The power of artificial intelligence will revolutionize patient care and operational efficiency, ushering in a new era of medical advancement. Yet, beneath the surface of this promising landscape lies a fundamental truth: AI's potential hinges upon the availability and quality of data. Without first focusing on the less glamorous—though foundationally important—priorities related to building, maintaining and leveraging good data architecture, the grand vision of AI-driven healthcare will remain an elusive dream. This transformation will take place as leaders effectively structure and utilize existing data to enhance patient outcomes and expedite return on investment.
The Problem—Data Quality and Utilization
Focusing on data quality is tedious work that lacks the allure of flashy AI tools, but optimizing patient outcomes while maintaining operational efficiency rests on this essential bedrock.
Amid the excitement and promise of AI, the gap between available data and its optimal utilization remains vast—Deloitte estimates that up to 97% of all data produced by hospitals remains unused. This gap inhibits the potential for immediate improvements and casts doubt on the feasibility of AI-powered healthcare. Put simply, failing to meet foundational data requirements diminishes the potential of AI-powered tools to create value.
To address this critical need, we provide several steps healthcare executives can take to prepare their organizations for sustainable AI adoption.
The Proposal—Boosting Patient and Operational Outcomes With High-Quality Data
Healthcare executives can capitalize on the potential of existing data to unlock a multitude of benefits, from heightened patient satisfaction and swift ROI to cost-effectiveness that can potentially transform healthcare. By combining insights gleaned from personal experience with a thorough analysis of existing data, organizations can implement operational changes that rapidly yield tangible ROI and substantial enhancements in key performance indicators such as patient outcomes and satisfaction scores.
The data-driven journey at Omada Health, as an example, brings crucial insights to light.
Meeting the Challenge
Omada Health is a virtual-first healthcare organization, offering services for individuals managing cardiometabolic and musculoskeletal conditions. At Omada, each service line works cross-functionally with data science, product design and engineering to leverage data to improve patient outcomes.
In late 2022, leaders of Omada’s MSK service line, which provides telehealth physical therapy, conducted an internal analysis of data to identify opportunities to improve patient outcomes. Prior analysis of Omada’s data demonstrated that follow-up visits were associated with increased exercise completion, which was associated with improved patient outcomes. Based on clinical experience, the leaders hypothesized that a follow-up visit within a week of the initial consultation would lead to improved patient engagement and outcomes.
A simple analysis confirmed this hypothesis, revealing that patients who completed a follow-up visit within eight days of their consultation were more likely to reach the thresholds for minimum clinically important differences in pain and function and report high satisfaction. Based on this analysis, Omada set a goal to improve the number of patients who have a follow-up visit within eight days by 50% within one year.
Omada then performed an analysis of “asynchronous assessments”—a feature that allows patients to send their physical therapist videos of themselves performing specific exercises—and found utilization of this tool also was associated with improved outcomes. A target was set to increase utilization by 30% over 12 months, and providers were given access to data about their patients’ utilization of the tool. In addition, the provider workflow was updated to make it faster and easier to assign asynchronous assessments and schedule follow-up visits.
Results
In the six months following these interventions, the goal to improve asynchronous assessment utilization by 30% was met, and the frequency at which follow-up visits occurred within eight days more than doubled. The Net Promoter Score for the MSK service line increased from 72 to 78, and patient outcomes for pain improved by 7%. Finally, presenting providers with patient outcome data strengthened their support for leveraging follow-ups and asynchronous assessments. According to internal survey data, 92% of providers reported that they found significant value in the use of asynchronous assessments. Similarly, 88% of providers reported making follow-up visits within eight days for the majority of their patients.
Steps for Building a Data-Driven, AI-Ready Culture of Care
These results highlight the importance of building a data-driven culture to create meaningful change, boost AI readiness and ensure ROI. It starts with leadership investing in personnel, tools, training and data architecture, including appropriate privacy and security controls, which enables the organization to extract value from its data. For example, this means making KPIs such as NPS and MCID available and personalized for staff so they can be aware of their own performance and so their managers can regularly offer actionable feedback.
High-quality data should be used to monitor performance, set goals, measure progress against those goals and ultimately drive improved performance across the organization. Achieving this level of performance hinges on leadership’s ability to communicate the significance of data and secure buy-in from individuals at all levels of the organization.
Unfortunately, few organizations are realizing the full potential of their existing data, which could be used to drive AI-powered initiatives and deliver better care. To overcome this hurdle, we developed 10 actionable KPIs that healthcare executives can use to locate, capture, refine and communicate the value of their organization's data (see Table 1). The framework is based on personal experience, proven best-practices such as Airbnb’s Data University and standards such as the UK’s Data Maturity Assessment for Government.
KPIs for Creating a Robust Data-Driven Culture
| Focus | Key Performance Indicator | Objective |
|---|---|---|
| Leadership | Percent of Senior Leadership With Data Skills | Set a desired value for the percentage of senior leaders who possess adequate data and analytical skills through training, hiring and data enablement. |
| Leadership | Data Literacy Promotion Score | Set a desired promotion score by conducting regular surveys to measure senior leadership’s perceived value and promotion of data literacy. |
| Leadership | Data-Informed Decision Ratio and Perceived Utilization Score | Set a desired value for the percentage of major decisions that are explicitly based on data, and establish a desired subjective score by conducting a survey of leadership’s perceived rate of data usage. |
| Leadership | Data Utilization Rate for Employee Performance | Evaluate the use of high-quality data in monitoring and improving employee performance, and set a value for the percentage of performance assessments and promotions based on data. |
| People | Data Accessibility Score | Set a desired accessibility score by conducting regular surveys to measure staff’s perceived access to necessary data, while protecting privacy and security. |
| People | Staff Data Literacy Index | Set desired scores for staff data literacy through regular objective tests of data literacy and subjective surveys of perceived ability. |
| People | Data Integration Score | Set a desired integration score by conducting regular surveys to measure staff’s perception of data integration with overall business strategies. |
| Processes & Systems | Data Privacy & Security | Track the number of incidents, policy violations (e.g. inappropriate access to PHI), or near-miss educational opportunities related to data privacy and security to evaluate areas where risk needs to be remediated. |
| Processes & Systems | Data Collection & Storage Efficiency | Measure the efficiency of data collection and storage processes and set a desired percentage target for a reduction in data collection time and a desired percentage target for data storage efficiency. |
| Processes & Systems | Patient Feedback Utilization Rate and Quality Score | Set a desired percentage for the objective use of patient feedback used to drive improvements in products or services and establish a subjective quality score by surveying patients’ perceived satisfaction with data-driven improvements. |
|
Note. Leaders should personalize these performance indicators to best suit their needs. |
||
Healthcare organizations have mountains of data available that can be leveraged to improve patient outcomes and efficiency, but extracting value from this data requires investment in building a solid foundation and ensuring processes protect the privacy and security of patient data. Employing the above KPIs should allow organizations to effectively leverage existing data across all levels of the organization. As demonstrated by Omada, this data-obsessed approach can yield meaningful results in as little as six months, and it’s a prerequisite for organizations to maximize any future investments in AI.
Todd Norwood, PT, DPT, is senior director, Clinical Services, Omada Health, San Francisco, where he leads a diverse nationwide team of physical therapists. Brian R. Spisak, PhD, is an independent consultant, a research associate at the National Preparedness Leadership Initiative at Harvard University and the Harvard T.H. Chan School of Public Health, and a faculty member at the American College of Healthcare Executives.