Creating ‘data aware’ culture: Quinte Heath Care

Background:

Quinte Health Care Corporation is comprised of four geographically dispersed hospitals in Hastings and Prince Edward counties of Ontario, which employ a total of 1,600 people. Each of these institutions serves patient regions featuring different demographic profiles, a point that has become more significant as Quinte strives to align delivery of patient care with the requirements introduced by the Ontario Health System Funding Reform of 2012.

QHC Belleville General Hospital
QHC Belleville General Hospital

Ontario’s shift four years ago towards patient-based funding introduced new payment models for the province’s hospitals. Instead of guaranteed budgets indexed to the rate of inflation, hospitals were faced with a new formula that calculated budgets based on three inputs: 30 percent of historic, global budgets (no longer indexed to inflation) would serve as the foundation, 30 percent would be based on the delivery of quality-based procedures (ex. hip replacements) that commanded a specified price point, and the remaining 40 percent would depend on regional demographics, including the number of people that live in the area, population age, virility, economic status and previous healthcare encounters, the disease types that are prevalent in the region, and population growth. Designed to encourage more financial accountability and greater efficiencies in the delivery of healthcare across Ontario as a whole, the reform also generated considerable uncertainty for individual hospitals tasked with providing patient care out of changing budgets. As Peter Papadakos, director of decision support & analytics at Quinte Health Care, explained, approximately 70 percent of hospital budgets could be in flux as institutions competed for the delivery of quality procedures, and as a result of year-to-year fluctuations in population demographics. The population in Quinte, for example, is not growing as quickly as in other regions: combined with the elimination of inflation indexing, this factor would serve to further shrink hospital budgets.

Today, the Ministry of Health estimates and publishes what funding each healthcare institution will receive; however, when the reform was first introduced, specific numbers were not provided. “It was a guessing game,” Papadakos noted. For its part, Quinte received a budget estimate from its local LHIN suggesting that the hospital would have to cut $12 million from its budget in year one of reform, while another estimate put that number at $7 million. To better understand what the actual reduction would be, Quinte Health Care opted to do its own analysis, and to deploy more sophisticated analytics tools to speed this process. According to Papadakos, while analysis could be done using SQL Server and Excel, this approach would have increased the time required to come up with a reliable estimate of budget cuts by a factor of four, time the hospitals did not have: “The original analysis of what we had to save would have been done either way; but it could have taken two months or two weeks.” Additionally, he added, “identifying where the savings were, and where we could find them, was definitely driven by the ability to slice and dice the data, get into it and actually use the data.”

QHC Trenton Memorial Hospital
QHC Trenton Memorial Hospital

Based on its own internal estimate of the impact of health system reform, Quinte was ultimately faced with the need to cut $10 million from the hospitals’ budget in year one, and Papadakos anticipates the biggest impacts of the reform will be felt in the next year or two. From this circumstance, a second set of drivers for Quinte’s analytics adoption followed: the need to improve processes, to build quality of care levels, to understand what services are being delivered, to become more financially insightful and responsible, and to automate processes as much as possible. As Papadakos explained, “now an additional driver is to protect the population in the surrounding area. If you aren’t using analytics to improve data quality and improve processes, the people in the surrounding region can be negatively impacted because if all other hospitals are improving and we are not, our market share could decrease.” In other words, future funding could be impacted by the failure to optimize care delivery: “It’s just a downward spiral from there,” he noted.

Health care procurement

Quinte Health Care’s analytics journey began with creation of the business case for senior finance officers, which included items that could be quantified – with qualifiers, since outcomes are difficult to predict with absolute certainty. The goal was to outline existing process, how time could be saved, what productivity improvements would mean in terms of reporting, as well as any additional output from time savings. As example, Papadakos pointed to the potential for taking operational/transactional data sources, such as admitting data, service utilization data, plus coded data sources residing in Excel and Microsoft Access files, drawing these together, and then outputting the data in ways that would be highly consumable by hospital workers. The original process involved use of a variety of tools, some of which were highly manual, including an old application that created static graphs, time consuming analysis through Excel, and manual creation of weekly reports for a number of departments. Due to staff turnover in finance, it took about two years for decision support to convince senior leadership that new analytics capabilities would expedite this process.

Peter Papadakos, director, decision support, Quinte Health Care
Peter Papadakos, director, decision support, Quinte Health Care

Since the eHealth scandal, Ontario healthcare institutions have been required to put out competitive bids in procurement of goods and services. Close to ten vendors responded to Quinte Health Care’s open RFP for analytics software, and team evaluations produced a short list of three candidates. These were invited in for demos and to engage in Q & A with the Quinte team. Information Builders (IB) won the contract based on technology capability and on their presentations. According to Papadakos, the IB group answered the questions on the RFP really well, and followed up with examples. “There’s an art to RFPs,” he explained, “and they’re very good at it. They went above and beyond responding to the RFP. Pretty much everything we wanted they could deliver: yes, they could create dashboards very quickly, and here’s how you do it. You drag stuff onto a canvass and you’re done.” The IB demonstration was also compelling because they used mocked up healthcare data from Quinte, showing not only end result, but also how quickly data could be manipulated and integrated with other sources.

