Business Intelligence Best Practices - BI-BestPractices.com

Collaboration. Communication. Community.

 
 
 Printer-friendly
 E-mail to friend
ADVERTISEMENT
BI Case Study: Dashboards for Doctors

by Stephen Swoyer
The analytic dashboard is often positioned as the must-have accessory of the savvy corporate executive, but as EMA’s success demonstrates, users at every level of an organization can benefit from dashboard-like presentations, too.

Jonathan Rothman knows a thing or two about doing BI on a budget. As the business intelligence guru for healthcare services provider Emergency Medical Associates (EMA), Rothman, the company’s director of data management, has tapped an application framework from Business Objects SA to deliver reporting, analysis, and—as of June, 2004—dashboard services to hospital administrators, doctors, and other healthcare professionals. Best of all, Rothman was able to implement EMA’s cutting-edge analytic dashboards in just over three months.

The analytic dashboard is often positioned as the must-have accessory of the savvy corporate executive, but as EMA’s success demonstrates, users at every level of an organization can benefit from dashboard-like presentations, too. More to the point, Rothman says, dashboards aren’t just for traditional consumers anymore: Research scientists, civil engineers and even doctors, nurses, and healthcare administrators need dashboards too.

In spite of EMA’s dashboard success, Rothman believes there’s still room for improvement—if Business Objects does its part, that is. EMA is also a user of Crystal Reports, an operational reporting solution Business Objects acquired in July 2003. In fact, much of the data that’s eventually fed into EMA’s data warehouse originates in Crystal repositories deployed at more than a dozen customer sites. Right now, Rothman uses an ETL solution to extract data from Crystal and load it into its Oracle database. With the enhanced integration Business Objects has enabled between Crystal and its own BI products, however, Rothman will be able to push data directly from Crystal out to his users.

Why a Dashboard?
Digital dashboards provide an at-a-glance view—by means of graphs, charts, gauges, speedometers, or any other instantly understood mechanism—of several predefined metrics. The idea, of course, is that the dashboard interface helps to mask the complexity of the underlying business analytic infrastructure by displaying the results of an ad hoc query or report—or just about any other kind of analysis, for that matter—in the context of an easily digestible view. Worried about whether you’re going to make your sales goals for the current quarter? If you have a dashboard on your desktop—perhaps one outfitted with a speedometer-like display, one glance and you'll know the answer. If you’re in the red, it's time to seriously motivate your sales managers; the yellow zone suggests you'll need firm prodding, and green means that you can breath a sigh of relief (at least for now).

That’s a typical usage scenario in a forprofit, sales-driven organization. So why is it that EMA—a physicianowned, privately held, not-for-profit organization with 16 (mostly) notfor- profit clients in New Jersey and New York—invested so heavily in a dashboard infrastructure?

As it turns out, Rothman argues, dashboards are well-suited to the frenetic pace of emergency department (ED) operations. “The dashboard is really the best [solution] for the ED. The speedometers are great for quick analysis of data, because once you understand the metric, and then once you understand from the end user what their targets and thresholds are, then all they essentially have to do is look at metrics and then look for something red,” he explains.

Taking the Plunge
EMA didn’t decide to invest in dashboards on a lark. Rather, Rothman was looking for a more intuitive way to expose much of the information stored in EMA’s Oracle data warehouse, which was previously accessible only through its bread-and-butter ED management application, Emergency Department Information Manager (EDIM).

EDIM—which EMA built in 1998 using Visual FoxPro—is a patienttracking and electronic medical records system. It’s used wherever EMA has an emergency department (ED) contract, and boasts more than 350 clinical templates designed for ED administrators, along with about 80 predefined reports. Because of its age and its Visual FoxPro underpinnings, EDIM could not easily be ported to a dashboard-type interface, Rothman says.

