Naming the "baby" or the "beast"? The importance of concepts and labels in healthcare safety investigation.

This paper focuses on concepts and labels used in investigation of adverse events in healthcare. The aim is to prompt critical reflection of how different stakeholders frame investigative activity in healthcare and to discuss the implications of the labels we use. We particularly draw attention to issues of investigative content, legal aspects, as well as possible barriers and facilitators to willingly participate, share knowledge, and achieve systemic learning. Our message about investigation concepts and labels is that they matter and influence the quality of investigation, and how these activities may contribute to system learning and change. This message is important for the research community, policy makers, healthcare practitioners, patients, and user representatives.

Extraglottic Airway Device Misplacement: A Novel Classification System and Findings in Postmortem Computed Tomography.

STUDY OBJECTIVE: Extraglottic airway devices are frequently used during cardiac arrest resuscitations and for failed intubation attempts. Recent literature suggests that many extraglottic airway devices are misplaced. The aim of this study is to create a classification system for extraglottic airway device misplacement and describe its frequency in a cohort of decedents who died with an extraglottic airway device in situ. METHODS: We assembled a cohort of all decedents who died with an extraglottic airway device in situ and underwent postmortem computed tomographic (CT) imaging at the state medical examiner's office during a 6-year period, using retrospective data. An expert panel developed a novel extraglottic airway device misplacement classification system. We then applied the schema in reviewing postmortem CT for extraglottic airway device position and potential complications. RESULTS: We identified 341 eligible decedents. The median age was 47.0 years (interquartile range 32 to 59 years). Out-of-hospital personnel placed extraglottic airway devices in 265 patients (77.7%) who subsequently died out of hospital; the remainder died inhospital. The classification system consisted of 6 components: depth, size, rotation, device kinking, mechanical blockage of ventilation opening, and injury. Under the system, extraglottic airway devices were found to be misplaced in 49 cases (14.4%), including 5 (1.5%) that resulted in severe injuries. CONCLUSION: We created a novel extraglottic airway device misplacement classification system. Misplacement occurred in greater than 14% of cases. Severe traumatic complications occurred rarely. Quality improvement activities should include review of extraglottic airway device placement when CT images are available and use the classification system to describe misplacements.

[Medical error in theory and practice - a review of the most important issues]. / Blad medyczny w teorii i praktyce ­ przeglad najwazniejszych zagadnien.

In recent years, in Poland, despite the lack of an adverse medical events monitoring system, a sharp increase in the number of complaints to various medical and legal institutions, as well as court cases with a suspicion of a medical error, was found, based on the available reports and statistics, which poses a serious medical and legal. The aim of this study was to review the theoretical and practical issues of medical errors in the medico-legal context on the basis of the current legislation in Poland. This paper presents the conceptual scope and the evolution of terminology, starting from "error in the medical art/craft" up to the currently defined and used concept of "medical error." The problem of medical errors in medico-legal categories, according to Polish legal regulations and ethical standards in medicine, was also considered. Different classifications, as well as the causes and consequence of various medical errors, were analyzed. Based on current literature, Polish judicial decisions were reviewed, and some examples of legal rulings with respect to different categories of medical errors were presented. Given the ambiguity, both in conceptual and categorizing terms, with regard to adverse medical events: errors, negligence, malpractice and omission, it would be justified to adopt an unambiguous definition and classification. Such an arrangement would expand the possibilities of research in the field of etiology of medical errors, and more importantly, prepare such procedures that would maximally protect the patient, and allow the maximum reduction of the number of medical errors and any other adverse events. In addition, specifying the medical, legal and economic standards in medical units, and determining the scope of personal and institutional responsibility for undesirable medical events, would, in turn, improve the processing of claims made by patients or their families, as well as the activities of medical and legal institutions, including doctors appointed as court experts. Med Pr. 2020;71(5):613-30.

Changes in error patterns in unanticipated trauma deaths during 20 years: In pursuit of zero preventable deaths.

