Big data provides true picture of diabetic population
New Zealand Ministry of Health improves diabetes policy planning with analytics
Diabetes is a chronic disease affecting many around the world. New Zealand’s Ministry of Health sought to more accurately understand the percentage of its population that suffers from the condition. With the help of SAS data analysis capabilities, the ministry created a register to accurately predict the prevalence of the condition and help design effective public health policies to support quality clinical improvements.
In diabetics, the pancreas fails to make enough insulin or the body is insensitive to the insulin present, resulting in an overabundance of glucose in the blood. Patient care for the condition is extremely important as diabetes often causes additional complications, including significant morbidity, mortality and high health care costs. While diabetes’ prevalence continues to rise around the world, New Zealand’s Ministry of Health found that it was difficult to accurately estimate the number of cases in the country since there was no consistent data collection across all general practices and hospitals.
Tracking the population of diabetics
The Ministry of Health, in collaboration with experts from the New Zealand Society for the Study of Diabetes (NZSSD), established a Virtual Diabetes Register (VDR) that combines and filters various sources of health information to more accurately determine how many people are diagnosed with the condition, as well as predicting who is likely to develop it in the future.
Dr. Paul Drury, Clinical Director of the Diabetes Auckland Centre and Medical Director of NZSSD; Dr. Sandy Dawson, Chief Clinical Advisor; and Emmanuel Jo, Principal Technical Specialist at Health Workforce New Zealand, Ministry of Health, have all been instrumental in the establishing the VDR.
“Previously, we would use national surveys to measure diabetes in the community; however, this proved to be slow, costly and have a high error rate,” Jo says. “Now, the quality of data within the VDR is offering the accuracy and immediacy we need.”
Six major databases were used to establish the VDR: hospital admissions coded for diabetes; outpatient attendances for diabetes and diabetes retinal screening; prescriptions of specific antidiabetic therapies; laboratory orders for HbA1c, which is a measure of diabetes management; and primary health (general practitioner) enrollments.
Using SAS software, an analytical model was created and modified to improve sensitivity and specificity, and then validated against primary care registers.
“Primary data was linked by the National Health Index number, which was available from six databases at the ministry,” Jo says.
He adds one challenge came from linking the six data sources and integrating them with a patient’s health number to work out if they needed to make contact with their local health center – a process he says took years to refine.
But Drury says the initial results were impressive. “The initial estimation before the VDR was refined showed that 210,679 people, or 4.88 percent, had diabetes in New Zealand as of Dec. 31, 2009. This compares with the country’s population of 4,315,355. The VDR yielded a final estimate of 189,256 people, or 4.39 percent, with diabetes.”He adds that diabetes prevalence also showed a clear difference between European/other versus non-European/other ethnicity. “Indian and Pacific people have the highest diabetes prevalence rate,” Drury says. “This means we can focus health policies
on this group.
“The beauty of the VDR is in combining many data sources and merging the database with other sources of data to look at the implication of diabetes in particular cohorts.”
“SAS’ analytical capabilities have been the key tool in developing the VDR to be accurate and robust, revealing a true representation of the diabetes population,” Jo says.
More effective health policy planning
In addition to more accurately predicting the number of people with diabetes, the VDR is used for planning health services and policies.
“We have 20 different District Health Boards (DHB), and the data can show them how many diabetic people are in their area,” Drury says. “GPs should know already how many they have, but the VDR is also able to help them predict who may be at risk so they can be prepared. By knowing the populations where diabetes is more prevalent, more resources can be directed at them to provide clinical quality improvements.”
Privacy in health is always an important factor, but Drury says access to the data in the VDR is regulated. “We've built the VDR to ensure there is privacy around patient information. Each DHB from this year can access this and see its data. GPs themselves don’t have open access to the register – it only goes to those directly involved in that individual person’s health care.
”100 million rows of data with SAS®
The VDR relies on SAS for its analytical data management. “SAS is the only tool that is able to deal with the amount of number crunching that we require,” Jo says. “Some of the data sets we use are huge with more than 1 million rows of data per month. Sometimes we need to go back 10 years, so we’re looking at working with over 100 million rows of data. SAS is the best tool to enable this.”
In the future, the VDR will continue to have more criteria added to it to monitor changes. “The criteria from five years ago will become less accurate as medical science changes, so our systems with SAS are built to be able to respond easily to these changes,” Drury says.
Jo adds that it's easy to change the analytical models in the SAS solution. “It is a well-structured system with robust coding,” he says.
Both Jo and Drury are extremely pleased with the support from SAS in setting up the VDR and the positive effects it's had on managing diabetes in the country.
“The VDR is invaluable for monitoring national prevalence and supporting clinical quality improvements,” Drury says. “It's also readily applicable to other areas and future projects to investigate the correlation between the two areas or among many other factors.
Using big data to treat diabetes
For another look at how data can be used to help patients with diabetes, consider the work of Brian Denton, a health care researcher at the University of Michigan. For the past five years, Denton’s team has been collaborating with investigators at the Mayo Clinic in Type 2 diabetes research funded by the National Science Foundation.
The goals are to help define better treatment guidelines for high blood pressure and high cholesterol – two major factors for diabetes – at a national level, and to provide decision support to physicians at the point of care.
“We’ve been using large amounts of historical data to develop models that can be used to simulate what happens when you apply guidelines for treatment to patients with Type 2 diabetes,” explains Denton.
The data comes from more than 100,000 people across the US who have Type 2 diabetes, and includes data on various medical factors, such as blood pressure, cholesterol, and blood-sugar levels, and how they’ve changed over a 10-year period, Denton said. “Twenty years ago, we just didn’t have access to data sets like this. It’s the availability of the data that’s making these kinds of studies possible.”