Embed Size (px)
Citation preview
Infectious disease mortality in two Outer Hebridean islands:1. measles, pertussis and influenza
E. J. Clegg
Department of Biomedical Sciences, University of Aberdeen, UK
Received 23 August 2001; in revised form 16 May 2002; accepted 21 May 2002
Summary. Objectives: The purpose of the study was to examine changes in mortality frommeasles, pertussis (whooping cough) and influenza (all epidemic diseases) in Harris andBarra, two Outer Hebridean islands, from 1855 to 1990, and to compare the findingswith those from Scotland as a whole over the same period. It was also intended to relatechanges in mortality to those in social and economic factors.Materials and methods: Ages and causes of death in Harris and Barra were ascertained fromcopies of death certificates held at the General Register Office, Edinburgh, and for Scotlandas a whole from the Annual Reports of the Registrars General for Scotland. Data werestandardized by calculating the Proportionate Mortality Ratio (PMR), the proportion ofdeaths due to a particular cause to all deaths over a given period. Spectral analysis wasemployed to examine the durations of epidemic cycles.Results: Ages at death increased slightly over the period of the study. For measles andpertussis, other than for the former in Harris, there were significant relationships betweennumbers of deaths per decade, and numbers of new susceptibles, estimated as the numbersof births. Epidemics of measles and pertussis in the islands occurred at intervals, usuallyseparated by years of no mortality. The highest PMRs were generally during the laterdecades of the 19th and first decade of the 20th centuries; this may have been related tothe economic problems of agriculture and fishing, and to increasing population density.Influenza epidemics were more frequent than those of the other two diseases. For allthree diseases in both islands, there were significant negative power relationships betweenepidemic size and frequency of occurrence; those in Harris were the stronger. The relation-ships between length and frequency were significant only in Harris. Generally, epidemiclengths seemed less variable than sizes, possibly because of the rather ‘coarse’ units of length(quarters) employed. Spectral analysis of the ‘detrended’ data for the period before theintroduction of specific immunoprophylaxis revealed that for measles the main epidemiccycle in all three populations was between 7.3 and 7.8 years’ duration. Barra and Scotlandhad additional 2.5- and 2-year cycles, respectively. For pertussis, Harris and Barra had maincycles of 7.4 years. Harris had an additional cycle of 3.2 years. Scotland had cycles of 4 and 2years. For influenza, Harris had a main cycle length of 7.4 years, and a less-defined one ofabout 2.6 years. Barra had a main cycle of 6.9 years, and a subsidiary one of 2 years.Scotland has a single cycle of 8 years. Cubic regressions of the spectral densities on cyclefrequencies showed large coefficients for Harris and Barra, but small ones for Scotland.Measles coefficients were closely similar in the two islands, but not those for pertussis.Conclusions: The findings demonstrate the episodic occurrence of epidemics of these threediseases in the two islands, as against their continual presence in the much larger populationof Scotland. They reveal also the decreasing importance of these causes of death in all threepopulations. The data from Harris and Barra suggest that measles is a more epidemiologi-cally ‘stable’ disease than pertussis. Both islands appear to obey Hamer’s law of ‘massaction’. The relatively long intervals between epidemics in the islands may be due partlyto their isolation, and partly to the slow accumulation of sufficient numbers of susceptiblesto enable an epidemic to occur.
1. Introduction
Historical changes in patterns of mortality in a population are good indicators ofthe adaptation of its members to their environment. Deaths from particular causesmay indicate the strength of particular adverse environmental and/or hereditaryfactors, and the presence of historical trends in such deaths indicates natural changes
Annals of Human Biology ISSN 0301–4460 print/ISSN 1464–5033 online # 2003 Taylor & Francis Ltdhttp://www.tandf.co.uk/journals
DOI: 10.1080/03014460210157420
ANNALS OF HUMAN BIOLOGY
JULY–AUGUST 2003, VOL. 30, NO. 4, 455–471
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
in the environment–organism relationship and/or the success of policies to amelio-
rate environmental effects, either through specific measures designed to combat par-
ticular diseases, or through more general actions, for example improvements in
housing, diets, access to health care, etc.
Deaths due to infectious disease tend to be commoner in the pre-reproductive and
reproductive years. Hence from the point of view of natural selection, they are
generally regarded as the more important than those due to non-infectious disease.
The continued presence of an infectious disease in a community depends on the
presence within it of a pool of susceptible individuals. Where a disease is highly
infectious and of easy transmission, for example by droplet or water-borne spread,
on first introduction it may affect all or nearly all susceptibles within a very short
time. What happens next depends on the number of infecteds introduced and on the
fate of those infected. If the disease confers no immunity to subsequent infection, it
will remain endemic, the numbers presenting depending on its mortality, the main-
tenance of infectivity, the mode of transmission, and any long-term trends in host
resistance. If some, but not total immunity is conferred, the same result will be seen,
but the numbers affected and particularly the severity of the disease will decline. If
total immunity is conferred by a single infection, as with, for example, measles or
pertussis, the incidence of the disease will fall; the extent of the fall depending on the
continued presence of new susceptibles, either by birth or immigration or by decrease
in immunity, from whatever cause. The former two factors are related to the size of
the population (Bartlett 1957), and the extent to which it is isolated from other
populations. In the simplest case, where the population is small and very isolated,
the disease may disappear completely until a new pool of susceptibles has been built
up and the isolation of the population is breached by an infected individual or
individuals. A good Hebridean example of this situation is the ‘boat cold’, which
occurred in the very isolated island of St Kilda, following the return of islanders
visiting the mainland (Steel 1976, p. 149).
Rhodes and Anderson (1996a,b), Rhodes, Jensen and Anderson (1997) and
Rhodes, Butler and Anderson (1998) have emphasized the irregularity of outbreaks
of communicable disease in isolated populations; such outbreaks tend to be sepa-
rated by periods when the disease is absent. However, they established that the
relationships between the sizes or lengths of epidemics and their frequencies could
be described by power laws.
When the population is larger, and a single infection confers lifelong immunity,
the occurrence of new cases may vary cyclically. A first appearance of the disease will
affect nearly all susceptibles (perhaps the whole population), and the epidemic will
run its course, leaving almost all the survivors immune. Even new introductions will
result in few cases, until a new reservoir of susceptibles has been built up, sufficient to
make transmission from infected to susceptible likely. When there are sufficient
susceptibles, a new introduction of the infective agent will be followed by a rapid
rise in the numbers of affected, and the number of susceptibles will fall commensu-
rately. Because of the diminishing number of susceptibles, transmission becomes
increasingly difficult, numbers of new cases decrease, and the epidemic comes to
an end: the disease again becomes sporadic in incidence among the remaining
small number of susceptibles until a sufficient pool of new susceptibles is built up
again, and the next epidemic ensues. Diseases with this cyclical pattern of behaviour
include measles (Anderson, Grenfell and May 1984, Anderson and May 1991,
456 E. J. Clegg
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
Mielke 1996, Sumi 1998), pertussis (Anderson et al. 1984) and influenza (Schild1977).
