lschaemic Heart Disease Mortality and The Business Cycle in Australia ALFRED REX BUNN, DSR, BCOM, MEc

Abstract: Trends in Australian heart disease mortality were assessed for association with the business cycle. Correlation models of mortality and unemployment series were used to test for association. An indicator series of "national stress" was developed. The three series were analyzed in path models to quantify the links between unemployment, national stress, and heart disease. Ischemic heart disease (IHD) mortality and national stress were found to follow the business cycle. The two periods of accelerating IHD mortality coincided with economic reces-

sion. The proposed "wave hypothesis" links the trend in IAD mortality to the high unemployment of severe recession. The mortality trend describes a typical epidemic parabolic path from the Great Depression to 1975, with a smaller parabolic trend at the 1961 recession. These findings appear consistent with the hypothesis that heart disease is, to some degree, a point source epidemic arising with periods of severe economic recession. Forecasts under the hypothesis indicate a turning point in the mortality trend between 1976 and 1978. (Am JPublic Health 69:772-781, 1979.)

In Australia as in other countries, mortality from ischaemic heart disease (IHD) rose by the 1960s to an epidemic incidence.' In the mid-1960s, a plateau occurred. Since 1968, IHD mortality has steadily declined. No satisfactory explanation of these trends exists; the post-1968 decline is particularly puzzling. Woodruff2 has pointed to the influence of competitive mortality, i.e., the decline in respiratory disease mortality after chemotherapy innovations and coincident with the increase in IHD mortality. However, Lancaster3 has observed that the greater part of mortality from infectious diseases had disappeared before the introduction of chemotherapy. Despite extensive research, the causal pathways of IHD are a "grey area". The risk factor paradigm built around hypertension, hypercholesterolemia, and cigarette smoking has shown associations between these factors and IHD. However, proof of a causal relation is lacking.4 Other risk factors are reported to include diabetes, sedentary habits, obesity, stress, water hardness, air pollution, other mineral factors, and a family history of the disease.4 Of central interest in Australia is the decline in IHD mortality after 1968. Two propositions have been advanced for this. Reader5 has suggested that coronary care units have been instrumental. However, as 60 per cent-70 per cent of deaths occur outside hospitals, this explanation seems an unlikely one.

Christie' noted that cigarette consumption increased as IHD mortality declined* and considered it unlikely that Australian serum cholesterol levels had fallen. It has been suggested that anti-hypertensive treatment may have had a beneficial effect on IHD mortality via a reduction in hypertensive mortality.' This view has received some support as a partial explanation.'0 However, other investigators have viewed the decline in hypertensive heart disease as a change in diagnostic fashion." It seems doubtful that antihypertensive treatment alone can account for the decline. That the decline is real is now beyond doubt.' Under these circumstances it seems reasonable to examine other reported risk factors. Recent research12 suggests that a relation exists between stress and IHD, in particular the socioeconomic stress associated with economic recession. This study examines the association between IHD mortality and the business cycle in Australia.

Address reprint requests to Alfred Rex Bunn, 27 Walker Road, Wyoming, 2250, N.S.W., Australia. Mr. Bunn is with the School of Economic and Financial Studies, Macquarie University, N.S.W., Australia. This paper, submitted to the Journal July 6, 1978, was revised and accepted for publication November 30, 1978. Editor's Note: See editorial, page 762, this issue, as well as accompanying articles, pages 782, 784, and 789. 772

Methods (a) Correlation Analysis Residual series of age-specific IHD mortality and unemployment were correlated (see Appendix I for models used). IHD data were bridged over successive revisions of the International Classification of Diseases (ICD) and based *Cigarette consumption in Australia continued to increase in the 1960s and for at least 4 years after the downturn in IHD mortality, i.e., until 1971.6 ' From 1974 to 1976 Gray and Hill found a slight decline in male cigarette smokers, (41 per cent to 40 per cent) and an increase in female cigarette smokers (29 per cent to 31 per cent). The 1977 Australian Bureau of Statistics survey classifies 43 per cent of males and 29 per cent of females as cigarette smokers.9 AJPH August 1979, Vol. 69, No. 8