Speed was critical in the Quinte decision to opt for the Information Builders platform. Papadakos noted: “In the RFP, I used ‘agile’ a couple of times because you need to be able to pull in data from any data source and build something around it very quickly. You don’t want to waste your time with steps that aren’t needed. [With IB] You could have an Excel sheet, upload and pull in all the data, create metadata on that, add your dimensions, and then build out dashboards and reports, and an OLAP cube – and it would literally take about ten minutes.”

Implementing new analytics capability

Quinte’s implementation was guided by the principle of ‘customer focus’. The process began with identifying who the end users would be, what internal “customers” would need, and what senior leaders who approved the project needed to see. “These are the things we always kept in mind because we needed to show results quickly, ensuring that we didn’t destroy what was in place as long-term vision in the process. We were always trying to balance these two things off – they are often not the same,” Papadakos explained.

To achieve this, the Quinte team started with information support for the acute inpatient department, an area that accounts for the majority of the hospital’s activity from a financial perspective. This involved analysing existing data sets, refining the data in a staging/sandbox area, integrating it with IB’s iWay DataMigrator into the main data warehouse, using that to create a massive data table from inpatient data, adding dimensions, and then reporting off the table using IB’s WebFOCUS business intelligence platform. The intent was to show rapid progress on the $10 million budget reduction target. As Papadakos described it, “How do we immediately satisfy, prove within a month that something is happening here? We had no time to go underground and build a complex data model where no one would see results for two years, so we built this nice, acute inpatient data set with the appropriate linkages so we wouldn’t impact long term data management. It had linkages, such as patient identifiers, locations, and periods of stay, which could be integrated with anything else we brought into the data.

According to Papadakos, “our senior leadership was really happy” with the fact that the initial project took only a month: “We knew what our data sources were, pulled them in, used some SQL knowledge to tie them together, iWay DataMigrator brought them all in, and WebFOCUS is pretty simple to use.” Altogether, the Quinte team took advantage of eight or so days of professional services support over the course of the first year and a half of deployment. IB training was provided on the technology, and its examples were delivered using Quinte data. “So the combination of technology understanding, which IB brought to the table, and our understanding of the data and the required output worked beautifully. It was very fast,” he added.

Training a critical element in continuous data-based innovation

Quinte’s four person implementation team consisted of a BI specialist and individuals who were familiar with the data. While comfort with existing processes led to some initial resistance in one case, team buy-in was encouraged by involving members in building the business case – an initial training ground that was expanded through additional coaching. According to Papadakos, onboarding decision support staff was relatively painless: “Now they love it [the new software]. It has totally changed what they do and how they do it. They were used to formatting Excel files into array formats – now they can do statistical analysis of changes in value over time based on PDSA cycles in the hospital. This means the team can participate in effecting real change within the institution, providing direct support for the project teams working to improve processes.” An engaged team, using advanced analytics to create additional value represented a first level of transformation to better patient outcomes.

QHC North Hastings Hospital
QHC North Hastings Hospital

Stakeholders from the hospital’s various business units – or departments – were also involved throughout the implementation process to ensure that the team retained its customer focus. According to Papadakos, the maturity of potential users in terms of their understanding of analytics was initially quite low. Ramping up these end users proved challenging as Quinte’s team had to cover both training on use of the technology, and also on use of the content. “We were actually teaching the end user about their data. We said to the end user, ‘here’s your patient, here's what the different matrices mean, here’s how you can interpret them, here’s how we can aggregate them, and here’s how you can make decisions based on them.”

A key to effective end user training involved understanding the different user types: who has access to the apps, and who has the time to learn various levels of analytics usage. Based on these profiles, the team would engage in broad education sessions, one-on-one training with more advanced users, and deliver email blasts on what applications were/are available. Access to different apps or dashboards was based on job description, but also on training achieved. “Some users are really easy to train because all they have to do is go to a dashboard, while others are using more complex reports,” Papadakos noted, but the team found that work with more advanced users had a tendency to spread: “They will talk to their friends. They will be at a meeting and have all this power because they have the data that they have pulled. Getting access to data, and being able to manipulate it is empowering for people.”

In Quinte’s case, change management also occurred top down. When a manager is hired, they have to meet with the decision support team to review what data is available to them. The hospital also requires a lot of reporting to the board, based on specific indicators. An individual responsible for readmission rates, for example, would have to go to the Compass portal to review that indicator, do analysis on it in self-service fashion, or if necessary, with the help of decision support staff. So rather than create incentives to encourage adoption, at Quinte, analytics use has become part of a manager’s job description – “It has become part of the culture,” Papadakos added, and training serves as the means to access information. “Someone that four years ago did not want to use a computer, now gladly uses it. They use email to find active reports, and to discuss the results.”