BACKGROUND: A fundamental goal of continuous process improvement programs is to evaluate and improve the ratio of actual to expected mortality. To study this, we examined contributors to error-associated deaths during two consecutive periods from 1996 to 2004 (period 1) and 2005 to 2014 (period 2). METHODS: All deaths at a level I trauma center with an anticipated probability of death less than 50% and/or identified through process improvement committees were examined. Demographics were assessed for trend only because period 1 data were only available in median and interquartile range. Each death was critically appraised to identify potential error, with subsequent classification of error type, phase, cause, and contributing cognitive processes, with comparison of outcomes made using χ test of independence. RESULTS: During period 1, there were a total of 44,401 admissions with 2,594 deaths and 64 deaths (2.5%) associated with an error, compared with 60,881 admissions during period 2 with 2,659 deaths and 77 (2.9%) associated with an error. Deaths associated with an error occurred in younger and less severely injured patients in period 1 and were likely to occur during the early phase of care, primarily from failed resuscitation and hemorrhage control. In period 2, deaths occurred in older more severely injured patients and were likely to occur in the later phase of care primarily because of respiratory failure from aspiration. CONCLUSION: Despite injured patients being older and more severely injured, error-associated deaths during the early phase of care that was associated with hemorrhage improved over time. Successful implementation of system improvements resolved issues in the early phase of care but shifted deaths to later events during the recovery phase including respiratory failure from aspiration. This study demonstrates that ongoing evaluation is essential for continuous process improvement and realignment of efforts, even in a mature trauma system. LEVEL OF EVIDENCE: Therapeutic/Care Management, level IV.

Frequency and classification of errors in laboratory medicine at a veterinary teaching hospital in the United States.

BACKGROUND: The accuracy of laboratory data is of utmost importance. Data regarding laboratory error in human laboratories are often extrapolated into veterinary settings. One study investigated the rate and type of errors in a European commercial veterinary laboratory, but that data might not directly apply to an educational setting. OBJECTIVES: This study determined the frequency and type of errors in laboratory medicine at a veterinary teaching hospital. METHODS: Errors associated with clinical pathology samples were recorded over two 60-day periods. The first period included a time when new students and house officers started at the veterinary school. The second time period was 6 months later. The errors were assigned to categories, and the frequency of each was calculated. Sample hemolysis, icterus, and lipemia were evaluated separately using an automated index, as these conditions could be pathologic or the result of error. Frequencies of error and hemolysis, icterus, and lipemia were assessed between the groups. RESULTS: Total error rates were 4.7% and 3.5% for the first and second periods, respectively. The frequency of each error subclassification was similar to those observed in the veterinary and human literature, with preanalytic error predominating. Statistically significant differences in the overall error rate and percentage of preanalytic errors that occurred outside of and within the laboratory were observed comparing differences between the two periods. CONCLUSIONS: The overall error rate in this veterinary teaching hospital was slightly higher than that previously reported in other settings, although a proportion of errors was as expected. Areas needing improvement were identified, and strategies to reduce error could be developed.

Using the WHO International Classification of patient safety framework to identify incident characteristics and contributing factors for medical or surgical complication deaths.

This study aimed to operationalise and use the World Health Organisation's International Classification for Patient Safety (ICPS) to identify incident characteristics and contributing factors of deaths involving complications of medical or surgical care in Australia. A sample of 500 coronial findings related to patient deaths following complications of surgical or medical care in Australia were reviewed using a modified-ICPS (mICPS). Over two-thirds (69.0%) of incidents occurred during treatment and 27.4% occurred in the operating theatre. Clinical process and procedures (55.9%), medication/IV fluids (11.2%) and healthcare-associated infection/complications (10.4%) were the most common incident types. Coroners made recommendations in 44.0% of deaths and organisations undertook preventive actions in 40.0% of deaths. This study demonstrated that the ICPS was able to be modified for practical use as a human factors taxonomy to identify sequences of incident types and contributing factors for patient deaths. Further testing of the mICPS is warranted.

Characterization of adverse events occurring during nursing clinical rotations: A descriptive study.

INTRODUCTION: Patient safety is a pillar of quality health care. Nursing students may commit errors during clinical practice, compromising patient safety. OBJECTIVE: Analyze the adverse events, as well as the factors associated with the errors, reported by students from a private university in Santiago, Chile during nursing clinical rotations. METHODS: Quantitative cross-sectional descriptive study. A total of 68 errors by first- through fifth-year nursing students were reported between 2012 and 2018. The data collection instrument was the Adverse Events Notification Form from the School of Nursing. This form documented information about the study as well as about the event. RESULTS: After this reporting system was established in 2012, the number of events reported increased steadily each year. The greatest numbers of reported errors were committed by fifth-year students (73.5%), and the most common type of error was associated with medication administration (94.2%), including incorrect dose (27.9%) and incorrect medication (17.6%). The major factors contributing to errors were failure to review the "10 rights of medication administration" (85.3%) or lack of critical judgment (7.4%). Most of the errors occurred in public institutions (72.1%). CONCLUSION: The results suggest that it would be beneficial to re-evaluate how safety and quality of care are taught at the school of nursing, with an emphasis on understanding the learning styles of students and teaching strategies of instructors. It is crucial that the academic institution remain actively involved in teaching safety-related skills to future nursing professionals. Furthermore, we suggest modifications to the adverse events reporting system that would avoid the need for personal interpretations of the event by the student.

