Page 178 - 臺大管理論叢第33卷第1期

The Effect of Applying Clinical Pathways on Medical Resource Utilization among Patients Undergoing a
Thyroidectomy

monitor and review the workflows for the developed CPs.

2.4 Model and Statistical Analysis
Using the CPs developed above, this study compares the differences in medical

resource consumption among the three groups as follows. The first one is the GS group
before the CPs’ implementation (GS-BCP group), the second one is the GS group after
the CPs’ implementation (GS-ACP group), and the third one is the ENT group without the

CPs. The comparisons are conducted in two ways: one involves a comparison between the
first and the second patient group within the GS department, i.e., GS-BCP vs. GS-ACP w/
and w/o CPs, respectively, and the other is that between the second and the third group of
the GS-ACP and ENT departments, w/ and w/o implementing the CPs, respectively. The
factors including comorbidities, and surgeon-related and disease-specific characteristics

are also incorporated into the model to control for their possible effects on medical
resource utilization.
Given the retrospective and observational nature of our study, there is a potential

risk of selection bias and confounding variables such as the provider’s behaviors that
may influence the significance of our findings. To address this endogeneity issue, we
thus estimate the impact of CPs on the LOS and medical care expenses of thyroidectomy
patients using the Difference-in-differences (DID) method (Athey and Imbens, 2017;
Wooldridge, 2012). To compare the mean outcomes for the treated and controlled

variables before and after the CPs’ implementation, the DID analysis requires data on
the outcomes in the group that receives the CPs’ implementation (GS patients) and the
group (ENT patients) that does not, both before and after the implementation. For this

purpose, in addition to the GS-BCP and GS-ACP referred to above, the ENT patients
are also divided into two groups, ENT-BCP and ENT-ACP, before and after the CPs’
implementation, respectively. The effect of the CPs on medical resource utilization, i.e.,
the DID, is estimated from a regression with terms for the group, period, group-by-period
interaction and other control variables. A dummy variable, i.e., Treatment, is employed

to distinguish GS patients (Treatment = 1) from the comparison group of ENT patients
(Treatment = 0). The coefficient of the Treatment dummy represents the differences in
medical resource utilization between the GS and ENT patients. Another dummy variable,

i.e., the CPs, is introduced to denote the time period during which the CPs are in place (July

170