Health care in China : the availability, utilization and cost of county hospitals

PUN Technical Note 86-16 HEALTH CARE IN CHINA: THE AVAILABILITY, UTILIZATION AND COST OF COUNTY HOSPITALS by Manon Muller and Mary Young April 1986 Population, Health and Nutrition Department World Bank The World Bank does not accept responsibility for the views expressed herein which are those of the author(s) and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsover on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city area, or of its authorities, or concerning the delimitations of its boundaries, or national affiliation. PHN Technical Note 86-16 HEALTH CAR IN CHINA: THE AVAILABILITY, UTILIZATION AND COST OF COUNTY HOSPITALS ABSTRACT This paper analyzes the data collected for the World Bank's first health project in China in 46 counties located in the provinces, Heilongjiang, Shandong, Sichuan and in Ningxia Hui Autonomous Region. The county hospital health care supply, utilization and cost data were analyzed and regression analyses were carried out using economic development indicators and other characteristics of the counties, The supply of health care facilities is correlated positively with income and urbanization. However, this positive income-health facilities availability relationship may have been partially off-set by the large investments allocated from the central government to the counties with large minority populations, which in general are also poor counties. The data also support the conjecture that demand for health care provided by the county hospital is a function of the per capita availability, which suggests an unmet demand for higher level care. Prepared by: Manon Muller Energy Department and Mary Young Consultant to the World Bank April 1986 HEALTH CARE IN CHINA The Availability, Utilization and Cost of County Hospitals Table of Contents Page Introduction.......************************ ********......... . 1. Health Care Facilities......*...................cl....s.....33 2. Health Care Utilization..............9...............,..........9 3. The Cost of Health Care**********************2**................22 4. Conclusion.*. .*.*.*..*.*. . *.*. ............... 26 INTRODUCTION The rurgL health care system for the Chinese counties is based on a three-tier system, with brigade health stations at the base, commune health centers in the middle and the county hospitals at the top. Additional institutions at the county level are the maternal and child health centers and the epidemic prevention stations. In this paper we analyze the county hospitals. The data was collected for the World Bank's first health project in China, in 46 counties locates in the provinces, Heilongjiang, Shandong, Sichuan and in Ningxia Hui Autonomous Region. Altogether, the counties surveyed represent a total population of 31.5 million, only a little over 3 percent of the total Chinese popualation. About 90 percent of the population in these 46 counties are rural. Because factors such as managerial capability and logistical feasibility may have influenced the selection of these 46 counties, the data set cannot be considered a random sample of China's 2,136 counties. Consequently, our findings should not be considered representative of the Chinese county hospital system as a whole. The results, however, are indicative of the variation among counties existing in China's health care system. Sichuan, located in the southwest region of China, is one of the most densely populated provinces in China, with income below the national average. Agriculture is the major activity of the province. Availability of beds and health workers is below the national average. Heilongjiang is China's northernmost province; its labor force is evenly split between agriculture and industry. Twenty-two percent of its crop is mechanically harvested, as compared to 3 percent nationwide. The -2- availabijity of its health facilities is above the national average. Ningxia is one of China's poorest ard least developed region. About 53 percent of the population is illiterate. Agriculture is the mainstay of the region. Compared with the national averages the number of beds per population is low but the number of health workers is high. Shandong, located in the east of China, is the third most populated province in China. The distributed collective income is about 95 percent of the national average. The availability of health facilities (including beds, and health workers) is below the national average. Using data from 1975, 1980 and 1982, this paper will analyze county hospital health care supply, utilization and cost data, and relate them to economic development indicators and other characteristics of the counties. Section 1 presents the summary statistics of the health care facilities of 46 counties. Section 2 and 3 analyze the health care utilization and cost and Section 4 sets out conclusions. -3- 1. Health Care Facilities The structure of the health system at the county level is similar for all four provinces. 