×

A high-fidelity model to predict length of stay in the neonatal intensive care unit. (English) Zbl 07549372

Summary: Having an interpretable, dynamic length-of-stay model can help hospital administrators and clinicians make better decisions and improve the quality of care. The widespread implementation of electronic medical record (EMR) systems has enabled hospitals to collect massive amounts of health data. However, how to integrate this deluge of data into healthcare operations remains unclear. We propose a framework grounded in established clinical knowledge to model patients’ lengths of stay. In particular, we impose expert knowledge when grouping raw clinical data into medically meaningful variables that summarize patients’ health trajectories. We use dynamic, predictive models to output patients’ remaining lengths of stay, future discharges, and census probability distributions based on their health trajectories up to the current stay. Evaluated with large-scale EMR data, the dynamic model significantly improves predictive power over the performance of any model in previous literature and remains medically interpretable.
Summary of contribution: The widespread implementation of electronic health systems has created opportunities and challenges to best utilize mounting clinical data for healthcare operations. In this study, we propose a new approach that integrates clinical analysis in generating variables and implementations of computational methods. This approach allows our model to remain interpretable to the medical professionals while being accurate. We believe our study has broader relevance to researchers and practitioners of healthcare operations.

MSC:

90-XX Operations research, mathematical programming
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] Aczon M, Ledbetter D, Ho L, Gunny A, Flynn A, Williams J, Wetzel R (2017) Dynamic mortality risk predictions in pediatric critical care using recurrent neural networks. Preprint, submitted January 23, http://arxiv.org/abs/1701.06675.Google Scholar
[2] Ansari SF, Yan H, Zou J, Worth RM, Barbaro NM (2018) Hospital length of stay and readmission rate for neurosurgical patients. Neurosurgery 82(2):173-181.Crossref, Google Scholar · doi:10.1093/neuros/nyx160
[3] Anthony Celi L, Mark RG, Stone DJ, Montgomery RA (2013) “Big Data” in the intensive care unit: Closing the data loop. Amer. J. Respiratory Critical Care Medicine 187(11):1157-1160.Crossref, Google Scholar · doi:10.1164/rccm.201212-2311ED
[4] Basques BA, Webb ML, Bohl DD, Golinvaux NS, Grauer JN (2015) Adverse events, length of stay, and readmission after surgery for tibial plateau fractures. J. Orthopaedic Trauma 29(3):e121-e126.Crossref, Google Scholar · doi:10.1097/BOT.0000000000000231
[5] Bender J, Koestler D, Ombao H, McCourt M, Alskinis B, Rubin LP, Padbury JF (2013) Neonatal intensive care unit: Predictive models for length of stay. J. Perinatology: Official J. California Perinatal Assoc. 33(2):147-153.Crossref, Google Scholar · doi:10.1038/jp.2012.62
[6] Chaou CH, Chen HH, Chang SH, Tang P, Pan SL, Yen AMF, Chiu TF (2017) Predicting length of stay among patients discharged from the emergency department—Using an accelerated failure time model. Plos One 12(1):e0165756.Crossref, Google Scholar · doi:10.1371/journal.pone.0165756
[7] Chen T, Guestrin C (2016) XGBoost: A scalable tree boosting system. Proc. 22nd ACM SIGKDD Internat. Conf. Knowledge Discovery Data Mining KDD’16 (Association for Computing Machinery, New York), 785-794. Google Scholar