IHD MORTALITY AND THE BUSINESS CYCLE

in the 7th revision (see Appendix III). Unemployment was measured as the average annual rate. In Australia, unemployment series have been the best single indicator series of the business cycle.'3 This approach therefore relates IHD to the business cycle. Three periods are assessed, each of a different socioeconomic experience, i.e., before, during, and after the Great Depression. IHD mortality clearly lagged behind unemployment in each period. The lag range, i.e., 0-5 years, follows past research. 14. 15 Considerable statistical investigation was undertaken to detect the appropriate lag as in preceding work by Armstrong, et al, 14 and Brenner. 15 The recent view of Eyer'6 questioning both the existence of lagged associations and the link between economic downturns and mortality was also tested (see Appendix II). (b) Path Analysis This is a special regression technique for examining relations among several variables.'7 Past research linking economic trends with IHD has been unable to measure the supposed link, i.e., economic stress.'2 In Australia there is one series which reflects national stress. This is a series of Pharmaceutical Benefits Prescriptions recorded by the Commonwealth Department of Health. These are mainly written by General Practitioners and have accounted for 85 per cent of all ethical drug prescribing.'8 There has been a strong one-to-one relationship between general practice consultations and general practice prescribing.'9 It is therefore possible to use the Parmaceutical Benefits Scheme (PBS) series as an indicator of consultations. By assuming that when people feel ill they are likely to consult their General Practitioner, then the PBS series stands as an indicator series of national stress or morbidity. It reflects the supply of medical assistance in terms of ethical drug prescriptions which, in turn, reflects the effective demand for such assistance (see Appendices I and IV for model details). (c) Cohort Analysis In a 50-year period of study, inter-generational differences are likely to appear.' To overcome this, the IHD mortality for 18 cohorts was calculated for the period 1920-

1975. (d) Curve Fitting under the Wave Hypothesis The results from approaches (a) (b) and (c) lead to the wave hypothesis. This links the trend in IHD mortality to the Great Depression and the 1961 recession. To elucidate this, second degree polynomial curves were fitted to the mortality series for the period 1931-1975 with interpolation across the period 1962-1%7. The latter period was also fitted with second degree polynomial curves.

Findings (a) Correlation Analysis In Table 1 below, are shown the correlation findings for six age/sex specific groups. AJPH August 1979, Vol. 69, No. 8

TABLE I -AssocIations between Unemployment and lschaemIc Heart Disase Mortality among Males & Females, 35-64 years, 1921-1975-Detrended, standardized series. Sex and Age

Data Series

Group

R* & Significance# limit

Number of Years

Years of

10 8 25

3 3 3

0.632 (0.05) 0.122 (N/S) 0.628 (0.001)

10 7 24

3 1 3

0.618 (0.05) 0.680 (0.05) 0.524 (0.01)

9 9 22

3 1 3

0.778 (0.05) 0.784 (0.05) 0.647 (0.01)

9 9 23

3 1 5

0.510 (0.1) 0.469 (N/S) 0.448 (0.05)

8 8 24

4 1

0.760 (0.05) 0.776 (0.05) 0.587 (0.01)

9 7 28

4 0 5

Lag

Males 35-44 years

1923-1932 1935-1942 1951-1975 45-54 1923- 1932 1935-1941 1951-1974

55-64 years

1924-1932 1933-1941 1953- 1 974 Females 35-44 years 1922-1930 1933-1941 1947-1969 45-54 years

1924-1931 1936-1943 1952-1975

4

55-64 years 1925-1933 1934-1940 194- 1 975

0.629 (0.1)

0.812(0.05) 0.424 (0.05)

*Correlation coefficient.

#Significance limits for r. Documenta Geigy.

In each period, unemployment is associated with IHD mortality. The association appears strongest in the older groups with higher mortality. It is interesting that the lag shortens during the Great Depression. This would be consistent with theories about precipitation of disease.20

(b) Path Analysis The findings in Table 2 extend the age/sex group findings in (a). The path coefficients** in the column headed "p21" show that 18 months after a change in unemployment, there is a corresponding change in drug prescribing and, by extension, general practice consultations. These events are consistent with a change in national stress. The "p31" coefficients show that three to five years after a change in unemployment, there is a similar change in IHD mortality. The "p32" coefficients relate prescribing to mortality. There are no consistent findings from this pathway. Were age/sex/disease specific prescribing data available, then this pathway approach could probably be developed to quantify any links between prescribing and mortality. **Path coefficients are standardized B coefficients.

773

BUNN TABLE 2-Relations between lschaemic Heart Disease Mortality, Prescribing and Unemployment, 1955-1975, Males and Females 35-64 years. Data Series Mortality

Age/sex Group

1955-1975

Males 35-44

21

1955-1975

45-54

1955-1974 1956-1975

55-64 Females 35-44

1956-1975 1957-1975

Years ofo Number Lag of years Males~~~~~~~~~~~~~~~~

p31V

R*

p21 #

p32s

3

0.654

33 0.543

0.034

0.635

21

3

0.558

0.543

-0.140

0.621

20

3

0.659

0.470

-0.060

0.685

20

5

0.525

0.535

0.154

0.427"

45-54

20

4

0.641

0.470

-0.448"

0.715'

