SEARCH

SEARCH BY CITATION

Keywords:

  • Life expectancy;
  • socioeconomic determinants;
  • stepwise regression model;
  • Beijing;
  • China

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Objective To explore the socioeconomic factors associated with life expectancy in Beijing, and to predict future trends.

Methods The linear stepwise regression model was used to construct the relationship between life expectancy and socioeconomic factors.

Results The model showed that there were four factors associated with life expectancy in Beijing. Floor space available per rural resident (P = 0.000) and GDP per capita (P = 0.022) correlated positively with life expectancy, while the rural population proportion (P = 0.010) and illiteracy rate (P = 0.001) correlated negatively with life expectancy.

Conclusion There is a close relationship between life expectancy and socioeconomic factors. The constructed model can be used as a rapid tool to project life expectancy in Beijing. It is possible to improve life expectancy continuously with sustained development of socioeconomic conditions in Beijing, China.

inline image

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Health is the basis for overall human development, while longer, high-quality life is a goal pursued by everyone. Especially concerning socioeconomic development, living longer and having healthier conditions has become a common aspiration shared by all. As shown in large-scale studies, several factors were associated with overall health status. such as the level of socioeconomic development, education, culture, environment, and lifestyle (1–3). The World Health Organization (WHO) launched a special commission in 2005 focusing on the social determinants of health. The commission has yielded a series of research reports of high quality, which have exerted positive impacts on the member states in the past three years. So far in China, very little domestic research has been done on social determinants of health. The vital statistical information system in Beijing, China, is very well developed. The Municipality of Beijing has published a vital report annually since the 1980s. A time series of annual life expectancy (LE) data has been available since 1987, which established a data foundation for analysis. The present study analyzes the socioeconomic factors associated with life expectancy in Beijing, and presents possible policy implications for future health quality improvements.

Data and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Data sources

Data used in this study were mainly from statistical reports and peer-reviewed papers. Life expectancy and socioeconomic data were collected from statistical yearbooks from 1988 to 2005 at municipal and national levels, including the China Statistical Yearbook, China Education Statistical Yearbook, China Energy Statistical Yearbook, China Health Statistical Yearbook, Beijing Statistical Yearbook, and the Annual Report of Beijing Center for Disease Prevention and Control.

Methods

Literature review

Through analyzing findings in domestic and international research literature we obtained information on socioeconomic factors associated with life expectancy. Several candidate variables were screened. We were concerned with the aspects of economy, environment, education, demography, and nutrition.

Variable adjustment

According to the requirements of multiple regression analysis, some variables were adjusted to fit a normal distribution. As reported, LEs in Beijing presented some changes that constituted abnormal fluctuations between some years. We hypothesized that Beijing held a similar growth trend in LE as that of Shanghai, so we adjusted LEs in Beijing from 1987 to 2004 based on the LE trend in Shanghai. The data series in Shanghai was chosen as the benchmark for adjustment because Shanghai and Beijing are very similar in terms of population, health, and socioeconomic development; moreover, the growth of LE in Shanghai changed smoothly and in line with the general trend.

Correlation analysis and multiple regression analysis

Pearson correlation analysis was adopted, taking factors primarily screened out as the independent variables and LE as the dependent variable. Multiple stepwise regression analysis was used to construct an econometric model between life expectancy and socioeconomic factors so as to explore the extent that factors were associated with LE. EXCEL and SPSS v13.0 software were used for dataset construction and regression analysis.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Life expectancy and its adjustment in Beijing

To eliminate the abnormal fluctuations in reported LE data in Beijing, we adjusted the time series of LE from 1987 to 2004 based on the same period LE's annual growth rate of Shanghai. The equation was followed in this paragraph and we have given one example of how to calculate the new LE for Beijing for the year of 1988. First, we calculated the average annual growth rate of LE in Shanghai from 1987 to 2004. Second, we hypothesized that LE in Beijing followed the same annual growth rate as that of Shanghai since Beijing and Shanghai shared almost the same characteristics in socioeconomic and demographic status. Third, we used the following equation to calculate the new time series of LE in Beijing.

