Abstract
Background
Epidemiological transition theory is based on a succession of specific “patterns” of causes of death in human societies. However, the reality and consistency of patterns of causes of death in a population at a given moment has never been formally and statistically evaluated.
Methods
Correlation analyses and principal component analysis were used to explore the correlation between age and sex cause‐specific death rates and to identify consistent patterns of mortality in France for two periods: 1968–79 and 1988–99.
Results
Cause‐specific death rates in France from 1988 to 1999 were found to be strongly and consistently correlated across space and time. The analysis outlines four specific patterns: mortality of 45–84‐year olds, mostly by neoplasms, cardiovascular and digestive diseases; mortality of the oldest old (>84 years); mortality of 25–64‐year‐old men, notably by HIV infection; and mortality by injury and poisoning of 15–44‐year olds. These patterns, which cover 96% of the total mortality during the period, differ from those for the period 1968–79 when respiratory diseases and conditions affecting children aged <1 year shaped mortality. They also differ substantially from those predicted by classical epidemiological transition theory.
Conclusion
This study provides evidence for an evolutionary structure of patterns of mortality in contemporary France and therefore suggests using the concept of epidemiological transition in a less simplistic way than is commonly the case. It also shows much stronger interrelationships between diseases leading to death than is usually believed and suggests that current categorisations of cause‐specific mortality in populations need reconsideration.
Approaching death remains the largest demographic driver of healthcare utilisation and costs.1 A comprehensive assessment of patterns of cause‐specific mortality and their trends is therefore essential for public health planning while most countries are undergoing profound demographic and epidemiological transformations.2,3 In 1971, Omran formulated the theory of “epidemiological transition” to explain worldwide changes in time and causes of human death: while developing and modernising, human societies experience a “shift in mortality and disease patterns”, transiting from a first “age of pestilence and famine” in which mortality is high and fluctuating owing to epidemics of infectious diseases and food crises, to a second “age of receding pandemics” and attains a third “age of degenerative and man‐made diseases” in which chronic diseases (cardiovascular diseases, diabetes, cancer, etc) and accidents predominate as causes of death.4 This theory, which was initially based on informal comparisons over time of the weight of broad causes of death (rank order, contribution to all‐cause mortality) in six countries (Ceylon, Chile, England, Japan, Sweden and Wales), was further revised by its author and his followers, notably to account for the epidemiological changes of the 70s–80s. In particular, a fourth “age of delayed degenerative diseases” and a fifth “age of emergent and re‐emergent infections” were added.5,6,7,8,9 After its endorsement by the World Heath Organization (WHO) and by the US National Research Council, the epidemiological transition theory became increasingly popular to a point where it shapes many scientists' and public health professionals' conceptions of changes in causes of death with time. However, criticisms of the simplistic determinism (modernisation) and oversimplicity of the theory, which does not fit with observations in many countries, accumulated during the 90s (summarised in Carolina and Gustavo10 and Caselli etal11). Only one formal analysis of epidemiological transition in the long term, using detailed mortality datasets and multivariate descriptive methods, has been reported.12 We used a similar approach to evaluate the reality of a short‐term structure of causes of death—that is, whether, in a population at a given time, causes of death consistently assemble to form a specific “pattern” (or a set of few patterns). This fundamental assumption of the epidemiological transition theory has never been formally and statistically investigated. We used correlation analyses and principal component analysis (PCA) of age and sex cause‐specific death rates to explore their correlation and to identify patterns of mortality in France for two periods: 1968–79 and 1988–99.