In this paper, mortality from selected infectious diseases will be considered in twoislands, Harris and Barra, which form part of the 120 km chain of islands off thenorth-west coast of Scotland known as the Outer Hebrides. Each island consists of amain portion, with detached islets, a good number of which are now uninhabited.Both are comparatively isolated: Harris, though conjoined to the north with the Isleof Lewis, is separated from it by inhospitable mountainous terrain, and from thenext southerly island, North Uist, by the difficult Sound of Harris. Barra, a trueisland, is separated from South Uist by the Sound of Barra. Each island has trans-port links with the Scottish mainland, those of Barra being the more direct.
The populations of both islands showed marked growth during the latter half ofthe 19th century and first decade of the 20th. Harris increased from 4183 in 1861 to5449 in 1911, and Barra from 1853 to 2620 over the same period. Subsequently bothpopulations declined precipitately, and in 1991 that of Harris was 2218, while that ofBarra was 1318.
Economically, both islands are good examples of the ‘crofting’ way of life, amixture, in varying proportions, of small-scale agriculture, fishing, tweed weavingpractised as a cottage industry, and various other types of paid employment (seeHunter (1976) for an authoritative account of the development of the crofting econ-omy).
During the latter part of the 19th century, the Highlands and Islands of Scotlandsuffered periods of severe economic stress, due partly to the population increases,and partly to the failure of agriculture and the fishing industry in the 1880s. Therewas further hardship after World War 1, which led to the massive depopulation ofthe 1920s and 1930s.
Patterns of mortality in these two islands will be related to these socio-economicchanges and compared with those in Scotland as a whole. The impacts of time, andnumbers of susceptibles (measures of population density) on numbers of deaths willbe estimated. In addition, in the two islands, the relationship between the sizes orlengths of epidemics and their frequencies (Rhodes and Anderson 1996a,b, Rhodeset al. 1997, 1998) will be studied, as well as the extent to which cyclical changes inmortality have occurred.
2. Materials and methods
Ages at and causes of death in Harris and Barra were ascertained from copies ofdeath certificates for the period 1855–1990, held at the General Register Office,Edinburgh. Since hospital facilities for Harris are in Lewis, and for Barra inSouth Uist, all Outer Hebridean registers were examined to ensure that deaths ofresidents occurring outwith the two islands were included. For comparative pur-poses, data for Scotland as a whole were obtained from the Annual Reports ofthe Registrars General for Scotland.
Some difficulty was experienced in determining precise causes of death, particu-larly during the 19th and early 20th centuries, when neither island had access topermanently resident medical care. To reduce this difficulty, acute infectious diseaseswhich are common and well known, and where the diagnosis is relatively easy for thelayman, were selected. In the end the choice was of measles, pertussis (whoopingcough) and influenza. The first two are epidemic diseases of childhood, generallyeasy for the layperson to recognize. Influenza is more difficult, as misdiagnosis is
Infectious disease in the Outer Hebrides 457
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
easier, and the taxonomy of the (highly mutable) virus more complex, but it is animportant epidemic disease, and it was felt that its inclusion would be of interest.Single episodes of measles or pertussis confer lifetime immunity; this is not the casewith influenza.
For each of the decades between 1856 and 1985, numbers of deaths from thesecauses were ascertained. For measles and pertussis, multiple linear regression ana-lyses were carried out, the independent variables being the accumulation of suscep-tibles, estimated as the total births in successive decades around Census years, andtime, measured as successive Census years. No similar analyses were carried out forinfluenza, because of the greater age ranges at onset, and the much more compleximmune responses to this disease.
In the study of the interaction between the frequencies of epidemics and their sizesand durations, total deaths in each quarter of each year (size), and the numbers ofconsecutive quarters in which deaths occurred (length) were ascertained. Given thesmall numbers of deaths during each year, it was felt that ascertaining monthlydeaths would not be appropriate. Curve-fitting analyses were performed, the depen-dent variables being either epidemic size or length, the independent variable epidemicfrequency. A power model (Rhodes and Anderson 1996a,b) was tested:
N ¼ b0xb1
where N is the expected number of epidemics of size or length x, b0 is a constant andb1 a negative coefficient.
For further analysis, numbers of deaths in each year were standardized by calcu-lating the Proportionate Mortality Ratio (PMR) (Mausner and Bahn 1974), the ratioof deaths from any one cause to total deaths in a given year.
In all cases the PMRs, as well as showing year-to-year variation, showed down-ward trends over the whole period of study. Since measles, pertussis and influenzaexhibit periodical epidemics, apart from long-term trends, it was of interest to ex-amine changes in mortality from this point of view. This necessitated initially‘detrending’ the data, which was accomplished by the ‘moving average’ method(Cliff and Haggett 1988, p. 149). The method used to examine possible cyclicalchanges in PMRs was that of spectral analysis (Priestley 1981), which determinesall possible values of f, the Fourier frequency in the series,
fj ¼ j=N
where j is the number of times the cycle repeats, and N the number of observations.The spectral density is the sum of the squares of the two weights (sine and cosine) ateach frequency. When the spectral density is plotted against frequency (cycles perunit time), random ‘noise’ can be reduced by using a ‘window’ in which a givennumber of successive observations are given different weights according to theirpositions in the width of the ‘window’. In the present study the width was of 15observations and the weighting method was the Tukey–Hanning one (Tukey 1949,Priestley 1981, p. 443).
In interpreting the spectral density plot, the frequencies with the greatest densitiesare those which represent peaks of cycles. At frequency 0 the density represents, inthe present context, the endemicity of the condition. At the highest frequency (0.5with annual observations) the spectral density expresses the significance of the fastestmeasurable cycle, in these circumstances, of 2 years.
458 E. J. Clegg
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
In more recent years, measles and pertussis have been subject to effective immu-noprophylaxis. Vaccination for measles became widespread after 1969, and for per-tussis after 1956. While vaccination against influenza has been available since the1960s, its efficacy is usually short-term, because of the antigenic mutability of thevirus. Also, the virus varies very considerably in its virulence, so that mortality mayvary greatly from this cause alone, irrespective of the degree of immunity in thepopulation.
Because mortality from measles and pertussis ceased well before the dates ofintroduction of effective immunoprophylaxis, the periods of study for the spectralanalyses were limited to the years before these dates, up to 1969 for measles, and upto 1956 for pertussis. For influenza, it was not felt possible to subdivide the periodunder study in this way.
Each spectral density plot was subjected to regression analysis of density onfrequency (cycles per year). Empirically it was found that the best fits were tocubic polynomials. Residuals were calculated for each regression.
