Characterising the nature of clinical incidents reported across a tertiary health service: a retrospective audit

Patient safety incidents (PSIs), sometimes referred to as adverse events, incidents or patient safety events, are defined as an unplanned event or circumstance that may have resulted in unnecessary harm to a patient.^1,2 PSIs are classified according to different characteristics, including severity and degree of preventability.^2,3 Across studies of preventable PSIs, estimates of the overall median incidence of PSIs leading to patient harm range from 9.2% to 12%.^4–8 According to the Australian Commission on Safety and Quality in Health Care (ACSQHC), some of the most common harmful PSIs relate to surgical procedures (e.g. wrong patient or site, retention of foreign object), haemolytic blood transfusion reactions and medication errors (drug or fluid administration).⁹ PSIs are associated with considerable morbidity comparable to cervical cancer or multiple sclerosis.⁸

Harmful PSIs also represent a major financial burden to healthcare systems globally, with estimates of between 10% and 15% of healthcare expenditure resulting from health care-related harm.⁸ Consequently, the early detection and prevention of PSIs has become an international policy imperative.^10,11 Early detection and prevention requires safety learning systems, inclusive of incident reporting and management processes.^6,12 Effective safety learning systems should include reliable methods for reporting, detecting, monitoring and managing the investigation of PSIs.^13,14 The use of voluntary incident reporting systems (IRS) and processes provides data about the nature of safety problems, allowing analyses and learning from PSIs, and facilitates the development and implementation of commensurate improvement strategies to address the contributing factors.^14,15

However, it is important to focus on gaining a better understanding of what types of PSIs are being reported and the context in which they occur rather than merely focusing on raw numbers of events. Further, PSIs occur within complex sociotechnical and adaptive systems.^7,16 A PSI is not necessarily the result of one person making a mistake at the frontline of care; rather, it is a combination of system-related conditions that culminate in such incidents.^7,17 This is why reporting should be viewed as ‘a process of social and participative learning, rather than as a mechanism of data collection and analysis’.¹⁸

Understanding the severity, nature and characteristics of reported PSI in an IRS is therefore the first important step to building strong and mature safety learning systems.^6,12 Such understandings provide a foundation and frame of reference for all subsequent work in this area. The aim of this study was to gain an in-depth understanding of the severity, nature and characteristics of PSIs from an analysis of their frequency. In addition, we wanted to describe the association between clinical division and type of PSI. Nevertheless, these learnings are not necessarily specific to the particular organisation. Trends in PSIs and their contexts can inform others of potential threats that may exist in their own organisations.

A descriptive study was undertaken using a retrospective audit and analysis of PSIs drawn from the electronic clinical incident management system used by the health service. The 12-month audit period was from 1 July 2017 to 30 June 2018. This study is part of a larger program of research intended to establish an effective safety learning system in a health service in Queensland, Australia.

The study was conducted in a large tertiary health service in Queensland, Australia, in a catchment area of over 620 000 residents.¹⁹ The health service included three hospitals: a tertiary-level hospital, a regional health facility and a six-theatre day surgery hospital. In all, there were approximately 1250 beds across these three facilities.²⁰

Table 1 lists the eligibility criteria, developed before identifying PSIs that occurred during the provision of acute medical and surgical health care. Documented clinical incidents were screened and reviewed systematically by two researchers (IS and RB).

Table 1.  Eligibility criteria

The health service where this study was undertaken uses the Safety Assessment Code (SAC) classification system^3,21 to code clinical incidents (PSIs), which includes adverse and sentinel events. The coding system used by the health services is based on state health department definitions of SAC categories and specific incidents (e.g. falls, pressure injuries, clinical processes, medication, consent etc.). SAC codes are assigned only to incidents that are judged to be related to the care patients received and not their underlying condition. Thus, some incidents in the IRS do not have SAC codes and the SAC we analysed are a subset of the incidents reported. The SAC scoring system is based on the consequence of a PSI (i.e. severity or level of harm) and the likelihood of its reoccurrence.^3,12 Table 2 provides working definitions for each SAC category.

