In May 2006, a family practice in Garland, Texas, switched on a new electronic health record. About three hours later, the server crashed, and the system stayed down for the rest of the day.
The clinic, Family Medical Center at North Garland, was the first pilot site for HealthTexas Provider Network, the outpatient arm of Baylor Health Care System. The network then had 88 primary care, specialty and senior health centers and more than 370 physicians, and it planned to move all of them onto one electronic record by December 2008. North Garland went first partly because it had used an earlier electronic system since 1997, so the switch was treated as an upgrade, while most other sites would be moving from paper for the first time. No major technical problems were expected.
For the transition, each doctor’s schedule had been trimmed by only two to four patients per session. When the server went down, the physicians faced an almost full caseload with no complete records for those patients. Test orders, patient reminders and referrals, which normally ran through the system, were disrupted. The clinic worked on paper for most of the day, kept going by a downtime plan it already had. Overnight, the database was rebuilt and a clean copy of the software installed.
The implementation team, with a co-author from North Garland, published what it had learned. Clinicians at North Garland had had only four hours of training in total, and the requirement for physicians and clinical staff at later sites went up to ten hours. Of the first few days of a switch, they wrote, “there is no such thing as too much help.”
The rollout went on. Network policy made the new record mandatory for every patient visit once a practice went live, and sites were scheduled by technical readiness and willingness. A later study estimated that getting ready took an average of 134 hours of staff time per physician, counting doctors, clinical staff and front-office staff.
Then researchers led by Neil Fleming of Baylor Health Care System measured what happened next in 26 primary care practices that went live between June 2006 and December 2008, using monthly billing, payroll and ledger data from 2004 to 2009.
In the first six months, the billed work per full-time doctor, measured in standard work units, fell about 8 percent, and so did visits. Staff per doctor rose about 6 percent. Net income per doctor dropped 16.5 percent. By twelve months, productivity had regained half the lost ground and sat about 4 percent below where it had been.
The average hid a split. Practices that went live in 2006 and 2007 saw productivity and net income recover completely after a year. Practices that went live in 2008, the year the recession hit, had not recovered by the end of the study. The authors wrote that they could not tell how much of that gap came from the recession and how much from differences between the early and late adopters.