As the COVID-19 pandemic quickly became our clients’ top priority many of them turned to us for support in tracking, managing, and reporting their growing COVID-19 population. We had emergency late-night and early morning meetings that augmented existing analytics dashboards with relevant data feeds ranging from very granular patient-level reporting to data aggregated by unit, hospital, county, and even state.
Two common examples of this were tracking ventilator usage and tracking COVID-19 test results. We added these feeds to our data warehouse and connected them to dashboards that were already in widespread use so that the learning curve for staff was minimal. By combining these new feeds with the existing data such as patient age and whether or not they were admitted from a nursing home, we were able to provide enhanced reporting for both internal staff and external / governmental entities.
As we enter our third month of COVID-19, there are some clear data management lessons we learned for our “new normal” and how to react to a possible second wave.
- Set up touchpoints: A new feed setup is not actually complete until there are plans to meet regularly to review the feed and update data handling best practices. Anything COVID-19 related is going to change — and break — and then change some more. A scheduled checkin may be the only opportunity to not only get us on the same page with clients, but for clients to get their own staff aligned. For better or worse, our insistence on a regular meeting cadence can be the primary communication pathway for client teams.
- For example, as hospitals first implemented COVID-19 plans, all elective surgeries were cancelled and COVID tests were scarce. Any patient with a pending test result was automatically considered a PUI (patient under investigation) and flagged as “COVID suspected”. But as clients are restarting elective surgeries and COVID tests are more available, not all patients with pending results are necessarily PUI and it’s a mistake to flag them as such. Some patients being tested are not suspected to have COVID-19, and the test is simply a confirmatory rule-out. For clients in this state, we need additional data to distinguish between pending results for PUI and non-PUI. Updating our data handling was easy — the challenge was getting the client aligned on best practices.
- This built-in check in is the opportunity to share updates and make sure we stay synchronized with hospital best practices. Another example is making sure scarce ventilators are not idle while demand still exists because systems are not communicating with each other.
- Automate data validation: Implement as much automating testing as possible. Setting up alerts if a scheduled feed fails to arrive is table stakes. We set up data monitors to make sure the number of COVID-19 tests per day or ventilator updates are within a preset range. Finally, we make sure the test results align to a client-defined dictionary of approved terminology and alert them if we get novel free-text that is not parsable into well-defined buckets of “COVID-POSITIVE” and “COVID-NEGATIVE”.
- When we see any unmapped test results, we reach out to the client to understand the correct data handling. Going back to the prior insight about communication, these alerts are sometimes the first or only way clinical staff gets informed that a process has changed — for example, a new lab with unique nomenclature has been added for testing that would have otherwise been underutilized or unchecked.
- Evaluate sufficient payoff: “Is the juice worth the squeeze?” Given that over time maintaining a COVID-19 feed and associated feature set will be much more work than implementing one, is the value worth the effort? Especially when considering the churn to staff associated with decommissioning a feed that turns out to be unmanageable. We need to ask if it’s worth it?
- This is challenging because restraint is rarely rewarded. But for some clients, the final COVID-19 status just had too much dependency on factors outside of data analytics. We can manage rules like “must have three consecutive negative tests after a positive test” to clear precautions, but we are unable to implement a data-driven policy when a client’s determination of “COVID suspected” includes clinical hunches by experienced frontline staff. While visual observation is an entirely legitimate data source, it’s not something that can be automated by our platform. Seeing that a patient has pending COVID-19 test results will never match the acumen of an experienced nurse.
- In some cases, we can confidently implement a scaled down version of automated reporting such as simply indicating that test results are available, but not trying to interpret them. Even this seemingly simple notification accumulates many saved minutes by not having to dive into the EHR only to discover nothing has changed.
- But sometimes it’s best to just decline a client request outright. When this happens, we’re clear about the data necessary for a successful deployment.
These strategies help us balance a flexible data handling policy with standardization appropriate for the fast-moving COVID-19 environment. We certainly hope a “second wave” doesn’t materialize but need to make sure hospitals can use their resources as efficiently as possible if it does.
Hospital IQ’s platform creates that single source of truth organizations need in order to know what has happened, is happening now and what is going to happen in the future so they can plan accordingly. Learn more on how our Perioperative, Staffing, and Inpatient solutions build a path forward during uncertain times.