14 November 2019

Creating a New Community Integrated Health System – Role of the Traditional Health Provider




By Leslie Wainwright, PhD, Chief Funding and Innovation Officer

Addressing the social determinants of health (SDoH) in communities is a hot topic of conversation in healthcare. The industry has bought into the theory that 20 percent of an individual’s health is determined by clinical care and the rest by social, economic, genetic and behavioral factors. But perhaps more importantly health systems need to recognize that they can’t solve this issue on their own.

From my perspective at PCCI, I’ve seen an increase in value-based contracting models in recent years, and health systems and physicians are looking beyond the four walls of their institutions to build relationships with outpatient, behavioral health, post-acute care, and now non-medical providers. The number and types of collaboratives between health systems and non-traditional providers has been growing over the past several years with a recent report gathering information on over 200 different partnerships between hospital and community-based organizations across the country.

But while health systems may be embracing community provider relationships, I believe that sustainable success in addressing social determinants of health requires a fundamental shift in the way health systems view their role in improving the health of their communities.

Over the past ten to fifteen years there has been an evolution in how health systems have approached improving health outcomes. Initially health systems focused on providing high-tech solutions for care delivery such as robotic surgery, and advanced imaging techniques. Then to meet the need for increased access and demand for outpatient services, health systems seeded service areas with ambulatory surgery centers, urgent care, retail clinics, and physician offices.

In each of these evolutions the strategies centered on a solution created by the health system alone. And one could argue that the main beneficiaries of these investments were often the health systems themselves – increased market share, improved reimbursements. But such a self-centered approach will not work when addressing social determinants where the root causes lie outside the four walls of the health system.

Effectively creating a System of Community will require a collaborative mentality from health systems. While they may have power and influence to gather partners to the table, execution of successful interventions lies with social services and community-based organizations that are the experts in understanding and helping individuals address social needs.  Even if not leading, health systems should still be active participants in this work. Indeed, there are areas where their contributions to the organization of partners is critical:

  • Community Health Needs Assessment

CHNAs, which all health systems are required to complete, can be a starting point for developing strategies to address social determinants of health by quantitatively and qualitatively identifying the needs of the local community. To supplement the CHNA, additional SDoH data should be incorporated to help identify needs at the block level which can help pinpoint exactly where an intervention will likely make the most impact.  These enhanced data should map and evaluate SDoH needs at the block, not zip code level, and should be supplemented with qualitative surveys to understand capacity for self-care, isolation, and learned helplessness across individuals and community.

  • Governance Structure

At the core of any collaborative with community partners should be a formal governance structure that defines the policies and documentation that will enable partners to execute and measure success of their strategic interventions. A formal governance structure can also ensure that all partners have a voice at the table and may help to mitigate any fears that community organizations have that the health system is in control of the initiative.

  • Legal Structure and Data Sharing

Now more than ever, technology, and indeed cloud technology, can connect disparate partners across multiple settings to exchange, share and report on data about the same community members. But there are significant legal and compliance requirements involved in sharing data across entities. Health systems have the expertise to ensure that policies around data sharing are in accordance with Medicare and HIPAA regulations. Health system experts in data privacy and security can provide advice and support to community-based providers in developing policies and procedures required to share data securely.

Improving patient engagement is at the top of the list of priorities for most health systems. The only way that a health system can achieve this is by creating strategies that start and end with the needs of the community. To find success in addressing social determinants of health, health systems will need to cede control and the notion that they need to create, lead and execute the strategy alone.

SPOTLIGHT ON PATIENT CARE: DALLAS CONNECTED COMMUNITIES OF CARE

The Connected Communities of Care (CCC) platform was first implemented in Dallas in 2014 and serves as a comprehensive foundation for partnership by leveraging a web-based information exchange/case management software platform providing seamless connection and coordination between healthcare providers and a wide array of community-based social service organizations.

Since its inception, more than one million services have been documented and more than 215,000 unique individuals who have been impacted by a network of six health care systems and over 100 community-based organizations. The novel approach to addressing SDoH and organizing cross sector information sharing through sophisticated connections has garnered national recognition and has made a lasting impact in Dallas.

Not only has this collaboration connected existing organizations in a new system of community health, it has also changed the way health systems define competitors vs. collaborators. Local health systems that may have viewed each other as competitors for services, have recognized that prioritizing the needs of the community through collaboration makes a stronger impact than any isolated intervention.

 

About Leslie Wainwright

Leslie Wainwright, PhD., is the Chief Funding and Innovation Officer at Parkland Center for Clinical Innovation (PCCI). She is passionate about entrepreneurship and innovation, and has experience that spans academic research, pharma/biotechnology and healthcare delivery.

Dr. Wainwright would like to thank Natasha Goburdhun from NDGB Advisors who contributed to this post.

