Impact of adverse events on patient outcomes in a Japanese intensive care unit: a retrospective observational study

Abstract Aim We investigated adverse events (AEs) in a Japanese intensive care unit (ICU) and evaluated the impact of cause‐specific AEs on mortality and length of stay. Design A retrospective observational study in the ICU of an academic hospital. Methods We reviewed medical records with the Global Trigger Tool. Results Of the 246 patients, 126 (51%) experienced one or more AEs with an incidence of 201 per 1000 patient‐days and 115 per 100 admissions. A total of 294 AEs were detected with 119 (42%) adverse drug events, 67 (24%) procedural complications, 63 (22%) surgical complications, 26 (9%) nosocomial infections, 5 (2%) therapeutic errors and 4 (1%) diagnostic errors. Adverse event (AE) presence was associated with length of ICU stay (β = 2.85, 95% confidence interval [CI]: 1.09–4.61). Adverse drug events, procedural complications and nosocomial infections were strongly associated with length of ICU stay (β = 2.38, 95% CI: 0.77–3.98; β = 3.75, 95% CI: 2.03–5.48; β = 6.52, 95% CI: 4.07–8.97 respectively).


| INTRODUC TI ON
Adverse events (AEs) are defined as "unintended physical injury resulting from or contributed to by medical care that requires additional monitoring, treatment or hospitalization or that results in death (Griffin & Resar, 2009)." Diverse studies from various countries reported that AEs developed in 12% of hospitalized patients (Panagioti et al., 2019) while the severe and unstable patients often seen in the intensive care unit (ICU) experienced more AEs than those in other general wards (Andrews et al., 1997). Up to 20%-25% of ICU patients experience an adverse event (AE), with 45.3-80.5 events per 1000 patient-days, and, within these events, 13% were lethal or life-threatening (Rothschild et al., 2005;Sauro et al., 2020).
Numerous international studies show that AEs increase ICU stay length by 8.9 days and the length of a hospital stay by 6.8 days (Ahmed et al., 2015).
The incidence of AEs varies according to national background and medical culture. For example, anaesthesia-related mortality is higher in developing countries than in developed countries (Bainbridge et al., 2012). Also, patient safety culture scores have been found to be negatively correlated with AEs incidence (Han et al., 2020;Mardon et al., 2010;Najjar et al., 2015), and Japan was reported to have a lower score for patient safety culture than the United States (Fujita et al., 2013). In Japan, medical policy promotes an error-based incident reporting system, but, on the other hand, few studies have focussed on AEs that are important for patients. Additionally, most of these studies are limited to drug-related AEs and do not report AEs within targeted ICU populations (Anzai et al., 2019;Chisaki et al., 2017;Fujiwara et al., 2016;Hatahira et al., 2018;Matsumura et al., 2018;Suga et al., 2019;Tsuchiya et al., 2020).
Furthermore, AEs may be subclassified by cause, such as adverse drug events, surgical complications, procedural complications, nosocomial infections and error (Forster et al., 2008). However, many studies have integrated and analysed AEs without comparing cause which obscures cause-specific AE impact on patient outcomes (Ahmed et al., 2015).
Therefore, the purpose of the present study was to investigate AEs in a Japanese general ICU and evaluate the impact of causespecific AEs on patient outcomes.

| Study design
We retrospectively reviewed electronic medical records with the Global Trigger Tool (GTT) developed by the Institute for Healthcare Improvement to detect AEs. Trigger tool methodology is a retrospective review of a random sample of medical records with triggers to identify possible AEs. Examples of triggers include acute dialysis, pneumonia onset or intubation. GTT consists of 53 triggers defined in six different modules (cares, medication, surgical, intensive care, perinatal and emergency department). When a trigger was found, a careful analysis was conducted to confirm whether an AE was related to the trigger event. The following definitions of AEs were used: unintended physical injury resulting from or contributed to by medical care that requires additional monitoring, treatment or hospitalization or that results in death. Psychological harm (by definition) is excluded as an AE in the GTT, which focusses on AEs related to the active delivery of care (commission) and excludes issues related to substandard care (omission) (Griffin & Resar, 2009).

| Setting
The study was conducted in the ICU of an 800-bed academic hospital in Japan. The unit itself is a general ICU with twelve beds and approximately 700-800 admitted patients per year. The system is an open ICU with an ICU nursing staff-to-patient ratio of 1:2.

| Record selection
For each month, 20 medical admission records were randomly selected between April 2016 and March 2017 in ICU using "RANDBETWEEN" as a randomization function (Griffin & Resar, 2009). Exclusion criteria were (a) under the age of 18 years, (b) a length of stay less than 24 hr or (c) readmission. Short stays were excluded because of a lack of information by which to determine AEs.