55-64

19

5

0.681

0.466

-0.708'

0.592

a

a

a

a

Olag between unemployment and mortality *coefficient of multiple correlation #p21 path between unemployment and prescribing Sp32 path between prescribing and mortality Vp3l path between unemployment and mortality 'Significance limit = 0.001 *Significance limit = 0.01 "Significance limit = 0.05 "Significance limit = 0.10

(c) Cohort Analysis Figure 1 shows the IHD mortality sustained by nine male cohorts born by decades between 1850 and 1930. The dotted line joins the cohorts' mortality in the year 1930. This period saw the beginning of the IHD epidemic. Each cohort whose members were 30 years of age or older at this time, sustained an exponential increase in mortality thereafter. The increase in deaths was most pronounced in the 1860-cohort. The mortality in this cohort increased by 4500 C

4000_ 3500_ 3000' 2500

2000

,L

1500;

a

Io

o' 1000 i

.C

0 30

35

40

45

50

55

60

65

70

75

80

COHORT AGE

FIGURE 1-Ischaemic Heart Disease Mortality in Nine Male Cohorts Born by Decades between 1850-1930

774

800 per cent between 1930 and 1940. This gave a death rate approximately 350 per cent greater than that of the preceding 1850 cohort, at the same age, i.e., 80 years. The 1850-cohort provides assurance that the exponential rise in IHD mortality was not solely related to aging within cohorts. Even at a cohort age of 80 years, peak IHD mortality reached a rate of 1,000 per 100,000 by 1930 when this cohort was nearly extinct. Between 1930 and 1940, the 1870-cohort mortality increased by 800 per cent, the 1880-cohort by 1500 per cent and the 1890-cohort by 1700 per cent. The sharp contrast between the 1850-cohort and later cohorts' mortality implies some short-term environmental change affecting the cohorts about the 1930 period, rather than any long-term life-style factors, e.g., diet or smoking. The 1900-cohort experienced the highest cohort mortality so far recorded between the ages of 60 and 70 years. The 1910-cohort between the ages of 45 and 58 years records the highest mortality of any cohort before or since. The 1920 and 1930-cohorts record increasing IHD mortality, although at lower levels. Figure 2 shows the nine comparable female cohorts' mortality. It is worth noting that until 1930 the IHD mortality for males and females is similar at each age. For example, the 1850-cohorts have virtually identical lifetime mortality experience. Similar increases occur in the female cohorts, although death rates are generally less than in comparable male cohorts. These increases after 1930 imply, as for the male cohorts, some environmental change about the 1930 period rather than factors operating over a lifetime. They also imply that the change had greater influence on males than females. The migration factor effect in cohorts is discussed in Appendix V. AJPH August 1979, Vol. 69, No. 8

IHD MORTALITY AND THE BUSINESS CYCLE

3500 3000

2500 2000

1500I

.9 3 1000

i S

A 500

250

iso. so 30

35

40

45

55

50

60

65

70

75

s0

COHORT AGE

FIGURE 2-Ischaemkc Heart Disease Mortality in Nine Female Cohorts Born by Decades between 185041930

Figure 3 shows the IHD mortality for males age 55 to 64 years from 1921 to 1975. It is apparent that mortality increases steeply after 1930 and appears to peak in the late 1950s. After 1961, there is a renewed increase reaching a peak in the mid-1960s and declining after 1968. Figure 4 shows the comparable data for 45 and 54-year-old males. Mortality again increases after 1930 and appears to

"plateau" in the late 1950s. There is a renewed increase to a peak in the mid-1960s and a decline thereafter. There appear to be two periods in the mortality series and these also have great economic significance. The 1930s were the years of the Great Depression while the early 1960s saw the next recession. Given a year-by-year association between the business cycle and IHD, it is reasonable to consider what association may exist between IHD and the larger swings in the business cycle, i.e., recession and depression. The "wave hypothesis" states that the scale of IHD mortality associated with the business cycle might be related to the severity of the economic downturn, as measured by unemployment. Extreme economic depression might be associated with epidemic scale mortality from diseases with a high stress component, high mortality rate, and a chronic, irreversible development. Where a significant section of the population is subject to severe economic stress, a phylogenetic maladaption, as postulated by Selye, may occur.2' IHD has a high mortality rate, a chronic development, and may have a high stress component. It is responsible for approximately 33 per cent of all deaths in Australia. Its causal pathways are still being investigated. A model to test the wave hypothesis was constructed on the following assumptions: (a) The small cyclical fluctuations in IHD mortality associated with yearly movements in the business cycle may have larger counterparts associated with recessionary and depression phases of the business cycle; (b) These larger fluctuations or "wave forms" would be linearly related to the small fluctuations; (c) The amplitude and period of the larger waves would

