ELSEVIER

Health Policy 33 (1995) l-14

Health and its determining factors in the Tokyo megacity

Sachiko Takeuchi, Takehito Takano*, Keiko Nakamura
Department Tokya $edical

of Public Health and Environmental

and Dental University, l-5-45

Science, School of Medicine, Yushima, Bunkya, Tokyo 113, Japan

Received 17 August 1994, accepted 20 September 1994

Tbis study was designed to (i) analyze community healthin post-industrial, denselypopulated Tokyo, and (ii) find a statistical-based predictionmodelfor the health level of the communities from the environment.For this purpose, compiledhealth-level we indicatorsfor each city-ward, and indicators for demography,infrastructure, land-use,amenities,education, working conditions, economics, and medicaland welfareservices. healthfactorsand the The environmental factors were obtained by using weighted principal factor analysis; their relation

wasexamined weightedcorrelation analysis by and weightedlinear regression analysis. One to six environmentalfactors correlating significantly to health were found for eachhealthlevel indicator or factor. The prediction model explainedup to 0.98 of the variance of the health-level factorsandindicators. Thisanalysis elicitedhealth-determinant factorsin various fields.The communityhealthlevel waswell reproduced the present by method;the modelwill be conduciveto decisionmaking pertainingto health policies.
Keywords:

Health determinants;

Prediction

model; Megacity

1. Intmduetion

The idea that health is governed to a considerable extent by the living environment is widely accepted and has been discussed in many articles [l-5]. The environments which have a major effect on community health may vary with the social and physical conditions. In post-manufacturing urban areas, where the service industry con*Corresponding author, Tel: +81 3 3813 6111, ext. 3177; Fax: +81 3 3818 7176.

0168-8510/95/SO9.50 0 1995 Elsevier Science Ireland Ltd. All rights reserved SD1 0168-8510(94)00701-F

S. Takeuchi

et al. /Health

Policy

33 (1995)

1-14

stitutes a major part of the industrial structure, most of the health fundamentals are usually provided, and the health status of communities within such an area is characterized by a low birth rate, low death rate, and prolonged life expectancy. The urbanization usually induces intensive land utilization, high land prices, and high living costs; the cities often expand with various inequity problems, including some in health. The central part of Tokyo consists of twenty-three city-wards. Each of the citywards covers an area of between 10 and 60 square kilometers and has a population of between 40 000 and 800 000. All together, the twenty-three wards constitute a densely populated urban area of 8 163 573 people in 618 km2, which is located in the center of the largest conurbation of the more developed regions defined by the United Nations [6]. The population density, 13 214 persons per square kilometer, far surpasses that of the national average of 327 persons per square kilometer, which itself is the fourth highest among the nations with a population larger than five million [6]. All the land has already been developed, and the living environments are principally artificial; only 0.2% of the land is registered as woodland, and park area is 5.2% of the total land area. Each of the twenty-three city-wards is a semi-independent municipality, with its own mayor and council elected by the citizens; each is responsible for making its own city health plan. In the rapidly changing city environments, however, their countermeasures often fall behind. It is an urgent issue to find effective policies aimed at promoting health and reducing inequity in health through the prudent use of the limited resources. It is interesting to study the relationship between health and health determinants in such a post-industrial service society. The present study investigated the indicators of health and those of the environments which may contribute to health; it analyzed their relationship in order to construct a model which can predict the health levels on a statistical basis. The prediction model provides information which helps identify the relevant health determinants and to what extent they contribute to health quantitatively in each community; it will be instrumental in deciding health policies which cover not only the health-care and welfare services but also various social and physical environments.
2. Method 2.1. Compilation and categorization of indicators

We compiled indicators which represent the level of health in each city-ward (health-level indicators) and various environmental indicators of each ward as candidates of the health determinants (environmental indicators). Data for the indicators were obtained by scrutinizing statistics from the national government and the local government, as well as from the government of each city-ward. We examined the following statistics for use in the present analysis: Vital Statistics Japan, Population Census of Japan, Housing Survey of Japan, Establishment Census of Japan, Report on School Basic Survey, Annual Report on Public Health in Tokyo, Statement of Accounts in Special Wards, Road Survey in Tokyo, Special Ward

S. Takeuchi

et al. /Health

Policy

33 (1995)

