HDM Feature: Assembling an Analytics Team
Providers of all stripes are trying to get analytics teams together, and there are recipes for success and disaster.
Analysis is nothing new in the science and data-heavy health care field, but what used to be a couple guys crunching numbers for Excel sheets and calling it "analytics" has become a high-profile, risky and expensive adventure aimed at the heart of the business of health care organizations. If you don't know your data in this stupendously complex market, you don't know your business. And enormous changes in payments and clinical quality standards mean organizations that don't know their business are at real of risk of not having one.
Intense, enterprise analytics has come about suddenly; most providers that have assembled analytics teams have done so recently, and in that quick ramp-up have learned some hard lessons about how to put together the structure and staff they need to provide actionable intelligence to financial and clinical business owners.
Smooth sailing it hasn't been. Executives who have spearheaded team-building efforts have struggled to find the right place for analytics teams to reside in the structure of their organizations. The work sometimes requires them to get beyond cultural and political obstacles and create something entirely new. And to a person, they have continual concerns about finding the right skill sets for their teams and fine-tuning the mix of experienced hands, greenhorns, communicators and programmers.
But in some cases, the success of analytics efforts depends on the hand you're dealt, says John Hendricks, the chief technology officer at Henry Ford Health System, Detroit. A big factor in predicting the odds of success is how comfortable an organization is with centralizing assets. Hendricks has first-hand experience on both sides of the coin. Henry Ford centralized many operations when he came on board earlier this year, and pulling together an enterprise analytics group was relatively easy. But at his previous post at an Iowa-based health system, few operations were centralized, and when Hendricks tried to put an enterprise team together some regional leaders slammed the brakes hard on his idea, he says.
First things
The good thing is that analytics is not an especially hard sell to the board room. Most C-suite executives feel the pressure of having too many clinical and financial unknowns in their operations and are willing to support enterprise analytics efforts, executives say. But that support can dry up quickly if the C-suite gets behind an effort and then doesn't see deliverables from those investments quickly, notes John Walton, solutions manager at CTG Health Solutions, a health care consultancy. "We've seen a lot of instances where there's a lot of enthusiasm and investment in analytics programs, but then it dies on the vine fairly quickly because it's just another bunch of reports and the whole effort gets overwhelmed by the sheer number of other initiatives," he says. "The reason that happens is because, right from the beginning, those initiatives didn't have a chief medical officer or CMIO evangelizing about it, and didn't set up a governance structure that supported an effort of that magnitude."
Something else many failed efforts often have in common? They committed what many now see as a cardinal sin when assembling analytics teams: They planned to have enterprise analytics reside in the I.T. department and report up to the CIO. "That's a recipe for disaster, there's really no other way to say it," Walton says.
The tricky part there is that the nuts and bolts of analytics efforts are data warehouses and data marts, which have been owned by the I.T. department. But Joe Kimura, M.D., medical director of analytics at Atrius Health, says that keeping analytics in I.T. creates degrees of separation from the business issues.
"Analytics used to live under the I.T. department here, but we moved it under the medical director because we wanted to reduce the cycle times for reports and get the group closer to our strategic plans and enterprise initiatives," he says.
"Here's the thing: the I.T. department is usually not the first department to hear about clinical initiatives around diabetes or other quality efforts, but that's exactly where analytics has to be focused. We have the analytics and business requirements up-front, which connects it much more tightly with the business."
The analytics team at Boston-based Atrius Health-which comprises six large practices with 1,000 physicians and 5,000 total employees-reports up to Kimura. It is separate from the data warehouse team, which builds the data marts and maintains the enterprise warehouse at the delivery system.
Team focus
The data warehouse team reports up through the CIO. Kimura's team is focused on enterprise analytics; each practice has a few analytics experts working on specific reports for those practices. Those analysts are not part of Kimura's group but are part of regular meetings about improving and maintaining the shared warehouse. In all, there are about 50 analysts at Atrius Health; 13 of those analysts are in Kimura's group.