Using convolutional neural networks to identify patient safety incident reports by type and severity.

OBJECTIVE: To evaluate the feasibility of a convolutional neural network (CNN) with word embedding to identify the type and severity of patient safety incident reports. MATERIALS AND METHODS: A CNN with word embedding was applied to identify 10 incident types and 4 severity levels. Model training and validation used data sets (n_type = 2860, n_severity = 1160) collected from a statewide incident reporting system. Generalizability was evaluated using an independent hospital-level reporting system. CNN architectures were examined by varying layer size and hyperparameters. Performance was evaluated by F score, precision, recall, and compared to binary support vector machine (SVM) ensembles on 3 testing data sets (type/severity: n_benchmark = 286/116, n_original = 444/4837, n_independent = 6000/5950). RESULTS: A CNN with 6 layers was the most effective architecture, outperforming SVMs with better generalizability to identify incidents by type and severity. The CNN achieved high F scores (> 85%) across all test data sets when identifying common incident types including falls, medications, pressure injury, and aggression. When identifying common severity levels (medium/low), CNN outperformed SVMs, improving F scores by 11.9%-45.1% across all 3 test data sets. DISCUSSION: Automated identification of incident reports using machine learning is challenging because of a lack of large labelled training data sets and the unbalanced distribution of incident classes. The standard classification strategy is to build multiple binary classifiers and pool their predictions. CNNs can extract hierarchical features and assist in addressing class imbalance, which may explain their success in identifying incident report types. CONCLUSION: A CNN with word embedding was effective in identifying incidents by type and severity, providing better generalizability than SVMs.

Where are we in patient safety in the ED in Turkey?

Near misses and unsafe conditions have become more serious for patients in emergency departments (EDs). We aimed to search the near misses and unsafe conditions that occurred in an ED to improve patient safety.This was a retrospective analysis of a 10-year observational period from January 1, 2007 to December 31, 2016. We gained access to the adverse event notification forms (AENFs) sent to the hospital quality department from the ED. Patient age, sex, and date of presentation were recorded. The near misses and unsafe conditions were classified into 7 types: medication errors, falls, management errors, penetrative-sharp tool injuries, incidents due to institution security, incidents due to medical equipment, and forensic events. The outcome of these events was recorded.A total of 220 AENF were reported from 294,673 ED visits. The median age of the 166 patients was 60 (21-95) years. Of these, 57.1% of the patients were females and 47.9% were males. The most commonly reported events were medication errors (32.7%) and management errors (27.3%). The median age of falling patients was 67.5 years. The nurse-patient ratio between 2007 to 2011 and 2011 to 2016 were 1/10 and 1/7, respectively. We found that when this ratio increased, the adverse events results were less significant (P < .003).This was the 1st study investigating the adverse events in ED in Turkey. The reporting ratio of 0.07% for the total ED visits was too low. This showed that adverse events were under-reported.

Combined SNA and LDA methods to understand adverse medical events.

OBJECTIVE: To compare primary medical adverse event keywords from reporters (e.g. physicians and nurses) and harm level perspectives to explore the underlying behaviors of medical adverse events using social network analysis (SNA) and latent Dirichlet allocation (LDA) leading to process improvements. DESIGN: Used SNA methods to explore primary keywords used to describe the medical adverse events reported by physicians and nurses. Used LDA methods to investigate topics used for various harm levels. Combined the SNA and LDA methods to discover common shared topic keywords to better understand underlying behaviors of physicians and nurses in different harm level medical adverse events. SETTING: Maccabi Healthcare Community is the second largest healthcare organization in Israel. DATA: 17,868 medical adverse event data records collected between 2000 and 2017. METHODS: Big data analysis techniques using social network analysis (SNA) and latent Dirichlet allocation (LDA). RESULTS: Shared topic keywords used by both physicians and nurses were determined. The study revealed that communication, information transfer, and inattentiveness were the most common problems reported in the medical adverse events data. CONCLUSIONS: Communication and inattentiveness were the most common problems reported in medical adverse events regardless of healthcare professional reporting or harm levels. Findings suggested that an information-sharing and feedback mechanism should be implemented to eliminate preventable medical adverse events. Healthcare institutions managers and government officials should take targeted actions to decrease these preventable medical adverse events through quality improvement efforts.

