Towards automated clinical coding

Abstract

Background. Patients' encounters with healthcare services must undergo clinical coding. These codes are typically derived from free-text notes. Manual clinical coding is expensive, time-consuming and prone to error. Automated clinical coding systems have great potential to save resources, and realtime availability of codes would improve oversight of patient care and accelerate research. Automated coding is made challenging by the idiosyncrasies of clinical text, the large number of disease codes and their unbalanced distribution.Methods. We explore methods for representing clinical text and the labels in hierarchical clinical coding ontologies. Text is represented as term frequencyinverse document frequency counts and then as word embeddings, which we use as input to recurrent neural networks. Labels are represented atomically, and then by learning representations of each node in a coding ontology and composing a representation for each label from its respective node path. We consider different strategies for initialisation of the node representations. We evaluate our methods using the publicly-available Medical Information Mart for Intensive Care III dataset: we extract the history of presenting illness section from each discharge summary in the dataset, then predicting the International Classification of Diseases, ninth revision, Clinical Modification codes associated with these.

Natural language processing algorithms for mapping clinical text fragments onto ontology concepts: a systematic review and recommendations for future studies

“…Out of the 256 publications, we excluded 65 publications, as the described Natural Language Processing algorithms in those publications were not evaluated. The full text of the remaining 191 publications was assessed and 114 a As reported by authors [11,30,33,35,37,41,44,46,49,55,58,64,68,69,72,73,75,76,83,87,90,95,98,100,104] Included reference to dataset 21 (27%) [11,30,35,37,41,44,46,49,55,58,64,72,75,76,83,87,90,95,98,100,104] Training of algorithm Trained 47 (61%) [11, 12, 29, 31, 32, 34, 37, 39, 41, 42, 44, 45, 48-53, 55-59, 62, 63, 65, 66, 68, 69, 73, 74, 76, 78-84, 87, 88, 90, 95, 96, 98, 99, 104] Not listed 3 (3.9%) …”

Formal Medical Knowledge Representation Supports Deep Learning Algorithms, Bioinformatics Pipelines, Genomics Data Analysis, and Big Data Processes

“…Section editors achieved a first selection of 100 papers based on title and abstract. After a second review of this set of papers, including full text reviews, a selection of 15 candidate best papers was established 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 . Five reviewers reviewed these papers and four papers were finally selected as the best papers 2 3 4 5 .…”

“…Two candidate best papers presented a research involving semantic formalization associated with NLP approaches to process clinical texts. The first paper by Catling et al , 13 explored methods for representing clinical text using hierarchical clinical coding ontologies. This study demonstrates that hierarchically-structured medical knowledge can be incorporated into statistical models to produce improved performance for automated clinical coding.…”

Artificial Intelligence (AI) in Rare Diseases: Is the Future Brighter?

“…Recently developed ML tools to analyze heterogeneous, sparse, and noisy clinical data have shed light on this neglected topic [73][74][75][76]97]. Nevertheless, most advances have occurred in unconcerted actions or in disease niches.…”

“…Additionally, author-specific medical terms, abbreviations, and grammatical errors in EHR further hinder biomedical NER. Recurrent NNs proved to be advantageous for automated clinical coding improving representation of RDs in hierarchically-structured medical knowledge [73]. For disease NER, Bhasuran et al implemented a stacked ensemble approach combined with fuzzy matching using both forward and reverse disease labeling.…”