90No

700 F

600 F

8 : 500 I 400 k

300* 200

l00

k

1910

1915

1920

1925

1930

1935

1940

1945

1950

1955

1960

1965

1970

1975

YEAR

FIGURE 3-Ischaemic Heart Disease Mortality in Males 55-4 Years of Age, 1921-1975 AJPH August 1979, Vol. 69, No. 8

775

BUNN

350 k

300

[

250 p

8 k 200 Ias

150

100

50

n

1920

1925

1930

1935

1940

1945

1950

1955

1960

1965

1970

1975

YEAR

FIGURE 4-Ischaemkc Hear Disa

(d) (e) (f) (g)

(h)

(i)

(j)

(k)

776

be determined partly by the level and duration of unemployment in the economy; The limiting period of any wave would be equal to an adult lifetime; Co-existent wave forms would be additive; Wave forms may be modified by other factors, e.g., medical intervention, environmental changes, diet, smoking, other contributing disease; Wave forms are hypothesized to be parabolic in shape where they arise with large scale unemployment and where morbidity is chronic, irreversible and the mortality rate is high and age related. Unemployment of individuals may be considered as a random event; Under this assumption, cohort effects within age groups in the labor force would contribute to a parabolic wave form in age/group mortality, characteristic of an epidemic. However, this would not arise from contagion per se, but rather be "triggered" by the acute level of economic stress at an extreme business cycle trough; The mortality waveform is transmitted through time by structural maladaption in individuals exposed to the economic stress among the cohorts ordinarily present in the labor force. Five have been identified for study, namely those cohorts born by decade and aged 20, 30, 40, 50 and 60 years on average at the time of the economic adversity; The precise shape of a depression waveform would be determined, inter alia, by the age distributions of, the labor force and the unemployed, the duration and level of unemployment, and the social attitude to, and support for, those affected; Intersexual differences in mortality may be related to participation rates in the labor force as an in-

in Males 4-s54 Years of Age, 1921-1975

dicator of direct economic stress. Indirect stress effects on families would be offset by any community support, e.g., unemployment benefits. In order to examine this hypothesis, the mortality data in Figures 3 and 4 were investigated for evidence of wave forms. The findings are shown in Table 3. The near perfect values for the coefficients indicate that the path of the epidemic for these age groups is parabolic from 1930 to 1975. This appears consistent with the hypothesis that IHD mortality is describing a typical epidemic parabolic path, triggered in part by the economic stress of the Great Depression and moving over time through the population. The 1962-1967 course of the epidemic is consistent with a smaller second wave consequent upon the 1961 recession. These findings are portrayed in Figures 5 and 6. The extremely close fit of the yearly mortality data about the parabolic trend lines offers the facility for forecasting under the (b) (c) and (e) assumptions listed above. By inspection of the amplitude and period of the two major waves, together with evaluation of the unemployment rate and duration around the start of each wave, and the unemployment rate and expected duration of the current recession, a third forecast wave is constructed. This recession is estimated to be associated with a mortality wave approximately thrice the amplitude and duration of the 1962 wave.

TABLE 3-Waveform Curve-Fitting Age Group

Male 55-64 years Male 55-64 years Male 45-54 years Male 4554 years

Period

1930/61-1969/75

1962-1968 1930/1961-1969/75 1962/1968

Number of years

Correlation Coefficients

46 7 46 7

0.993 0.924 0.991 0.888

AJPH August 1979, Vol. 69, No. 8

IHD MORTALITY AND THE BUSINESS CYCLE

0L 500

400

,7/1931 Wave

300

200

100 1930

1935

1945

1940

1950

19S5

19W0

1965

1970

1975

1980

1985

1990

1995

xx)X YEARLY

FIGURE 5-Ischaemic Heart Disease Mortality in Males 55-64 Years of Age, 1930-1995 Waveforms

On this basis, the " 1976 wave" for 55 to 64-year-old males is estimated to have a mortality amplitude of 350/100,000 deaths and a period of 21 years.*** Similarly the 45 to 54year-old group amplitude is estimated to be in the order of 120/100,000 deaths with a 21-year period. Under assumption (e) above, the 1976 wave is combined with the forecast paths of the 1931 and 1962 waves. Thus the incidence of IHD can be forecast toward the year 2000. ***Under the 7th revision of the ICD.