Critical Review, Development, and Testing of a Taxonomy for Adverse Events and Near Misses in the Emergency Department.

OBJECTIVES: An adverse event (AE) is a physical harm experienced by a patient due to health care, requiring intervention. Describing and categorizing AEs is important for quality and safety assessment and identifying areas for improvement. Safety science suggests that improvement efforts should focus on preventing and mitigating harm rather than on error, which is commonplace but infrequently leads to AEs. Most taxonomies fail to describe harm experienced by patients (e.g., hypoxia, hemorrhage, anaphylaxis), focusing instead on errors, and use categorizations that are too broad to be useful (e.g., "communication error"). We set out to create a patient-centered, emergency department (ED)-specific framework for describing AEs and near misses to advance quality and safety in the acute care setting. METHODS: We performed a critical review of existing taxonomies of harm, evaluating their applicability to the ED. We identified and adopted a classification framework and developed a taxonomy using an iterative process categorizing approximately 600 previously identified AEs and near misses. We reviewed this taxonomy with collaborators at four medical centers, receiving feedback and providing clarification. We then disseminated a set of representative scenarios for these safety experts to categorize independently using the taxonomy. We calculated interrater reliability and performance compared to our criterion standard. RESULTS: Our search identified candidate taxonomies for detailed review. We selected the Adventist Health Systems AE taxonomy and modified this for use in the ED, adopting a framework of categories, subcategories, and up to three modifiers to further describe events. On testing, overall reviewer agreement with the criterion standard was 92% at the category level and 88% at the subcategory level. Three of the four raters concurred in 55 of 59 scenarios (93%) and all four concurred in 46 of 59 scenarios (78%). At the subcategory level, there was complete agreement in 40 of 59 (68%) scenarios and majority agreement in 55 of 59 instances (93%). Performance of individual raters ranged from very good (88%, 52/59) to near perfect (98%, 58/59) at the main category level. CONCLUSIONS: We developed a taxonomy of AEs and near misses for the ED, modified from an existing framework. Testing of the tool with minimal training yielded high performance and good inter-rater reliability. This taxonomy can be adapted and modified by EDs seeking to enhance their quality and safety reviews and characterize harm occurring in their EDs for quality improvement purposes.

Can We Avoid Implant-selection Errors in Total Joint Arthroplasty?

BACKGROUND: Implant selection in the operating room is a manual process. This manual process combined with complex compatibility rules and inconsistent implant labeling may lead to implant-selection errors. These might be reduced using an automated process; however, little is known about the efficacy of available automated error-reduction systems in the operating room. QUESTIONS/PURPOSES: (1) How often do implant-selection errors occur at a high-volume institution? (2) What types of implant-selection errors are most common? METHODS: We retrospectively evaluated our implant log database of 22,847 primary THAs and TKAs to identify selection errors. There were 10,689 THAs and 12,167 TKAs included during the study period from 2012 to 2017; there were no exclusions and we had no missing data in this study. The system provided an output of errors identified, and these errors were then manually confirmed by reviewing implant logs for each case found in the medical records. Only those errors that were identified by the system were manually confirmed. During this time period all errors for all procedures were captured and presented as a proportion. Errors identified by the software were manually confirmed. We then categorized each mismatch to further delineate the nature of these events. RESULTS: One hundred sixty-nine errors were identified by the software system just before implantation, representing 0.74 of the 22,847 procedures performed. In 15 procedures, the wrong side was selected. Twenty-five procedures had a femoral head selected that did not match the acetabular liner. In one procedure, the femoral head taper differed from the femoral stem taper. There were 46 procedures in which there was a size mismatch between the acetabular shell and the liner. The most common error in TKA that occurred in 46 procedures was a mismatch between the tibia polyethylene insert and the tibial tray. There were 13 procedures in which the tibial insert was not matched to the femoral component according to the manufacturer's guidelines. Selection errors were identified before implantation in all procedures. CONCLUSIONS: Despite an automated verification process, 0.74% of the arthroplasties performed had an implant-selection error that was identified by the software verification. The prevalence of incorrect/mismatched hip and knee prostheses is unknown but almost certainly underreported. Future studies should investigate the prevalence of these errors in a multicenter evaluation with varying volumes across the involved sites. Based on our results, institutions and management should consider an automated verification process rather than a manual process to help decrease implant-selection errors in the operating room. LEVEL OF EVIDENCE: Level IV, therapeutic study.

