Rule-Based Natural Language Processing Pipeline to Detect Medication-Related Named Entities: Insights for Transfer Learning

We document the procedure and performance of a rule-based NLP system that, using transfer learning, automatically extracts essential named entities related to drug errors from Japanese free-text incident reports. Subsequently, we used the rule-based annotated data to fine-tune a pre-trained BERT model and examined the performance of medication-related incident report prediction. The rule-based pipeline achieved a macro-F1-score of 0.81 in an internal dataset and the BERT model fine-tuned with rule-annotated data achieved a macro-F1-score of 0.97 and 0.75 for named entity recognition and relation extraction tasks, respectively. The model can be deployed to other, similar problems in medication-related clinical texts.

Assessing the Impact of COVID-19 on Incident Reporting and Elucidating the Characteristics of Incident Reports for COVID-19 Patients at the Critical Care Center of a Tertiary-Care Teaching Hospital in Japan

The global coronavirus disease 2019 (COVID-19) pandemic has necessitated the establishment of new medical care systems worldwide. Medical staff treating COVID-19 patients perform their care duties in highly challenging and psychologically demanding situations, raising concerns about their impact on patient safety. Therefore, this study aimed to investigate and characterize incident reports related to COVID-19 patients to clarify the impact of COVID-19 on patient safety. The study included data from 557 patients admitted to the Critical Care Center of a tertiary-care teaching hospital in Osaka, Japan, from April 2020 to March 2021. The patients were divided into two groups: COVID-19 (n = 106) and non-COVID-19 (n = 451) and compared based on various characteristics, incident reporting rates, and the content of incident reports. The findings indicated a significantly higher rate of patients with incident reports in the COVID-19 group compared to the non-COVID-19 group (49.1% vs. 24.4%, P < 0.001). In addition, quantitative text analysis revealed that the topic ratio, consisting of "respiration," "circuit," "settings," "connection," "nursing," "ventilator," "control," "tape," "Oxylog®," and "artificial nose" was significantly higher in the incident reports of the COVID-19 group (P = 0.003). In conclusion, COVID-19 patients are more susceptible to adverse incidents and may face a higher risk of patient safety issues. The characteristic topics in incident reports involving COVID-19 patients primarily revolved around ventilator-related issues. In the future, the methodology used in the current study may be utilized to identify incident characteristics and implement appropriate countermeasures in the event of unknown patient safety issues.

The Nature, Causes, and Clinical Impact of Errors in the Clinical Laboratory Testing Process Leading to Diagnostic Error: A Voluntary Incident Report Analysis

OBJECTIVES

Diagnostic errors, that is, missed, delayed, or wrong diagnoses, are a common type of medical errors and preventable iatrogenic harm. Errors in the laboratory testing process can lead to diagnostic errors. This retrospective analysis of voluntary incident reports aimed to investigate the nature, causes, and clinical impact of errors, including diagnostic errors, in the clinical laboratory testing process.

METHODS

We used a sample of 600 voluntary incident reports concerning diagnostic testing selected from all incident reports filed at the University Medical Center Utrecht in 2017-2018. From these incident reports, we included all reports concerning the clinical laboratory testing process. For these incidents, we determined the following: nature: in which phase of the testing process the error occurred; cause: human, technical, organizational; and clinical impact: the type and severity of the harm to the patient, including diagnostic error.

RESULTS

Three hundred twenty-seven reports were included in the analysis. In 77.1%, the error occurred in the preanalytical phase, 13.5% in the analytical phase and 8.0% in the postanalytical phase (1.5% undetermined). Human factors were the most frequent cause (58.7%). Severe clinical impact occurred relatively more often in the analytical and postanalytical phase, 32% and 28%, respectively, compared with the preanalytical phase (40%). In 195 cases (60%), there was a potential diagnostic error as consequence, mainly a potential delay in the diagnostic process (50.5%).

CONCLUSIONS

Errors in the laboratory testing process often lead to potential diagnostic errors. Although prone to incomplete information on causes and clinical impact, voluntary incident reports are a valuable source for research on diagnostic error related to errors in the clinical laboratory testing process.

