Skip to main content

Table 4 Correlates of outcome quality with limited control for case-mix, general-acute hospitals

From: Comparing public and private hospitals in China: Evidence from Guangdong

  Inpatient mortality rate
  Model 1 Model 2
Non-government non-profit (indicator variable) 0.003
(0.48)
-0.005
(0.9)
Private for-profit hospital (indicator variable) -0.012 -0.002
  (1.92) (0.34)
ln (beds) 0.003 -0.001
  (1.3) (0.34)
ln (emergency patients) -0.005 -0.001
  (3.67)** (1.2)
Contract with social insurance ("appointed" indicator variable) 0.012
(2.50)*
0.014
(3.33)**
University hospital (indicator variable) -0.006 -0.012
  (0.64) (1.31)
Rural hospital (indicator variable) -0.004 -0.002
  (0.49) (0.33)
Internal medicine visits as % of op visits   0.02
   (2.03)*
Surgery visits as % of op visits   -0.028
   (1.55)
Ob/gyn visits as % of op visits   0.011
   (0.42)
Pediatrics visits as % of op visits   0.041
   (1.33)
TCM visits as % of op visits   0.044
   (3.58)**
Internal medicine ip beds as % of beds   0.054
   (5.36)**
Surgery ip beds as % of beds   -0.027
   (2.40)*
Pediatrics ip beds as % of beds   -0.05
   (1.4)
Obgyn ip beds as % of beds   -0.045
   (2.49)*
Psychiatry dept ip beds as % of beds   -0.013
   (0.34)
Infectious disease dept ip beds as % of beds   -0.041
   (0.83)
Tumors ip beds as % of beds   0.292
   (4.11)**
TCM ip beds as % of beds   -0.007
   (0.92)
Hospital accreditation level 2 (indicator variable)   0.006
(1.04)
Hospital accreditation level 3 (highest; indicator variable)   0.002
(0.21)
Not classified into a level (indicator variable)   0.004
   (0.79)
Constant 0.044 0.019
  (3.54)** (1.24)
Observations 265 265
R-squared 0.12 0.44
  1. Note:Absolute value of t statistics in parentheses
  2. All dependent variables are analyzed in ln (.) form
  3. op = outpatient; ip = inpatient
  4. * significant at 5%; ** significant at 1%
  5. Model 1 and Model 2 are separate OLS regressions, with the explanatory variables as listed in the first column; each cell of the table reports the estimated regression coefficient, with the absolute value of the associated t statistic in parentheses below the estimated coefficient. A blank cell for Model 1 indicates the row variables (e.g., "Internal medicine visits as % of op visits") that were not included in the Model 1 regression. We take the natural logarithm of the dependent variable (e.g., ln (inpatient mortality rate)) and use that converted form for the econometric analysis to avoid an artificial bias in the statistical results from the skewed distribution of mortality rates.