 

28 October 2019

SDoH: From Theory to Action – Making Social Determinants of Health a Reality




By Steve Miff, PhD, President & CEO of PCCI

The term “social determinants of health” is far more than a trendy new buzzword in health care. Serving the physical, mental and social needs of the community is not just the right thing to do but can mean substantial improvement in care and reduction in unnecessary healthcare costs.

Several studies have shown that addressing social needs, such as food or housing insecurity, can have a significant impact on a person’s healthcare outcomes and costs. Individuals experiencing housing insecurity or homelessness have higher rates of chronic diseases such as high blood pressure, heart disease, diabetes, asthma, chronic bronchitis, and HIV.  This in turn leads to higher utilization of healthcare services such as emergency room visits, inpatient hospitalization and longer lengths of stay compared to those individuals with secure housing. Similar results are seen in those experiencing food insecurity.

Hospitals often state that part of their mission is to provide high quality care and improve the community’s health, or community benefit. A recent study of hospital mission statements in three states (Ohio, Florida and Texas) found that while quality was cited most often (65%), the second most frequently used term was community benefit (24%).[1]  If community benefit or community health is part of your health system’s mission statement, how much are you really doing to address the whole health of a community vs. just addressing their “sickness” needs?

At PCCI, our combination of data scientists and expert clinicians believe that health systems have an obligation to address social determinants of health to ultimately remove the disparities and inequality that we see in our community’s health. Yet this is tricky because success requires outreach skills, community relationships and data insights that extend beyond the traditional promise of health-related services. That said, there are three key elements that can assist health systems in making an investment in social determinants of health a reality. In order to move from theory to action, my suggestion is that health systems do the following:

1. Leverage the board’s community presence to align on areas of greatest need

As part of health system leadership, board members ensure alignment between mission and a defined SDoH strategy at all levels of the organization. As community representatives themselves, board members can also create the momentum and connections that health systems need to bring community and business partners together to create a governance structure for launching a connected community of care.  Such governance structure will guide the strategy, legal and policy needs, and the investment and execution of a connected and aligned SDoH strategy.

2. Invest in long-term partnerships to ensure sustainability

Recognize that as health systems, you alone cannot solve for social determinants. To truly meet the social, behavioral and emotional needs of some of the most vulnerable individuals in your community, you need to identify community partners with expertise in these areas. With the assistance of board members, assemble a partnership collaborative, with a formal governance structure, to build community-based strategies around SDoH needs. Support the sustainability of this collaborative with technology and data science techniques to identify specific root causes of social need in target populations, share data, and measure impact of interventions.  Identify an independent partner to evaluate the effectively of the SDoH initiatives and measure the cost, savings and impact across the community and for the health system.

3. Develop your own financial models that demonstrate the impact of SDoH

Between 2000 and 2017, hospitals and health systems across the country spent $620 billion in uncompensated care. We propose that health systems create an internal “at-risk” ACO-like model for their uninsured population and invested just five to ten percent of their annual uncompensated care dollars in developing community engagement programs to address social determinants of health.  These systems would see a three to four-fold return by addressing upstream, root causes in the community and increasing preventive, social and emotional support services to individuals in the community.

SPOTLIGHT ON PATIENT CARE

Texas Health Resources (THR), a 29-hospital faith-based non-profit health system based in Arlington, Texas has supported their mission “To improve the health of the people in the communities we serve” by creating a ten-year strategic plan to move from a hospital- to a patient-centric to a population health-focused organization. The THR board of directors has been an integral part of overseeing every step of the strategy to ensure that there is measurable and sustainable improvement in their community’s health.

Data and information gathered from regular Community Health Needs Assessments, has led to the creation of more than 200 non-profit partners across the region, including formal agreements with the American Cancer Society, American Diabetes Association and the March of Dimes, to increase health and well-being through programs focused on behavioral health, chronic disease management, child automobile safety, healthy eating, and provision of low-cost mammograms.

Board committees regularly monitor progress toward strategic goals and receive input from local community health councils and entity boards. But this commitment to community health is not just at the leadership level, employees of THR can spend between 8-12 hours of paid time annually to volunteer at local or THR sponsored community organizations to support community health efforts.

Most recently, THR announced a new initiative called Texas Health Community Impact which employed a data-driven, outcome-focused approach to identify areas of need in their communities. Mental health was indicated as a priority through the community health assessment. Zip code level data analysis and qualitative research helped them identify specific areas where seniors and youth lacked access to food and were also isolated, which led to depression and physical problems. As a result of this work, THR will distribute $5.2 million in grants to twelve agencies that will focus on interventions for these issues.

These grants only represent a portion of the financial investment that THR has made to its communities. In 2017 they provided $362.5 million in charity care, $31 million to community benefit programs and in volunteer hours, and $456.6 million in unreimbursed Medicare services.[2]

[1] Cronin, CE, Bolon, DS. Comparing Hospital Mission Statement Content in a Changing Healthcare Field. Hosp Top. 2018 Jan-Mar;96(1):28-34.

