There were some important limitations to this study. We did not try other methods of directly contacting patients after their procedures to identify adverse events. We did not use direct mailings, and we did not arrange for follow-up visits in our unit after every procedure. In our practice, doing so is not feasible, although we recognize that for many practices this may be a viable process to assist in adverse event tracking. We also did not use an electronic mailing system because, at the time the study was conducted, there was not a secure patient portal to do so, an obstacle that has since been cleared. Had these elements been part of the study protocol, we likely would have captured more adverse events.
Our study did not include a claims analysis. Although our health system does not use claims analysis to track adverse events, we recognize that many systems have an insurance arm or payer partner that may provide adverse event data based on claims. Medicare claims analysis suggests a higher rate of complications than shown with either one of the processes presented here. A claims process to identify adverse events should identify events no matter where the patient presents for care and also would identify patients who have events as part of the procedural preparation process (e.g., complications of colon purge or hypoglycemia in diabetic patients). In the future, it is likely that the best tracking of adverse events will be via claims data, and it is also this information that will be used to identify high-value endoscopy units. A significant problem we encountered was that our study personnel were able to contact only a portion (17.5%) of the patients we had intended for them to contact. We grossly underestimated the challenges of contacting patients by telephone during regular business hours (a limitation of our institution). Our study personnel also were surprised with the length of time they were kept on the phone with study participants when they did make contact. Although we provided instructions to stick to the telephone interview script, participants often would stray from the subject of the interview and discuss issues such as upcoming appointments, messages they wanted conveyed to their providers, or feedback about their experience. As a result of these issues, we were not able to contact as many patients as required by our a priori sample size calculation. However, the statistical power for the per-protocol analysis (which included only those patients for whom there was an attempt to be contacted: 1,999 patients) was 80.7%.
The funding for this study was directed solely at salary support for the personnel conducting the telephone interviews, and when those funds were exhausted, the study was closed, which is why we were forced to accept the low proportion of patients contacted by telephone.
There are several ways this may have affected our study results. First, patients who can be contacted during business hours may be different from those who could not be contacted. Certainly, if a patient is in the hospital or deceased, contact by telephone would not be successful. This notion is supported by our data, which shows a trend toward adverse events that were identified in the telephone query group being of a lower severity than those identified with the MDI.
Second, the inability to connect by telephone with most of the patients who underwent a procedure may have biased our results in favor of the MDI, assuming that the MDI could capture events in the 82.5% of those patients not contacted by telephone.
It is disappointing that contacting patients directly by telephone was more of a challenge than we anticipated, but it underscores the reality that this one modality may not be sufficient for an adverse event tracking program.
It should be noted that only seven adverse events overlapped between the two systems, which indicates that they were complementary. Not only did we learn that querying by telephone likely is not enough as a singular modality for adverse event tracking, we also learned that our MDI could use another method of contacting patients directly. Our practice does not have the resources to employ personnel to contact all patients by telephone postprocedurally (because we perform approximately 150 procedures daily), so another modality could be considered, such as an electronic mailing to patients, which could be automated and nearly without cost.
This study showed a method, besides direct patient contact, that can track adverse events effectively, in a way that many practices can adopt. We believe our practice has benefited from this infrastructure, not only because of our ability to track adverse events and use trends to improve our practice,20,21 but also because it has promoted a distinct cultural shift toward a focus on quality and safety. Our professional societies and the Accreditation Council for Graduate Medical Education have focused on adverse event tracking as an attempt to develop a culture of reporting and transparency. In this study, two-thirds of all the adverse events identified were reported by physicians (including fellows) or nurses in our own practice. To us, this shows that we not only have developed a viable program for adverse event tracking but that we have created a culture in which providers feel encouraged and empowered to report and discuss quality and safety.