3. Results
3.1. Ages at deathTable 1 shows median ages at death for the three diseases in successive time
periods (generally 20 years, but 15 in the last) for Harris and Barra (data on agesat death were not available for Scotland, but numbers of deaths are presented).Measles and pertussis were generally diseases of early childhood; there was sometendency for medians to increase with time. Influenza showed wide variations in ageat death, possibly due to inconstant diagnostic criteria: in Harris and Barra thereappears to have been no excess mortality during the 1916–1935 period, whichencompasses the global pandemic of 1918–1920.
In the islands, these infectious diseases showed marked tendencies for numbers ofdeaths to diminish, after 1915 for measles and pertussis, and after 1935 for influenza.
Infectious disease in the Outer Hebrides 459
Table 1. Median ages at death (n) during successive time periods.
Period Population Measles Pertussis Influenza
1856–1875 Harris 2.00 (19) 1.75 (34) 2.00 (14)Barra 0.46 (4) 1.50 (18) 1.67 (7)Scotland (25 509) (36 592) (3898)
1876–1895 Harris 1.42 (29) 1.46 (40) 32.00 (48)Barra 1.17 (11) 2.00 (35) 60.00 (3)Scotland (31 413) (45 805) (9912)
1896–1915 Harris 4.00 (13) 1.08 (25) 62.00 (58)Barra 13.00 (16) 2.58 (10) 9.00 (24)Scotland (32 075) (41 179) (15 370)
1916–1935 Harris 58.50 (4) 3.00 (7) 64.50 (16)Barra 5.50 (2) 0.75 (1) 60.00 (13)Scotland (18 864) (21 022) (21 619)
1936–1955 Harris 53.00 (1) 1.75 (1) 65.00 (9)Barra – (0) 0.25 (1) 80.00 (1)Scotland (2487) (4103) (8446)
1956–1975 Harris – (0) – (0) 81.00 (1)Barra – (0) – (0) – (0)Scotland (151) (96) (5288)
1976–1990 Harris – (0) – (0) – (0)Barra – (0) – (0) 80.00 (1)Scotland (27) (24) (1006)
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
In Scotland, measles and pertussis deaths also declined after 1915, while deaths frominfluenza reached maxima between 1896 and 1935.
3.2. The impact of numbers of susceptibles on mortality from measles and pertussisTable 2 shows numbers of deaths from these two diseases, together with numbers
of births per decade, from 1856 to 1945, after which time there was no mortality inHarris and Barra from these two causes.
Table 3 shows the results of the multiple regression analysis. Apart from measlesin Harris, all populations showed significant and positive regressions of numbersdying on numbers of births. In addition, Scotland showed a significant and negativeregression of numbers on time. Values for r2 were generally quite high, especially forScotland.
3.3. The relationships between epidemic size or length and frequencyPower regression coefficients of epidemic frequency on size and of epidemic fre-
quency on length are shown in table 4.Table 4 shows that the association between frequency and epidemic size was much
stronger in Harris than in Barra. The exponent (b1) was always significantly differentfrom 0 in the former, for all three causes of death, while in the latter the levels ofsignificance of the exponents were much lower.
For the association between frequency and epidemic length (table 4), the differ-ences between Harris and Barra were still marked. In the former, the exponents werestatistically significant, but in the latter they were not. Comparison of the numbers of
460 E. J. Clegg
Table 2. Numbers of deaths from measles and pertussis per decade from 1856–1945, together withnumbers of births per decade.
Decade
Harris Barra Scotland
Measles PertussisBirths/decade Measles Pertussis
Births/decade Measles Pertussis
Births/decade
1856–1865 4 11 1338 0 1 617 13 730 17 564 1 070 7481866–1875 15 23 2387 4 17 639 11 779 19 028 1 173 8611876–1885 10 15 1396 3 14 657 12 723 23 117 1 262 4731886–1895 19 25 1372 8 21 792 18 690 22 688 1 248 8861896–1905 3 19 1419 16 5 769 16 715 22 219 1 313 0811906–1915 10 6 1293 0 5 552 15 360 18 960 1 234 3271916–1925 0 6 1007 2 0 459 12 506 13 887 1 102 0791926–1935 4 1 914 0 1 473 6 358 7 135 9 29 8021936–1945 1 1 671 0 1 417 2 095 3 357 8 96 662
Table 3. Stepwise regression coefficients of numbers of deaths per decade on decades (Census years) andbirths per decade.
Population Disease Decade Numbers of births r2
Harris measles – – –pertussis – 0.0274 (0.0075) 0.6062
Barra measles – 0.0308 (0.0097) 0.5823pertussis – 0.0421 (0.0161) 0.4222
Scotland measles – 0.0306 (0.0061) 0.7518pertussis –82.3388 (4.9551) 0.0370 (0.0028) 0.9808
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
cases (N) in the two tables suggests that epidemic sizes were more variable than theirlengths.
3.4. Changes in PMRs with timeThese are shown in figures 1 and 2, in which PMRs for Harris or Barra are
represented by bars, and for Scotland by continuous lines. Generally speakingboth islands had periodical years of mortality, separated by years in which nodeaths occurred (influenza was a partial exception, with several prolonged periodsduring which deaths occurred in every year). By contrast, the data for Scotland showthat all three diseases caused deaths in every year, with periodic years of highermortality.
3.4.1. MeaslesThese are shown in figure 1. In Harris (figure 1a), mortality reached a maximum
PMR of 0.163 in 1871, followed by a decline. The last death was in 1942. In Barra(figure 1b), there was a similar rise to a maximum (PMR ¼ 0:182) in 1887, with asubsequent fall: the last death was in 1920. In both islands, the intervals betweenepidemics were irregular.
For Scotland (figure 1a,b), the disease was endemic, with periods of rising mor-tality, especially to 1890–1893 and in 1922. These latter two periods coincidedapproximately with epidemics in Harris and Barra, respectively. The last deathsfrom measles in Scotland occurred in 1950.
3.4.2. PertussisThese results are also shown in figure 1. In Harris (figure 1c), epidemics were more
frequent and more regular in occurrence than those of measles, The highest mortality(PMR ¼ 0:196) was in 1856, with further high values between 1890 and 1920: the lastdeath was in 1945. Barra (figure 1d) also exhibited regular epidemics, with a veryhigh maximum PMR of 0.344 in 1892. The last death was in 1944.
For Scotland (figure 1c,d), mortality fluctuated, with perhaps a slight rise duringthe period 1880–1910, followed by a gradual fall: there was evidence of a 2–3-year
Infectious disease in the Outer Hebrides 461
Table 4. Coefficients (standard errors) of power regression of frequency of occurrence on epidemic sizeand epidemic length.
