Jing Liu
P207
Project Paper
Healthcare Access and Health Outcomes for Rural Populations in China
Introduction
Because of rising inequality in China, mainly between the coastal urban regions and rural central
and western Chinese provinces, this project focuses on the relationship between healthcare
access and outcomes for provinces with high proportions of rural populations in China. How do
healthcare inputs affect health outcomes of these people? Are rural populations really more
vulnerable than urban populations? Are they more susceptible to catastrophic healthcare
expenses? It is my hope that mapping the relationships above will help inform policymaking.
Given the size of China, both geographically and in terms of population, policymaking can be a
serious challenge. The goal of this mapping exercise is not to identify specific areas for targeted
interventions but rather to provide policymakers with a broad overview of population needs and
trends.
Data Sources
The two main sources of data I used were official statistics yearbooks published by the Chinese
government.
1) China Yearbook 2011 compiled by the National Bureau of Statistics of China
Accessible online at: http://www.stats.gov.cn/tjsj/ndsj/2011/indexeh.htm
The yearbook is published annually and available in excel tabular format for download.
2) China Health Yearbook 2011 compiled by China’s Ministry of Health
This year book is also published annually but is only available in book format in
simplified Chinese only.
The base layer for mapping was obtained from the Global Administrative Areas website
accessible at http://www.gadm.org/.
Data preparation and analysis
Given that the majority of the data I required was not available as premade GIS shape files, I was
compelled to use data tables for my analysis. Once they were manually entered into an excel
Jing Liu
P207
Project Paper
spreadsheet in the ArcGIS recognizable format, it was only a matter of joining the excel
spreadsheets to the appropriate layers.
As for the analysis, I chose a simple overlay of two variables in each map. The base map
illustrates the percent of rural population in each province and the second layer consists of one
additional variable that affects either healthcare access or shows health outcomes. After testing
different types of segmentation under the symbology, I decided that the clearest maps were made
with just four sizes of symbols. Natural breaks were chosen to best identify naturally occurring
trends. It was my hope that these patterns would be highlighted when the finished maps were
juxtaposed.
According to the literature, there should be great inequalities in income between coastal
provinces with large urban populations and the central and western provinces with high rural
populations. The maps produced for health outcomes show that higher infant mortality rates,
higher chronic diseases rates, and lower life expectancy rates are associated with provinces with
greater percentage of rural people.
Jing Liu
P207
Project Paper
The maps for number of medical personnel and healthcare facilities are intended to be a proxy
for healthcare inputs. Greater investment in medical infrastructure and care should pay off with
healthier populations as indicated by the health outcomes above. The maps illustrate as expected
that areas with more hospitals and medical personnel (generally with higher urban populations
and coastal) are healthier than their rural counterparts.
Finally, the health care expenditure maps are intended to compare rural and urban populations
directly in terms of their out-of-pocket expenditures on healthcare. This indicator is a ratio
calculated by dividing the out of pocket expenses of rural/urban population by disposable income.
Because of limited health insurance coverage in China and the known high catastrophic expense
rate, I expected to find much larger expenditures in the rural regions, but the results of the data
do not reflect this hypothesis. In fact, from the map, it is difficult to discern correlation.
Jing Liu
P207
Project Paper
Difficulties encountered
The most difficult part about this final project was finding an appropriate research question with
adequate data. Although I started out with an ambitious research question about migrant workers
and vulnerability to health care issues, I realized that there was virtually no information on the
shadow population that could be used in ArcGIS.
Switching tracks, I decided to look more broadly at health trends in general populations that may
affect health outcomes. I found interesting resources such as China Health and Nutrition Survey
and the Chinese Family and Health Survey. Unfortunately the data was in STATA and SAS
formats that were not set up for GIS and also did not include data for all provinces. These data
sets required intense prepping that was not appropriate for the scale of this project. I do not
recommend attempting to use any unfamiliar formats that require additional software platforms
for a project such as this.
Jing Liu
P207
Project Paper
In the end, with the guidance of professors from the Harvard School of Public Health and
support of bibliographers at the Yenching Library at Harvard University, I finally found the
China Statistics yearbooks. The Health Yearbook was frustrating to use because it was in
Chinese only and I needed to have it translated and then manually entered into excel.
With the data I gathered, I initially hoped to access correlation between healthcare access and
vulnerability of rural populations. However, because most of the data was aggregated and did not
distinguish between rural and urban, I modified my methods to the one currently on display.
Jing Liu
P207
Project Paper
Conclusion
In the end, I believe my approach was incredibly inefficient because I struggled to identify an
interesting spatial question on health in China that had the available shape files and data layers
that I needed. From my experience, it seems that in the fields of health and social sciences,
simple maps and GIS analysis are sufficient for performing country-level analysis. However,
GIS can be useful for taking detailed data such as those in the China health surveys, mapping
them and using the outcomes of the analysis to predict overall trends.
