Vous êtes sur la page 1sur 7

Brit. J. prev. soc. Med.

(1966), 20, 165-171



Research Scholar, Department of Social Medicine, University of Edinburgh

The purpose of this article is to offer a develop- Sutherland (1962) developed a form of demo-
ment of the demographic map as an alternative to the graphic diagram, using the same principle of area
geographical base-map for the presentation of areal proportional to population, called an isodemic
data in epidemiology. The results of geographical representation. As a basis he split Scotland into its
investigations into disease morbidity or mortality five hospital regions and showed them separated.
are often presented cartographically and invariably He thus lost geographical contiguity between the
the base-nap used is the normal geographical administrative units, a condition it would seem
one, showing the relevant administrative areas. The desirable to preserve.
relating of disease rates to area is useful, in that the An attempt has been made in developing the age-
recognition of areas characterized by high or low sex specific cartograms illustrated below (Figs 1,
rates may lead to clues of aetiological significance. 3, and 4) to relate disease rates to both the popula-
When considering areal patterns of disease, tion at risk and to geographical position in the 57
however, the epidemiologist requires information public health districts of Scotland.* A basic aim was
about the size of the population at risk in the areas also to try to produce cartograms similar to each
concerned. Sutherland (1962) drew attention to the other in local outline as well as in overall shape to
principal deficiency of the geographical base-map make them readily comparable and hence of greater
in this respect. Referring to Scotland, he showed potential utility. Efforts to simplify their construc-
that on the normal map correct weighting could not tion and final shape have also been made. To date
be given to the large urban populations which occupy demographic maps have presented complicated
small areas, whilst small rural populations, often construction problems and their final appearance
sparsely distributed over large areas, could appear has often been of considerable complexity. These
to be over-represented. Thus, base-maps which factors may well have obscured their advantages and
would relate disease rates to the local populations contributed to their limited application so far.
at risk as well as to geographical position might
prove useful epidemiological tools. Development of METHOD
the demographic map, in which the area of each Levison and Haddon (1965) outlined their
administrative unit is made proportional to its procedure for constructing a demographic map or,
population whilst contiguity of geographical boun- as they called it, an "area adjusted map" of New
daries and the relative geographical positions of the York State. As there is no unique solution for a
units are maintained as far as is possible (Hollings- given area and population, and because conventions
worth, 1966) offers interesting possibilities in this are developed as one proceeds which require ex-
direction. planation, an account of the method of construction
The use of such maps in epidemiology is not new, used here is given below. Reference to the locational
Levison and Haddon (1965), for example, having cartogram (Fig. 1, overleaf) based on the 1961 female
used the technique for plotting cases of Wilm's
tumour in New York State. To date, however, * The insular portions of Inverness and Ross and Cromarty Counties,
these maps have not been used to present data at the treated as have separate public health districts by the Registrar General for
been considered as part of their parent counties here,
national level. thus reducing the number of districts from 59 to 57.

Oy C.
Sutherlan|d C.ICaith
gess C. r-
BOUNDARY Cromarty l
Aberdeen C.ofC.
nverness Bans
L _L.B. C C. C
& r .% . r. w -% -vL
I nverness C.
C.ot C. COUNTY OF Bue Perth C. t L B- Angus C Arbro8th LB6
CITY LJArgyll C. Perth L.B. -
L.B. LARGE BURGH Dunbarton C. Stirling C. L Dundee C.ot C.
CIa kmar non
I C. C.-

L_ L ' irdriL.B. LL B-L L_

Cootbrldge I
IRutherglen L.SB
r - - - I Motherwell L _
Greenock I
L. B. CG losggow C.of C. I and Wishow C Edinburgh C-ot C.
Pf rfSI^ow L.B. Hamilton

Lanark C. o

- - - - - - - - Midlothian C. B ick C.
Rentrew C.
Pee bI e s

Ki narnock
10,000 PERSONS
L -r

2,500 ...

FIG. 1.-Locational cartogram, Public health districts of Scotland.