4t the county level, there is at least one general hospital, an epidemic prevention station, a maternal child health center and a health training school. Dependent on the population size of the county, there may also be a second general county hospital, and other specialized state hospitals. On average in our sample, a county hospital has 205 beds in 1982 but the variation among counties is large. The smallest hospital has 69 beds and the largest has 300 beds. While the average number of senior western doctors per hospital is 49, their number ranges from 20 to 90. The number of County hospital staff ranges between 81 and 352, with an average of 210. Although the magnitude of the growth of the health facilities varies among provinces and counties, improvement in the avilability of the health facilities is noted in all the counties. The number of health professionals increased 44 percent while the number of hospital beds increased 25 percent. Mean, 1982 Percent Change, 1975-1982 Beds 205 +25 Total staff 210 +44 Western doctors 49 +49 Nurses 72 +45 Work area in mu 6,563 +33 Between 1975 and 1982 almost half the county hospitals acquired one or more '-4- new x-ray machines, thereby increasing the average number of x-ray machines per hospital to two. Table 1 provides information on the growth of hospital beds and health workers in relation to population growth. The average number of beds per thousand population increased from 0.290 to 0.342. Not only did the number of beds/population ratio increase, but the standard deviation also decreased from 0.160 to 0.143, thus reducing the difference in hospital bed availability among counties. Table 1.1: Health Care Availability and Utilization Per 1,000 Population; Averages over 46 Counties; 1975, 1980, 1982 (standard deviation in parentheses) 1975 1980 1982 County hospital beds/ 0.290 0.322 0.342 000 population (0.160) (0.138) (0.143) CHC beds/ 0.588 0.702 0.702 000 population (0.205) (0.232) (0.221) Total county 1.205 1.487 1.580 health staff/000 (0.345) (0.336) (0.338) Doctors/ 0.085 0.106 0.120 000 population (0.126) (0.134) (0.157) Outpatients visits/ 261.852 282.772 314.707 000 population (132.437) (133.146) (161.708) Inpatients/ 7.656 9.023 9.570 000 population (3.675) (3.710) (4.045) -5- While total county health staff per thousand population irreased by 31 percent betweeR 1975 and 1982, the number of county hospital staff and county hospital doctors per population increased by 59 percent and 41 percent, respectively. This points to a major health policy effort towards strengthening the county hospitals. The utilization indicators, such as inpatient and outpatient visits per population, also show a steady growth. Assuming that per capita demand based on the actual need for health services has not changed markedly between 1975-1982, it seems likely that the increased supply of health services has met a previously latent demand. Using agricultural and industrial output per capita as income and development level indicators, we tested the hypothesis that "counties have the health care system they can afford". We estimated regression equations with per capita output and other county level indicators (percentage of rural population, illiteracy rate and prootion-of minorities) as explanatory variables and each of the health care facilities (per 100,000 population) as the dependent variables. The squared income terms were added to the equations in search for nonlinearities. And finally, dummy variables were used to capture province-specific effects. The findings of these regression estimations were consistent over time, the regressions for 1975, 1980 and 1982 reveal the same pattern. Therefore only the 1982 results are shown in Table 1.2 below. Table 1.2: Regression Results: The Availability of Health Care Facilities (1982) (standard errors in parentheses) Hosp. beds/ flosp. beds/ Hosp. staff/ Hosp. staff/ No. doctors/ No. doctors/ 100,000 pop. 100,000 pop. 100,000 pop. 100,000 pop. 100,000 pop. 100,000 pop. Agr PC 0.170 -- 0.204 -- 0.049 -- (0.064) (0.073) (0.019) Agr. PC -0.148 -- -0.216 -- -0.052 -- Sq/1000 (0.097) (0.110) (0.028) Ind. PC 0.087 -- 0.097 -- 0.029 -- (0.048) (0.05) (0.014) Ind PC -0.132 -- -0.175 -- -0.047 -- Sq/000 (0.081) (0.092) (0.024) Distributed -- 0.164 -- 0.148 -- 0.040 income (0.059) (0.069) (0.018) Distributed -- -0.150 -- -0.142 -- -0.039 income (0.098) (0.113) (0.029) or Sq/1000 Percent rural -0.540 -0.290 -0.989 -0.793 -0.233 -0.199 (0.313) (0.281) (0.357) (0.327) (0.091) (0.083) Percent illiteracy 0.058 -0.325 0.087 -0.174 0.071 -0.012 (0.335) (0.341) (0.382) (0.396) (0.098) (0.101) Percent minority 0.471 0.438 0.490 0.371 0.173 0.155 (0.206) (0.221) (0.234) (0.256) (0.060) (0.065) D1 (Heilongjiang) 4.035 4.493 3.193 4.126 0.083 0.186 (4.206) (4.739) (4.797) (5.413) (1.225) (1.396) D2 (Shandong) 712.921 -9.821 -4.340 -3.119 -2.358 -1.664 (4.595) (4.664) (5.241) (5.413) (1.338) (1.374) D4 (Ningxia) 10.364 -3.680 14.648 3.520 -0.220 -3.572 (11.227) (13.509) (12.805) (15.679) (3.270) (3.979) C 35.729 43.250 69.542 85.589 14.396 20.082 R2 0.698 0.612 0.675 0.566 0.649 0.538 -7- Agricultural and industrial output per capita both have a significant positive effect unnn the availability of health care (county hospital beds, county hospital staff and number of doctors, all per 100,000 population). If we use distributed income as an income indicator instead of output, we get the same positive effect.!' In all cases, the quadratic income variable shows a negative coefficient. The quadratic functions imply that, for instance, the per capita availability of hospital beds will increase with economic growth until agri- cultural output per capita reaches about 570V and until industrial output reaches about 330V. Since the averages of agricultural output and industrial output respectively 281V and 174Y in 1982, we are well in the ascending portion of the curve. Turning to the "social" development indicators, we can see that the percentage of rural population is a powerful negative predictor of the availability of health facilities. A country with a high percentage of rural population shows fewer hospital beds, and hospital staff per p6pulation, than a more urban county. This finding is not surprising. County hospitats are located in towns and their catchment population is essentially urban. However county hospitals do not only perform the role of urban health care facilities, they are also the top of the pyramid health system and handle referrals from CHCs in the rural areas. Moreover, health care staffing as a whole, as opposed to 1/ A different result from the one that J. van der Gaag obtained from an analysis showing that lower level facilities (CHCs, MCH, BFDs) are not dependent upon income. (See J. van der Gaag, Commune Health Care in China. Technical Paper No. 11, China Health Sector Report, World Bank,- June 1983.) -8- county hospital staff alone, shows the same pattern: rural counties have fewer health dare facilities. No significant pattern appears concerning the effect of illiteracy on the availability of health care facilities. But the proportion of minority population appears to be an essential factor influencing health care facilities, with a large, significant and positive effect. This can be explained by the specific features of Mingxia that are different. The five counties in Mingxia are different from other counties in that Ningxia has a large minority population, about 54 percent are Hui (versus 7 percent in the total sample). It is also one of China's poorest and least developed areas. The level of mortality and morbidity is very high, hence the central government allocates extra funds to assist its development. The per capita state expenditures on health in Ningxia are higher than the national average. The salaries paid to health workers in Ningxia are higher for comparable positions than the national average in an attempt to encourage health staff to work in the poorer areas. Likewise the central government invests a much Larger amount of resources per capita to improve the health facilities. Hence the variable percentage of minority population accentuates the very special features of Ningxia and thus cannot be generalized to other provinces. The province specific variations expressed by the dummy variables (Dl =,Heilongjiang, D2 = Shandong, D4 = Ningxia, while