Rural and Urban Disparity in Health Services Utilization ...people.ucalgary.ca/~lu/MedicalCare_ChinaHealthInsurance.pdf · ORIGINAL ARTICLE Rural and Urban Disparity in Health Services

ORIGINAL ARTICLE

Rural and Urban Disparity in Health Services Utilizationin China

Meina Liu, MD, PhD,* Qiuju Zhang, MD, MPH,* Mingshan Lu, PhD,†¶ Churl-Su Kwon, MD,§and Hude Quan, MD, PhD‡�

Objectives: To describe patterns in physician and hospital utiliza-tion among rural and urban populations in China and to determinefactors associated with any differences.Methods: In 2003, the Third National Health Services Survey inChina was conducted to collect information about health servicesutilization from randomly selected residents. Of the 193,689 respon-dents to the survey (response rate, 77.8%), 6429 urban and 16,044rural respondents who were age 18 or older and reported an illnesswithin the last 2 weeks before the survey were analyzed. General-ized estimating equations with a log link were used to assess therelationship between rural/urban residence and physician visit/hos-pitalization to adjust for respondents clustered at the householdlevel.Results: About half of respondents did not see a physician whenthey were ill. Rural respondents used physicians more than urbanrespondents (52.0% vs. 43.0%, P � 0.001) and used hospitals less(7.6% vs. 11.1%, P � 0.001). Factor associated with increasedphysician utilization included residing in rural areas among majorityChinese (ie, Han) �rate ratio (RR), 1.21; 95% confidence interval(95% CI), 1.16–1.26�, residing �3 km away from the medicalcenter (RR, 1.16; 95% CI, 1.12–1.21), or being uninsured (RR, 1.38;95% CI, 1.30–1.46). Rural minority Chinese visited physicianssignificantly less than urban minority Chinese (RR, 0.90; 95% CI,0.83–0.98). Hospital utilization was significantly lower among ruralmales (RR, 0.84; 95% CI, 0.72–0.98), rural seniors (age, �65; RR,0.64; 95% CI, 0.53–0.77), rural respondents with low education(RR, 0.70; 95% CI, 0.57–0.86 for illiterate), or rural insured respon-dents (RR, 0.86; 95% CI, 0.69–0.99) than hospitalization amongurban counterparts.

Conclusions: Three national approaches should be considered inreforming the healthcare system in China: universal insurance cov-erage, higher amounts of insurance coverage, and increasing thepopulation’s level of education. In addition, access issues in remoteareas and by rural minority Chinese population should be addressed.

Key Words: health services, healthcare system, health insurance,China

(Med Care 2007;45: 767–774)

During the last 2 decades, the population health status inChina significantly improved along with its dramatic

economic development. In the 21st century, life expectancy atbirth reached 71.4 years.1 However, a large disparity in ruraland urban health status exists. Compared with urban residents(about 561 million), infant mortality in the rural population(about 745 million) is over 2 times higher (11.3 of 1000 vs.28.7 of 1000) and life expectancy is 6 years less (75.2 yearsvs. 69.5 years).1,2 Although many factors, particularly socio-economic status, account for such a large gap, access to thehealthcare system is a pivotal factor.3

The healthcare system was profoundly reformed in the1980s under the national climate of economic profit as pri-ority. A publicly funded and government-managed systemwas transited to a market-oriented system,4 resulting in thecurrent Chinese healthcare system serving a small proportionof the population that has or can afford health insurance. Forthe urban population, the dominate form of health insuranceis employment-based, including the Labor Insurance Programand government employee insurance. These 2 forms of in-surance provide partial or no coverage for dependents.5 Fur-thermore, with increasing utilization of costly diagnostics,high-technology surgery, and imported drugs, many employ-ers no longer provide full insurance coverage even to theemployee. In fact, the Medical Saving Account scheme isbecoming increasingly popular in urban China, wherebyemployers provide employees with a fixed amount of moneyper month to cover basic health services and employees areresponsible for the remaining expenditures.6,7 For the ruralpopulation, the Cooperative Medical System, a form of com-munity-based health insurance, has collapsed, leaving privatehealth insurance the only option.5,8 Although different de-grees of coverage are available through private health insur-ance, the premiums for extensive coverage are expensive and

From the *Department of Biostatistics, School of Public Health, HarbinMedical University, Harbin, China; Departments of †Economics, ‡Commu-nity Health Sciences, University of Calgary, Calgary, Alberta, Canada;§Lister Hospital, Corey’s Mill Lane, Stevenage, Hertfordshire, United King-dom; ¶Research Institute of Economics and Management, SouthwesternUniversity of Finance and Economics, Chengdu, China; and �Centre forHealth and Policy Studies, University of Calgary, Calgary, Alberta, Canada.

Supported by a Population Health Investigator Award from the AlbertaHeritage Foundation for Medical Research, and by a New InvestigatorAward from the Canadian Institutes of Health Research, Canada (toH.Q.), and Institute of Health Economics, Alberta (to M.L.).

