Patient Journey Mapping with Graph Databases for Powerful Clinical Insights

The goal of patient journey mapping is to improve the patient experience, optimize clinical operations, and eliminate patient care gaps, Davies (2022). Incorporating graph data science into patient journey mapping has been shown to surface more meaningful clinical insights, leading to a better understanding of a patient’s journey and outcomes. The potential value in patient journey mapping is finally obtainable at scale with graph databases.

What is Patient Journey Mapping?

Patient journey mapping can be done using many approaches, resulting in different formats relevant to their intended use case. For example, if the patient journey map intends to optimize clinical operations, then the output likely will be a map of patient interactions with a healthcare system. The goal in this context is to establish any gaps in the patient experience and the supporting clinical operations at each touchpoint. Figure 1 is an example of an interactive patient journey map created as part of a custom clinical application (see Streamlit tutorial for more) for patients and clinicians.

Patient journey maps, often referred to as patient experience maps or patient experience journey mapping, of this nature can also be the product of a clinical pathway. Clinical pathways establish the standard of care for a patient’s clinical presentation with a specific disease and are often linked to a companion patient journey map.

Clinical pathways, like the breast cancer patient journey map shown above, are in fact graphs that model a patient, their treatment, clinical decisions, and the clinical terrain they journey through. Together, the patient journey mapping process, and the creation of a clinical pathway, can be a powerful combination to capture a comprehensive view of the complex nature of a disease, its effect on a patient, and the actions of a healthcare system.

In this article, I will focus on how patient journey maps (or patient experience mapping) can surface meaningful clinical insights by modeling patient interactions with a healthcare system in a clinical knowledge graph. While clinical knowledge graphs can be developed to explore domains like target protein interaction exploration, new drug development, clinical trial administration, and other research and healthcare use cases, they can also explore a patient’s care with a healthcare system, their disease progression, and treatments. It also drives insights into improving the Triple Aim of healthcare- the improvement of patient experience, raising the health of populations and reducing health care cost.

Modeling Clinical Care in a Graph

Figure 2 establishes a digital twin of a single patient’s journey with their disease. This digital twin was established by loading the patient’s clinical data using graph etl (graph orchestration) into a clinical knowledge graph, and complimentary clinical analytic systems using a well-designed schema create a digital twin of a patient’s journey within a healthcare system.

Digital twins are an essential concept across many industries that need to create digital representations of their users, resources, and systems. Within the healthcare setting, a digital twin approach can help clinicians and analysts model past and current events and simulate how changes to operations, care, procedures, and programs may influence future outcomes. Digital twins traversing a patient journey map help to reveal real-world obstacles, gaps in care and processes, and other challenges faced by patients and their families.

To understand the clinical care a patient receives and explore concepts such as disease progression and cost, a graph schema is developed to support these explorations. Figure 3 demonstrates a graph schema for the journey seen in Figure 2. The schema models a patient (red node) with a temporal chain of clinical encounters as (the yellow nodes).

Each clinical encounter is connected to an insurance claim event and further linked to the claim’s procedures, diagnosis, drugs, imaging, clinical provider, and other relevant aspects of care to provide a comprehensive view of a patient’s journey through complex clinical terrain.

This type of graph is created using all available clinical data, which might provide insights into patient care, cost, and outcomes. It may not appear intuitive at the outset, but a graph like this can deliver unusually helpul and meaningful clinical inghts by leveraging the clinical data’s connections. For example, patients with a specific diagnosis tend to have the same viral infection earlier in their clinical history.

Procedures, imaging, and labs may be found in the Clinical Procedure Terminology (CPT) codes. At the same time, patient diagnoses are captured in the World Health Organization (WHO) International Classification of Diseases (ICD) codes which cover common and uniquely rare diagnoses.

ICD and CPT codes, as is frequently the case with other data sources when modeled in a graph database, provide an opportunity to reason semantically about patients and populations. ICD-10 codes for example are comprised of chapters, blocks, and sections. ICD-10-CM Chapter 2 concerns neoplasms, Section C15-C26 malignant neoplasms of digestive organs, and C16 the malignant neoplasm of the stomach.

