As Seen in HBR: How to Use Digital Health Data to Improve Outcomes
September 16, 2022 · Andy Reynolds
NCQA’s Peggy O’Kane, Brad Ryan, and Eric Schneider and board member John Glaser co-authored an illuminating essay for Harvard Business Review, How to Use Digital Health Data to Improve Outcomes.
We’ve reprinted the text below, with permission from Harvard Business Press.
We recommend this essay for its:
- Grounding in Quality Basics: Peggy, Brad, Eric and John do a nice job acquainting readers with the animating Why? and How? of quality measurement.
- Inspiring Vision for Digital’s Promise: The essay paints a powerful picture of what patients, payers and providers stand to gain from digital quality measurement.
- Wit and Accessibility: Relatable examples and pithy phrases are sprinkled throughout this essay.
How to Use Digital Health Data to Improve Outcomes
The amount of health information generated by digital tools is rapidly growing. It can and should be used to improve the quality of health care. This article makes the case and explains what will be required to make it happen
We hear a lot about “digital health” these days. As data about our health piles up — thanks to sources like electronic health records, personal fitness apps and gadgets, and home genome test kits — we should understand a lot more than we used to about what’s wrong with our health and what to do about it. But having a lot of data is not enough. We have to be aware of what we have, understand what it means, and act on that understanding. While the challenges are in some ways more acute in the United States because of its fragmented system of care, they exist in health care across the globe.
Here’s an all-too-common scenario:
June, aged 67, is in the emergency department with abdominal pain and rectal bleeding. Tests reveal inoperable colon cancer that’s probably been developing for years. After several difficult and unsuccessful courses of chemotherapy, she enters hospice care and passes away several weeks later.
Colon cancer is largely curable and often preventable if it’s caught early enough to detect and remove precancerous growths. With recommended screening, June might be alive today. What happened? She had colonoscopies on schedule at 50 and 60 but thought she was in the clear until 70 because no one flagged the radiologist’s note about a few small irregularities that meant she should come back at 63. It wasn’t the radiologist’s job to make sure June acted on the finding, which was hidden away in the “Test Results” tab of her electronic health record (EHR). She missed it. Her primary care doctor missed it. The whole health care system missed it.
Too many Junes are lost too soon. These small failures with big consequences are everywhere in the U.S. health care system, costing Americans years of healthy life and billions of dollars in avoidable treatment costs. When clinicians had to depend on landline phones, multipart forms, and paper folders to execute and track these tasks, it isn’t surprising that such failures to occur. Now that computers, smartphones, and the internet exist, they (theoretically at least) can be used to send consistent and timely reminders to patients like June to get those early colonoscopies.
But digital tools don’t use themselves: We have to tell them what to do. In June’s case, the right combination of systems would have had to detect and analyze the data, send it to her and her physician, track their responses, make it easy for her to “click here” to schedule her procedure once she turned 63, and follow up the suspicious result with recommendations for tests and treatment. Though the dangers of “alert fatigue” are very real and must be avoided, clinicians and staff will welcome properly designed reminders that help avert a missed or delayed diagnosis and the regrets that come with it.
Figuring out how to develop systems to use a growing quantity and variety of digital information is perhaps the most important, and formidable, health care mission of our time. Since the 1990s, our organization, the National Committee for Quality Assurance (NCQA) has been using data to measure and improve health care quality, originally to accredit health plans and more recently to gauge the performance of providers. When the NCQA began, the challenge was to compile enough data and to make inferences to fill in the blanks where there wasn’t any good information. Now the challenge is the overwhelming amount of data that needs to be mined for its essentials. But NCQA’s mission remains the same: to put data to work to increase the effectiveness of the resources devoted to health care.
In this article we will outline the steps needed to close the loop that connects digital information to action.
Measuring Quality: Basic Principles
Health care quality measurement rests on three questions:
- Are we doing the right things to manage health and health care?
- Are we getting the outcomes we want?
- If not, what do we need to change?
