Agile Decision Making: Combining Insights with Instinct to Navigate the Unknown

Decisions are being made today that may alter a healthcare organization’s trajectory for years to come. The pandemic has revealed how unprepared we are as an industry to use real-time and predictive data to tell us “what is happening”, “what will happen” and “how do we make it happen.”

COVID-19 has upended conventional leadership thinking and challenged even the most astute to admit “we don’t know what we don’t know.”

 

We Must Think Differently

In recent years, the healthcare industry has made a significant investment in advanced analytics and encouraged a transition to insight driven decision making. A survey of healthcare executives conducted by Black Book in January 2020, indicated that the vast majority (93%) feel data analytics is crucial to helping meet future healthcare demands and make strategic decisions. Yet, utilization of advanced analytics was described as “negligible” by 80% of the respondents. 71% of surveyed executives said they were too busy to learn such systems. In our conversations with healthcare leaders, they indicate they prefer to delegate analysis to others, aren’t really aware of what data analytics options they have, and often prefer to rely on their instinct – after all, it’s what made them successful.

For many, they have relied on the hundreds of static spreadsheets and score cards providing weekly, daily, monthly, and quarterly reports on “what happened” to validate these instinctual decisions. Yet, in today’s world much of this history is not pertinent. Comparative benchmarks, based on pre-COVID operating models, are no longer valid. Even the most sophisticated data scientists cannot take this historic data, identify new patterns and build a model to predict the future based on our current circumstances in a rapid fashion. Even the fastest computer processor can’t generate the algorithms we need when the situation changes as rapidly as it does today. Advanced analytics tools such as machine learning and predictive modeling are at their best when the future environment is similar to the historic environment – and that’s not the case today!

Healthcare decision makers must think differently about the role of data in their decisions and their approach to making decisions.

Agile Analytics, Combining Insights & Instinct for Decision Making

In the emerging COVID world, decision makers need practical, short term, results driven answers to their questions. Influenced by interrelated factors such as numbers of positive cases, bed availability, medication and PPE supply, treatment protocols, deployment of virtual tools and the impact of social determinants on population segments, the problems we are trying to solve change constantly.

Agile Analytics is a new paradigm for healthcare decision makers and is focused on rapidly finding value in data.

COVID brings the need for quick turnaround of information and rapid decisions. Historic data trends and traditional, large scale analytics projects are not dynamic enough nor nimble enough to support today’s decision-making environment. Since “we don’t know what we don’t know”, we must learn along the way. Jumping to solutions, based on traditional best practice or the play books of the past may lead us down the wrong path. Yet, complex architectures and advanced analytics platforms may not be responsive enough to meet the challenges of the current environment.

Agile Analytics is a style of working and problem solving. Originally created by software developers in 2001, agile is a collaborative approach built on short one to three-week iterations resulting in a data output that can be used to solve a problem or make a decision. Each targeted iteration starts simply, explores key components of the problem and initial hypothesis, and provides decision makers directional insights to drive data stories and resulting conversations.

The Minimum Viable Analysis

A Minimum Viable Analysis (MVA), is similar to the Minimum Viable Product concept developed by Eric Reis of Lean Start-Up fame. “Rather than spending months even years perfecting a product (analysis) without ever showing it to a customer (decision maker)…we start with a simple prototype of the product (analysis) get feedback from the customer (decision maker) and learn, then build it out further through multiple feedback sessions and iterations.”

The MVA starts simply. As in the case of the minimum viable product, an initial problem to be solved or decision to make is identified by a small collaborative team of decision makers and data analysts. Using a travel analogy, rather than building an entire automobile to solve a transportation problem, the team starts with a skate-board as an initial prototype. This prototype defines data output, visualization or presentation expected from the MVA.

For problems the healthcare delivery organization is trying to solve during these times of uncertainty, a minimum viable analysis can be used to answer questions such as:

  • When will we run out of ICU beds/ventilators?
  • What staffing constraints will we have in acute care? in ambulatory? in post acute?
  • What is changes should we make in our care model given the mix of virtual and in-person visits?
  • What triggers will change our plans for rescheduling visits and procedures?
  • Will we have PPE limitations or other supply chain issues?