Methods
Death certification
The health system in France follows the WHO guidelines on death certification. Death certification (compulsory with 100% coverage) uses the international medical certificate, which allows the underlying and, since 1968, two contributory causes of death and an immediate cause of death to be recorded. Causes are further coded according to the International Classification of Diseases (ICD): between 1968 and 78, the eighth revision was used13 and, between 1979 and 99, the ninth revision.14 These classifications are sufficiently similar (identical structure and headings and few different diseases) to allow comparisons of data coded with the two systems by broad categories.14
Cause‐specific death rates
Cause‐specific death rates were computed for
the 17 headings (groups of diseases) of the ICD or the 46 diseases on the European shortlist of causes of death,15
those diseases either recorded as an underlying cause or as “any” cause (immediate, underlying or contributory cause),
sex and age category (<1, 1–14, 15–24, 25–44, 45–64, 65–84 and >84 years),
each week of the two 12‐year study periods, 1968–79 and 1988–99 (time units, each n = 626), and for each district of France (spatial units, n = 348) for the period 1988–99. The two study periods (1968–79 and 1988–99) were chosen long enough to provide sufficient time units (two or three times the number of rates, see later) and as distant as possible to observe differences between them.
Denominators of the rates (populations at risk of death) were obtained for each week (time unit) and district (spatial unit) from the Institut National de la Statistique et des Études, the French national demographic institute, which continuously updates population data in France.
Statistical analysis
Changes in age and sex cause‐specific rates of death between 1968–79 and 1988–99 were assessed not only from the differences in rates between the two periods but also from the differences in ranks—this allows the general decrease in mortality over the whole study period to be taken into account. Correlation analyses and PCA were used to show patterns of death consistent across time (weeks) and space (districts). Firstly, correlation matrices (Pearson's coefficients) of sex and age cause‐specific mortality were constructed for time units (for both the 1968–79 and the 1988–99 periods) and space units (for the period 1988–99). These matrices were further studied by PCA, an exploratory factor analysis method that allows identification of unobservable latent factors that underlie or structure a set of variables (here, sex and age cause‐specific mortality). PCA includes three successive steps16:
Condensation of the data by the computation of principal components. Principal components are linear functions of observed correlated variables (death rates) generated in a sequential manner. The first component is the linear function of the original variables, the variance of which is the maximum over all possible linear functions. Each following component is the linear combination uncorrelated with the preceding principal components and having the largest possible variance. Principal components are usually computed using correlation matrices (or standardised variables) to avoid unit scale effects and ensure that variables are on an equal footing. In this case, the coefficients defining these linear functions, referred to as the loadings, are interpretable as correlation coefficients between principal components and original variables and are used to identify the nature of the principal components (loadings>0.40 usually indicate substantial correlation).
Extraction of principal components or selection of the number of components to retain. These components correspond to the underlying latent variables or factors. The number of factors to extract was determined using a previously described methodology.17
Rotation of retained components to facilitate the interpretation. Usually, many of the principal components computed are difficult to interpret and a second linear transformation of the principal components is therefore useful. This transformation is referred to as a rotation. Rotations are often carried out to obtain a simple structure—that is, to obtain, as with the varimax method used in this study, rotated components with variables that have loadings near either 0 or 1 (in which case the given variable participates in the definition of a single factor).
To avoid uninformative statistical fluctuations and limit the number of variables considered in the correlation matrices and PCA (<150), low death rates (<10 per 100 000) were grouped together, for each sex and age category, in a residual cause‐of‐death category. The primary analysis considered rates formed using the underlying cause categorised according to the 17 ICD headings. Secondary analyses were carried out using rates formed “any” causes (underlying, immediate or contributory) and those using a more detailed list of 46 diseases for the underlying cause of death to evaluate the sensitivity of results to classification rules.
Results
Over 13 million deaths (about 530 000/year) and their causes (underlying, immediate or contributory) were considered in the analysis. Table 1 shows the top 50 list of age and sex cause‐specific death rates (underlying cause, coded using ICD headings) for the period 1988–99 and changes relative to the period 1968–79. Several rates increased massively between 1968–79 and 1988–99 (diseases of the nervous system, mental disorders, diseases of the skin, disease of the musculoskeletal system for 65–84‐year olds and especially >84‐year olds), whereas many others decreased sharply or even disappeared from the top 50 (notably those about children <1 year). Ill‐defined conditions remained stable overall (about 15% in both periods).