Island Disease N b0 b1 r2
Epidemic sizeHarris measles 8 7.678 (1.099) 71.024 (0.225) 0.823
pertussis 10 8.643 (0.911) 71.250 (0.211) 0.875influenza 9 20.626 (1.502) 71.156 (0.130) 0.951
Barra measles 7 1.910 (0.353) 70.337 (0.160) 0.447pertussis 7 3.437 (0.661) 70.603 (0.198) 0.663influenza 4 12.778 (1.695) 71.733 (0.494) 0.916
Epidemic lengthHarris measles 4 14.025 (0.693) 71.936 (0.188) 0.994
pertussis 4 12.007 (1.051) 71.271 (0.010) 0.999influenza 5 27.546 (1.955) 71.468 (0.192) 0.972
Barra measles 3 4.300 (1.675) 70.667 (0.725) 0.517pertussis 4 5.272 (1.118) 70.962 (0.401) 0.707influenza 2 16.000 (–) 71.678 (–) –
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
462 E. J. Clegg
Figure 1. Proportionate Mortality Ratios for measles in (a) Harris and (b) Barra; and pertussis in(c) Harris and (d) Barra. Bars—Harris and Barra; continuous line—Scotland.
Figure 2. Proportionate Mortality Ratios for influenza for (a) Harris and (b) Barra. Bars—Harrisand Barra; continuous line—Scotland.
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
cycle of mortality, the amplitude of the cycles decreasing with time. The last death inScotland was in 1953.
3.4.3. InfluenzaThese are shown in figure 2. In all three populations, epidemics were much more
frequent than those of measles and pertussis. In Harris (figure 2a), epidemics wereespecially frequent during 1860–1875, 1900–1917 and 1924–1932. Maximum mortal-ity (PMR ¼ 0:159) was in 1892, and the last death was in 1963. In Barra (figure 2b),there was, overall, much less mortality than in Harris, but the maximum(PMR ¼ 0:356), in 1901 was especially high. The last death was in 1976.
In Scotland (figure 2a,b), there were clusters of high mortality during 1890–1905and 1916–1925. The last death occurred in 1990.
3.5. Evidence for cyclical changesFigures 3–5 show the changes in spectral density, plotted against the number of
cycles per year for measles (1861–1969), pertussis (1861–1956) and influenza (1861–1990). Cubic regression coefficients for the three diseases in the three populationsare shown in table 5 and the residuals from these regressions are also shown infigures 3–5.
Infectious disease in the Outer Hebrides 463
Figure 3. Measles: spectral densities (continuous lines) and residuals (dashed lines) from cubicregression (table 5), plotted against numbers of cycles per year.
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
3.5.1. Measles (figure 3)For the period 1861–1969, all three populations showed cycles of 7.3–7.8 years,
the least marked being that of Scotland. In addition, Barra had a poorly markedcycle of 2.5 years, and Scotland a pronounced 2-year cycle.
3.5.2. Pertussis (figure 4)Between 1855 and 1956, Harris and Barra each had cycles of 7.4 years. Harris had
a less defined cycle of 3.2 years. Scotland had an ill-defined cycle of about 4 years,and a well-marked one of 2 years.
464 E. J. Clegg
Figure 4. Pertussis: spectral densities (continuous lines) and residuals (dashed lines) from cubicregression (table 5), plotted against numbers of cycles per year.
Table 5. Coefficients (standard errors) of cubic regressions of spectral densities on frequencies.
Disease Population Constant Frequency Frequency2 Frequency3 r2
Measles Harris 0.0363 (0.0270) 4.4774 (0.4761) 722.6220 (2.2248) 28.1326 (2.9771) 0.799Barra 0.0453 (0.0239) 3.8861 (0.4210) 720.0150 (1.9451) 25.3883 (2.6326) 0.782Scotland 0.0051 (0.0008) 0.0763 (0.0119) 70.5036 (0.0562) 0.7914 (0.0745) 0.794
Pertussis Harris 0.0231 (0.0255) 4.8162 (0.4463) 721.4660 (2.0867) 25.0286 (2.7418) 0.790Barra 0.0394 (0.0558) 10.1044 (0.9757) 745.0920 (4.5616) 52.9349 (5.9939) 0.764Scotland 0.0051 (0.0004) 0.0321 (0.0158) 70.2107 (0.7392) 0.3436 (0.0971) 0.432
Influenza Harris 0.2274 (0.0298) 1.3143 (0.5223) 710.6090 (0.4418) 15.0848 (3.2085) 0.750Barra 0.1130 (0.0300) 7.1475 (0.5251) 734.6900 (2.4550) 42.1480 (3.2259) 0.890Scotland 0.0105 (0.0012) 0.1643 (0.0210) 70.8858 (0.0982) 1.1078 (0.0291) 0.845/
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
3.5.3. Influenza (figure 5)During the years 1861–1990, Harris had a main cycle of 7.4 years, and Barra one
of 6.9 years. In addition, Harris had a subsidiary cycle of about 2.6 years, and Barraone of 2 years. Scotland had a main cycle of 8 years.
The cubic regression coefficients (together with standard errors), are shown intable 5. All equations are of the form:
y ¼ b0 þ b1x� b2x2 þ b3x
3
but it will be seen that while the respective coefficients for Harris and Barra aregenerally large, those for Scotland are much smaller. For measles, coefficients forHarris and Barra are closely similar, but this is not the case for pertussis or influenza:for both diseases all coefficients other than b0 are greater (in absolute terms) in Barra.
The residuals from these regressions, shown in figures 3–5 generally show peakscorresponding to the major cycle frequencies. In all cases, normal probability plotsof the standardized residuals (not shown) gave good approximations to straightlines, suggesting that the regression equations are valid ones.
4. Discussion
The results show that for these two islands, the period 1855–1990 was one ofdiminishing mortality from these three acute infectious diseases. This general ame-lioration is reflected in the increased ages at death during the later part of the period,
Infectious disease in the Outer Hebrides 465
Figure 5. Influenza: spectral densities (continuous lines) and residuals (dashed lines) from cubicregression (table 5), plotted against numbers of cycles per year.
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
as shown in table 1. However, caution in interpretation must be exercised here, aslatterly, numbers of deaths were small.
The results of the regression analyses (table 3), suggest that considerable propor-tions of the total numbers of deaths from measles and pertussis can be explained bythe independent variables of time and/or numbers of susceptibles (births over thedecade encompassing a Census year). For measles, the only significant coefficientswere for births in Barra and Scotland. For pertussis, all populations had significantcoefficients for numbers of births, and Scotland had an additional significant (andnegative) relationship with time, persisting after stepwise regression; for the othertwo populations these coefficients, although negative, were eliminated after the firststep. These results suggest that the principle of ‘mass action’ in disease transmission(Hamer 1906) applies, at least in respect of population density of susceptibles.
The findings confirm to some extent those of Rhodes and Anderson (1996a,b) andRhodes et al. (1997, 1998) that (negative) power relations exist in small isolatedpopulations between epidemic frequencies and their sizes or durations. However,the sizes of the exponents in this study (table 4) are generally less than those ofthese workers.