"Our medical groups have their own analysts who have their own fiefdoms, but we're all hitting on the same shared database," Kimura notes.
When struggling with the decision on where to put analytics in the corporate structure, leaders should ask themselves a simple question: When bad data occurs, who pays the costs? The answer is ... not the information technology department. The observation should give organizations a clue to what direction to take, says Peter Aiken, associate professor at Virginia Commonwealth University's School of Business Information Systems Department and the owner of Data Blueprint, a data management and I.T. consulting firm.
"It's the business that pays the price of bad data, so the emphasis should be on getting analytics as close to the business as possible," Aiken says. "While any analytics effort needs those database experts to work closely with them, this isn't an I.T. project."
Aiken, like Walton from CTG Health Solutions, has seen many efforts go awry for this exact reason. But new strategies haven't necessarily been better. Aiken says in the past couple years he's seen a cross-industry trend of separating the database and extract, transform, load (ETL) specialists in I.T. from the analysts in an effort to move the initiatives partially out from under the I.T. umbrella. But without strong leadership and an emphasis on collaboration, the strategy often has yielded the same results as keeping them in I.T.-a disconnection from business owners and no front-row seat for enterprise planning.
Better strategy
A better strategy, he says, is to move both those groups out of information technology. "You can't separate the groups because you need the natural synergy that's there, but you have to establish this is about data and not a purely technological issue."
To do so requires executives willing to break new ground, a place where it's not easy to figure out how the pieces fit together. In September 2011, massive HCA Corp., based in Nashville, Tenn., created an enterprise, analytics team led by a director of clinical data management and a counterpart from the I.T. and services division.
Anna Daly, the director on the clinical side, says there was some "role confusion" when the team was first formed, but the effort had very strong support from the C-suite. "This is part of our larger clinical data management program, where we decide what our business needs are and how we need analytics to bear on those needs."
HCA, which operates 164 hospitals and hundreds of ambulatory facilities, didn't want to use the traditional ad hoc strategy for analytic projects, pulling people out of different departments for one-off projects, Daly says.
While that still occurs, the core team on an analytics project is pulled from the group. Another significant decision around that analytics team was to put all 70-plus members in one office, which required some team members to re-locate but created more esprit de corps and cross-pollination.
"You don't have to schedule a meeting. You don't have to wait a couple days. Our team is sitting right next to each other, and the analysts can reach over and touch the database administrators and data modelers," Daly says. "That interaction and collaboration is crucial when you have to match up complex thinking and programming." The success of the analytics team has led HCA to use the model for the development of a consumer/patient portal for the health system.
Sniff test
Yet another crucial factor that's moving analytics teams away from I.T. is market shifts. These underlying changes are making it ever more important that the "consumers" of analytics-typically physicians-get reports that make clinical sense and map out a way to apply those findings to the practice of medicine, says Shawn Griffin, M.D., chief quality and informatics officer at Memorial Hermann Physician Network.
The network comprises more than 3,500 physicians in the greater Houston area. Many of those practices are shooting for medical home certification from the NCQA and restructuring their operations for population health management and accountable care. Memorial Hermann Physician Network as a result has been on a care manager hiring binge the past 18 months, adding more than 20, and setting up a core analytics group plus bringing in outside help from health I.T. vendors.
Griffin says the problem for many analytics efforts is straightforward-they forget that someone has to look at the data. Teams can bring all kinds of data-crunching firepower to bear, but if they don't make clinical sense, they're essentially worthless-even if technically accurate. "People who process data don't see it from a clinical standpoint. It has to be vetted by someone with a clinical eye, because if you send something to a clinician that simply doesn't make sense, they're going assume you simply don't know what's going on. Reports have to pass our clinical sniff test before we send it out."
People, people
Carving out a niche for an analytics team can be a grueling battle. But then comes the really hard part-finding the right people. And if your organization is struggling to get the right players in the analytics positions, take heart in that you're not alone-everyone has staffing challenges. Everyone.
