Dependent variables (technical efficiency score)
The dependent variable was a technical efficiency score calculated using data envelopment analysis (DEA). In this study, an input-oriented Charnes-Cooper-Rhodes model was used according to previous research. In DEA, a decision-making entity (DMU) produces outputs using inputs. In this study, a DMU was defined as a hospital listed in the “Management Strategies of Public Enterprises and Development Status of New Public Hospital Reform Plan.” The inputs were the number of beds, and costs of staff salaries and materials in FY2017, while the outputs were hospital and outpatient revenues in FY2017. These data were obtained from the 2017 Local Public Enterprise Yearbook (LPEY),17 published by the MIC. These items were based on previous studies,13, 18 which used this reference material. Since it takes a certain amount of time for the effects of a reform plan to become evident, the data used to calculate the technical efficiency scores were those from one period later than the data used to define the independent variables. In addition, all inputs and outputs were assumed to be under DMU control.
The technical efficiency score of each public hospital was calculated using DEA SolverPRO™ 15.1 (SAITECH, Holmdel, New Jersey, USA). Technical efficiency ranged from 0 to 1, whereby the greatest hospital technical efficiency was indicated by a score of 1.
Control variables
Nine control variables were collected in this study. These were predominately those shown in previous studies to affect the efficiency of public hospitals. To prevent reversal of causality, these data were the most recently available before 2017, consistent with previous studies.13, 18
(1) Hospital location conditions
This variable indicated whether the hospital was an unprofitable district hospital. An unprofitable district hospital refers to a general hospital that meets the criteria described below. A general hospital is a hospital whose beds are predominately general beds or medical treatment beds, as opposed to a hospital that primarily provides rehabilitation, or a hospital that is primarily a child welfare facility. Among general hospitals, a hospital with fewer than 150 beds and a distance of 15 km or more to the nearest general hospital is defined as a Type 1 unprofitable district hospital. In addition, if the number of hospital beds is fewer than 150 and the population within a radius of 5 km of the hospital is fewer than 30,000 persons according to the latest national government survey, it is defined as a Type 2 unprofitable district hospital. In this study, the variable listed in the 2016 LPEY19 was used: Type 1 unprofitable district hospitals were assigned a dummy variable value of 1 and Type 2 unprofitable district hospitals a value of 2; otherwise a value of 3 was assigned.
This number represented the number of patients per nurse. These data were also obtained from the 2016 LPEY.
(3) Transfer to other accounts ratio
According to the definition of the MIC, a transfer from other accounts is a transfer from accounts such as the general account. It is basically a subsidy from the government. It is used to pay expenses that, due to the nature of the company, are not appropriate to allocate income from management. It is also used to pay expenses that are objectively recognized as difficult for an enterprise to cover via its income only, even if it operates efficiently.20 The variable used was the ratio between other accounts and the current account. The figures listed in the 2016 LPEY were used.
(4) Disaster base hospital designation
Disaster base hospitals are designated by the Ministry of Health, Labor and Welfare as facilities that satisfy the following conditions:211. Emergency response to disasters is possible 24 hours a day, and there is a system that can accept and treat victims in the affected areas. 2. Seriously ill patients can be accepted and transported by helicopter. 3. There is a medical relief team dispatch system that cooperates with the fire department (emergency fire assistance team). 4. In addition to being able to dispatch doctors by helicopter, the hospital is equipped with sufficient medical equipment, medical systems, and information collection systems to support this dispatch, as well as heliports, emergency vehicles, and self-contained equipment that can be dispatched to medical teams. In this study, disaster base hospitals were assigned a dummy variable value of 1 and other hospitals a value of 0.
(5) Clinical training hospital designation
A hospital that conducts clinical training can be designated as a clinical training hospital if it meets several criteria.22 In this study, we assigned a dummy variable value of 1 to facilities that were designated as a clinical training hospital, and 0 otherwise.
(6) Population under 15 years old (in secondary medical zone)
This figure is the population under the age of 15 in the secondary medical zone (a medical zone that can provide medical care related to general hospitalization) to which the hospital belongs. These figures were calculated by the authors with reference to the correspondence table of local governments and secondary medical areas published by the Japan Health Economics Research Organization (JHERO)23 and the 2015 national census.24
(7) Number of hospitals per 1,000 km (in the secondary medical zone)
This figure represents the number of hospital facilities per 1,000 km in the secondary medical zone. These figures were independently calculated by the author with reference to the correspondence table published by the JHERO and the 2016 Healthcare Facility Survey.25
(8) Number of beds per 100,000 population (in secondary medical zone)
This figure is the number of beds per 100,000 people in the secondary medical area. These figures were independently calculated by the author with reference to the correspondence table published by the JHERO and the 2016 Healthcare Facility Survey.
(9) Population density (in secondary medical zone)
This figure represents the population density in the secondary medical area. This figure was independently calculated by the author, referring to the correspondence table published by the JHERO and the 2015 census.
Data analysis
The presence of significant between-groups differences was assessed for each demographic data using t-tests and Mann-Whitney’s U-tests.
DEA was performed using variables collected for inputs and outputs, and a technical efficiency value (θ) was calculated. Since 0 ≦ θ ≦ 1 and were censored data with 0 as the lower limit, the Tobit model was used for both univariate and multivariate analysis, following Chilingerian (1995)26 and Maddala (1983).27
We performed a univariate analysis. A univariate analysis was performed with the existence of the reform plan as the independent variable and as the dependent variable. Subsequently, another univariate analysis was performed in which each of the control variables was the independent variable and was the dependent variable.
Finally, to correct for potential confounding bias, multivariate analysis was performed that adjusting for the control variables described above. In all of the above analyses, a two-tailed significance level of 5% was adopted. STATA 14.0 (State Corporation, Lake Way, Texas, USA) was used to conduct all analyses.