Understanding Adverse Events and Death: Root Cause Analysis of California Department of Public Health Case

by Dr. Bradley Truax

(The Truax Group consults with hospitals to improve patient safety and procedures.)

PCA (Patient Controlled Analgesia) safety has been one of our most frequent topics. Serious incidents related to PCA continue to occur frequently. Another case with some valuable lessons learned comes from the California Department of Public Health (CDPH 2012).

In that case an opioid-naïve patient received an estimated total dose of 67.8 mg of morphine over an 8-hour period via PCA following a surgical procedure on his heel. He was found unconscious and not breathing at 3AM and could not be resuscitated.

There were multiple lessons learned in this case. The most obvious one is the total opioid dose in a patient who was apparently opioid-naïve. The patient had a surgical procedure on his heel from 2:15 PM to 6:50 PM. The PCA order, written by a CRNA in the PACU, was for 3 mg morphine per injection with a lockout interval of 10 minutes and a basal rate of 1 mg morphine per hour. The PCA computerized log showed that a total of 20 injections were given plus the basal infusion between 7:10 PM and 3 AM when the patient was found unresponsive and not breathing. The last self-administered dose had been given at 1:27 AM.

Generally in opioid-naïve patients we would begin with a much lower starting dose of morphine and titrate upward as needed for pain control. Also we avoid basal infusions in opioid-naïve patients.

The other obvious problem in this case had to do with lack of appropriate monitoring. The patient was apparently monitored via the medical-surgical post-operative protocol and not monitored via a PCA protocol or policy. It appears the patient was evaluated by nursing every 30 minutes for the first hour on the med/surg floor, then hourly until midnight. There was another nursing evaluation at 2:00 AM. The last evaluation (when the patient was found unresponsive and not breathing) was about 3:00 AM and apparently the nurse entered the patient’s room to check on him because the patient could no longer be heard snoring. The monitoring apparently included blood pressure, pulse, respirations, pain level, and level of arousal. The patient’s oxygen saturation was not monitored (pulse oximetry equipment apparently was available in the room but not utilized). It also does not appear that electrocardiographic monitoring or capnography were used. Note also that it is not clear in the report what specific monitoring parameters had been ordered.

But those are only the more proximate causes in this case. The root causes are much more revealing.

It appears that numerous providers were unfamiliar that the hospital even had a PCA policy. The CRNA who wrote the PCA orders was not familiar with the hospital PCA protocol. The surgeon stated he does not write PCA orders and that all he wrote was “pain control per PCA” and that anesthesia took responsibility for initiating it. An RN and a CNA taking care of the patient on the med/surg unit were both unfamiliar with the PCA policy. Despite everything that has been written about PCA safety issues we still find that hospitals underappreciate the potential dangers and don’t put the appropriate policies, procedures, monitoring and education in place to prevent adverse events related to PCA.

The relatively high dose of morphine being administered by PCA in this case also was not questioned by the pharmacists at the hospital. We certainly recommend that all hospitals have dosage range limits for their high alert drugs that trigger alerts when an ordered dose falls outside those set limits. Sometimes, however, such limits are set wide to accommodate doses sometimes given. For PCA those dose limits might be set high for opioid-tolerant patients, who might need a dose as high as used in this case. They key is having a trigger that forces the pharmacist to verify whether the patient is opioid-tolerant or opioid-naïve. The easiest way to do that is to have a required field on a standardized order set for PCA (computer-based or paper-based) that makes the ordering provider state whether the patient is opioid-tolerant or not. However, because a provider might bypass such order set you still need a flag for the pharmacist to verify that status. (Note also that this really should apply to opioids administered via any route, not just PCA.)

PCA pump setup is one procedure that should require an independent double check by nursing. Though we’ve written a lot about both the pros and cons of the independent double check, when dealing with high alert medications it probably does not hurt to have another set of eyes that might say “hey, I think the dose of this drug might be too high”. It’s not clear if an independent double check was done when the PCA was set up in this case.