These questions almost never have easy answers. People are not widgets, and the outcome of a particular episode of care depends on multiple factors: the performance of clinicians, the attentiveness of caregivers, the patient’s initial state of health and motivation to get better, and the patient’s overall circumstances (income, environment, access to food or transportation, availability of help around the house). Outcomes include not only whether patients are now healthier but also how they felt about their care and how it compares with the same care rendered elsewhere or with different treatment approaches that might cost less and/or deliver a better outcome.
While measuring the quality of care is difficult, we do know that the current report card for the United States paints a mixed picture. Its best available care is often truly the best in the world. However, it is primarily famous in health care circles for paying the most (19.7% of GDP, twice as much as most peer nations) and getting poor value for its money. For example, the U.S. maternal mortality rate is an international disgrace: more than twice as high as Canada’s and four times that of Sweden (not to mention the gaping, and worsening, disparities by race). And the gap between average life expectancy in the United States and peer countries is widening.
This mixed and incomplete picture of the quality of care poses a significant problem for health care stakeholders. Health plans and employers need to know that they’re getting the value they are paying for. As payer contracts shift from rewarding more services to rewarding better outcomes, providers need to track their own performance. Quality should guide patients’ choices among providers and health plans, to the extent they have choices. Lawmakers and regulators need to understand the effectiveness of providers and medical services to help them allocate resources where they’ll have the most impact.
There are several reasons that the measurement of health care quality has been underdeveloped. One is that quality-based reimbursement still accounts for a minority of most providers’ revenue. Second, consumers have not demanded them. Instead, they trust the recommendations of their doctor or friends and family who have been treated for the condition in question.
However, the primary reason for the limited state of quality measurement is its reliance on insurance claims as the foundation for measurement.
Claims Data: An Incomplete Foundation for Measuring Quality
For the three decades since the health care industry began a serious, data-driven effort to measure quality, it has relied heavily on analyzing insurance claims — the only large and relatively consistent digital data source across all providers. While claims data can provide some insights, data collected for one purpose — in this case, getting the provider paid — is often not well suited for other purposes.
For one thing, it’s often months old by the time it’s available for analysis. For another, it’s clinically incomplete. A claim shows whether something was done but not the effect it had. A list of completed tasks — blood sugar tests, eye exams, weight and blood pressure checks — shows that a diabetic patient received care but not whether her blood sugar is under control. Claims also won’t contain vital information on the patient’s full health picture — unless that information gets the provider more money. He or she can bill for a diagnosis that pays at a higher rate if the patient has a comorbidity: for example, treating a heart attack for a patient who also has diabetes. But linking the patient’s other claims together may be the only way to discover that she also has arthritis and reflux disease and eczema.
And finally, each claim is a partial snapshot of one service or episode of care delivered at a moment in time, and even a pile of snapshots is not the same as a movie. Improving health or worsening illness take place between the snapshots. By the time we take the picture, it’s too late to affect the course of events, and all we can do is look at the result and think about how to do better next time.
The Era of Digital Measures
Fortunately, we need no longer rely on claims data. The tide began to turn with the mass adoption of electronic health records, driven by federal government incentive payments that started in 2010. The Office of the National Coordinator for Health Information Technology, which oversaw this herculean effort, continues to initiate and promote ways to leverage EHR data.
More recently, that data has been joined by information streams from monitoring devices, fitness trackers and smartphones, patients’ own assessments of their health, genomic data, and readily accessible population-level data on social factors that profoundly affect health: employment status, income level, environmental quality, level of community support, and so on. Advanced analytics can potentially allow us to combine all of these data sources to start developing a clearer picture of health status and the effectiveness of care at all levels — from individuals to groups of patients with the same diagnosis to entire communities.
That’s the supply side. On the demand side, the Centers for Medicare & Medicaid Services (CMS), the single largest payer in U.S. health care, is actively advancing the use of digital data to measure the quality of care. Commercial payers, too, are seeking better ways to gauge value, since it’s difficult to do “value-based” contracts without reliable measurements. Our own organization is developing digital measures to track the performance of the health plans we accredit, which collectively insure more than half the U.S. population. Every organization with a stake in measuring health care quality is preparing for a new era.
