Machine learning models to detect social distress, spiritual pain, and severe physical psychological symptoms in terminally ill patients with cancer from unstructured text data in electronic medical records

Abstract

Background: Few studies have developed automatic systems for identifying social distress, spiritual pain, and severe physical and phycological symptoms from text data in electronic medical records. Aim: To develop models to detect social distress, spiritual pain, and severe physical and psychological symptoms in terminally ill patients with cancer from unstructured text data contained in electronic medical records. Design: A retrospective study of 1,554,736 narrative clinical records was analyzed 1 month before patients died. Supervised machine learning models were trained to detect comprehensive symptoms, and the performance of the models was tested using the area under the receiver operating characteristic curve (AUROC) and precision recall curve (AUPRC). Setting/participants: A total of 808 patients was included in the study using records obtained from a university hospital in Japan between January 1, 2018 and December 31, 2019. As training data, we used medical records labeled for detecting social distress ( n = 10,000) and spiritual pain ( n = 10,000), and records that could be combined with the Support Team Assessment Schedule (based on date) for detecting severe physical/psychological symptoms ( n = 5409). Results: Machine learning models for detecting social distress had AUROC and AUPRC values of 0.98 and 0.61, respectively; values for spiritual pain, were 0.90 and 0.58, respectively. The machine learning models accurately identified severe symptoms (pain, dyspnea, nausea, insomnia, and anxiety) with a high level of discrimination (AUROC > 0.8). Conclusion: The machine learning models could detect social distress, spiritual pain, and severe symptoms in terminally ill patients with cancer from text data contained in electronic medical records.

The use of natural language processing in palliative care research: A scoping review

“…Clinical notes and electronic medical records were the most common primary data sources, used in 57 studies (69.5%). 21,23–27,29,30,33–36,40,42–46,48–52,54,55,57–64,66–68,70–73,75,79–93,95 Other primary sources included audio recordings ( n = 6, 7.3%), 6,28,32,38,39,65 administrative data ( n = 5, 6.1%), 37,47,53,77,...…”

“…These applications offer the possibility of several toolkits for natural language processing and also the possibility to develop codes based on the researchers' needs. Another interesting finding was that many of the studies used a combination of more than one 27 Elhazmi et al, 33 Ganguli et al, 36 George et al, 37 Hu et al, 44 Kehl et al, 48 Laios et al, 50 Lin et al, 57 Manz et al, 66 Agarwal et al, 74 Santos et al, 77 Sung et al, 82 Ye et al 93 Assessment of the impact of interventions 9 (10.9) Ando et al, 22 Greer et al, 42 Lakin et al, 51 Lefèvre et al, 56 Macieira et al, 64 Santarpia et al, 76 Steiner et al, 80 Udelsman et al, 87 Uyeda et al 89 Social and spiritual health 8 (9.7) Gray et al, 40 Johnson et al, 46,47 Masukawa et al, 67 Yoon et al, 94 Ando et al, 96 Ando et al 99 Topic identification 8 (9.7) Sarmet et al, 5 Ando et al, 21 Chan et al, 26 Davoudi et al, 29 Agaronnik et al, 52 Lucini et al, 61 Seale et al, 78 Wang et al 90 Advance care planning/EOL process measures/Code-status clarification/Goals of care documentation 8 (9.7)…”

“…Clinical notes and electronic medical records were the most common primary data sources, used in 57 studies (69.5%). 21,[23][24][25][26][27]29,30,[33][34][35][36]40,[42][43][44][45][46][48][49][50][51][52]54,55,[57][58][59][60][61][62][63][64][66][67][68][70][71][72][73]75,[79][80][81][82][83][84][85][86][87][88][89]…”

Opportunities for computational tools in palliative care: Supporting patient needs and lowering burden

“…3 Further, mortality-based referral does not recognize the varied sources of patient need set forth by nationally-recognized clinical practice guidelines for quality palliative care. 4 Computational methods have potential beyond prognosis; they offer an opportunity to help determine which patients could most benefit from palliative care by efficiently detecting and measuring relevant features such as those identified by Masukawa et al, 2 including evidence of social distress, spiritual pain, and severe physical and psychological symptoms documented in the medical record. Further development of measurements such as these make it possible to introduce alternative thresholds and criteria for referral which are grounded in data and scalable, and which depend on more than a single outcome that may not be most relevant to patient experience.…”

“…Computational methods (e.g., machine learning, natural language processing) present one way to capture patient-centered information in a low-burden way, offering increased means for precision palliative care. The article that accompanies this editorial by Masukawa et al 2 mentions the burden of data collection to patients, and emphasizes the importance of making the best use of existing data collected as a part of routine clinical care. Through aggregating the extensive data available in the electronic health record, and using novel methods for leveraging the content of naturally-occurring clinical conversations, computational tools can help us better understand complex and traditionally hard-to-measure phenomena.…”