[2] https://www.texashealth.org/news/system-awards-5.2-million-in-community-impact-grants

 

About Steve Miff

Steve Miff, PhD., is the President and CEO of Parkland Center for Clinical Innovation (PCCI). He is a seasoned executive with more than 20 years of experience in healthcare analytics and consulting. He has served in various leadership positions in technology/consulting start-ups and on multiple boards. Dr. Miff is a recognized national thought leader with over 100 peered-reviewed and independent thought leadership publications.

Dr. Miff would like to thank Natasha Goburdhun from NDGB Advisors who contributed to this post.

 

 

 

 

23 October 2019

Vikas Chowdhry Presents at the Root Cause Coalition’s 2019 National Summit on the Social Determinants of Health




PCCI’s Chief Analytics and Information Officer, Vikas Chowdhry, participated in a panel at the Root Cause Coalition’s National Summit on the Social Determinants of Health in San Diego. The panel, “Making the Value – Equity Connection,” was moderated by Priya Bathija, Vice President, The Value Initiative. The panel included healthcare leaders from across the country discussing how each address some of the social root causes of inequity and adverse outcomes in healthcare.

Vikas presented on the Breast Cancer Equity initiative at Parkland – a program that with a goal of reducing late breast cancer diagnosis in the most under-served communities in Dallas. Click on the image to see Vikas’ full presentation:

 

 

17 July 2019

Reducing “Misbehaving” In Healthcare Operations Through Data and Optimal System Design




By Manjula Julka, MD, MBA and Albert Karam, MS

Every year, people throw away millions of dollars when they decide to fill up their car tanks with more expensive premium gas when regular unleaded will do just fine for their cars.

There are several reasons for that kind of sub-optimal behavior. Nobel Laureate and University of Chicago professor Richard Thaler calls it — “misbehavior”. Thaler, in his book “says that the optimization problems that ordinary people confront are often too hard for them to solve, or even come close to solving. Thaler’s two friends and mentors, Amos Tversky and Daniel Kahneman (himself a Noble Laureate) have illuminated several pathways on how we make decisions. Kahneman’s book “” articulates some of them and one of their decision theories may have applicability here. They say that when people make decisions, they do not seek to maximize utility. They seek to minimize regret. So, in this case, among other things, perhaps people are thinking that “better not regret causing any damage to the car for a few pennies”, not realizing that those pennies add up and that there’s no damage being caused.

Regardless of the root cause though, one of the ways to minimize these behaviors is to use data to educate and frame choices (the famous “nudge”) to make the optimal decision the easier one to make (through appropriate defaults etc.). And we see these kinds of suboptimal behaviors play out in every walk of our lives and healthcare is no exception.

At PCCI, we recently had an opportunity to work with a group of passionate clinicians at Parkland Health and Hospital System regarding a very similar issue. Magnesium is a key mineral for body functionality especially for heart, nerve, muscle and protein synthesis. Monitored in most hospitalized patients, it is often replenished to maintain normal levels. With very few exceptions, oral Magnesium is as effective as intravenous (IV) Magnesium medication with the added value of being significantly less expensive and more comfortable for patient (think premier gas versus regular unleaded). However, for a variety of reasons, the primary route of ordering Mg was through IV. To understand the magnitude of the problem at hand (and potential savings), we used Parkland’s EHR system (Epic) to identify instances where oral Mg could be as effective as IV Mg and realized that simply by changing the route for appropriate patients, the system could save hundreds of thousands of dollars. This analysis led to system-wide effort to provide informational messages to clinicians at point of care in the ordering process via the EHR so that they could make a more informed choice.

This initiative is a great example of how innovation, changes in behavior and optimal choices happen at the intersection of analytics, data and human behavior and psychology. Every care team member wants to provide the best care for patients, but sometimes the cumulative impact of individual decisions is lost. The conversion of one single IV order at a time to oral magnesium multiplied across many clinicians is now saving thousands of dollars to the hospital system while improving evidence-based care.

For additional technical details, please see 

12 June 2019

The Time Is Now for Health Systems to Get Serious About Social Determinants of Health




A fundamental question continuing to face many health system executives is: How do we comprehensively address the needs of patients when those needs extend beyond the boundaries of traditional clinical care?  As President and CEO of PCCI, we have been focusing on this very challenge since 2012.  And while there has been much talk and excitement about social determinants of health (SDOH), we believe that ~90% of the health system market still does not leverage social/economic information when designing population health programs, developing patient-specific treatment plans, locating new services, or conducting community needs assessments. But before health system executives can design an effective SDOH strategy for their organizations, they must first assess where they are and where they would like to be based on the insights and advantages a progressive SDOH strategy would offer. PCCI’s Social Determinants of Health Maturity Model can help executives take this critical step.