Learning from Others
The United States can learn from other developed countries that are employing their digital data to improve health care and health. Denmark, for example, has patient registry data dating back to the 1960s as well as a single shared system of electronic health records for the whole country. Its national digital health strategy focuses on all the things that the United States wants: timely knowledge, partnership with patients, prevention, equity. Denmark has a more manageable task than the United States, with a compact geography and fewer than 6 million people, but it shows us what’s possible.
The European Union is pursuing similar goals: In May it introduced a proposal for the European Health Data Space, to set up a single digital health market for its 450 million people.
In turn, the efforts in the United States to advance digital measures are of interest and value to other countries that are grappling with similar challenges of health care costs, quality, and access.
A To-Do list for Digital Measures
We see at least four imperatives for getting the United States where it needs to be:
Reduce the cost of data collection and improve its timeliness. This may sound like two objectives, but digital measures achieve both. Many traditional measures use data (such as insurance claims) that lags care delivery by up to a year, which in some areas can make them all but irrelevant. If we design them correctly, systems such as electronic health records and wearable devices can generate data as a byproduct of managing care not only more cheaply but also much faster. When data collection stops being a separate step from delivering care, we can go right to analysis and results.
Expand the range of usable data. All the new sources we mentioned above — EHRs, wearable health monitors, patients’ feedback on their own health (known in the trade as Patient-Reported Outcome Measures, or PROMs) — can potentially be combined with data on the patient’s environment such as water and air quality, crime rates, green space, access to transportation, and the density of grocery stores or social services.
NCQA is examining how to account for patients’ social circumstances — homelessness, poverty, isolation, access to nutritious food or places to exercise — in assessing the quality of their care. A physician may recommend that a patient take a daily walk — a great idea for a patient who lives near a park but bad advice for one who lives in a high-crime area and is afraid to leave the house. More data on more patients will allow us to develop measures that more accurately reflect the care needs and best treatments for specific groups or even individual patients. We will be able to account for the differences in care needs depending on economic circumstances, patients’ ability to manage their own care, and the quality of their social supports.
Leverage the broad adoption of electronic health records, mobile devices, and artificial intelligence to provide real time feedback and guide care. Electronic health records are evolving from being a record of the patient’s condition and the care they received to providing real-time support: alerts, reminders, computer-based guidelines for managing chronic disease, and logic that (tactfully) critiques a physician’s orders for tests and medication, comparing them against standard practice and checking for inconsistencies. Such an intelligent EHR would have reminded June and her doctor to schedule that follow-up colonoscopy when she turned 63.
As our systems for measuring care quality become more sophisticated, we will be better able to incorporate intelligence that is more personalized to the needs and desires of patients. A really intelligent EHR would notice that June likes to schedule her medical appointments on Tuesdays and would, with her approval, go ahead and schedule the procedure for the next available Tuesday.
Integrated health systems such as Salt Lake City-based Intermountain Healthcare or Pennsylvania’s Geisinger have developed digital tools to improve care for their patients, though both have the twin advantages of advanced IT capabilities and the financial incentive, as both provider and insurer, to focus on improving their patients’ health rather than simply on delivering more services. These organizations and others have leveraged their electronic health records to provide real time feedback to clinicians and patients. By expanding the range of data collected and reducing the cost to gather the data, the feedback that can be provided by these systems can be more tailored to the patient and hence lead to more effective care and health decisions
Establish a digital foundation for the ongoing production processes of gathering, analyzing, and reporting quality measures. Developing digital measures is not a one-and-done venture but a continuous transformation. Creating this foundation involves the following:
Devising a process for standardizing the many measures now in use. This process has to be rigorous enough that there’s general agreement on, for example, what level of blood pressure constitutes hypertension or what range of test results show well-controlled diabetes, but at the same time flexible enough to accommodate a degree of adjustment based on the population or individual being measured. Currently payers, regulators, and professional societies all have slightly different approaches to designing measures. The variation creates more work for the providers being measured, but almost certainly isn’t delivering commensurate value.
