{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,3]],"date-time":"2026-04-03T06:59:29Z","timestamp":1775199569403,"version":"3.50.1"},"reference-count":26,"publisher":"Springer Science and Business Media LLC","issue":"S11","license":[{"start":{"date-parts":[[2020,12,1]],"date-time":"2020-12-01T00:00:00Z","timestamp":1606780800000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2020,12,30]],"date-time":"2020-12-30T00:00:00Z","timestamp":1609286400000},"content-version":"vor","delay-in-days":29,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["BMC Med Inform Decis Mak"],"published-print":{"date-parts":[[2020,12]]},"abstract":"Abstract<\/jats:title>\n \n Background<\/jats:title>\n The availability of massive amount of data enables the possibility of clinical predictive tasks. Deep learning methods have achieved promising performance on the tasks. However, most existing methods suffer from three limitations: (1) There are lots of missing value for real value events, many methods impute the missing value and then train their models based on the imputed values, which may introduce imputation bias. The models\u2019 performance is highly dependent on the imputation accuracy. (2) Lots of existing studies just take Boolean value medical events (e.g. diagnosis code) as inputs, but ignore real value medical events (e.g., lab tests and vital signs), which are more important for acute disease (e.g., sepsis) and mortality prediction. (3) Existing interpretable models can illustrate which medical events are conducive to the output results, but are not able to give contributions of patterns among medical events.<\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n \n In this study, we propose a novel interpretable\n P<\/jats:bold>\n attern\n A<\/jats:bold>\n ttention model with\n V<\/jats:bold>\n alue\n E<\/jats:bold>\n mbedding (PAVE) to predict the risks of certain diseases. PAVE takes the embedding of various medical events, their values and the corresponding occurring time as inputs, leverage self-attention mechanism to attend to meaningful patterns among medical events for risk prediction tasks. Because only the observed values are embedded into vectors, we don\u2019t need to impute the missing values and thus avoids the imputations bias. Moreover, the self-attention mechanism is helpful for the model interpretability, which means the proposed model can output which patterns cause high risks.\n <\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n We conduct sepsis onset prediction and mortality prediction experiments on a publicly available dataset MIMIC-III and our proprietary EHR dataset. The experimental results show that PAVE outperforms existing models. Moreover, by analyzing the self-attention weights, our model outputs meaningful medical event patterns related to mortality.<\/jats:p>\n <\/jats:sec>\n \n Conclusions<\/jats:title>\n PAVE learns effective medical event representation by incorporating the values and occurring time, which can improve the risk prediction performance. Moreover, the presented self-attention mechanism can not only capture patients\u2019 health state information, but also output the contributions of various medical event patterns, which pave the way for interpretable clinical risk predictions.<\/jats:p>\n <\/jats:sec>\n \n Availability<\/jats:title>\n \n The code for this paper is available at:\n https:\/\/github.com\/yinchangchang\/PAVE<\/jats:ext-link>\n .\n <\/jats:p>\n <\/jats:sec>","DOI":"10.1186\/s12911-020-01331-7","type":"journal-article","created":{"date-parts":[[2020,12,30]],"date-time":"2020-12-30T02:02:24Z","timestamp":1609293744000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":21,"title":["An interpretable risk prediction model for healthcare with pattern attention"],"prefix":"10.1186","volume":"20","author":[{"given":"Sundreen Asad","family":"Kamal","sequence":"first","affiliation":[]},{"given":"Changchang","family":"Yin","sequence":"additional","affiliation":[]},{"given":"Buyue","family":"Qian","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4601-0779","authenticated-orcid":false,"given":"Ping","family":"Zhang","sequence":"additional","affiliation":[]}],"member":"297","published-online":{"date-parts":[[2020,12,30]]},"reference":[{"key":"1331_CR1","doi-asserted-by":"crossref","unstructured":"Cheng Y, Wang F, Zhang P, Hu J. Risk prediction with electronic health records: a deep learning approach. In: Proceedings of the 2016 SIAM international conference on data mining. London: SIAM; 2016. p. 432\u201340.","DOI":"10.1137\/1.9781611974348.49"},{"key":"1331_CR2","doi-asserted-by":"crossref","unstructured":"Zhang X, Qian B, Li Y, et\u00a0al. KnowRisk: an interpretable knowledge-guided model for disease risk prediction. In: 2019 IEEE international conference on data mining (ICDM). New York: IEEE; 2019. p. 