“Learn from Errors”: Post-traumatic growth among second victims | BMC Public Health

Design

We used a descriptive, cross-sectional study design with cluster sampling. The current study utilized the Strengthening Reporting of Observational Studies in Epidemiology (STROBE) cross-sectional checklist [32].

Setting

This study was conducted in a large, comprehensive hospital, which is a major health care provider in the region. The hospital comprises 38 clinical departments and 8 medical technology departments, with a total of 4616 beds. In 2023, the hospital served an impressive 3.4564 million outpatient visits, discharged 197,100 inpatients, and performed 89,000 surgeries. The patient population is diverse, with individuals coming from all over the country. This diversity in the patient population also translates to a wide range of cases and experiences for health care professionals, thereby adding to the richness and complexity of the data collected in our study.

Participants

Inclusion and exclusion criteria

Drawing upon prior studies [33] and definitions [2], the participants, namely, the second victims, are identified as follows: (a) health care professionals who are directly engaged in patient care or patient management; (b) those who have experienced or witnessed a patient safety incident within the past year; (c) individuals who self-assess as having been negatively impacted during these events; and (d) those who have provided informed consent and demonstrated a willingness to participate.

Our study aimed to capture the experiences of those with a greater degree of responsibility and autonomy in their roles. Given the limited experience of interns and trainees and the supervised nature of their roles, they may not engage with or respond to these incidents in the same depth as our target population. Therefore, interns and trainees were excluded from the study.

Sample size

The sample size was calculated with a confidence interval of 95% at a proportion of second victims of 76.88% [33], a marginal error of 5%, and a permissible error of 5%. Considering a loss to follow-up rate of approximately 30%, the required sample size was 357. To ensure the reliability of SEM, a priori power analysis based on the recommendation of Kline et al. [34] of a 20:1 sample size-to-parameter ratio was used for estimating the sample size. Based on a maximum of 20 parameters, the priori-determined sample size was 400.

Measures

Patient safety incidents (PSIs)

PSIs are primarily evaluated based on their frequency and severity. First, participants were asked to report the frequency with which they had experienced or witnessed PSIs in the past month. The responses were scored as follows: 3 (always, ≥ 11), 2 (often, 5 ~ 10), 1 (seldom, 1 ~ 5), or 0 (none). Second, participants were required to indicate the type of their most recent PSI based on the definition and classification provided by the Chinese National Health Commission for medical adverse events [35]. The categories were defined as follows: Level I (events with factual errors leading to consequences), Level II (events without factual errors but still leading to significant consequences), Level III (events without factual errors but leading to minor or no consequences), and Level IV (events without factual errors and without any resulting consequences). A cumulative score was calculated based on these two parameters. A higher total score suggests a greater level of stress induced by the PSIs. The reliability of this scoring system was confirmed with a Cronbach’s alpha of 0.712 in this study.

Perceived threat

The Perceived Threat Scale (PTS), molded after the Perceived Life Threat Scale [36], was employed to assess the perceived threat. This 4-item tool assessed the perceived threat of PSIs to one’s life or work, the likelihood of similar future incidents, the potential disruption to work or life, and the severity of the incident’s consequences. A 5-point Likert scale was used to quantify the responses. The scale demonstrated acceptable reliability in this study, with a Cronbach’s alpha of 0.756.

Coping style

The coping style was assessed using the Trait Coping Style Questionnaire (TCSQ), which is divided into two dimensions: negative coping (NC) and positive coping (PC). Each dimension contains 10 items, and a 5-point Likert scale was used for scoring (5 points for “strongly agree” and 1 point for “strongly disagree”). A higher score in each dimension indicates more pronounced positive or negative strategies. The TCSQ is widely used in the Chinese population, and the two dimensions are typically analyzed separately [37]. In this study, the Cronbach’s alpha was 0.814 for the positive coping dimension and 0.805 for the NC dimension, demonstrating good reliability.

Social support

Social support was measured using the Second Victim Experience and Support Tool (SVEST). The SVEST was developed by Burlison et al. [38] and is the first tool designed to evaluate the experiences of second victims and the quality of support resources available to them. This tool has been widely used in China [13, 33]. The social support part includes five dimensions and 18 items, with a Cronbach’s alpha of 0.854, indicating good reliability. The responses are scored using a 5-point Likert scale (5 points for “strongly agree” and 1 point for “strongly disagree”). A higher score on this scale indicates a greater level of perceived social support.

Second-victim symptoms (SVS)

The negative outcome, second victim symptoms (SVS), was also assessed using the SVEST of [38]. This aspect of the tool includes two dimensions and eight items, with a Cronbach’s alpha of 0.901, indicating excellent reliability.

Posttraumatic growth (PTG)

PTG was measured using the Chinese Posttraumatic Growth Inventory (C-PTGI). The C-PTGI was adapted and translated into Mandarin by Wang [39] from the Posttraumatic Growth Inventory [40]. The C-PTGI consists of 20 items distributed across 5 dimensions. Responses are scored on a six-point Likert scale, with 6 points given for “very much” and 1 point for “not at all”. Based on the participants’ scores, they were categorized into three levels of growth: low (less than 60 points), middle (60–65 points), and high (66–100 points). In this study, the C-PTGI demonstrated excellent reliability, with a Cronbach’s alpha of 0.953.

Data collection

The online survey was created through a free website ( In November 2021, the survey link was disseminated to qualified participants via WeChat groups. This distribution was facilitated by department heads and head nurses to ensure that the link reached the intended audience. The groups were chosen based on the cluster sampling method; each group represented a ‘cluster.’ The clusters were formed based on the departments, specifically including medical departments, surgery departments, and technology-related departments. We carried out the survey across all regular employee within these departments, encompassing three professional categories: nurses, doctors, or medical technical staff. To ensure data quality, we excluded the following questionnaires: (a) those with identical answers on both reverse and forward questions; (b) those completed in less than nine minutes, as per the pilot study; and (c) those in which all of the answers were the same.

Data analysis

Data analysis was performed using IBM SPSS and AMOS version 26.0 (IBM Corp, Armonk, NY, USA). We used descriptive statistics to describe the demographic information. Continuous data, such as the scores from the C-PTGI, are reported as the means along with their standard deviations (SDs). To ensure the reliability of our model, we first checked for multicollinearity among the independent variables by calculating the variance inflation factor (VIF). The results showed that the VIF for all variables was less than 3, indicating no significant multicollinearity issues among the variables. Bivariate analyses were conducted to explore the relationships among all of the variables. To identify the underlying mechanisms of PTG among second victims, we employed structural equation modeling (SEM). We assessed the model fit using several indices [41]: a relative chi-square (χ2/df) less than 5, a goodness-of-fit index (GFI) over 0.9, a comparative fit index (CFI) above 0.90, and a root mean squared error of approximation (RMSEA) below 0.08. The bootstrap method (with 5000 samples) was used to calculate the 95% confidence intervals (CIs) for direct effects, indirect effects, and total effects. A P value less than 0.05 (two-sided) was considered to indicate statistical significance.

Ethical consideration

Ethical approval for the study was obtained from the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University (reference number 2019-067). Participants were assured that their decision to participate or not would not affect their professional status or opportunities within the institution. Only those who provided written consent on the first page of the survey were allowed to proceed with answering the questions. They were informed that they could withdraw from the study at any time. To ensure confidentiality, any data sharing or reporting was conducted in a manner that protected the privacy of individual participants. To prevent any unnecessary administrative coercion and to ensure the integrity of the data, each questionnaire was anonymized. Additionally, measures were taken to permit each IP address to submit the survey only once.