Reprints: Dr. Hude Quan, Department of Community Health Sciences,University of Calgary, 3330 Hospital Drive NW, Calgary, Alberta,Canada T2N 4N1. E-mail: [email protected].

Copyright © 2007 by Lippincott Williams & WilkinsISSN: 0025-7079/07/4508-0767

Medical Care • Volume 45, Number 8, August 2007 767

cannot be afforded by most of the rural and even the urbanpopulations. This leaves the vast rural population, children,and those who are unemployed or retired uninsured.5,8 Be-cause rural residents are much poorer than urban residentson average, financial barriers may widen the disparitybetween urban and rural residents’ access to healthcareservices even further. In recent years, the Chinese govern-ment has been exploring new financing models in ruralareas, hoping to reestablish community-based health insur-ance; however, these efforts largely remain at the researchand experimental stages, with some pilot projects beinglaunched and evaluated.8,9

Previous studies about the Chinese healthcare systemhave been limited to examining health insurance and health-care in certain areas in China, with little empirical evidenceon the rural and urban disparity in health service utiliza-tion.6,10–19 A growing number of studies have been con-ducted to address geographic variation in access to healthcareservices and outcomes because geographic location mayinfluence a patient’s chance of receiving healthcare20–22; tofill this knowledge gap in China, we conducted this uniquestudy using a Chinese national survey to describe patterns ofrural and urban physician and hospital services utilization,and to determine factors associated with any observed differ-ences. Our study provides evidence to health policy makerson whether rural and urban residents in China access thehealthcare system equally, and proposes approaches to elim-inate unequal access. The elimination of disparities in health,including healthcare, has been identified as a target to im-prove population health status in several countries.23

METHODS

Study PopulationWe used data from the China Third National Health

Services Survey, which collected data through face-to-face in-terviews from September 18 to October 20, 2003. Of the193,689 respondents surveyed, we excluded individuals whowere �8 years old and included only those with a demonstratedneed for healthcare, that is, those reporting an illness within thelast 2 weeks before the survey (Fig. 1). This resulted in a total of22,473 (6429 urban and 16,044 rural) respondents in our sampleafter excluding 168 respondents with missing values.

The survey used a multiple-stage cluster sampling methodto randomly select the sample. The entire country, except HongKong and Macau, was clustered by the government administra-tive geographic system (ie, city, county, town, and village). Atotal of 95 cities or counties, 475 towns, and 950 villages wererandomly selected. The selected cities were further divided intoneighborhood communities (the smallest unit of administrationin an urban area). In each community or village, 60 householdswere randomly selected, resulting in about 57,000 households.All family members were invited to participate in the face-to-face interview. The same sampling method had been used in theprevious 2 National Health Services Surveys. Analyses of pre-vious surveys suggest that this sampling method is adequate togenerate a nationally representative sample.24 The survey re-spondent age and sex composition was comparable with the2000 census.

Medical doctors conducted the survey. Before the sur-vey, interviewers were trained and practiced interviewing.After training, their understanding and knowledge about thesurvey method and content were examined through testing.During the survey, interviewers visited each household up to3 times on different days. Interviewers explained the pur-poses and confidentiality of the survey, and then invitedfamily members to participate. Respondents could choose notto participate; their participation in the survey was acceptedas oral consent.

The completeness of questionnaires was checked by adistrict survey manager at the end of every day. If there wasmissing information on the survey, individuals were resur-veyed if possible. The survey response rate was 77.8%. Afterthe survey, 5% of households were randomly selected andresurveyed on 14 questions to examine survey quality; theagreement was 95%.25

Dependent VariablesThe dependent variables included physician visit(s)

within the last 2 weeks and hospitalization(s) within thelast 1 year. The survey asked, “Have you received anytreatment during the last 2 weeks?” Those who receivedtreatment were further asked, “What type of treatment?”and provided 3 choices: (1) self-treatment without physi-cian visit, (2) physician visit, or (3) a combination ofself-treatment and physician visit. We grouped respon-dents who chose either physician visit or a combination ofself-treatment and physician visit as the physician users.The remaining individuals, those who chose either self-treatment only or not receiving any treatment, were clas-sified as nonphysician users.

FIGURE 1. Sample size.

Meina et al Medical Care • Volume 45, Number 8, August 2007

© 2007 Lippincott Williams & Wilkins768

Hospitalization in the last year was determined basedon the survey question, “How many times were you hospi-talized during the last one year due to any reasons, such asillness, injury, physical check-up, delivery, et al?” From thisquestion, we categorized respondents into hospital users (1 ormore hospitalization) and nonhospital users. Hospitalizationwas defined as staying in hospital at least 1 night.