"It's the same story all over, be it in the finance or logistics or health care industries-everyone is begging us to help them find data scientists and analysts," says Aiken, the VCU professor and consultant. "I think the industry has become pretty good at recruiting I.T. talent, but what's really needed for analytics programs is people who have a multi-disciplinary backgrounds and aren't just loyal to the I.T. portion of the job-technology people tend to stay loyal to the technology and not necessarily interested in the data. Many do eventually become interested, but that's only after 10 years or so in the industry and then they have an 'aha' moment and get really interested in the data and understand when they get the data right, everything else becomes easier."
Health I.T. vendors often can ease the pain by delivering technology and expertise. But analytic leaders warn that while solution vendors can help, they can't pull all the weight. "I've been a part of both builds and buys, and my experience has been that vendors can take you pretty far down the analytics path, but they can't take you as far as you need to go," says Hendricks from Henry Ford. "Organizations know their data and have to be in the driver's seat. So the mix of skill sets might change if you buy off-the-shelf, but you're still going to need your own highly-trained staff to work on your data."
But health care stands alone in the complexity of its data and operations, myriad nomenclatures, the number of different stakeholders and the layers of security surrounding nearly every byte of data. As a result, many health care organizations find there are never enough workers on the market who have the skill sets and temperament to fit into the health care environment, or at least do it quick enough to satisfy executives.
Data scientists
"One real issue I've seen is that health care organizations recruit data scientists and assume that, since medicine is based on science, they can make everything work," Aiken says. "But you can't recruit data scientists and expect to lay out 90-day deliverables-it takes time for analysts to figure out the business of health care. I think too many organizations have unrealistic expectations about the speed which these efforts can move."
To be a successful health care analyst requires a great deal of human skills and a big dash of patience, traits that analytics leaders say are actually harder to find than the technology experience that is typically considered a prerequisite for the job.
With the sophisticated tools available now, analytics don't necessarily need to be coding wizards; but they need to know how to communicate with team members and the clinical staff who are pushing the business forward.
Work ethic needed
"I go after biomedical degrees and workers with backgrounds in visual sciences, but I have a fantastic recent grad whose background is in advertising but has great user interface skills and an eye for detail," says Daly at HCA Corp. "You can't teach personality, and to work in this environment you have to communicate constantly. And I've also learned that if you put someone with a very strong work ethic with others with the similar trait, you can really, really move mountains."
Kimura at Atrius Health takes the same approach. The group practice federation is headquartered in Boston, where a couple large health systems are snatching up experienced analysts and competition is fierce for similarly skilled workers.
But Kimura says he's been successful in recruiting a team of freshly-minted grads and seasoned veterans by putting a premium on finding people who have a gift for breaking down technical elements for a non-technical audience and are comfortable interfacing with business users and executives, most of whom are much farther up the corporate ladder.
"We hire on four tiers, E1 being entry level and E4 an experienced health care analyst, and I can tell you that those E4s are getting harder and harder to find," Kimura says. "So we've shifted our focus somewhat on hiring E1s and E2s with the expectation that it will take a couple years to train them and get them enculturated into our environment. We're not so much concerned about how much SQL they know as whether they can deconstruct a problem and have the confidence to not blindly follow what the medical director tells them to do, to be innately curious about the data and the business issue they're working on."
Kimura looks for diverse skill sets when building an analytics team. Think big when it comes to evaluating talent, he suggests.
"This is really important when you start moving into very complex stuff like variation analytics," he says. "Certain people can hold their own in this environment, and others can't. What's interesting is that we've found people with very diverse backgrounds who have those skills and have done great work here. There's no one set of work experiences and academic training that creates a really good health care analyst."
Recruiting tool
Kimura adds that Atrius Health's data quality also provides the group with a potent recruiting tool. Atrius has a data steward responsible for ensuring all code sets are up to date with a standard nomenclature and a data quality committee that resolves conflicts with data going into the electronic health record.