Quinte began its analytics implementation with the acute inpatient department because that was the largest group; since then, IB-based decision support has been rolled out to Emergency Room, Day Surgery, Finance, Lab and Pharmacy – most of the units in the organization.

Assessing implementation outcomes

The care taken in introducing the IB platform has generated significant benefits. In particular, it has helped Quinte staff to become data aware: “It’s become really personal,” Papadakos noted. With initial rollout, he explained, departmental units claimed they didn’t have access to enough data; in next versions, they complained they had too much, and so the team engaged in more user education. Now the pendulum has swung back, and users are again complaining that they don’t have enough data; however, what they are asking for now is much more complex. Hospital users of analytics capabilities have become more sophisticated, an evolution the Quinte team is hoping to address through rollout of version 8 of WebFOCUS, which Papadakos believes will allow decision support to provide the level of detail users are asking for – in digestible forms. Improved transparency for data access, improved graphics for data visualization, HTML formatting, and a more user-centric portal are specific features that he anticipates will enhance usability for both the developer and the consumer.

QHC Trenton Memorial Hospital
QHC Trenton Memorial Hospital

While accurate ROI measures on software implementation are often difficult to achieve, Quinte considers adoption rates and time saved through use of the IB tools as key success indicators. So far, the hospital has experienced rapid increases in user adoption, and user surveys tell the team users are happy with the results. In terms of productivity improvements, reports that formerly took an hour and a half to generate through largely manual processes are now generated automatically and sent to the appropriate users. This allows the decision support team to work on more detailed analysis, where hard dollar returns may be easier to quantify, as opposed to simple reporting.

As example, Papadakos pointed to a more recent project aimed at identifying new sources of revenue within the hospital budget. A key input to establishing hospital funding in the province’s new model has been the Cost per Weighted Case, an estimate of the intensity of resource usage for an individual patient in medical cases where this is likely to be greater than for standard patients – pneumonia patients, for example. But the real cost on a patient-by-patient basis may vary from this provincial standard, resulting in lost revenue if case numbers alone are reported. To explore this issue, the Quinte team used WebFOCUS to build an application which allowed them to identify cases or diagnosis groups where Quinte costs were above the Ontario average of cost per weighted case, or cost per patient day. A second application was developed to slice and dice data at the physician and intervention levels to figure out why costs might be higher than the Ontario standard. According to Papadakos, to establish that certain areas were not receiving the appropriate funding, high data quality would be needed, a problem the team tried to solve for by implementing a program to identify information coders for specific hospital areas that might have recurring data quality input issues.

But the team also used machine learning, predictive algorithms to improve data quality in order to better compare Quinte and Ontario costs. With COPD data, for example, the team looked at the specified patient pathways for treatment/encounters, downloading 4 million records of acute inpatient data for Ontario, as well as a subset of COPD diagnosis data for the hospital to identify outliers, relationships between data points, points where issues may emerge and where recoding data or turning up the numeric value on an indicator might make a difference. The team also applied different data models to assess data accuracy, which Papadakos explained could be deployed out of WebFOCUS and brought into AppStudio for visualization/reporting, as any other function. The results were significant: in one particular case, the team discovered an expense difference of .5 – a gap that represented $2,800 in the cost of care. Aggregated, these gaps added to considerable new revenue when reported to Health Records for review, analysis, and additional reimbursement, based on missing data elements. For one type of COPD case, Papadakos estimates a savings of approximately $100,000 to 150,000 per year for the institution.

Going forward, the team is looking to refine its data quality model and make it more precise before investigating the next group, or any disease cohort that the hospital may be losing money on, and also to continue to optimize the system. As Papadakos has described it based on information derived from analysis, the questions have changed: they have evolved from how many COPD patients do we see per month, to what are the ordering practices of this physician group on these COPD patients; from what is the patient flow, by day, by hour, of this particular type of patient, to what is the frequency of this type of lab test being ordered, after another lab test was ordered. So new data sets, new levels of integration, new data models, and new ways of reporting have helped to develop very advanced information consumers, posing increasingly-complex questions that can uncover new sources of revenue, and for the hospital, the ultimate ROI – continuous process improvement.

Lessons Learned: advice for practitioners

Know your data, and know the intended outcomes.

Build your data model so that it is expandable and agile; make sure you have a long term vision in mind as you build it.

Build a table with the right identifying fields in it so that you can link back to other tables if needed.

Don’t neglect training.

Implement Resource Analyser functionality, which tracks who is using what reports to assess adoption of the tool, and identify which groups may require additional support. Quinte created two kinds of user groups: to speed rollout and adoption, it created a public user for public content in addition to named users. But better policy is have all users named with identifying information in order to track adoption rates and spread.

 

 

 

 

 

 

1 COMMENT

LEAVE A REPLY