Every hospital in the U.S. is being pushed to improve patient experience, health outcomes, and total costs. Not every hospital has a data analyst –let alone a team of analysts—dedicated to measuring the progress. Fortunately, the 3M Client Experience Summit provided plenty of opportunities to learn from presenters and trend-setting hospitals.
This year, for the first time, several sessions at the Summit focused on population health. Amirav Davy, senior clinical analyst at Allina Health, talked about how to “provide information that matters” in improving transitions of care.
Following his presentation, 3M met with Amirav to learn more about how Allina uses analytics to improve the delivery of healthcare. Here are some excerpts from the conversation:
What does a data analyst do?
I help support transitions in healthcare programs with data analysis, process metrics, outcome metrics, data definitions, and data mining. A data analyst is someone who can provide guidance on outcome metrics and what they mean for patient quality, cost, and experience—for the Triple Aim.
What is an example of an outcome metric?
Readmission rates are a great example. We’re using readmissions to understand the quality of care patients receive when they transition out of the hospital and into the outpatient setting and back to the community. We want to see whether patient health is being managed in a way that it doesn’t deteriorate into a return within 30 days to the hospital.
Why is it important to reduce readmissions?
It’s one thing for hospitals to avoid readmission penalties from the CMS or to reduce healthcare costs. But I think at the end of the day it’s about the patient spending more nights at home, sleeping in their own bed, and feeling well enough to be able to pursue what they love in life, rather than spending that time sick in a hospital. That’s really what it’s about.
How do you use outcome metrics to influence the delivery of care?
We look for trends. We can look for common characteristics among patients, such as diagnosis or financial class. Similarities in where patients go when they’re discharged. We take those similarities and identify what types of patients readmit versus those who are less likely to readmit.
We have embedded the metrics in interactive dashboards where users can easily go and identify where there’s opportunity for improvement. They can see opportunities to improve processes to reduce readmissions. On the population health side, we can identify which population is more commonly using the hospital for inpatient services or who is more commonly using the emergency department, and then be able to say these are the types of populations that we want to target for care management or community engagement.
What metrics are you using and how have they helped?
The metrics we use include potentially preventable readmissions (PPR), potentially preventable complications (PPC), and population-focused preventables, which include: potentially preventable admissions (PPA), emergency department visits (PPV), and ancillary services (PPS).
The right metrics provide a more focused look on the type of patients who are readmitting more often. We can identify complications that are more commonly occurring. We can also see populations we want to avoid admitting in the first place.
On the readmission side in particular, the PPR metric allows us to stratify by diagnostic group. We can see which types of patients are more frequently readmitting than others. It allows for tracking the data over time, in a risk-adjusted way, to see if we’re really making progress towards lower readmission rates.
The other key piece where the PPR metric helps is that it allows us to benchmark all the hospitals within our health system to identify which hospitals are performing better than others. We can also benchmark ourselves to know how we are performing against the state average by the data we get through state PPR reports.
In short, the PPR metric helps with the overall visibility of the metric, ability to identify rates over time, to set organizational goals over time, and to have a retrospective tool to identify the type of patients who are readmitting more often than others. In that way it has helped us redesign care:
- Interdisciplinary meetings that allow for better discharge planning
- Standard follow-up appointment within five days
- Clear recommendations from inpatient clinicians to outpatient providers so they understand what happened while the patient was hospitalized
- Standardized hand-offs with other discharges into the community (skilled nursing facilities or home health), such as a more standard process of instructions as well as providing what medications were used in the hospital and need to be managed in the post-discharge setting
What have you done with your findings?
A good example of how we have used predictive modeling to change the process of care is what we call a transition care conference. It is an interdisciplinary meeting between the care givers, the patient, and the patient’s family, identifying their goals of care and their needs after discharge. The high-risk and moderate to high-risk patients are the target patient population to have these transition care conferences, because of their resource intensity, and how time consuming it is for the care team and physicians if the patient readmits.
Historically we’ve seen that high-risk patients perform at 78 percent or above in what we’d expect in terms of their readmission rates. That is our baseline historical performance. We can measure the readmission rate for patients who have undergone a transition care conference and compare their performance against the historical benchmark. We discovered an 11 percent reduction in PPRs for high-risk patients who underwent a transition care conference. With moderate-high-risk patients, they’re 23 percent below with fewer readmissions than what the historical performance has been for that group of patients.
What would you recommend to another hospital that wanted to reduce readmissions?
In order to understand why patients readmit, a facility would have to do a full process mapping of the hand-off between the inpatient and outpatient environment. Then they would need to look at the type of patients that are readmitting, what they are readmitting for, how quickly they’re readmitting back to the facility, and the characteristics that are common among patients who are readmitting.
The PPR data is important in that context because it provides more structure to the readmission data. It provides a standardized way to measure readmissions within 30 days. It identifies the clinical chain of readmissions following an index admission in a consistent way. It helps in identifying clinically related reasons for readmission. And it screens out a lot of readmissions that many clinicians would consider to be not really preventable or related to the index admission. We can focus on those readmission events where we might be able to prevent a possible return to the hospital.
Kristine Daynes is a marketing manager at 3M Health Information Systems.