Presentation outline CNAC N A Comparison of Medical Error Reports Submitted to a Voluntary Patient Safety Reporting System by Different Types of Reporters: A report from the ASIPS Collaborative Daniel M. Harris, PhD, The CNA Corporation, Wilson Pace, MD & Doug Fernald, MA, Univ of CO DFM, Linda Marr, MS, The CNA Corporation ¾Background ¾Research objective ¾Study design ¾Findings ¾Conclusions & implications AcademyHealth Annual Research Meeting June 6, 2004 2 1 Background Research objective ¾ Applied Strategies for Improving Patient Safety (ASIPS) is a 3-yr AHRQ-funded demonstration project based at the Univ of CO Dept of Fam Med ¾ Developed patient safety reporting system (PSRS) ¾ Medical error reporting systems are advocated as a strategy for improving patient safety Collect voluntary narrative reports of medical errors from providers, other clinical staff, and non-clinical staff at primary care practices in two CO PBRN Anonymous or confidential-to-anonymous reports Instructed to report “any event you don’t wish to have happen again that might represent a threat to patient safety” ¾ Overall purpose of project is to test ability of the PSRS to collect incident reports and to use them to design interventions to improve patient safety 3 Study design Reports are consensually coded into multi-axial taxonomy of 400+ attributes by teams of coders Attributes characterize events by type of participants & their contribution, complexity, setting, clinical intent, activities, underlying causes, patient outcomes, interventions, and “discoverer” Attributes are coded “1” if present in report; “0” if not present ¾ Reporters self-classified as provider, other clinical staff, or non-clinical staff Learning from errors requires receiving reports on representative range of errors ¾ Limited evidence exists re: nature of events reported to such systems by different reporter types, especially in ambulatory primary care settings Do different types tend to report different types of events? ¾ Our objective is to (1) analyze reports submitted to ASIPS PSRS to ascertain “who reports what,” (2) identify similarities and differences in reports, and (3) characterize differentiating event attributes by reporter type 4 Findings ¾ Error incident reports are received from 34 primary care practices throughout Colorado ¾ Analysis of report Reports by reporter type Nature of report Event “discoverer” ¾ Analysis of event reported Participants Patient harm Communication errors Other event characteristics Discriminant analysis 522 (85.8%) of 608 reports received thru Aug 2003 identified a reporter type ¾ Report content compared by reporter type using cross tabs, ANOVA, and discriminant analysis 5 6 1 Reports by reporter type ¾ Most reports (68.6%) from providers Unchanged from yr1 to yr2 ¾ About 25% from other clinical staff Increased from yr1 to yr2 ¾ Fewest reports (6.9%) from non-clinical staff Deceased from yr1 to yr2 despite project effort to increase participation Nature of report Percent of Reports within Year, by Reporter Type ¾ No difference by reporter type: 80 ¾ Differences by type: Patient 40 20 0 Prov Oth Clin Non-Clin Total Yr 1 age • Mean lowest for nonclin; highest for provider Patient gender Perceived preventability Perceived pt knows Mean taxonomy, event activity, and “don’t know” codes 60 At least 1 “insufficient information” code • Non-clin Noted most likely in Med Rec’d • Non-clin least likely Perceived • Oth Yr 2 pt harm clin least likely 7 8 Event “discoverer” Event participants ¾ Providers & other clinical staff likely to report events discovered by caregivers like themselves ¾ Non-clinical staff likely to report events discovered by office staff like themselves ¾ Providers somewhat less likely to report events discovered by patients or patient’s families ¾ Each reporter type is more likely to report an event with a participant of the same type Percent of reports discovered by various parties, w/in reporter type 100 80 60 All relationships statistically significant by χ2 ¾ No significant association by reporter type for other types of participants: 40 20 0 Caregvr Prov Ofc Staff Oth Clin Pt/Fam Percent of reports with selected type of participants, within reporter type 60 50 40 30 20 10 0 Patient/family or 3rd party Number of participants or number of participant types (by ANOVA) Prov Prov Non-Clin Oth Clin Oth Clin NonClin Non-Clin 9 Patient harm Communication ¾ Overall, 26.4% of reported events coded w/some form of patient harm Additional 7.7% of reports, coded w/patient may have been harmed, but too early to tell ¾ Significant differences in type of harm (by χ2): Provider reports most likely coded w/clinical harm Non-clinical staff reports most likely coded with nonclinical harm Other clinical staff reports least likely coded with any harm 10 ¾ Over half (57%) of reports involve a communication error between parties Percent of reports coded with patient harm, w/in reporter type 35 30 25 20 15 10 5 0 Clin Harm Non-Clin Any Harm Harm Prov Oth Clin Non-Clin 11 Each reporter type likely to report communication involving their own type ¾ Oth clin staff less likely to report within office communication errors ¾ Overall, non-clin staff less likely to report communication errors ¾ Providers more likely to report communication involving patients 50 Percent of reports with communication errors involving selected parties, w/in reporter type 40 30 20 10 0 Att Prov Party Prov Non-Phy Party Oth Clin Ofc Staff Party Non-Clin 12 2 Other event characteristics Discriminant Analysis (1) ¾ No difference by reporter type: ¾ Identify event attributes that differentiate be tween reports submitted by each reporter type Missing information General proc issue Medication error Supervision error Intervention following error detection Judgement error Clinical knowledge or skill error Resource function or availability error ¾ Differences by type: Documentation • Oth clin most likely Specific in Dx and Tx – prov most likely • Delay in testing – oth clin staff most likely • Dx testing error – non-clin least likely Pt mgmt error • Non-clin • Prov • Prov • • Group centroids are significantly different Provider reports are most different from other 2 types Canonical correlations (~ANOVA Eta) for the functions are .455 & .346; jointly account for 30% of variance in group scores Discriminant scores correctly classify 63% of reports Classify unknown reports in similar proportion 14 most likely Distraction/inattention least likely System Stepwise method: attributes enter that maximize distance between 2 closest groups (16 of 32 entered) Resulting 2 discriminant functions differentiate between the 3 reporter types proc issues • Delay issue least likely 13 Discriminant Analysis (2) Conclusions ¾ The analysis identified attributes that best discriminate between events reported by each reporter group ¾ Provider reports most differentiated by: ¾ Different types of reporters tend to submit reports of different kinds of medical errors to a voluntary PSRS provider participant, communication involving a provider, disclosure to a patient, delay in diagnosis, diagnostic testing error, and problem with resource (availability or function) ¾ Other clinical staff reports most differentiated by: Non-physician provider participant, third party participant, communication involving a provider, communication within the office, delay in performing a test, diagnostic testing error, and misuse of a system ¾ Non-clinical staff reports most differentiated by: Non-clinical staff participant, non-clinical harm, error in patient management, distraction/inattention error, misuse of a system, and malfunction of a system Each type of reporter tends to report errors involving the kinds of activities and participants they can be most expected to observe and believe should be reported For example: • • Providers tend to report clinical harm events while nonclinical staff tend to report non-clinical harm events Providers tend to report events involving clinical procedures while non-clinical staff tend to report patient management and system issue events 15 Implications The CNA Corporation ¾ To be effective in identifying a full and repre sentative range of errors and threats to patient safety, a voluntary PSRS needs to assure that it receives incident reports from all types of staff (and patients?) who are in a position to observe errors 16 Questions? Provider exclusive or dominated reporting systems will restrict this range ¾ Our experience demonstrates the difficulty of obtaining full participation of non - providers More effort necessary to achieve their participation 17 18 3