Social Determinants of Health Maturity Model

Level Zero: Incomplete Picture of an Individual’s Environment

Realistically, this is the starting baseline for most organizations. Often, teams will attempt to use clinical and claims data ALONE as a means to segment patient populations and project the impact on a patient or cohort. This rarely works; rather, it often leads to late treatment in acute environments, sub-optimal interventions, and erroneous insights about specific patients, patient populations, or geographic markets.

Level 1: High-Level View of SDOH, Using Specific Social and Economic Indicators as Proxies

Teams can extract basic information from claims or clinical data that could serve as effective SDOH proxies. An example would be to look at the number of changes in addresses in a specific record, over a 12-month period, as a strong indicator of housing instability.  At the highest level, teasing out information from existing records can begin to illuminate some of the critical social and economic challenges that may present for individuals in a given community. This level of insight also allows health- system teams to test basic assumptions about a market. We’ve seen teams fooled when the employment level appears to be relatively stable, only to subsequently discover that much of the employment is via low-wage jobs with very poor benefits.  If you begin to see that people are moving around even though the employment statistic looks stable, you begin to realize that the actual stability of your community might not be what you perceived it to be.

If at Level 1, Leadership Teams Should Be:

  • Developing high-level proxy indicators to reflect underlying social and economic challenges that could play a significant role in health status or the ability to access services.
  • Understanding the payer mix; who you serve and, even within the insured population, understand the wage/income levels because there is a high percentage of employed, low-wage individuals that have vulnerabilities associated with transportation, housing, affordable daycare, etc.
  • Becoming familiar with existing local or state connected communities of care programs or activities aligning providers and community-based organizations, such as food pantries, to streamline assistance efforts, reduce repeat crises and emergency funding requests, help address disparities of care, and improve the health, safety, and well-being of residents.

Level 2: Root Causes Understanding of Poor Outcomes at the Population Level

The rubber hits the road at level 2 and teams begin leveraging local data that directly reflect variation in social determinants. We believe that to understand root causes and build actionable models for patient engagement and support, you must evaluate data at the block level. Zip-code level aggregation often masks important details. This is particularly true in highly populated municipalities that can see a tremendous amount of social determinant variation within a 0.1 mile distance. For example, if I had block-level information providing insight that a six-block neighborhood within my market was having transportation-oriented issues and concentrated pockets of non-violent crime, I would model these insights into the deployment of my mobile diagnostic clinics or my development of innovative models to improve access.  Also, if I was discharging a patient who resided in that neighborhood, I would rethink how to schedule follow-up appointments, since the chances of the patient keeping the visits are extremely low. This level of insight and actionability would be missed at the zip-code level.

In collaboration with DFWHC Foundation, Community Council of Greater Dallas, and the University of Texas at Dallas, PCCI built a platform for Dallas called Dallas Community Data for Action and/or Community Data Insights [CDI].  CDI ingests and organizes multiple, publicly available data inputs, such as housing, education, food availability, and 911 and 311 data to generate real-time, actionable dashboards containing over 60 factors that all point to specific social determinants.  In Dallas, use of this data has been vital in understanding pockets of need and in locating areas where the impact of interventions can be the most profound.  You can also use this data more broadly to generate support to build community cross-sector collaboration, by enabling health systems to effectively  engage and coordinate with local municipality officials on community-based support services and planning, and also by helping philanthropic organizations to better understand (and track) community needs in order to invest in/prioritize funding areas that will produce the greatest impact.  In addition to having a detailed and dynamic picture of social and economic needs (demand for services), the CDI dashboard can quickly map out where support services are available/delivered and map/model the interdependencies and concentration of chronic health conditions with social support needs.  As this model is rapidly scalable, PCCI is already working with others across the country.

If at Level 2, Leadership Teams Should Be:

  • Integrating SDOH market insights into your strategic planning process and your community engagement plan
    • Use block-level SDOH in community needs assessments
  • Anticipating and predicting the correlations between multiple social and economic factors to inform your patient flow and access strategy (including your telehealth strategy). Start conducting trend analyses to anticipate and forecast the changes in local-market dynamics that will impact utilization, payer mix, and social/economic barriers to health.
  • Crafting a data-driven engagement plan to align more directly with local municipalities and local philanthropic organizations.

Level 3: Comprehensive Partnership Between a Community’s Clinical and Social Sectors

Participating organizations across a community are collaborating on one Information Exchange Platform and are connected through an innovative closed-loop referral system allowing them to communicate and share information with each other. Success at this highest level requires both a strong technology infrastructure and consistent programmatic deployment [at scale] across a community. This is what we’ve done in Dallas with our technology partners at Pieces Technology Inc.; effectively managing the right balance of people, processes, and technology has allowed us to achieve the positive results that we’ve seen.