"Our tools are pretty laid out and we sell the fact that it's easier to break into our organization and do more interesting analytics because analysts won't get tripped up by the intricacies of health care data.
"They can pull clean data instead of spending the bulk of their time trying to clean it up, which I think a lot of analysts at other organizations are being forced to do."
Intense, enterprise analytics has come about suddenly; most providers that have assembled analytics teams have done so recently, and in that quick ramp-up have learned some hard lessons about how to put together the structure and staff they need to provide actionable intelligence to financial and clinical business owners.
Smooth sailing it hasn't been. Executives who have spearheaded team-building efforts have struggled to find the right place for analytics teams to reside in the structure of their organizations. The work sometimes requires them to get beyond cultural and political obstacles and create something entirely new. And to a person, they have continual concerns about finding the right skill sets for their teams and fine-tuning the mix of experienced hands, greenhorns, communicators and programmers.
But in some cases, the success of analytics efforts depends on the hand you're dealt, says John Hendricks, the chief technology officer at Henry Ford Health System, Detroit. A big factor in predicting the odds of success is how comfortable an organization is with centralizing assets. Hendricks has first-hand experience on both sides of the coin. Henry Ford centralized many operations when he came on board earlier this year, and pulling together an enterprise analytics group was relatively easy. But at his previous post at an Iowa-based health system, few operations were centralized, and when Hendricks tried to put an enterprise team together some regional leaders slammed the brakes hard on his idea, he says.
First things
The good thing is that analytics is not an especially hard sell to the board room. Most C-suite executives feel the pressure of having too many clinical and financial unknowns in their operations and are willing to support enterprise analytics efforts, executives say. But that support can dry up quickly if the C-suite gets behind an effort and then doesn't see deliverables from those investments quickly, notes John Walton, solutions manager at CTG Health Solutions, a health care consultancy. "We've seen a lot of instances where there's a lot of enthusiasm and investment in analytics programs, but then it dies on the vine fairly quickly because it's just another bunch of reports and the whole effort gets overwhelmed by the sheer number of other initiatives," he says. "The reason that happens is because, right from the beginning, those initiatives didn't have a chief medical officer or CMIO evangelizing about it, and didn't set up a governance structure that supported an effort of that magnitude."
Something else many failed efforts often have in common? They committed what many now see as a cardinal sin when assembling analytics teams: They planned to have enterprise analytics reside in the I.T. department and report up to the CIO. "That's a recipe for disaster, there's really no other way to say it," Walton says.
The tricky part there is that the nuts and bolts of analytics efforts are data warehouses and data marts, which have been owned by the I.T. department. But Joe Kimura, M.D., medical director of analytics at Atrius Health, says that keeping analytics in I.T. creates degrees of separation from the business issues.
"Analytics used to live under the I.T. department here, but we moved it under the medical director because we wanted to reduce the cycle times for reports and get the group closer to our strategic plans and enterprise initiatives," he says.
"Here's the thing: the I.T. department is usually not the first department to hear about clinical initiatives around diabetes or other quality efforts, but that's exactly where analytics has to be focused. We have the analytics and business requirements up-front, which connects it much more tightly with the business."
The analytics team at Boston-based Atrius Health-which comprises six large practices with 1,000 physicians and 5,000 total employees-reports up to Kimura. It is separate from the data warehouse team, which builds the data marts and maintains the enterprise warehouse at the delivery system.
Team focus
The data warehouse team reports up through the CIO. Kimura's team is focused on enterprise analytics; each practice has a few analytics experts working on specific reports for those practices. Those analysts are not part of Kimura's group but are part of regular meetings about improving and maintaining the shared warehouse. In all, there are about 50 analysts at Atrius Health; 13 of those analysts are in Kimura's group.
"Our medical groups have their own analysts who have their own fiefdoms, but we're all hitting on the same shared database," Kimura notes.
