone calls
- Traditional mail
- In person
This is where Salesforce Health Cloud and Salesforce Marketing Cloud can elevate the patient engagement to a whole new level. By personalizing outreach in a way that is relevant and timely—and in the modality of choice—the patient is much more likely to be engaged and informed with their health care.
Based on better statistical knowledge about a patient, that patient can receive communications that are highly relevant to them. Often, in the absence of an engagement engine and the data to power the engine, communications lack focus and fail to engage the target audience. Newsletters and broadcast emails may cast a wide net, but they are unlikely to generate interest or change behavior.
Marketers divide a population into what they refer to as segments. In population health, segments are further classified into specific registries. Customer journeys—in this case patient or member journeys—are tied to segments and registries. By identifying the potential conditions, the level of risk and the best way to engage a given patient, these groupings can greatly elevate engagement.
Individual patients can take forks along the journey’s road—and they do. Each patient’s journey will be influenced by data that continues to be collected. The level of engagement is also a data point which can be used to further classify the patient. Depending on the responses of patients to specific digital and non-digital outreach efforts, outputs can lead to additional inputs. It’s a branching feedback loop.
For example, is a patient engaging with content around nutritional recommendations? This can be measured by whether they are clicking through to healthy eating articles and how much time they are spending on website pages. Are they following calls to action on the pages they land on? Are they responding to questionnaires? Are they engaging with information about the benefits of medication adherence? Are they even opening emails?
If certain patients aren’t opening emails and the patient journey therefore suggests outreach by phone, are they answering the phone? If they do answer the phone, what information are they providing and what actions are they committing to? For example, “Yes, I will take that medication as prescribed.” “Yes, I will go over to the clinic for my fasting blood test.”
The feedback loop can move a patient up, down and across the risk strata.
Moving toward “a segment of one”
Amazon Prime has over one hundred million members. Amazon is able to treat each member of a large population as a segment of one because the company has the advantage of perfect information about past member buying behavior on its platform. The recommendations a member sees when they log into Amazon.com are unique to them.
The output side of population health initiatives inhabits the middle ground between presenting every patient with the same content and presenting each patient with unique advice, ideas, recommendations and calls to action.
Over time, population health initiatives will result in smaller and smaller patient segments. This means more meaningful and therefore more impactful information can be communicated to each patient.
Amazon’s segmentation and associated content gets people to buy more things. More granular population health segmentation may get people to live healthier lives. This, in turn, will result in healthier communities and better outcomes for patients and their families.