Spatiotemporal characteristics of health poverty and its associates among middle-aged and elderly populations in China: a CHARLS and statistical data analysis

  Objective  To explore spatiotemporal non-stationarity in health poverty prevalence and its influencing factors among middle-aged and elderly residents in China for providing evidence to promote comprehensive and effective management on health poverty alleviation in the populations.

  Methods  The data of the analysis were extracted from four rounds of China Health and Retirement Longitudinal Survey (CHARLS) conducted in 2011, 2013, 2015, and 2018 across China, which collected the information on 37 296 households with family members aged ≥ 45 years. Other relevant data were extracted from the China Statistical Yearbook of 2012, 2014, 2016, and 2019, the China Health and Family Planning Statistical Yearbook of 2012 and 2016, the China Tertiary Industry Statistical Yearbook of 2014 and 2019, and from the Global Burden of Disease Network in 2019. Based on the self-developed indexes for measurement of multidimensional health poverty (MHP), we used the ordinary least squares model (OLS), the geographically weighted regression model (GWR), the temporally weighted regression model (TWR), and the geographically and temporally weighted regression model (GTWR) to model the impacts of influencing factors on health poverty among the middle-aged and elderly residents during the period. The spatiotemporal modes of the estimation coefficients in the optimal model were further described.

  Results  After adjusting for fixed effects of year and region, the global OLS regression results showed that the MHP index increased with the increment of the prevalence of non-communicable diseases (NCDs) and disabilities, gross domestic production (GDP) per capita, average out-of-pocket medical expenditure, and mean annual concentration of particulate matter less than 2.5 μm in aerodynamic diameter (PM_2.5) in previous 3 years; but the index decreased with the increment of population density, the coverage of basic medical insurance, and the number of nurses per 10 000 population. Among all partial regression models constructed, GWR fitted the survey data best, with the increased R² from 70.57% to 88.68% and the decreased sum of squares of the residuals from 797.791 to 585.277 compared with the global OLS regression model. The impacts of the eight covariates mentioned above on MHP index were non-stationary based on the results of GWR modeling; the aggravating impact of NCDs prevalence rate on MHP index was much stronger for Inner Mongolia Autonomous Region (Inner Mongolia) and the four provinces including Hebei, Shanxi, Yunnan, and Guizhou; the aggravating impact of disability prevalence rate on MHP index was much more stronger for Jiangsu, Anhui, and Henan province; the alleviating impact of population density on MHP index was stronger for the five provinces of Heilongjiang, Jilin, Liaoning, Shandong, and Shanxi and much more stronger for Inner Mongolia, Gansu and Hebei provinces, and Beijing municipality; both the alleviating impact of the coverage rate of basic medical insurance and the number of nurses versus 10 000 population and the aggravating impact of GDP per capita on MHP index were the strongest for Xinjiang Uygur Autonomous Region, Qinghai province, and Gansu province; the aggravating impact of average out-of-pocket medical expenditure on MHP index was much more stronger for Henan, Anhui and Jiangsu province; the aggravating impact of mean annual PM_2.5 concentration in previous 3 years on MHP index attenuated gradually from a central region including Beijing municipality, Hebei province and Inner Mongolia to the surrounding area of whole Bohai rim region.

  Conclusion  There is a spatial non-stationarity for the impacts of multiple factors on health poverty among middle-aged and elderly residents in China. Partial weighted regression analysis could be adopted to examine the non-stationarity in studies on improving administrative efficiency of regional health poverty reduction.