Construction of an index system for big data application in health care enterprises --A study based on the Delphi method

This study explored the theory of medical enterprise management and big data. Based on the Delphi method, two rounds of expert opinions were consulted on the capability of a health care enterprise big data application index system covering 11 dimensions, 46 first-level indicators and 111 second-level indicators. The index system includes two categories of input and capacity. The input category includes five dimensions: human resources, material resources, financial resources, government policies, and social service system or social environment. The capabilities category includes six dimensions: data integration capabilities, service capabilities, data analysis capabilities, information security, profitability, and innovation capabilities. This index system aims to appraise the application capability of big data scientifically and systematically, and fills the gaps of such research in China so far, which has positive significance for further research in the future.

1. Scientific principle: The indicator conforms to the criteria of objective facts and has a scientific theoretical basis. The definition of each indicator is clear and precise, and the specific calculation formula is expressed for indicators that are prone to ambiguity.
2. Comprehensive principle: Through systematic investigation and demonstration, all indicators that measure the big data application capability are covered as much as possible.
3. Principle of operability: The index system should reflect the direction of big data development and have guiding significance for health care enterprises to achieve further big data application. Therefore, the indicators should be universally defined. The meaning of the indicators should be observable, measurable, and operational.

Indicator screening method
Each indicator was rated on a scale of 1 to 5 where 1 represents very unimportant and 5 represents very important. Scores for each dimension and indicator were summed. The experts' evaluation of each indicator was measured using three statistical measures: the percentage of experts who held 'very important' and 'important' opinions, the mean score, and the coefficient of variation. An indicator was deleted if the percentage of experts that felt it was important was less than 75%, the mean score was less than 4.0, and the coefficient of variation was greater than 1.0.
If only one or two of these statistical measures were met, the retention of the indicator was determined after discussion by the research team.

Characteristics of Experts
This study pre-selected 17 experts and received feedback from 16. The 16 experts were senior teachers from various universities in Guizhou province, heads of big data companies, or persons from relevant departments. They all had systematic and unique insights in the area of big data. The basic information of the 16 experts is shown in Table 1.

Results and discussion
First round of inquiry

Questionnaire setting and distribution
The purpose of the first round of questionnaires was to ask experts to comment on the importance of the researcher's initial setting of the indicators for big data application capability. The questionnaire asked for evaluation of each dimension, the first-level indicators, and the second-level indicators in the index. In each part, the items of modification opinions and other indicators were set up, and experts were requested to propose amendments.
The first round of questionnaires was issued to 17 experts, of which 16 were recovered resulting in a recovery rate of 94%.

Reliability
Chronbach's alpha values of the whole system and for the first and second levels were 0.96, 0.85, and 0.96, respectively, indicating that the reliability of the index system was quite high. Table 2 shows the importance percentage for each dimension and indicator selected by the 16 experts, together with the mean, standard deviation, and coefficient of variation. As shown in Table 2, the average score of the eight dimensions was above 4.0 (important), the important percentage was higher than 90%, and the coefficient of variation was less than 1.0 indicating that the experts believed these eight dimensions were highly and consistently important.  Table 3 shows that in the COSTS category, the mean of 'Personnel' and 'Other

Statistical Analysis
Costs' was less than 4.0. In the capabilities category, the mean of 'Service Coverage', 'Government Big Data Subsidy Income', and 'Other Income' were below 4.0. The percentage of experts who felt that these five indicators were important was less than 75. Apart from '3 Information Security', the standard deviation of other first-level indicators was less than 1.0. The coefficient of variation for all indicators was less than 1.0.  Indicator screening results -first round 1

. Deleting
There were no indicators that met the criteria for deleting proposed by the research team, so no indicator was deleted.

Adding
In In terms of the first-level indicators, one expert suggested adding 'Synergy' in

Reliability
The Chronbach Alpha values of the index system, the first level indicator, and the second level index were 0.98, 0.90, and 0.96 in the second round, indicating that the reliability of the index system was high.