The differences between Harris and Barra in these results are interesting. Thereseems to be in both islands a relationship between epidemic size and frequency,although it is much stronger in Harris, possibly because of the greater number ofepidemics. But for the associations between epidemic length and frequency, only inHarris are they statistically significant. Even in Harris the associations are weakerthan between size and frequency, possibly due to the smaller range of epidemiclengths as compared with sizes. This in turn may be a consequence of the rather‘coarse’ units (quarters) employed. Differences in size between Harris and Barra mayalso explain, at least in part, the differing results of curve-fitting.
In considering epidemic sizes and durations, PMRs and spectral analyses, it mustbe remembered that the data are of mortalities, not infections. Whether an individualeither dies of or recovers from a particular disease may depend not only on theseverity of infection, but on the many other factors, including intercurrent illness,the state of nutrition, or the quality of care.
The values of PMRs for measles and pertussis (figure 1) and influenza (figure 2)show (with the reservation that the issue is one of mortality, not morbidity) thatneither Harris nor Barra had populations large enough to sustain the diseases inendemic form (see Bartlett 1957, Cliff and Haggett 1988, p. 245). Epidemics of eachdisease show mortality levels generally higher than those in Scotland, almost cer-tainly due to the small sizes of the islands, with consequent ease of spread so that allor nearly all of the populations were exposed to the infection. Generally, epidemiclevels were higher in Barra than in Harris, again possibly reflecting that island’ssmaller size.
Influenza PMRs (figure 2) in both Harris and Barra show peaks around 1890 and1900—periods when major outbreaks occurred (Schild 1977)—but not particularlyduring 1918–1920, when a global pandemic occurred. Scotland showed increasedmortality at all these three time periods, with the maximum in 1922. It is difficultto understand why Harris and Barra appear to have escaped the 1918–1920 pan-demic. Certainly they were relatively isolated, but contact with the mainland wascontinuous, and the potential for disease transmission must have been ever-present.
Deaths from these three diseases are associated with poor standards of living,especially poverty, overcrowding and inadequate housing. As mentioned in the
466 E. J. Clegg
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
Introduction, the populations of both islands increased rapidly before World War 1,despite substantial emigration, and decreased equally rapidly subsequently, resultingfrom continued emigration and decreasing nuptiality (Clegg 1999) and fertility(Clegg and Cross 1995). Up to 1914, the demographic pattern was of fairly lownuptiality, but high (although decreasing) fertility within marriage (Anderson andMorse 1993, Clegg and Cross 1995). This period was also characterized by recurrentfluctuations in agriculture and in the fishing industry, especially during the 1880s,when the price of wool fell steeply, and the lucrative Baltic market for herringcollapsed (Hunter 1976, p. 171, Gray 1979). There was considerable hardshipthroughout the Highlands and Islands, which resulted in the establishment in 1883of the Napier Commission, and in 1886 in the first of the Crofters Acts, to give somesecurity of tenure to crofters. The final administrative step was the setting up in 1897of the Highland Congested Districts Board, which while effecting significantimprovements in the infrastructure of the Highlands and Islands, was less successfulin reducing the congestion.
Many contemporary writers have given graphic accounts of the economic depri-vation of the Highlands and Islands during the latter part of the 19th century.Several factors combined to cause this state of deprivation—the consequences ofthe Clearances of the earlier part of the century, when crofters were removed fromgood arable land to make way for large farms, and good grazing was expropriatedfor sheep, the result being that crofting townships were often relocated to poor,circumscribed sites; the potato blight of 1847, which had a severe effect on a popula-tion already suffering from the effects of the Clearances; and the insecurity of tenureand the absence of compensation for improvements to a which a crofter might maketo his croft. It seems likely that overcrowding, poverty, ignorance, lack of medicalcare and large families all contributed to the high mortality rates from measles andpertussis.
The ‘land hunger’ seen in the decades around the turn of the century reappearedafter World War 1. However, population density was decreasing, and fertility wasfalling, and therapeutic and especially public health measures reduced the impact ofinfectious disease on populations. Hence, although there was still economic depriva-tion in the Highlands and Islands at this time, the impact of infectious disease onmortality was much reduced.
The results of spectral analyses (figures 3–5) show that Harris and Barra had cyclelengths for all three diseases of between 6.9 and 7.8 years. Scotland had a 7.3-yearcycle for measles, an 8-year cycle for influenza, and marked 2-year cycles for measlesand pertussis. These results differ in some ways from those of some previous authors.For populations large enough for measles to be endemic (e.g. Fiji, French Polynesia(Cliff and Haggett 1988, p. 196), England and Wales, Manchester, Aberdeen,Frankfurt-am-Main (Anderson et al. 1984, Anderson and May 1991)), intervalsbetween epidemics are generally of 1 or 2 years, the predominant one varying. InChile (1962–1988), Canals (1992) found a cycle of approximately 4 years, while inPortugal, Gomes and Paulo (1999) found a 3-year interval in the pre-vaccinationperiod, and Duncan, Duncan and Scott (1997) found intervals declining from 5 to 2years in London between 1647 and 1837. In Copenhagen, Olsen, Truty and Schaffer(1988) found cycles of mortality of 2.5 years, as well as an annual cycle. From thepresent study, in which the annual nature of the data precludes the identification ofan annual cycle, Scotland has a 2-year periodicity, for measles, but its additional 7.3year cycle is unexpected. Where population sizes are small, with some years with no
Infectious disease in the Outer Hebrides 467
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
mortality, the interval between epidemics may be considerable. In the Aland Islands,
Mielke (1996) found a main cycle of 7.24 years, not too different from the present
findings, together with a less marked one of about 3 years. Perhaps the 7-year cycle
of Scotland is a reflection of its geographical and demographic heterogeneity, the
Central Belt and main cities behaving like Bartlett’s (1957) cities, with measles
always present (Cliff and Haggett 1988, p. 245, Type I), while the much less densely
populated rural areas (especially the Highlands and Islands) have no endemic
measles, and fall into Cliff’s and Haggett’s Type II, where periods of no measles
are interspersed with regular epidemic cycles, or Type III, where epidemics occur
irregularly. The present evidence suggests that both Harris and Barra, despite their
small populations, fall into Type II.
For pertussis, the pre-vaccination 7.4 year cycle seen in Harris and Barra is not
seen in the Scottish data. Instead there is a 3.2-year cycle, not dissimilar to the 3-year
cycle in England and Wales, noted by Anderson et al. (1984) and Anderson and May
(1991). In addition, there is a better-marked 2-year cycle. Both these last cycles are
seen to a minor extent in Harris.