Evaluation of an electronic health record structured discharge summary to provide real time adverse event reporting in thoracic surgery.

BACKGROUND: The reporting of adverse events (AE) remains an important part of quality improvement in thoracic surgery. The best methodology for AE reporting in surgery is unclear. An AE reporting system using an electronic discharge summary with embedded data collection fields, specifying surgical procedure and complications, was developed. The data are automatically transferred daily to a web-based reporting system. METHODS: We determined the accuracy and sustainability of this electronic real time data collection system (ERD) by comparing the completeness of record capture on procedures and complications with coded discharge data (administrative data), and with the standard of chart audit at two intervals. All surgical procedures performed for 2 consecutive months at initiation (Ti) and 1 year later (T1yr) were audited by an objective trained abstractor. A second abstractor audited 10% of the charts. RESULTS: The ERD captured 71/72 (99%) of charts at Ti and 56/65 (86%) at T1yr. Comparing the presence/absence of complications between ERD and chart audit demonstrated at Ti a high sensitivity and specificity, positive predictive value (PPV) of 95.5%, negative predictive value (NPV) of 93.9% with a kappa of 0.872 (95% CI 0.750 to 0.994), and at T1yr a sensitivity, specificity, PPV and NPV of 100% with a kappa of 1.0 (95% CI 1.0). Comparing the presence/absence of complications between administrative data and chart audit at Ti demonstrated a low sensitivity, high specificity and a kappa of 0.471 (95% CI 0.256 to 0.686), and at T1yr a low sensitivity, high specificity of 85% and a kappa of 0.479 (95% CI 0.245 to 0.714). CONCLUSIONS: We found that the ERD can provide accurate real time AE reporting in thoracic surgery, has advantages over previous reporting methodologies and is an alternative system for surgical clinical teams developing AE reporting systems.

Reporting medical device safety incidents to regulatory authorities: An analysis and classification of technology-induced errors.

The European Union Medical Device Directive 2007/47/EC1 defines software with a medical purpose as a medical device. The implementation of health information technology suffers from patient safety problems that require effective post-market surveillance. The purpose of this study was to review, classify and discuss the incident data submitted to a nationwide database of the Finnish National Competent Authority with other forms of data. We analysed incident reports submitted to the authority database by users of electronic health records from 2010 to 2015. We identified 138 valid reports. Adverse events associated with electronic health record vulnerabilities, clustered around certain error types, cause serious harm and occur in all types of healthcare settings. The low rate of reported incidents raises questions about not only the challenges associated with medical software oversight but also the obstacles for reporting.

Preventable deaths in a French regional trauma system: A six-year analysis of severe trauma mortality.

BACKGROUND: Analyzing mortality in a mature trauma system is useful to improve quality of care of severe trauma patients. Standardization of error reporting can be done using the classification of the Joint Commission on the Accreditation of Healthcare Organizations (JCAHO). The aim of our study was to describe preventable deaths in our trauma system and to classify errors according to the JCAHO taxonomy. METHODS: We performed a six-year retrospective study using the registry of the Northern French Alps trauma network (TRENAU). Consecutive patients who died in the prehospital field or within their stay at hospital were included. An adjudication committee analyzed deaths to identify preventable or potentially preventable deaths from 2009 to 2014. All errors were classified using the JCAHO taxonomy. RESULTS: Within the study period, 503 deaths were reported among 7484 consecutive severe trauma patients (overall mortality equal to 6.7%). Seventy-two (14%) deaths were judged as potentially preventable and 36 (7%) deaths as preventable. Using the JACHO taxonomy, 170 errors were reported. These errors were detected both in the prehospital setting and in the hospital phase. Most were related to clinical performance of physicians and consisted of rule-based or knowledge based failures. Prevention or mitigation of errors required an improvement of communication among caregivers. CONCLUSIONS: Standardization of error reporting is the first step to improve the efficiency of trauma systems. Preventable deaths are frequently related to clinical performance in the early phase of trauma management. Universal strategies are necessary to prevent or mitigate these errors.

Adverse events identified by the global trigger tool at a university hospital: A retrospective medical record review.

