Patient Characteristics Associated With Voluntary Safety Event Reporting in the Acute Care Setting

This study was performed in a 160-bed tertiary-care academic hospital with >4500 pediatric admissions per year. We performed a retrospective cohort study of all pediatric patients 2 to 17 years old seen for an acute care encounter between February 2015 and February 2016. Acute care encounters were defined as any patient admitted to the emergency department, operating room, or inpatient pediatric unit (including ICUs). This study was approved by the local institutional review board.

Patients aged <2 years were excluded because overweight and obesity are not consistently defined in this age group.¹⁵  Patients admitted to outlying hospitals were also excluded because these facilities do not use institutional VER for event reporting. We included the first encounter during the study period for each patient.

For all patient encounters, age, sex, height, weight, race or ethnicity, insurance provider, and length of stay (LOS) were collected. These variables were chosen on the basis of previous literature examining patient characteristics and socioeconomic disparities in safety events and their ability to be reliably extracted from the medical record.^12,13  We accepted heights and weights within 3 months of the index encounter; patients without height and weight in that time frame were excluded. BMI category was defined as underweight, healthy weight, overweight, and obese according to the Centers for Disease Control and Prevention definitions.^15,16  Race or ethnicity was patient reported as documented within the electronic medical record and were defined as white, African American, Hispanic, and other. Insurance provider was taken at the time of chart review and categorized as either public or private.

All reports submitted to the hospital VER system during the same time period were reviewed. Encounters were matched to reports submitted to the VER system by using medical record number and encounter date. This institution uses the VER system RL Solutions as its primary method of recognizing adverse safety events.

Descriptive statistics were used to define these demographic categories. Our primary outcome was any report of a safety event and was compared between patient characteristics by using Pearson’s χ² analysis for proportions and univariate logistic regression. All demographic characteristics found to be significant in the univariate analysis were incorporated into multivariable logistic regression models of event reporting. A P value < .1 was considered significant for inclusion in the model; otherwise, P < .05 was used to define statistical significance. All analyses were performed with SAS version 9.4 (SAS Institute, Inc, Cary, NC).

During the study period, 22 062 patients aged 2 to 17 years with an acute care encounter were identified for inclusion (Fig 1). Eight were excluded because of missing BMI data, resulting in a final cohort of 22 056 patients.

Our cohort was 52.4% boys and had a median age of 8 years (interquartile range [IQR] 4–13). Median LOS was 0.09 (IQR 0.05–1.9) days. A total of 5.4% were classified as underweight, 53.6% as normal weight, 15.6% as overweight, and 25.4% as obese. A total of 48.3% of patients reported their race or ethnicity as white non-Hispanic, 28.6% as African American non-Hispanic, 19.9% as Hispanic, and 3.1% reported as other.

Three hundred and forty-one patients (1.5%) had at least 1 safety event reported during the study time period. Demographics of patients with and without reported adverse events are shown in Table 1. Patient characteristics with significant differences between those with and without a reported safety event included BMI classification (P = .03), age (P < .01), and race or ethnicity (P < .01). LOS differed significantly between patients with and without a reported safety event (P < .01); it is unknown if this was a predictor or consequence of the safety event.

In Fig 2, we show results of multivariable logistical regression model incorporating significant patient characteristics. In this model, patients with obesity (odds ratio [OR] 0.69; 95% confidence interval [CI] 0.49–0.97) and African American race (OR 0.65; 95% CI 0.46–0.93 relative to white race) remained significantly less likely to have a safety event reported. LOS was positively associated (OR 1.51; 95% CI 1.46–1.55) with safety event reporting.

In this study, we demonstrated decreased reporting of safety events in patients with obesity and those of African American race using a VER system. To our knowledge, this is the first study used to identify associations between demographic characteristics of pediatric patients and voluntary safety event reporting. This result is not consistent with research on actual frequency of adverse events and quality indicators in the literature, which suggests patients with obesity and those of African American and Hispanic race are more likely to experience a safety event during their hospital encounter.^12,13  By using the VER system, we only detected events reported by providers, which could be influenced by many factors, including the perceived severity of the event, the awareness of different medical team members of the event, and conscious or unconscious bias of the provider toward different patient populations. Provider bias has been demonstrated toward patients with obesity and minority race, although there is mixed evidence on bias influencing patient care.^17–22 

We did not detect a statistically significant difference in event reporting for patients with public versus private insurance. In looking at the frequency of safety events in the literature, results are mixed. Authors of some large studies using hospital administrative data did find significantly higher rates of safety events in patients with public insurance, but authors of studies using more thorough chart review suggest this may be attributable to patient health characteristics.^23,24 

We recognize limitations to this study. The results from this single-center study may have limited generalizability. It should be noted that the baseline event reporting rate of 1.5% is similar to studies used to examine VER previously.^5,25,26  Because our study did not include chart review of all encounters for the actual frequency of safety events, we cannot say if this difference in frequency of reported events is equal to a difference in frequency of actual safety events. We also recognize that excluding patients <2 years of age likely skews the age data and excludes a significant patient population in pediatrics. We did not include medical complexity or primary language in this study, although these have been associated with adverse events. Further studies should be conducted in which authors use trigger tools or other less subjective forms of event detection to compare provider reporting with the true frequency of safety events in different patient populations.

Given that VER is currently the industry standard in safety event detection for most children’s hospitals, it is important to examine the validity and weaknesses of this system. It is suggested in this study that VER may underreport safety events in socioeconomically vulnerable populations. More research is needed to determine if discrepancies in event reporting between different demographic groups correlate with the true frequency of patient safety events.

We acknowledge the data extraction services of the Wake Forest Clinical and Translational Science Institute, which is supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through grant award UL1TR001420. We also acknowledge Hannah Hulshult for assistance with data collection.