AI-DRIVEN CLINICAL SUPPORT TO MANAGE TYPE-2 DIABETES

Mr. Andreas Krug spoke about artificial intelligence (AI) -driven decision support in the safe management of diabetes in patients in hospitals as well as their experiences, challenges and the benefits at the Selangor International Healthcare Conference 2022 at the Kuala Lumpur Convention Centre on 7 October 2022.

Krug is Chief Executive Officer of Decide Clinical Software GmbH in Vienna, Austria, a scientifically-driven, health technology spin-off from the Medical University, Graz, Austria which developed GlucoTab, an automated, client/server-based insulin dosage suggestions system for basal bolus therapy of patients in hospitals with type 2 diabetes, which is accessed through a web browser on tablets devices and PC screens.

He has more than 25 years of experience in the hospital and healthcare industry, including as manager and consultant on a global level for Bosch, T-Systems, Siemens and Atos.

“It’s well known that many people suffer from diabetes and up to 30% of patients admitted to hospitals for any health ailment are also diagnosed as diabetic, with blood glucose levels abnormally high and out of control”, said Krug.

According to the U.S. Centers for Disease Control and Prevention (CDC), insulin is a hormone which regulates the blood sugar level in the body in a complicated process which includes helping blood sugar enter the cells where it is used as energy, as well to signal the liver to store blood sugar for later release when needed by the body, whilst excess blood sugar in the liver is converted into fat, and the person gains weight.

When this finely tuned system fails to function properly, the sugar level in the person’s bloodstream rises to abnormal levels for prolonged periods, resulting in the person becoming diabetic, which is very damaging to the body, so it has to be lowered and controlled with insulin administered externally.

Hypoglycaemia is the opposite condition where the blood-glucose level is abnormally low, which can cause the person to faint, suffer brain damage and in worst cases can result in death. Hypoglycaemia can result naturally due to starvation or can be induced due to an excessive amount of insulin administered to diabetics, causing their blood-sugar level to drop precipitously.

So far, there is no known cure for diabetes, so it needs to be monitored and controlled for the rest of the diabetic’s life.

Various factors influence blood glucose levels, such as the illness itself, different diets and so forth, and effective glucose management in diabetics determines different outcomes such as complications, mortality rate and so forth.

For example, a study in China showed that the mortality rate of COVID-19 patients who are also diabetic with their blood sugar out of control was 10 times that of non-diabetic patients.

 Also, one in three patients in hospitals suffer from medical errors in the treatment of diabetes, since such treatments are a complex processes which need to be very individually-oriented, and according to the evidence and the results of many studies, this is a problem for diabetics in hospitals.

“However, it’s good to know that there are clear guidelines on how to avoid these problems and keep blood glucose levels under control, by achieving basal therapy goals helped by electronics”, said Krug.

However, most hospitals worldwide lack standardised workflows, responsibilities are not clear and fears of excess insulin delivery resulting in hypoglycaemia in patients hinders provision of the right does of insulin.

One of the major problems is lack of knowledge, as diabetics are not always treated by specialists but by different departments in hospitals, and evidence-based guidelines are not implemented.

So, the results of the studies found that each patient was provided the same amount of insulin, which is nonsense, which led Decide Clinical Software (DCS) to develop a clinical support system which provides different options in diabetes treatment, where clinically-proven algorithms calculate the right amount of insulin to be given to a patient according to his or her specific situation at particular times.

“The benefits of digitalisation of data have enabled us to create realistic solutions for the proper control of diabetes in patients in hospitals”, said Krug.

Their solution is integrated into the hospital’s information system and provides the data which can be displayed on a mobile device, so when a nurse goes to the patient’s bedside and measures the blood sugar level, the system display the amount of insulin which should be given to the patient on the same mobile device.

According to the rules of insulin therapy, diabetic patients are usually given a long-acting insulin one time followed by short acting insulin three times and the patient’s condition needs to be monitored and understood each time.

Also, with caregivers having to perform many different tasks, there is a tendency for them to make errors, and all the readings, correct or otherwise are traditionally recorded on paper charts.

