When is a death ... not a death? In thousands of cases, medical device manufacturers have been allowed to classify patient fatalities as less serious injuries, because the U.S. Food and Drug Administration allows companies to make their own determinations on causality.
One patient died a few hours after being implanted with a cardiac stent that was "damaged and stretched." The stent maker called it a "malfunction."
In another instance, a patient developed a fatal, septic infection after receiving a knee injection. The maker of the injection product labeled the death an "injury."
In another case, a patient died on the floor after falling out of a hospital safety bed. Hospital staff said the bed's alert system failed, but the bed manufacturer labeled the incident a "malfunction" instead of a "death."
Even though those patient fatalities happened in close proximity to medical device malfunctions and problems, Angela Stark, an FDA spokeswoman, said, "The reports were likely classified correctly."
How could a patient's death be correctly classified as an injury?
It is a question perplexing to family of Oklahoma's Shelly Wilhite.
She was an epilepsy patient who died in 2010 after being implanted with the Vagus Nerve Stimulator (VNS), a device that emits small shocks intended to block seizures. Less than 48 hours after complaining of abnormally painful shocks, Wilhite's family says she collapsed and died on the bathroom floor.
Since then, her loved ones have come to learn Cyberonics, the manufacturer of the nerve stimulator, has classified dozens of patient deaths as less serious injuries and malfunctions.
"I would like to say to the company to start being honest and to let all of the numbers really reflect the people who have died from this," said Connie Walton, Wilhite's mother.
Wilhite's family filed a lawsuit against Cyberonics, which is a unit of LivaNova, alleging the VNS device caused her death. But the case was dismissed because the VNS passed through the FDA's most stringent approval process, which insulates the company from product liability suits.
In 2001 and 2004, the FDA sent Cyberonics warning letters, accusing the manufacturer of failing to investigate and report patient deaths in a timely manner. The company said all the issues in those letters were corrected.
But after NBC New York and CNBC asked Cyberonics about a number of more recent patient deaths that were classified as injuries and malfunctions, the company did an internal review, finding a total of 108 misclassified deaths. Karen King, a spokeswoman for Cyberonics parent company LivaNova, said in an email the misclassifications were a mix of unintentional human and procedural errors. Cyberonics said the errors are being fixed and noted they make up less than 1 percent of the 34,728 safety reports filed for alleged problems with the VNS.
"Aside from the rare instances described above, there have not been any significant misclassifications of reports," King wrote. She declined to address allegations in the Wilhite death suit, but said more than 90,000 patients have been implanted with the VNS and Cyberonics categorically denies manipulating safety data.
Madris Tomes, a former consultant to the FDA, said the labels "death," "injury" and "malfunction" matter a great deal, because public health analysts must prioritize their work.
"The way that the FDA reviews the [safety] reports is in order of importance by the classification that is checked on the form," Tomes said. "So if it is an injury or malfunction, it could take them weeks to months to get to reading them."
The FDA said analysts do often prioritize death reports, but they also use other criteria to review safety data and identify patient fatalities.
In 2009, an investigation by the Office of Inspector General of the Department of Health and Human Services, found the vast majority of unread safety reports were events classified as "malfunctions" and the FDA's software only assigned about 10 percent of malfunction reports to analysts for review.
Since the inspector general report, the FDA has made investments to better track safety problems with medical devices. The agency is currently upgrading its software platform for reporting problems. The new system would link data on health risks from several different sources including clinical registries, electronic health records, medical billing claims and data transmitted from devices themselves.
After leaving her work consulting to the FDA, Tomes launched a company called Device Events, which offers clients a streamlined way to search FDA safety reports about every medical contraption — from tongue depressors to pace makers.
According to data compiled by Tomes, between 2011 and 2015, device makers classified more than 4,000 patient deaths as "injuries" or "malfunctions."
FDA reporting rules say manufacturers are supposed to report any event that "reasonably suggests a device may have caused or contributed to a death or serious injury."
Just because a patient death occurs doesn't mean a medical device is at fault, and the FDA says there is often insufficient evidence linking a device to a fatality.
But who decides what is reasonable evidence of causality?
Typically the FDA allows manufacturers the discretion to decide when and if their product is related to the death.
For example, in five separate cases, VNS patients were found to have died of pneumonia. Although the VNS instruction manual warns the nerve stimulator could cause patients to aspirate liquid into their lungs — a known pneumonia risk — Cyberonics reasoned three of the patient deaths were properly classified as "injuries" or "malfunctions" because medical professionals attested the device was not related to the pneumonia.
In the other two fatal pneumonia cases, Cyberonics said human errors resulted in reporting mistakes.
Some lawmakers have criticized the FDA for relying on the judgment of industry players when those companies — and the doctors who use their products — have a financial interest in minimizing the appearance of medical risk.
"Are physicians misreporting? Are companies misclassifying? I think the FDA is in the best position to answer those questions, and they should. And I don't think they have, to date," said Rep. Michael Fitzpatrick (R-Pa.).
Specifically, Fitzpatrick has criticized the FDA for its handling of complaints about the permanent birth control implant Essure. Bayer, the device maker, has reported more than 300 miscarriages experienced by patients who became pregnant in spite of the implant. But instructions on filling out the FDA safety reports explicitly say not to check the "death" box if "a fetus is aborted because of a congenital anomaly, or is miscarried."
Catrice Jones, an Essure patient who experienced two miscarriages, said the reporting rules obscure patient harm and favor companies over consumers.
"Once a woman is pregnant, she's pregnant. That's a baby. And if she loses it, whether it's 10 weeks, or 15 weeks, or 24 weeks, that's a life that was taken away," Jones said. "And they're not being labeled as deaths."
Tara DiFlumeri, a spokeswoman for Bayer, said in an email the company reports all adverse events according to FDA rules. She stressed that just because a pregnancy fails, it doesn't mean the Essure implant was the cause.
"All of the Essure data from post-marketing monitoring show that pregnancy-related adverse events are within the background rate for similarly aged populations," DiFlumeri wrote in the email.
Jones and other Essure patients have tried to launch product liability lawsuits against Bayer with limited success because, like the nerve stimulator, the birth control device has passed through the toughest approval process and has broad immunity from most legal claims.
Without the ability to sue, both Essure patients and VNS patients must rely on health regulators to ensure the products are safe.
The FDA said when it comes to balancing the interests of the medical device industry and consumers, regulators favor the public.
"The FDA considers patients to be our first and primary customer," Stark said.