[8] Chipman HA, George EI, McCulloch RE (2010) BART: Bayesian additive regression trees. Ann. Appl. Statist. 4(1):266-298.Crossref, Google Scholar · Zbl 1189.62066 · doi:10.1214/09-AOAS285
[9] Collett D (2015) Modelling Survival Data in Medical Research, 3rd ed. (Chapman and Hall/CRC, New York).Crossref, Google Scholar · doi:10.1201/b18041
[10] Dobson G, Lee HH, Pinker E (2010) A model of ICU bumping. Oper. Res. 58(6):1564-1576.Link, Google Scholar · Zbl 1238.90083
[11] Fang HB, Wu TT, Rapoport AP, Tan M (2016) Survival analysis with functional covariates for partial follow-up studies. Statist. Methods Medical Res. 25(6):2405-2419.Crossref, Google Scholar · doi:10.1177/0962280214523586
[12] Ghassemi M, Celi LA, Stone DJ (2015a) State of the art review: The data revolution in critical care. Critical Care 19(1):118.Crossref, Google Scholar · doi:10.1186/s13054-015-0801-4
[13] Ghassemi M, Pimentel MA, Naumann T, Brennan T, Clifton DA, Szolovits P, Feng M (2015b) A multivariate timeseries modeling approach to severity of illness assessment and forecasting in ICU with sparse, heterogeneous clinical data. Proc. AAAI Conf. Artificial Intelligence, 446-453.Google Scholar
[14] Gomella TL, Cunningham MD, Eyal F (2013) Neonatology, 7th ed. (McGraw-Hill Education/Medical, New York).Google Scholar
[15] Harsha SS, Archana BR (2015) SNAPPE-II (score for neonatal acute physiology with perinatal extension-II) in predicting mortality and morbidity in NICU. J. Clinical Diagnostic Res. 9(10):SC10-SC12.Google Scholar
[16] Hauck K, Zhao X (2011) How dangerous is a day in hospital? A model of adverse events and length of stay for medical inpatients. Medical Care 49(12):1068-1075.Crossref, Google Scholar · doi:10.1097/MLR.0b013e31822efb09
[17] Hoogervorst-Schilp J, Langelaan M, Spreeuwenberg P, de Bruijne MC, Wagner C (2015) Excess length of stay and economic consequences of adverse events in Dutch hospital patients. BMC Health Services Res. 15(1):531.Crossref, Google Scholar · doi:10.1186/s12913-015-1205-5
[18] Ishwaran H, Kogalur UB (2018) randomForestSRC: Random forests for survival, regression, and classification (RF-SRC). Accessed November 28, 2018, https://CRAN.R-project.org/package=randomForestSRC.Google Scholar
[19] Ishwaran H, Kogalur UB, Blackstone EH, Lauer MS (2008) Random survival forests. Ann. Appl. Statist. 2(3):841-860.Crossref, Google Scholar · Zbl 1149.62331 · doi:10.1214/08-AOAS169
[20] Kaboli PJ, Go JT, Hockenberry J, Glasgow JM, Johnson SR, Rosenthal GE, Jones MP, Vaughan-Sarrazin M (2012) Associations between reduced hospital length of stay and 30-day readmission rate and mortality: 14-year experience in 129 Veterans Affairs hospitals. Ann. Internal Medicine 157(12):837-845.Crossref, Google Scholar · doi:10.7326/0003-4819-157-12-201212180-00003
[21] Kim SH, Chan CW, Olivares M, Escobar G (2014) ICU admission control: An empirical study of capacity allocation and its implication for patient outcomes. Management Sci. 61(1):19-38.Link, Google Scholar
[22] Lee HC, Bennett MV, Schulman J, Gould JB, Profit J (2016) Estimating length of stay by patient type in the neonatal intensive care unit. Amer. J. Perinatology 33(8):751-757.Crossref, Google Scholar · doi:10.1055/s-0036-1572433
[23] Louppe G (2014) Understanding random forests: From theory to practice. Unpublished PhD thesis, Université de Liège, Liège, Belgique.Google Scholar
[24] Lowsky DJ, Ding Y, Lee DKK, McCulloch CE, Ross LF, Thistlethwaite JR, Zenios SA (2013) A K-nearest neighbors survival probability prediction method. Statist. Medicine 32(12):2062-2069.Crossref, Google Scholar · doi:10.1002/sim.5673
