Risk Analysis - Events which may cause to Data Loss

[eyalhend]

Hello All,
Would appreciate the forum aid with the following subject:
As part of risk analysis process, we are dealing with events which may cause to data loss (due to, for example, file corruption, deletion of data, communication issues, etc.)

As said, the harm was determined as data loss.
Yet, Clinical personnel claims data loss does not cause harm to patient, as the physician shall not relay on previous recorded data for its decision making and suggest not to include risks related to data loss in the risk analysis file, or at least, to separate it to a different category not related to clinical risks (suggested a term of regulatory risks, which I personally don’t except).

The same question raised from security issues, for example, exposing personal data – is that a risk which may cause harm to a patient, and accordingly shall be introduced to the risk analysis?

Thanks a lot in advance!

 

[yodon]

We've bumped into that numerous times. I think if you look at the definition of 'harm' in 14971 it offers a little more flexibility:

physical injury or damage to the health of people, or damage to property or the environment

While it might be a stretch, you could maybe argue that data loss is "damage to property." If you consider the ramifications of the GDPR, it certainly makes sense to take data protection pretty seriously.

We've typically expanded our 'severity' definition to include these aspects.

 

[mihzago]

I would recommend keeping the Security risk assessment separate from the safety risk, since each has a different definitions of harms and different approaches to estimating probability.
Then, any security risks with potential to result in injury could be included in the safety risk assessment. For example, loss of data, depending on the intended use may by a hazardous situation that could lead to a harm. If you have a system to provides diagnosis or treatment for a serious condition, a loss of data could be a bad thing. Similarly, data corruption, for example in a closed loop system would be a very serious problem.
A disclosure of personal or health info, probably very little or no safety risk.

 

[pkost]

Data loss can affect many areas of the manufacturing and developing process. you may also want to consider data corruption that could lead to incorrect decisions at any stage that can then lead to more obvious risks

If you lose data from the manufacturing process, you may struggle to demonstrate the device has been manufactured to specification
If you lose data from the design processes, how can you prove that your device is safe and effective and in compliance with regulations?

You mention that clinical personal disagree - is this with information that is presented to the clinician? If so, why is it being presented if they don't use it? it may not be required, but they may come to use it to support a decision that they otherwise would not have made, in which case there is a risk of harm

It is worth noting that ISO13485:2016 now specifically has a requirement relating to data protection!

 

[Ronen E]

Good responses above.

To add another general comment, I wouldn't necessarily try to force data loss or data breach directly into the "harm" slot. It may contribute or lead to harm in a more indirect way, in which case you need to continue your analysis until you reach an actual harm (or not - that's part of the analysis). Remember that ISO 14971 is about scenarios or combinations of events/conditions that may lead to harm and thus pose a risk.

I agree with the viewpoint that clinical data loss may pose a risk even if that data is not routinely used for making clinical decisions or it's just a support for more important sources. The risk might be lower, but saying that there is absolutely no risk? I don't think so. Every bit of clinical data available to the clinician that might affect their actions is significant. Otherwise there's no point in having that device/service in the first place.

 

[eyalhend]

Thanks all for the responses!
Maybe I need to emphasize some issues:
1) By 'Data Loss', I mean only to clinical data acquired during clinical procedure. It has no relevance neither to manufacturing nor to design
2) An event of Data Loss may occur due to file corruption (for example, due to network issue during transferring it to PACS) or deletion of a file mistakenly by user (with no backup).
3) The clinical team argue/challenge that such Data Loss poses no risk to patient (and therefore - there is no potential harm), as the decision making post-procedure is not (solely) based on the data which is lost. Actually, the system just measures values and present it to the physician. If a redo is required (not due to the data loss), the physician may and will measure again anyway.

Given the above, the risk-management team is challenged to define whether Data Loss is posing a risk.

Hope the above clarifies more the concern.
Thanks!

 

[Ronen E]

the decision making post-procedure is not (solely) based on the data which is lost.

Not solely based on it, but it does contributes to the decision (based on your wording). Correct? So the decision may not be as robust without it. Maybe the effect is small, maybe even negligible, but not zero. The significance should be analysed and the results documented, as a minimum.

the system just measures values and present it to the physician.

What does it mean?... Presents the values for what purpose? If no purpose, why do they need this presentation in the first place?

If a redo is required (not due to the data loss), the physician may and will measure again anyway.

Are you trying to say that data loss does not pose a risk because a redo can be had? That would represent some wasted resources. Maybe not a big risk, but again, also not zero.

 

[eyalhend]

Thanks again.
Decision making: well, clinical personnel may say that a redo may be interpenetrated as a complete new procedure, in the event of data loss
System measurements: I can't get into details, but the system build maps based on measured values. One may say that it only present data on a convenient manner to the physician. But the decision making is only on him. The system don't suggest/diagnose.

Hope it clarifies more.
The inputs I get from you people are extremely helpful.
Thanks!

 

[mihzago]

loss of data may result in 'delay of diagnosis', or 'delay of treatment', or 'prolonged condition'.
the severity will depend on the intended use of your product, i.e. how critical is the information or its timing, how severe the condition, etc.

if you still determine that the loss of data does not result in any harm, then you can still capture it in a separate section that describes limiting factors; or, a section that lists various hazards (one example could be Annex E of ISO14971) where you can indicate that you considered these hazards, but they do not cause harm.

 

[Jen Kirley]

Of all the disciplines that use data, I am baffled the most by medical clinicians declaring loss of data is not a risk to process or patients.

As a patient, I can assure you that my doctors do refer to my clinical history.

Moreover, the ability to securely share data is critical for other health care providers to consider in order to avoid adverse reactions to treatments for patients with more than one condition.

Lastly, having patient clinical history helps medical personnel avoid repeating a complex case's diagnosis that previously was found to be incorrect.