Table 1 Top 50 annual age and sex cause‐specific death rates in France for the period 1988–99, classified according to the 17 headings of the International Classification of Diseases, and changes since the period 1968–79.
| ICD heading of underlying cause of death for age and sex group | Rank 1988–99 | Change in rank relative to 1968–79*† | Rate per 100 000 (1988–99) | Change in rate (%) relative to 1968–79* |
|---|---|---|---|---|
| Diseases of the circulatory system, male, >84 years | 1 | 0 | 6478.18 | −29 |
| Diseases of the circulatory system, female, >84 years | 2 | 0 | 5949.67 | −23 |
| Neoplasms, male, >84 years | 3 | 2 | 3028.39 | 26 |
| Diseases of the respiratory system, male, >84 years | 4 | 3 | 2158.65 | −5 |
| Neoplasms, female, >84 years | 5 | 5 | 1524.97 | 4 |
| Neoplasms, male, 65–84 years | 6 | 5 | 1503.01 | 3 |
| Diseases of the circulatory system, male, 65–84 years | 7 | −1 | 1413.66 | −41 |
| Symptoms, signs, and ill‐defined conditions, female, >84 years | 8 | −4 | 1320.30 | −54 |
| Diseases of the respiratory system, female, >84 years | 9 | 0 | 1315.20 | −14 |
| Symptoms, signs, and ill‐defined conditions, male, >84 years | 10 | −7 | 1218.82 | −62 |
| Diseases of the circulatory system, female, 65–84 years | 11 | −3 | 912.32 | −47 |
| Injury and poisoning, male, >84 years | 12 | 1 | 899.31 | −11 |
| Injury and poisoning, female, >84 years | 13 | −1 | 794.07 | −33 |
| Neoplasms, female, 65–84 years | 14 | 0 | 669.14 | −8 |
| Diseases of the digestive system, male, >84 years | 15 | 1 | 656.33 | 8 |
| Diseases of the digestive system, female, >84 years | 16 | 2 | 590.02 | 21 |
| Mental disorders, female, >84 years | 17 | 26 | 555.94 | 293 |
| Endocrine, nutritional, metabolic diseases, and immunity disorders, female, >84 years | 18 | 1 | 529.34 | 10 |
| Endocrine, nutritional, metabolic diseases, and immunity disorders, male, >84 years | 19 | 2 | 490.22 | 9 |
| Mental disorders, male, >84 years | 20 | 28 | 440.95 | 256 |
| Diseases of the urogenital system, male, >84 years | 21 | −6 | 440.29 | −37 |
| Neoplasms, male, 45–64 years | 22 | 0 | 428.20 | 1 |
| Diseases of the respiratory system, male, 65–84 years | 23 | −3 | 348.48 | 0 |
| Diseases of the nervous system and sense organs, male, >84 years | 24 | 3 | 321.43 | −3 |
| Diseases of the nervous system and sense organs, female, >84 years | 25 | 3 | 264.77 | −12 |
| Infectious and parasitic diseases, male, >84 years | 26 | 11 | 251.40 | 34 |
| Diseases of the urogenital system, female, >84 years | 27 | 9 | 231.12 | 21 |
| Symptoms, signs and ill‐defined conditions, male, <1 year | 28 | 5 | 223.89 | −8 |
| Injury and poisoning, male, 65–84 years | 29 | 2 | 218.42 | −17 |
| Certain conditions originating in the perinatal period, male, <1 year | 30 | −13 | 208.47 | −63 |
| Diseases of the digestive system, male, 65–84 years | 31 | −6 | 201.02 | −44 |
| Infectious and parasitic diseases, female, >84 years | 32 | 14 | 192.38 | 46 |
| Neoplasms, female, 45–64 years | 33 | 2 | 191.89 | −14 |
| Diseases of the circulatory system, male, 45–64 years | 34 | −8 | 190.47 | −45 |
| Certain conditions originating in the perinatal period, female, <1 year | 35 | −12 | 158.47 | −62 |