[25] Ma J, Lee DKK, Perkins ME, Pisani MA, Pinker E (2019) Using the shapes of clinical data trajectories to predict mortality in ICUs. Critical Care Explorations 1(4):e0010.Crossref, Google Scholar · doi:10.1097/CCE.0000000000000010
[26] Meira-Machado L, de Uña-Álvarez J, Cadarso-Suárez C, Andersen PK (2009) Multi-state models for the analysis of time-to-event data. Statist. Methods Medical Res. 18(2):195-222.Crossref, Google Scholar · doi:10.1177/0962280208092301
[27] Pallin DJ, Allen MB, Espinola JA, Camargo CA, Bohan JS (2013) Population aging and emergency departments: Visits will not increase, lengths-of-stay and hospitalizations will. Health Affairs (Project Hope) 32(7):1306-1312.Crossref, Google Scholar · doi:10.1377/hlthaff.2012.0951
[28] Pollack MM, Holubkov R, Reeder R, Dean JM, Meert KL, Berg RA, Newth CJL, et al. (2018) PICU length of stay: Factors associated with bed utilization and development of a benchmarking model. Pediatric Critical Care Medicine 19(3):196-203.Crossref, Google Scholar · doi:10.1097/PCC.0000000000001425
[29] Probst P, Wright MN, Boulesteix AL (2019) Hyperparameters and tuning strategies for random forest. WIREs Data Mining Knowledge Discovery 9(3):e1301.Crossref, Google Scholar · doi:10.1002/widm.1301
[30] Rajkomar A, Oren E, Chen K, Dai AM, Hajaj N, Hardt M, Liu PJ, et al. (2018) Scalable and accurate deep learning with electronic health records. NPJ Digital Medicine 1(1):18.Crossref, Google Scholar · doi:10.1038/s41746-018-0029-1
[31] Richardson DK, Corcoran JD, Escobar GJ, Lee SK (2001) SNAP-II and SNAPPE-II: Simplified newborn illness severity and mortality risk scores. J. Pediatrics 138(1):92-100.Crossref, Google Scholar · doi:10.1067/mpd.2001.109608
[32] Rinne ST, Graves MC, Bastian LA, Lindenauer PK, Wong ES, Hebert PL, Liu CF (2017) Association between length of stay and readmission for COPD. Amer. J. Managed Care 23(8):e253-e258.Google Scholar
[33] Romero-Brufau S, Huddleston JM, Escobar GJ, Liebow M (2015) Why the C-statistic is not informative to evaluate early warning scores and what metrics to use. Critical Care 19(1):285.Crossref, Google Scholar · doi:10.1186/s13054-015-0999-1
[34] Rosenberg MA, Browne MJ (2001) The impact of the inpatient prospective payment system and diagnosis-related groups. North Amer. Actuarial J. 5(4):84-94.Crossref, Google Scholar · Zbl 1083.62542 · doi:10.1080/10920277.2001.10596020
[35] Sanchez-Pinto LN, Luo Y, Churpek MM (2018) Big data and data science in critical care. Chest 154(5):1239-1248.Crossref, Google Scholar · doi:10.1016/j.chest.2018.04.037
[36] Shmueli A, Sprung CL, Kaplan EH (2003) Optimizing admissions to an intensive care unit. Healthcare Management Sci. 6(3):131-136.Crossref, Google Scholar · doi:10.1023/A:1024457800682
[37] Verburg IWM, Atashi A, Eslami S, Holman R, Abu-Hanna A, de Jonge E, Peek N, de Keizer NF (2017) Which models can I use to predict adult ICU length of stay? A systematic review. Critical Care Medicine 45(2):e222-e231.Crossref, Google Scholar · doi:10.1097/CCM.0000000000002054
[38] Verghese A, Shah NH, Harrington RA (2018) What this computer needs is a physician: Humanism and artificial intelligence. JAMA 319(1):19-20.Crossref, Google Scholar · doi:10.1001/jama.2017.19198
[39] Verma A, Okun NB, Maguire TO, Mitchell BF (1999) Morbidity assessment index for newborns: A composite tool for measuring newborn health. Amer. J. Obstetrics Gynecology 181(3):701-708.Crossref, Google Scholar · doi:10.1016/S0002-9378(99)70516-8
[40] Witten DM, Tibshirani R (2010) Survival analysis with high-dimensional covariates. Statist. Methods Medical Res. 19(1):29-51.Crossref, Google Scholar · doi:10.1177/0962280209105024
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.