Machine learning approaches to predicting no-shows in pediatric medical appointment

Dianbo Liu; Won-Yong Shin; Eli Sprecher; Kathleen Conroy; Omar Santiago; Gal Wachtel; Mauricio Santillana

doi:10.1038/s41746-022-00594-w

Machine learning approaches to predicting no-shows in pediatric medical appointment

NPJ Digit Med. 2022 Apr 20;5(1):50. doi: 10.1038/s41746-022-00594-w.

Authors

Dianbo Liu^{1

2}, Won-Yong Shin³, Eli Sprecher⁴, Kathleen Conroy⁴, Omar Santiago⁴, Gal Wachtel^{4

5}, Mauricio Santillana^{6

7

8}

Affiliations

¹ Boston Children's Hospital, Boston, MA, USA. dianbo@mit.edu.
² Department of Pediatrics, Harvard Medical School, Boston, MA, USA. dianbo@mit.edu.
³ School of Mathematics and Computing (Computational Science and Engineering), Yonsei University, Seoul, Republic of Korea.
⁴ Boston Children's Hospital, Boston, MA, USA.
⁵ Department of Pediatrics, Harvard Medical School, Boston, MA, USA.
⁶ Boston Children's Hospital, Boston, MA, USA. msantill@g.harvard.edu.
⁷ Department of Pediatrics, Harvard Medical School, Boston, MA, USA. msantill@g.harvard.edu.
⁸ Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA. msantill@g.harvard.edu.

Abstract

Patients' no-shows, scheduled but unattended medical appointments, have a direct negative impact on patients' health, due to discontinuity of treatment and late presentation to care. They also lead to inefficient use of medical resources in hospitals and clinics. The ability to predict a likely no-show in advance could enable the design and implementation of interventions to reduce the risk of it happening, thus improving patients' care and clinical resource allocation. In this study, we develop a new interpretable deep learning-based approach for predicting the risk of no-shows at the time when a medical appointment is first scheduled. The retrospective study was conducted in an academic pediatric teaching hospital with a 20% no-show rate. Our approach tackles several challenges in the design of a predictive model by (1) adopting a data imputation method for patients with missing information in their records (77% of the population), (2) exploiting local weather information to improve predictive accuracy, and (3) developing an interpretable approach that explains how a prediction is made for each individual patient. Our proposed neural network-based and logistic regression-based methods outperformed persistence baselines. In an unobserved set of patients, our method correctly identified 83% of no-shows at the time of scheduling and led to a false alert rate less than 17%. Our method is capable of producing meaningful predictions even when some information in a patient's records is missing. We find that patients' past no-show record is the strongest predictor. Finally, we discuss several potential interventions to reduce no-shows, such as scheduling appointments of high-risk patients at off-peak times, which can serve as starting point for further studies on no-show interventions.