| Symptoms, signs and ill‐defined conditions, male, 65–84 years | 36 | −12 | 156.63 | −60 |
| Diseases of the respiratory system, female, 65–84 years | 37 | −3 | 155.59 | −34 |
| Symptoms, signs and ill‐defined conditions, female, <1 year | 38 | 1 | 152.17 | −16 |
| Congenital anomalies, male, <1 year | 39 | −9 | 139.13 | −53 |
| Injury and poisoning, female, 65–84 years | 40 | −2 | 128.90 | −31 |
| Diseases of the skin and subcutaneous tissue, male, >84 years | 41 | 18 | 128.63 | 72 |
| Congenital anomalies, female, <1 year | 42 | −10 | 119.13 | −52 |
| Diseases of the digestive system, female, 65–84 years | 43 | –3 | 117.51 | −33 |
| Diseases of the blood and blood‐forming organs, male, >84 years | 44 | 19 | 112.47 | 80 |
| Symptoms, signs, and ill‐defined conditions, female, 65–84 years | 45 | −16 | 111.85 | −62 |
| Injury and poisoning, male, 45–64 years | 46 | −4 | 100.87 | −30 |
| Diseases of the nervous system and sense organs, male, 65–84 years | 47 | 3 | 99.37 | −14 |
| Diseases of the musculoskeletal system and connective tissue, female, >84 years | 48 | 20 | 92.78 | 97 |
| Diseases of the skin and subcutaneous tissue, female, >84 years | 49 | 23 | 92.05 | 102 |
| Injury and poisoning, male, 25–44 years | 50 | 3 | 87.83 | −14 |
ICD, International Classification of Diseases.
*Rate for the period 1988–99 minus the rate for the period 1968–79 divided by the rate for the period 1968–79.
†Rank for the period 1988–99 minus the rank for the period 1968–79.
The correlation between death rates was high (Pearson's r>0.40), both for time and spatial units (eg, table 2 shows extracts of the correlation matrix of 1988−99 rates for time units, upper right corner, and spatial units, lower left corner). Noticeably, the highest levels of correlation included rates of males and females for a same disease and age category (eg, diseases of the circulatory system in 65−84‐year olds) and also both rates across age categories for individual diseases (eg, diseases of the circulatory system in 45−64‐year olds and 65−84‐year olds) and rates of different diseases (eg, diseases of the circulatory system, diseases of the respiratory system and diseases of the digestive system in 65−84‐year olds).
Table 2 Extract of the correlation matrix* of age and sex cause‐specific death rates.
| Neoplasms, male, 45–64 years | Neoplasms, male, 65–84 years | Neoplasms, female, 65–84 years | Diseases of the circulatory system, male, 45–64 years | Diseases of the circulatory system, male, 65–84 years | Diseases of the circulatory system, female, 45–64 years | Diseases of the circulatory system, female, 65–84 years | Diseases of the respiratory system, male, 45–64 years | Diseases of the digestive system, male, 45–64 years | Diseases of the digestive system, male, 65–84 years | Neoplasms, male, >84 years | Diseases of the circulatory system, male, >84 years | Diseases of the circulatory system, female, >84 years | Diseases of the respiratory system, male, >84 years | Diseases of the digestive system, male >84 years | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Neoplasms, male, 45–64 years | 0.66 | 0.64 | 0.79 | 0.74 | 0.72 | 0.80 | 0.40 | 0.66 | 0.63 | ‐0.01 | 0.27 | 0.37 | ‐0.01 | 0.20 | |