Underreporting of work-related low back pain among registered nurses: A mixed method study

BACKGROUND

Identifying and addressing work-related health problems early is crucial, but workers often perceive barriers in reporting these to management. This study aimed to investigate the factors associated with nurses' reporting of work-related low back pain to their managers and explored the reasons why nurses with patient handling injuries did not report them.

METHODS

This study is a concurrent mixed-method analysis of data from two statewide cross-sectional surveys of California registered nurses conducted in 2013 and 2016. The reporting of work-related low back pain to management (n = 288) was examined for associations with individual, occupational, and organizational factors. For qualitative analysis, the reasons for not reporting patient handling injuries were explored using open-ended responses (n = 42).

RESULTS

Reporting was associated with BIPOC (Black, Indigenous, and People of Color) men (adjusted odds ratio [AOR]: 1.31, 95% confidence interval [CI]: 1.07-1.59) compared to non-Hispanic White women; being a non-US educated nurse (AOR: 0.90, 95% CI: 0.80-1.01); experiencing greater low back pain (AOR: 1.07, 95% CI: 1.02-1.12); missing work (AOR: 1.38, 95% CI: 1.21-2.62); perceiving high physical workload (AOR: 0.89, 95% CI: 0.81-0.98); perceiving high people-oriented culture (AOR: 1.14, 95% CI: 1.04-1.25); and perceiving high ergonomic practices (AOR: 0.89, 95% CI: 0.81-0.98). Identified themes on the reasons for not reporting injuries included organizational-culture attitudes toward work-related injuries and injury characteristics of musculoskeletal disorders.

CONCLUSIONS

The findings indicate a need for management to remove structural barriers and improve organizational practices, and for a culture that promotes trust and open communication between workers and management.

Nurses' Medication Administration Workarounds when Using Electronic Systems: An Analysis of Safety Incident Reports

Electronic medication management systems (EMMS) have been implemented in most acute care settings in Australia to reduce medication error rates. One of the key challenges related to the introduction of EMMS in hospitals is the uptake of informal "workarounds" by clinicians, including nurses. In this study, we aimed to examine one workaround in depth, nurses not documenting medication administration in the EMMS at the time of administration. We conducted a review of incident reports to identify the factors that contribute to this workaround occurring and the consequences or potential consequences of this workaround on patients. We identified a range of contributing factors, with factors relating to the user (e.g. nurses being time poor) occurring most frequently in incident reports. The most frequently seen consequence of this workaround was the patient receiving an additional dose. This research revealed that strategies to reduce the uptake of this workaround should consider user and organisational factors rather than just EMMS design alone.

Medication incidents and medication errors in Danish healthcare: A descriptive study based on medication incident reports from the Danish Patient Safety Database, 2014-2018

In Denmark, reporting of safety incidents to the nationwide Danish Patient Safety Database (DPSD) is mandatory. Medication incident reports constitute the largest category of safety reports. We aimed to provide numbers and characteristics of medication incidents and MEs reported to DPSD focusing on medication, their severity and the trends therein. This is a cross-sectional study of medication incident reports for individuals ≥18, submitted to DPSD in 2014-2018. We performed analyses on the (1) medication incident and (2) ME levels. Out of 479 814 incident reports, 61.18% (n = 293 536) were related to individuals ≥70 and 44.6% (n = 213 974) to nursing homes. Most of the events were harmless (70.87%, n = 340 047) and 0.8% (n = 3859) had caused severe harm or death. ME-analysis (n = 444 555) revealed that paracetamol and furosemide were the most frequently reported drugs. The most common drugs for severe and fatal MEs were warfarin, methotrexate, potassium chloride, paracetamol and morphine. When the reporting ratio for all MEs and harmful MEs was considered, other drugs than the most frequently reported ones were found to be associated with harm. We found a large proportion of harmless medication incident reports and reports from community healthcare services and identified high-risk medicines associated with harm.

Bus passenger injury prevention: Learning from onboard incidents

OBJECTIVE

Bus travel is relatively safe: however there remains a lack of understanding of passenger injury incidents onboard buses. The objective of this study was to understand more about onboard passenger incidents to help inform injury mitigation.