  • image

where LEn+1, adjusted LE in Beijing in the year n + 1; LEn, LE in Beijing in the year of n; and X, annual growth rate of LE in Shanghai. For example, LE in 1988 was calculated using the following equation:

  • image

The reported and adjusted LEs in Beijing are shown in Table 1.

Table 1.  Reported and adjusted life expectancies in Beijing in 1987–2004
YearReported LE in Beijing (years)Reported LE in Shanghai (years)Adjusted LE in Beijing (years)
  1. LE, life expectancy.

198772.8974.4672.89
198872.8374.6373.23
198972.6174.9873.56
199072.4775.4673.90
199174.4975.7974.24
199274.4275.9774.59
199374.4775.9774.93
199474.0776.2675.27
199574.0476.0375.62
199673.7876.1175.97
199773.8877.2076.32
199872.8577.0376.67
199974.3378.4477.03
200073.8178.7777.38
200175.8579.6677.74
200277.9379.5278.10
200379.6279.8078.46
200479.8780.2978.82

Correlation analysis

As shown in the Pearson correlation analysis, there were significant effects of correlation (R > 0.9) between LE and 17 indicators (Tables 2 and 3), such as GDP per capita, annual net income per rural capita, daily water consumption per capita, illiteracy rate, public green areas per capita, floor space per urban resident, floor space per rural resident, proportion of female population, and proportion of rural population.

Table 2.  Original data of 17 indicators associated with life expectancy from 1987 to 2004
Indicators198719881989199019911992199319941995199619971998199920002001200220032004
GDP per capita (RMB)333841274497488153956806823710,26513,07315,04416,73518,47819,84622,46025,52328,44932,06137,058
Annual disposable income of urban households per capita (RMB)1187.81437.01597.11787.12040.42363.73296.04731.25868.46886.07813.18472.09182.810,349.711,577.812,463.913,882.615,637.8
Annual net income of rural households per capita (RMB)916.41062.61230.71297.11422.31568.81854.82422.13223.83580.03762.44029.04316.44687.04525.65880.16496.37172.1
Proportion of rural household recreation, education and culture articles expenditure (%)5.36.97.46.48.38.111.410.710.610.211.013.114.714.414.314.715.415.8
Daily consumption of tap water for residential use per capita (liters)160.4157.9155.2165.4178.2184.2198.9212.5219.2237.3268.1238.2250.1240.2260.0237.0248.0160.4
Park green areas per capita (m2)5.15.86.06.16.46.66.86.67.57.57.89.09.19.710.110.711.411.5
Using space of residential buildings in urban areas per capita (m2)6.827.177.457.728.018.318.518.739.039.339.6610.0310.6311.1511.6411.9318.6719.10
Floor space of residential building in rural areas per capita (m2)18.3819.2320.0920.6221.9222.6723.7024.4224.7425.7427.3927.6428.6528.9031.0032.6033.9534.20
Number of public lavatories per 10,000 people (unit)11.2711.0710.8210.7210.5310.3910.3810.029.028.759.198.948.597.216.856.996.816.55
Science expenditure per capita (RMB)5.894.743.885.016.456.386.416.758.569.9612.1213.4916.1720.5524.0029.1335.9744.11
Education expenditure per capita (RMB)48.4949.5147.4753.8650.9061.4564.8571.6284.31100.26121.84136.16162.26198.83238.45284.68330.22403.96
Illiteracy rate (%)13.9013.4012.2211.0310.1910.9810.419.857.957.647.336.516.454.234.935.354.614.48
Proportion of female population (%)48.9849.0249.0549.1349.1749.2049.2149.2449.2749.2649.349.3749.3849.4349.4550.0749.4349.51
Proportion of rural population (%)39.1738.6538.2437.9937.6237.1936.3935.6034.8934.1533.4232.7832.0631.3130.4928.9927.6826.50
Proportion of non-permanent resident population (%)7.397.365.966.466.837.107.568.808.568.9810.7810.7812.0113.3417.8824.0024.4522.70
Number of hospital beds per 1000 people (unit)4.494.875.085.375.655.735.936.076.006.026.066.136.156.256.316.465.896.54
Healthcare coverage rate of 0–6 year-old children (%)65.3467.6369.9159.4069.0080.1582.7088.6081.7892.6693.4695.6396.1197.7898.2398.0697.1796.58
Table 3.  Pearson correlation results between life expectancy and 17 indicatorsa
 RP
  1. aAll indicators were converted to normal distributions before correlation analysis.