1492\u20137.","DOI":"10.1109\/ICDM.2019.00196"},{"key":"1331_CR3","doi-asserted-by":"crossref","unstructured":"Baytas IM, Xiao C, Zhang X, et\u00a0al. Patient subtyping via time-aware lstm networks. In: Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining; 2017. p. 65\u201374.","DOI":"10.1145\/3097983.3097997"},{"key":"1331_CR4","doi-asserted-by":"crossref","unstructured":"Yin C, Liu R, Zhang D, Zhang P. Identifying sepsis subphenotypes via time-aware multi-modal auto-encoder. In: Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining; 2020. p. 862\u201372.","DOI":"10.1145\/3394486.3403129"},{"key":"1331_CR5","doi-asserted-by":"publisher","first-page":"61","DOI":"10.1146\/annurev-publhealth-031811-124606","volume":"34","author":"TA Glass","year":"2013","unstructured":"Glass TA, Goodman SN, Hern\u00e1n MA, Samet JM. Causal inference in public health. Annu Rev Public Health. 2013;34:61\u201375.","journal-title":"Annu Rev Public Health"},{"key":"1331_CR6","doi-asserted-by":"crossref","unstructured":"Liu R, Yin C, Zhang P. Estimating individual treatment effects with time-varying confounders. arXiv preprint; 2020. arXiv:200813620.","DOI":"10.1109\/ICDM50108.2020.00047"},{"key":"1331_CR7","doi-asserted-by":"crossref","unstructured":"Zhu Z, Yin C, Qian B, et\u00a0al. Measuring patient similarities via a deep architecture with medical concept embedding. In: 2016 IEEE 16th international conference on data mining (ICDM). London: IEEE; 2016. p. 749\u201358.","DOI":"10.1109\/ICDM.2016.0086"},{"key":"1331_CR8","unstructured":"Choi E, Bahadori MT, Sun J, et\u00a0al. Retain: an interpretable predictive model for healthcare using reverse time attention mechanism. In: Advances in neural information processing systems; 2016. p. 3504\u201312."},{"key":"1331_CR9","doi-asserted-by":"crossref","unstructured":"Choi E, Bahadori MT, Song L, et\u00a0al. GRAM: graph-based attention model for healthcare representation learning. In: Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining; 2017. p. 787\u201395.","DOI":"10.1145\/3097983.3098126"},{"key":"1331_CR10","doi-asserted-by":"crossref","unstructured":"Ma F, You Q, Xiao H, et\u00a0al. KAME: knowledge-based attention model for diagnosis prediction in healthcare; 2018. p. 743\u201352.","DOI":"10.1145\/3269206.3271701"},{"key":"1331_CR11","doi-asserted-by":"crossref","unstructured":"Ma F, Chitta R, Zhou J, et\u00a0al. Dipole: Diagnosis prediction in healthcare via attention-based bidirectional recurrent neural networks. In: Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining; 2017. p. 1903\u201311.","DOI":"10.1145\/3097983.3098088"},{"key":"1331_CR12","doi-asserted-by":"crossref","unstructured":"Yin C, Qian B, Cao S, Li X, Wei J, Zheng Q, et\u00a0al. Deep similarity-based batch mode active learning with exploration\u2013exploitation. In: 2017 IEEE international conference on data mining (ICDM). IEEE; 2017. p. 575\u201384.","DOI":"10.1109\/ICDM.2017.67"},{"issue":"1","key":"1331_CR13","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s41598-017-17765-5","volume":"8","author":"Z Che","year":"2018","unstructured":"Che Z, Purushotham S, Cho K, et al. Recurrent neural networks for multivariate time series with missing values. Sci Rep. 2018;8(1):1\u201312.","journal-title":"Sci Rep"},{"issue":"1","key":"1331_CR14","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/sdata.2016.35","volume":"3","author":"AE Johnson","year":"2016","unstructured":"Johnson AE, Pollard TJ, Shen L, et al. MIMIC-III, a freely accessible critical care database. Sci Data. 2016;3(1):1\u20139.","journal-title":"Sci Data"},{"key":"1331_CR15","unstructured":"Razavian N, Marcus J, Sontag D. Multi-task prediction of disease onsets from longitudinal laboratory tests. In: Machine learning for healthcare conference; 2016. p. 73\u2013100."},{"key":"1331_CR16","doi-asserted-by":"crossref","unstructured":"Ma F, Gao J, Suo Q, You Q, Zhou J, Zhang A. Risk prediction on electronic health records with prior medical knowledge. In: Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery and data mining; 2018. p. 1910\u20139.","DOI":"10.1145\/3219819.3220020"},{"issue":"2","key":"1331_CR17","doi-asserted-by":"publisher","first-page":"361","DOI":"10.1093\/jamia\/ocw112","volume":"24","author":"E Choi","year":"2017","unstructured":"Choi E, Schuetz A, Stewart WF, Sun J. Using recurrent neural network models for early detection of heart failure onset. J Am Med Inform Assoc. 