METHODS

The UK national STATS19 data and Transport for London bus incident data (IRIS) were used to determine the size of the problem in Greater London. Other data including onboard incident reports from two bus operators and CCTV footage of 70 incidents were used to understand passenger injury in more depth and identify common themes and challenges.

RESULTS

The STATS19 and IRIS analysis showed that there was a difference between nationally reported bus incidents compared to locally reported bus incidents. Non-collision incidents are prevalent in the data suggesting there is a large problem to tackle. The CCTV and bus incident data identified braking to be the single largest problem in onboard bus passenger injury incidents. Inconsistent reporting of passenger incidents and injury descriptions make it difficult to identify injury patterns and trends. Areas on the bus appear to contribute to higher injury incidents namely those seats facing and closest to the wheelchair area. Other challenges relating to expected passenger and driver behaviors were noted where blame for the incident and outcome can be attributed to both parties.

CONCLUSIONS

This combined analysis of incident reports and CCTV footage has enabled a better understanding of the events leading to on-board passenger injury incidents. Preventing harsh braking would appear to be the most effective way of reducing passenger injuries. Additionally improved data collection would assist both transport authorities and bus operators to identify and monitor the effect of bus safety improvements.

Management and reporting of safety incidents by residential care facilities in Ireland: A thematic analysis of statutory notifications

The prevention of safety incidents (SI) in health and social care settings is an ongoing undertaking. Limited research has been conducted on SIs outside of acute care. Internationally residential care facilities (RCFs) are typically regulated to promote quality and safeguarding. A part of this regulation is the statutory responsibility of RCFs to notify the regulator about SIs. Notifications include details surrounding SIs and are used to inform the regulatory monitoring approach. The recent development of the Database of Statutory Notifications from Social Care in Ireland facilitates in-depth analysis of notifications which can be used to inform the management of SIs and thus, improve quality and safety. The aim of this study was to analyse narratives provided in statutory notifications for older persons and people with disability, in order to identify current management of SIs, system vulnerabilities and reporting practices. A Qualitative Descriptive approach was taken. A random sample of notifications received in 2018 was drawn and stratified by service-type and notification-type. Data extraction was conducted against priori agreed target areas of management, system vulnerabilities and reporting practices. Inductive thematic analysis was used identifying two parent themes: 'chronology' and 'regulatory input'. 'Chronology' subthemes included 'pre-event', 'immediate response' and 'continued response'. Measures that are resident focused and follow policies and protocols in RCFs to prevent or mitigate the seriousness of SIs were evident in the immediate response and continued response. The actions taken in the immediate and continued response in turn became part of the pre-event of future SIs. Under 'regulatory input' subthemes included 'inaccurate reporting', 'lines of inquiry', 'requests for further information', 'identification of repetitive patterns' and 'satisfactory conclusion'. In conclusion, RCFs manage SIs with short and longer term actions focused on resident wellbeing. These actions in turn become part of the pre-event of future SIs. Regulatory input highlighted regulatory burden.

Frequency and characteristics of medical emergencies in an Australian dental school: A retrospective study

PURPOSE/OBJECTIVES

The frequency of medical emergencies in a dental setting is relatively low. However, most dental treatment occurs outside of a hospital setting; thus the time to respond to a life-threatening situation is crucial. The aim of this study was to determine the frequency and characteristics of medical emergencies that occurred at the Griffith University Dental Clinic over a 6-year period.

METHODS

Data involving medical emergencies was collected at a dental school between January 2014 and December 2019. Data was obtained from the risk incident reporting system (GSafe), and descriptive statistics were analyzed using IBM SPSS.

RESULTS

The frequency of medical emergencies in a dental setting over the 6-year retrospective period based on the number of dental services provided was 0.037% (n = 108). The three most common medical emergencies were syncope (25.0%), hypoglycemia (16.7%), and foreign body ingestion (13.9%). These happened more often during dental extractions (26.9%), followed by local anesthesia (16.7%) and restorative procedures (13.0%). A larger portion of these incidents occurred during dental treatment (62.0%) as opposed to before (12.0%) or after (26.0%). Most medical emergencies happened within the dental student clinic (72.2%) followed by the dental waiting room (19.5%) and dental private clinic (8.3%).