1-14

Statistics, Statistics of the Metropolitan Police Department in Tokyo, and others. We used data taken as close to 1990, the year of the most recent population census, as possible. The health-level indicators we selected include the age-adjusted death rate (standard mortality ratio; SMR), the extent of inequity in health, the absenteeism rate for students and workers, per capita cigarette tax revenue, the number of traffic accidents and fires, and the number of criminals. The environmental indicators we selected include demographic indicators, infrastructure conditions, educational level, working and living environments, Iinancial indicators, and welfare and health-care services. Indicators which are uniform over the concerned area were not used in this analysis; i.e., the availability of piped water and the rate of children attending elementary or junior high schools. To obtain clear meaning for each factor, we classified the environmental indicators into six categories; i.e., demography, infrastructure, amenities, education, local activities, and health-care and welfare services.
2.2. Inequity indicators

Among the indicators, we here specifically mention two which represent the degree of inequity in health within a ward. The first one is age-adjusted PMI, the proportional mortality indicator directly age-adjusted by using a five-year-interval death rate of the observed area with the Japanese standard population. The divided age of the commonly used PM1 is 50 years old. In addition to that, we used PMIs whose divided ages were 65,80, and 85 years old (defined as the age-adjusted PM165 the age-adjusted PM180, and the age-adjusted PM185, respectively). In Tokyo, where few people die before 50 years of age, the age-adjusted PM180 and PM185 are considered to be appropriate indicators representing inequity in health. The longevity differential index is another indicator which more directly represents the inequity in health; it is defined by the standard deviation of the ageadjusted death rate as a function of age, which is also calculated by using a tive-yearinterval death rate with the Japanese standard population. To make the indicator reflect the inequity more clearly, we omitted the deaths of those under five years of age.
2.3. Principal factor analysis

We performed the principal factor analysis weighted by the population of each ward to obtain the factors which express the comprehensive characteristics of each city-ward in health or in the environments we considered. The factors were taken from the health-level indicators or indicators in each environmental category separately and were rotated with the varimax operation. The number of factors for each category was determined as follows. First, we imposed the condition that the eigenvalue of the correlation matrix should be more than one. We then adjusted the number of factors in the categories where the meaning of the factors was not clear according to that condition. The sign of health-level factors were taken so that a positive value means a good health level. Because many of the indicators had similar distributions, it was not necessary to

S. Tokeuchi et 01./Health Policy 33 (1995) 1-14

use all of the indicators. As for the environments, we further selected a set of six to nine indicators for each category so that the selected indicators would give similar factors to those achieved by the full-set analysis; we did this by checking Pearson correlation coefficients between the factors and the full-set indicators. In the following we used the factors obtained from the selected indicators.
2.4. Correlation analysis

The correlation analysis was performed to find the environmental factors which correlate significantly to health (the correlating factors). The analysis was weighted by the population of each ward. We looked into Pearson correlation coefficients with the associated probability (P-value) between the environmental factors and the health-level factors, as well as the health-level indicators. By looking into each distribution, we also determined whether the assumption used in the following regression analysis - i.e., that the health-level factors depend linearly on the environmental factors - is correct.
2.5. Linear regressionanalysis

To find a model which can predict each health-level indicator or factor from the environmental factors, we performed linear regression analysis weighted by the pop ulation. The health-level factors and indicators were assumed to depend linearly on the independent variables. We restricted the number of the independent variables by checking the eigenvalues of the correlation matrix for all of the environmental factors together; it should be less than six, the number of the eigenvalues greater than one. The independent variables were chosen from the environmental factors by examining the rate of the variance explained by the model, R2, and the significance probability for each independent variables, P, with the following procedure: (i) find a n-variable model which gives the best R2 for each n I 6; (ii) choose a model of the largest n which satisfies the two conditions: the sum of P-values of all n independent variables should be less than 0.05, and the model does not contain the environmental factors which correlate to each other with P < 0.001; (iii) when the n chosen above is 5 or 6, we further looked into whether the R2 was improved significantly from the model of the next largest n. In the following, we call the environmental factors chosen by this procedure the predicting factors, which are considered to be the health-determinant factors. 3.Result.s
3.1. Indicators

We found twenty-seven indicators which represented the health level and one hundred and seven indicators for the environments (Table 1).
3.2. Principal factor analysis