The CDPH investigation and hospital RCA/POC also mention several other items that bear comment and there are other potential contributory factors we’ll speculate about.

The policy at the hospital in this incident did apparently require nursing to verify that the patient’s cognitive function was adequate to understand and participate in the PCA process. Confusion or impaired cognition are clearly contraindications to PCA. However, we also feel there are several other relative contraindications. These include sleep apnea, COPD, severe neuromuscular weakness, skeletal conditions that may restrict respiration, and others. Perhaps the most important consideration in this case would be sleep apnea. We know from the report that the patient was obese and loud snoring was heard on several occasions. So it is a distinct possibility this patient may have had sleep apnea and we’ll comment on that further below. That is one of the reasons we advocate use of a screening tool for sleep apnea, such as the STOP-Bang questionnaire, in all patients undergoing surgery. If such a tool suggests that sleep apnea is likely, we recommend special precautions if opioids are administered via any route (eg. continuous monitoring in an ICU or stepdown setting).

Who should be able to order PCA? In this case the order was written by a CRNA. It is not clear in the report whether or not that CRNA was part of a pain management team that would be following the patient throughout the hospitalization. If not, wouldn’t it make more sense if the order were written by a physician who would be following that patient on an ongoing basis? Who would nursing call if a question arose at night related to the PCA (the surgeon or the CRNA?)? Quite frankly, we actually recommend that providers should require specific hospital privileges to order PCA. Just as we privilege providers to do moderate sedation, those desiring to use PCA should undergo appropriate educational sessions and be certified to order PCA. We’ve simply seen too many errors related to PCA to allow all providers to order PCA without knowing they are proficient in doing so.

The orders for PCA depend on whether the patient is opioid-naïve or opioid-tolerant. That is an issue often neglected in writing the order for PCA, as it apparently was in this case. Using a standard order set (paper or computer) is a good way to force attention to opioid tolerance level. But that still is only good if the provider uses that form. In the current case a standard order form is alluded to in the hospital policy but it was obviously not used in this case.

Monitoring patients on PCA remains problematic. Any form of monitoring that is intermittent is obviously very risky. Pulse oximetry is recommended for all and it should be continuous pulse oximetry. But we’ve warned many times in the past that pulse oximetry may also provide false reassurances. Opioid-induced respiratory depression leads to hypercarbia before significant hypoxemia occurs. That is especially true in patients who are receiving supplemental oxygen (so supplemental oxygen should be used with caution in patients on PCA).

Patients on PCA (or receiving opioids via other route) may have normal ventilation and oxygenation when you arouse them (as opposed to what happens when they are sleeping). So if you only check their pulse oximetry when you arouse them you will almost never document hypoxemia.

Capnography has really now become the standard of care for avoiding post-operative opioid-induced respiratory depression and has been endorsed by numerous specialty organizations. However, a recent article does point out that the capnorgraphy used in unintubated patients is sidestream capnography and this is not particularly good for detecting hypercarbia (ASGE Technology Committee 2013). However, it can be used to monitor the respiratory rate and detect apnea and both of those are critical to monitor in the post-op patient on opioids.

Back to the issue of possible sleep apnea. Apparently no one was bothered by the loud snoring in the case at hand. Particularly in an obese patient it should have been a red flag that this patient might have had sleep apnea. Opioids in patients with sleep apnea may not cause the typical picture we associate with CO2 narcosis, i.e. a progressive hypoventilation resulting in CO2 retention. Rather, patients with sleep apnea undergo periodic hypoventilation broken up by “arousals”. Eventually the opioids blunt the ability to have physiological arousals and that is when deaths result.