Sichuan is our reference province) are large but not consistently significant. On the whole it appears that Heilongjiang enjoys more health facilities, while Shandong has less facilities. Results are indeterminate for Ningxia, but the variable--percen- tage of minority population--has probably accounted for much of the variation in the availability of health facilities specific to Ningxia. -9- The analysis has focussed so far on existing facilities. A Look at future trends in health care provided by total county capital construction and county hospital construction seems to confirm the general pattern discussed above. The more urban counties and the counties with the highest proportion of minority population enjoy the largest investments in capital construction (see Tables 1.3 and 1.4). The income variables however show a mixed picture: only the results with respect to industrial production imply an unambigous increase in investment for health-care facilities when income increases. 2. Health Care Utilization A look at the pattern of county hospital health care utilization (Table 2.1) reveals a much lower number of outpatient visits per capita (0.315 on average or one visit every three yea-s per capita) than what has been previously noted in the CHCs!/ (1.8 in Qufu, 3.7 in Pingluo county). This reflects the fact that the catchment area of county hospitals does not include the whole county and that most outpatient care can be handled at the CHC and brigade health Etation level. The outpatient per staff statistic, an indicator of the workload of the health staff can be used as an indirect measure of quality of services provided by health workers. The average of 3.1 outpatient visits per staff 1/ See J. van der Caag, "Rural Health Care in China, The Case of Qufu County", mimeo, World Bank, January 1983. Table 1.31 Total County Capital Construction per/000 Popul ion Distrib. Agr. Ind. Sq. agr. Sq. ind. Rural/ Agr. I % lieilong- Shan- inc. prod. PC prod. PC PC/1000 PC/1000 tot. pop. prod/mu illit. minor. iang dong C 1982 0.853 4.643 -1.303 -7.083 -2135.776 -- -- -- -- -- 1565.509 0.455 (2.975) (1.738) (4.431) (2.953) (791.553) 1982 0.795 -- -- -- -- -1868.666 -- 8.490 76.778 -- -- 1519.461 0.636 (0.451) (683.579) (7.679) (15.158) Table 1.4. County Hospital Capital Construction per/000 Population Distrib. Agr. Ind. Sq. agr. Sq. ind. Rural/ % % lieilong- Shan- inc. prod. PC prod. PC PC/1000 PC/1GO tot. pop. illit. minor. jiang dong C &2 1975 0.208 -0.,482 0.410 1.719 -32.652 -- -- -- -- 35.425 -0.072 (0.982) (0.943) (2.359) (3.517) (166.028) 1980 2.950 2.662 -5.136 -5.062 -1012.613 -- -- -- -- 458.304 0.284 (3.748) (1.731) (7.776) (3.410) (545.424) 1982 3.102 -1.123 -3.666 4.061 -2137.868 -- -- -- -- 1568.987 0.338 (3.004) (1.755) (4.474) (.2.982) (805.274) 1982 0.742 -- -- -- -- -157.544 8.602 96.094 -- -- -131.285 0.794 (0.311) (471.032) (5.292) (10.445) - 11 - Table 2.1 Health Care Utilization-Statistics 1975 1980 1982 Mean Min Max Mean Min Max Mean Min Max Outpatient 0.262 0.076 0.585 0.283 0.103 0.658 0.315 0.094 0.772 per pop Inpatient 7.656 2.740 16.170 9.023 3.268 20.608 9.570 3.880 21.411 per av pop % Referral 49.5% 49.6% 52.1% among inpatients Average 11.35 7.2 16.5 11.44 8.2 17.5 11.9 7.9 14.4 duration of stay Outpatients 4.5 1.7 8.2 3.3 1.7 5.8 3.1 1.5 6.1 per staff per day* Outpatients 23.5 8.8 93.9 16.3 7.3 43.6 13.7 6.4 33.5 per doctor per day Inpatient 28.2 15.8 46.9 28.4 16.6 40.7 28.4 18.8 51.3 per bed per year Bed occu- 310.1 153.3 449.9 316.8 200.7 380.3 329.0 253.7 405.3 pancy (days per year) 48 weeks, 6 days a week. - 12 - per day, in the county hospitals is about the same as the commune health centers,1/ but thg range is very wide, from 1.5 to 6.1. The work Load is surprisingly low for a county hospital which receives referrals in addition to patients who use the outpatient clinic as the primary source of health care. Increased staffing has brought the workload down from 4.5 visits in 1975 to 3.1 in 1982. Although there is no predetermined ideal patient to staff ratio, the present ratio of three patients per health staff per day suggests inefficient utilization of available health staff. A drastic change has been achieved in the outpatient per doctor per day ratio. The patient to doctor ratio was 23.5 in 1975; the ratio decreased to an average of 13.7 in 1982. The change in work load has been particularly marked in the "doctor-poor" hospitals where work Load dropped from'93.9 