Then, as specific questions are answered, an iterative approach can be used to add to the knowledge base and analysis to produce answers to more targeted questions or broader concerns.

We term the gap between the data output from our agile analytics effort and the change in behavior required to solve the original problem as the “Last Mile.” Unfortunately, the Last Mile is often an afterthought in analytics projects. Frequently, leading edge analytics initiatives don’t generate value or results due to ignoring the last mile. Resistance, lack of trust and transparency, siloed based behaviors, and a variety of last mile issues can often be anticipated and planned for early as problems and hypotheses are defined. Decision makers have a unique understanding of the environment where the desired insight-driven changes should occur. Instinct driven decision making plays an important role in defining Last Mile issues, early in the decision process.

COVID-19 brings a great deal of uncertainty, and decision makers can’t delegate data analysis as they did in the past. Multiple perspectives, judgement and experience are needed to collaboratively define the problems to be solved, potential hypotheses, explore the implications of analyses findings and communicate insight implications.

While data driven insights can provide new “aha’s”, they are best when combined with the decision maker’s instinct to navigate these uncharted waters.

The Story of Data

Untangling Complexity: Agile Decision-Making During the COVID19 Crisis

“The system is not really geared to what we need right now… let’s admit it,” said Dr Anthony Fauci from the National Institute of Allergy and Infectious Diseases.

The outcry for testing from patients and the media has challenged the US government, public health leaders and clinicians in primary care and hospital emergency rooms as they the battle the COVID19 pandemic. It is clear, the complexities of test development and deployment are only the tip of the iceberg as our fragmented healthcare system ramps up our response.

In an ideal world, the public health, care delivery, and payment systems are all components of a unified system – a three legged stool of sorts, guaranteed not to wobble and each carrying its respective weight in managing the health of the US population. Yet, in reality the three legs of the stool rarely collaborate. Each with their own ingrained cultural, political, regulatory and economic incentives. In this world, distinct responsibilities, bureaucratic processes and information systems burden decision making and slow down response.

COVID19, knows no boundaries. In just a few short days, it is smashing the complicated mixture of federal, state, local, public and private organizational siloes and accomplishing more than many of us who have spent our careers trying to improve the system.

The March 17th expansion of telehealth benefits for Medicare recipients by CMS provides a tangible example. Relaxation of HIPAA rules gives the country’s older population access to medical care (both virus related and for other services) without having to leave their homes. Providers can use personal video chat applications like Apple FaceTime, Facebook Messenger video chat, Google Hangouts video, or Skype.  Rather than deploying technology for technology’s sake as we have done in the past, this step demonstrates:

  • Coordination across the silos of public health, care delivery and payment
  • Design of a new way of working – based on the needs of the at-risk population and their care providers
  • An agile decision that simultaneously untangled the complexity of the current system
  • Rapid communication of the change through multiple media channels
  • Quick tools forwarded to patients and their clinicians from health care industry associations, data and technology partners and advisors

The 9-11 Commission reported, “The 9-11 attacks revealed four kinds of failures: in imagination, policy, capabilities, and management’ (National Commission on Terrorist Attacks upon the United States 2004, p.339). Today, as we face this threat each step we take provides an opportunity to untangle the complexity, remove barriers and set the course for more agile and unified decision making. What other steps should we take?