Independent VariableIndependent variables were selected based on the be-

havioral model of health service utilization developed byAday and Andersen.26,27 This model is one of the mostfrequently used frameworks in North America as well as inChina28–30 for analyzing the factors associated with patientutilization of healthcare services and access to healthcare.Using this model, we classified age, sex, ethnicity, andmarital status as predisposing factors; rural/urban, geographicregion, education, distance from home to the closest medicalcenter, and health insurance status (insured or not) as en-abling factors; and self-perceived severity of illness andpresence of physician-diagnosed chronic diseases within thelast 6 months as need factors.

China was geographically grouped into urban (ie, cit-ies) or rural areas (ie, town or villages) after the governmentaladministration system, as well as Eastern China, mid-China,and Western China based on economic development status.Eastern China is the most developed region, mid-China is lessdeveloped, and Western China the least.

Ethnicity was grouped into Han or minority. Amongthe 56 ethnicities in China, Han ethnicity is the largest,accounting for 95% of the population. The survey askedrespondents: “How far is the nearest medical center or clinicfrom your home?” Possible responses were as follows: �1, 2,3, 4, and 5 km or more. Health insurance status was definedaccording to survey questions about coverage by social wel-fare, private health insurance, or both.

Presence or absence of individual chronic disease wasrecorded. Chronic disease referred to disease diagnosed bymedical doctors as chronic disease in the last 6 months beforethe survey, or chronic disease that was diagnosed more than6 months before the survey but reoccurred with the last 6months and received treatment. Nonphysician-diagnosedchronic disease was not included because the validity ofself-diagnosed medical conditions depends on the level of therespondent’s knowledge and their perceptions on the defini-tion of “disease” and “health.” Physician-diagnosed chronicdisease was further confirmed through self-reported type ofhospital where the diagnosis was received (including com-munity clinics, county hospital, city hospital, provincial hos-pital, military hospital, and others). Self-perceived illnessseverity was measured within the categories of minor, mild,severe, and unsure.

Statistical AnalysisDescriptive statistics were used to test the statistical

differences in sociodemographic characteristics, physicianvisits, and hospitalizations between rural and urban residents.Frequencies of variables in the survey were not weightedbecause sampling weights were not available. Stratified anal-

yses were conducted for the independent variables to deter-mine whether the association between rural/urban residenceand health services utilization differed across strata (ie, effectmodification). Finally, multiple binomial regression with alog link was used to assess whether the relationship betweenrural/urban residence and physician visit/hospitalization wasconfounded or modified by other characteristics. Clusteringof individuals within family was adjusted for using therepeated measure function in SAS 9.1 Proc GENMOD.31,32

Two multiple binomial regression models, 1 for physi-cian visit and another for hospitalization were fitted after a3-step modeling strategy.

Step 1 was to form a main effect model with physicianvisit/hospitalization as the dependent variable, rural/urban as theexposure independent variable, and sex, age, ethnicity, maritalstatus, education, distance from home to medical center, healthinsurance, severity of illness, chronic disease, and geographicregion as the remaining independent variables.

Step 2 was to assess effect modifications. All 2-factormodification terms involving rural/urban and other indepen-dent variables were produced and fitted into the main effectmodel. Exclusion of the 2-factor modification terms from themodel in sequential fashion was based on the log-likelihoodtest and biologic plausibility.

Step 3 was to remove variables that did not meaning-fully alter the rate ratios (RRs) (also called prevalence ratio orrisk ratio) for rural/urban residence or were not significantlyassociated with physician visit/hospitalization. The retainedvariables and modification terms in the parsimonious modelwere used to assess the association between rural/urban andphysician visit/hospitalization.

The data were analyzed at the health information centerof the Ministry of Health in Beijing. Confidentiality of thesurvey was protected by storing the data on password-pro-tected computers at the Ministry, removal of personal iden-tifiable information (such as name and address) from thedatabase available for researchers, and examining analysisoutputs for release of aggregated data by the center staff.

RESULTS

Descriptive AnalysisCompared with urban respondents, rural respondents were

more likely to be male, younger, an ethnic minority versus Han,unmarried, less educated, and residents in a remote area orWestern China (Table 1), but were less likely to have healthinsurance (12.6% vs. 62.5%, P � 0.001) and physician-diag-nosed chronic disease (52.3% vs. 72.0%, P � 0.001).

About half of the respondents did not see a physicianwhen they were ill (Table 2), but many respondents treatedillnesses by themselves without physician consultations(32.3% for rural and 47.0% for urban). The proportion ofself-treatment was correlated to level of education (49.7% forcollege or above, 42.3% for senior high, 38.6% for juniorhigh, 34.8% for elementary, and 32.1% for illiterate).