| Neoplasms, male, 65–84 years | 0.68 | 0.68 | 0.64 | 0.68 | 0.61 | 0.67 | 0.48 | 0.55 | 0.60 | 0.27 | 0.50 | 0.56 | 0.26 | 0.33 | |
| Neoplasms, female, 65–84 years | 0.38 | 0.61 | 0.61 | 0.65 | 0.57 | 0.65 | 0.44 | 0.51 | 0.52 | 0.19 | 0.42 | 0.50 | 0.21 | 0.24 | |
| Diseases of the circulatory system, male, 45–64 years | 0.73 | 0.38 | 0.21 | 0.85 | 0.78 | 0.89 | 0.55 | 0.68 | 0.71 | 0.03 | 0.43 | 0.54 | 0.15 | 0.27 | |
| Diseases of the circulatory system, male, 65–84 years | 0.59 | 0.52 | 0.32 | 0.73 | 0.80 | 0.95 | 0.69 | 0.68 | 0.77 | 0.14 | 0.67 | 0.78 | 0.42 | 0.37 | |
| Diseases of the circulatory system, female, 45–64 years | 0.60 | 0.38 | 0.24 | 0.72 | 0.65 | 0.82 | 0.52 | 0.63 | 0.65 | 0.05 | 0.42 | 0.53 | 0.16 | 0.26 | |
| Diseases of the circulatory system, female, 65–84 years | 0.57 | 0.51 | 0.35 | 0.69 | 0.86 | 0.65 | 0.66 | 0.71 | 0.78 | 0.06 | 0.57 | 0.70 | 0.32 | 0.33 | |
| Diseases of the respiratory system, male, 45–64 years | 0.67 | 0.42 | 0.20 | 0.62 | 0.49 | 0.55 | 0.49 | 0.43 | 0.54 | 0.23 | 0.65 | 0.72 | 0.60 | 0.35 | |
| Diseases of the digestive system, male, 45–64 years | 0.84 | 0.55 | 0.32 | 0.71 | 0.57 | 0.55 | 0.54 | 0.65 | 0.61 | 0.06 | 0.35 | 0.45 | 0.13 | 0.24 | |
| Diseases of the digestive system, male, 65–84 years | 0.61 | 0.57 | 0.37 | 0.50 | 0.55 | 0.39 | 0.53 | 0.51 | 0.61 | 0.15 | 0.54 | 0.63 | 0.34 | 0.34 | |
| Neoplasms, male, >84 years | 0.38 | 0.48 | 0.33 | 0.20 | 0.29 | 0.21 | 0.33 | 0.26 | 0.35 | 0.28 | 0.46 | 0.42 | 0.45 | 0.28 | |
| Diseases of the circulatory system, male, >84 years | 0.38 | 0.36 | 0.20 | 0.40 | 0.59 | 0.39 | 0.57 | 0.28 | 0.36 | 0.27 | 0.58 | 0.89 | 0.77 | 0.50 | |
| Diseases of the circulatory system, female, >84 years | 0.27 | 0.25 | 0.13 | 0.34 | 0.51 | 0.35 | 0.59 | 0.20 | 0.23 | 0.19 | 0.43 | 0.70 | 0.78 | 0.48 | |
| Diseases of the respiratory system, male, >84 years | 0.37 | 0.37 | 0.17 | 0.27 | 0.37 | 0.33 | 0.43 | 0.46 | 0.32 | 0.37 | 0.50 | 0.50 | 0.38 | 0.42 | |
| Diseases of the digestive system, male, >84 years | 0.28 | 0.15 | 0.13 | 0.26 | 0.25 | 0.21 | 0.29 | 0.20 | 0.23 | 0.35 | 0.30 | 0.35 | 0.31 | 0.35 |
Upper right corner, time units; lower left corner, spatial units.
*Pearson's correlation coefficient between selected rates in lines and columns are shown in cells.
Bold characters indicate high correlation (r>0.40).
These correlated rates assemble consistently in PCA to form four main factors or patterns of mortality (tables 3 and 4). For the period 1988–99, these patterns can be labelled “mortality between ages 45 and 84 years, mostly by neoplasms, cardiovascular and digestive diseases”; “mortality of the oldest old (>84 years old)”; “mortality by infectious diseases, neoplasms, mental and nervous system disorders, cardiovascular, respiratory, urogenital system and digestive disease”); “mortality of 25–64‐year‐old men” and “mortality by injury between ages 15 and 44 years”. These patterns explain 31–39% of the total variance of rates, but cover 96% of all mortality during the period (as most of the top‐ranked death rates are constituents of these four patterns).