  2. LE, life expectancy; R, coefficient of correlation; P, probability.

LE and GDP per capita 0.9320.000
LE and annual consumer income per urban capita 0.9630.000
LE and annual net income per rural capita 0.9190.000
LE and proportion of rural entertainment, culture and education services expenditure 0.9680.000
LE and daily water consumption per capita 0.9260.000
LE and public green areas per capita 0.9790.000
LE and using space per urban resident 0.9940.000
LE and floor space per rural resident 0.9950.000
LE and No. of public lavatories per ten thousand population−0.9730.000
LE and science expenditure per capita 0.8930.000
LE and education expenditure per capita 0.9100.000
LE and illiteracy rate−0.9770.000
LE and proportion of female population 0.8220.000
LE and proportion of rural population−0.9870.000
LE and proportion of non-permanent resident population 0.8800.000
LE and the number of hospital beds per 1000 population 0.8680.000
LE and health coverage rate of 0–6-year-old children 0.9160.000

Regression model of life expectancy

According to the above analysis results, these 17 indicators were put into the multiple stepwise regression model as independent variables with the LE as a dependent variable. The regression model was as follows and the results are illustrated in Table 4.

  • image

where Y, adjusted life expectancies in Beijing;

Table 4.  Results of stepwise regression model analysisa
 βStandardized BetatP
  1. aAR2= 0.999, F = 4783.012, P < 0.05.

Constant 75.200  48.3740.000
Proportion of rural population−0.006−0.304−3.0330.010
Floor space per rural resident−0.003−0.437−6.0650.000
Illiteracy rate 0.003 0.162 4.4350.001
GDP per capita 0.396 0.103 2.5920.022

X1, transformed proportion of rural population, X1= (proportion of rural population × 100 – 21.75)2;

X2, transformed floor space per rural resident, X2= (housing space for rural resident per capita × 100 – 55.75)2;

X3, transformed illiteracy rate, X3= (illiteracy rate × 100–22.86)2;

X4, transformed GDP per capita, X4= ln(GDP per capita).

Regression results revealed that there were four factors significantly associated with LE: (i) proportion of rural population; (ii) floor space per rural resident; (iii) illiteracy rate; and (iv) GDP per capita. The rural population and housing space correlated negatively with LE, while the GDP per capita and illiteracy rate had positive effects.

Based on these findings, we could identify the relationship between LE and the four associated factors. Proportion of rural population, if over 21.75, correlated negatively with LE, which means the larger proportion of rural population, the lower LE. Floor space per rural resident, if less than 55.75 m2, correlated positively with LE; thus, the increase of LE follows the increase of floor space per rural resident. Illiteracy rate correlated negatively with LE, while GDP per capita correlated positively with LE.

Sensitivity analysis of four socioeconomic factors

The sensitivity analysis showed the variation in LE in 2004 when the four socioeconomic factors varied. Given the other factors were fixed, the increasing speed of LE presented a downward accelerating trend with the decrease in proportion of rural population and the increase of floor space per rural resident. However, the increasing speed of LE presented an upward accelerated trend when illiteracy rate dropped, while marginal LE remained smooth, increasing when GDP per capita increased (Table 5).