Rural respondents visited physicians more than urbanrespondents (52.0% vs. 43.0%, P � 0.001), but hospitals less(7.6% vs. 11.1%, P � 0.001; Table 3). The stratified analysisby rural/urban and remaining independent variables is pre-

Medical Care • Volume 45, Number 8, August 2007 Health Services Utilization in China

© 2007 Lippincott Williams & Wilkins 769

sented in Table 3. Differences between rural and urbanresidents in physician visits depended on ethnicity, maritalstatus, and distance from home to medical center; differencesin hospital utilization depended on age, ethnicity, maritalstatus, education, distance from home to medical center,insurance status, and presence of chronic disease.

Multivariate AnalysisAfter adjusting for the independent variables listed in

Table 1, rural respondents were still more likely to visit a

physician �risk-adjusted RR, 1.16; 95% confidence interval(95% CI), 1.12–1.21, Table 4), but tended to use the hospitalless than urban respondents (RR, 0.94; 95% CI, 0.84–1.05;Table 4).

There were significant subpopulation disparities in ruraland urban health services utilization (Tables 5 and 6). Ruralresidents visited physicians significantly more than urbanresidents among respondents who were Han Chinese (RR,1.21; 95% CI, 1.16–1.26; Table 6), resided less than 3 kmaway from a medical center (RR, 1.16; 95% CI, 1.12–1.21),or uninsured (RR, 1.38; 95% CI, 1.30–1.46). Rural minorityChinese respondents visited physicians significantly less thanurban minority Chinese respondents (RR, 0.90; 95% CI,0.83–0.98). Hospital utilization was significantly loweramong rural male (RR, 0.84; 95% CI, 0.72–0.98), ruralseniors (age �65; RR, 0.64; 95% CI, 0.53–0.77), and ruralrespondents with low education (RR, 0.70; 95% CI, 0.57–0.86 for illiterate) or insured (RR, 0.86; 95% CI, 0.69–0.99)than hospitalization among urban counterparts.

DISCUSSIONOur analysis of the China Third National Health Ser-

vices Survey highlighted that about half of respondents didnot visit physicians when they were ill in the 2 weeks beforethe survey; rural residents used physician services more buttended to use hospital services less than urban residents. Therural and urban disparity in hospital service utilization wasparticularly evident in male, senior, illiterate, or insuredpopulations.

Physician services utilization has been declining in thelast decade in China. The previous 2 National Health ServicesSurveys revealed that of urban residents who reported illnessin the last 2 weeks, 59% saw a physician in 1993 and 50% in1998.16 The percentage dropped to 43% in the 2003. Ofmultiple factors related to the decline, increases in fees forhealthcare services and low insurance coverage may be themost essential.

Patients generally see physicians at outpatient hos-pital clinics for diagnostic tests and treatment, and atcommunity medical centers for follow-up, rehabilitation,and treatment of minor medical conditions. With the tran-sition in national policy from social equality to capitalism,the traditional health policy of “public funding to health-care, prevention first and health for all” has been ne-glected. Medical clinics and hospitals set economic profitsas their priority, and the fee for physician visits greatlyvaries by hospitals and by physicians’ qualifications. Topursue revenue, hospitals purchase expensive diagnosticequipment and build costly wards, whereas physiciansprescribe high-profit and expensive imported pharmaceu-ticals as well as high-profit diagnostic and laboratory tests.For urban residents, the average cost was 219 Yuan (Chi-nese currency) per physician visit (including diagnosticsand medication) and 7606 Yuan per hospitalization in2003, an increase of 85% and 88% over the average cost in1998. For rural residents, the average cost in the same yearwas 91 Yuan per physician visit (including diagnostics and

TABLE 1. Characteristics of Survey Respondents WhoReported Illness in the 2 Weeks Before the Survey

VariablesRural (%)N � 16,044

Urban (%)N � 6429 P

Gender, male 7276 (45.4) 2734 (42.5) �0.001

Age, yr

18–49 7714 (48.0) 2045 (31.8) �0.001

50–64 5062 (31.6) 1917 (29.8)

�65 3268 (20.4) 2467 (38.4)

Ethnicity, Han Chinese 13,381 (83.4) 5840 (90.8) �0.001

Marital status

Married 13,373 (83.3) 5096 (79.3) �0.001

Unmarried 700 (4.4) 251 (3.9)

Divorced 110 (0.7) 147 (2.3)

Widow 1861 (11.6) 935 (14.5)

Education

Illiterate 5726 (35.7) 1086 (16.9) �0.001

Elementary school 5529 (34.5) 1382 (21.5)

Junior high school 3833 (23.9) 1717 (26.7)

Senior high school 726 (4.5) 966 (15.0)

College/university orhigher

230 (1.4) 1278 (19.9)

Distance from home tomedical center�3 km

14,105 (87.9) 6165 (95.9) �0.001

Having health insurance 2023 (12.6) 4019 (62.5) �0.001

Self perceived illness asbeing severe

4120 (25.7) 1607 (25.0) 0.2881

Presence of physician-diagnosed chronicdisease

8398 (52.3) 4627 (72.0) �0.001

Geographic region

Eastern China 5014 (31.3) 2473 (38.5) �0.001

Mid-China 3853 (24.0) 1818 (28.3)