CONCLUSION

Overall, the number of medical emergencies at the Griffith University Dental Clinic was low. The most common medical emergencies were syncope, hypoglycemia, and foreign body ingestion. Dental education in preventative strategies and training in basic life support is necessary to ensure dental practitioners can manage the acute deterioration of a patient promptly.

Content Analysis of Patient Safety Incident Reports for Older Adult Patient Transfers, Handovers, and Discharges: Do They Serve Organizations, Staff, or Patients?

OBJECTIVE

The aim of the study was to analyze content of incident reports during patient transitions in the context of care of older people, cardiology, orthopedics, and stroke.

METHODS

A structured search strategy identified incident reports involving patient transitions (March 2014-August 2014, January 2015-June 2015) within 2 National Health Service Trusts (in upper and lower quartiles of incident reports/100 admissions) in care of older people, cardiology, orthopedics, and stroke. Content analysis identified the following: incident classifications; active failures; latent conditions; patient/relative involvement; and evidence of individual or organizational learning. Reported harm was interpreted with reference to National Reporting and Learning System criteria.

RESULTS

A total 278 incident reports were analyzed. Fourteen incident classifications were identified, with pressure ulcers the modal category (n = 101,36%), followed by falls (n = 32, 12%), medication (n = 31, 11%), and documentation (n = 29, 10%). Half (n = 139, 50%) of incident reports related to interunit/department/team transfers. Latent conditions were explicit in 33 (12%) reports; most frequently, these related to inadequate resources/staff and concomitant time pressures (n = 13). Patient/family involvement was explicit in 61 (22%) reports. Patient well-being was explicit in 24 (9%) reports. Individual and organizational learning was evident in 3% and 7% of reports, respectively. Reported harm was significantly lower than coder-interpreted harm (P < 0.0001).

CONCLUSIONS

Incident report quality was suboptimal for individual and organizational learning. Underreporting level of harm suggests reporter bias, which requires reducing as much as practicable. System-level interventions are warranted to encourage use of staff reflective skills, emphasizing joint ownership of incidents. Co-producing incident reports with other clinicians involved in the transition and patients/relatives could optimize organizational learning.

Impact of the initial response to COVID-19 on long-term care for people with intellectual disability: an interrupted time series analysis of incident reports

BACKGROUND

The lockdown-measures in response to COVID-19 taken by long-term care organisations might have impacted problem behaviour and behavioural functioning of people with intellectual disability. This study tested changes in reported incidents, in particular regarding aggression, unexplained absence and, for contrast, medication errors.

METHODS

Metadata on weekly incident and near-incident reports from 2016 to June 2020 involving over 14 000 clients with mild to serious intellectual disability of 's Heeren Loo, a long-term care organisation for people with intellectual disability, were subjected to interrupted time series analysis, comparing the COVID-19 with the pre-COVID-19 period.

RESULTS

The imposition of lockdown-measures coincided with a significant drop in incidents (total, P < .001; aggression, P = .008; unexplained absences, P = .008; and medication errors, P < .001). Incidents in total (P = .001) and with aggression (P < .001) then climbed from this initial low level, while medication errors remained stably low (P = .94).

CONCLUSION

The rise in incidents involving aggression, against the background of generally lowered reporting, underlines the need for pandemic control measures that are suitable for people with intellectual disability in long-term care.

Evaluation of Medication Error Incident Reports at a Tertiary Care Hospital

Background: Medications errors (MEs) have been a major concern of healthcare systems worldwide. Voluntary-based incident reporting is the mainstay system to detect such events in many institutions. However, the number of reports can be highly variable across institutions depending on their adoption of the safety culture. This study aimed to evaluate and analyze medication error incidents that were submitted through the hospital safety reporting system in 2015 at a tertiary care center in the western region of Saudi Arabia, and to explore the most common types of harmful MEs in addition to the risk factors that led to such harmful incidents. Methods: This is a descriptive study that was conducted utilizing 624 medication error reports extracted from the hospital safety reporting system. Reports were analyzed based on the medication name, event type, event description, nodes of the medication use process, harm score (adapted from the National Coordinating Council for Medication Error Reporting and Prevention harm index), patients' age/gender, incident setting, and time of occurrence as documented in the Safety Reporting System (SRS). Furthermore, all errors that resulted in injury or harm to patients had a deeper review by two senior pharmacists to find contributing factors that led to these harmful incidents and recommend system-based preventive strategies. Results: This study showed that most reported incidents were near misses (69.3%). The pediatric population was involved in 28.4% of the incident reports. Most of the reported incidents were categorized as occurring in the inpatient setting (57.4%). Medication error incidents were more likely to be reported in the morning shift versus evening and night shift (77.4% vs. 22.6%). Most reported incidents involved the dispensing stage (36.7%). High-alert medications were reported in 281 out of 624 events (45%). Conclusions: The hospital medication safety reporting program is a great tool to identify system-based issues in the medication management system. This study identified many opportunities for improvement in the medication use system, especially in management of chemotherapy and anticoagulant agents.