Table 3 Principal component analysis results for age and sex cause‐specific death rates classified according to the 17 headings of the International Classification of Diseases 8–9: period 1988–99, time units*.
| Age and sex‐specific mortality | Factor 1 | Factor 2 | Factor 3 | Factor 4 |
|---|---|---|---|---|
| Infectious and parasitic diseases | ||||
| Male, <1 year | ||||
| Male, 25–44 years | 0.76 | |||
| Male, 45–64 years | 0.65 | |||
| Male, 65–84 years | 0.54 | |||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 65–84 years | 0.52 | |||
| Female, >84 years | ||||
| Neoplasms | ||||
| Male, 25–44 years | ||||
| Male, 45–64 years | 0.67 | –0.59 | ||
| Male, 65–84 years | 0.74 | |||
| Male, >84 years | 0.49 | |||
| Female, 25–44 years | ||||
| Female, 45–64 years | 0.54 | –0.42 | ||
| Female, 65–84 years | 0.68 | |||
| Female, >84 years | ||||
| Endocrine, nutritional and metabolic diseases and immunity disorders | ||||
| Male, 45–64 years | ||||
| Male, 65–84 years | 0.47 | 0.44 | ||
| Male, >84 years | 0.51 | |||
| Female, 65–84 years | 0.72 | |||
| Female, >84 years | 0.64 | |||
| Diseases of the blood and blood‐forming organs | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Mental disorders | ||||
| Male, 45–64 years | 0.43 | |||
| Male, 65–84 years | 0.73 | |||
| Male, >84 years | 0.64 | |||
| Female, 65–84 years | 0.72 | |||
| Female, >84 years | 0.80 | |||
| Diseases of the nervous system and sense organs | ||||
| Male, <1 year | ||||
| Male, 45–64 years | ||||
| Male, 65–84 years | 0.40 | 0.53 | ||
| Male, >84 years | 0.74 | |||
| Female, <1 year | ||||
| Female, 65–84 years | 0.68 | |||
| Female, >84 years | 0.83 | |||
| Diseases of the circulatory system | ||||
| Male, <1 year | ||||
| Male, 25–44 years | 0.43 | |||
| Male, 45–64 years | 0.78 | –0.45 | ||
| Male, 65–84 years | 0.91 | |||
| Male, >84 years | 0.78 | 0.46 | ||
| Female, 45–64 years | 0.74 | |||
| Female, 65–84 years | 0.89 | |||
| Female, >84 years | 0.88 | |||
| Diseases of the respiratory system | ||||
| Male, <1 year | 0.46 | |||
| Male, 45–64 years | 0.76 | |||
| Male, 65–84 years | 0.77 | 0.42 | ||
| Male, >84 years | 0.61 | 0.66 | ||
| Female, 45–64 years | 0.56 | |||
| Female, 65–84 years | 0.75 | 0.42 | ||
| Female, >84 years | 0.62 | 0.64 | ||
| Diseases of the digestive system | ||||
| Male, 25–44 years | ||||
| Male, 45–64 years | 0.66 | |||
| Male, 65–84 years | 0.77 | |||
| Male, >84 years | 0.46 | |||
| Female, 45–64 years | 0.52 | |||
| Female, 65–84 years | 0.72 | |||
| Female, >84 years | 0.61 | |||
| Diseases of the urogenital system | ||||
| Male, 65–84 years | 0.65 | |||
| Male, >84 years | 0.53 | |||
| Female, 65–84 years | 0.49 | |||
| Female, >84 years | 0.42 | |||
| Diseases of the skin and subcutaneous tissue | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Diseases of the musculoskeletal system and connective tissue | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | 0.51 | |||
| Female, >84 years | ||||
| Congenital anomalies | ||||
| Male, <1 year | ||||
| Female, <1 year | ||||
| Certain conditions originating in the perinatal period | ||||
| Male, <1 year | ||||
| Female, <1 year | ||||
| Symptoms, signs and ill‐defined conditions | ||||
| Male, <1 year | 0.65 | −0.50 | ||
| Male, 25–44 years | ||||
| Male, 45–64 years | ||||
| Male, 65–84 years | 0.74 | |||
| Male, >84 years | 0.68 | 0.41 | ||
| Female, <1 year | 0.62 | −0.47 | ||
| Female, 45–64 years | ||||
| Female, 65–84 years | 0.82 | |||
| Female, >84 years | 0.72 | 0.48 | ||
| Injury and poisoning | ||||
| Male, <1 year | ||||
| Male, 1–14 years | −0.53 | |||
| Male, 15–24 years | −0.65 | 0.47 | ||
| Male, 25–44 years | −0.58 | 0.44 | ||
| Male, 45–64 years | 0.40 | −0.49 | ||
| Male, 65–84 years | 0.60 | |||
| Male, >84 years | 0.47 | |||
| Female, <1 year | ||||
| Female, 15–24 years | −0.41 | 0.40 | ||
| Female, 25–44 years | ||||
| Female, 45–64 years | −0.41 | |||
| Female, 65–84 years | 0.74 | |||
| Female, >84 years | 0.76 | |||
| “Residual” | ||||
| Male, <1 year | ||||
| Male, 1–14 years | 0.52 | |||
| Male, 15–24 years | 0.40 | |||
| Male, 25–44 years | 0.42 | |||
| Male, 45–64 years | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 1–14 years | 0.44 | |||
| Female, 15–24 years | ||||
| Female, 25–44 years | 0.53 | |||
| Female, 45–64 years | 0.58 | |||
| Female, 65–84 years | ||||
| Female, >84 years |
*Only rates that are substantially correlated with factors (loading>0.40) and help interpret these factors are indicated.