Replacing the paper-based descriptions of quality measures and the data they need. These descriptions must be manually entered into electronic health records and reporting software, a process that is expensive and error prone. The remedy is to replace the paper with software-based descriptions that can be easily added to clinical systems.
Creating software tools that facilitate collaboration in developing, testing, and maintaining measures. Neither illnesses nor treatments are static, and each new one will require its own measures. Payers, regulators, providers, and patient groups must participate in this effort in order to accelerate the development and testing of new measures and arrive at a consensus on which ones to adopt.
Automating the extraction of data from electronic health records rather than using human data abstracters (still a common practice). This will reduce the cost associated with collecting clinical data and improve its accuracy. We already have a solid tool for doing this: the Fast Healthcare Interoperability Resources (FHIR) standard, which is a standardized API for exchanging information among systems. Starting next year, CMS will require providers to use FHIR-enabled systems.
Automating the process of auditing and cleaning data. Much of the data in EHRs and other clinical systems, though not all, is entered by humans, and is subject to errors, omissions, and inconsistent entry practices. Without excellent underlying data, digital measures will have no value.
Along with creating the infrastructure to use digital information, every health care stakeholder has its part to play:
- The quality measurement community needs to intensify and expand its efforts to determine which new data elements are the most important for identifying best practices and explaining variations in outcomes.
- Both hospitals and insurers harbor legacy computer systems that struggle to support the need to exchange data with other systems. They need some combination of upgrades, application of standards, or workarounds in order to serve the new needs of digital measurement.
- Physicians and hospitals are still primarily paid on the basis of care volume rather than care quality, which reduces their motivation to reengineer their approach to care delivery. Both providers and payers must embrace data-driven payment models based on effectiveness and value.
- Since employers and government pay for the vast majority of health care, they have a critical role to play in using their clout (e.g., contracts and their ability to move their provider and health plan business elsewhere) to demand that providers, health plans, and the quality measurement community accelerate the development and adoption of digital quality measures. In addition, employers and governments could use their talents to help the industry understand how they will use the measures to enhance their health-care-benefit offerings, and their staff should participate in forums that define health data standards and appropriate uses of data.
- These insights need to be easily available to patients in a way they can interpret and evaluate as they make decision about their health and health care.
The Impact of Digital Measures
What would it mean to be able to harness this overwhelming mass of data to measure and manage the quality of our health care?
Providers could more accurately and effectively assess and improve their performance. They would catch the patients due for screenings, manage the patients whose chronic illnesses land them in the hospital periodically if they’re not managed, and maybe even head off some of those chronic illnesses with strategically applied attention and education.
Patients could make better choices for themselves and their families. They could find the best care by employing the same digital methods that now suggest where they should have dinner or get their oil changed.
Insurers and employers could refine health benefit coverage to better serve the needs of their employees and members, pay for services proven to keep them healthier, and identify the best providers for those services. And they could do it in real time, or close to it, instead of relying on data from last year.
In short, health care could become the same kind of data-driven powerhouse as retailing or financial services — except in the service of saving lives and keeping everyone healthy.
John Glaser is an executive in residence at Harvard Medical School. He previously served as the CIO of Partners Healthcare (now Mass General Brigham), a senior vice president at Cerner, and the CEO of Siemens Health Services. He is the co-chair of the HL7 Advisory Council and a board member of the National Committee for Quality Assurance.
Margaret O’Kane is the founder and president of the National Committee for Quality Assurance, a nonprofit organization in the United States that works to improve health care quality through the administration of evidence-based standards, measures, programs, and accreditation.
Brad Ryan, MD, is the chief product officer of the National Committee for Quality Assurance, a nonprofit organization in the United States that works to improve health care quality through the administration of evidence-based standards, measures, programs, and accreditation.
Eric Schneider, MD, is the executive vice president who heads the Quality Measurement and Research Group at the National Committee for Quality Assurance, a nonprofit organization in the United States that works to improve health care quality through the administration of evidence-based standards, measures, programs, and accreditation.