Allina Health boosts clinical operations with big data partner
The large Minneapolis health system works with Health Catalyst on developing consistent and concise data that can be used to identify and analyze medical variances across specialties.
Minneapolis-based Allina Health System works in a risky environment. Like other healthcare providers, it’s caught up in the tectonic shifts in healthcare reimbursement that require providers, under accountable care arrangements, to meet certain thresholds for higher quality and lower cost care, or lose potentially millions in reimbursements if they don’t hit those thresholds.
And, like its peers, it must rely heavily on information technology to try to identify where and when it can cut costs while patients and data flow through its operations—in Allina Health’s case, through 15 hospitals, 61 clinics and a network of more than 3,000 employed or affiliated physician offices.
Gleaning any type of intelligence from that massive, relentless data flow required Allina Health to make big, risky bets on IT infrastructure and analytic tools. But it found a vendor partner, Salt Lake City-based Health Catalyst, also willing to take risks.
Also See: Johns Hopkins Uses Big Data to Narrow Care
The vendor, which has developed a healthcare-specific data warehouse architecture, was willing to form an unusual partnership in which Allina Health committed to building its data warehouse platform and analytics infrastructure on Health Catalyst technology. For its part, the vendor put a sizable portion of its contract with Allina Health on the line through a risk-sharing arrangement that stipulates that if the health system doesn’t hit certain clinical and financial outcomes, the company forfeits revenue.
The organizations signed a 10-year agreement in January 2015, valued at more than $100 million, to combine technologies, clinical content and front-line personnel. They will rely on a joint committee that makes a list of clinical improvement projects and sets outcomes goals for those projects. About 20 percent of the contract is dependent on whether Allina Health hits those costs and quality goals.
The partnership has already paid out significant dividends: Over the past year, Allina Health hit its goal of $100 million in improved outcomes in six categories targeted by the joint committee. The list includes a $43 million improvement in productivity, $25 million in site-specific initiatives and $30 million in reducing clinical variations. The improvements in reducing variations range across the care spectrum, including cardiovascular ($10 million), neuroscience and spine ($7 million), orthopedics ($3 million), laboratory ($1.5 million) and pharmacy ($1 million).
“We saw a win-win situation where we could take some of Allina Health self-developed tools and integrate those products with our technology, and they could rely on our expertise with building an infrastructure that could meet their unique needs,” says Health Catalyst CEO Dan Burton. “The evolving model across the healthcare industry is about risk sharing, and we felt that as a partner it was time to have some skin in the game.”
The health system’s overriding information goal is pretty straightforward, says Brian Rice, Allina’s vice president of network/accountable care integration. It needs consistent and concise data to identify clinical variances across all its medical specialties and analyze how providers can bring consistency to treating patients while optimizing their clinical outcomes.
“Core to what we do is to look at the costs of production—how much does it truly cost to provide, say, a knee replacement, and how we need to change behavior to do it optimally every time while also providing the best patient experience we can,” he says. “To do so requires insights from data from our electronic health record as well as information from numerous data sources—we have 50 different source systems, including workforce productivity systems, patient experience databases, ancillary clinical applications and a wide variety of electronic health systems used at physician offices.”
It’s a massive and complex big data initiative, but it’s a healthcare big data initiative, which makes it even more of a challenge. “Healthcare data is different, that’s the simple truth,” Burton says. “The database architecture used in traditional models is too inflexible and slow in ingesting data, and requires too much data transformation—such as formatting it for specific purposes—for our market.
“In other industries, you build a data warehouse designed to manage data that has a predictable set of relationships—such as deposits and withdrawals from bank accounts—that don’t change over time. In healthcare, you’re managing a living organism that yields data with ever-changing relationships. The science changes, a patient ages, they have new conditions and family situations … the information is magnitudes more complex. In addition, you’re ingesting data that uses numerous different terms to say the same thing. For example, if you’re managing data about a pregnant mother, there might be 10 or more different ways in a source system that the gestational age of the baby is recorded.”
Health Catalyst’s data warehouse architecture is based on what the company calls a “late-binding” approach to information, compared with the more commonly used “early-binding” strategy.