Bold figures indicate a very high correlation between the rates and factors (loading>0.75).
Table 4 Principal component analysis results for age and sex cause‐specific death rates classified according to the 17 headings of the International Classification of Diseases 8–9: period 1988–99, spatial units*.
| Age and sex‐specific mortality | Factor 1 | Factor 2 | Factor 3 | Factor 4 |
|---|---|---|---|---|
| Infectious and parasitic diseases | ||||
| Male, <1 year | ||||
| Male, 25–44 years | −0.42 | 0.56 | ||
| Male, 45–64 years | 0.50 | |||
| Male, 65–84 year | ||||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Neoplasms | ||||
| Male, 25–44 years | 0.51 | |||
| Male, 45–64 years | 0.84 | |||
| Male, 65–84 year | 0.64 | |||
| Male, >84 years | 0.50 | 0.42 | ||
| Female, 25–44 years | ||||
| Female, 45–64 years | 0.60 | |||
| Female, 65–84 year | 0.41 | |||
| Female, >84 years | 0.46 | |||
| Endocrine, nutritional and metabolic diseases and immunity disorders | ||||
| Male, 45–64 years | 0.57 | |||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | 0.56 | |||
| Female, >84 years | ||||
| Diseases of the blood and blood‐forming organs | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Mental disorders | ||||
| Male, 45–64 years | 0.67 | |||
| Male, 65–84 years | 0.54 | |||
| Male, >84 years | ||||
| Female, 65–84 years | 0.46 | |||
| Female, >84 years | −0.43 | |||
| Diseases of the nervous system and sense organs | ||||
| Male, <1 year | ||||
| Male, 45–64 years | 0.47 | |||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 65–84 years | 0.41 | |||
| Female, >84 years | 0.45 | |||
| Diseases of the circulatory system | ||||
| Male, <1 year | 0.54 | |||
| Male, 25–44 years | 0.76 | |||
| Male, 45–64 years | 0.74 | |||
| Male, 65–84 years | 0.56 | |||
| Male, >84 years | ||||
| Female, 45–64 years | 0.67 | |||
| Female, 65–84 years | 0.75 | |||
| Female, >84 years | 0.47 | 0.43 | ||
| Diseases of the respiratory system | ||||
| Male, <1 year | ||||
| Male, 45–64 years | 0.71 | |||
| Male, 65–84 years | 0.67 | |||
| Male, >84 years | 0.57 | |||
| Female, 45–64 years | 0.54 | |||
| Female, 65–84 years | 0.56 | |||
| Female, >84 years | 0.49 | |||
| Diseases of the digestive system | ||||
| Male, 25–44 years | 0.61 | |||
| Male, 45–64 years | 0.82 | |||
| Male, 65–84 years | 0.72 | |||
| Male, >84 years | 0.42 | |||
| Female, 45–64 years | 0.71 | |||
| Female, 65–84 years | 0.64 | |||
| Female, >84 years | 0.41 | |||
| Diseases of the urogenital system | ||||
| Male, 65–84 years | 0.42 | |||
| Male, >84 years | 0.42 | |||
| Female, 65–84 years | 0.48 | |||
| Female, >84 years | 0.51 | |||
| Diseases of the skin and subcutaneous tissue | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | 0.42 | |||
| Female, >84 years | ||||
| Diseases of the musculoskeletal system and connective tissue | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Congenital anomalies | ||||
| Male, <1 year | ||||
| Female, <1 year | ||||
| Certain conditions originating in the perinatal period | ||||