Western China 7177 (44.7) 2138 (33.2)

TABLE 2. Illness Management Among Rural and UrbanResidents Who Reported Illness in the 2 Weeks Before theSurvey

Rural (%)N � 16,044

Urban (%)N � 6429

Self-treatment without physician visit 5180 (32.2) 3022 (47.0)

Physician visit 6958 (43.4) 2111 (32.9)

Self-treatment and physician visit 1379 (8.6) 652 (10.1)

No treatment 2527 (15.8) 644 (10.0)



medication) and 2649 Yuan per hospitalization in 2003, a103% and 73% increase over 1998.33 To put this intoperspective, the fee for 1 hospitalization is approximatelyequivalent to 6 months’ average income of a labor workerin China. Such expensive healthcare compels low-incomepatients to stay away from hospitals. Also, health insur-ance premiums are expensive and out of reach for low-income patients, further restricting access. In our sample,87% of rural and 37% of urban residents were uninsured.Our findings of low physician visits among seniors, mi-

nority Chinese, and residents in less-developed regions(mid and Western China) as well as high physician visitsamong the insured support the argument that financialbarriers restrict access to healthcare (Table 4).

Our study demonstrated that many patients (47% ruraland 32% urban residents) took self-treatment (such as takingmedication stored at home or purchased at drug stores)without consulting a physician. Prescribed drugs have beenintroduced in China just recently. Before that, a prescriptionwas not required to purchase medication at a pharmacy.

TABLE 3. Physician Visit in the Last 2 Weeks and Hospitalization Within 1 Year Among Rural and Urban Residents WhoReported Illness in the 2 Weeks Before the Survey

Physician Visit Hospitalization

Ruraln (%)

Urbann (%) P

Ruraln (%)

