Healthcare.AI™ Product Suite from Health Catalyst

Failing to use SPC means making errors both in not detecting improvement when it has happened and in concluding improvement has happened when it has not. Implementing even basic SPC techniques typically means using different and daunting tools. Spreading the tools or using more recent developments has been almost impossible for most organizations.

Within minutes, your analysts can click a button to apply SPC within any BI tools that enable you to click a button to insert a line chart. Algorithms automatically select everything from time periods (e.g., weekly, monthly, etc.) to SPC chart types (e.g., i-chart, p-chart, o:e-chart, etc.). We augment traditional rules with AI algorithms to detect change over time. Expert users can interrogate and override algorithmic choices.

Business leaders and analysts will draw more accurate and consistent conclusions more quickly and transparently, reducing time to insight by months and building confidence to lead change or sustain improvement.

A bewildering array of techniques exist for trying to identify change in time series data. Often this leads to using nothing at all (e.g., “I’ll know it when I see it”) or inappropriate method selection (e.g., linear regression).

We make a variety of time series outlier detection algorithms available directly in your BI and other standard analytics tools. Analysts and decision makers get to shift their discussion from “which tool should we use?” to “how do we weigh early notification against the possibility of a false alert?” The algorithms can run in the background against thousands of measures and levels of granularity across the system. At any point, any person can look at the algorithm’s outputs to determine if they wish to alter the triggers.

Business leaders and analysts can leverage data to monitor vastly more aspects of their business with much more granularity. This serves the dual purpose of reacting more quickly to positive and negative outliers but also having the confidence to retain focus elsewhere knowing the system has your back.

Forecasting techniques are well studied and tested but remain difficult to implement in standard analytic tools. This leads to “rear window” decision making.

We provide a set of forecasting techniques within standard BI and other analytic tools that allow both fully automated forecasting and direct control of the parameters.

Business leaders and analysts can focus more on performance trajectory and achievement rather than a single, historical point in time.

Using statistical power analysis usually requires learning both statistical techniques and a statistical programming language. The bar for including these tools in usual decision making is very high.

The most common statistical power analysis techniques are embedded directly into BI tools and specifically into reports and analytics where they are most likely to be useful. For instance, if you would like to know how long it will take to detect a 10 percent improvement compared to a baseline forecast, this is a button click away in your BI tool of choice.

Business leaders and analysts can set goals, expectations, and accountability targets based on rigorous and transparent methods. Doing so is both more likely to result in actual improvement and motivating to those responsible for the change.

Analytic techniques are hard enough to deploy in isolation. System decisions require bringing many tools together at the same time. Doing so often exceeds staff knowledge, tool integration, and computational ability.

Wherever you can insert a bar chart, a forest plot is now one click away. A bar chart can only show you nominal performance at a fixed point in time. A forest plot uses statistics, machine learning, and forecasting to show you real performance differences, highlight clusters of similar/dissimilar performance, and identify where performance is most likely to be in the future.

Business leaders and analysts can understand system performance and direction in ways they could not before. This routinely leads to insights that were otherwise hidden in plain sight. These insights have changed the nature of discussions and focus from department managers to Boards of Directors.

Only reporting data elements such as counts of positive COVID-19 cases does not inform decisions of whether data are changing or whether there is cause for concern or relief.

We augment raw data with algorithms to calculate and detect changes in measures of spread. We can embed results directly into standard data loading scripts or BI reports and configure selective alerting to ensure that decision makers have the information they need when they need it.

Community and hospital leadership teams have the information they need to make decisions ranging from (re)establishing situation management teams through specific decisions to alter services and supply chain management.

Dashboards of a few simple measures, perhaps with current performance compared to last year, appeal to our human need for simplicity. They also throw away most of our data. How likely is it that we could choose a single, best indicator for diabetes management, let alone all our other clinical, operational, and financial measures? How do we choose between year-to-date and rolling quarters as an indicator of progress, stability, or trouble? What is the appropriate level within an organization to report?

Population health managers can use AI algorithms to find and tailor outreach to members. Healthcare leaders can use those same algorithms to find and tailor interventions. These algorithms ingest hundreds or thousands of measures over time and at different levels of the organization to help you find patterns in what would otherwise be an impossible task or set of arbitrary (and likely wrong) decisions.

Business leaders and analysts can gain insight across a previously unusable spectrum of data to ensure that focus is applied correctly.

Once people agree the inevitable data quality issues have been resolved, the next obstacle to accepting the need for change is “my patients are sicker” or “my market is different.” If these concerns are correct, they need to be dealt with. If these concerns are incorrect but not addressed, sustainable change will not occur. Either way, we should remember Churchill: “I am always ready to learn, although I do not always like being taught.”

Peer finding helps guide risk adjustment and stratification so that quantitative comparisons make sense. We do so in a transparent way to build trust—different from Care Compare. We go a step further and leverage AI to help identify peers at the individual, department, or organizational level to foster learning and facilitate change.

Business leaders or analysts can derive purpose-driven benchmarking results either within or across organizations.