to 33.5 patients per day. The contrast between the doctor to patient and the health staff to patient ratio distinctly shows that there is still a shortage of higher level technical staff as compared to the availability of junior level, senior level and supportive staff. The admission rate (number of inpatients per thousand population) ranges between 3.88 and 21.41 with an average of 9.57. About 52 percent of inpatients are referred from Lower Level rural health care facilities. The mean duration of stay is 11.9 days which is much higher than in CHCs (5.1 observed in Qufu CHCs). This must be i reflection upon the seriousness of cases treated in county hospitals as compared to CHCs and also maybe upon the availability of health facilities (this Latter issue will be addressed below). 1/ Jacques van der Gaag, "Rural Health Care in China, The Case of Qufu County", mimeo, World Bank, January 1983. - 13 - We tested the hypothesis that health care utilization is d;ictated by the socioeconomic level and by the availability of health care facilities. To do this we estimated regression equations with agricultural and industrial output (or, alternatively, distributed income), percentage of rural population, of minority and of illiteracy as well as availability of facilities as explanatory variables to determine health care utilization. We also introduced province dummy variables (see Tables 2.2 and 2.3). Availability proves to be a more powerful predictor of health care utilization than socioeconomic demand factors (Income level has a small significant positive effect only for outpatient visits). Staffing (staff per thousand population) is a very strong predictor of outpatient visits (see Table 2.3). The addition of one county hospital health staff per thousand population will increase outpatient visits per thousand population by about 800. This means that each new staff would see 2.8 -outpatients per day which is barely lower than the workload of 3.1 outpatients per existing staff. The number of inpatients and hospital days are strongly determined by bed availability (see Tables 2.3 and 2.5). As much as 90 percent of the variation in hospital days among counties and 74 percent of the variation of inpatients per thousand population are explained by availability variables only. Each additional bed will be used 287 days per ye,ar (regression equation with agricultural and industrial outputs, Table 2.5) and accommodate 27 patients per year. However, the number of staff per bed--a more qualitative predictor--did not prove to have any significant effect upon the demand for inpatient care (the results are not shown here). -14- Table 2.2: Reg-ession Results: Outpatient Visits and Availability (1982) Outpatients/000 Outpatients/000 Outpatients/000 population population population Agr. PC -- 0.492 -- (0.255) Ind. PC -- 0.096 -- (0.171) Distrib. income -- -- 0.408 (0.228) Percent rural -- -2.146 0.829 (3.966) (3.243) Percent minority -- 3.796 1.478 (4.043) (3.676) Percent illiteracy -- 8.070 7.836 (2.411) (2.347) Staff/000 883.354 729.692 799.868 (129.405) (162.666) (149.057) D -123.234 -150.543 -144.767 (49.392) (51.334) (50.730) D2 -91.311 -173.746 -144.658 (37.846) (56.423) (47.536) D4 -88.442 -464.576 -524.200 (62.636) (136.038) (141.040) C 70.262 83.263 -73.070 R2 0.577 0.653 0.651 - 15 - Table 2.3: Reg-resrion ResiIts: Inpatient Visits and AvailabiLity (1982) Inpatients/000 Inpatients/000 Inpatients/000 Inpatients/000 population population population population Agr. PC -- -0.004 -- -- (0.005) Ind. PC -0.003 -- -- (0.003) Distrib. income -- -- 0.0004 -- (0.0048) Percent rural -- -0.059 -0.074 -- (0.072) (0.060) Percent illiteracy -- -0.089 -0.050 -- (0.080) (0..073) Percent mi.nority 0.011 0.022 -- (0.048) (0.047) Hospital beds/000 26.999 27.507 25.071 24.565 (2.587) (3.770) (3.383) (2.137) Dl -1.648 -2.537 -2.481 -- (0.861) (1.017) (1.009) D2 -1.734 -0.152 -1.213 -- (0.677) (1.017) (0.973) D4 -3.054 -3.288 -3.202 -- (1.095) (2.667) (2.776) C 1.599 10.165 10.161 1.175 20.787 0.785 0.782 0.745 - 16 - Table 2.4: Regression Results: Mean Stay and Availability (1982) Average Duration Average Duration Average Duration of stay in of stay in of stay in county hospital county hospital county hospital Agr. PC -- 0.004 -- (0.004) Ind. PC -- 0.002 -- (0.003) Distrib. income - -- 0.001 (0.004) Perceat rural -- 0.033 0.048 (0.060) (0.050) Percent illiteracy 0.120 0.090 (0.067) (0.061) Percent minority -0.023 -0.030 (0.040) (0.039) Hospital