Urbann (%) P

Overall 8337 (52.0) 2763 (43.0) �0.001 1223 (7.6) 713 (11.1) �0.001

Age, yr

18–49 4114 (53.3) 899 (43.5) �0.001 544 (7.1) 134 (6.6) 0.430

50–65 2623 (51.8) 789 (41.2) �0.001 433 (8.6) 178 (9.3) 0.335

�65 1600 (48.9) 1085 (44.0) �0.001 246 (7.5) 401 (16.3) �0.001

Gender

Male 3706 (50.9) 1158 (42.4) �0.001 564 (7.8) 360 (13.2) �0.001

Female 4631 (52.8) 1605 (43.4) �0.001 659 (7.5) 353 (9.6) 0.001

Ethnicity

Han Chinese 7076 (52.9) 2415 (41.4) �0.001 975 (7.3) 660 (11.3) �0.001

Minority Chinese 1261 (47.4) 348 (59.1) �0.001 248 (9.3) 53 (9.0) 0.812

Marital status

Married 6970 (52.1) 2199 (43.2) �0.01 1046 (7.8) 562 (11.0) �0.001

Unmarried 377 (53.9) 107 (42.6) 0.002 44 (6.3) 17 (6.8) 0.787

Divorced 49 (44.6) 57 (38.8) 0.3525 8 (7.3) 15 (10.2) 0.415

Widow 941 (50.6) 400 (42.8) �0.001 125 (6.7) 119 (12.7) �0.001

Education

Illiterate 2996 (52.3) 520 (47.9) �0.007 407 (7.1) 119 (11.0) �0.001

Elementary school 2890 (52.3) 617 (44.7) �0.001 441 (8.0) 172 (12.5) �0.001

Junior high school 1969 (51.4) 700 (40.8) �0.001 282 (7.4) 174 (10.1) �0.001

Senior high school 366 (50.4) 411 (42.6) 0.001 64 (8.8) 100 (10.4) 0.290

College or university 116 (50.4) 516 (40.3) 0.004 29 (12.6) 148 (11.6) 0.656

Distance from home to medical center

�3 km 7328 (52.0) 2627 (42.6) �0.001 1064 (7.5) 691 (11.2) �0.001

�3 km 1009 (52.0) 136 (51.5) 0.8735 159 (8.2) 22 (8.3) 0.941

Health insurance

Insured 1059 (52.4) 1879 (46.8) �0.001 158 (7.8) 514 (12.8) �0.001

Uninsured 7278 (51.9) 884 (36.7) �0.001 1065 (7.6) 199 (8.3) 0.260

Self perceived illness

Severe 2638 (64.0) 961 (59.8) 0.003 560 (13.6) 313 (19.5) �0.001

Nonsevere 5699 (47.8) 1802 (37.4) �0.001 663 (5.6) 400 (8.3) �0.001

Presence of physician diagnosed chronic disease

Absence 4098 (53.6) 776 (43.1) �0.001 331 (4.3) 94 (5.2) 0.102

Presence 4239 (50.5) 1987 (42.9) �0.001 892 (10.6) 619 (13.4) �0.001

Geographic region

Eastern China 2691 (53.7) 1152 (46.6) �0.001 339 (6.8) 266 (10.8) �0.001

Mid-China 1946 (50.5) 573 (31.5) �0.001 273 (7.1) 217 (11.9) �0.001

Western China 3700 (51.6) 1038 (48.6) 0.0148 611 (8.5) 230 (10.8) 0.001



Individuals would purchase drugs based on their own knowl-edge of their illness and the effectiveness of these drugs intreating their illness. There was no mechanism in place tomonitor the self-administration of medications and safety ofindividuals choosing to do so. When individuals did notbecome better after self-treatment or perceived their illness asbeing severe, they would go to visit a physician. This issupported by our findings that self-perceived severity ofillness is a strong predictor of physician visit and hospital-ization. Moreover, the correlation between the level of edu-cation and self-treatment further suggests that patients with ahigh level of health knowledge were more likely to turn toself-treatment.

Long-term consequences of self-treatment on popula-tion health could be numerous, such as misusing medication,worsening disease severity, causing complications, delayingthe optimal timing of effective treatments, and impairingquality of life. These consequences could increase disease

burden and impair population health. Self-treatment couldalso cause challenges in monitoring infectious and epidemicdiseases.

Our results highlighted that rural residents used physi-cians more and hospitals less than urban residents. The firstpossible explanation for these findings is that drugs aregenerally available only at clinics in rural areas. Therefore,residents in these areas see a physician to obtain a prescrip-tion for medication. The relatively lower cost of physicianvisits in rural areas may promote physician utilization as well.The second possible explanation is that the cost to travel from

TABLE 4. Rate Ratio in Physician Visit in the Last 2 Weeksand Hospitalization Within 1 Year Among Respondents WhoReported Illness in the 2 Weeks Before the Survey

Physician VisitRate Ratio (95% CI)

HospitalizationRate Ratio (95% CI)

Rural/urban 1.16 (1.12–1.21) 0.94 (0.84–1.05)

Gender

Female/male 1.02 (0.99–1.04) 0.94 (0.86–1.02)

Age, yr

50–64/18–49 0.95 (0.92–0.98) 1.11 (1.00–1.24)

�65/18–49 0.91 (0.88–0.95) 1.40 (1.24–1.58)

Ethnicity

Minority/Han 0.94 (0.90–0.98) 1.09 (0.96–1.24)

Marital status

Unmarried/married 1.02 (0.96–1.08) 0.95 (0.74–1.21)

Divorced/married 0.88 (0.77–1.00) 1.00 (0.69–1.47)

Widow/married 0.98 (0.94–1.03) 0.86 (0.75–0.99)

Education

Elementary/illiterate 1.00 (0.97–1.03) 1.30 (1.15–1.45)

Junior high/illiterate 0.96 (0.93–1.00) 1.35 (1.18–1.55)

Senior high/illiterate 0.96 (0.91–1.02) 1.48 (1.23–1.78)

College oruniversity/illiterate

0.92 (0.86–0.98) 1.52 (1.26–1.84)

Distance from hometo medical center

�3 kilometer/�3 kilometer

0.99 (0.95–1.03) 1.00 (0.86–1.16)

Health insurance

Insured/uninsured 1.09 (1.05–1.13) 1.14 (1.03–1.27)

Self perceived severity ofillness

Severe/unsevere 1.42 (1.38–1.45) 2.22 (2.03–2.42)

Presence of physiciandiagnosedchronic disease

Presence/absence 0.90 (0.88–0.93) 2.15 (1.92–2.40)

Geographic region

Mid/Eastern 0.87 (0.84–0.91) 1.08 (0.96–1.21)

Western/Eastern 0.95 (0.92–0.98) 1.12 (1.01–1.25)

TABLE 5. Rate Ratio for Rural vs. Urban Residents inPhysician Visit in the Last 2 Weeks Among Respondents WhoReported Illness in the 2 Weeks Before the Survey

Physician VisitRate Ratio (95% CI)*

Rural/urban among Han Chinese respondents 1.21 (1.16–1.26)

Rural/urban among minority Chineserespondents

0.90 (0.83–0.98)

Rural/urban among respondents with distance�3 kilometer from home to medical carecenter

1.16 (1.12–1.21)

Rural/urban among respondents with distance�3 kilometer from home to medicalcare center

0.99 (0.86–1.15)

Rural/urban among respondents insured 0.99 (0.93–1.05)

Rural/urban among respondents uninsured 1.38 (1.30–1.46)

*Rate ratio for rural vs. urban was adjusted for sex, age, ethnicity, martial status,education, distance from home to medical center, health insurance, severity of illness,chronic disease, and geographic region.