When struggling with the decision on where to put analytics in the corporate structure, leaders should ask themselves a simple question: When bad data occurs, who pays the costs? The answer is ... not the information technology department. The observation should give organizations a clue to what direction to take, says Peter Aiken, associate professor at Virginia Commonwealth University's School of Business Information Systems Department and the owner of Data Blueprint, a data management and I.T. consulting firm.
"It's the business that pays the price of bad data, so the emphasis should be on getting analytics as close to the business as possible," Aiken says. "While any analytics effort needs those database experts to work closely with them, this isn't an I.T. project."
Aiken, like Walton from CTG Health Solutions, has seen many efforts go awry for this exact reason. But new strategies haven't necessarily been better. Aiken says in the past couple years he's seen a cross-industry trend of separating the database and extract, transform, load (ETL) specialists in I.T. from the analysts in an effort to move the initiatives partially out from under the I.T. umbrella. But without strong leadership and an emphasis on collaboration, the strategy often has yielded the same results as keeping them in I.T.-a disconnection from business owners and no front-row seat for enterprise planning.
Better strategy
A better strategy, he says, is to move both those groups out of information technology. "You can't separate the groups because you need the natural synergy that's there, but you have to establish this is about data and not a purely technological issue."
To do so requires executives willing to break new ground, a place where it's not easy to figure out how the pieces fit together. In September 2011, massive HCA Corp., based in Nashville, Tenn., created an enterprise, analytics team led by a director of clinical data management and a counterpart from the I.T. and services division.
Anna Daly, the director on the clinical side, says there was some "role confusion" when the team was first formed, but the effort had very strong support from the C-suite. "This is part of our larger clinical data management program, where we decide what our business needs are and how we need analytics to bear on those needs."
HCA, which operates 164 hospitals and hundreds of ambulatory facilities, didn't want to use the traditional ad hoc strategy for analytic projects, pulling people out of different departments for one-off projects, Daly says.
While that still occurs, the core team on an analytics project is pulled from the group. Another significant decision around that analytics team was to put all 70-plus members in one office, which required some team members to re-locate but created more esprit de corps and cross-pollination.
"You don't have to schedule a meeting. You don't have to wait a couple days. Our team is sitting right next to each other, and the analysts can reach over and touch the database administrators and data modelers," Daly says. "That interaction and collaboration is crucial when you have to match up complex thinking and programming." The success of the analytics team has led HCA to use the model for the development of a consumer/patient portal for the health system.
Sniff test
Yet another crucial factor that's moving analytics teams away from I.T. is market shifts. These underlying changes are making it ever more important that the "consumers" of analytics-typically physicians-get reports that make clinical sense and map out a way to apply those findings to the practice of medicine, says Shawn Griffin, M.D., chief quality and informatics officer at Memorial Hermann Physician Network.
The network comprises more than 3,500 physicians in the greater Houston area. Many of those practices are shooting for medical home certification from the NCQA and restructuring their operations for population health management and accountable care. Memorial Hermann Physician Network as a result has been on a care manager hiring binge the past 18 months, adding more than 20, and setting up a core analytics group plus bringing in outside help from health I.T. vendors.
Griffin says the problem for many analytics efforts is straightforward-they forget that someone has to look at the data. Teams can bring all kinds of data-crunching firepower to bear, but if they don't make clinical sense, they're essentially worthless-even if technically accurate. "People who process data don't see it from a clinical standpoint. It has to be vetted by someone with a clinical eye, because if you send something to a clinician that simply doesn't make sense, they're going assume you simply don't know what's going on. Reports have to pass our clinical sniff test before we send it out."
People, people
Carving out a niche for an analytics team can be a grueling battle. But then comes the really hard part-finding the right people. And if your organization is struggling to get the right players in the analytics positions, take heart in that you're not alone-everyone has staffing challenges. Everyone.