Level 3 means a significant investment and a multi-year commitment, not only by the anchoring healthcare system or systems, but also by the local community.  It requires an initial investment and a robust sustainability plan that can ensure that the platform capabilities evolve with the changing needs of the community.  Deployment requires not only new technology, but an engaged local governance structure, new legal and data sharing agreements, and further refinement of data integration and advanced analytics at the individual level.  Integrating these into new/updated clinical and community workflows enables teams to proactively predict specific health and social/economic needs, the complexity and co-dependency of needs, and the ability to act real time at the point of care to address these needs.  This can facilitate making real-time referrals for community support services, tracking whether individuals accessed suggested medical or community resources (and what specific services were provided), and measuring and tracking the impact to individual/community resiliency, self-sustainability, health outcomes, and cost.  In Dallas, we’ve also started to leverage advanced data algorithms to risk-stratify individuals based on their health and social/economic needs to better prioritize and tailor resources and to proactively target high-risk individuals for engagement and follow-up via digital technology.

At levels 2 and 3, a health system must also think about how to leverage its foundation resources and internal employee community-outreach volunteer programs.  Once you better understand the patients that you’re serving in your market and the community-based services they access, you can better deploy employee-based efforts and philanthropic activities that align with the strategic efforts and provide maximal impact.

If at Level 3, Leadership Teams Should Be:

  • Crafting the information exchange platform governance infrastructure to delineate key roles, essential participants, and shared objectives.
  • Committing to cross-community collaboration [potentially including competitors] and a long-term effort; recognizing that your health system might be an anchor organization, but it cannot independently solve the entire problem.
  • Selecting and deploying the technology infrastructure [Pieces Iris™, TAVHealth, Unite Us, etc.] to enable cross-community engagement.  Develop updated clinical and community-based workflows.

In summary, if you’re just starting to address SDOH, you’re late.  It is critical for health systems to begin their SDOH journey today, especially if you serve a vulnerable population and/or operate in a market dominated by uninsured and Medicaid patients.  Addressing SDOH is also equally important for organizations managing a lower-wage, commercially insured population and for any health system that is actively managing or considering taking on risk-based contracts.

If you’re well on your way up the SDOH curve and actively integrating SDOH into your strategic and care-delivery models, then start working on new models to bridge social isolation (physical and mental) and to better understand (and develop strategies to address) challenging behaviors, including chronic helplessness.

To learn more about our Dallas journey, please visit our website and see what our team of PCCI experts is doing to make a difference or visit our technology partners at Pieces Technology to experience the Pieces IRIS™ technology.

14 November 2018

World Diabetes Day




Today, November 14, is designated as World Diabetes Day to unite the global diabetes community to produce a powerful voice for diabetes awareness and advocacy. According to the World Health Organization (WHO), over 425 million people are currently living with diabetes, prevalence is continuing to rise, and one in two people currently living with diabetes is undiagnosed.

Living with diabetes is a daily struggle, but many organizations have worked to create programs to decrease the struggle of those impacted. Ms. F, a 62-year-old African-American female with diabetes who relies on getting her nutrition from a food pantry, is a great example of someone that has benefited from these programs. Ms. F struggled with making proper food choices, adherence to proper medication, and transportation to make regular doctors’ appointments.

Through part of PCCI’s Connected Communities of Care program which shares patient’s information between providers and community-based organizations, Ms. F’s health and social service providers were able to connect and share information regarding her condition. When Ms. F visited the food pantry, staff members were aware of her diabetes. This knowledge enabled the staff to effectively guide her through her diet choices. This pilot program between three food pantries and Parkland Health & Hospital system helped many patients in taking the steps needed to control their disease.

In addition to limited access to healthy food choices, many patients in underserved communities have limited access to transportation. This challenge has made the remote monitoring of patients a critically important component in managing diabetes. PCCI is partnering with Parkland Health & Hospital System’s Global Diabetes Initiative to explore innovative approaches to improving the care of diabetic patients with foot ulcers which can lead to amputations if unresponsive to care. By acquiring data from home glucose monitoring devices and making real-time changes to treatment without physically having to see the patient, the (soon to be launched) study aims to create a sustainable remote glucose monitoring care system. This system will improve glucose control, promote faster healing of foot wound, and reduce long-term healthcare utilization and ultimately, reduce the burden cost of care for individuals and families.

Resources:

https://www.worlddiabetesday.org/

https://www.idf.org/e-library/epidemiology-research/54-our-activities/455-world-diabetes-day-2018-19.html

https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf

4 September 2018

Family Doc to “Design Doc”!




“Hey doc, don’t you miss being a family doctor?”  is a frequently asked question over cocktails and during client meetings. My response is always the same, “Actually, I am still serving patients but in a very innovative way and on a much larger scale for better health and social impact. I am now a “Design doc.”

Positive Promotion

After 15 gratifying years of service as a traditional family doctor, I now enjoy taking care of patients by designing healthcare solutions that result in better patient experiences, lower costs, and increased quality. My new career as a “design doc” has been very rewarding.