There are five factors obtained (Hl -H27), three factors (FDl-FD3)

(FHl-FH5) for the health level indicators for demography from six selected indicators

Table 1 Pearson correlation coefficients between the factors and the original indicators for the health determinants FHl -O.SO** FH2 FH3 FH4 FH5

Health level (C = 0.87)

-0.98** -0.98** -0.90** -0.93. -0.92** -0.96** 0.82** 0.84** -0.41. -0.50* 0.71** 0.76** 0.71** -0.72** -0.96** -0.94** -0.83* -0.59** 0.77+* 0.83** 0.52* 0.74** 0.58** -0.56** -0.59** -0.93;; -0.93** -0.86** -0.84** -0.55** FDI FD2 -0.47* FD3 -0.47*

HI H2 H3 H4 H5 H6 H7 H8 H9 HlO HI1 HI2 H13 HI4 HI5 HI6 HI7 HI8 H19 H20 H21 Hi22 H23 H24

H25 H26 H27

Death rate SMR (male) SMR (female) SMR (malignant neoplasm) SMR (malignant neoplasm of stomach) SMR (ischemic heart disease) SMR (cerebrovascular disease) PM1 PM165 PM180 PM185 Birth rate Stillbirth rate Longevity differential index (both sides) Longevity differential index (half) SMR (suicide) Student absenteeism rate (elementary school) Student absenteeism rate (junior high school) Worker absenteeism rate Marriage rate Divorce rate Per capita cigarette tax revenue Traffic accidents per kilometer of road Number of persons injured or killed by traffic accidents per kilometer of road Number of criminals per 1000 people (traffic violation excluded) Number of vicious criminals per loo0 people Number of fires per 10 000 buildings

ul

Demography (C = 0.84)

S. Takeuchi

et al. /Health

Policy

33 (1995)

1-14

: z d : P d

t 8 d

: 2 d

0.72** 0.89 0.82**

0.80** -0.80 FW2 FW4 FW3 Fws 0.97** -0.84

El E2 E3 E4 ES E6 E7 E.8 FWl

Number of students per elementary school Number of students per junior high school Number of students per high school Employment rate of junior high school graduates 0.88** 0.91. Employment rate of high school graduates -0.91.. Age-adjusted educational level Sex ratio of age-adjusted educational level (younger than 50 years of age) Number of libraries per 100 000 people

Local activites (C = 0.95)

0.91.. 0.89** 0.90** -0.47 0.77.. -0.62** 0.78** FSl 0.92** 0.95** 0.972 0.52 0.78 -0.75** -0.46 0.74++ 0.93** FS2 0.46* FS3 FS4 0.95** -0.54**

Wl w2 w3 w4 W5 W6 w7 W8

w9

Unemployment rate Number of employees as percentage of population Tertiary industry workers as percentage of total work force Managers and officials as percentage of total work force Number of employees as percentage of total work force Number of establishments per 1000 people Growth rate of total work force Small-scale establishments (less than 4 employees) as percentage of all businesses Per capita resident tax

Health care and welfare services (C = 0.89)

Sl s2 s3 s4

S5 S6 s7

Number of medical doctors per 100 000 people Number of hospital or clinic beds per 100 Ooo people Welfare-assistance recipients as percentage of total households Percentage of medical-assistance expenditures of the total welfareassistance expenditures Percentage of people attending health examinations held by clinics Emergency ambulance calls per 100 000 people Percentage of dwelling units located more than 500 meters from medical facilities

P < 005.I P

< 001 *

S. Takeuchi

et al. /Health

Policy

33 (1995)