Table 2.  Safety Assessment Code (SAC) categories and definitions^11,22

Not all PSIs in the total sample originated from medical and surgical care units. As part of the analysis, we extracted a subsample of incidents based on wards and units within service streams according to whether they were from a medical or surgical location. For instance, PSIs occurring in the service stream ‘Cancer and Speciality Services’ were coded under ‘medical’, whereas PSIs occurring in ‘General Surgery and Gastroenterology Services’ were coded under ‘surgical’ care.

Across the health services, all PSIs are reported in accordance with relevant legislation, regulations and best practice guidelines. Following entry of the PSI into the IRS by the person who was involved or witnessed the incident, its management involves staff at different clinical and managerial levels. If the PSI is coded by the reporting staff member as an SAC incident (i.e. related to a process of care rather than related to a patient’s medical condition), the nurse unit manager and attending physician are required to review the incident within 24 h of its occurrence. Based on their initial assessment of the PSI, they may write additional notes in the IRS to supplement what the person involved has reported. Where death or permanent harm has occurred (SAC1 incident), the physician and treating team engage in the ‘clinician disclosure’ process and document the outcome in the patient’s digital record. Then, the SAC incident is reviewed by the patient safety coordinator and discussed at the executive triage meeting, where it is confirmed as an SAC1 (completed within 90 days). If confirmed, a decision is made on further analysis methodology (i.e. root cause analysis, human error/patient safety review, clinical/comprehensive review or external review) of investigations to be conducted according to health services policy.

This study was approval by the hospital (LNR/2019/QGC/46977) and Griffith University (NRS/08/19/HREC) human research ethics committees. Following approval, we sought permission from the Director-General, Queensland Health, as required by the Public Health Act (2005), to obtain deidentified clinical incident data.

Electronic data were exported from the health service’s clinical incident management system by the data custodian and given to the lead author as an encrypted Microsoft (Bellevue, WA, USA) Excel database. The Excel database included records of only PSIs coded as SAC1, SAC2 or SAC3 and reported between 1 July 2017 and 30 June 2018. Data extraction was undertaken based on date and incident report, using the following data fields: incident ID; incident date, time and date entered; ward/unit; clinical division (e.g. Surgical, Anaesthetic and Procedural Services, Emergency and Assessment Services); facility (hospital within health service); primary incident type; classification (e.g. skin injury, medication, falls, clinical process); details of incident; summary of incident; and confirmed level of harm (SAC category).

Data were cleaned and analysed using SPSS ver.25 (IBM, New York, NY, USA). Descriptive statistics using numbers and percentages were calculated relative to SAC category, PSI type and service or stream division. The total number of PSIs with an SAC category assigned to them over the 12-month audit period was used as the denominator. The denominator was based on the total number of eligible SAC (across categories 1–3), whereas the numerators were based on the absolute frequencies of SAC events according to type of PSI (e.g. clinical process, medication incident, skin injury, falls, clinical communication) and clinical division. Raw data (e.g. SAC category, PSI classification, clinical location) were transferred from SPSS to the program R (R Foundation for Statistical Computing, Vienna, Austria) to generate illustrative circular barplots, in which each bar represents a category (e.g. SAC classification, PSI type, clinical department) and the bars are displayed along a circle instead of a line that uses an x- and y-axis. The subsample of PSIs that occurred in medical and surgical units was subsequently analysed using Chi-squared analyses to assess the relationship between the location and type of PSI. Statistical significance was set at two-tailed P < 0.05.

Fig. 1 illustrates a flow diagram of the number of incidents during the extraction and exclusion process. In all, 5791 PSIs were identified in the database; of these, 4385 satisfied our eligibility criteria and were included in the analysis.

Fig. 1.  Flow diagram of incident case numbers that satisfied the eligibility criteria.

