Jewish General uses self-service platform, synthetic data to democratize analytics

The Jewish General Hospital in Montreal had a problem with its patient data: despite having sufficient data available, the process of gaining insights that improved patient care was difficult and burdensome.

THE PROBLEM

It has a large number of staff who are end users of data at the Jewish General Hospital and the maternal health system, CIUSSS du Centre-Ouest-de-l'Île-de-Montréal. These stakeholders include operational teams focused on improving efficiency and reducing delays, clinical teams engaged in clinical quality improvement projects, and a thriving research community at the research organization, the Lady Davis Institute.

“The need for data is constant, and our ecosystem of data sources is quite varied and in some cases difficult to access,” said Dr. Justin Cross, Chief Digital Health Officer at Jewish General Hospital. “As in many healthcare systems, the data from these source systems has been brought into a centralized data warehouse. From there, they can be granted to the appropriate stakeholders if the access conditions are met.

“This process unfortunately requires technical knowledge and database skills that most of our data end users simply do not have,” he continued. “Because of this technical need, the data warehouse is managed by a small team of data engineers and data analysts who have these skills and can build out views for those with the permissions to access source data.”

The process works, but the lead times to formulate a query, send the query to the appropriate approver, and then structure a technical request for the internal data team has created a long road for data access, especially for researchers.

PROPOSAL

The Jewish General Hospital decided to address this issue with healthcare IT vendor MDClone's Adamas platforms. The platform offered a tool that ensured that anyone with an idea within the institution could ask a question and achieve results.

“We know that good ideas can come from anywhere, and especially from those who may not traditionally be in roles focused on operational efficiency or quality improvement,” Cross said. “Staff working on the front lines may notice things that lead them to view questions from a very different lens than that of an internal data team.

“It is absolutely critical that a multi-disciplinary team is involved in the planning, design and implementation of a system like this to get maximum value from the platform.”

Dr. Justin Cross, Jewish General Hospital

“We wanted to leverage the power of diverse points of view from across the organization and focus them on our common goal: improving the quality of the services we provide to our patients,” he continued. “We wanted our healthcare teams to be able to quickly access the platform, ask a question about something they saw in real time, and quickly get answers back that would indicate a real signal is present.”

Similarly, the researchers' staff wanted them to be able to quickly search the institution's data store to see if a research cohort of interest exists.

“The special element that makes this possible in a way that protects patient privacy is the synthetic data mode of the MDClone Adams platform,” he explained. “In response to a query, the Adams platform is able to dynamically generate a fully synthetic dataset from the original data. This synthetic dataset is statistically similar to the original data, but fully synthetic – and not simply anonymized.

“This allows us to use the synthetic results to identify whether a signal of interest exists, and if so, the end user of the data can then go through the required approvals to access the real underlying data,” he continued. “This greatly accelerates the initial validation of ideas and focuses the time involved in the real data approval process for those initiatives that have shown signals in the synthetic analysis.”

MEETING THE CHALLENGE

The Jewish General Hospital is in the early stages of going live with the platform; it launched the system two months ago. To implement the system, staff created a multidisciplinary working group consisting of the MDClone team, staff from the hospital's digital health team, the IT team, and subject matter experts who knew the source systems inside and out.

This team led both the technical implementation and the data extraction and loading process. The staff focused primarily on clinical data at this time, but will soon expand to other data sources.

“The Adams platform will democratize the use of data by enabling the exploration of data using basic human language to build data sets in a simple and intuitive way; knowledge of SQL or other database skills are no longer necessary,” explains Cross out. “After receiving appropriate training and authorization, the tool will be available to both our clinical and administrative teams.”

RESULTS

The Jewish General Hospital is only in the early stages of training the community on the tool and identifying pilot users for the new platform.

“We already have teams looking to look at several factors that may influence: length of stay, delays in access, home and community care optimization, hospital-acquired infections, STEMI care, mental health optimization and more” , Cross said.

ADVICE FOR OTHERS

“When starting a major data project like this, it's not just about the technology, but also the people and processes involved,” Cross advised. “It is absolutely crucial that a multi-disciplinary team is involved in the planning, design and implementation of a system like this to get maximum value from the platform.

“Data must be validated and usable by the end users, and a feedback loop must exist for continuous improvement of the platform,” he concluded. “Finally, adequate support must be provided to clinical teams wishing to use a tool like this. They will often have fantastic ideas and need to be supported with appropriate project management support as an actionable idea becomes reality.”

Follow Bill's HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.

Related Post