Hey Folks
Is anybody out there storing their radiology data in a way which is more cost efficient than just having “everything available immediately”?
The central region is considering how we might be able to make some cost savings in radiology storage.
A proposal of tiered levels of information has been put forwards: Radiology images would be avaiable at one of four tiers:
immediately available
available with a delay of up to an hour to retrieve data if requested.
cached - delay of up to a day to retreive data.
deleted.
It is anticipated that patients who have current or recent activity in the hospital system would have all their data moved into category 1
dead people’s data would be in groups 3 and 4
the rest would be in group 2
Makes sense to have a graded retrieval if the cost saving is significant but I would wonder how you would apply deleted?
In GP we used to think of using archiving stuff when we started in 1990 odd but disk space just got cheaper and cheaper and bigger and bigger and in the end we did not archive anything but retrieval times get slower as the item becomes more historic. Nothing like an hour though. cant see clinicians looking if it took that long. We live in a world at ED for example where if it is not available in 4 seconds HCP’s just go bare (ie without data)
I wonder if you need a category between 3 and 4- you could put it data into deep archive storage with longer delays in retrieval- easy to do with cloud storage that has low cost deep archiving functionality.
And then look at deletion at the end of a person’s lifespan- using 120 years as an estimated maximum lifespan.
This ensures you retain radiology data for future use cases, such as the current Royal Commission of Inquiry into Historical Abuse in State Care and in the Care of Faith-based Institutions that requires health data and information availability back to 1950.
There are a number of issues that need to be factored into any decision around image caching. The simple ones you’ve already identified, i.e. archiving patients that are deceased.
I’m assuming that if you’re being presented with an option to have hour long retrieval then that is via tape archive (or as Karen says, super low cost cloud archives). It is probably cheaper, however should really only be for imaging that is very old and therefore likely of little to no clinical relevance.
Most PACS/RIS systems should have the ability to do some smart caching, i.e. when an order is placed for imaging to be done and hits a modality (i.e. the CT/MRI etc) then the PACS should already be pre-retrieving images for the reporting radiologist to have instant access to. These pre-retrieval rules are usually configurable but…reflect a storage model that is hybrid i.e. tape/disk. Disk storage is now so cheap that most of the cost benefits for tape are now no longer relevant although low cost slow access cloud archive is the modern equivalent.
Its not really a good idea to use recent activity as the only model for near-time vs archived storage (although for elective imaging pre-retrieval can help). It’s better to think about a moving window of near time (i.e instantly retrievable) imaging versus archived. You could use 7 years as a reasonable benchmark, with rules that keep images on disk if a patient has had recent imaging…i.e. the window is set at the per patient level, not the organisation level.
So…only 2 tiers.
7 years on instant retrieval per patient.
Archive rest.
PACs purges data from instant to archive storage if cost benefit makes sense.
Again - most good PACS vendors should be able to give good advice re this. I know CDHB made the decision to dump tape archives a long time ago so Sharyn MacDonald down there will definitely be able to explain their strategy. Karen’s comment re deep archive storage on cloud is almost certainly a better model than tape.
As clinical user in an imaging dense specialty (stroke), I am not keen on waiting an hour for an image. I am often going back a number of years to see comparative images during my clinical work. And having just done a 3 year audit that included reviewing imaging, an hour delay would have been awful. We recently changed our PACS system. Images over 6 months old initially had to be called from an another archive meaning a delay of up to a minute. That was problematic enough. In terms of destroying images, you have the Privacy Act requiring records to be kept for at least 10 after last patient contact, but also it seems we are under the Public Records Act where destroying records requires authorisation of the Chief Archivist. So while archiving really old imaging and imaging for deceased is probably OK, deleting is likely more trouble than it is worth in an environment where storage is getting so cheap. Otherwise I want rapid access to images of any patients I am seeing.
This is actually a pretty common problem in IT generally, and has been solved several times over. When I was with Canterbury DHB we planned on using cloud native technologies to achieve exactly as you describe:
All images stored in shared local edge for upto X hours
Disk images replaced with references to the same images on remote storage with only seconds delay for longer period, configurable to bringing back locally if accesses more then Y times.
Remote images moved to remote ‘cold storage’ after X months, access time in the hrs
cold storage moved to glacier storage after X years. access time in the days.
Each of these is an native cloud service. you don’t have to worry about anything regarding the underlying physical technology, the provider will ensure that 50 year old images are still accessible tomorrow.
The only constraint you have is that you have to make a trade of between local storage and access risk, the more you store locally the less likely you are to need a 18 month old image in the middle of an earthquake , when the internet is down.
Frankly I would cap it at 2 years, or 2 terrabytes, but I’m not a radiologist.
Thanks to all who have replied, this is very helpful - and I’ll pass your comments on to the rest of our governance group. Good to know that this has been thought about in some detail - at least in the south lsland.
Just to clarify the plan - if any image was taken, or if teh patient has a planned outpaitient, or recent outpatient event, then ALL their data would be pre-fetched. i.e. the MRI Brain from 2006 would be available immediatly to the patient you are currently seeing.
Cheers
Mat