Design with a Cause

Like any designer, impactful ideas put both the big picture and intricate details into context simultaneously. Taking into consideration questions like: How does a doctor think? What are a patient’s expectations, needs, and goals? And what are the high-precision treatment options available? Begin the innovative process of designing scalable healthcare solutions.

Care and Collaboration

By approaching solutions from a “design doc” perspective, I collaborate with healthcare executives, frontline care teams, services providers, and members of the communities we serve to combine the “art” of medicine with clinically engineered artificial intelligence. The result of these collaborations and insights are solutions that can augment clinical decision making at the point of care and facilitate timely coordination of care beyond the walls of service providers and into the community.

Same Goal, Different approach

Quality of patient care has always been top of mind both during my time in the clinic and my new role at the Parkland Center for Clinical Innovation (PCCI). I went from providing patient care to enabling providers to better care for their patients and the community through PCCI’s innovative solutions.

Learn more about PCCI’s careers, or stay up-to-date with our recent news by following us on FacebookTwitter and LinkedIn!

Photo via Thinkstock by Getty Images. Item number: 857015410.

28 August 2018

My PCCI Internship – Synthetic Data Project




As my internship at Parkland Center for Clinical Innovation (PCCI) comes to an end, it feels great to look back and ponder over what I had the opportunity to work on, achieve and experience over the past three months. I arrived at PCCI with high expectations and am happy to say that I wasn’t disappointed. The project I worked on is called “Synthetic Data.” As the name suggests, the goal of the project is to create synthetic data from real medical datasets.

Why do we need synthetic medical data sets?

Real medical data is expensive and seldom released for research due to various privacy issues connected to it. Regulations exist because by looking at the medical data, a hacker could identify the name of a patient, thereby gaining access to sensitive information. The synthetic data project at PCCI aims to alleviate these problems by creating synthetic datasets, which are as close to the real medical datasets as possible without compromising a patient’s privacy.

Generative Machine Learning Algorithms and Challenges

Generative machine learning algorithms, specifically, Generative Adversarial Networks opens in a new window (GANs), proposed in 2014 [1] were used in this project. GANs have gained huge popularity within the machine learning community with a wide variety of GAN models being proposed.

There were quite a few challenges along the way in realizing the goal of synthetic data generation. First, the proposed GAN model has not been applied to real medical datasets before us, as it was mainly designed for image generation tasks. It also tends to not perform well with different modalities of data, which are naturally present in a real medical dataset. Modifying the network to work with real medical data or modifying the data (mostly getting it into a single modality) for it to work with GANs was a major challenge.

Second, there is an explosion in the number of GAN-based architectures being proposed and thus coming up with a novel architecture is a huge challenge in itself. After a lot of deliberation, we came up with an approach that allows us to incorporate domain knowledge into the GAN architecture. Below were our three possible approaches:

  1. Advice on just the discriminator.
  2. Advice on just the generator.
  3. Advice on improving the zero-sum game between the discriminator and the generator.

Improving the Zero-sum Game

Taking the third approach, we decided to incorporate reconstruction error [2] into the discriminator and the generator loss functions. The simple intuition is this: train the network with a mini-batch of data and generate the synthetic counterpart for each mini-batch.  Since reconstruction errors, as the name suggests, measure the errors between the real data and the “reconstructed” data, adding this to the loss functions can penalize either the discriminator or the generator depending on which performs the worse (i.e. has a higher loss difference) with respect to the mini-batch in question.

Initial Experiments Using the MIMIC III Dataset

For our initial experiments, we made use of the MIMIC III dataset [3], which is a dataset incorporating clinically relevant data for all admissions to an ICU at the Beth Israel Deaconess Medical Center between 2001 and 2012. Figure 2 shows the selected features.

Synthetic Data

Figure 2: Selected MIMIC III features

Experiments Using GAN-based Methods

We then ran a couple of experiments. First, we used three different GAN-based methods, without the reconstruction error, to see how the network performed with the defined loss functions.  Figure 3 shows the original and the generated data as obtained from these networks.

Synthetic Data

Figure 3: Snippets of the original and generated data

An Imbalanced Dataset

The original and synthetic datasets were then used to train a machine learning model. For the synthetic data to be useful, it needs to be as close to the real data distribution as possible, which should be captured by a machine learning classification model. Our dataset was highly imbalanced. We were predicting mortality rate and since a majority of patients come out of the ICU alive, we had nearly a 90%-10% split between negative examples i.e. patients who are alive after the ICU treatment and positive examples i.e. patients who die in the ICU. We used a cost-sensitive support vector machine classifier, constructed for such imbalanced data, to report the F1 score and the area under the curve (AUC-ROC) in Figure 4.

Synthetic Data

Figure 4: Comparison between the machine learning model performance for real and synthetic data. (W-GAN, GAN and MA-GAN are GAN models used to generate synthetic data)

As it can be seen, the results proved our hypothesis that real healthcare data was going to be a challenge for techniques, like GANs, which rely on many samples of very predictable data types since healthcare data tends to be more diverse and are difficult to compose into higher-order features. (Credit: David Watkins opens in a new window, my supervisor). We then used the reconstruction error to “indirectly” capture the relationships between the data points and used “real” hospital data to test on and create a synthetic dataset from it. The work was currently still in progress at the time this blog was written.