beds/000 -2.965 -3.014 -1.342 (2.214) (3.156) (2.823) Dl 1.850 2.636 2.602 (0.737) (0.851) (0.843) D2 0.876 -0.674 0.079 (0.580) (1.010) (0.813) D4 2.111 2.513 2.271 (0.937) (2.232) (2.320) C 11.996 5.192 5.003 0.086 0.116 0.109 - 17 - Table 2.5: Regression Results: Hospital Days and Availability (1982) Hospital Hospital Hospital Hospital days/000 days/000 days/000 days/000 population population population population Agr. PC -- 0.017 -- -- (0.033) Ind. PC -- -0.007 -- -- (0.021) Distrib. income -- -- 0.023 -- (0.029) Percent rural -- -0.381 -0.490 -- (0.444) (0.364) Percent illiteracy -- 0.368 0.525 -- (0.492) (0.444) Percent minority -- -0.088 0.035 -- (0.299) (0.285) Hospital beds/000 778.367 286.951 269.437 286.206 (15.590) (23.322) (20.477) (13.325) Dl 5.638 3.096 3.146 -- (5.187) (6.291) (6.107) D2 -9.291 -7.346 -13.387 -- (4.081) (7.466) (5.888) D4 -13.372 -16.102 -19.664 -- (6.597) (16.496) (16.802) C 19.356 48.207 51.408 13.420 R2 0.931 0.926 0.929 0.911 - 18 - These results seem to point to a shortage in county hospital facilities or, at jeast, to an unmet latent demand. Yet the mean duration of stay (Table 2.4) does not depend upon bed availability (or has an unsig- nificant negative response to availability). This is surprising, because, if the bed shortage was really acute, the mean stay would be shorter when avail- ability is small so that the facilities could accommodate more inpatients. This might mean also that the lower acceptable limit of mean stay has been reached and cannot be shortened any more. These results1differ markedly from what has been found at the CHC level, where the mean stay increases with availability. The general following mathematical expression is always true: (Inpatients/Population) x (Mean stay/Patient) - Hospital days/Population. The log-log form is: Log admissions + Log stay = Log hospital days. The log-log regressions (excluding the dummy variables) show that hospital stay elasticity is 0.946 (i.e., almost unity). They show also that 0.929 of it is due to admission elasticity with mean stay elasticity close to zero (0.017) (results are not shown here). This reinforces our suspicion that bed shortage is acute in the 46 counties studied. Province specific factors have a significant negative effect on outpatient and inpatient visits in all three provinces (Heilongjiang, Shandong and Ningxia) compared to our reference province, Sichuan. Yet duration of stay of hospitalized patients is significantly longer in Heilongjiang. - 19 - Referral figures can test whether inpatients bypass CHCs in favor of county hospitals. The results of our analysis suggest that neither income nor county hospital senior staff to staff ratio (a "quality" indicator) are sig- nificant in explaining referrals. Actually, distributed income has a negative (though insignificant) effect upon referrals, which tends to show that Lower level facilities improve with income and can handle more inpatient cases. Once again availability is most important: referrals increase with the number of beds in county hospitals and decrease with the number of beds in the CHCs (see Table 2.6). Table 2.6: Referrals and Availability Distributed Senior/staff Hospital income ratio beds CHC beds C Referrals -0.012 0.535 17.973 -5.299 4.626 0.098 per 000 (0.014) (24.905) (8.282) (5.126) 1982 Since (i) Health facilities are function of socioeconomic variables as we saw in the first section of this paper; Fac. = E (Income) and (ii) Utilization is mainly a function of availability of health care facilities UtiL. f (Fac.), - 20 - we can use a reduced form equation.whereby utilization of health care utilities is specified as a function of socioeconomic variables: Util. = f (Income). We estimated this reduced function by regressing socioeconomic variables and province specific dummies against utilization data. The results show that the number of outpatient and inpatient visits per thousand population seem to depend mainly on the three factors pointed out in the previous section (Table 2.7). They depend namely 1. positively on distributed income; 2. negatively on the proportion of rural population; 3. positively on the proportion of minority population (Table 2.2). In particular, the positive effect of distributed income on visits to county hospitals might support the widespread concern that lower level health care facilities are being more and more bypassed in favor of county hospitals as income grows. We would