Other populations show cycle lengths varying from 2.5 to 5 years. In pre-vaccina-
tion Portugal (Gomes and Paulo 1999), the cycle was of 2.5–4 years’ duration. In
London, Duncan, Duncan and Scott (1996) found cycles varying between 3 and 5
years in length between 1720 and 1812, and the same authors (1998) found cycles in
Liverpool between 1863 and 1900 to be between 2.9 and 3.4 years in length. Olsen
et al. (1988) found that in Copenhagen, pertussis was largely endemic, but there was
a small annual cycle.
The 7.4-year cycles in Harris and Barra might be explained, as with measles, by
the length of time needed to build up a sufficient population of susceptibles for an
epidemic to occur. The approximately 3-year and 2-year cycles in Harris and
Scotland seem in agreement with previous work.
The spectra for pertussis, although of the same general shape in Harris and Barra,
have very different cubic regression coefficients (table 5), suggesting that the two
islands are in this respect, epidemiologically distinct. However, for measles, the
coefficients of the regressions are closely similar. This is a surprising finding, given
the geographical separation between the two islands, and their different routes of
communication with the mainland. Perhaps it is due to an inherent ‘stability’ of
measles epidemics.
The spectra for influenza show main peaks at between 6.9 and 8 years for all three
populations, with subsidiary peaks at 2.6 and 2 years, respectively, for Harris and
Barra. Generally, influenza epidemics occur on an annual cycle, the peak being
reached in the winter. In addition there are continental or global pandemics at
intervals from 10 to 40 years, the results of major changes in the surface antigens
of the virus (Stuart-Harris, Schild and Oxford 1985). A model explaining the occur-
rence of epidemics (Kilbourne 1973) hypothesizes that after a major antigenic shift,
resulting in a pandemic, increasing numbers in a population come to possess the
appropriate antibodies, and subsequent epidemics, decreasing in intensity and occur-
ring at increasing intervals, are due to minor antigenic changes. Eventually another
major antigenic shift occurs, and the cycle is repeated. The present findings suggest a
rather shorter inter-epidemic interval for Scotland.
468 E. J. Clegg
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
Acknowledgements
I am grateful to Dr J. F. Cross for her help in the acquisition of the data, to theRegistrar General for Scotland, for permission to examine the registers, and to hisstaff for their willing and courteous assistance. I am also indebted to the Economicand Social Research Council, the Leverhulme Trust, and the Carnegie Trust for theScottish Universities, for their generous financial support.
ReferencesAnderson, R. M., Grenfell, B. T., and May, R. M., 1984, Oscillatory fluctuations in the incidence of
infectious disease and the impact of vaccination: a time-series analysis. Journal of Hygiene, 93, 587–608.
Anderson, R.M., and May, R.M., 1991, Infectious Diseases of Humans: Dynamics and Control (Oxford:Oxford University Press).
Anderson, M., and Morse, D. J., 1993, High fertility, high emigration, low nuptiality: adjustmentprocesses in Scotland’s demographic history. Population Studies, 47, 5–25 and 319–343.
Bartlett, M. S., 1957, Measles periodicity and community size. Journal of the Royal Statistical Society,120A, 48–70.
Canals, M., 1992, Measles in Chile: a dangerous upsurge. Revisto di Medicina de Chile, 120, 585–588.Clegg, E. J., 1999, Probabilities of marriage in the Outer Hebrides. Journal of Biosocial Science, 31, 167–
193.Clegg, E. J., and Cross, J. F., 1995, Religion and fertility in the Outer Hebrides. Journal of Biosocial
Science, 27, 79–94.Cliff, A. D., and Haggett, P., 1988, Atlas of Disease Distributions: Analytic Approaches to
Epidemiological Data (Oxford: Blackwell).Duncan. C. J., Duncan, S. R., and Scott, S., 1996, Whooping cough epidemics in London, 1701–1812:
infection, dynamics, seasonal forcing and the effects of malnutrition. Proceedings of the RoyalSociety, 263B, 445–450.
Duncan, C. J., Duncan, S. R., and Scott, S., 1997, The dynamics of measles epidemics. TheoreticalPopulation Biology, 52, 155–163.
Duncan, C. J., Duncan, S. R., and Scott, S., 1998, The effects of population density and malnutritionon the dynamics of whooping cough. Epidemiology and Infection, 121, 325–334.
Gomes, M. C., and Paulo, A. C., 1999, Diphtheria, pertussis and measles in Portugal before and aftermass vaccination: a time series analysis. European Journal of Epidemiology, 15, 791–798.
Gray, M., 1979, The Fishing Industries of Scotland, 1790–1914: a Study in Regional Adaptation (Oxford:Oxford University Press).
Hamer, W. H., 1906, Epidemic disease in England. Lancet 733–739.Hunter, J., 1976, The Making of the Crofting Community (Edinburgh: James Thin).Kilbourne, E. D., 1973, The Influenza Viruses and Influenza (New York: Academic Press).Mausner, J. S., and Bahn, A. K., 1974, Epidemiology (Philadelphia: Saunders).Mielke, J. H., 1996, Historical epidemiology of measles and scarlet fever in Aland, Finland. Rivista di
Antropologia, 74, 127–138.Olsen, L. F., Truty, G. L., and Schaffer,W.M., 1988, Oscillations and chaos in epidemics: a nonlinear
dynamic study of childhood diseases in Copenhagen, Denmark. Theoretical Population Biology, 33,344–370.
Priestley, M. B., 1981, Spectral Analysis and Time Series, vol. 1, Univariate Series, (London: AcademicPress).
Rhodes, C. J., and Anderson, R. M., 1996a, Power laws governing epidemics in isolated populations,Nature, 381, 600–602.
Rhodes, C. J., and Anderson, R.M., 1996b, Scaling analysis of measles epidemics in a small population.Philosophical Transactions of the Royal Society, 351B, 1679–1688.
Rhodes, C. J., Jensen, H. J., and Anderson, R. M., 1997, On the critical behaviour of simple epidemics.Proceedings of the Royal Society, 264B, 639–646.
Rhodes, C. J., Butler, A. R., and Anderson, R. M., 1998, Epidemiology of a communicable disease insmall populations. Journal of Molecular Medicine, 76, 111–116.
Schild, G. C., 1977, Influenza. In A Geography of Human Disease, edited by G. Howe (London,Academic Press), pp. 339–376.
Steel, T., 1976, The Life and Death of St Kilda (Glasgow: Fontana/Collins).Stuart-Harris, C. H., Schild, G. C., and Oxford, J. S., 1985, Influenza—the Viruses and the Disease,
3rd edn (London: Edward Arnold).Sumi, A., 1998, Time series analysis of surveillance data of infectious diseases in Japan. Hokkaido Igaku
Zash, 73, 343–363.
Infectious disease in the Outer Hebrides 469
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
Tukey, J. W., 1949, The sampling theory of power spectrum estimates. In Proceedings of a Symposium onApplications of Autocorrelation Analysis to Physical Problems, Office of Naval Research,Washington, pp. 47–67.
