The City of Health

Masood Gheasi, Noriko Ishikawa, Peter Nijkamp , Karima Kourtit

ABC Contribution: The City of Health

• Human Health and the Urban-Rural Dichotomy: An Overview

•A meta-analysis of health effects of urban and rural living patterns

Urbanization and quality of Life

Paper overviewThe health outcomes are well known to be affected by not only

individual socioeconomic status, but also by the physical environment, and socioeconomic and political circumstances including “urbanization”.

This paper aims to offer an inventory of studies that have addressed the relationship between urban conditions and human health and to examine why the estimated effects of urbanization vary among countries or studies.

2


The definition of urbanityThe definition of urbanization varies across countries and during

different periods reflecting their geographical and socio-economic differences.

Some studies follow a classification scheme from the National Census (e.g., UIC or RUCC in the U.S., ONS or ACORN in England).

Criteria of urban areas is often determined by population size, density or whether people are living in a core city or not.

3


Step1: InventoryUrban/rural effects on health outcomesOverview of several papers addressing health inequality between

urban and rural areas…(99 papers)

29 papers: advantage of urban living for health

42 papers: advantage of rural living for health

13 papers: different results depending on health outcomes

5 papers: no significant differences in health outcomes between urban/rural living

4


6 contextual effects on health outcomes (1)

Housing conditions (direct/ indirect)Access to safe water and adequate sanitation facilitiesStructural housing characteristics (housing type, floor level, quality)Housing tenure (proxy of residential stability or spatial deprivation)

Environmental hazardsThe exposure to air pollution (whether living near a major road,

highway, or noxious facilities)

5



Land use mix and spatial segregationProximity to facilities (amenities, retail services, work place)Degree of urban sprawl (mixed land use, neighbourhood diversity,

population density, automobile dependency, proximity to metropolitan area)The degree of walkability (residential density, street connectivity, extent of

mixed land use, retail floor area ratio)The degree of social/racial segregation

Residential proximity to CBDCommuting distance, time (proximity to metropolitan area)Intra-area relationship (spatial autocorrelation)Degree of spatial (geographical)/socioeconomic isolation

6



Accessibility to and availability of medical services(Accessibility)Time and distance from home to health care facilities(Availability)The number of physicians (GPs/specialists), health care facilities, and some

related facilities (acute/chronic services, medical/nursing care services)

Population densityA proxy of living conditions (narrow houses, a lack of open/green spaces,

traffic congestion, high exposure to pollution, lower transportation costs, accessibility to services and amenities, neighbourhood walkability, employment opportunities, and human capital accumulation)

7


Inconsistency in the urban/rural effects

A series of studies have described the defects in urban/rural dichotomy (Higgs, 1999; McDade and Adair, 2001, Vlahov and Galea, 2002; Hall et al. 2006; Laditka et al., 2007; Berke et al., 2009; Peen et al., 2010)

Inconsistency in findings may come from (1) a not-unified (inappropriate) classification scheme, and (2) an intra-regional heterogeneity.

It is important to find a classification scheme that minimizes the heterogeneity within regions.

Inconsistency in the results from different studies appears to be due the ambiguous definition of urbanity and an intra-regional heterogeneity (e.g., inner-city areas).

8


Various measures of heath statusSubjective health statusSRH (self-rated health) scores

LLTI (limiting long-term illness)

Objective health statusICD (International Classification Diseases)

GHQ (General Health Questionnaire)

mortality rate/life expectancy

9


Step 2: Meta AnalysisStudy Characteristics We included research papers that take first ‘self-rated’ or ‘self-

reported’ physical health and secondly include ‘urban’, ‘rural’, ‘city residence’ differences since 1999 into account.

12 studies that yielded altogether 221 point estimations.

In this study, we focus on good-excellent physical health.

Unadjusted vs adjusted estimations

Journal impact factor as an identification factor for the quality

10


The numerical attributes include the calculated standard error, OR for urban, semi-urban, semi-rural and rural, number of observation per regression, and t-values.

The study characteristics are coded as dummy variables, and equals 1 in each regression that includes that particular attribute, 0 otherwise.