In the early-binding approach used by most industries, such as banking and retail, data flowing into the warehouse is quickly related based on business rules and vocabularies. There are known dimensions of data, such as a type of product, the size and color of that product, and where it’s sold. Those data dimensions are relatively simple to identify and link, and in early-binding strategies all the relationships would be resolved after the data hits the warehouse so it can be provisioned and used to support analytics efforts.
While that approach works well when relationships are stable and predictable, healthcare is too volatile an environment for such an architecture, Burton says. Blood pressure readings, for example, require numerous variables to be taken into account, such as if the patient was standing or sitting, their age, weight, when the reading was taken, in what care setting, and more, as well as changing science around what is a “good” blood pressure reading.
At its heart, the late-binding approach is a strategy that enables organizations to move data from source systems into a warehouse without trying to transform the data upfront by changing its formatting to make it usable for specific purposes and committing it to a relationship. That transformation and binding happens later, when the specific data is needed for an analytics effort.
Health Catalyst moves data from source systems into source marts, which serve as a type of staging area for analytics. Minimal data transformation occurs until that data is then linked to a “data bus,” comprising a small number of core data elements that are common to almost all analytics use cases in healthcare (patient ID, provider ID, date and time, facility ID and so on).
Binding source mart data to these stable, core data elements enables organizations to query across disparate source system content in the data warehouse.
The next element of the architecture is subject-area marts that are created to address particular analytic use cases. An example of a subject area mart would be a diabetes registry, where data could be taken from a source mart and then associate with data across numerous, additional variables, such as eye exam history, A1c test results, medication history, age and other data points.
To utilize the information, Allina Health has devised an “open” enterprise data warehouse that enables staff to use commercial tools to analyze the information, Rice says. Administrators and even physicians can create dashboards using software from QlickView. “The tools are very easy to use and enable staff to get the data views they need to solve problems on the front line,” Rice says. “That’s eliminated bottlenecks around data access and lets staff solve even small problems by analyzing discrete data sets, instead for someone with advanced SQL skills to run queries for them.” For more complex, static reports, staff can use Crystal Reports, a business intelligence application from SAP.
The dashboards are pushed along clinical service lines and used for various purposes, including clinical workflow. For example, hospital discharge staff use a dashboard that lists patients ready to go home along with analytics that indicate which ones might be at high risk for readmission.
Readmissions, Rice says, are a focus industrywide because of the health problems and high costs of putting patients back in the hospital. When the Allina Health analytics flag those patients, specific protocols kick in, such as scheduling follow-up visits before discharge and educating family members about the immediate risks the patient is facing.
“It’s not about reducing care; it’s about cutting out inappropriate care that can be avoided if we take specific steps,” Rice says. “The success we’re having is through getting this type of information in front of the physician community and letting them act on it. Physicians are smart—they see the data and understand the problem. And they’re also very competitive, so when see data about care variations and how much we can improve by eliminating those variations, they want to perform better than their peers. Getting data into their hands gets them engaged.”
The Allina Health/Health Catalyst partnership itself is based on a level of engagement not typical in most customer/vendor relationships. Instead of bringing in all of its own technology, Health Catalyst purchased applications for predictive analytics, stroke treatment and readmissions that Allina Health had built. Health Catalyst also addressed Allina Health’s challenge with recruiting and retaining highly-skilled data scientists and architects by opening an office in Minneapolis and moving more than 60 Allina Health workers onto the Health Catalyst payroll.
“There’s a lot of competition for those types of skills, and we can offer Allina’s staff more competitive compensation through options and company equity than Allina can as a not-for-profit,” Burton says. “There was a lot of concern at first that the move was made to outsource those jobs outside the United States, but we didn’t eliminate any jobs, and those team members work exclusively on Allina Health’s infrastructure, in the same office they worked before. Since the move, our employee satisfaction scores have been very high and we’ve had extremely low turnover, so it’s been another case of win-win for this relationship.”