Wrong-patient incidents during medication administrations

AIMS AND OBJECTIVES

To describe the factors pertaining to medication being administered to the wrong patient and to describe how patient identification is mentioned in wrong-patient incident reports.

BACKGROUND

Although patient identification has been given high priority to improve patient safety, patient misidentifications occur, and wrong-patient incidents are common.

DESIGN

A descriptive content analysis.

METHODS

Incident reports related to medication administration (n = 1,012) were collected from two hospitals in Finland between 1 January 2013-31 December 2014. Of those, only incidents involving wrong-patient medication administration (n = 103) were included in this study.

RESULTS

Wrong-patient incidents occurred due for many reasons, including nurse-related factors (such as tiredness, a lack of skills or negligence) but also system-related factors (such as rushing or heavy workloads). In 77% (n = 79) of wrong-patient incident reports, the process of identifying of the patient was not described at all.

CONCLUSIONS

There is need to pay more attention to and increase training in correct identification processes to prevent wrong-patient incidents, and it is important to adjust system factors to support nurses.

RELEVANCE TO CLINICAL PRACTICE

Active patient identification procedures, double-checking and verification at each stage of the medication process should be implemented. More attention should also be paid to organisational factors, such as division of work, rushing and workload, as well as to correct communication. The active participation of nurses in handling incidents could increase risk awareness and facilitate useful protection actions.

Do failures in non-technical skills contribute to fatal medical accidents in Japan? A review of the 2010-2013 national accident reports

OBJECTIVES

We sought to clarify how large a proportion of fatal medical accidents can be considered to be caused by poor non-technical skills, and to support development of a policy to reduce number of such accidents by making recommendations about possible training requirements.

DESIGN

Summaries of reports of fatal medical accidents, published by the Japan Medical Safety Research Organization, were reviewed individually. Three experienced clinicians and one patient safety expert conducted the reviews to determine the cause of death. Views of the patient safety expert were given additional weight in the overall determination.

SETTING

A total of 73 summary reports of fatal medical accidents were reviewed. These reports had been submitted by healthcare organisations across Japan to the Japan Medical Safety Research Organization between April 2010 and March 2013.

PRIMARY AND SECONDARY OUTCOME MEASURES

The cause of death in fatal medical accidents, categorised into technical skills, non-technical skills and inevitable progress of disease were evaluated. Non-technical skills were further subdivided into situation awareness, decision making, communication, team working, leadership, managing stress and coping with fatigue.

RESULTS

Overall, the cause of death was identified as non-technical skills in 34 cases (46.6%), disease progression in 33 cases (45.2%) and technical skills in two cases (5.5%). In two cases, no consensual determination could be achieved. Further categorisation of cases of non-technical skills were identified as 14 cases (41.2%) of problems with situation awareness, eight (23.5%) with team working and three (8.8%) with decision making. These three subcategories, or combinations of them, were identified as the cause of death in 33 cases (97.1%).

CONCLUSIONS

Poor non-technical skills were considered to be a significant cause of adverse events in nearly half of the fatal medical accidents examined. Improving non-technical skills may be effective for reducing accidents, and training in particular subcategories of non-technical skills may be especially relevant.

Outpatient falls prevention program outcome: an increase, a plateau, and a decrease in incident reports

OBJECTIVE

We implemented an outpatient falls guideline in 2008 in the department of radiology. Here, we describe our multiyear experience.