Population base, 1961 Census, females aged 45 to 54 years.
population in the age group 45-54 years, will facilitate In reality, no two large burghs in any county are
understanding of the following explanation: more than 15 miles apart and in Lanark, which has
(1) 1961 census population figures were taken for five large burghs, both Airdrie and Coatbridge, and
the 57 public health districts of Scotland for the Motherwell and Hamilton are adjacent. Grouping
10-year age groups 35-44, 45-54, and 55-64, and does not therefore cause such a distortion as it
for 35-64 combined for males and females separately. might at first appear.
There were thus eight age-sex groups in all. RESULTS
(2) A scale of 01l" square to the nearest 100 The same mortality data, calculated for all causes
persons was selected as suitable for plotting the of death among females aged 45-54 years for the
cartogram outlines onto arithmetic graph paper. period 1959-63, is presented on the conventional
(3) The firths of Moray, Tay, Forth, and Clyde, map (Fig. 2, overleaf) and the cartogram (Fig. 3,
were selected as prominent geographical features
overleaf). Clearly, impressions gained from maps are
around which to build the cartograms. (These subjective and opinions will vary as to the relative
features penetrate markedly into Scotland and do value of different forms of presentation. Attention is
much to determine the country's shape). drawn below to only the major points of difference
between the two methods shown here in an attempt
(4) The cities of Glasgow, Edinburgh, Aberdeen, to illustrate the characteristics and, it is felt, the
and Dundee were given constant positions fronting essential advantage of the demographic map.
onto their respective firths. As these cities are As a first step it will be helpful to compare the
important population concentrations, their positions locational cartogram (Fig. 1) with the geographical
and size partially determined the positions of the map (Fig. 2) so that familiarity can be gained with
other units. The firths and cities thus provided a the conventions of the former and the alterations
framework of lines and areas common to each in the areal weighting of Scotland that result. In
cartogram which helped to achieve similarity of particular (Fig. 1):
final outline. (1) The north-west highland counties, north of
(5) To prevent the cartograms from becoming too Dumbarton and Stirling Counties, and the
complex in shape, a factor which can lead to diffi- southern ones from Berwickshire to Wigtown,
culties of assessing the relative weighting to be with their numerically small populations, are
given to individual units, the following conventions reduced in area, relative to Scotland as a whole.
were adopted: (2) Conversely, the main areas of population
(a) No attempt was made to retain the geograph- concentration in west central Scotland and in
ical shape of administrative units; instead they were eastern Scotland from Edinburgh northwards
formalized into shapes based on rectangles. Con- assume increased proportions.
sideration of Argyll with its small population, (3) As a result of these changes, it is easier to
considerable north-south extent and often insular appreciate the extent to which the population
shape will show why this was necessary. Any group used as a base for the cartogram is
attempt to retain its shape on a demographic map resident in the larger urban centres (Table)
would have resulted in a fragmented narrow zone, and in Glasgow in particular.
the demographic weighting of which would have
been difficult to assess. TABLE
(b) Geographical positions were modified where SELECTED POPULATION TOTALS, 1961 CENSUS,
it was felt that this helped towards simplification of
shape and construction, whilst as far as possible Area Population
Per cent. of
Scottish Age Group
geographical contiguity of boundaries was main-
tained. For example, relative to Ayrshire, Wigtown Glasgow .. .. 72,885 20-44
Edinburgh .. .. 35,683 10-00
has been given a north-westerly, as opposed to its Aberdeen
true south-westerly, position.
Counties of City 135,083 37-86
(c) Where two or more large burghs occurred in a Large Burghs 50,317 14*11
county, they were grouped and indicated as non-
contiguous. The alternative, to have distributed Cities and Large Burghs 185,400 51*97
them throughout their county area, would have County Areas .. 171,295 48*03
fragmented the latter and again have contributed to TOrAL .. .. 356,695 100
the problem of assessing demographic weighting.
Comparison of the geographical distribution of the (1) Glasgow County of city and Kirkcudbright
mortality rates on the two presentations (Fig. 2, and County, both being high mortality rate areas.
Fig. 3, opposite) can usefully be made at two levels. (2) Dundee County of city and Angus County,
At the local level it is interesting to note the way in respectively high rate and low rate areas.
which the following are represented: (3) The variable mortality experience in the large
RATES/ 100.000

OVER 600
500< 550
450<500 A400
a &
[I - 400< 450
[fl] UNDER 400 6% 40
0 30

Fia. 2.-Average mortality rates 1959-63, all causes of death. Females aged 45 to 54 years. (Conventional map).

