OK, before I get rolling, let’s back up a bit. To those that didn’t see my earlier feature, I’ve been dredging up the days when Kaiser caught a lot of heat for what was reputed to be a $3 billion EMR installation. Today, after four more years, Kaiser’s EMR rollout is old news. But even though it hit full stride in 2006 or so, it was such big news that the echoes still remain. So here you have what may be some data from those tumultuous times.
Below, consider the first set of data from (what appears to be) a Kaiser report on its Epic EMR performance. This coincides with the period during the period when whistleblower Justen Deal took his complaints about its performance. Of course, a little bird gave it to me, and as noted previously, I’m fairly sure it wasn’t Justen.
This report, which spans August through November of 2006, looks at a bunch of measurements of network and application performance. I’m not a technical expert, so I can only guess, but truthfully, it looks like the organization did pretty well, especially since nobody, more or less, knew how to scale an EMR for such as large installation.
Not only that, it seems to me that if only 580,000 user hours were blacked out during those four months, vs. almost 63 million potential hours, it’s pretty good performance.
My main question here, having seen this doc, is whether these are cherry-picked network stats. Personally, I’d like to know more about how the application performed on the ground, what latency/response times were, whether the interface took eleventy-odd months of training to use, whether Kaiser did a good job of integrating other data silos, and perhaps most critically, whether clinical care took a disproportionate hit.
What data would you have wanted to see if you were running the show? Check below and tell me what you think.
P.S. By the way, if you want to lighten things up, feel free to check out this video of George Halvorson looking august and scholarly. But I digress…back to the data.
Topline Data from August through early November 2006 for KP HealthConnect
Usage Based User Availability: 99.09%
This represents (Potential User Hours- User Impact Hours)/Potential User
Unique Incident reports: 429
These are incidents which affected the deployment, regional or national totals.
Average Concurrent Users: 8,481
Average number of users on the system during a given month
Potential User Hours: 62,895, 096
Average concurrent users * the total hours in the month
User Impact Hours: 572,241
Calculated for every incident by multiplying the actual number of users affected by the duration of the incident.