(23) Estimation and Prediction of Length of Stay using Artificial Neural Networks in Rheumatoid Arthritis in a Tertiary Care Hospital

Presenting Author(s)

RD

Riyona D Silva (she/her/hers)

Pharm.D Student
Manipal Academy of Higher Education, Manipal, United States

Author(s)

BH

Bhavna Harikumar, Pharm D (she/her/hers)

Student
Department of Pharmacy Practice, Manipal College of Pharmaceutical Sciences, Manipal Academy of Higher Education, India
RV

Rajesh V, PhD (he/him/his)

Associate Professor
Department of Pharmacy Practice , Manipal College of Pharmaceutical Sciences ,Manipal Academy of Higher Education
Manipal College of Pharmaceutical Sciences
Udupi, Karnataka, India

ABSTRACT:

Background: Rheumatoid arthritis (RA) is a chronic autoimmune disease with articular and extra-articular manifestations resulting in long-term pain, disability, increased socioeconomic burden, and early death. Severe RA is linked to a shorter life expectancy and a higher chance of hospitalization, both of which affect the length of hospital stay (LOS). In terms of calculating the length of stay, the ability of ANNs to seize knowledge through instances rather than rules is a highly intriguing and innovative factor. The proposed method expands the scope of the possibilities of developing a full-scale knowledge-based model for forecasting length of stay.

Objective: The objective of this study is to predict the factors affecting the length of hospital stay in patients diagnosed with RA patients using Artificial Neural Networks (ANN).

Methodology: A post hoc study was conducted using the data from 321 in-patients diagnosed with Rheumatoid Arthritis in the Medicine department in a tertiary care hospital setting. Length of stay was selected as the dependent variable. The independent variables were chosen after performing linear regression. In addition, the independent and dependent variables were considered as input and output layers, respectively. The output layer was the length of stay, while the input layers were the factors affecting length of stay as identified from the regression analysis. In this study, ANN using Multiple Layer Perception (MLP) was used to explore the length of stay. The SPSS statistical software was used for carrying out the statistical analyses (version 20).The best conformity model which calculated the least number of errors and the best approximations were determined in the analysis.

Result: The females made up to 79.8% of the patients, while the males made up 20.2 %. The mean age of the included patients was 53.7±12.3 years. The duration of illness for the majority of the patients ranged from 1 to 6 years. Furthermore, the mean LOS was 6.3 ± 3.1 days. According to this ANN analysis, the variable that had the highest impact on the length of stay was BMI, followed by Age. The presence of infection, the use of DMARDS such as methotrexate, sulphasalazine, leflunomide, and HCQ, and the steroid used were the other variables that showed significant normalized importance on the LOS.

Conclusion: Artificial Neural network has significant advantages over other regression models and can be used as an operative tool for predicting factors influencing LOS, replacing the existing classical and statistical models.