OBJECTIVES: The purpose of this study was to describe the level, preventability and categories of adverse events (AEs) in Chinese geriatric patients identified by medical record review using the Global Trigger Tool. The applicability of the GTT was also assessed to explore possible modifications for trigger tools. METHODS: The study was conducted at a 4300-bed tertiary teaching hospital. Twenty randomly-selected medical records for patients over 60 were reviewed every 2 weeks from January 1 2015 to December 31st, 2015. We studied 480 medical records in total. Two trained specialists reviewed the presence of AEs using 43 triggers, and a physician reviewed and validated the findings. The outcome measures included the number of AEs per 1000 patient days, AEs per 100 admissions, the percentage of entries with at least 1 AE and AE categorisation. Also, we carried out a descriptive analysis of the suspected factors of AEs, such as age, gender, length of stay, surgery. RESULTS: A total of 610 AEs were detected in the 480 medical records reviewed, corresponding to 127 injuries per 100 admissions. The number of AEs per 1000 patient days was 22.43. AEs occurred at least once in 329 (68.54%) patients. The rate of care harms ranked highest of all AEs, followed by the rate of medication harms and surgical harms. Patients with a more extended hospital stay or surgery was more likely to experience AEs. However, there was a negative correlation between age and the rate of AEs. CONCLUSION: The Global Trigger Tool was a useful method for detecting the characteristics of AEs in geriatric patients in a Chinese tertiary teaching hospital. To improve patients' safety, this tool should be incorporated into routine screening systems.

Unstructured brainstorming is not enough: structured brainstorming based on four verification and validation questions yields better hazard identification in healthcare.

OBJECTIVES: (1) To introduce the Methodical Hazard Identification Checklist (MHIC) for structured brainstorming and the four V&V categories on which it is based, and (2) to compare its efficacy with that of brainstorming (BS) in identifying hazards in healthcare. DESIGN: Comparative analysis of MHIC and team BS results. SETTING: Baruch Padeh Medical Center, Poriya, Israel. STUDY PARTICIPANTS: Quality engineering students, facilitators, validation teams and hospital staff who were familiar with the specific processes. INTERVENTION(S): The number of hazards identified by team BS were compared with those deduced by applying the four V&V hazard categories to each step (the MHIC) of 10 medical and 12 administrative processes. MAIN OUTCOME MEASURE(S): The total number of hazards (1) identified by BS, (2) identified by MHIC, (3) validated by the validation team and (4) hazards identified by both methods that the validation team deemed unreasonable. RESULTS: MHIC was significantly more successful than BS in identifying all hazards for the 22 processes (P < 0.0001). The estimated probabilities of success for BS for administrative and medical processes were 0.4444, 95%CI = [0.3506, 0.5424] and 0.3080, 95%CI = [0.2199, 0.4127], respectively. The estimated probabilities of success for MHIC for administrative and medical processes were 0.9885, 95%CI = [0.9638, 0.9964] and 0.9911, 95%CI = [0.9635, 0.9979], respectively. CONCLUSIONS: Compared to traditional BS, MHIC performs much better in identifying prospective hazards in the healthcare system. We applied MHIC methodology to administrative and medical processes and believe it can also be used in other industries that require hazard identification.

Classifying Safety Events Related to Diagnostic Imaging From a Safety Reporting System Using a Human Factors Framework.

PURPOSE: The aim of this study was to measure diagnostic imaging safety events reported to an electronic safety reporting system and assess steps at which they occurred within the diagnostic imaging workflow and contributing sociotechnical factors. METHODS: The authors evaluated all electronic safety reporting system reports related to diagnostic imaging during calendar year 2015 at an academic medical center with 50,000 admissions, 950,000 ambulatory visits, and 680,000 diagnostic imaging studies annually. Each report was assigned a harm score ranging from 0 to 4 by the reporter; scores of 2 (minor harm) to 4 (death) were classified as "potential harm." Two reviewers manually classified reports into steps involved in the diagnostic imaging chain and sociotechnical factors per the Systems Engineering Initiative for Patient Safety framework. The κ coefficient was used to measure interreviewer agreement on 10% of reports. The percentage of reports that could cause "potential harm" was compared for each step and sociotechnical factor using χ2 analysis. RESULTS: Of 11,570 safety reports submitted in 2015, 854 (7%) were related to diagnostic imaging. Although the most common step was imaging procedure (54% of reports), potential harm occurred more in result communication (odds ratio, 2.36; P = .05). Person factors most commonly contributed to safety reports (71%). Potential harm occurred more in safety reports that were related to tasks compared with person factors (odds ratio, 5.03; P < .0001). The κ coefficient was 0.79. CONCLUSIONS: Safety events were related to diagnostic imaging in 7% of reported events. Potential harm occurred primarily during imaging procedure and result communication. Safety events were attributed to multifactorial sociotechnical factors. Further work is necessary to decrease safety events related to diagnostic imaging.