Address for correspondence: E. J. Clegg, Department of Biomedical Sciences, Marischal College,Aberdeen AB9 1AS, UK. Email: [email protected]
Zusammenfassung. Ziele: Zweck der Untersuchung war erstens die Analyse derSterblichkeitsveranderungen, die im Zeitraum 1855 bis 1990 durch Masern, Keuchhusten und Grippe(alle epidemischen Krankheiten) auf den beiden Inseln der Außeren Hebriden Harris und Barra verursachtwurden und zweitens der Vergleich dieser Ergebnisse mit denen, die im selben Zeitraum fur ganzSchottland ermittelt werden konnten. Ebenso war beabsichtigt, die Mortalitatsveranderungen mit denVeranderungen sozialer und okonomischer Faktoren in Beziehung zu setzen.Material und Methoden: Alter und Todesursachen auf Harris und Barra wurden mit Hilfe von Kopien derSterbeurkunden, die im General Register Office in Edinburgh und fur ganz Schottland in den Jahres-berichten des allgemeinen Standesamtes Schottlands archiviert sind, ermittelt. Die Standardisierung derDaten erfolgte durch Berechnung des Proportionalen Mortalitatsverhaltnisses (PMR, Verhaltnis der Zahlder an einer bestimmten Ursache Gestorbenen zur Gesamtzahl Gestorbener uber einen gegebenenZeitraum). Zur Analyse der Dauer epidemischer Zyklen wurde die Spektralanalyse angewandt.Ergebnisse: Das Sterbealter erhohte sich leicht uber den untersuchten Zeitraum. Fur Masern undKeuchhusten gab es in Harris, anders als fur den fruheren Zeitraum, signifikante Beziehungen zwischender Anzahl Gestorbener pro Dekade und der Anzahl der neuen Krankheitsgefahrdeten, geschatzt aus derAnzahl der Geburten. Masern- und Keuchhustenepidemien treten auf den Inseln in Intervallen auf,normalerweise durch Jahre ohne Mortalitat getrennt.Die hochsten PMR traten im allgemeinen wahrend der spateren Jahrzehnte des 19. und der ersten Dekadedes 20. Jahrhunderten auf; dies kann im Zusammenhang mit den wirtschaftlichen Problemen in derLandwirtschaft und dem Fischfang und der erhohten Populationsdichte stehen. Influenza - Epidemienwaren haufiger als jene der beiden anderen Krankheiten.Fur alle drei Krankheiten es gab auf beide Inseln signifikant negative Zusammenhange zwischen demAusmaß der Epidemie und der Haufigkeit des Auftretens; jene in Harris waren die starkeren. DieRelationen zwischen Lange und Haufigkeit waren nur in Harris signifikant. Im Allgemeinen schienendie Langen der Epidemien weniger zu variieren als das Ausmaß, vielleicht wegen der eher ‘groberen’angewendeten Langeneinheiten (Quartale).Die Spektralanalyse der ‘ungerichteten’ Daten fur den Zeitraum vor der Einfuhrung der spezifischenImmunprophylaxe zeigte, dass in allen drei Bevolkerungen bei den Masern der hautsachliche Zyklusder Epidemien zwischen 7,3 und 7,8 Jahren Dauer betrug. Barra und Schottland hatten zusatzliche 2,5-beziehungsweise 2-Jahres-Zyklen. Fur Keuchhusten wiesen Harris und Barra Hauptzyklen von 7,4 Jahrenauf. Harris hatte einen zusatzlichen Zyklus von 3,2 Jahren, Schottland Zyklen von 4 und 2 Jahren. FurGrippe ließen sich fur Harris eine Hauptzykluslange von 7,4 Jahren und ein wenig definierbarer Zyklusvon ungefahr 2,6 Jahre nachweisen. Barra hatte ein Hauptzyklus von 6,9 und einen unterstutzenden von 2Jahren. Schottland wies einen Einzelzyklus von 8 Jahren auf. Kubische Regression der Spektraldichten aufdie Zyklusfrequenzen zeigte große Koeffizienten fur Harris und Barra, aber kleinere fur Schottland. DieKoeffizienten fur Masern waren fur die zwei Inseln relativ ahnlich, aber nicht die fur Keuchhusten.Schlussfolgerungen: Die Befunde demonstrieren das episodische Auftreten von Epidemien dieser dreiKrankheiten auf den zwei Inseln, verglichen mit ihrer dauernden Prasenz in der wesentlich großerenBevolkerung Schottlands. Die Ergebnisse zeigen auch die verminderte Bedeutung dieser Todesursachenin allen drei Bevolkerungen. Die Daten von Harris und Barra legen nahe, dass Masern eine mehr epide-miologisch ‘stabile’ Krankheit als Keuchhusten ist. Beide Inseln scheinen Hamer’s Gesetz von‘Mengenaktion’ zu gehorchen. Die relativen langen Intervalle zwischen den Epidemien auf den Inselnkonnte zum einen auf ihre Isolierung zuruckzufuhren sein und zum anderen auf die langsame Zunahmeeiner ausreichenden Anzahl von Krankheitsgefahrdeter, durch die erst das Auftreten einer Epidemiemoglichen wird.
