Visualizing Drivers of Inpatient Admissions Using Wavelet Analysis

1University of California‐San Diego, San Diego, CA

Meeting: Hospital Medicine 2008, April 3-5, San Diego, Calif.

Abstract number: 11

Background:

Although inpatient admissions to hospitals appear to be a random process, identifying and understanding any underlying patterns and drivers could have a significant impact on medical staffing and organizational planning.

Methods:

The number of daily admissions to the medical service of an academic medical center was analyzed using the wavelet transform, a method of deconstructing signals as a compilation of stretched and shifted versions of a basic pattern, the mother wavelet. For any given day, the relative strength of all underlying periodicities can be computed and displayed as a topographic map. Patterns whose strength were greater than that expected for a purely random process were extracted for further deterministic analysis.

Results:

The wavelet transform identified locally significant trends of 2‐, 4‐, 7‐, and 30‐day periodicities/delays throughout the admission data set. A 2‐day delay was seen immediately after Martin Luther King Day, suggesting a postholiday effect on admissions. The 7‐ and 30‐day cycles stimulated exploration of day‐of‐week and time‐of‐month effects on admissions. Indeed, in a univariate analysis, Sunday, Monday, Tuesday, Thursday, Saturday, ends of the month, and day after holiday were significantly correlated with inpatient admissions. A simple multivariate linear model based entirely on binary variables could explain 26% of the variance in patient admissions.

Conclusions:

Wavelet analysis systematically uncovered the effects of societal schedules on patient behavior. It is a powerful and robust tool for understanding some of the forces driving complex processes.

Author Disclosure:

T. Chau, none.

Figure 1. (a) Plot of daily admissions by date. (b) Wavelet power spectrum using the Morlet 6 mother wavelet. Hatched region is the cone of influence, which is susceptible to edge effects. Black contours denote the 5% significance level compared to a random white‐noise background spectrum. Courtesy of http://www.researchsystems.com

To cite this abstract:

Chau T. Visualizing Drivers of Inpatient Admissions Using Wavelet Analysis. Abstract published at Hospital Medicine 2008, April 3-5, San Diego, Calif. Abstract 11. Journal of Hospital Medicine. 2008; 3 (suppl 1). https://www.shmabstracts.com/abstract/visualizing-drivers-of-inpatient-admissions-using-wavelet-analysis/. Accessed May 26, 2019.

« Back to Hospital Medicine 2008, April 3-5, San Diego, Calif.