TABLE 6. Rate Ratio for Rural vs. Urban Residents inHospitalization Within 1 Year Among Respondents WhoReported Illness in the 2 Weeks Before the Survey

HospitalizationRate Ratio (95% CI)*

Rural/urban among males 0.84 (0.72–0.98)

Rural/urban among females 1.06 (0.91–1.23)

Rural/urban among respondents age18–64

1.20 (1.03–1.39)

Rural/urban among respondents age�65

0.64 (0.53–0.77)

Rural/urban among illiteraterespondents

0.70 (0.57–0.86)

Rural/urban among respondents withelementary education

0.85 (0.70–1.03)

Rural/urban among junior or seniorhigh school education

1.18 (0.98–1.42)

Rural/urban among respondents withcollege/university or highereducation

1.37 (0.93–2.00)

Rural/urban among respondentsinsured

0.83 (0.69–0.99)

Rural/urban among respondentsuninsured

1.07 (0.91–1.25)

*Rate ratio for rural vs. urban was adjusted for sex, age, ethnicity, martial status,education, distance from home to medical center, health insurance, severity of illness,chronic disease, and geographic region.



a village to a town with a pharmacy is a deterrent, comparedwith the cost of seeing a physician in the local clinic andfilling a prescription there. In contrast, residents in urbanareas have drugs available to them at local pharmacies andthe cost of seeing a physician is much higher compared withthat in rural areas. Therefore, urban residents with insurancecoverage are as likely as rural residents to see physicians, buturban residents without insurance may choose drugs at phar-macies without physician consultation. The third possibleexplanation is that rural residents primarily rely on outpatienttreatment and accept hospitalization only as a last resortbecause of its high cost. Our study demonstrated that unin-sured rural residents used physician services significantlymore but used hospitals equally compared with uninsuredurban residents.

Of variables associated with hospitalization (ie, age,marital status, education, insurance, severity of disease, pres-ence of chronic disease, and geographic region), educationand insurance are 2 modifiable factors (Table 4). Thesefindings indicated that promoting access to healthcare shouldfocus on reforming the Chinese healthcare system towardsuniversal healthcare coverage, and promoting a higher na-tional educational level. Education is a proxy of socioeco-nomic status.34 Patients with a high level of education wouldlikely be employed and insured because health insurancepremiums are partly covered by employers. However, amajority of rural residents are less educated and uninsured.We found that rural insured residents used physician servicesequally but used hospitals significantly less compared withinsured urban residents (Tables 5 and 6). The amount ofinsurance might be lower for rural residents than urbanresidents, and thus be insufficient to cover expensive hospitalservices. Unfortunately, our data do not allow us to quantifythe actual coverage of insurance.

Our study has some important limitations. Data werecollected through a survey and therefore subjected toerrors in recall. Also, the cross-sectional nature of thesurvey makes it impossible to establish a causal relation-ship between urban/rural residence and health servicesutilization. We studied physician and hospital utilizationand did not assess quality of care, or access to other sectorsof the healthcare system, such as preventive services.Health services utilization is influenced by supply. Wemeasured the supply factor using distance from home tothe nearest medical center. However, other supply factors (suchas number of doctors, community health programs, and type ofmedical center) were not assessed.

In conclusion, rural and urban disparities in utilizationof healthcare exist in China. Rural residents use physiciansmore but hospitals less than urban residents, particularlyamong uninsured populations. Although our study was un-able to directly assess the role of insurance quantity on theseutilization patterns, reforming the healthcare system towardsuniversal insurance coverage and increasing the amount in-sured, while increasing the population’s education level, is anapproach that should be considered at the national level. Inaddition, access issues in remote areas and by rural minority

Chinese population should be addressed. This study onlyanalyzed utilization of healthcare services; assessment ofquality of care is imperative for future research.

ACKNOWLEDGMENTSThe authors thank the China Ministry of Health for

providing the data for the analysis.

REFERENCES1. Information Centre, Health Ministry of China. China Health Develop-

ment Report, 1997–2001 �in Chinese�. Beijing, China: Health Ministryof China; 2003.

2. China Statistics Department. Statistics of 1% National Sample, 2005.Beijing, China: China Statistics Department; 2006.

3. Zimmer Z, Kwong J. Socioeconomic status and health among olderadults in rural and urban China. J Aging Health. 2004;16:44–70.

4. Research Project Team of China Development Research Center (CDRC)under the State Council. Evaluation and recommendation to Chinahealthcare system reform. China Development Evaluation and Discus-sion [in Chinese]. 2005;(Suppl 1):1–254.

5. Hsiao WC. The Chinese health care system: lessons for other nations.Soc Sci Med. 1995;41:1047–1055.

6. Liu Y, Hsiao WC, Eggleston K. Equity in health and health care: theChinese experience. Soc Sci Med. 1999;49:1349–1356.

