Fixing Data Overload in Health Care
March 16, 2022 at 12:00:00 AM
Many health care organizations are plagued by data overload. The result is their boards, leaders, and managers don’t have the data they need to identify areas in need of improvement and track progress in addressing them. A five-step approach, however, can solve this problem. It was derived from work with thousands of health care organizations.
Harvard Business Review
Deirdre E. Mylod and Thomas H. Lee
Health care leaders are both addicted to and overwhelmed by data. In search of simplicity, leaders at the top of organizations often look for one or two measures that summarize everything — and then find that those measures do not guide their managers in their efforts to improve their particular function.
These organizations need an enterprise-wide data strategy that can overcome these challenges and get managers the right data with the right analyses at the right time for them to translate the information into action. In our firm’s work with thousands of health care organizations, we have found that these strategies are best pursued by taking five steps.
1. Segment the data’s consumers
The first step is identifying “customers” — those who will consume the data — and deciding how value will be created for them. These customers can be grouped based on their level in the organization and the type of measures that match their responsibilities.
Boards and CEOs need summary data on measures. For example, they should see one summary measure of safety for the overall organization rather than rates of every type of adverse event. And they should track a summary measure of patient experience (e.g., percent of patients who gave their caregivers a top rating when asked the likelihood that they would recommend them to others) but not measures for individual service lines. They need this data at regular intervals to monitor progress in achieving strategic goals, and they need benchmarking data to understand their relative performance in the market.
Senior leaders need more granular information that connects high-level outcomes with the key process that influences those outcomes. For example, they need data on the different types of safety problems and the specific drivers of patient experience. They need this information more regularly than board level updates, and they need data that’s segmented to reflect the performance of particular geographies or service lines.
Frontline staff need feedback about performance that is directly relevant to the behaviors expected of them — such as feedback that patients are concerned about response time when they press a call button for help or that likelihood of patient falls is higher on a particular unit. They need such feedback frequently and quickly, so they can seize opportunities to reduce risk.
2. Figure out how to use the data to create value
The next step is to determine how the organization will create value for each level of the customers of the data. Simply providing a “data dump” is not enough for any customer. Instead, the chain of activities for value creation should be based on doing these eight things:
• Select your key performance indicator (KPI), or top-level metric, and the data format you will use to track your performance at the highest level (e.g., patient ratings of coordination of care)
• Know your status, i.e., your current level of performance, usually expressed as percentile compared to that of the relevant benchmarking group
• Track your trend in performance and understand your trajectory (e.g., worsening while others are improving)
• Identify priorities (i.e., the key drivers, behaviors, and practices) to focus on (e.g., joint rounding of physicians and nurses)
• Investigate variation in performance across subgroups (e.g., site, unit, provider, patient cohort)
• Set goals that align with your strategy and are challenging but realistic, ideally incorporating benchmarking into the goal metric so that you know you are maintaining your relative performance against peers even as external factors such as Covid-19 impact national trends
• Take action with your data by setting expectations for review and response (e.g., that all units whose performance is below the median will develop and implement an improvement plan)
• Be consistent in how outcomes are reported by connecting the metrics to be reported to your board with your larger strategy and sharing them regularly (e.g., using a balanced scorecard)
3. Integrate data to generate more insights
Within health care, outcomes measures tend to be grouped and managed by leaders who have expertise in specific areas: a chief safety officer focuses on safety-related adverse events, a chief experience officer focuses on patient feedback, a head of HR focuses on staff engagement and retention, a chief medical officer focuses on brand, and so on. As a result, the natural state is for measures to be siloed based on the category of outcome rather than connected across the enterprise to represent holistic performance. Three types of integration efforts are necessary to get full value out of data relevant to any category of quality.
Level 1: Integrate the full complement of data within a category of quality performance within one patient care setting. Health care leaders are problem solvers. When a metric indicates that their organizations is underperforming, the first inclination is to immediately create an action plan. Before doing so, it’s important to make sure you have the full story or context so you truly understand the current state.
