Introduction
Burns are a major source of morbidity, accounting for over 16 million injuries annually. Approximately 8% of burns are intentional. Information about intent is essential to inform preventative interventions. Most intentional injuries occur in low- and middle-income countries, particularly in South Asia. There is a lack of high-quality data on injury intent in many parts of the world. Where data are collected, such data are prone to bias due to social and legal factors. The core research question that this thesis aimed to address was: “How can international burn injury morbidity surveillance data be strengthened to better differentiate burns that are: unintentional; due to self-harm; or due to interpersonal violence?”.

Methods
The thesis research question was addressed through five studies. Each corresponds to a chapter of the thesis. In chapter two (papers 1 and 2), I investigated data that are collected about injury intent across 13 international burn registers. Data dictionaries were compared, and potential sources of bias were considered. In chapter three (papers 3 and 4), I used a systematic scoping review to investigate injury intent data from South Asia. Terminology and methods used to differentiate intent of hospital presenting burns were investigated. In chapter four (paper 5), I used process mapping to design a robust case ascertainment strategy for a new self-harm register that includes intent information at two hospitals in south India. In chapter five (paper 6), I describe how to assess the quality of existing hospital-based surveillance data on injury intent and then digitise these data for epidemiological analysis. I demonstrate this using an example of six years of handwritten data from a hospital in south India. I analysed these data using exploratory data analysis techniques to investigate patient groups that may be at risk of misclassification of injury intent (chapter six – paper 7).

Results
A lot of information is being collected about burn injury intent internationally. Data dictionaries from national-level burn registers include 43 variables about intent across 12 registers. The systematic scoping review showed that 89 primary research studies from South Asia included information on injury intent. Lack of definitions for key terms limit data comparisons. Where method of assessment for intent is described, it is primarily based on clinician documentation. In South Asia, this is likely to be influenced by medicolegal processes where patient reported intent must be documented. Process mapping showed that multiple routes to emergency care in hospitals in south India could introduce selection bias into data. I showed that a government hospital in south India collects high-quality handwritten burn injury information. Digitisation of these data was possible and yielded very low error rates (0.06% per field). Exploratory analyses of these data indicate that some groups (e.g. women with large total body surface area burns) may be misclassified, and that overwriting of intent data may be a useful predictor of misclassification.

Conclusions
I have identified numerous ways that the international burns community can strengthen intent data. Firstly, digitisation of high-quality handwritten routinely collected data is a feasible way to begin to address data inequity in low- and middle-income countries. Secondly, there needs to be development of a common data element for injury intent to reduce the risk of misclassification. This should include standardised definitions and method of assessment, and be acceptable to clinicians and patients. It can be implemented in all burn registers and used as a reporting standard in publications. It is likely that an explicit reduction of the concept of intent in injury surveillance to ‘Who did the act that resulted in the burn injury?’ would minimise confusion amongst professional groups, but still provide valuable public health data. Thirdly, process mapping is a useful technique to explore and document potential sources of selection bias in registers that include intent information. Finally, exploratory data analyses are a promising method to identify misclassification in existing intent data, and should be encouraged in burn injury studies.