That’s why he decided to bypass the middleman, so to speak, by directly exposing the information stored in EMA’s Emergency Medicine Analysis and Reporting Systems (eMARS) data warehouse in a dashboard interface. Even then, Rothman says, he probably wouldn’t have gone the dashboard route if he hadn’t been intrigued by the promise of the Business Objects Application Foundation (AF) Dashboard Manager, an analytic framework that’s designed to let customers develop rich analytic dashboards with—or so Business Objects officials promise, anyway—a modicum of effort. “Application Foundation starts with a top-level view of key metrics, but it also gives the user the ability to drill down from any of these [metric] to the operational data beneath,” explains Lance Walter, director of BI product platform marketing with Business Objects.

Using AF Dashboard Manager, Rothman saw that he could provide ED administrators and healthcare clinicians with the level of detail that they required, both in the form of an ad hoc query and reporting facility— using Business Objects’ native reporting tool—as well as customized analytic dashboards.

“AF allows the end user to launch a Business Objects report and the prompts are there to allow the end user to set their start date, their end date, set their hospital, [and] set their physician,” he explains.

Rothman admits he wasn’t sold on the idea until he contracted with an outside integrator, Sapphire Consulting, to develop an AF-based dashboard proof of concept for him. His work with Sapphire allowed Rothman—a BI autodidact with an MBA in risk management—get rapidly up to speed on dashboards. In the process, he says, it also helped him to tweak his data warehouse environment to better take advantage of AF. “During that proof-of-concept process, I learned a great deal about how data moves from the database level to the tables used by AF to display metrics using their [Sapphire’s] ‘gadgets,’” he confirms. “As a result, I developed some additional ETL processes to create new tables in Oracle in order to modify how the data, used by AF, is stored. These new ETL processes, and new summarized tables, have had a positive impact on data quality and the speed in which daily data is uploaded and then used by the AF tables.”

Sapphire spent all of eight weeks at EMA, during which time its integrators constructed a base infrastructure for analytic dashboards. From there, Rothman says, he was able to hit the ground running.

“I was able to look at what they had developed and learn how to use the AF product at a level that would have taken me months, within about two weeks,” he explains. “As a result of their work, I was able to re-design all dashboards and, with minimal outside help, launch a whole new set of dashboards by the beginning of June. If you look at what Sapphire completed by the end of February versus what is in place today, there is no resemblance.”

Dashboarding Wellness, One Metric at a Time
What metrics do hospital EDs need to keep track of? After all, it’s not as if they’re monitoring late-breaking supply chain information or daily sales figures, as would traditional, forprofit, sales-driven organizations. Rothman and EMA aren't putting vital stats—blood pressure, respiratory, cardio-vascular, and other readings— up on their dashboards. Instead, you might say that EMA is helping administrators and clinicians monitor the health of the ED itself.

The company’s EMARS application provides an analytical breakdown of important ED criteria—e.g., total revenues, total expenses, total patient visits, overall ED admission percentages— and e-mails reports (in the form of conventional PDF documents) to hospital and ED administrators, as well as other healthcare professionals. Thanks to its integration with Business Objects’ AF Dashboard Manager, EMARS also provides a dashboard display of dozens of key emergency room metrics, giving doctors and other emergency personnel a stat view of the ER’s wellness—in both financial- and service-oriented metrics.

“We have two different types of dashboards. We have a set of dashboards that are specific to emergency department operations, and we have a set of dashboards associated with billing and collections,” Rothman says. “Within these, you’ve got ED operations, you’ve got payroll, you’ve got billing. And then you’ve got syndromic surveillance”—the monitoring of specific illnesses.

How do the dashboards differ? Consider the billing and collecting dashboard, which is designed for both hospital bean counters and ED doctors. This is possible, Rothman notes, because the AF Dashboard Manager framework allows him to create different dashboard views for different classes of users with very little effort. “Within billing and collections, one [dashboard] is associated with how we get paid, the other is associated with what the doctors actually get paid, so we have a set of metrics associated with how much money did [an ED] get in, what were [its] expenses, things like that,” he explains.