The decision to code characteristics of each study is not an easy taskwe coded the habits such as smoking, and alcohol consumption as

one variable called “addiction” social trust, social support and social capital under another dummy

variable, called “social”. 11


Study Time period

Country # estimation

Mean

Gerdtham and Johannesson (1999)

1991-1991 Sweden 4 0.968

Hyyppä and Mäki (2001) - Finland 6 1.35

Leinsalu (2002) 1996-1997 Estonia 4 1.19

Auchincloss and Hadden (2002)

1989-1991 US 54 0.866

Kim and Kawachi (2006) - US 24 0.956

Maas et al. (2006) - NL 24 0.991

Riva et al. (2009) 2000-2003 UK 13 0.865

Sharkey et al. (2011) 2006-2006 US 2 1.505

Monnat and Pickett (2011)

2002-2008 US 16 1.113

Bthea et al. (2012) 2006-2006 US 6 1.03

Lankila et al. (2012) 1997-1997 Finland 48 1.317

Haraldsdóttir et al. (2014)

2007-2007 Iceland 8 1.39

Study Characteristics (1)

12



13

Negative and

statistically significant

Negative and

statistically not

significant

Positive and statistically

not significant

Positive and statistically significant

Total

Urban 30% (29) 33 % (19) 19 % (11) 18 % (10)100% (57)

Semi-urban 13 % (7) 48% (25) 12% (6) 27% (14)100% (52)

Semi-rural 23% (9) 46% (18) 21% (8) 10% (4)100% (39)

Rural 22% (19) 32% (13) 3% (18) 17% (10)100% (60)



14


Meta-regression model

15

We aim to investigate to what extent the variation in health outcomes between and within studies can be related to the study characteristics.

If we assume that the true effect size varies from one study to the other, a random effect meta-regression model is applicable (we apply the algorithm of restricted maximum likelihood with the Knapp-Hartung modification, see Harbord and Higgings, 2000).

If we assume that there is no variation between the studies, then we use the FE meta-regression. weighted least square (WLS) with the weights of variables equal1/(𝛼𝑖𝑗). The standard errors of the regression are to be adjusted with the cluster of

observations defined by 12 studies.


Key findings (I)

16

(1) (2) (3)

VARIABLES OLS FE RE

standard error 1.129 (0.228)*** 0.259 (0.478)

urban -0.0154 (0.0331) -0.0111 (0.0356) -0.0597 (0.0268)**

semi-urban 0.0962(0.0319)*** 0.0386 (0.0345) 0.0616 (0.0280)**

semi-rural -0.0349 (0.0343) -0.0226 (0.0356) -0.0241 (0.0330)

before 2000 -0.274 (0.0674)*** -0.354 (0.0987)*** -0.316 (0.0784)***

female 0.0621(0.0369)* 0.0762(0.0343)** 0.0539(0.0416)

adjusted -0.125 (0.0511)** -0.125 (0.0575)** -0.144 (0.0436)***

journal impact factor -0.120 (0.100) -0.340 (0.187)* -0.269 (0.102)***


Key findings (II)

17

(1) (2) (3)

VARIABLES OLS FE RE

family income 0.154(0.104) 0.112 (0.0491)** 0.146 (0.0836)*

education -0.0800 (0.0786) -0.0880 (0.0816) -0.0809 (0.0792)

employment -0.0624(0.0909) -0.0822(0.0510) -0.0792(0.0799)

length of residence -0.0979 (0.0807) -0.0709 (0.0775) -0.0790 (0.0818)

material condition 0.0842 (0.157) 0.529 (0.186)*** 0.299 (0.147)**percentage of green place 0.127 (0.0770) 0.116 (0.0541)** 0.138 (0.0585)**

religious -0.0505(0.0953) -0.0188(0.141) -0.0374(0.0745)

physical activity 0.0509(0.0807) 0.0644(0.0722) 0.0710(0.0845)

Urbanization and quality of Life 18

(1) (2) (3)

ordinal 0.451(0.110)*** 0.576(0.158)*** 0.577(0.101)***

multilogit -0.0245(0.0399) -0.0455(0.0434) -0.0220(0.0367)

race 0.162(0.0735)** 0.130(0.0786) 0.147(0.0615)**

community predictor -0.140(0.0896) -0.308(0.0961)*** -0.223(0.0767)***

occupation 0.0494(0.128) 0.479(0.245)* 0.262(0.130)**

migratory behavior 0.592(0.290)** 0.921(0.389)** 1.061(0.266)***Health personal shortages -0.160(0.111) 0.0591(0.101) -0.0779(0.0907)

loneliness feeling -0.149(0.0807)* -0.122(0.0991) -0.160(0.0853)*

constant 1.071(0.0118)*** 1.071(0.0144)*** 1.063(0.0118)***

Observations 209 209 209

R-squared 0.733 0.684


ConclusionA meta-analysis approach is one of the effective methods for

estimating the effect size across countries/studies.

There is also ambiguity regarding the direction of the relationship.

Further research on direction and statistical significance of empirical estimated of self-rated/reported health.

19

Education

The City of Health