Study sample and data collection
Questionnaire Survey. A cross-sectional study was carried out between June 2018 and July 2019 in 4 eastern cities (Shanghai, Qingdao, Ningbo and Changzhou), 1 central city (Zhengzhou), and 1western city (Chengdu) of China. Objective Sampling and convenience sampling were performed, and 25 care facilities that accepting the elderly with dementia were selected from the 6 cities. Questionnaire survey was conducted among these 25 care facilities. The questionnaire was developed (Additional file 1) and consisted of the basic information of the care facility, the human resources allocation, the setting of the D-SCUs, and the satisfaction of the elderly with dementia. The satisfaction involves 6 items: living environment, recreational and fitness equipment, dietary conditions, attitude of nursing staff, knowledge and skills of nursing staff, service items. Satisfaction of each item divided into five levels (5 points for very satisfying, 4 points for satisfying, 3 points for neutral, 2 points for dissatisfying, 1 point for very dissatisfying). The contents of the basic information of the care facility, the human resources allocation and the setting of the D-SCUs were filled in by the facility managers. Given the mental state of the elderly with dementia, the items of satisfaction were filled in by their relative who came to see them frequently. Finally, the satisfaction scores of 1046 elderly people with dementia were effectively collected.
Semi-structured in-depth interviews. "Interview Outline for Facility Manager" (Additional file 2) and "Interview Outline for Civil Affair or Medical Insurance Department Manager" (Additional file 3) were developed based on literature review and expert consultation. ①"Interview Outline for Facility Manager" was applied to interview the managers of the 25 facilities surveyed, and the contents involved the management mode, operation status, and care status of the elderly with dementia. ②Three cities (Shanghai, Qingdao, and Chengdu) out of the 6 selected cities have formulated standards for D-SCUs setting at the city level. "Interview Outline for Civil Affair or Medical Insurance Department Manager "was applied to interview the Manager of Civil Affair or Medical Insurance Department of these 3 cities, and the contents involved the setting standards, management, service models of the D-SCUs. Each interview lasted 20 to 40 minutes, and two researchers participated simultaneously, one in charge of the interview and the other in charge of recording.
Analysis
Qualitative analysis. All of the interview data were analyzed by the first author (Xiaoxin Dong) using grounded theory approach[11]. The analysis was done manually. In the first step, close reading of the interview records was done to familiarize and identify the key themes emerging from the interviews. In the second step, a line-by-line coding of the interview records was performed using a three-stage coding process involving open coding, axial coding and selective coding, in order to identify and name concepts and categories and to determine their relationships[12]. The coding and categorization were done by using the Microsoft Office Word.
Quantitative analysis. In quantitative analysis, we followed a two-stage process. First, we took each care facility surveyed as a decision-making unit (DMU), and the efficiency value of each DMU was calculated using DEA method. Second, the care efficiencies were compared by statistical significance tests between care facilities with and without D-SCUs. A similar approach was employed in previous studies, such as Ozcan et al.[13] and Björkgren et al.[14]. Means and Medians were used to indicate the central tendency of the basic information of inputs and outputs and the efficiency values. In view of a normal distribution of efficiency scores cannot be assumed in DEA model[13], the differences between the two groups were tested by the nonparametric Wilcox test of median efficiency levels. Deep2.1 software and SPSS25.0 were applied for DEA analysis and statistical significance tests, respectively, with a significance level of a=0.05.
DEA Model
DEA is a quantitative analysis method that uses linear programming to evaluate the relative effectiveness of comparable DMU based on multiple inputs and multiple outputs.DEA does not require specification of functional form[14], can also handle efficiency evaluation under the conditions of multiple inputs and multiple outputs, which has become a common analysis tool for enterprise resource allocation[13-16]. Since DEA is a cutting-edge analysis technology that uses linear programming to identify production frontiers and is completely driven by data, using DEA to analyze the efficiency of care facilities can avoid the problem of subjectively assigning index weights, minimize arbitrariness, and improve the scientificity of decision-making and the objectivity of evaluation.
The DEA method mainly includes C2R model and B2C model. C2R model is mainly used to evaluate the relative effectiveness of the DMU to determine whether the technology and the scale are effective at the same time. The final comprehensive efficiency value or technical efficiency (TE) is θ, which can be effected by resource utilization and resource allocation. When θ=1, the DMUj technology is valid, indicating that the management model and scale efficiency have reached the best level under the current inputs and outputs. when 0<θ<1, the DMUj technology is invalid. The larger the θ value, the higher the efficiency of the DMUj relative to other DMUs. BC2 model decomposes the TE of DMU into pure technical efficiency (PTE) and scale efficiency (SE). PTE refers to production efficiency that affected by management technology and production technology, and SE generally refers to whether a DMU is operating with optimal production size for producing a defined output. TE= PTE×SE.
Inputs and outputs
In previous DEA studies of nursing homes, researchers tended to choose the number of different types of nursing home staff[15, 17-19], bed number[14, 17, 20], value of fixed assets[14] as input variables, and the resident number[17], the satisfaction of residents[16], the days the resident stay[14, 20, 21], and income[16] as output variables. Based on these studies and the common assumption of labor and capital as the basic inputs for a production function [14], in this DEA models, we used building area, number of open beds and nursing staff as the inputs. The building area and number of open beds was included as proxies for capital, and the number of nursing staff was included as proxy for labor. In terms of outputs, we think that service quality can directly affect the service effects and the development of facilities. Therefore, outputs can be selected from the perspective of residents, and then we choose the number and the care satisfaction of the elderly resident with dementia as outputs in this model.