Multicentre validation of a sepsis prediction algorithm using only vital sign data in the emergency department, general ward and ICU

BMJ Open - Tập 8 Số 1 - Trang e017833 - 2018
Qingqing Mao1, Melissa Jay1, Jana Hoffman1, Jacob Calvert1, Christopher Barton2, David Shimabukuro3, Lisa Shieh4, Uli K. Chettipally2,5, Grant Fletcher6, Yaniv Kerem7,8, Yifan Zhou1,9, Ritankar Das1
1Dascena, Inc, Hayward, California, USA
2Department of Emergency Medicine, University of California San Francisco, San Francisco, California, USA
3Department of Anesthesia and Perioperative Care, University of California, San Francisco, San Francisco, California, USA
4Department of Medicine, Stanford University School of Medicine, Stanford, California USA
5Kaiser Permanente South San Francisco Medical Center, South San Francisco, California, USA
6Division of Internal Medicine, University of Washington School of Medicine, Seattle, Washington, USA
7Department of Clinical Informatics, Stanford University School of Medicine, Stanford, California, USA
8Department of Emergency Medicine, Kaiser Permanente Redwood City Medical Center, Redwood City, California, USA
9Department of Statistics, University of California Berkeley, Berkeley, California, USA

Tóm tắt

Objectives

We validate a machine learning-based sepsis-prediction algorithm (InSight) for the detection and prediction of three sepsis-related gold standards, using only six vital signs. We evaluate robustness to missing data, customisation to site-specific data using transfer learning and generalisability to new settings.

Design

A machine-learning algorithm with gradient tree boosting. Features for prediction were created from combinations of six vital sign measurements and their changes over time.

Setting

A mixed-ward retrospective dataset from the University of California, San Francisco (UCSF) Medical Center (San Francisco, California, USA) as the primary source, an intensive care unit dataset from the Beth Israel Deaconess Medical Center (Boston, Massachusetts, USA) as a transfer-learning source and four additional institutions’ datasets to evaluate generalisability.

Participants

684 443 total encounters, with 90 353 encounters from June 2011 to March 2016 at UCSF.

Interventions

None.

Primary and secondary outcome measures

Area under the receiver operating characteristic (AUROC) curve for detection and prediction of sepsis, severe sepsis and septic shock.

Results

For detection of sepsis and severe sepsis,InSightachieves an AUROC curve of 0.92 (95% CI 0.90 to 0.93) and 0.87 (95% CI 0.86 to 0.88), respectively. Four hours before onset,InSightpredicts septic shock with an AUROC of 0.96 (95% CI 0.94 to 0.98) and severe sepsis with an AUROC of 0.85 (95% CI 0.79 to 0.91).

Conclusions

InSightoutperforms existing sepsis scoring systems in identifying and predicting sepsis, severe sepsis and septic shock. This is the first sepsis screening system to exceed an AUROC of 0.90 using only vital sign inputs.InSightis robust to missing data, can be customised to novel hospital data using a small fraction of site data and retains strong discrimination across all institutions.

Từ khóa


Tài liệu tham khảo

Murphy, 2013, Deaths: final data for 2010, Natl Vital Stat Rep, 61, 1

10.1097/00003246-200107000-00002

10.1097/CCM.0000000000000026

Pfuntner A , Wier LM , Steiner C . Costs for Hospital Stays in the United States, 2010: Statistical Brief #146. In Healthcare Cost and Utilization Project (HCUP) Statistical Briefs: Cost H , Project U , Rockville; MDAgency for Healthcare Research and Quality USA, 2006. https://www.ncbi.nlm.nih.gov/books/NBK121966/.

O’Brien J . The Cost of Sepsis. CDC Safe Healthcare Blog 2015 https://blogs.cdc.gov/safehealthcare/the-cost-of-sepsis/#ref.

10.1097/CCM.0b013e31827c09f8

10.1056/NEJMoa010307

10.1097/01.CCM.0000259463.33848.3D

10.1097/01.CCM.0000217961.75225.E9

Levy, 2001, sccm/esicm/accp/ats/sis international sepsis definitions conference, Intensive Care Med, 2003, 530

10.1001/jama.2016.0287

10.1136/emj.2006.035816

10.1001/jama.1995.03520260039030

10.1007/BF01709751

Johnson, 2016, MIMIC-III, a freely accessible critical care database, Sci Data, 3, 160035, 10.1038/sdata.2016.35

Lowe, 2009, STRIDE--An integrated standards-based translational research informatics platform, AMIA Annu Symp Proc, 2009, 391

PostgreSQL Global Development Group. https://www.postgresql.org/ (5 Oct 2017).

Van Rossum G . The Python Language Reference Manual: Network Theory Ltd, 2003. https://www.python.org/.

10.1126/scitranslmed.aab3719

10.1016/j.resuscitation.2012.01.004

10.1007/s10994-009-5152-4

Calvert, 2016, A computational approach to early sepsis detection, Comput Biol Med, 74, 69, 10.1016/j.compbiomed.2016.05.003

Desautels, 2016, Prediction of Sepsis in the Intensive Care Unit With Minimal Electronic Health Record Data: A Machine Learning Approach, JMIR Med Inform, 4, e28, 10.2196/medinform.5909

Calvert, 2016, High-performance detection and early prediction of septic shock for alcohol-use disorder patients, Ann Med Surg, 8, 50, 10.1016/j.amsu.2016.04.023

10.1093/jamia/ocw014

Calvert, 2016, A computational approach to mortality prediction of alcohol use disorder inpatients, Comput Biol Med, 75, 74, 10.1016/j.compbiomed.2016.05.015

Calvert, 2016, Using electronic health record collected clinical variables to predict medical intensive care unit mortality, Ann Med Surg, 11, 52, 10.1016/j.amsu.2016.09.002

10.1097/MLR.0b013e318268ac86

10.1136/amiajnl-2013-001815

Brause R , Hamker F , Paetz J . Septic shock diagnosis by neural networks and rule based systems. In: Schmitt M , Teodorescu HN , Jain A , eds. et alComputational intelligence techniques in medical diagnosis and prognosis. New York: Springer, 2002:323–56.

Thiel, 2010, Early Prediction of Septic Shock, J. Hosp. Med, 1, 19, 10.1002/jhm.530

Horng, 2017, Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning, PLoS One, 12, 10.1371/journal.pone.0174708

Thông tin
Thông tin xuất bản

BMJ Open

Tập 8 Số 1

e017833

Thông tin tác giả