Don't Let Notes Be Misunderstood: A Negation Detection Method for Assessing Risk of Suicide in Mental Health Records

Abstract

Mental Health Records (MHRs) contain freetext documentation about patients' suicide and suicidality. In this paper, we address the problem of determining whether grammatic variants (inflections) of the word "suicide" are affirmed or negated. To achieve this, we populate and annotate a dataset with over 6,000 sentences originating from a large repository of MHRs. The resulting dataset has high InterAnnotator Agreement (κ 0.93). Furthermore, we develop and propose a negation detection method that leverages syntactic features of text 1 . Using parse trees, we build a set of basic rules that rely on minimum domain knowledge and render the problem as binary classification (affirmed vs. negated). Since the overall goal is to identify patients who are expected to be at high risk of suicide, we focus on the evaluation of positive (affirmed) cases as determined by our classifier. Our negation detection approach yields a recall (sensitivity) value of 94.6% for the positive cases and an overall accuracy value of 91.9%. We believe that our approach can be integrated with other clinical Natural Language Processing tools in order to further advance information extraction capabilities.

Proceedings of the Fourth Workshop on Computational Linguistics and Clinical Psychology –- From Linguistic Signal to Clinical Reality

“…Resnik et al (2015a) proved that such approaches can be successfully used in identifying users with depression, who have self-disclosed their mental illnesses on Twitter. In general, a clear distinction in the lexical and syntactic structure of the language used by individuals with different mental disorders, as well as between individuals within a control group, can be identified throughout the literature mentioned above, as well as from the explorative analysis conducted by Gkotsis et al (2016). Due to the reliability of the lexical and behavioral features used in many of the models mentioned above, our proposed solution also focused on these feature categories.…”

Classifying the reported ability in clinical mobility descriptions

“…However, as Action Polarity is defined in terms of the interaction between an individual and a specific environment, it adds a layer of complexity to noninteractive physiological observations. Gkotsis et al (2016) investigate using parsing-based scoping limitations for negation detection in complex clinical statements, though their focus is specifically on mentions of suicide.…”

A fast, accurate, and generalisable heuristic-based negation detection algorithm for clinical text

“…Negation detection determines whether a clinical finding mentioned in a narrative is 12 present or absent, usually using the sentence mentioning the concept as input [1]. Many 13 methodologies have been applied to the task of negation detection, traditionally using rule-based 14 methods, with a more recent focus on machine learning algorithms. The performance of rule-based 15 methods for negation in comparison to machine learning methods is hotly contested.…”

“…46 One such approach, DEEPEN [11], operates upon concepts that NegEx determines to be 47 negated [11]. Other dependency-based algorithms make no use of NegEx, such as NegBio, 48 negation-detection, and DepNeg [12][13][14], reported to exhibit an improved precision over syntactic 49 approaches. However, an independent assessment showed that ConText maintained its performance 50 over a novel dataset, while the other approaches did not [2].…”