| Data collection
Patient data were retrospectively collected from electronic medical records. Demographic data collected include age, gender, the disease for ICU admission, mechanical ventilation status, admission category (medical, elective surgery, emergency surgery), location prior to ICU and Acute Physiology and Chronic Health Evaluation II (APACHE II) score (scale range, 0-71) as a marker of illness severity. The APACHE II score is calculated by the most abnormal values obtained during the first 24 hr of an ICU stay (Knaus et al., 1985).
Outcome data collected include 28-day mortality, hospital mortality, length of ICU stay and length of hospital stay.

| Review process
The review team consisted of two registered intensive care nurses (primary reviewers) and one intensivist (secondary reviewer). All reviewers had more than 5 years of ICU experience and general knowledge about the ICU. The review process was performed in two-stages following GTT guidelines (Griffin & Resar, 2009). In stage 1, primary reviewers used GTT to independently conduct reviews of individual electronic medical records within 20 min or less.
All reviews were conducted manually by looking through text fields that included the physician's diagnosis, treatment records, surgical records and nursing care records from ICU admission until 2 days after ICU discharge to clarify AEs occurring in the ICU. The primary reviewer screened for one or more of the 53 triggers then marked any such triggers on the GTT worksheet and described suspected AEs with a one-to two-paragraph summary. Primary reviewers then compared findings and came to a consensus.
In stage 2, the secondary reviewer did not review patient medical records directly but only performed a review of the primary reviewers' summaries. Any suspected AE for which the secondary reviewer disagreed was discussed, and a final consensus was reached on the presence, severity and preventability of suspected AEs.
Since review training was not mentioned in the Japanese GTT, the GTT manual was distributed to the reviewers and, after one training session, we started the review.

| Preventability and severity of adverse events
Preventability of AEs was assessed as "an error in management due to failure to follow accepted practice at an individual or system level (Rodziewicz et al., 2021)." The degree of preventability was scored on a modified three-grade scale: grade 1 was "virtually no evidence for preventability," grade 2 was "preventability not likely, less than 50%," and grade 3 was "preventability more likely than not, more than 50%." AEs were judged to be preventable if scored as grade 3 (Schwendimann et al., 2018

| Statistical analysis
Quantitative variables are reported as median (interquartile range, IQR) and qualitative variables as number (%). The incidence of the AE rate was calculated as the number of AEs per 1000 patient-days and per 100 admissions.
Characteristics and outcomes were compared using the Mann-Whitney U test for quantitative variables and Pearson's chi-squared test or Fisher's exact test for qualitative variables. Associations between the presence of the AE and patient outcome (28-day mortality, hospital mortality, length of ICU stay and length of hospital stay) were examined using linear regression analysis or logistic regression analysis, and results are reported as coefficients with 95% confidence intervals. These multivariate analysis models included the following covariates most likely to affect patient outcomes: presence of the AE, APACHE II score and mechanical ventilation required. The p values <.05 indicated statistical significance. All analyses were conducted with IBM SPSS Statistics version 25 (IBM Corp.).

| Ethics approval
The institutional review board of our hospital approved this study (H30-126). The need for informed consent from each patient was waived due to the retrospective design and use of anonymized patient data.

| Characteristics and patient outcomes
Out of 772 patients admitted to the ICU within the study period, 257 patients were excluded while the remaining 515 eligible patients were randomly sampled at 20 per month. Finally, 246 patients were selected (Figure 1).
Patient characteristics and outcomes are given in Table 1. The median age was 68 (55-75) years, 157 (64%) patients were male, 128 (52%) patients required mechanical ventilation, and the median APACHE II score was 13 (9-20). Patients were divided into two groups based on if they had one or more AEs. Significant differences were found in mechanical ventilation status, APACHE II score, neuromuscular, trauma, 28-day mortality, hospital mortality, length of ICU stay and length of hospital stay.

| Incidence and category of adverse events
A total of 246 patients' electronical medical records were reviewed. We identified 284 AEs and judged that 56 (20%) AEs were preventable. The changes over time per month in the incidence of AEs and disease severity are shown in Appendix 1. Of the total, 126 (51%) patients experienced one or more AEs and incidence was 201 events per 1000 patient-days and 115 events per 100 F I G U R E 1 Flow chart of the study population admissions. According to severity, 185 (65%) events were category E, 87 (31%) events were category F, 2 (1%) events were category G, 6 (2%) events were category H, and 4 (1%) events were category I

| Association between adverse events and patient outcomes
Presence of an AE was strongly associated with length of ICU stay after adjusting for APACHE II score and mechanical ventilation status (β = 2.85, 95% CI: 1.09-4.61) but was not associated with 28-day mortality, hospital mortality or length of hospital stay (Tables 2 and   3).
The impact of cause-specific AE classification on length of ICU stay is shown in