Discussion The correlation findings support those obtained by Brenner in the United States.22 There is no precedent for the path analysis findings. In particular, the use of a prescribing series as an indicator series of national stress appears to be the first such attempt. However, the use of indicator series and path analysis is well documented in other disciplines.'3' 17 The link between unemployment and prescribing patterns gives further shape to the relation between

400 1976 Wave

350

300

250

,200

a

1931 Wave

1930

1935

1940

1945

1950

1955

1965 XXX YEARLY

1960

1970

1975

1980

1985

1990

1995

FIGURE 6-Ischaemic Heart Disease Mortality in Males 45-54 Years of Age, 1930-1995 Waveforms AJPH August 1979, Vol. 69, No. 8

777

BUNN

the business cycle and IHD mortality. Other writers have described the Australian trends in IHD, including the renewed acceleration in the early 1960s and the post-1968 decline.' However, the economic implications of the coincidental 1930s and 1960s accelerations in mortality have not been explored. The wave hypothesis model explains the recent decline in IHD mortality by viewing IHD partly as a point source epidemic23 with its roots in the Great Depression and 1961 recession. In the absence of other adequate explanations for the decline, ways of testing this hypothesis ought to be pursued. Ideally, a prospective study is needed of individuals from the labor force cohorts who experienced the Great Depression and/or the 1961 recession or the 1975 recession. Such a study would seek a higher incidence of IHD among those individuals subject to unemployment or other economic loss, while controlling for other risk factors. It is pertinent that the recent Framingham results link IHD with Type A behavior and "lend support to the notion that the American workplace may be somewhat responsible for the full development of the Type A behavior.' '24 This observation may be viewed against Kasl and Cobb's prospective study into unemployment which "points to an almost permanently damaging effect of the job loss experience."25 In the absence of long-term individual studies, however, there are at least two other ways of testing the wave hypothesis. Under assumption (c) above, an international survey, relating the severity of Great Depression unemployment to later IHD incidence might be expected to disclose a rank association of countries. Within a country, e.g., Australia, statistical forecasting of IHD mortality could be attempted as outlined above. There is no precedent for this and the results, although plausible on the evidence, should be received with caution. The different structural features of the present recession need to be kept in mind, i.e., the progressive industrialization since the Great Depression, the now significant migrant participation in the labor force, the unemployment benefits now available, and the present high youth unemployment. The possible effect of migration is of special interest (see Appendix V). The forecasts obtained predict a turning point in IHD mortality in the 1976-1978 period. Several factors may impinge upon this. First, the correlation analysis indicates a current three-year lag. On this basis, IHD mortality would expectedly increase in 1978. However, in previous recession and depression the lag period shortened. Opposing this is the improved medical care available during this current recession. It has been calculated that the post-myocardial infarction use of beta-blocking drugs could reduce mortality by a figure corresponding to five per cent of the current Australian rate.26 Second, the current recession continues and it will not be possible to fully compare the 1961 and 1975 recessions until the latter resolves. Third, there may be an element of negative skewness in the estimated waveforms, as discussed by Boulding.27 Fourth, there are inevitable measurement errors in the series used and limitations to the statistical models. With these factors in mind it is nevertheless possible to 778

forecast a clear discontinuity in the IHD series in the 19761978 period. It may then be possible to further establish the significance of economic factors in IHD.

