The Data Migration Blues

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon

12:30 AM Pity the poor imaging informaticist who has a data migration in their future. That potentially includes everyone who owns a PACS.

“PACS has a 5 to 7 year life cycle,” explained Richard H. Wiggins III, MD, University of Utah Health Sciences Center. Wiggins sat on an afternoon panel, PACS/RIS Replacement Debate, with William Gregg Jr, MS, LSU Health Sciences Center; Richard “Skip” Kennedy, MSc, Kaiser Permanente Medical Group; and David Wild, Fujifilm Medical Systems. Charles Socia, RT, Empiric Systems/Fujifilm, moderated.

Forewarned is forearmed, so Socia laid out the ugly truth with a discourse on the nature of data structure and the specificity of HL7 message types. For instance, a DFT (data financial transfer) message can be hundreds of characters long, and there is an ADT (admission discharge transfer) message for every type of patient transaction. Needless to say, the opportunities for error are many in a field where data integrity is crucial for everything from taking care of patients to getting paid.

So this is what you need to know: painstaking upfront preparation will mitigate headaches on the backend in the form of corrupted and unmatched data. And when conducting a data migration, patience is not only virtuous, it is required. Kennedy just completed a 1 petabyte migration—it only took a year and a half.

Key takeaways included:

• Insure vendor cooperation in data migration by writing it in the contract when you buy the system. No one likes losing customers, and without a contract that binds the vendor to migration support in a timely manner, you are likely to find your entreaties for help unanswered.

• Everything has its price, and a big data migration can run hundreds of thousands of dollars. By negotiating up front, you can protect yourself from price inflation.

• In planning your communications protocols, be sure to have one piece of information that is key to every record in your database or risk an unacceptable level of data corruption.

• Vendor neutral archives enable a switch from one system to the other without having to rewrite the databases, but it’s not magic and you pay for the service. These VNA vendors simply know the languages of all of the other vendor databases.

• Even with storage costs dropping, it is not practical to keep every bit of data forever. The more you have the more it costs to migrate.

• Knowing your data retention policy helps drive your migration policies, so be sure to review it before you begin a migration. There are legal and clinical ramifications to what you dump, so be sure to engage legal and clinical minds in these decisions.

• Vendors are hungry for business: as you are going through the acquisition process, don’t hesitate to ask for proof that they can assist in a migration with a test migration of one month’s data.

• At the start of a data migration, schedule daily meetings to look at the percentage of corrupt data for the purpose of figuring out what the problems are.

• In addition to IT and clinical representatives, include people who are expert at data management on your migration team. People who understand the life cycle of data through multiple system and where HL7 from one system matches up to another can reduce the problem of unmatched data.

The sites that drive the data management project from their side and use the vendor to support those efforts are the ones that are most successful.