| D ISCUSS I ON
We found that 51% of ICU patients suffered from an adverse event, 20% of which were preventable. Most AEs were classified as "temporary harm" and became non-preventable as the severity increased.
Compared to a prospective study with direct ICU observation that reported an AE incidence of 20% at a rate of 80.5 events per 1000 patient-days (Rothschild et al., 2005), our study had a higher fre-  (Resar et al., 2006). In terms of disease classification, AEs were found to be less frequent in neuromuscular (7/25:28%) and more frequent in trauma (8/9:89%) categories, in line with previous studies reporting AEs incidence in 24% of stroke and 29% of trauma patients (Daud-Gallotti et al., 2005;Forster et al., 2008). The incidence of AEs in neurology patients was similar but more common in trauma patients.
The reason for this is unclear, but we cannot rule out the possibility that, in addition to the aforementioned reasons, it was a coincidence due to the small case number.
In our study, AEs were not associated with mortality but were associated with length of ICU stay. Although AE occurrence has been reported to increase mortality (Roque et al., 2016;Sauro et al., 2020), meta-analysis consistently shows that the AE is not statistically relevant (Ahmed et al., 2015). Abbreviations: AE, adverse event; APACHE, Acute Physiology and Chronic Health Evaluation; CI, confidence interval; OR, odds ratio.
The mortality was a binary variable, and logistic regression analysis was performed and values were expressed as ORs. could be prevented if medical staff understand and standardize goal-oriented circulation management methods (Rivers et al., 2001).

TA B L E 3 Multivariate analysis of factors associated with length of stay
Although prevention of adverse drug events is a difficult task, early detection of adverse drug events is important to prevent them from becoming serious. In order to create a new system for early detection, it would be useful to share knowledge of drugs frequently used in the ICU and to enact team-wide checks for major adverse reactions when administering drugs. In addition, a multi-component educational intervention that includes knowledge and practice for nurses could likely improve knowledge and adherence to the Standard Precaution Guidelines, thereby reducing nosocomial infections (Gomarverdi et al., 2019). Other patient safety interventions, such as increased support staff, interdisciplinary team interventions, clinical pathways, and catheter reminder and stop orders, have also been reported to be effective (Zegers et al., 2016 A prospective study using combinations of conventional methods, such as GTT, direct observation and voluntary reporting (Thomas & Petersen, 2003), although costly in time and funding, might be the most suitable for future studies. Trigger tools may be ideal for this purpose as they are both time and cost-effective with higher sensitivity than conventional methods (Classen et al., 2011;Naessens et al., 2009). Second, these results may not extrapolate to regional hospitals or other organizations because our research was conducted at a single ICU of one university hospital. It would thus be necessary to increase the number of study facilities in the future. Third, there may have been a selection bias because we did not survey all admitted patients. However, this bias was minimized by random sampling, a method that has been performed since the Harvard Medical Practice Study (Brennan et al., 1991), and was also reported with GTT (Classen et al., 2011;Naessens et al., 2009). Random selection is a recommended sampling approach sufficient to observe the incidence of AEs and temporal changes (Griffin & Resar, 2009). Fourth, since AEs are classified by GTT as physical injuries, psychological suffering (such as anxiety, depression and fear) was not judged as an AE. Since psychological suffering is a common ICU symptom, GTT should be modified to detect these events (Puntillo et al., 2010).
Fifth, judgments of preventability depended on the subjectivity of the reviewer as there was no consensus process in the secondary review. However, we tried to maintain objectivity by defining and scaling preventability.

| CON CLUS IONS
Adverse events were common in the Japanese ICU, and most adverse events were classified as "temporary harm." AEs were associated with length of ICU stay, and in particular, adverse drug events, procedural complications and nosocomial infections were strongly associated with length of ICU stay.

ACK N OWLED G EM ENTS
We are grateful to Mr. Bryan J. Mathis, Medical English Communications Center, University of Tsukuba Hospital, for the grammatical revision of this paper.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest. All authors read and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets used and/or analysed during the current study are available from the corresponding author on reasonable request.

A PPE N D I X 2
Severity and preventability of adverse events Preventable and non-preventable adverse events categorized between E and I. Category E is temporary harm requiring intervention, category F is temporary harm requiring initial or prolonged hospitalization, category G is permanent harm, category H is life-threatening harm, and category I is harm causing or contributing to death. AE, adverse event.

A PPE N D I X 3
Examples of adverse events with high severity

Cause Severity Preventability Event description
Therapeutic error Permanent More than 50% Tooth damage caused by the use of mouth openers during oral care.
Adverse drug event Life-threatening Less than 50% A new cerebral haemorrhage occurred due to the use of anticoagulants following the implementation of an extracorporeal circulation device.

Procedural complication
Death Virtually no evidence After glycerine enema, the patient developed intestinal perforation and died from septic shock.

Nosocomial infection Death
Less than 50% A tracheostomy was performed, but the infection at the tracheostomy site spread and the patient died of mediastinitis.

A PPE N D I X 4
Adverse event category