Resume. Objectifs: Cette etude a pour but d’etudier les changements de la mortalite par oreillons,coqueluche et grippe (toutes des maladies epidemiques) de 1855 a 1990, dans les ıles d’Harris et deBarra qui appartiennent aux Hebrides Exterieures et d’en comparer les resultats pour la meme periodeavec ceux de l’ensemble de l’Ecosse. On a egalement tente de relier les changements de la mortalite avec leschangements economiques et sociaux.Materiels et methodes: Les ages et les causes de deces dans les ıles d’Harris et Barra ont ete etablis a partirde copies de certificats de deces conserves au General Register Office d’Edimbourg. Pour l’Ecosse engeneral, ils proviennent des Registres generaux d’Ecosse. Les donnees ont ete standardisees par le calculd’un Rapport Proportionnel de Mortalite (RPM): la proportion des deces due a une cause particuliere parrapport a tous les deces pendant une periode donnee. La duree des cycles epidemiques a ete examinee paranalyse spectrale.Resultats: Les ages au deces augmentent legerement sur l’ensemble de la periode etudiee. Il y a uneassociation entre le nombre de deces par oreillons et coqueluche par decennie et le nombre potentiel de
470 E. J. Clegg
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
nouveaux malades, estime par le nombre de naissances, mais seulement pour la coqueluche dans Harris.Les epidemies d’oreillons et de coqueluche se produisaient par intervalles dans les ıles, habituellementseparees par des annees sans mortalite. Les RPM les plus eleves se rencontrent generalement au cours desdernieres decennies du 19eme siecle et de la premiere decennie du 20eme siecle, peut etre en liaison avec lesproblemes economiques de l’agriculture et de la peche, ainsi qu’avec l’accroissement de la densite humaine.Les epidemies de grippe sont plus frequentes que celles des deux autres maladies. On trouve dans les deuxıles pour chacune des trois maladies, une association negative significative entre l’importance et la fre-quence des epidemies, celles qui frappaient Harris paraissant les plus fortes. L’association entre duree etfrequence n’est significative que dans Harris. D’une maniere generale les durees des epidemies paraissentmoins variables que leur etendue, peut etre a cause de la grossierete de l’unite d’evaluation: le trimestre.L’analyse spectrale des donnees pour la periode anterieure a l’introduction d’une immunoprophylaxieappropriee, une fois elimine l’effet de tendance diachronique, revele que le principal cycle epidemique desoreillons dans les trois populations, avait une duree s’echelonnant entre 7,3 ans et 7,8 ans, Barra et l’Ecosseayant un cycle additionnel respectif de 2,5 et 2 ans. Pour la coqueluche, Harris et Barra avaient un cycleprincipal de 7,4 ans et Harris avait un cycle additionnel de 3,2 ans tandis que l’Ecosse avait des cycles de 4et 2 ans. Pour la grippe, Harris avait un cycle principal de 7,4 ans et un autre moins bien defini d’environ2,6 ans. Barra avait un cycle principal de 6,9 ans et un cycle subsidiaire de 2 ans. L’Ecosse avait un seulcycle de 8 ans. Des regressions cubiques de densites spectrales sur les frequences des cycles montrent degrands coefficients pour Harris et pour Barra et des coefficients petits pour l’Ecosse. Les coefficients desoreillons sont proches dans les deux ıles, mais ce n’est pas le cas de ceux de la coqueluche.Conclusions: Les resultats montrent l’apparition episodique d’epidemies de ces trois maladies dans les deuxıles, qui contrastent avec leur presence permanente dans la population bien plus grande que representel’ensemble de l’Ecosse. Ils revelent aussi que ces causes de deces ont vu leur importance diminuer dans lestrois populations. Les donnees de Harris et de Barra suggerent que les oreillons sont une maladie epide-miologiquement plus stable que la coqueluche. Les deux ıles semblent suivre la loi de Hamer ‘‘d’action demasse’’. Les intervalles entre epidemies relativement plus longs dans les ıles, sont peut etre dus pour unepart a leur isolement et pour une autre part a la lenteur de l’accumulation d’un nombre suffisant de‘‘susceptibles’’ qui permette a l’epidemie d’eclater.
Resumen. Objetivos: El proposito del estudio fue examinar los cambios en la mortalidad por sarampion,tos ferina y gripe (todas ellas enfermedades epidemicas) en Harris y Barra, dos islas Outer Hebridas, desde1855 a 1990, y comparar los resultados con los del conjunto de Escocia durante el mismo periodo.Tambien se intento relacionar los cambios en la mortalidad con los de los factores sociales y economicos.Material y metodos: Se obtuvieron las edades y causas de fallecimiento en Harris y Barra a partir de copiasde los certificados de defuncion depositados en la Oficina del Registro General, en Edimburgo y, para elconjunto de Escocia, a partir de los Informes Anuales de los Archiveros Generales de Escocia. Los datos seestandarizaron mediante el calculo de la Razon de Mortalidad Proporcional (PMR), es decir, la propor-cion de muertes debida a una causa particular frente a todas las muertes durante un periodo concreto. Seempleo el analisis espectral para examinar la duracion de los ciclos epidemicos.Resultados: Las edades de muerte aumentaron ligeramente durante el periodo de etudio. En relacion alsarampion y a la tos ferina, aparte de para las previas en Harris, existıan relaciones significativas entre elnumero de fallecimientos por decada, y el numero de nuevos casos susceptibles, estimado como el numerode nacimientos. Las epidemias de sarampion y tos ferina en las islas ocurrıan a intervalos, separados engeneral por anos de no mortalidad. Las mayores PMRs se obtuvieron generalmente durante las ultimasdecadas del siglo XIX y primera decada del XX; esto puede haber estado relacionado con los problemaseconomicos de la agricultura y pesca, y con el incremento de la densidad de poblacion. Las epidemias degripe fueron mas frecuentes que las de las otras dos enfermedades. En las tres enfermedades se dieronimportantes relaciones negativas entre la magnitud de la epidemia y la frecuencia de aparicion en ambasislas; las de Harris fueron las mas fuertes. Las relaciones entre duracion y frecuencia solo fueron signifi-cativas en Harris. Generalmente, las duraciones epidemicas parecıan menos variables que los tamanos,posiblemente debido al empleo de unidades de duracion (cuartos) bastante ‘‘groseras’’. El analisis espectralde los datos ‘‘suavizados’’ para el periodo anterior a la introduccion de inmunoprofilaxis especıfica reveloque, para el sarampion, el principal ciclo epidemico en las tres poblaciones tuvo una duracion de entre 7,3y 7,8 anos. Barra y Escocia tuvieron ciclos adicionales de 2,5 y 2 anos, respectivamente. Para la tos ferina,Harris y Barra tuvieron ciclos principales de 7,4 anos. Harris tuvo un ciclo adicional de 3,2 anos. Escociatuvo ciclos de 4 y 2 anos. Para la gripe, Harris tuvo un ciclo principal de 7,4 anos de duracion, y otromenos definido de unos 2,6 anos. Barra tuvo un ciclo principal de 6,9 anos, y otro subsidiario de 2 anos.Escocia tuvo un unico ciclo de 8 anos. Las regresiones cubicas de las densidades espectrales sobre el ciclode frecuencias mostraron grandes coeficientes para Harris y Barra, pero pequenos para Escocia. Loscoeficientes del sarampion fueron muy similares en las dos islas, pero no los de la tos ferina.Conclusiones: Los resultados demuestran la aparicion episodica de epidemias de estas tres enfermedades enlas dos islas, frente a su presencia continuada en la mayor parte de la poblacion de Escocia. Asimismo,revelan la importancia decreciente de estas causas de muerte en las tres poblaciones. Los datos de Harris yBarra sugieren que el sarampion es una enfermedad mas ‘‘estable’’ epidemiologicamente que la tos ferina.Ambas islas parecen obedecer a la ley de Hamer de ‘‘accion de masas’’. Los intervalos relativamente largosentre epidemias en las islas pueden ser debidos en parte a su aislamiento, y en parte tambien a la lentaacumulacion de un numero suficiente de susceptibles como para permitir la aparicion de la epidemia.
Infectious disease in the Outer Hebrides 471
Ann
Hum
Bio
l Dow
nloa
ded
from
info
rmah
ealth
care
.com
by
UB
Kie
l on
11/0
9/14
For
pers
onal
use
onl
y.