7. Liu G, Nolan B, Wen C. Urban Health Insurance and Financing inChina. Washington, DC: The World Bank; 2004.

8. Zhang L, Wang H, Wang L, et al. Social capital and farmer’s willing-ness-to-join a newly established community-based health insurance inrural China. Health Policy. 2006;76:233–242.

9. Liu Y, Rao K. Providing health insurance in rural china: from researchto policy. J Health Polit Policy Law. 2006;31:71–92.

10. Gao J, Qian J, Tang S, et al. Health equity in transition from planned tomarket economy in China. Health Policy Plan. 2002;17(suppl):20–29.

11. Wang H, Yip W, Zhang L, et al. Community-based health insurance inpoor rural China: the distribution of net benefits. Health Policy Plan.2005;20:366–374.

12. Henderson G, Jin S, Akin J, et al. Distribution of medical insurance inChina. Soc Sci Med. 1995;41:1119–1130.

13. Liu GG, Zhao Z, Cai R, et al. Equity in health care access to: assessing theurban health insurance reform in China. Soc Sci Med. 2002;55:1779–1794.

14. Akin JS, Dow WH, Lance PM. Did the distribution of health insurancein China continue to grow less equitable in the nineties? Results from alongitudinal survey. Soc Sci Med. 2004;58:293–304.

15. Gu X, Bloom G, Tang S, et al. Financing health care in rural China:preliminary report of a nationwide study. Soc Sci Med. 1993;36:385–391.

16. Gao J, Tang S, Tolhurst R, et al. Changing access to health services in urbanChina: implications for equity. Health Policy Plan. 2001;16:302–312.

17. Wu J, Liu Y, Rao K, et al. Education-related gender differences in health inrural China. Am J Public Health. 2004;94:1713–1716.

18. Henderson GE, Akin JS, Hutchinson PM, et al. Trends in health servicesutilization in eight provinces in China, 1989–1993. Soc Sci Med.1998;47:1957–1971.

19. Akin JS, Dow WH, Lance PM, et al. Changes in access to health care inChina, 1989-1997. Health Policy Plan. 2005;20:80–89.

20. Seidel JE, Beck CA, Pocobelli G, et al. Location of residence associatedwith the likelihood of patient visit to the preoperative assessment clinic.BMC Health Serv Res. 2006;6:13.

21. Piette JD. Moos RH. The influence of distance on ambulatory care use,death, and readmission following a myocardial infarction. Health ServRes. 1996;31:573–591.

22. Meden T, St John-Larki C, Hermes D, et al. MSJAMA. Relationshipbetween travel distance and utilization of breast cancer treatment in ruralnorthern Michigan. JAMA. 2002;287:111.

23. Institute of Medicine. Unequal Treatment: What HealthCare ProvidersNeed to Know About Racial and Ethnic Disparities in Health Care.Washington, DC: National Academy of Sciences; 2002.

24. China Ministry of Health. The Third National Health Services SurveyDesign �in Chinese�. Beijing, China: China Ministry of Health; 2003.



25. Chai M, Yao K, Qian J, et al. Three National Health Services Surveys:trend analysis of major diseases �in Chinese�. China Hosp Stat. 2005;12:107–111.

26. Aday LA, Andersen R. A framework for the study of access to medicalcare. Health Serv Res. 1974;9:208–220.

27. Andersen RM. Revisiting the behavioral model and access to medicalcare: does it matter? J Health Soc Behav. 1995;36:1–10.

28. Lin JD, Wu JL, Lee PN. Healthcare needs of people with intellectualdisability in institutions in Taiwan: outpatient care utilization and im-plications. J Intellect Disabil Res. 2003;47:169–180.

29. Guo S, Wang L, Yan R. Utilization of health service in women withreproductive tract infections in urban and rural areas �in Chinese�.Zhonghua Liu Xing Bing Xue Za Zhi. 2002;23:40–42.

30. Ruy H, Young WB, Kwak H. Differences in health insurance and health

service utilization among Asian Americans: method for using the NHISto identify unique patterns between ethnic groups. Int J Health PlannManage. 2002;17:55–68.

31. McNutt LA, Wu C, Xue X, et al. Estimating the relative risk in cohortstudies and clinical trials of common outcomes. Am J Epidemiol.2003;157:940–943.

32. Spiegelman D, Hertzmark E. Easy SAS calculations for risk or preva-lence ratios and differences. Am J Epidemiol. 2005;162:199–200.

33. The Centre for Statistics and Health Information MoHC. Abstract of thereport on the 3rd national health services investigation and analysis �inChinese�. Chin Hosp. 2005;9:3–11.

34. Lahelma E, Martikainen P, Laaksonen M, et al. Pathways betweensocioeconomic determinants of health. J Epidemiol CommunityHealth. 2004;58:327–332.



Documents

Rural and Urban Disparity in Health Services Utilization ...people.ucalgary.ca/~lu/MedicalCare_ChinaHealthInsurance.pdf · ORIGINAL ARTICLE Rural and Urban Disparity in Health Services