Health Catalyst has other risk-sharing partnerships in place and plans to use the model going forward, Burton says. In September 2015, it inked an agreement with Boston-based Partners Healthcare. Health Catalyst licensed applications and content from Partners and the organizations jointly established a population health center. The agreement also includes a risk-based contract based on outcomes improvement at Partners, and the health system has a full subscription to Health Catalyst technology.
And, like its peers, it must rely heavily on information technology to try to identify where and when it can cut costs while patients and data flow through its operations—in Allina Health’s case, through 15 hospitals, 61 clinics and a network of more than 3,000 employed or affiliated physician offices.
Gleaning any type of intelligence from that massive, relentless data flow required Allina Health to make big, risky bets on IT infrastructure and analytic tools. But it found a vendor partner, Salt Lake City-based Health Catalyst, also willing to take risks.
Also See: Johns Hopkins Uses Big Data to Narrow Care
The vendor, which has developed a healthcare-specific data warehouse architecture, was willing to form an unusual partnership in which Allina Health committed to building its data warehouse platform and analytics infrastructure on Health Catalyst technology. For its part, the vendor put a sizable portion of its contract with Allina Health on the line through a risk-sharing arrangement that stipulates that if the health system doesn’t hit certain clinical and financial outcomes, the company forfeits revenue.
The organizations signed a 10-year agreement in January 2015, valued at more than $100 million, to combine technologies, clinical content and front-line personnel. They will rely on a joint committee that makes a list of clinical improvement projects and sets outcomes goals for those projects. About 20 percent of the contract is dependent on whether Allina Health hits those costs and quality goals.
The partnership has already paid out significant dividends: Over the past year, Allina Health hit its goal of $100 million in improved outcomes in six categories targeted by the joint committee. The list includes a $43 million improvement in productivity, $25 million in site-specific initiatives and $30 million in reducing clinical variations. The improvements in reducing variations range across the care spectrum, including cardiovascular ($10 million), neuroscience and spine ($7 million), orthopedics ($3 million), laboratory ($1.5 million) and pharmacy ($1 million). “We saw a win-win situation where we could take some of Allina Health self-developed tools and integrate those products with our technology, and they could rely on our expertise with building an infrastructure that could meet their unique needs,” says Health Catalyst CEO Dan Burton. “The evolving model across the healthcare industry is about risk sharing, and we felt that as a partner it was time to have some skin in the game.”
The health system’s overriding information goal is pretty straightforward, says Brian Rice, Allina’s vice president of network/accountable care integration. It needs consistent and concise data to identify clinical variances across all its medical specialties and analyze how providers can bring consistency to treating patients while optimizing their clinical outcomes.
“Core to what we do is to look at the costs of production—how much does it truly cost to provide, say, a knee replacement, and how we need to change behavior to do it optimally every time while also providing the best patient experience we can,” he says. “To do so requires insights from data from our electronic health record as well as information from numerous data sources—we have 50 different source systems, including workforce productivity systems, patient experience databases, ancillary clinical applications and a wide variety of electronic health systems used at physician offices.”
It’s a massive and complex big data initiative, but it’s a healthcare big data initiative, which makes it even more of a challenge. “Healthcare data is different, that’s the simple truth,” Burton says. “The database architecture used in traditional models is too inflexible and slow in ingesting data, and requires too much data transformation—such as formatting it for specific purposes—for our market.
“In other industries, you build a data warehouse designed to manage data that has a predictable set of relationships—such as deposits and withdrawals from bank accounts—that don’t change over time. In healthcare, you’re managing a living organism that yields data with ever-changing relationships. The science changes, a patient ages, they have new conditions and family situations … the information is magnitudes more complex. In addition, you’re ingesting data that uses numerous different terms to say the same thing. For example, if you’re managing data about a pregnant mother, there might be 10 or more different ways in a source system that the gestational age of the baby is recorded.”
Health Catalyst’s data warehouse architecture is based on what the company calls a “late-binding” approach to information, compared with the more commonly used “early-binding” strategy.
In the early-binding approach used by most industries, such as banking and retail, data flowing into the warehouse is quickly related based on business rules and vocabularies. There are known dimensions of data, such as a type of product, the size and color of that product, and where it’s sold. Those data dimensions are relatively simple to identify and link, and in early-binding strategies all the relationships would be resolved after the data hits the warehouse so it can be provisioned and used to support analytics efforts.