An organization focusing on staff engagement might be interested in improving engagement in the hospital’s emergency department (ED) after noticing that its staff’s response in the most recent annual engagement survey was that their intention to remain at the organization was low. But leaders should look beyond a single summary score to assess the specific features in the ED environment that influence staff wellness and resilience such as how employees feel about the ED’s commitment to safety and the inclusiveness of the ED’s culture. Both quantitative and qualitative data from multiple sources (e.g., formal surveys, pulse surveys, leader rounding, and social media) should be brought together to paint a full picture of this staff group’s needs.
Similarly, hospitals focusing on improving safety should go beyond data on outcomes like inpatient falls or the rates of central-line-associated bloodstream infections (CLABSI) and also assess the overall safety culture as reported by staff working in the inpatient care setting. And organizations working to enhance the patient experience in a clinic setting should look beyond global measures such as their likelihood to recommend the practice to others and review the extent to which various patient needs such as communication and coordination of care are being met. Comments and other narrative data can be analyzed with artificial intelligence and natural language processing to extract insights that might not be captured by structured surveys.
In each of these examples, the key takeaway is that organizations should do more than track high-level KPIs. Instead, they should paint the full picture of performance for any specific issue (e.g., safety, employee engagement) using all sources of quantitative and qualitative data.
Level 2: Integrate data within a category of quality performance across settings. Organizations that have identified areas in need of improvement must understand whether the issue is relevant to just one site or population of patients or is part of a larger pattern across the organization. If engagement is low among physicians, is it also low among nurses and non-clinical staff? When performance issues are found to be systemic, so must the efforts to address the root cause. On the other hand, if only one patient care unit is exhibiting lower performance, much more focused support tailored to the unique context of care would be necessary.
Level 3: Integrate data across all categories of quality performance within one patient care setting. There are often correlations among the different kinds of performance. For example, organizations that have excellent safety culture are more likely to have a highly engaged workforce and provide a first-rate patient experience. These correlations suggest that “cross-domain analytics” that look at multiple performance issues are often valuable for addressing challenges that cut across them.
Intermountain Healthcare took such an approach with its work to improve equity in outcomes across racial and ethnic groups by bringing together data on safety, quality, operations, and patient experience. Experienced managers understand the interrelationships among the culture, safety performance, quality record, patient experience, and efficiency of a particular part of the organization (e.g., a specific patient unit). Integrated data allows these interrelationships to be detected and analyzed systematically.
4. Establish priorities
At this point, the explicit goal should be to establish priorities, so that managers at all levels can consider the question, “Which three things should I focus upon?” This approach acknowledges that not all issues are of equal importance to the overall organization, and that the highest priority issues may vary in different settings — a fact of life that emphasizes the need for a data strategy that does not try to overly simplify organizational needs. For example, patient experience may be considered an enormous concern in the outpatient setting and improving safety the highest priority for inpatient care. Of course, safety and patient experience matter in every context of care delivery, but this framework acknowledges that leaders or managers may invest time and resources differently in different settings.
The goal of this work is to identify the priorities that are relevant at key units of analysis (e.g., service lines or other clinical programs that are especially critical to organizational success). After identifying the issue and the location that are the focus of analysis, the next step is to uncover key drivers of performance for those locations, establish the priority measures you will seek to influence, and select the best practices to implement.
For example, Houston Methodist has streamlined the patient experience data it shares with clinicians in the practice setting by focusing only on the top three drivers of patient outcomes of trust and loyalty (i.e., whether patients felt that their doctors listened carefully to them, showed respect for what they had to say, and knew important information about their medical histories). Robert Phillips, its executive vice president and chief physician executive, refers to this approach as “success in a minute” because caregivers have a shared knowledge of the most critical behaviors needed to create an environment of trust with their patients.
5. Provide information in the format most helpful to its users
While boards and CEOs are looking for report cards with red, yellow, and green coding to highlight strengths and problem areas for the overall organization, senior managers want figures with trend lines that reveal whether the most recent results reflect performance is improving, worsening, or normal variation. Graphic displays of real-time data (as opposed to spreadsheets of numbers) can help frontline managers identify problems as they arise.
How will organizations know if their data strategy is working? Leaders and managers feel that they are getting the data they need to set priorities, make decisions, and track progress. They aren’t wasting time searching for insights amid a numbing array of numbers. Their investment in data collection generates returns in the form of new insights that are surfaced on a regular basis. And they spend more time discussing problems with performance and less time discussing problems with data.