This patient was described as being “arousable” at 2AM and was apparently alert enough to self-dose with the PCA pump at 1:27AM. That suggests to us this was not simply a case of opioid-induced CO2 narcosis. Rather, it suggests the third pattern of unexpected in-hospital death described in the excellent article by Lynn and Curry (Lynn 2011) on alarms and their failure to identify deteriorating patients early. Our February 22, 2011 Patient Safety Tip of the Week “Rethinking Alarms” highlighted that article in which they describe 3 patterns of unexpected in-hospital deaths and demonstrate the problems with threshold-based alarms (almost all currently used alarm systems use threshold-based principles) in detecting early deterioration. Indeed, they posit that threshold-based alarms themselves often cause us to miss signs of early deterioration.

The second pattern of deterioration they describe is the classic CO2 narcosis. As CO2 rises one sees further central depression of respiration and a vicious cycle. It is notoriously difficult at the bedside to predict who is hypoventilating. This is a pattern for which we have advocated using a combination of sedation scales and capnography. However, the authors point out that even such monitoring may be inadequate because many of these cases are complicated by the third pattern described below. The important point is that there may be a huge difference when the patient is awake and when he/she is asleep.

The third pattern they describe is one that is typically seen in sleep apnea. In this pattern one sees repetitive reductions in airflow and oxygen saturation during sleep followed by arousals. The arousals rescue the patient but eventually the capacity or reserve of the patient to recover with arousals becomes impaired (often in response to narcotics or sedatives) and the patient may experience sudden death during sleep. The authors discuss the inability of currently used oximeters to recognize this pattern. They even imply that this pattern may give rise to oximeters alarming and being interpreted as “false” alarms attributed to motion artifact, etc. because when staff respond to the alarm the patient is now awake, breathing normally and has a normal oxygen saturation.

See the list of our previous columns on sleep apnea in the perioperative period at the end of today’s column.

Education at several levels is imperative. At the medical staff level we’ve already mentioned our recommendation about requiring specific privileges to order PCA and having them have appropriate training prior to granting those privileges. At the nursing level all staff should receive inservices on PCA and be required to demonstrate competency on PCA annually.

At the patient level there are several considerations. When PCA was set up in the PACU nursing noted that PCA was explained to the patient and questions were answered. It is not clear what the patient’s capability of understanding was at that time since he may well have had residual effects from anesthesia and other medications given during and immediately after his surgery. Patient education should begin before the day of surgery, either in the surgeon’s office or at a pre-op visit, and should include both verbal and printed materials. Such education should also be given to any family members likely to be with the patient in the hospital.

And there is also the issue of education at the family level. It is also not clear whether any education was provided to family members in this case. The report does state that the patient’s wife left around 11PM or 12AM so she should have received education about the PCA process. Because of the well-known problem of “PCA by proxy” (where family or others push the PCA button to administer analgesic) it is imperative that family and visitors be educated not to manipulate the PCA pump in any way. We also recommend specific signage be affixed to the PCA pump to warn family and visitors not to manipulate the pump.

Communication and handoffs are critical in most incidents that result in patient harm and this case is no different. In this case there was an apparent handoff from one CRNA to another CRNA. Did the second CRNA have adequate knowledge of the patient’s history and current status? We don’t like to see any of the key players in surgery (physicians, residents, nurses, surgical techs, anesthesiologists, CRNA’s) change in the middle of a procedure. Scheduling and timing of cases should take into account the need to maintain a full complement of personnel through the entire procedure.

And there was a handoff from the PACU nursing staff to the nursing staff on the med/surg floor. Were any standardized handoff tools used? We recommend that PCA be a specific item that needs to appear on such nursing handoff tools.

And there are other potential contributing factors we cannot adequately assess from the CDPH report. For example, the surgical case was of fairly long duration and ended around 7PM. We don’t know if ending that late in the day put any time pressures on staff. For example, was there any pressure to move the patient from the PACU to the med/surg floor so the PACU staff could go home? Did the timing and/or duration of the case have anything to do with the change in CRNA’s noted above?

For further resources and patient safety tips on PCA by Dr. Truax, please click here.

Leave a Comment, if You Care About Patient Safety