REFERENCES 1. Christie, D: Mortality from Cardiovascular Disease. Med. J. Aust. 1:393, 1974. 2. Woodruff, P: Changing Mortality-Triumph and Challenge in Preventive Medicine. Med. J. Aust. 2:397, 1963. 3. Lancaster, HO: The Causes of the Declines in the Death Rates in Australia. Med J Aust 2:939, 1967. 4. Australian Academy of Science: Diet and Coronary Heart Disease. Report No. 18, 1975. 5. Reader, R: Incidence and Prevalence of Ischaemic Heart Disease in Australia. Med. J. Aust. Special Supp. 2:5, 1972. 6. Beese, DH: Tobacco Consumption in various countries, Tobacco Research Council, London, 3rd edition, 1972. 7. Egger, GJ: The Economics of Smoking in Australia. Health Commission of New South Wales, 1974, p.8. 8. Gray, NJ and Hill, DJ: Patterns of Tobacco Smoking in Australia 2. Med. J. Aust. 2:327-328, 1977. 9. Australian Bureau of Statistics: Alcohol and Tobacco Consumption Patterns. February 1977, p.9. 10. Editorial: Brit. Med. J. 1:58, 1976. 11. Editorial: Brit. Med. J. 2:537, 1977. 12. Bunn, AR: Drane, NT: Economic Change as a Factor in Heart Disease. New Doctor 5:53-55, 1977. 13. Beck, MT, Bush, MG, Hayes, RW: The Indicator Approach to the Identification of Business Cycles. Reserve Bank of Australia Occasional Paper No. 2, 1973. 14. Armstrong, BK, Mann, JI, Adelstein, AM, et al: Commodity Consumption and Ischaemic Heart Disease Mortality, With Special Reference to Dietary Practices. J. Chron. Dis. 28:455469, 1975. 15. Brenner, MH-: Estimating the Social Costs of National Economic Policy: Implications for Mental and Physical Health and Criminal Aggression. Joint Economic Committee, U.S. Govt. Printing Office, Paper No. 5, 1976. 16. Eyer, J: Prosperity as a Cause of Death. Int. J. Health Serv. 1:135-136, 1977. 17. Kerlinger, FN, Pedhauzur, EJ: Multiple Regression in the Behavioural Sciences. New York: Holt Rinehart and Winston, 1973. 18. Pharmacy Guild of Australia: The Pharmacy Guild Digest, 1977, p.21. 19. Intercontinental Medical Statistics: Australian Morbidity Index, April-June, 1977, Vol. 1. 20. Andrews, G, Tennant, C: Being Upset and Becoming Ill: An Appraisal of the Relation between Life Events and Physical Illness. Med. J. Aust. 1:324-327, 1978. 21. Selye, H: Stress without Distress. London: Hodder and Stoughton, 1974, pp. 38-39. 22. Brenner, MH: Economic Change and Heart Disease Mortality. Am. J. Public Health, 61:606-611, 1971. 23. Austin, DF, Werner, SB: Epidemiology for the Health Sciences. Springfield: Thomas, 1974, p. 60. 24. Haynes, SG, Feinleib, M, Levine, S, et al: The Relationship of Psychosocial Factors to Coronary Heart Disease in the Framingham Study. Am. J. Epidemiol 107:399-400, 1978. 25. Kasl, SV, Cobb, S: Blood Pressure Changes in Men Undergoing Job Loss: A Preliminary Report. Psychosom Med. 32:36, 1970. 26. O'Rourke, M: The Role of Long-Term Beta-Blockade After Myocardial Infarction. Aust. Fam. Phys. Spec. Issue, 1978, p. 33. 27. Boulding, KE: Conflict and Defence-A General Theory. New York: Harper, 1963, pp. 125-127. 28. Yeomans, KA: Statistics for the Social Scientist: Harmondsworth: Penguin, 1973, p. 233. 29. Bass, FM, Clarke, DG: Testing Distributed Lag Models of Advertising Effect. J. Marketing Res. 9:298-308, 1972.

AJPH August 1979, Vol. 69, No. 8

IHD MORTALITY AND THE BUSINESS CYCLE 30. Goldsmith, SB: The Status of Health Status Indicators. Health Serv. Reports 87:212-220, 1972. 31. Anderson, NA and Bridges-Webb, C: Prescribing Med. J. Aust. Spec. Suppl. 2:23-25, 1976. 32. Borrie, WD: Population and Australia Parlt. Paper #6, 1975. 33. Stenhouse, NS, and McCall: Differential Mortality from Cardiovascular Disease in Migrants from England and Wales, Scotland, Italy and native-born Australians, J. Chron. Dis. 23:423, 1970. 34. Commonwealth Employment Service Advisory Committee: Parliamentary Paper #265, 1973, p.40. 35. Haywood, E: Relations between C.E.S. registered unemployed and filled vacancies and other statistical series, in Commonwealth Employment Advisory Committee: Parliamentary Paper #265. 1973, p.107.

ACKNOWLEDGMENTS This article owes a great deal to Professor N. Drane and Dr. J. Brennan. Sandoz Australia Pty Ltd provided calculator facilities and encouragement.

APPENDIX I STATISTICAL MODELS (a) Correlation Models A standard time series model was used, i.e. y =T xR where y = data series of unemployment or mortality T = Secular trend R = Residual component As in any credible time series correlation model28 evidence of correlation was initially sought between the residual components. A multiplicative model was used in view of the series' length. Mathematical de-trending was used for each series lest the use of moving averages inject artificial oscillatory components into the residual series. The curves used for detrending are shown in Table A-l, below: TABLE A-1 -Secular Trend Removal Curves Series

Detrending Curves

35-44 year l.H.D. male 35-44 year I.H.D. female 45-54 year I.H.D. male 45-54 year I.H.D. female 55-64 year I.H.D. male 55-64 year l.H.D. female

logistic no trend logistic

logistic logistic logistic linear logistic

Unemployment-annual rate Prescriptions-General Benefit

The residual series obtained were transformed into standard units. There was no evidence of a cyclical component. (b) Path Models These are special regression models for theory testing. 17 The model used is shown below.