EMARS also lets healthcare providers track common ED service metrics, including total pediatric visits; total patient admissions; percentage of patients who left without being seen (LWOB); and percentage of patients who left against medical advice (AMA). But that’s only the beginning: the EMARS application also tracks more granular service-oriented metrics, such as the time that elapses between when a patient is first registered and actually seen by a physician; turnaround time for admitted patients; and turnaround time for discharged patients. “Think of it as a speedometer for visits at a site that sees 300 patients per day, for example. If I want to know when my 'left without being seen' is greater than two percent, which is the national benchmark, I just look and see if it’s in the red,” Rothman says.

Above all else, says Rothman, the goal was to make eMARS—in both its ad hoc query/reporting and dashboard manifestations—as easy as possible to use for a range of non-traditional consumers. “We’ve been able to use AF Dashboard Manager to set up reports that are more interactive, so eMARS doesn’t require [users] to even be familiar with Business Objects at all. It just requires them to answer some questions,” he explains.

In this respect, Rothman can configure predefined thresholds for many metrics, such that users are alerted when thresholds are exceeded. “That’s where AF has helped us, because by setting all of these targets and thresholds, we pump all of these metrics through these and fire off alerts if one of them is exceeded. We can say [to the recipient], ‘Here’s your package, but you should look at page two of the report because there’s a problem here,” he explains. “That’s where Dashboard Manager is really helping us. It also does a lot more analytics for us, so rather than you produce the report as a human being and sit down with the data and say, ‘Your trend over the last six months is X, and you should be worried about it,’ you program it into the machine.”

It’s an impressive achievement, but eMARS as it’s presently constituted has one obvious shortcoming: The Crystal data used by its EDIM application— which feeds the eMARS data warehouse—isn’t directly accessible from within the Business Objects environment. For that piece of the puzzle, EMA must wait on Business Objects. The good news, Rothman says, is that help is on the way.

Bringing Crystal Into the Fold EMA isn’t a Business Objects-only shop, of course. The company taps an Oracle database as its data warehouse repository, along with an ETL tool from the former Data Junction, which was recently acquired by Pervasive Software Inc. Elsewhere, EMA—like many Business Objects customers— deploys Crystal Reports as an operational reporting solution for EDIM, in this case for use at more than a dozen distributed customer sites. As if that’s not enough, the company taps Crystal for its help desk reporting solution.

Much of the data eventually fed into EMA’s data warehouse originates in Crystal repositories deployed at 16 customer sites throughout New York and New Jersey. Right now, Rothman uses an ETL solution to suck data out of Crystal and load it into its Oracle database. Because this is typically done on a nightly basis, the data in the eMARS repository can be up to 24 hours old. With the enhanced integration that Business Objects has promised between Crystal and its own BI products, however, Rothman will be able to publish data directly from Crystal out to his users.

“What my vision long-term is not just to suck out the data, but actually post the Crystal Reports that are generated,” he says.

Aside from the hassle of transferring data between Crystal and Business Objects, Rothman says there are other drawbacks to his existing approach. “I’m trying to set up an infrastructure so that if somebody was in California, and they wanted to see what the activity was in their emergency department yesterday, they can use AF, and they can see the speedometers over the Web,” he explains.

Later this year, Business Objects is expected to ship a new version 11 release of its eponymous BI suite, thus providing Rothman and other users with the answer to their prayers: backend integration between its own products and Crystal. Rothman, for his part, says that he can’t wait: “I am excited about the fact that Crystal Reports can take advantage of the Business Objects universes, but I’m even more excited about being able to post Crystal Reports directly on AF.”

Conclusion
EMA’s new dashboards have been live for just a few months, but Rothman says that they’re already starting to make a difference—especially as a strategy for helping users deal with information overload.

“One of the things that happens— and it should be included in the life cycle of the data warehouse—is that all of the sudden you became so good at analyzing everything that you’re collecting and reporting on everything,” he observes. “We got to where once a month, we’d e-mail users a standard [report] package that was 32 pages long. So we can still give everybody everything, but now they have their dashboards and alerts, so they know to go into the report if [the dashboard] tells them something’s wrong.”


Recent articles by Stephen Swoyer

Stephen Swoyer -

Stephen Swoyer is a technology writer based in Athens, Ga. You can contact Stephen via E-mail at swoyerse@percipient-analytics.com.