Working at PCCI

The work culture at PCCI, in my opinion, towers above other places. The work hours are flexible, the team’s ethics and bonding are strong and people are always willing to help you regardless of their schedule. The company truly values its employees and creates a work environment where every employee gives his/her best. You will never feel out of place (not even on the first day) as all your tasks are defined and everyone is so welcoming. A great thing about PCCI is the absence of an implicit hierarchy. Everyone from the CEO to your respective manager(s) (thank you Albert Karam opens in a new window) is always accessible. I never felt any different than a PCCI employee and this says a lot about the values of the company and how these values are being nurtured by PCCI’s CEO Steve Miff opens in a new window and all employees of this amazing organization.

Another important quality about PCCI is that it values and encourages all feedback that any employee may have and any grievances are then actually addressed. I am proud to say that Steve himself makes sure that any such issues are addressed.

Nothing is perfect and PCCI has a few areas where it can improve. One area of improvement is getting access to the real data. This is currently a very is a slow process, which makes sense as it is sensitive medical information of real patients, but it can be sped up. Another area of improvement, that I have actually raised to Steve during a meeting is that PCCI should focus on publishing research papers. It is a company that is capable of doing amazing research and has access to real medical datasets that are difficult to find. I hope PCCI becomes more active in this regard.

What PCCI does is super important to the community and it makes sure that all the employees realize this fact. Creating an impact in the real world, on real lives is a great morale booster for anyone and since PCCI values are centered around this motto, working here has been a great experience. It’s been a pleasure interning at PCCI and I am happy to be taking fond memories with me back to school.

Learn more about PCCI’s careers opens in a new window, or stay-up-to-date with our recent news by following us on FacebookTwitter and LinkedIn!

References:

[1] Goodfellow, Ian, et al. “Generative adversarial nets.” Advances in neural information processing systems. 2014.

[2] Borji, Ali. “Pros and Cons of GAN Evaluation Measures.” arXiv preprint 2018.

[3] Johnson, Alistair EW, et al. “MIMIC-III, a freely accessible critical care database.” Nature 2016.

8 August 2018

My Summer as a Data Science Intern at PCCI




For the short duration of returning to my hometown Dallas for the summer, I’ve been interning at Parkland Center for Clinical Innovation (PCCI) as a Data Science Intern. During my interview with Albert and Vikas, we discussed some issues with the representation of data in the current healthcare system. Hospitals use different coding systems in their electronic medical records (EMRs), making communication between hospitals and care providers difficult. A while ago, a new health data standard called FHIR (Fast Healthcare Interoperability Resource, pronounced “fire”) was proposed. My project this summer aimed at identifying whether data could be easily transformed into the new FHIR format, carrying out the transformation, and creating predictive models using the new FHIR data.

Situated on the 11th floor of the building, PCCI is a very chill place to work. Quiet spaces are easily found at desks and conference rooms scattered around the office. As an intern, I sit on the “Intern Island” with (usually) 6 other interns. I like this space because we get two monitors and a Lenovo Thinkpad.

 

Emily Wang, PCCI Data Science Intern

As for work, each PCCI project usually consists of one project manager, a clinical expert, and a data scientist. The intern projects are no different; Aaron was the FHIR Project Manager Intern, and Mila was the FHIR Clinical Intern. Both had important but separate duties that helped our project succeed.

As the Data Science Intern on the FHIR project, I was responsible for first converting the data into FHIR resources.  This involved bringing back Java knowledge from several years ago! There were definitely some issues figuring out how to add the right dependencies because Java can get complicated very quickly. A few days were spent just trying to get oriented with Java and Eclipse, and making sure all the necessary packages for FHIR were installed.

We were working with two years of data. This roughly translates into 27 million (!) vitals and 17 million labs, and each vital and lab was converted into its own separate file. I quickly realized that there would be no space on my laptop to hold all of these files, so we decided to enlist the help of Microsoft Azure. With Azure, the task became less difficult, but still, the hardest part of my summer was working with such huge numbers of files.

Caught up in the huge task of transforming vast amounts of data to FHIR resources, I left very little time in my internship to work on actual data science. Out of the approximately 13 weeks total, about six weeks were spent converting the table format EMR data into FHIR resources, five weeks were spent on parsing the FHIR resources into a format for machine learning, and the remaining two weeks were dedicated to model building. Reflecting back, I would definitely work harder to cut short the resource conversion in favor of more time for data science.