If additional time were available, I would like to incorporate both levels of analysis. After
mapping out general patterns and trends for the country overall, I would to choose an interesting
example to perform deeper analysis. In fact, I was able to find province-specific statistics
yearbooks, but am unable to incorporate it due to the lateness in finding it. Once data was located,
I discovered a whole new challenge in determining the most important variables and finding an
effective method of presenting a large number of data variables concurrently. In the end, it was
necessary to limit myself to a manageable data set and I recommend future learners of GIS to do
the same.
I would also like to perform more sophisticated analysis with the variables. Econometric analysis
would have been greatly helpful in teasing out the most important variables before mapping. It
would also have opened up my project to more original research that hitherto may not have been
mapped. The time required to mine the data and find the appropriate indicators are impossible
under the current time constraint, but is worth consideration. Overall, I believe my approach does
not showcase the best strengths of ArcGIS as a geo-spatial tool. However I am satisfied with the
simplicity of my approach which is better suited to support academic papers such as my thesis
writing in the future.
References that influenced my methods
Ebenstein, Avaraham Y. & Sharygn, Ethan Jennings (2009). The consequences of “missing girls”
in China. World Bank Economic Review. 23(3): 399-425. doi: 10.1093/wber/lhp012
Accessible at: http://wber.oxfordjournals.org/content/23/3/399.full.pdf+html
Jing Liu
P207
Project Paper
This article looks at the long term consequences of gender selection of children in China. The
author examines the health outcomes of high gender ratios. It is an interesting article for me
because it looks at a Chinese social phenomenon and looks at health outcomes as a result of this
cultural behavior. On page 404-405, he uses the 2000 census to construct a basic map of sex
ratios in children and sex ratios in Chinese of marriageable ages. Although this is very basic use
of GIS tools, he is able to use these maps as a basis of more sophisticated analysis of social
impacts in the long run such as power dynamics between men and women of marriageable age,
investments in health and education for each, and insecurity for those men who never find mates.
Rosero-Bixby, Luis (2004). Spatial access to health care in Costa Rica and its equity: a GIS
based study. Social Science and Medicine, 38, 1271-1284.
Accessible at: http://ac.els-cdn.com/S0277953603003228/1-s2.0-S0277953603003228main.pdf?_tid=d28eaf10eb16f9f93641009620cab580&acdnat=1333730085_8d89c04c02e3b0f07
3214a4a35a3252e
This author examines the impact of health care reform on equity of access in Costa Rica. Relying
on the 2000 census, the author performs traditional measurement of access according to distance
to health care facilities. Because the reforms were not rolled out simultaneous, he compared the
Jing Liu
P207
Project Paper
differences of access between areas that had implemented reform measures in the mid and late
1990’s and areas where reform had not occurred by 2001. This is an interesting study for my
project because it looks at equity of access on a national level in the context of reforms. China
implemented heavy reforms earlier on in the decade. I would have liked to model my project
after this author’s, but unfortunately, the size of the two countries is vastly different and China is
not suitable for raster density analysis. Moreover, the same level of data detail was not available.
Lafferty, Sara L. (2003). GIS and health care. Annual Review of Public Health. 24. 25-43. doi:
10.1146/annurev.publhealth.24.012902.141012
Accessible at:
http://www.annualreviews.org/doi/pdf/10.1146/annurev.publhealth.24.012902.141012
This is a great reference article that looks at the adoption of GIS by health care researchers in
analyzing health care need, access, and utilization. There is a section on analyzing access to
health care and evaluating inequalities. Because of China’s size I will likely use area based
analysis in the form of ratios or indicators such as doctors per 1000 people. However, provincial
level ratios are likely to mast patterns of inequalities and diversity within the provinces. On a
provincial level, it will be impossible to measure geographic distance and derive meaningful data.
Liu Yan, Wong Shuang Yann, Jin Tao (2009). Equality of spatial access to primary health
services for Singapore’s baby boomers. Asia Population Studies. Vol 5. 171-188.
Jing Liu
P207
Project Paper
Accessible at: http://www.tandfonline.com/doi/full/10.1080/17441730902992091
The authors assess equality of health services for a specific population (baby boomers) in
Singapore. The authors look at service capacity of health care measured by weighted number of
physicians of each zone. They also have a graph that displays population density as a shape file
in the background and then other ratios graphics in the foreground. The maps that I ended up
creating are very similar to the ones used by these authors but simplified. I found that even their
maps were too “busy” and contained variables. Alternatively, I hold a single dependent variable
constant (rural population) and then overlaid various indicators in the hopes of finding
correlations.