RATES/ 100,000
OVER 600
550 < 600
500 < 550 ......-

450< 500
400 E< 450
U NDER 400

.17 -= -
= ---=,
--L. ..

10.000 PERSONS

2.500 ..M.

100 Ad

Fso. 3.-Average mortality rates 1959-63, all causes of death. Females aged 45 to 54 years. (Cartogram).

I | IrI~- LJ

J X --L-J

-- - - - -

-_- -

J |_ I -
I _ _ _ -- _

_ _ L --
L - _- - - _ _ _.T
- _ _ _ __

FIG. 4.-Ago-ex specific cartosram (outline). Population base,

1961 census, male agd 35 to 44 yeas.
burghs in geographical proximity to Glasgow. base-map which is simple enough in arrangement
At the wider regional level, the following summary readily to allow geographic patterns to emerge.
of the geographical pattern of mortality is obtained Further, the utility of the device is considerably
and, it is suggested, is easily recognizable on both increased by the fact that the eight different age-sex
presentations. cartograms produced are easily visually comparable
(1) Glasgow is the most prominent area of un- and that they all share a common set of conventions
favourable mortality experience. Together with (cf. Fig. 1, and Fig. 4, opposite). It will be clear that
the large burghs of Stirling, Airdrie, Hamilton, such maps can be prepared for any area and any sub-
Paisley, and Lanark County, it forms a high section of the total population, given the relevant
rate area in West Central Scotland. Within population figures. It might be useful to consider the
this area the large burghs of Greenock and production of a national series of age-sex specific
Coatbridge stand out in marked contrast by base-maps of this type having common shapes and
virtue of their low mortality rates. conventions, for use in the presentation of areal data
(2) No other concentrations of high mortality are in epidemiological studies.
found. Rather, a sporadic distribution of
high rate areas occurs, including the rural SUMMARY
counties of Orkney, Caithness, Banff, and Presenting disease rates on the conventional
Ross and Cromarty in the North, Wigtown geographical base-map does not allow of weighting
and Kirkcudbright in the south, and the for local population differences. Development of the
urban area of Dundee on the east coast. demographic map offers possibilities for relating
(3) A zone of moderate to low mortality experi- disease rates both to local populations at risk and to
ence extends from Argyll and Bute, north- geographic position. In an attempt to achieve these
eastwards to Aberdeen County and south- effects, age-sex specific cartograms were developed
eastwards to Berwickshire. for Scotland based on 1961 census figures. Their
(4) All of eastern Scotland south of Aberdeen construction is described and, by presenting the
County, with the exception of Dundee, is thus same mortality data on both the cartogram and the
a favourable area in terms of mortality geographic base-map, an attempt is made to indicate
experience in this sex and age group. the advantage of the former over the latter for the
presentation of areal data in epidemiology.
The principal value of a map is that it enables the I am indebted to the following: Dr Mary Fulton and
essential features of a distribution pattern to be Mrs Elspeth Semple for their kind permission to use the
quickly appreciated. The cartographic needs of epidemiological data portrayed in this article, Prof. S. L.
different disciplines do, however, vary, and maps are Morrison for his encouragement, Mr T. Edge of J.
of greatest value when designed to serve specific Bartholomew and Company, Limited, for advice regard-
requirements. The cartograms presented above are ing the use of cartographic equipment, and Prof. Wreford
illustrative devices especially suited to the needs of Watson for kindly placing at my disposal the facilities
of the Department of Geography in the University of
the epidemiologist working with areal data for, by Edinburgh.
their very nature, they correct the principal defect
of the conventional map by illustrating the required REFERENCES
dimension of population. Hollingsworth, T. H. (1966). Map of Results of General
Perhaps the most encouraging feature is that, as Election, The Times, April 4, 1966. Late London
a result of careful initial consideration of the major Edition, p. 8.
demographic and geographic features of Scotland, Levison, M. E., and Haddon, W. (1965). Publ. Hlth Rep.
(Wash.), 80, 55.
it has been possible to produce a demographic Sutherland, I. N. (1962). Brit. J. prev. soc. Med., 16, 30.