MATERIALS AND METHODS

This was a retrospective study conducted between April 2006 and September 2013 to investigate outpatient falls. The span of the study was divided into eight periods. The incident reporting system was searched for the falls and the fall-related variables.

RESULTS

A total of 327 falls occurred during 5,080,512 radiology examinations (rate, 0.64/10,000 total examinations). The highest rate was in period 6 (0.83/10,000 examinations). The average for periods 1 and 2 is 0.39/10,000 examinations (37 falls/945,427 examinations), and the average for periods 3-6 is 0.77/10,000 examinations (204 falls/2,656,805 examinations). The average rate for periods 7 and 8 is 0.58/10,000 examinations (86 falls/1,478,280 examinations). There was a statistically significant increase in the total number of falls reported between period 2 and period 3 (p = 0.02). There was a statistically significant decrease in outpatient falls between period 6 and period 7 (p = 0.01). The number of falls among patients 60 years old or older was 177 falls/2,180,093 examinations (rate, 0.81/10,000 examinations), and that among patients younger than 60 years was 150 falls/2,900,419 examinations (rate, 0.52/10,000 examinations), with a statistically significant difference (p = 0.007). Although the rate of falls was higher among female patients, there was no statistically significant difference between the sexes (p = 0.18).

CONCLUSION

The outcome of the outpatient falls guideline was characterized by an increase, a plateau, and a decrease in incident reports. The initial increase may be due to the Hawthorne effect. The plateau may represent the value closest to the true incidence. The decrease may represent the effect of the program.

Police reporting practices for sexual assault cases in which "the victim does not wish to pursue charges"

Prior research examining sexual assault case attrition has focused on the processing of cases across the justice system. Studies have examined arrest decision making and prosecutorial decision making in an attempt to better understand where and when cases drop out of the system. Less explored are police reporting practices during the initial stage of processing for cases in which the officer stated that the victim chose to drop her case. We addressed this gap in the literature by reviewing law enforcement incident reports at their onset, specifically; we examine reports of cases in which the officer reported the victim chose to drop the case. Results indicated that of the 125 cases of sexual assault reported to the police, 41 reports (32.8%) stated that the victim decided to no longer pursue charges. However, few police reports (30.2%) provided a clear rationale for why the victim decided to no longer pursue charges. Results of this study call for more standardized police reporting practices and point to the need for future research into the initial stage of law enforcement involvement in adult sexual assault cases.

Detecting inpatient falls by using natural language processing of electronic medical records

BACKGROUND

Incident reporting is the most common method for detecting adverse events in a hospital. However, under-reporting or non-reporting and delay in submission of reports are problems that prevent early detection of serious adverse events. The aim of this study was to determine whether it is possible to promptly detect serious injuries after inpatient falls by using a natural language processing method and to determine which data source is the most suitable for this purpose.

METHODS

We tried to detect adverse events from narrative text data of electronic medical records by using a natural language processing method. We made syntactic category decision rules to detect inpatient falls from text data in electronic medical records. We compared how often the true fall events were recorded in various sources of data including progress notes, discharge summaries, image order entries and incident reports. We applied the rules to these data sources and compared F-measures to detect falls between these data sources with reference to the results of a manual chart review. The lag time between event occurrence and data submission and the degree of injury were compared.

RESULTS

We made 170 syntactic rules to detect inpatient falls by using a natural language processing method. Information on true fall events was most frequently recorded in progress notes (100%), incident reports (65.0%) and image order entries (12.5%). However, F-measure to detect falls using the rules was poor when using progress notes (0.12) and discharge summaries (0.24) compared with that when using incident reports (1.00) and image order entries (0.91). Since the results suggested that incident reports and image order entries were possible data sources for prompt detection of serious falls, we focused on a comparison of falls found by incident reports and image order entries. Injury caused by falls found by image order entries was significantly more severe than falls detected by incident reports (p<0.001), and the lag time between falls and submission of data to the hospital information system was significantly shorter in image order entries than in incident reports (p<0.001).

CONCLUSIONS

By using natural language processing of text data from image order entries, we could detect injurious falls within a shorter time than that by using incident reports. Concomitant use of this method might improve the shortcomings of an incident reporting system such as under-reporting or non-reporting and delayed submission of data on incidents.