1-14

(Dl-D6), three (FIl-F13) for infrastructure from seven indicators (Il-17), three (FAl -FA3) for amenities from nine indicators (Al-A9), four (FE1 -FE4) for education from eight indicators (El-E@, five (FWl-FWS) for local activities from nine indicators (Wl-W9), and four (FSl-FS4) for health-care and welfare services from eight indicators (Sl-S8). The selected indicators with their Pearson correlation coefficients to the factors are listed in Table 1. The cumulative proportion of variation by the factors, C, is also listed for each category. The indicators are well represented by the obtained factors: they explain from 84 to 95% of the total variance of the selected indicators in each category. 3.3. Correlation analysis The environmental factors which have significant correlation (P < 0.05) to the health-level factors and health-level indicators are found in Table 2. Each healthlevel factor correlates differently to the environmental factors. All the health-level factors and indicators except for FH5 have correlating factors in the environments. FH I, FH2 and FH3 correlate to the factors of all six categories, while FH4 correlates to five categories. Twenty-three of the twenty-seven healthlevel indicators have correlating factors in more than one category. With only one exception, H25, there are no health-level factors or indicators which correlate only to the factors of health-care and welfare services. The correlating factors for PM180 and PM185 (HlO and Hl l), as well as those for the longevity-differential index (the inequity indices newly introduced in this work), distribute to more than four of the six categories. SMR has significant correlation to all categories considered. In Table 3, we also list the environmental factors which have correlation to each other with P c 0.001 (Pearson coefficients > 0.65). 3.4. Linear regression analysis Among the health-level factors and indicators, those more than half of whose variance is explained by the obtained model are listed in Table 4 together with their predicting factors and R2. The model could predict from 58 to 98% of the variance of the health-level factors. All health-level indicators except for three are well predicted: up to 95% of the variance is explained. 4. Discussion The geographical heterogenity the health of people can be seen clearly in Tokyo and it is growing. The estimates from the death rate show that, in 1990, the average male life span differed by 2.7 years between the best and the worst city-ward; it was 2.4 years in 1985. For females, it grew from 1.7 years to 2.4 years. These differences in life span should be considered serious, since they are comparable to the causedeleted life expectancy for malignant neoplasms, the most frequent cause of death in Tokyo; that is 3.9 years for males and 2.9 years for females. This inhomogeneity is not just a fluctuation, as the differences are far larger than statistical error allows for which is at most 0.2 years. They are not considered to be an artifact either, because a corresponding difference is also seen in the standard mortality ratio (SMR).

S. Takeuchi et al. /Health Policy 33 (1995) 1-14

++z
+

I I
I I

I I $Z

$$$ $T ++ ++
I

I I $

+ I I I I I I

+
I I I I I+ I++

I +
I +

++

: +

I +

I I I

++++++ ++++++
I

I I

$ + $ $+ + I +$$
+++

$$$$$$
I I + I

+$
+ +
:

+:+
++ I

+ :

I I ! !

10 Table 3 The health

S. Takeuchi et al. /Health Policy 33 (1995) 1-14

determinant factors

factors

correlated

to each other

The predicting FDI FD2 F12 F13 FA2 FA3 FE1 The factors

Correlated

factors

F13, FA2, FW2, FS2 FA3, FEI, FWI, FS4 FAl FDl, FA2, FSI FDl, F13, FW3 FD2, FWl FD2, FWI, FS4 in the right column correlates to the factor in the left column with P -c 0.001.

Table 4 The predicting

factors

found

by the linear Health-determinant F13, FDl, FIl, F13, FDI, FDI F13, F13, FA3, FEl, FD3, FA3, FA4, FDI, FDI, FDI, FDI, F13, F13, FW4, FD2, F12, FEl, FEI, FSl, FDI, FDl, FD3, FD2,

regression factors

analysis

R2
FHI FH2 FH3 FH4 FHS HI H2 H3 H4 H5 H6 H7 HE H9 HI0 HI1 HI2 HI3 H14 H15 H16 H17 HI8 HI9 H22 H23 H24 H26 H27

0.98
0.63 0.82 0.58 0.68 0.77 0.95 0.94 0.89 0.88 0.91 0.92 0.86 0.89 0.90 0.85 0.94 0.61 0.57 0.50 0.73 0.55 0.72 0.83 0.82 0.88 0.89 0.76 0.91

FA3, FA4, FE1 FD3 FWl, FW3, FW4 FW4 FD3, F12, FW4 FA3, FA3, FEl, FEZ, F13, FEl, FW2, FD2, FD3, FD3 FD2, FW5 FW4 FSl FII FE3 FE2, FS2 FS2, FII, FD2, F12, FIl, FEl, FW4 FE1 FW2, FW3 FE4, FS2 FEI, FS3 FW2 FS4 FII FS4 FAl, FW3

