What Determines the Medical Expenses in China?–Evidence from Outpatients Case - Innovative Management in Information and Production - page 323

The rest of research is organized as follows. Section 2 establishes the
semi-parametric regression approach to confirm determinants and presents data
description. Section 3 identifies the decisive and control variables of average
medical expenses for outpatients. Section 4 draws the conclusion.

2 Model And Data

2.1 Model Specification

The research employs time-varying semi-parametric coefficient model ( Fan and
Huang 2005 ) to explore the determinants of average medical expenses for both
outpatients and inpatients in China. It is known to all that average drug expense
and medical care expense are two major parts of average medical expense; the
research established the initial semi-parametric model with all possible determinants
as follows:

Y i = a ( t i ) x 1 i + b ( t i ) x 2 i + g ( z 1 i , z 2 i ,..., z qi ) , i = 1 , 2 ,..., n , (1)

where Y is average medical expenses, x 1 is average drug expense, x 2 is medical care
expense, z 1 , z 2 i ,..., z q refer to all possible control variables, and ε is a random error

series with zero mean.
The relationships between z 1 , z q and Y are unknown and remain to be

confirmed as irrelevant variables and determinants of average medical expenses.
According to difference in the impact mechanism, the determinants are divided
into decisive variables and control variables. Decisive variables refer to the relevant
variables that exert direct linear impacts on average medical expenses while control
variables refer to the ones who exert indirect nonlinear impacts. Take the Taylor
expansion of a

z 2 ,...,

( t i ) , b ( t i ) at constant t 0 and asymptotically obtain

a ( t i )= a 1 + a 2 ( t i − t 0 ) , b ( t i )= b 1 + b 2 ( t i − t 0 ) . (2)

b 1 b 2

Estimate ( a 1 , , a 2 , ) ,then

b

= ( ) + ( ) + ( ) .

y i a t i x 1 i t i x 2 i g z i (3)

b

Let y i =

− ( ) + ( )

y i a t i x 1 i t i x 2 i , then Eq. ( 3 )comesto

y i =

( ) .

g z i (4)

The determinants of average medical expenses are searched by 6 steps.
Step 1. Apply LSCV approach to obtain the bandwidths of all possible determi-
nants in nonparametric regression model ( 4 ).