"It's the same story all over, be it in the finance or logistics or health care industries-everyone is begging us to help them find data scientists and analysts," says Aiken, the VCU professor and consultant. "I think the industry has become pretty good at recruiting I.T. talent, but what's really needed for analytics programs is people who have a multi-disciplinary backgrounds and aren't just loyal to the I.T. portion of the job-technology people tend to stay loyal to the technology and not necessarily interested in the data. Many do eventually become interested, but that's only after 10 years or so in the industry and then they have an 'aha' moment and get really interested in the data and understand when they get the data right, everything else becomes easier."
Health I.T. vendors often can ease the pain by delivering technology and expertise. But analytic leaders warn that while solution vendors can help, they can't pull all the weight. "I've been a part of both builds and buys, and my experience has been that vendors can take you pretty far down the analytics path, but they can't take you as far as you need to go," says Hendricks from Henry Ford. "Organizations know their data and have to be in the driver's seat. So the mix of skill sets might change if you buy off-the-shelf, but you're still going to need your own highly-trained staff to work on your data."
But health care stands alone in the complexity of its data and operations, myriad nomenclatures, the number of different stakeholders and the layers of security surrounding nearly every byte of data. As a result, many health care organizations find there are never enough workers on the market who have the skill sets and temperament to fit into the health care environment, or at least do it quick enough to satisfy executives.
Data scientists
"One real issue I've seen is that health care organizations recruit data scientists and assume that, since medicine is based on science, they can make everything work," Aiken says. "But you can't recruit data scientists and expect to lay out 90-day deliverables-it takes time for analysts to figure out the business of health care. I think too many organizations have unrealistic expectations about the speed which these efforts can move."
To be a successful health care analyst requires a great deal of human skills and a big dash of patience, traits that analytics leaders say are actually harder to find than the technology experience that is typically considered a prerequisite for the job.
With the sophisticated tools available now, analytics don't necessarily need to be coding wizards; but they need to know how to communicate with team members and the clinical staff who are pushing the business forward.
Work ethic needed
"I go after biomedical degrees and workers with backgrounds in visual sciences, but I have a fantastic recent grad whose background is in advertising but has great user interface skills and an eye for detail," says Daly at HCA Corp. "You can't teach personality, and to work in this environment you have to communicate constantly. And I've also learned that if you put someone with a very strong work ethic with others with the similar trait, you can really, really move mountains."
Kimura at Atrius Health takes the same approach. The group practice federation is headquartered in Boston, where a couple large health systems are snatching up experienced analysts and competition is fierce for similarly skilled workers.
But Kimura says he's been successful in recruiting a team of freshly-minted grads and seasoned veterans by putting a premium on finding people who have a gift for breaking down technical elements for a non-technical audience and are comfortable interfacing with business users and executives, most of whom are much farther up the corporate ladder.
"We hire on four tiers, E1 being entry level and E4 an experienced health care analyst, and I can tell you that those E4s are getting harder and harder to find," Kimura says. "So we've shifted our focus somewhat on hiring E1s and E2s with the expectation that it will take a couple years to train them and get them enculturated into our environment. We're not so much concerned about how much SQL they know as whether they can deconstruct a problem and have the confidence to not blindly follow what the medical director tells them to do, to be innately curious about the data and the business issue they're working on."
Kimura looks for diverse skill sets when building an analytics team. Think big when it comes to evaluating talent, he suggests.
"This is really important when you start moving into very complex stuff like variation analytics," he says. "Certain people can hold their own in this environment, and others can't. What's interesting is that we've found people with very diverse backgrounds who have those skills and have done great work here. There's no one set of work experiences and academic training that creates a really good health care analyst."
Recruiting tool
Kimura adds that Atrius Health's data quality also provides the group with a potent recruiting tool. Atrius has a data steward responsible for ensuring all code sets are up to date with a standard nomenclature and a data quality committee that resolves conflicts with data going into the electronic health record.
"Our tools are pretty laid out and we sell the fact that it's easier to break into our organization and do more interesting analytics because analysts won't get tripped up by the intricacies of health care data.
"They can pull clean data instead of spending the bulk of their time trying to clean it up, which I think a lot of analysts at other organizations are being forced to do."