 

"As an intern, I sit on the “Intern Island” with (usually) 6 other interns. I like this space because we get two monitors and a Lenovo Thinkpad." said Emily Wang

“As an intern, I sit on the “Intern Island” with (usually) 6 other interns. I like this space because we get two monitors and a Lenovo Thinkpad.” – Emily Wang

As a Data Science Intern at PCCI, you have the freedom to work in any language you want; the full-time Data Science team is very evenly divided between R and Python. There’s also a lot of freedom in dictating which path your project will go. Your supervisor will point you in a very general direction of where to go and state goals and expectations, but is otherwise very lenient!

Don’t be shy about asking around people for advice and help, even if they’re not on your project team! Even though most people are busy with various meetings, they will gladly schedule a 30-minute or even hour-long block to discuss your project privately with you.

When presenting your project, whether it’s a progress update or final presentation, expect multiple questions from the audience. It’s not that they want to quiz you on your knowledge and preparation on your project, but because they’re genuinely curious and care about understanding what you’re doing over the summer.

A mandatory 30-minute lunch is required every day. I recommended bringing lunches that can stay in the fridge for several days (like salad) or not bringing anything because there are often team lunches and random outings during the day. Occasionally there’s leftover pizza or sandwiches from lunch meetings in the big conference room or leftover burritos from breakfast.

I enjoy the diverse atmosphere at PCCI the most. The three teams: Data Science, Project Management, and Clinical teams collaborate and work together so well. It’s a very fluid system. A data scientist with a question about the best intervention methods for patients with diabetes can easily walk over to a clinical team member and get an answer within minutes. Despite being employed as a data scientist, you have access to an entire host of medical knowledge from the clinical team and connections from the project management team.

My biggest takeaway from this internship is learning about long-term time management and collaboration. Manage your time well and you’ll be able to at least touch on everything you wanted to learn during your internship. Collaborate with as many people as you can, so not only can you learn so much more but also gain friends and connections while doing so.

31 July 2018

Prevention of Sepsis Through Machine Learning




Sepsis opens in a new window is a word that can evoke feelings of dread for most people and yet, many fail to understand its meaning and implications fully. It is a life-threatening condition that arises when the body’s response to infection results in organ dysfunction or failure. It is one of the leading causes of deaths with 6 to 9 million deaths per year worldwide and 250,000 deaths per year in the United States. To make matters worse, sepsis is also one of the leading causes of preventable deaths worldwide. The key to prevention is early detection since every hour of delay increases the odds of death by 20%. However, sepsis can be very hard to detect especially in its early stages when it is reversible. The symptoms often point to other conditions and patients can deteriorate rapidly. Traditional risk models such as SIRS criteria to detect sepsis generate a lot of false positives leading to inefficient or ineffective care and nursing fatigue.

sepsis

Leveraging Machine Learning 

Due to PCCI’s close collaboration opens in a new window with Parkland Health & Hospital System, we recognized sepsis detection and prevention as an area that could benefit from an advanced machine learning-based algorithm. To be effective, the model needed to fulfill the following criteria:

  • A high enough positive predictive value (PPV) to ensure that false positives do not end up causing frustration for clinicians. It is very challenging to have a high PPV for use cases with low prevalence rates in the target population. Despite the high mortality numbers, the overall prevalence rate for sepsis in an inpatient setting (our target population) is 3% to 4%.
  • A prediction interval (how far ahead in the future can the model predict) that gives clinicians a sufficient heads-up to intervene.

PCCI developed a predictive model to fulfill the criteria identified as effective. This real-time model is designed to predict an individual’s risk of becoming septic within the next 12 hours.

Instead of relying on traditional ways of detecting the risk of sepsis through a handful of physiologic variables, we cast our net wide and started out with approximately 120 variables as potential predictors, such as socio-demographic, vital signs, co-morbidities, hospital utilization, medical conditions, clinical history, and lab results.

Testing the Lasso Logistic Regression Model

We trained and tested a Lasso Logistic Regression model, a decision tree and a neural net on 27 months-worth of Inpatient encounters from Parkland. Initial data sets included 54,629 encounters and 30,922 patients. We chose the Lasso Logistic Regression model as the final model because of its high performance and interpretability by clinical users. The resulting model that compared favorably to other community models resulted in:

  • 95% Accuracy
  • 30% PPV
  • 55% Sensitivity

Incorporating the Model into Clinical Workflows

A model is useful only if it is actionable and available at point of care. PCCI’s Sepsis model is incorporated into clinical workflows through industry-standard APIs. The model accesses EHR data in real-time every five minutes and alerts the clinician if the risk is above a certain threshold. The risk threshold can be tailored to local populations and clinical best practices of an organization. The alerting mechanism can vary based on EHR functionality, but it can include a simple alert, queuing up of order sets based on the risk threshold, and sending a real-time alert to a secure mobile device or pager for a member of the rapid response team.

Learn more about PCCI’s work, opens in a new window or stay up-to-date with our recent news opens in a new window by following us on Facebook opens in a new windowTwitter opens in a new window and LinkedIn opens in a new window!