P.B.S. prescribing in year T-m

Where T = a base year of mortality n = 0-5 years n-m = 18 months* p2 1, p32, p31 = path coefficients = beta coefficients

(c) Unemployment Series as a Business Cycle Indicator No single data source on unemployment exists for the 1921-1975 period. The composition of the unemployment series used is shown below: Period Source 1906-1945 Trades Union data 1946-1963 Commonwealth Employment Service Statistics 1964-1975 Commonwealth Bureau of Census and Statistics Quarterly Survey In common with the United States, there are no age/sex group data prior to 1946. Some age/sex specific data are available after 1964 in Australia. This is not necessarily a deficiency as age/sex specific unemployment rates have been shown to be highly correlated with average unemployment for persons.'5 The series of annual unemployment levels listed below gives the best single indication of unemployment in Australia from 1921 to 1975.34 It is -also the best single indicator series of business cycle .turning points over the period. '3 The series correlates, in phase with real gross Domestic Product.35 Average Annual Unemployment levels in Australia 1921-1975 Year % Rate 1921 9.6 1922 11.2 1923 7.1 1924 8.9 1925 8.8 1926 7.1 1927 7.0 1928 10.8 1929 11.1 1930 19.3 1931 27.4 1932 29.0 1933 25.1 1934 20.5 1935 16.5 1936 12.2 1937 9.3 1938 8.7 1939 9.7 1940 8.0 1941 3.7 1942 1.6 1943 1.1 1944 1.2 1945 1.2 1946 1.4 1947 1.2 1948 0.9 1949

Unemployment in year T-n p.21 1,

p.31 p.32

AJPH August 1979, Vol. 69, No. 8

Mortality in Year T

0.5

*Unemployment and mortality are measured in calendar years. Prescribing is measured in fiscal years, hence a 6-month phase difference is present. 779

BUNN 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975

0.3 0.3 1.2 1.4 0.6 0.5 0.9 1.3 1.6 1.6 1.2 2.3 2.2 1.8 1.4 1.3 1.5 1.6 1.5 1.3 1.5 1.6 2.1 1.9 2.3 4.3

Australian unemployment rates have been lower than the USA rate. This reflects the social sensitivity to unemployment in Australia. For example the pronounced 1961 recession, where the unemployment rate reached a post-war peak, coincided with a federal election. The 2.3 per cent unemployment rate resulted in the near defeat of the sitting Government.

APPENDIX 11 THE USE OF LAGGED MODELS Lagged time series regression models are most developed as econometric marketing models.29 Recently, Brenner has applied a distributed lag model to morbidity and economic series.'5 Brenner's conclusion that morbidity is related to economic downturns has been challenged. Eyer has put the view that mortality is related to economic peaks and has questioned the existence of lags.'6 Under this view an inverse association would be expected if the IHD series and unemployment series in this study were correlated at a zero lag. This would indicate that low unemployment at a boom period is accompanied by peak mortality reflecting in Eyer's view, migration and overwork factors. The three male IHD post-war series in Table 1 were analyzed to explore this view. There was no evidence of inverse association at zero lag, the correlation values, 0.055, 0.190 and 0.226, rather showed an absence of association. The business cycle is not a true cyclical series. Rather it describes an oscillatory series as does the residual mortality series. When correlating such series in this study, there was typically only one lag which resulted in a significant association. The lag structure used in the present research varies within the 0-5 year range by age group, sex, and economic 780

period. This may arise from variable exposure to risk factors of varying intensity at different economic periods by the age and sex groups examined.

APPENDIX III BRIDGING CALCULATIONS Australia and the USA began using the 8th revision of the ICD in 1968. The Australian Bureau of Statistics (ABS) has computed sex-specific comparability ratios for IHD by recording 1967-1968 IHD mortality under both the 7th and 8th Revisions. It estimates the male ratio as 0.94 and the female ratio as 0.90, to apply when converting 8th revision mortality to a 7th revision base. These correspond to the USA comparability ratio of 0.87 for persons. Earlier periods bridging calculations were determined after discussions with the ABS and Australian epidemiologists. Details are available from the author.

APPENDIX IV THE USE OF PRESCRIBING SERIES AS AN INDICATOR OF NATIONAL STRESS OR MORBIDITY The development of "'activity counts" as health status indicators is well documented by Goldsmith30 who cites seven criteria for such indicators. First, the purpose of the indicator. In this study the purpose is to quantify variations in prescribing as an indicator series of the changing demand for medical care and hence changes in population health status. This indicator series expands the unemployment/mortality model, to test for any temporal change in health status between a change in unemployment and a later change in mortality. If unemployment is associated with both changes in the health status indicator and in mortality, then the practical significance of the Table 1 correlation findings is increased. Second, the data composition (described in the text). Third, the data must be available with minimal modification. In this case the prescription data were readily obtainable in suitable form. Fourth, the indicator computation must be clear. The approach used is standard and is described in Appendix I. Fifth, the indicator components must be identifiable and distinguishable. There are two components in the PBS series, i.e., the General Benefit and Pensioner Benefit components. No more detailed age or sex specific tabulations are available. The chief interest in this study is with middle years, labor force age groups. Hence, the General Benefits component was selected and pensioners excluded. Therapeutic drug class data are available but difficulties in relating IHD to drugs prescribed from different therapeutic classes precluded their use. Sixth, the data must be reliable and valid. The PBS data AJPH August 1979, Vol. 69, No. 8