Table 5.  Sensitivity analysis of the four factors to life expectancy
Variation in factorsProjected figureLE (years)Net increment of LE (years)
  1. LE, life expectancy; -, not applicable.

Proportion of rural population (%)
 Figure in 2004 26.5078.851-
 First -5% 25.1778.9160.065
 Second -5% 23.8478.9600.044
 Third -5% 22.5178.9830.023
 Fourth -5% 21.1878.9850.002
Floor space per rural resident (m2)
 Figure in 2004 34.2078.851-
 First +5% 35.9179.0630.212
 Second +5% 37.6279.2580.195
 Third +5% 39.3379.4350.177
 Fourth +5% 41.0479.5950.160
Illiteracy rate (%)
 Figure in 2004  4.4878.851-
 First -5%  4.2678.8750.024
 Second -5%  4.0478.9000.025
 Third -5%  3.8278.9250.025
 Fourth -5%  3.6078.9500.025
GDP per capita (RMB)
 Figure in 200437,05878.851-
 First +5%38,91178.8700.019
 Second +5%40,76478.8880.018
 Third +5%42,61778.9060.018
 Fourth +5%44,47078.9230.017

Model projection of life expectancy in Beijing

Taking the figures of four socioeconomic factors in 2005–2007 into the model, LE in Beijing could be projected. The model showed that the projected LEs in 2005–2007 would be 79.23, 79.52, and 79.69 years, respectively, which are similar to 80.09, 80.07, and 80.24 years reported by the Beijing Health Authority. The errors were only 1.07%, 0.69%, and 0.69%, respectively. This is a strong indicator that the model has a very good projection capacity in LE in Beijing. Therefore, this model can serve as a tool for rapid projection of LE in Beijing and can support public health decision-making in this field (Table 6).

Table 6.  Model projection for life expectancy in Beijing for 2005–2007
YearRural population (%)Housing space per rural resident (m2)Illiteracy rate (%)GDP per capita (RMB)Model projected LEReported LE by authority
  1. Sources: China Statistical Yearbook 2006–2008; Annual Report of Beijing Center for Diseases Prevention and Control 2005–2007.

  2. LE, life expectancy.

200516.3836.433.9245,44479.2380.09
200615.6739.834.4750,46779.5280.07
200715.5 39.793.3458,20479.6980.24

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

New time series of life expectancy in Beijing

Historically, the information system for vital statistics has developed well in Beijing. Since 1987, life expectancy has been tracked annually by the Beijing Center for Disease Prevention and Control and reported by the Beijing Health Bureau. However, due to unknown reasons, the time series of LE in Beijing in recent years had undergone great fluctuations, even more than the two-year average difference between two adjacent years. These fluctuations are not consistent with the general rules of LE growth. Shanghai shares similar socioeconomic and health characteristics with Beijing, and it also has a very good record in terms of vital statistical information. For these reasons, we take Shanghai LE growth as the benchmark for Beijing. After adjustment, a new time series of LE in Beijing was generated upon the annual growth rate of Shanghai between 1987 and 2004. From this approach, we can eliminate the abnormal fluctuations in LE in Beijing, and the new series could be used in the regression model to establish the relationship between LE and socioeconomic factors. This new time series of LE reflects the evolving trend of LE in Beijing over the last 20 years.

Socioeconomic factors associated with life expectancy in Beijing

As shown in large-scale domestic and international research, health status is affected by many factors. The existing related research articles are mostly focused on a single factor (4, 5) or single disease (6, 7). Few studies are concerned with macro-level factors like socioeconomic development. Our research is based on multi-factor analysis and identified factors associated with life expectancy in Beijing from 1987 to 2004. The results indicate that some socioeconomic factors exert significant influence on LE, such as housing conditions, education, economic levels, and urbanization. These four factors can explain more than 99% of the variation in LE, and housing conditions contribute more than other factors to LE growth in sensitivity analysis. This study provides insights on how to improve the health status from macro viewpoint.

