Predictive Analytics Could Have Prevented Disaster – Heartbreak

On the weekend of March 3rd 2017 a cryogenic storage freezer failed. 2,100 embryos and eggs were destroyed and hundreds of families were left devastated. It took only a few hours for the finger pointing and allegations to begin. The Ohio fertility clinic blamed the equipment manufacturer; the OEM blamed the hospital. The families, who trusted their potential progeny to be stored, had little recourse but to file lawsuits. What could predictive analytics via artificial intelligence have done to prevent this? Let’s dig in to the details.

 

What happened?

After a lengthy investigation it was determined that the storage tank was having trouble for weeks. An alarm system had been turned off failing to indicate that the tank’s temperature began to rise. The tank was also undergoing preventive maintenance at the time because of a problem with a system that automatically fills the liquid nitrogen, which keeps the embryos frozen. The manufacturer of the tank, Custom Biogenic Systems, said it didn’t have anything to do with the remote alarm system being turned off. It said the tank functioned properly by indicating a high-temperature condition and activating a local alarm. For potentially weeks, that alarm was alerting staff locally that temperatures were rising out of spec. The staff eventually became annoyed and disabled the alarm. University Hospitals said it doesn’t know who shut off the remote alarm, which should have alerted staff again to changes in the storage tank’s temperature on the weekend of March 3 when no one was at the lab. Because that was turned off and no redundant alert system was in place, 2,100 embryos and eggs were lost.
 

Multiple Failure Points

1. Preventative Maintenance

   It’s fairly clear that the tank’s trouble began with a failure to carry out standard, necessary preventative maintenance. After 5-7 years of service life, the Custom Biogenic Systems tank was known to experience ice build up on the solenoid valves that automatically refill the nitrogen. A defrost cycle was necessary to prevent the valve from sticking. The defrost was not performed and the University Hospital staff was filling the tank manually.

2. Local Alarm Only

   When the tank temperature rose to an alert status, the understaffed hospital clinic was closed for the weekend AND a potentially non-clinical staff member silenced the alarm. Critical alarms with no centralized reporting can (and did) result in catastrophic failures.

3. No Reporting to the OEM

   Though Custom Biogenic Systems was not named in the lawsuit and appears to not be culpable, there was likely significant damage to their reputation in the market. The UK issued a warning after similar incidents had come to light. Googling the trade name results in pages of headlines about their relation to the destroyed embryos. There is little indication, however, that the University Hospital system is the one being held accountable unless one clicks through and reads the entire article.

 

What could have happened?

If the OEM had central monitoring of its deployed assets that indicated whether preventative maintenance procedures (like the defrost) had occurred, it could have alerted the hospital that best practices were not being followed. Furthermore, if the OEM had been able to monitor the individual components of its assets, they could have known a failure was imminent. A stuck valve has a very pronounced electrical signal and can fairly easily be identified as non-nominal behavior.

Predictive Analytics via Artificial Intelligence is the Answer

Sigsense allows equipment manufacturers to unobtrusively monitor deployed assets at the component level. An artificial intelligence algorithm constantly monitors the behavior of the device and compares it to non-nominal behavior. If maintenance procedures are not completed or motors or valves deviate from normal alerts are generated. The problems are then addressed before thousands of headlines are published. It’s true preventative maintenance and in this case, a really great reputation management tool.

Sigsense allows OEMs to understand why components fail before they do. By implementing remote monitoring capabilities OEMs can reduce service calls, downtime and reactive maintenance costs. In high-stakes applications like this, Sigsense could have enabled this manufacturer to protect its customers from a devastating disaster.