IHD MORTALITY AND THE BUSINESS CYCLE

are census data and are accepted as reliable. Their validity as a health status indicator is perhaps the main question. There is evidence that the rate of prescribing per disease contact is constant.3' The close link between consultations and prescriptions is mentioned in the text. Therefore prescribing appears to be a valid indicator of disease contacts and consultations. It could be considered that other factors-for example, changes in health insurance or secular changes in population distribution-might interfere with the use of the prescription indicator series. This series extends from 1953 to 1975 and deliberately coincides with the period of voluntary health insurance in Australia. This controls for health insurance system changes. This period was also chosen for the continuous operation of the Pharmaceutical Benefits Scheme, a Federal Government-financed program subsidizing a wide range of pharmaceuticals. Residual path models were constructed to test for yearby-year association while removing the influence of longterm factors, such as changes in population age distribution. There were statistically significant mid-range correlations in the age and sex groups in Table 2, between short-term business cycle and prescribing fluctuations. It could also be speculated that unemployed persons have more time to visit doctors for conditions which might not otherwise have stopped them from working. This does not appear to have occurred to any extent as the lag between unemployment and prescribing is 18 months while the mean duration of unemployment, even in the current recession, is of 6 to 7 months. Seventh, there must be a mechanism to correlate the health status indicator with other indicators. As discussed above, there are correlation between prescribing and consultations, disease contacts, and unemployment. As measured against Goldsmith's seven criteria,30 the general benefits prescription series appears to be a valid

I

health status indicator series, for comparing national health status with business cycle changes.

APPENDIX V MIGRATION AND IHD Thirty five per cent of the Australian population growth since 1788 has been by migration.32 This occurs in waves, following the business cycle. Ninety eight per cent of migration is from Europe.32 IHD mortality in migrants is lower than in the Australian population but rises over the 20 years after arrival, toward the population incidence.33 In the absence of any cohort studies investigating migration and IHD, it is important to consider what effect, if any, migration might have on IHD cohort analysis. Given an initially lower migrant IHD mortality, and the association between the business and migration cycles, it might be expected that some cohort IHD mortality might fall following a migration wave. The 1920s saw sustained high migration, yet the 1930s saw accelerating IHD mortality in each cohort. The next post-war migration wave occurred as IHD mortality was increasing towards the epidemic incidence of the 1960s. This seems to preclude any migration wave impact on the cohort analysis. However, this does not exclude smaller underlying effects from long term immigration. Interestingly, past migration in some respects tended to offset age groups depleted by war losses and periods of lowered birthrate. Immigration has been a common element in every Australian cohort for the past 200 years. Multiple decrement analysis of migrant and locally born cohorts may eventually clarify any effect. Until such data are available, the chronic development of IHD and the tendency for migrant IHD mortality to increase toward the population incidence through time appears to support the established use of cohort analysis in Australian studies of heart disease.

Conference on Community Health Promotion and the Hospital

I

The Montefiore Hospital and Medical Center, Bronx, NY, will sponsor an all-day conference on "Community Health Promotion and the Hospital" on Saturday, November 3, 1979, from 9:30 am-4:30 pm, just prior to the opening of the Annual Meeting of the American Public Health Association in New York City. The conference, to be held at Montefiore Hospital, will bring together representatives from hospitals and other health facilities and from community groups to address the role of the hospitals and of the community in promoting health. The afternoon session will be devoted to workshops and discussion groups. There is no registration fee for the conference, which is limited to 150 participants on a first-come/ first-served basis. The deadline for registration is October 1, 1979. For further information, contact Sally Kohn, Director, Community Health Participation Program, Montefiore Hospital and Medical Center, 111 East 210th Street, Bronx, NY 10467. Tel: 212/920-4184.

AJPH August 1979, Vol. 69, No. 8

781

Ischaemic heart disease mortality and the business cycle in Australia.

lschaemic Heart Disease Mortality and The Business Cycle in Australia ALFRED REX BUNN, DSR, BCOM, MEc Abstract: Trends in Australian heart disease mo...
1MB Sizes 0 Downloads 0 Views