need more data on CHC utilization to confirm this hypothesis. - 21 - Table 2.7: County jbspital Utilization; Reduced Form Equations Outpatients/000 population Inpatients/000 population Distributed income 0.940 0.019 (0.270) (0.006) Percent rural -6.089 -0.168 (3.915) (0.091) Percent illiteracy 0.092 -0.131 (4.826) (0.113) Percent minority 10.189 0.112 (3.034) (0.071) Dl -101.961 -1.030 (65.923) (1.539) D2 -152.403 -3.104 (62.522) (1.460) D4 -456.613 -2.815 (184.846) (4.316) C 729.118 24.908 R2 0.396 0.474 22 - 3. The Cost of Heatth Care We are unfortunately unable to breakdown health care costs between outpatients and inpatients. Our data shows a total recurrent hospital expenditure of 1.6Y per capita in 1982, 26 percent of which was for staff wages and 44 of which was for drugs (Table 3.1). Data on pharmaceutical prices in the different counties, price evolution, and wage differences are not available. * Tables 3.2 and 3.3 take a globaL Look at the variation of drug and wage costs. They reveal that the main factor of cost increases is the number of inpatients. Over time (since 1975) the cost of .rugs have been increasing more and more sharply relative to the increase in outpatient visits. This means that the number of prescriptions per outpatient has been increasing steadily. However, staff wage costs per patient have been rather stable over time. And income indicators do not yield any significant results. Both the cost of drugs and wages are lower when agricultural output per capita increases--which tends to show that agricultural, rural areas have lower wages and give out less prescriptions than industrial urban areas. Finally, it must be pointed out that less than half the variation in health care costs is explained by the number of outpatients and inpatients and by income variables. Therefore important institutional and soc.ioeconomic factors have been left out of the analysis and require further study. -23- Table 3.1: Health Care Cost Summary Statistics ------------------------m---m--wommemewommewo--o-mme-ewomm--we--*------------ 1975 1980 1982 County hospital 0.879 1.247 1..636 recurrent expenditureis per capita of which % wages 21.5% 26% 25.9% % drugs 46.1% 46.0% 44.3% Cost of drugs 1416.9 1751.7 2062.2 per bed Cost of.drugs 1810.0 1943.8 1985.9 per staff Cost of drugs 8928.2 9186.3 8764.8 per senior doctor Table 3.2; Regression Results: Cost of Drugs Agricultural Industrial Outpatient Inpatient Output Output Distributed per 1000 pop. per 1000 pop. per capita per capita Income C K Cost of drugs per 1000 population 1975 0.536 13.865 -- -- -- 140.515 O108 (0.518) (18.678) 1975 0.511 17.067 -0.995 0.755 -- 222.741 0.112 (0.518) (20.206) (0.816) 0.755 -- 222.741 0.112 1980 0.696 38.442 -- -- -- 14.643 0.430 (0.501) (17.981) 1980 0.756 35.004 -0.428 0.381 -- 64.725 0.422 (0.511) (18.361) (0.545) (0.375) 1982 0.900 33.357 -- -- -- 99.093 0.456 (0.416) (16.613) 1982 0.921 31.027 -0.392 0.534 -- 132.103 0.454 (0.417) (16.746) (0.452) (0.406) 1982 0.859 32.251 -- -- 0.529 25.000 0.457 (0.417) (16.637) (0.515) Table 3.3: Regression Results: Wages Expenditures Agricultural Industrial Outpatient Inpatient Output Output Distributed per 1000 pop. per 1000 pop. per capita per capita Income C Cost of wages per 1000 population 1975 -0.014 20.608 -- -- -- 32.029 0.213 (0.244) (8.387) 1975 -0.023 21.816 -0.370 0.278 -- 62.742 0.199 (0.246) (9.609) (0.388) (0.355) 1980 0.391 20.969 -- -- -- 13.525 0.503 (0.241) (8.645) 1980 0.419 19.658 -0.202 0.116 -- 43.944 0.489 (0.247) (8.879) (0.264) (0.181) 1982 0.288 29.528 -- -- -- 36.021 0.427 (0.263) (10.512) 1982 0.317 27.107 -0.503 0.425 -- 117.513 0.458 (0.256) (10.296) (0.278) (0.250) 1982 0.273 29.120 -- -- 0.194 8.828 0.419 (0.266) (10.615) (0.328) - 26 - 4. Conclusions This paper surveyed county level health care in 46 Chinese counties and reached the following conclusions: -- Supply of health care facilities is an increasing function of income and urbanization. At the same time, counties with a large urban population show relatively large investments for health care facilities. On the other hand, counties with a large minority population also show a large supply of health care facilities. This may partially off-set the positive income-availability relationship. - The demand for health care provided by the county hospital is primarily a function of the per capita availability, which in the current context, suggests an unmet latent demand for higher level care.