“However, with our system, all the decisions and steps in the process are fully digitalised, with the results presented on an electronic chart, so nurses can measure patients’ blood sugar levels at their bedside and provide them with the right does of insulin without having to refer back to the physician”, said Krug.

Krug then invited Associate Professor, Dr. Julia Mader, a leading diabetologist at the Medical University Graz, Austria to speak about their scientific work and studies of GlucoTab which have helped DCS to develop its decision support system.

In a healthy person, the pancreas provides a base insulin level even in the fasting state and when the person ingests food, the pancreas provides the right amount of insulin naturally according to the requirements of the meal.

Insulin therapy mimics this natural bodily function, with doses of long-acting insulin administered once or twice a day, and does of short-acting insulin administered when a diabetic person has a meal, or whenever the blood glucose level is out of range.

“The algorithm works to mimic the physiologic insulin profile and to administer the right amount of insulin to the patient, which perhaps is one of the most challenging aspects in hospitals”, said Dr. Mader.

Normally, medicines are prescribed to be given to patients once a day, which nurses take care of after their ward round, but since diabetic patients have several meals a day and their blood glucose level may be abnormally high at certain times, the system automatically informs nurses about the right amount of insulin to deliver based upon the blood glucose measurement, without nurses having to refer back to physicians, which greatly reduces the workloads of both, and also saves them time, and greatly helps to improve the treatment process.

The Medical University, Graz’s project compared the traditional diabetes therapy based upon paper records, also widely used in Malaysia, against results of algorithm-based therapy and they found that glucose levels using algorithm-based based therapy was initially a bit out of range at lunchtime.

They then compared the results of treatment using the algorithm against treatment not using it and found that treatment using the algorithm achieved the desired glucose level target range.

After having finalised the first algorithm, they proceeded to clinical studies and tested the electronic version and amongst the first set of 15 patients, the algorithm showed the same results as the paper-based version, and they modified their treatment each week until they finally achieved the desired blood-glucose level in most patients around lunchtime, and also achieved better blood-glucose control over a 24-hour period.

They then introduced the system to different departments such as surgery which was more challenging. They tested it on 99 diabetic patients over a 10-day period and managed to achieve the desired blood-glucose level on patients 67% of the time on the first day, and in most patients after the fourth day without increasing instances of hypoglycaemia in patients.

“The European and American Diabetes Association recommend that we achieve 70% blood-glucose level target, so we came close on the first day and we managed to have 92% of patients safely within target on the last day”, said Dr. Mader.

Her team then conducted a user-friendliness analysis, which is important to gauge the feelings of the users in their interaction with the system and they found that over 95% of the steps were performed according to the system, 95% of the dosing recommendations were accepted, and also that errors could be avoided.

Also, very relevant was the error analysis performed which enabled them to differentiate between errors in terms of documentation, insulin candles, calculations, and simulations from workflow, as well as the frequency of occurrence and the impact on blood-glucose values.

“With that, we had a reduction of insulin dose calculation errors when using the electronic version compared to the paper-based version, the effects of dosing errors were higher in the paper-based version, and they had a clinically meaningful effect on outcomes”, said Dr. Mader, who them handed the floor back to Andreas Krug.

Scientist-technologist collaboration

“I believe this shows a future where scientists and technologists will scientists and technologists will work much more closely together, which is slightly missing today when it comes to digital health”, said Krug.

“In the last study we made during a pilot project, one really tremendous highlight which came out of it was that physicians and nurses said that it enabled them to save around 10 to 15 minutes per patient with diabetes, treatment over the day.

“But in the same time, if we introduce new technologies into hospitals, we must aware that we cannot put additional burdens on people who will not work with and not accept them, which is a major challenge to overcome with the introduction of such products or solutions.

“Last but not least a little is the need for a bit of integration, since the glucose management application is a client server solution integrated with the hospital information system, which sits not in the cloud (a remote data centre) but right in the hospital for data privacy and so forth. It’s fully internationalised, so can be used everywhere according to regulations.

So just to remember, our achievements are better results, fewer errors made in the treatment of diabetes and the highest acceptance. We already serve 7,000 beds across several hospitals in Europe and we are growing further”, Krug concluded.