While that approach works well when relationships are stable and predictable, healthcare is too volatile an environment for such an architecture, Burton says. Blood pressure readings, for example, require numerous variables to be taken into account, such as if the patient was standing or sitting, their age, weight, when the reading was taken, in what care setting, and more, as well as changing science around what is a “good” blood pressure reading.
At its heart, the late-binding approach is a strategy that enables organizations to move data from source systems into a warehouse without trying to transform the data upfront by changing its formatting to make it usable for specific purposes and committing it to a relationship. That transformation and binding happens later, when the specific data is needed for an analytics effort.
Health Catalyst moves data from source systems into source marts, which serve as a type of staging area for analytics. Minimal data transformation occurs until that data is then linked to a “data bus,” comprising a small number of core data elements that are common to almost all analytics use cases in healthcare (patient ID, provider ID, date and time, facility ID and so on).
Binding source mart data to these stable, core data elements enables organizations to query across disparate source system content in the data warehouse.
The next element of the architecture is subject-area marts that are created to address particular analytic use cases. An example of a subject area mart would be a diabetes registry, where data could be taken from a source mart and then associate with data across numerous, additional variables, such as eye exam history, A1c test results, medication history, age and other data points.
To utilize the information, Allina Health has devised an “open” enterprise data warehouse that enables staff to use commercial tools to analyze the information, Rice says. Administrators and even physicians can create dashboards using software from QlickView. “The tools are very easy to use and enable staff to get the data views they need to solve problems on the front line,” Rice says. “That’s eliminated bottlenecks around data access and lets staff solve even small problems by analyzing discrete data sets, instead for someone with advanced SQL skills to run queries for them.” For more complex, static reports, staff can use Crystal Reports, a business intelligence application from SAP.
The dashboards are pushed along clinical service lines and used for various purposes, including clinical workflow. For example, hospital discharge staff use a dashboard that lists patients ready to go home along with analytics that indicate which ones might be at high risk for readmission.
Readmissions, Rice says, are a focus industrywide because of the health problems and high costs of putting patients back in the hospital. When the Allina Health analytics flag those patients, specific protocols kick in, such as scheduling follow-up visits before discharge and educating family members about the immediate risks the patient is facing.
“It’s not about reducing care; it’s about cutting out inappropriate care that can be avoided if we take specific steps,” Rice says. “The success we’re having is through getting this type of information in front of the physician community and letting them act on it. Physicians are smart—they see the data and understand the problem. And they’re also very competitive, so when see data about care variations and how much we can improve by eliminating those variations, they want to perform better than their peers. Getting data into their hands gets them engaged.”
The Allina Health/Health Catalyst partnership itself is based on a level of engagement not typical in most customer/vendor relationships. Instead of bringing in all of its own technology, Health Catalyst purchased applications for predictive analytics, stroke treatment and readmissions that Allina Health had built. Health Catalyst also addressed Allina Health’s challenge with recruiting and retaining highly-skilled data scientists and architects by opening an office in Minneapolis and moving more than 60 Allina Health workers onto the Health Catalyst payroll.
“There’s a lot of competition for those types of skills, and we can offer Allina’s staff more competitive compensation through options and company equity than Allina can as a not-for-profit,” Burton says. “There was a lot of concern at first that the move was made to outsource those jobs outside the United States, but we didn’t eliminate any jobs, and those team members work exclusively on Allina Health’s infrastructure, in the same office they worked before. Since the move, our employee satisfaction scores have been very high and we’ve had extremely low turnover, so it’s been another case of win-win for this relationship.”
Health Catalyst has other risk-sharing partnerships in place and plans to use the model going forward, Burton says. In September 2015, it inked an agreement with Boston-based Partners Healthcare. Health Catalyst licensed applications and content from Partners and the organizations jointly established a population health center. The agreement also includes a risk-based contract based on outcomes improvement at Partners, and the health system has a full subscription to Health Catalyst technology.
More for you
Loading data for hdm_tax_topic #better-outcomes...