The impact of healthcare reform on the dynamic changes in health service utilization and equity: a 10-year follow-up study | Scientific Reports – Nature.com

Study design and sample

Data were obtained mainly from the project ‘Innovating payment system and improving health benefits’, which was jointly carried out by Harvard University, Oxford University, Fudan University and Ningxia Medical University (2009, 2011 and 2012). The data from 2015 and 2019 were extracted from the National Natural Science Foundation of China (from a follow-up study of project). Questionnaires were completed and verified by trained teachers and graduate students. The survey was answered by one adult per household who provided information on all members.

A multi-stage stratified cluster randomized design was used to obtain a representative sample from each county. We selected a total of five counties in Ningxia Province, including two project counties (Haiyuan and Yanchi) and three control counties (Tongxin, Pengyang and Xiji). In each county, all the villages were divided into three economic levels, 40% of the sample villages were selected. Then, using the household head roster, 33 households (20 households in the control counties) in each village were selected by systematic sampling. Members of the sample households who had been living there for more than 6 months were selected as respondents. The survey conducted in the baseline year (2009) included 30,384 observations, and there were 28,886, 30,583, 28,697, and 23,821 observations in follow-up years 2011, 2012, 2015, and 2019, respectively . Sample data for Tongxin county is missing in 2019.

Concentration index

Equity in inpatient and outpatient utilization: The concentration index (CI) was used to measure equity in healthcare use. The CI is one of the most commonly used indexes to measure health service equity.

The specific steps of calculating the concentration index are as follows: ① sorting by economic income group and giving the relevant rank X (0–1); ② calculating the health level or disease prevalence rate h of each economic income group according to economic income; ③ calculating the average level M of health or disease of the whole population (such as 2-week prevalence rate and hospitalization rate in the past year); ④ calculating the correlation rank X and each economic income group (5) calculation of concentration index16:

$${text{CI }} = {text{ 2cov }}left( {{text{X}},{text{ H}}} right)/{text{M}}$$

$${text{cov }}left( {{text{X}},{text{ H}}} right) , = {text{E}}left( {{text{XH}}} right) , – {text{ E}}left( {text{X}} right){text{ E}}left( {text{H}} right)$$

X is the relevant rank given by economic income grouping, H is the health level or disease prevalence calculated by different economic income grouping, and M is the average level of health or disease of the whole population (such as 2-week prevalence and hospitalization rate in the past year). CI was between − 1 and 1. For health CI, “CI = − 1” means that health is concentrated in the hands of the people with the lowest socio-economic status group, “CI =  + 1” means that health is concentrated in the hands of the people with the highest socio-economic status group; for disease CI, if it is negative, it means that disease is concentrated in the hands of the lower socio-economic status group, and disease CI is positive, it means that disease is concentrated in the hands of the higher socio-economic status group. When CI tends to 0, it indicates that the distribution of health status or diseases is fair.

Statistical analysis

In this study, the sample data were checked for missing data and outliers and were cleaned prior to data analysis. We used scatter plots and Spearman’s rank correlation coefficient to assess the association between the disposable income of rural residents per capita or their educational level (represented as the proportion of people with an education level of senior high school or higher) and the chronic illness prevalence rate or hospital admission rate.

In the evaluation of health equity, socio-economic grouping is commonly used. Education level and income level are the main indicators to reflect the socio-economic situation1. We constructed multivariate random-effects generalized least squares regression models with our panel data to test whether the rate of receiving a medical consultation in the last 2 weeks or the rate of hospital admission was associated with education level, rural disposable income per capita, or the chronic illness prevalence rate. In those models, the group variable was county, and the time variable was year. The SE were adjusted for the clustering at the province level. The dependent variables were the rate of receiving a medical consultation in the last 2 weeks and the rate of hospital admission. The three independent variables were defined as follows: the percentage of individuals who had attended senior secondary school, which is the proportion of people with a senior secondary education level; per capita disposable income, which is the income that residents can freely spend; and the prevalence of chronic disease, which is the ratio of the number of individuals with a chronic illness surveyed in the first half of the year to the total number of people surveyed. The linear models included county data for 2009, 2011, 2012, 2015 and 2019. Tongxin County was excluded from the models for 2019 because of missing data. Our threshold for statistical significance was 0.05. We used Stata Version 14.0 for the analysis.

Ethics approval and consent to participate

Ethical approval was granted (the Ethics Committee of Ningxia Medical University, Approval number, No. 2018-114). All methods were carried out in accordance with the relevant standards and regulations. All the residents (including all members of the family) of the sample households were invited to participate in the investigation, and the questionnaire was answered by the head of the household or another adult in the household. Informed consent was signed by the head of the household or an adult in the family. Each family only needed to sign one informed consent form.