 

Healthcare Analytics: Test the Waters Before You Dive In

Analytics as a Differentiator

Many healthcare organizations see the use of analytics as a primary differentiator in the journey to value-based care. The current analytics vendor marketplace includes:

  • Niche and point analytics systems (solutions focused on one functional area – marketing, finance, clinical, claims, quality measurement, human resources, etc.)
  • Enterprise analytics tools that are part of larger offerings such as ERP and EHR
  • Enterprise data repositories, data lakes and tools to support data normalization, aggregation, visualization, etc.
  • Extended enterprise systems such as care management platforms to support population health management and consumer activation
  • External industry comparative data sets

Most health systems use these systems for a variety of analyses including:

  • Descriptive analytics – to answer questions such as what happened; how many, how often, where, who…; what exactly is the problem; how did we perform, etc.?
  • Diagnostic analytics – to answer why is this happening and what are the trends?
  • Predictive analytics – to answer what will happen?
  • Prescriptive analytics – to answer how do we make it happen?

Advanced organizations are moving into Continuous Intelligence (CI). Rather than depending on traditional score cards, which require people to orchestrate every step in the analysis, CI uses artificial intelligence and machine-based algorithms to automatically interpret and harmonize the data. CI continuously discovers patterns and learns what’s of value in the data, and immediately sends insights to decision makers. In healthcare, this will be extremely valuable at the point of care, for care managers working across the continuum and for strategists who need agility in their decision-making processes.

Demonstration Projects Help “Test the Waters”

Yet, healthcare leaders often struggle with basic analytics projects. In fact, Gartner has uncovered an 85% failure rate on data projects. Ironically, this may be in part the result of executives trusting their gut rather than the insights derived from the data. Other problems include political, data quality challenges, limited data interpretation skillsets, resistance to change, governance, etc. It is often helpful, prior to jumping in the deep end of the analytics pool for the organization to first:

  • Learn new ways to think about data
  • Collaborate on new ways to use data
  • Leverage existing analytics tools and resources prior to making significant analytics investments
  • Develop consistent methods for defining the problem to be solved, defining the data, cleaning up data inconsistencies and improve on data presentation skills

In other words, “test the waters” prior to diving in. A Demonstration Project provides the ability to learn about insight-driven decision making through the execution of a clearly defined project. The Demonstration Project works best with a limited scope, limited resource commitment and within a short time frame. An Analytics Demonstration Project focuses on data quality, building trust and deploying a consistent data management methodology. Insights can be channeled into performance improvement projects, strategic and operational plans or other pertinent initiatives.

Leading Demonstration Projects

Often a “See One”, “Do One”, “Lead One” approach works well — an outside analytics advisor partners with the organization and leads the first project, the advisor partners with the organization on the second project and in the third project, the organization leads the project with the advisor serving as a mentor. Preparation and starting the project well are essential elements for success and key steps include:

  • Understand the problems you are trying to solve: Many organizations acquire solutions before they have identified what the organizational “points of pain” are. This results in multiple tools or platforms competing for leadership attention, budget dollars and resources. Identifying the problems you are trying to address, or the questions you are trying to answer will help prioritize the use of limited analytics capacity.
  • Evaluate your existing investments in analytics tools, systems and methodologies: It is essential to understand your current analytic tools footprint. Rather than jumping to acquire a new analytics system or service, you may find you are able to leverage existing investments. Over time, the lessons learned from successful Demonstration Projects will identify opportunities to improve your overall Analytics Strategy. Specifics may include rationalization of  duplicate systems and addition of new capabilities to close gaps based on your existing vendor’s development road map.
  • Break Data Analytics projects into smaller waves versus a big bang deployment: Focused sprints, using an agile methodology rather than a traditional IT waterfall project implementation plan, shifts the focus to performance improvement. For the purpose of planning a Demonstration Project there are many ways to break the problem into smaller, targeted efforts. So for example, a Sepsis prevention and reduction effort can be divided into multiple smaller Demonstration Projects including descriptive and diagnostic analysis of sepsis rates, use of data driven screening tools or protocols, sepsis prediction tools, stakeholder education through data, etc.
  • Ensure data integrity: Source systems (such as your EMR or niche systems) have varying quality of data. Oftentimes data are missing or stored in different places depending on clinical documentation practices. In addition, definitions of data may vary from system to system or even from stakeholder to stakeholder. A common “Data Dictionary” becomes essential. Demonstration Projects can be useful defining organizational steps in developing the data dictionary, mining data, validating data, creating data baselines, analyzing data, presenting and visualizing data, and developing a go-forward data improvement plan.
  • Communicate benefits to the organization, including executives: Use techniques to model value and even Return on Investment/Value of Investment. By showing how Data Analytics initiatives impact strategic goals as well as daily operations, the organization can begin take a holistic view and integrate insight-driven decision making at all levels of the enterprise.

