The second most common source of significant inpatient injuries is postoperative deep wound infection (Gawande et al., 1999; Leape et al., 1991). The same hospital-based improvement team as that in case study 1 recognized that infection detection within its hospital, based upon recommendations developed and widely distributed by the Centers for Disease Control and Prevention (CDC), represented a data-based clinical trigger system. The team enhanced the hospital’s ability to detect postoperative deep wound infections by implementing careful patient follow-up after hospital discharge through calls to attending physicians’ offices and, occasionally, directly to patients (injury and near-miss detection). The improvement team also created a working model for infection prevention (hypotheses for change generation), then used expert opinion to focus its model on the timing of delivery of prophylactic antibiotics for clean or clean-contaminated surgery cases (prioritization). For most case types, postoperative infection rates were significantly lower if the antibiotics were started within 2 hours before the initial surgical incision was made, which produced high antibiotic levels in the patient’s blood and tissue at the time the surgery started (Classen et al., 1992).

Having established a strong link between a key intermediate outcome (timing of antibiotic prophylaxis) and the primary outcome of interest (postoperative deep wound and organ space infection rates), the improvement team was able to use a process factor (whether the antibiotic prophylaxis was started within the ideal 2-hour time window) to drive change. Failure to deliver antibiotic prophylaxis within the ideal time window is usually a near miss; only in a minority of cases does the process failure produce an outcome failure. However, use of a process step as a primary performance measure greatly increased the sample size (compared with infection rates) and enhanced the improvement team’s ability to tell when a change had resulted in improvement.

The improvement team then devised (hypotheses for change generation) and tested (rapid-cycle testing) a series of process change hypotheses to bring the hospital closer to the established clinical ideal. Table 5-3 shows on-time antibiotic prophylaxis rates and associated postoperative deep wound and organ space infection rates over time as the hospital’s process improved. The process change that finally worked best in this hospital’s care delivery environment involved fully preparing the intravenous prophylactic antibiotic, then having the anesthesiologist start the medicine immediately after initial induction of surgical anesthesia.