FE3 FS4 FWl, FW4 FD3, FW4 FW2, FW4 F13

S. Takeuchi et al. /Health P&y

33 (1995) 1-14

II

A megacity with inhomogeneity in health offers a valuable opportunity to analyze systematically the relationships between health and its determinants, as well as the dynamics between health and health determinants. The results obtained in the present study show that a major part of the inhomogeneity in health in Tokyo can be explained by the differences in the broad spectrum of environmental characteristics between city-wards. Our analysis identified five components of community health. SMRs significantly correlate to each other and to the age-adjusted PM1 divided at young-old age, to make the factor FHl . FH2 consists of the age-adjusted PM1 divided at old-old age, which correlates negatively to the crude death rate and positively to the student absenteeism rate and the marriage rate. FH3 principally relates to traffic accidents. FH4 is consists mainly of the longevity-differential indices. The stillbirth rate correlates to the number of criminals; this makes the factor FHS. We used health-level indicators which include physical, mental and social health, because the features of urban health may not be properly expressed by health-level indicators in a narrow sense. The five obtained factors explain 87% of the total variance of twenty-seven health indicators. Therefore, the results obtained by using five health-level factors effectively document overall community health from a wide perspective. Furthermore, when evaluating community health, the inequity of health should be addressed. The inequity within a city is one of the key concepts in achieving high health status for the city 171. In our present analysis, we used original indicators which may reflect inequity in health within a city: PMI, PM165, PMISO, PM185, and the longevity-differential index. Results of the factor analysis of health-level indicators, including both the above mentioned original indicators and other health indicators, identify two factors which behave differently from other health-level indicators. One is the age-adjusted PM1 divided at old-old age, and the other is the longevity-differential index. It was interesting to discover that the age-adjusted PMI divided at young-old age shows a different distribution in Tokyo from the one divided at old-old age. The distribution of the age-adjusted PM1 for young-old age has a similar distribution to that of SMRs; that means that age-adjusted PM1 for youngold age and the SMR represent the same aspect of health in Tokyo. We identified two components of health-level factors, FH2 and FH4, as health-inequity factors. Besides facilitating higher longevity and lower mortality, achieving a low degree of inequity within an area would be an objective target to better attain community health. The individual factors are independent of each other. Each health-level factor had its own set of health-determining factors. The health determinants belong to all the categories we employed. This fact shows us a broad spectrum of inter-relations between health and the environments. Age-adjusted mortality indicators are low where the educational level of the community is high. The correlation between health and educational level has been discussed with regard to various conditions [7]. An aspect of the relation between health and educational level is seen in the present study. However, the educational level represented by FE1 does not correlate to the other health-level factors. Nordstrom et al. indicated that maternal education has no significant influence on the

I2

S. Takeuchi

et al. /Health

Policy

33 (1995)

1-14

decrease of infant mortality in Sweden [8]; this is understood to mean that where the basic infrastructure of health care and education has already been equally provided, the impact of education itself has only a minor influence. Regarding the effect of the size of schools and their distribution in an area, large schools housing a large number of students tended to induce a lower level of inequity in community health represented by FH2; areas with schools of this type can provide uniform education and may reflect some other social environments which relate to the reduction of inequity in health. On the other hand, areas having schools with a large number of students showed a high student absenteeism rate. Further investigation is necessary to elucidate the relationship between school size, uniform education, and the mental health of children. Our present results show strong evidence that housing size and its quality, as well as amenities have an impact on the mortality indicators. The retail price of all commodities in Tokyo is higher than in other cities: New York, Hamburg, London, and Paris prices are 79%, 76%, 92% and 86% of those of Tokyo, respectively [9]. Among other living costs, the price of house rent is noteworthily high: that in the above cities is W??, 56O, 90% and 88% of Tokyo, respectively [9]. Housing size and amenities have great significance in Tokyo, more so than in the other cities. Good housing conditions, therefore, may reflect high socio-economic status. Housing problems are basically understood to be economic and political issues. A public health policy which assures quality of life in the housing environment is considered to be essential. Marmot et al. have reported that differences between mortalities due to selected causes of death in manual workers and in non-manual workers in Great Britain have widened [lo]. As for the effect of occupational status on womens health, KritzSilverstein et al. have demonstrated that middle-aged women employed in managerial positions are healthier than unemployed women [l 11. In the present study, the large proportion of technical and professional workers, large proportion of tertiary industry employment, and high income level are all inter-related. Correlations between FHl and FWl, and between FHl and FS4, are seen. However, it is of note that FWl and FS4 did not correlate with health inequity within city-wards. The citywards having diverse outdated commercial areas which consist of small-scale retail shops tend to display large health inequity, probably due to the fact that they are behind in the preparation and development of the residential environment. An increase in the number of medical doctors or the number of hospitals did not correlate to FHl. It is believed that the impact of the provision of medical-care resources has already reached the marginal level in Tokyo. Nevertheless, accessibility to technically advanced tertiary-level medical care facilities seems to have a relation to community health. However, it is of interest that health inequity in an area is found to increase where the number of doctors per resident is large, or the number of medical care facilities per resident is large. Further insight relating to the influence of the quality of medical care is necessary. Regarding the choice of categories for environmental indicators in the present analysis, we can validate as follows. The choice of categories divided from a different aspect does not matter in providing final structure of the relation between health and health determinants, provided that the factors of each category represent the in-