The WHO launched the Commission on Social Determinants of Health (CSDH) in Santiago, Chile in 2005. The CSDH had been collecting evidence for socioeconomic and cultural factors associated with health, as well as related policies and interventions. Its target is to improve the health status and equity in health through public campaigns (8). This study has identified some key socioeconomic factors associated with LE in Beijing. The results have a great deal to contribute to policy improvement and related research.

Model for rapid projection of life expectancy in Beijing

The regression model was constructed on the basis of identified factors associated with LE and predicting future LE. When comparing the projected figures with the reported LE by the Beijing Health Bureau, there were only minor gaps that are accepted in terms of projection. There is strong evidence that the model has a very good projection capacity for forecasting LE in Beijing. Thus, the model can be a valuable tool for rapid projection of LE and provides references for decision-making in public health. For other provinces and cities in China, it is usually difficult to calculate LE annually due to the poor vital statistical information systems, so there is a great need to develop a rapid tool for LE forecasting using limited numbers of available indicators. This paper also provides a reference for this purpose.

Continuous increase of life expectancy with development of socioeconomic conditions in Beijing

Japan currently has the world's highest LE of 83 years in 2006(9). There is still a gap of 3 years between LE in Beijing and Japan. Fortunately, the factors associated with life expectancy are in favorable conditions in Beijing. For example, the level of urbanization is rising and the proportion of rural residents is decreasing. Other factors will also contribute to further improvement of LE, such as the sustained and rapid development of the economy, the continuous enhancement of education levels and quality, as well as the gradual improvement of rural housing conditions. As for health care in Beijing, it ranks first in terms of the number of doctors and nurses as well as delivery infrastructure in China; this will help improve people's health in Beijing. Currently, health services are accessible to urban and rural residents, the emergency network has developed very well, prevention and health education has been carried out universally for residents, and multi-layer health insurance schemes are developing quickly. All this will lead to continuous improvement of resident health status. In general, there are great chances for further improvements of LE.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Thanks to the Health Statistical Information Center in Beijing, Center for Disease Prevention and Control for providing life expectancy data in Beijing. Thanks to Ms SH Lin. Institute of Medical Informatics, Chinese Medical Academy, for editing the paper. This study was funded by the Beijing Natural Science Foundation, China.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Data and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Howard Preece. Life's statistic. Finance week 2004; 49.
  • 2
    Schatzkin A. How long can we live? A more optimistic view of potential gains in life expectancy. American Journal of Public Health 2004; 70(11): 1199200.
  • 3
    Shaw JW, Horrance WC, Vogel RJ. The determinants of life expectancy: an analysis of the OECD health data. Southern Economic Journal 2005; 71(4): 76883.
  • 4
    Wang XW, Liu JW, Duan YF. The impact of the level of rural residents income on health status in Ningxiang county. China's Health Economics 2004; 23(4): 235.
  • 5
    Ha Q, Gao Y, Mu Y, Shi XR, Xiao XM. The Impact of Cerebrovascular Diseases to the Death Probability and Life Expectancy. Chinese Journal of Prevention and Control of Chronic Non-communicable Diseases 2005;13(2): 867.
  • 6
    Dever GEA. An epidemiological model for health policy analysis. Social Indicators Research 1976;2(4): 45366.
  • 7
    Okamoto K. Life Expectancy at the Age of 65 years and Environmental Factors: An Ecological Study in Japan. Archives of Gerontology and Geriatrics 2006; 43(1): 8591.
  • 8
    WHO Executive Board. The 115th provisional agenda item – Social Determinants of Health Committee. Geneva : WHO, 2004; 16.
  • 9
    World Health Organization. World Health Statistics. Geneva : WHO, 2008; 38.