Study questions safety of popular robotic surgical device
Katie Carr, CSL
- An increase in the number of adverse events associated with surgical robots prompted Illinois researchers to study their safety.
- The team turned to state-of-the-art techniques in text mining to analyze reports submitted to the FDA on adverse events in surgery.
- The researchers believe that substantial improvements in safety practices and development of advanced safety control mechanisms in the robots could prevent some of the adverse events.
Intuitive Surgical Inc., the manufacturer of the da Vinci surgical robot.
However, the number of adverse events associated with surgical robots has risen steadily over recent years, causing concerns about safety. While the rise in the absolute number of adverse events has resulted, in part, from an increasing number of robotic procedures and greater awareness of the need to report accidents, Illinois researchers - including several from ECE ILLINOIS - are looking to see if there’s more behind this rise in adverse events and recalls of the da Vinci system.
“There are several studies in the medical literature on experiences of different surgical teams and the outcomes of robotic surgery in comparison to traditional techniques,” said ECE graduate student Homa Alemzadeh. “We looked at the problems reported by the manufacturers and users, from an engineering perspective, to assess the safety of the system and its effectiveness across different classes of surgery.”
The da Vinci surgical device came on the market in 1999 and Intuitive Surgical Inc. is currently the only active FDA-approved company in the world that makes robots for minimally invasive surgeries.
A closer look at FDA reports
Alemzadeh and Ravishankar K. Iyer, the George and Ann Fisher Distinguished Professor of ECE, along with Dr. Jai Raman, Professor of Surgery and the Chief of Adult Cardiac Surgery at Rush University Medical Center in Chicago, and Professor Nancy Leveson, a Massachusetts Institute of Technology safety expert, studied reports on the da Vinci system collected by the U.S. Food and Drug Administration over 13 years (2000–2012). They analyzed patient outcomes and causes of adverse events reported for different classes and types of surgeries, as well as the recalls issued by the company. Over that time period, 5,374 adverse events were reported, including 455 patient injuries, 86 deaths, 3,933 device malfunctions and 19 recalls issued by the company; the events involved 109,709 devices and instruments on the market.
“The big message is that we are taking data and information that is being reported to the FDA, but is seldom analyzed at this level of detail, and what we are finding out is that the results are quite startling,” said Iyer, who recently received a 2013 IBM Faculty Award.
In fact, the researchers discovered that while the overall rate of adverse events per procedure, including malfunctions, declined over the 13-year period, the injury and death rates have stayed constant since 2006 and, in fact, have increased for more complex types of surgery.
The team did observe that the vast majority of procedures were successful and did not have any problems. However, they found that robotic surgery is similar to non-robotic minimally invasive techniques in terms of reported adverse events, but with higher complication rates, for complex procedures such as cardiothoracic surgery.
Using natural language processing to analyze big data
The challenge the team faced is that human reporters mainly write the reports submitted to the FDA in unstructured natural language text. To study the reports, the team needed to either read all the data manually or teach a machine how to interpret the semantics of the reports. Using state-of-the-art techniques in text mining, they were able to extract several features from the reports, such as what kind of malfunctions happened, what type of surgery was being done, if there was really an injury or malfunction, and whether the surgery was interrupted because of a robot malfunction. Those distinctions are important in analyzing the data. Often cases were classified only as malfunctions, but they actually also included some sort of injury to patients, such as burning of tissues or puncture of organs. Other cases were incorrectly reported as involving injuries.
“The key to this research is the seamless integration of natural language processing and statistical learning techniques into an automated data analysis framework that Homa has developed, where you can go and look at enormous amounts of data and extract the accurate pieces of information on the accidents,” Iyer said.
This method also evaluates the context in which the decisions were made and the actions taken. These were analyzed to uncover any potential safety hazards or violated safety constraints. All of this helps to identify safety controls that should be added to the design of the system to prevent accidents in the future. Alemzadeh is now working with Raymond Hoagland, an Illinois undergraduate student in electrical engineering, on extending the data analysis framework. The goal is to use the system-theoretic accident model and processes (STAMP) formalism used by CAST, to automatically extract from the reports common causal factors leading to accidents. The hope is that others, such as the FDA, could use this framework to study big data on accidents and safety issues as well.
“What was found is that while the robot works well the majority of the time, when the malfunctions occur, the surgeon and the people around it have little control, and serious safety consequences can result,” Iyer said.
For example, a malfunction in the system could necessitate rescheduling of the surgery to another day, pausing of the surgery to restart the da Vinci system and troubleshoot the errors. It could also involve conversion of the procedure to open or conventional laparoscopic surgery. The last two consequences may increase the length of the surgery and cause other complications.
Where to go next
The team’s analysis of accidents reported across different classes of surgery led them to conclude that surgery classes for which the robots are most commonly used, such as gynecologic and urologic, have much lower rates of death and conversion than non-robotic procedures. More complex procedures, such as those in head and neck and cardiothoracic surgery, are associated with higher rates of complications, which result in higher rates of adverse outcomes.
The researchers believe the likely reasons for many of the accidents include inadequacy of safety controls, limited safety and training practices, and limited surgical experience, but they say additional analysis and investigation is necessary to fully understand the causes of the events.
Their analysis shows that from a technology perspective, substantial improvements of safety practices and development of advanced safety control mechanisms in the robots could prevent some of the adverse events.
The team has already received attention for their findings, including selection of their paper as the Maxwell Chamberlain Memorial Paper in Adult Cardiac Surgery by the Society of Thoracic Surgeons (STS), which was presented at the opening session of the 50th annual meeting of STS in January.
The Wall Street Journal also recognized the importance of the study and published an article last November discussing the team’s work and results. Additionally, the FDA, which is responsible for monitoring the safety of medical devices, as well as the robot’s manufacturing company, have reached out to the researchers to discuss the findings and concerns that were brought to light in the paper.
“Robotic surgery holds great promise. However, presently used robotic surgical systems utilize dated technology and have been plagued by problems with the human-machine interface,” said Dr. Raman. “This is more so in complex procedures. This presents a great opportunity to improve safety, flexibility, and consequently, adaptability of robots in surgery.”
Moving into the future, the findings will help the team and others understand complex human-machine interfaces in medical devices. Studies such as theirs will help develop the next generation of robots, which will be safer, more flexible and more responsive.