Organizations who leverage the strategic value of data to make informed decisions will be the ones who not only survive, but thrive.

 

Data – The Star of the Show

One-week post #HIMSS19, blogs and articles are using phrases like “no one unifying theme”, “something for everyone”, “the invasion of non-healthcare high tech” and “the end of the EHR movement” to summarize the global conference key take-aways. Even though the exhibition showcased interoperability, artificial intelligence, telehealth, security, the internet of things, precision medicine and more, the focus was not on these technologies. In fact, a more subtle underlying reality was clear in both the education sessions and vendor booths. Data is the star of the show!

Data in the Spotlight at HIMSS

Data collection, data integrity and quality, data access, data for benchmarking and comparative analysis, data protection and safety, data as part of care management platforms, data analysis and prediction, data visualization and ultimately, the transformation of data into information each had the spotlight. Data and analytics vendors were certainly a major focus of the show. These vendors were not outdone by solutions which offered data a by-product of a much more comprehensive offering such as Enterprise Resource Planning or data embedded within cloud-based services designed to solve specific business and clinical problems such as readmissions. Moreover, CMS took center stage in multiple forums to discuss the Interoperability and Patient Access Proposed Rule. With a goal to touch all aspects of healthcare, from patients to providers to payers to researchers, “CMS hopes to break down existing barriers to important data exchange needed to empower patients by given them access to their health data,” CMS Administrator Seema Verma indicated. HIMSS even gotten directly into the act. Recognizing the importance of data and information, they have changed their vision statement from “better health through IT” to “better health through information and technology.”

Siloed Organizations and Turf Battles

Yet, even with all of this attention and applause its hard to reconcile the maturity of data initiatives in many health and healthcare organizations today and the challenges faced by many. For data to be valuable, data must be converted into information, information into insights, insights into decisions, and decisions into action. Unfortunately, many health and healthcare still manage their data assets within siloed organization structures. Turf battles are common. Data-related issues for decision makers often include:

  • Confusion over who “owns” data and analytics
  • Questions regarding centralization versus decentralization of analytics resources and tools
  • Limited trust in the data and reports created by other parts of the organization often resulting in considerable rework
  • Inconsistent data definitions, duplicate data sources and systems, and costly resource requirements
  • Analysts who don’t understand the problem to be solved when data requests are “thrown over the fence”
  • Proliferation of spreadsheets, manual manipulation of score cards, and limited automation
  • Delays and slow turnaround of data requests
  • Data repositories that create many but rarely used reports
  • An inability to fully appreciate much less realize the benefits of big data, and predictive and prescriptive analysis

According to thought leaders at SAP, “less than 1% of the world’s data in business is analyzed and turned into benefits”.

The Enterprise Analytics Management System

Creation of an organizational approach to standardizing management of data, or Enterprise Analytics Management System (EAMS), results in a defined, documented and deliberately managed set of priorities, polices, procedures and processes. The EAMS should address the collection, definition, analysis, interpretation, translation and presentation of data to a wide variety of audiences.

Objectives for the EAMS are to:

  • Transition from a data and analytics departmental/siloed approach to a clear, consistent enterprise approach to managing data assets
  • Ensure key stakeholders understand enterprise analytics assets and have a consistent methodology for working together across the organization
  • Clearly define an enterprise analytics organization and operating model
  • Build a culture of collaboration and accountability to support data and analytics
  • Design and implement an oversight process or governance process
  • Develop an Enterprise Analytics Strategy and Pragmatic Road Map for the next few years

For data to truly be the star of the show, we must improve our ability to govern and manage this critically important asset.