S. Takeuchi

et al. /Health

Policy

33 (1995)

1-14

13

dicators well. The obtained set of environmental factors explain between 84% and 95% of the total variance by the selected environmental indicators, i.e., the major part of the variance is taken into account. To handle many indicators at one time in the prediction model obtained by the linear regression analysis, we employed the following steps: categorizing indicators, performing the principal factor analysis, and correlation analysis. The obtained model has a high predicting power: 0.58-0.98 of the variance for each health-level factors is explained by the model; we may call it a prediction model. In each of the individual health indicators, SMRs (H2-H7) are predicted very well; their R2s are between 0.84 and 0.95. The longevity differential indices (H14 and H15) have a smaller R2 than the SMRs; they vary more widely probably because they are the deviation of the age-adjusted death rate rather than its mean value. The model for them, however, still explains more than half of the variance of the original indicators. The other inequity index, the age-adjusted PM180 and PM185 (HlO and Hll), are found to be well predicted: their R2s are 0.90 and 0.85. The absolute values of the basic data of the divorce rate and the number of criminals are small in each case and scattered, which made predicting FH5 rather difficult. Our analysis on the statistical basis makes it possible to predict the health and health determinants of communities quantitatively. Since the predicted community health level is actually compatible with the observed data, our prediction model for community health by the employed environmental indicators is worthwhile for use in practical applications including decision making in health planning and health policies for communities.
Acknowledgments

This work is supported in part by Grant-in-Aid for Scientific Research by the Japanese Ministry of Education, Science and Culture (No. 05670339 and No. 04202116) and by the Health Sciences Promotion Foundation Research Project.
References
[l] [2] [3] Doll, R., Health and the environment in the 199Os, American Journal of Public Health, 82 (1992) 933-941. Walster, SD., The ecologic method. 1. Overviews of the method, Environmental Health Perspectives, 94 (1991) 61-65. Wing, S., Bamett, E., Casper, M. and Tyroler, H.A., Geographic and socioeconomic variation in the onset of decline of coronary heart disease mortality in white women, American Journal of Public Health, 82 (1992) 204-209. Marmot, M.G., Kogevinas, M. and Elston, M.A., Social/economic status and disease, Annual Review Public Health, 8 (1987) Ill-135 World Heaith Organization, The Urban Health Crisis, World Health Organization, Geneva, 1993. Department of International Economic and Social Affairs, World Urbanization Prospects 1990, United Nations, New York, 1991. McMahon, L.F., Wolfe, R.A., Griffith, J.R. and Cuthbertson, D., Socioeconomic influence on small area hospital utilization, Medical Care, 31 (1993) YS29-YS36. Nordstrom, M.L., Cnattingius, S. and Haglund, B., Social differences in Swedish infant mortality by cause of death 1983-86, American Journal of Public Health, 83 (1993) 26-30.

[4] [S] [6] [7] [8]

14

S. Takeuchi et al. /Health Policy 33 (1595) I-14

191 Price Bureau, Price Report 1992 Summary, Japanese Economic Planning Agency, Tokyo, 1992. [IO] Marmot, M.G. and McDowell, M.E., Mortality decline and widening social inequalities, Lancet, 8501 (1986) 274-276. [l l] Kritz-Silverstein, D., Wingard, D.L. and Barrett-Connor, E., Employment status and heart disease risk factors in middle-aged women: the Ranch0 Bemerdo Study, American Journal of Public Health, 82 (1992) 215-219.