2017;24(2):361\u201370.","journal-title":"J Am Med Inform Assoc"},{"issue":"1","key":"1331_CR18","doi-asserted-by":"publisher","first-page":"36","DOI":"10.1186\/s12911-019-0775-2","volume":"19","author":"M Beeksma","year":"2019","unstructured":"Beeksma M, Verberne S, Van den Bosch A, et al. Predicting life expectancy with a long short-term memory recurrent neural network using electronic medical records. BMC Med Inform Decis Mak. 2019;19(1):36.","journal-title":"BMC Med Inform Decis Mak"},{"key":"1331_CR19","doi-asserted-by":"crossref","unstructured":"Che Z, Kale D, Li W, Bahadori MT, Liu Y. Deep computational phenotyping. In: Proceedings of the 21st ACM SIGKDD international conference on knowledge discovery and data mining; 2015. p. 507\u201316.","DOI":"10.1145\/2783258.2783365"},{"key":"1331_CR20","unstructured":"Lipton ZC, Kale D, Wetzel R. Directly modeling missing data in sequences with RNNS: improved classification of clinical time series. In: Machine learning for healthcare conference; 2016. p. 253\u201370."},{"key":"1331_CR21","first-page":"576","volume-title":"An interpretable fast model for predicting the risk of heart failure","author":"X Zhang","year":"2019","unstructured":"Zhang X, Qian B, Li X, Wei J, Zheng Y, Song L, et al. An interpretable fast model for predicting the risk of heart failure. London: SIAM; 2019. p. 576\u201384."},{"key":"1331_CR22","unstructured":"Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et\u00a0al. Attention is all you need. In: Advances in neural information processing systems; 2017. p. 5998\u20136008."},{"key":"1331_CR23","doi-asserted-by":"crossref","unstructured":"Gligorijevic D, Stojanovic J, Satz W, Stojkovic I, Schreyer K, Del\u00a0Portal D, et\u00a0al. Deep attention model for triage of emergency department patients. In: Proceedings of the 2018 SIAM international conference on data mining. London: SIAM; 2018. p. 297\u2013305.","DOI":"10.1137\/1.9781611975321.34"},{"key":"1331_CR24","doi-asserted-by":"publisher","first-page":"1250","DOI":"10.1097\/01.CCM.0000050454.01978.3B","volume":"31","author":"M Levy","year":"2003","unstructured":"Levy M. 2001 SCCM\/ESICM\/ATS\/SIS international sepsis definitions conference. Crit Care Med. 2003;31:1250\u20136.","journal-title":"Crit Care Med"},{"key":"1331_CR25","doi-asserted-by":"crossref","unstructured":"Cho K, van Merrienboer B, Bahdanau D, Bengio Y. On the properties of neural machine translation: encoder\u2013decoder approaches; 2014.","DOI":"10.3115\/v1\/W14-4012"},{"issue":"8","key":"1331_CR26","doi-asserted-by":"publisher","first-page":"735","DOI":"10.1162\/neco.1997.9.8.1735","volume":"9","author":"S Hochreiter","year":"1997","unstructured":"Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput. 1997;9(8):735\u20131780.","journal-title":"Neural Comput"}],"container-title":["BMC Medical Informatics and Decision Making"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/link.springer.com\/content\/pdf\/10.1186\/s12911-020-01331-7.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"http:\/\/link.springer.com\/article\/10.1186\/s12911-020-01331-7\/fulltext.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"http:\/\/link.springer.com\/content\/pdf\/10.1186\/s12911-020-01331-7.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2021,4,18]],"date-time":"2021-04-18T08:42:17Z","timestamp":1618735337000},"score":1,"resource":{"primary":{"URL":"https:\/\/bmcmedinformdecismak.biomedcentral.com\/articles\/10.1186\/s12911-020-01331-7"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,12]]},"references-count":26,"journal-issue":{"issue":"S11","published-print":{"date-parts":[[2020,12]]}},"alternative-id":["1331"],"URL":"https:\/\/doi.org\/10.1186\/s12911-020-01331-7","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2020.07.26.20162479","asserted-by":"object"}]},"ISSN":["1472-6947"],"issn-type":[{"value":"1472-6947","type":"electronic"}],"subject":[],"published":{"date-parts":[[2020,12]]},"assertion":[{"value":"10 November 2020","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"12 November 2020","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"30 December 2020","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The proprietary EHR data used in this study was extracted from Xi\u2019an Jiaotong University and the study has been approved by Xi\u2019an Jiaotong University. Requirement for individual patient consent was waived since all protected health information was de-identified. De-identification was performed in compliance with Health Insurance Portability and Accountability Act (HIPAA) standards. Deletion of protected health information (PHI) from structured data sources (e.g., database fields that provide patient name or date of birth) was straightforward. Research use of publicly available MIMIC-III has been approved by the Institutional Review Board of Beth Israel Deaconess Medical Center and Massachusetts Institute of Technology [\n \n ].","order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethics approval and consent to participate"}},{"value":"Not applicable.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Consent for publication"}},{"value":"PZ is the member of the editorial board of BMC Medical Informatics and Decision Making. The authors declare that they have no other competing interests.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"307"}}