Application of a human factors classification framework for patient safety to identify precursor and contributing factors to adverse clinical incidents in hospital

Highlights

• The HFCF for patient safety identified patterns in the causation of clinical incidents.

• Human error played a leading causal role in clinical incidents, with error type varying by type of action.

• Targeted approaches to prevention of incidents, based on an understanding of how and why they occur, are needed.

Abstract

This study aimed to identify temporal precursor and associated contributing factors for adverse clinical incidents in a hospital setting using the Human Factors Classification Framework (HFCF) for patient safety. A random sample of 498 clinical incidents were reviewed. The framework identified key precursor events (PE), contributing factors (CF) and the prime causes of incidents. Descriptive statistics and correspondence analysis were used to examine incident characteristics. Staff action was the most common type of PE identified. Correspondence analysis for all PEs that involved staff action by error type showed that rule-based errors were strongly related to performing medical or monitoring tasks or the administration of medication. Skill-based errors were strongly related to misdiagnoses. Factors relating to the organisation (66.9%) or the patient (53.2%) were the most commonly identified CFs. The HFCF for patient safety was able to identify patterns of causation for the clinical incidents, highlighting the need for targeted preventive approaches, based on an understanding of how and why incidents occur.

Introduction

There are many diverse approaches that have been adopted to identify and investigate the causal and contributing factors surrounding adverse clinical events (i.e. clinical incidents) in health care settings. These have included human reliability techniques that focus on task analysis and examine the probability of the occurrence of clinical incidents (Lyons et al., 2004), critical incident techniques based on observations and interviews (Flanagan, 1954), root cause analyses (RCAs) that involve team-based retrospective analysis of what when wrong (Bagian et al., 2002), and variations on ‘human factors systems’ classification approaches to identify the causal factors involved in the incident, such as the Human Factors Analysis and Classification System (HFACS) (Shappell and Wiegmann, 2000), the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) (Chang et al., 2005), the System Error and Risk Analysis (SERA) (Hendy, 2007) and the Eindhoven (Van Vuuren and Van der Schaaf, 1997) classification frameworks. Several of these approaches were initially developed for use elsewhere, such as in aviation or industrial safety, and were later adapted for use in health care, such as HFACS (ElBardissi et al., 2007). None of the existing human factors classification taxonomies take into account the temporal sequence of events leading to the clinical incident. Being able to examine the causal sequence of events leading up to an clinical incident enables relationships between factors to be identified and can highlight common sequences of events that lead to the occurrence of an incident The consideration of a linear sequence of events, combined with the ability to identify contributing factors to any of the sequential events, could provide an in-depth causal description of clinical incidents. In addition, for existing taxonomies inter-rater reliability is not often assessed, or when assessed, is not high.

While the human factors taxonomies that have been used in health care are diverse in their structure, what is common across the literature is that human factors, particularly human error, plays a leading role in clinical incidents (Mitchell et al., 2014). Errors, defined as ‘the failure of a planned action to proceed as planned'(US Institute of Medicine, (2000)), have been retrospectively analysed in the health care setting, but studies differ in the way that medical errors are classified. Many analyses used job-related descriptions of the nature of errors. For example, in a study of errors in radiology, the error classification included ‘request for wrong patient’, ‘illegible request or duplicate request’(Martin, 2005) and, in a study of errors in otolaryngology, classification of errors included ‘errors in testing’ (e.g., specimen lost/test not done) and ‘technical errors’ (e.g., cranial or other nerves or endoscopic sinus surgery) (Shah et al., 2004). This type of approach is informative in providing direction in which task or job areas where errors are most likely to occur, but it is not descriptive in terms of the type of cognitive failure that explains why the particular error type occurred. Other analyses of medical errors have used cognitive failure classifications, such as errors of omission or commission (Tang et al., 2004). The obvious advantage of cognitive classifications of error is that they provide insight into the nature of error itself which is helpful in understanding why it occurred.

Despite the advantages of being able to examine human error and its role in the causation of clinical incidents, human error classifications were not included in the International Patient Safety Classification (IPSC) conceptual framework (World Health Organization, 2009). While the IPSC framework appears comprehensive as an informational model, its role as a classification taxonomy has been questioned (Schulz et al., 2009). In its current form, the IPSC does not provide unique identifiers nor definitions for each item, key ‘concepts’, ‘classes’ of items and hierarchical links and relationships between items in the IPSC are not clearly identified. Moreover not all of the classification categories in the IPSC are mutually exclusive or exhaustive, and the effectiveness of the framework to classify causal and contributing factors to clinical incidents is yet to be tested extensively in the field (Schulz et al., 2009, Feijter et al., 2012).

Within the public healthcare system in New South Wales (NSW) Australia, adverse clinical incidents involving patients that are categorised as serious, and are given a Severity Assessment Code (SAC) of 1, have an RCA investigation conducted by health care teams not involved in the occurrence of the incident. These SAC 1 events represent 0.04% of all clinical incident notifications in NSW (NSW Department of Health, (2011)) and include instances where a patient dies unrelated to the natural course of their illness or has an immediate differing outcome from what would otherwise be expected from patient management, or are nationally reportable sentinel events (NSW Health Department, 2007). While RCA investigations can be useful in identifying the causal factors of adverse clinical events, the experience of the team involved in the RCA, the quality of the analysis conducted, time pressures and resource availability can all have an effect on the quality of the investigation performed (Wu et al., 2008, Nicolini et al., 2011a, Nicolini et al., 2011b). Each RCA investigation attempts to reconstruct the circumstances of an adverse clinical incident by reconstructing the temporal sequence of events (Nicolini et al., 2011a, Nicolini et al., 2011b), yet, in NSW, when the analysis of the causal factors of the RCA are later reported by the RCA investigative team, the temporal sequence of events are often ignored. There is also a limited selection of possible factors recommended to an RCA investigative team in NSW for use in identifying the causal factors of a clinical incident (i.e. communication, knowledge, skills and competence, work environment/scheduling, patient factors, equipment, policies/procedures, safety mechanisms) (NSW Clinical Excellence Commission, 2015). In addition, the focus of RCA recommendations has often been on corrective actions to address local issues that could very well be useful to implement system-wide, if aggregated analysis of RCA findings and recommendations were conducted (Wu et al., 2008, Nicolini et al., 2011a, Nicolini et al., 2011b).

The Human Factors Classification Framework (HFCF) for patient safety was adapted from an existing framework that was developed to identify the role of human factors in work-related fatalities, in terms of the type and nature of human factors involvement in safety-related incidents and how they interact with other causes (Williamson and Feyer, 1990, Feyer and Williamson, 1991). This framework classifies the temporal sequence of events leading to the incident, includes the identification of human error using a cognitive classification structure, has high reliability (Feyer and Williamson, 1991) and has also been successfully adapted to classify child driveway fatalities (Williamson et al., 2002a, Williamson et al., 2002b), aircraft maintenance errors (Hobbs and Williamson, 2003) and swimming pool drowning events involving young children (Williamson et al., 2002a, Williamson et al., 2002b). The aim of this research is to identify temporal precursor and associated contributing factors to clinical incidents in a hospital setting using the HFCF for patient safety.