Fig. 2 illustrates the number of all types of PSI within each SAC category. Of the 4385 events across all SAC categories, 24 (0.5%) were classified as SAC1. Of the 24 SAC1 incidents, almost one-third (29.2%; 7/24) were clinical process incidents. Nearly one-fifth of SAC2 incidents (temporary harm) involved falls (20/107; 18.7%), whereas almost one-third (1238/4254; 29.1%) of SAC3 incidents (minimal harm or no harm) were related to skin injury (i.e. pressure injuries, skin tears). The most common causes of incidents, representing 80.2% of all PSIs, related to skin injury (n = 1255; 28.6%), medication (n = 1017; 23.2%), falls (n = 872; 19.9%) and clinical process (n = 373; 8.5%).

Fig. 2.  Number of types of PSIs across SAC categories, with the percentage of the PSI for each bar relative to the SAC category to which it belongs.

Fig. 3 shows absolute frequencies of all PSIs across clinical divisions. Almost one-third (7/24; 29.2%) of SAC1 PSIs occurred in Emergency and Assessment Services. Nearly one-quarter of SAC2 incidents occurred in Women’s Services (24/107; 22.4%), whereas almost one-third (1245/4254; 29.3%) of SAC3 incidents happened in Speciality and Ambulatory Services.

Fig. 3.  Number of PSIs across clinical divisions according to SAC categories, with the percentage of the PSI for each bar relative to the SAC category to which it belongs.

To conduct a subgroup analysis of 4142 of 4385 events (94.5%), data were dichotomised into two categories by allocating clinical departments to either medical or surgical divisions. PSIs outside acute care (e.g. physiotherapy, speech pathology) were excluded from the analysis (n = 243; 5.5%). PSI type was collapsed into five major categories: (1) clinical process; (2) falls; (3) medication; (4) skin injury; and (5) other (e.g. deterioration, behaviour, infection). The first four categories represented 80.2% of data.

Table 3 details the individual relationships between PSI types and clinical division. PSI incidents of falls and skin injury were significantly associated with clinical division. The proportion of falls was higher in the medical than surgical division. Conversely, the proportion of skin injury incidents was higher across the surgical than medical division.

Table 3.  Subanalysis of relationship between PSI type and medical and surgical clinical divisions
Unless indicated otherwise, data are given as n (%)

In this study, clinicians of a single tertiary healthcare service reported 4385 incidents assigned to an SAC category over the course of a calendar year. Our analysis showed that the proportion of incidents leading to severe harm or death (SAC1 incidents) was 0.5% and the proportion of incidents leading to moderate harm (SAC2) was 2.4%. In Australia, the national incidence of reported SAC1 incidents (i.e. severe harm or death) ranges from 0.35 to 12.3 per 10 000 hospital admissions.⁹ However, international data indicate that the actual incidence of iatrogenic harm is significantly higher than these estimates. In a recent scoping review of patient safety in 27 countries across six continents, Schwendimann et al.⁷ found a median of 10% of all hospitalised patients were affected by at least one adverse event, of which 7.3% (range 0.6–30%) were fatal.

Knowing the most common setting for specific incident types can help health systems create targeted quality improvement strategies. Our results suggest the relationship between setting and frequency (or number) of skin injury PSIs is statistically significant, proportionally 8% higher in surgical division compared with medical units. The higher proportions of skin injuries across the surgical division may be linked to longer surgeries and the increased use of medical devices causing device-related injuries. These injuries may also include hospital-acquired pressure injuries (HAPI). The 2019 international clinical practice guidelines on pressure injury (PI) prevention recommend risk assessment be completed as soon as possible after hospital admission or transfer.²² However, patients may develop PI in an hour or two after hospital admission, depending on their physical condition.²² Moreover, skin injury PSIs that result in Stage 3 and Stage 4 HAPI can attract financial penalties. As such, the accurate reporting of these injuries is paramount. Findings from a recent national report by the ACQSHC suggest a nearly twofold increase in the number of HAPIs reported in IRS over the past 5 years: in 2014–15 across Australian health services, 2831 HAPI were reported compared with 4369 HAPI reported during 2017–18.⁹ In practice, clinicians and patient safety professionals can use this finding to prioritise their improvement interventions accordingly.