Patients with various digestive cancers can be considered a clinical cohort of patients with malignant neoplasms of digestive organs, or at a higher semantic level, as patients with neoplasms. This is a powerful capability of linking data sources in a graph.

Another benefit of working with clinical data in a graph environment is the opportunity to link the CPT and ICD codes with insurance claims data, so healthcare organizations can explore optimizations to care delivery, mitigate cost, intervene early in a patient’s disease progression, identify risk, and many other aspects of health care through patient journey mapping and graph data analytics.

Clinical Insights from Patient Graphs

As an example of the meaningful insights that can be surfaced from a patient journey map and clinical knowledge graph, I present a brief exploration of disease progression seen in the clinical diagnosis captured in a patient’s ICD codes from several healthcare institutions’ electronic medical records systems.

The temporal progression of a cohort of patients who ultimately were diagnosed with the same disease can be seen in Figures 4 and 5. Figure 4 explores the ICD-10 codes and their distribution over one year for this cohort. Figure 5 examines the same data within a polar plot of the ICD-10 categories. Both approaches result in an understanding of how the disease evolves and what components of an illness and its accompanying sequel contribute to the patient’s disability as their diagnosis progresses. This is a crucial concept to better understand the disease and its evolution over time for these clusters of patients.

From this kind of analysis of the graph data, several clinically relevant insights and inference capabilities can be developed. For example, a risk stratification could be established based on the progression and frequency distribution of diagnosis. This risk category could then be used to create earlier interventions to a likely diagnosis that result in a reduced burden of disease, reduce cost and need for aggressive treatments of a later-stage disease, and mitigate the chronic effects of a disease caught at a later stage of progression. Risk stratification can also be used in partnerships with insurance programs to elevate services to populations at risk of developing a disease or chronic condition.

Healthcare systems can better use information revealed from hospital patient journey mapping (and other kinds of patient mapping) and graph analytics to design preventive care opportunities and supportive care programs. In the disease progression seen in the chronology above, a health analyst team could identify that, e.g., patients with this specific disease frequently develop dental infections and poor oral health. A preventative intervention of prescribing a dental consult and supportive education regarding oral care could mitigate this associated health issue, improve the patient’s disease journey, reduce overall cost, and improve the patient’s overall outcomes.

Wrapping up our Journey

In this article, we explored the benefits of patient journey mapping, and we provided the beginnings of what can become your patient journey mapping template leveraging graph databases. We discussed several types of journey maps and focused on how they can be used to surface clinical insights from a wide variety of clinical data.

Further, we examined the unusually impactful role a graph database can play in modeling clinical data and in generating a digital twin of patients. This enables us to both develop novel insights as well as to better understand the touchpoints in a patient’s clinical journey with a disease, across a complex healthcare terrain.

Let us know if you want to further explore patient journey mapping healthcare improvements, graph analytics, or other healthcare topics. Together, we can start our journey of clinical data discovery towards improving the patient experience, advancing population health, and mitigating costs in your healthcare domain.

Read here for more specifics on clinical data trial analytics, and follow these links on the related clinical trials Streamlit application (Streamlit tutorial), and here for the article on graph etl / Neo4j etl with a focus on clinical trial data, and here for clinical trial data quality, as well as this on graph analytics.

Check out a powerful new natural language interface to your graph database / knowledge graph that enables your non-technical users to ask natural language questions right to the database itself (meet Sherlock™ from Graphable).

More Reading

• AI in Drug Discovery – Harnessing the Power of LLMs
• Large Language Models (LLMs) for Enterprises: A Comprehensive Guide to Navigating the New AI Frontier
• What is Prompt Engineering? 5 Ways to Unlock the Potential of Large Language Models
• Natural Language Interface for Knowledge Graphs
• What is ChatGPT? A Complete Explanation
• Resolution for 3 Errors: ‘ChatGPT Internal Server Error’, ‘ChatGPT is at Capacity Right Now’, ‘ChatGPT Network Error on Long Responses’
• Domo custom apps using Domo DDX Bricks with the assistance of ChatPGT
• Understanding Large Language Models (LLMs)