Exploring misclassification of injury intent: A burn register study

Abstract

IntroductionBurn registers are an important source of surveillance data on injury intent. These data are considered essential to inform prevention activities. In South Asia, intentional burn injuries are thought to disproportionately affect women. Assessment of injury intent is difficult because it is influenced by personal, family, social, and legal sensitivities. This can introduce misclassification into data, and bias analyses. We conducted a descriptive, hypothesis generating study to explore misclassification of injury intent using data from a newly digitised single centre burn register in south India. MethodsData from 1st February 2016 to 28th February 2022 were analysed. All patients in the data set were included in the study (n = 1930). Demographic and clinical characteristics for patients are described for each classification of injury intent. All data cleaning and analyses were completed using RStudio. ResultsInjury intent data were missing for 12.6% of cases. It was the most commonly missing variable in the data set. “Accidental” injuries had a similar distribution over time, age, and total body surface area (TBSA) for males and females. “Homicidal” injuries were more common in females. Injuries reported as “Suicidal” affected men and women equally. A decrease in reporting of “Suicidal” injuries in females corresponded to an increase in high TBSA injuries classified as ‘Other’ or with missing data. Overwriting of injury intent was present in 1.5% of cases. The overwritten group had a greater proportion of females (62.1% vs. 48.5%) and higher median TBSA (77.5% vs. 27.5%) compared to the group where intent was not overwritten. ConclusionOur findings indicate that some subgroups, such as females with high TBSA burns, appear to be more likely to be misclassified and should be the focus of future research. They also highlight that quality of surveillance data could be improved by recording of clinical impression, change in patient reported intent, and use of a common data element for intent to standardise data collection. We also recommend that injury intent is recorded as a unique variable and should not be mixed with other elements of injury causation (e.g. mechanism). Although this is a single centre study, the methods will be of interest to those who utilise routinely collected data and wish to reduce misclassification of this important variable.