The results of our analysis indicate that proportionally nearly double the number of falls occurred in medical than surgical division units (23% vs 13% respectively). The results of other studies support this finding.^23–25 The results of an earlier population based study found that a higher proportion of patients in geriatric and internal medicine units fell compared with patients in surgical units (33% combined vs 2% respectively).²⁵ The relatively lower number of falls among surgical patients may be due to a greater emphasis on bed rest with mobilisation only under nursing supervision, or these patients may simply have fewer falls risk factors. Notably, in our study three falls (0.5%) reported resulted in permanent injury (SAC1). Other studies report that up to 42% of in-patient falls result in injury, with around 8% of these resulting in permanent injury or death.²⁴ Most falls are related to either intrinsic factors (i.e. patient related, such as age, weight, prior fall and sex) or extrinsic (i.e. physical environment, medications, staffing ratios, delayed or missed care) factors.²⁶ Clearly, there is rarely a single cause for a fall,^24,25 although many of the falls reports we reviewed contained insufficient information relative to patient location and activities at the time of fall, and particular characteristics of the environment (e.g. lighting, noise, layout). Therefore, a more nuanced analysis was not possible. Although intrinsic and extrinsic factors cannot always be controlled, they can be managed and strategies implemented to mitigate patients’ falls risks.

Characterising the types of PSIs and describing their frequencies relative to their clinical division is helpful but provides only a superficial understanding of their aetiologies. PSIs are the product of complex adaptive systems in which the prevailing culture, the quality and timeliness of communication and the degree of teamwork can (and usually does) contribute.²⁷ Understanding error is important, but it is equally important to value how clinicians manage complex, dynamic situations throughout the day, constantly modifying their responses to get so much right.¹⁶ Therefore, identifying factors and conditions that reinforce success is integral to building these understandings. Reporting rates reflect the safety culture of an institution.^6,28 Transitioning from a blame culture that may incentivise people to cover up to an ethos of safety management underpinned by a just culture may improve reporting of PSI.¹³

Voluntary IRS are not intended to be an accurate picture of the incidence or severity of PSIs that occur across a health services district. Rather, they serve as a valuable resource to understand and act on latent and contributing factors of a representative sample of PSIs.²⁸ The ACQSHC recommends that healthcare services ensure their incident management and investigation systems provide adequate surveillance so that major safety lapses and risks are reliably detected, and that appropriate and timely corrective actions are implemented in response.⁹ This needs mature safety learning systems that enable and encourage incident reporting from all the healthcare settings in which patients present. In addition, the role of patients as health consumers in patient safety efforts has been recognised for over a decade. Patients and carers are important partners in health improvement and are able to provide valuable insights,²⁹ but their input needs to be valued.

The results of this study are subject to several well-recognised limitations of IRS, including selective disclosure of incidents resulting in underreporting of PSIs, variable clinician engagement and the estimated harm rates lacking reliability and validity. Our findings provide a solid rationale for developing robust safety learning systems, because PSIs occurred commonly in acute care settings and a small but significant minority are associated with moderate or severe patient harm. Another study limitation is that the mean number of incidents per admission/bed day/procedure/patient could not be calculated. Conceivably, Speciality and Ambulatory Services and Emergency and Assessment Services have higher workloads; therefore, the mean number of reported incidents may actually be lower than in other areas. We were unable to describe temporal trends in PSIs and thus are unable to speculate on their aetiologies. The IRS where the reported PSIs were drawn operates as a ‘standalone’ repository of incidents and is not linked to other health services databases. To enable identification of temporal trends, linkage of the data is essential.³⁰ Finally, analysis of PSIs focuses on ‘what went wrong’, whereas contemporary wisdom has advocated for also focusing on events when ‘things go right’.¹⁶

Our results suggest that preventable patient harm, particularly falls, skin injury and medication events, remain a serious problem across all health services contexts. The number of PSIs involving skin injury and falls appears to be associated with clinical division. Nonetheless, a deeper understanding of the nature and location of preventable patient harm may lead to more targeted quality improvement strategies with greater acceptability to clinicians and increase the likelihood of their normalisation. The lack of detailed information in most incident reports precludes in-depth analysis and the generation of more nuanced insights.

The authors declare no competing interests.