| Male, <1 year | ||||
| Female, <1 year | ||||
| Symptoms, signs and ill‐defined conditions | ||||
| Male, <1 year | ||||
| Male, 25–44 years | 0.64 | |||
| Male, 45–64 years | 0.78 | |||
| Male, 65–84 years | 0.76 | |||
| Male, >84 years | 0.44 | 0.52 | ||
| Female, <1 year | ||||
| Female, 65–84 years | 0.70 | |||
| Female, >84 years | 0.42 | 0.43 | 0.55 | |
| Injury and poisoning | ||||
| Male, <1 year | ||||
| Male, 1–14 years | ||||
| Male, 15–24 years | 0.76 | |||
| Male, 25–44 years | 0.75 | |||
| Male, 45–64 years | 0.50 | −0.41 | 0.45 | |
| Male, 65–84 years | 0.47 | |||
| Male, >84 years | 0.41 | |||
| Female, <1 year | ||||
| Female, 15–24 years | 0.53 | |||
| Female, 25–44 years | 0.50 | |||
| Female, 45–64 years | ||||
| Female, 65–84 years | 0.42 | |||
| Female, >84 years | ||||
| “Residual” | ||||
| Male, <1 year | ||||
| Male, 1–14 years | ||||
| Male, 15–24 years | 0.42 | |||
| Male, 25–44 years | 0.59 | |||
| Male, 45–64 years | 0.48 | |||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 1–14 years | ||||
| Female, 15–24 years | ||||
| Female, 25–44 years | 0.57 | |||
| Female, 45–64 years | 0.51 | 0.50 | ||
| Female, 65–84 years | ||||
| Female, >84 years |
*Only rates that are substantially correlated with factors (loading>0.40), and help interpret these factors are indicated.
Bold figures indicate a very high correlation between the rates and factors (loading>0.75).
Four main patterns of mortality were also observed during the period 1968–79 (table 5), but they were not the same as those for the period 1988–99: “mortality between ages 45 and 84 years, mostly by respiratory and cardiovascular diseases”; “mortality of children aged <1 year”; “mortality of those aged >64 years, mostly by neoplasms, nervous diseases and mental disorders”; and “mortality by injury at all ages”. These patterns explain 40% of the variance of the rates, but they cover only 85% of all mortality during the period.
Table 5 Principal component analysis results for age and sex cause‐specific death rates classified according to the 17 headings of the International Classification of Diseases 8–9: period 1968–79, time units*.
| Age and sex‐specific mortality | Factor 1 | Factor 2 | Factor 3 | Factor 4 |
|---|---|---|---|---|
| Infectious and parasitic diseases | ||||
| Male, <1 year | ||||
| Male, 45–64 years | 0.76 | |||
| Male, 65–84 years | 0.61 | |||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 45–64 years | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Neoplasms | ||||
| Male, <1 year | ||||
| Male, 25–44 years | ||||
| Male, 45–64 years | ||||
| Male, 65–84 years | 0.70 | |||
| Male, >84 years | 0.46 | |||
| Female, 25–44 years | 0.40 | |||
| Female, 45–64 years | 0.50 | |||
| Female, 65–84 years | 0.49 | |||
| Female, >84 years | ||||
| Endocrine, nutritional and metabolic diseases, and immunity disorders | ||||
| Male, <1 year | ||||
| Male, 45–64 years | 0.49 | |||
| Male, 65–84 years | 0.42 | 0.50 | ||
| Male, >84 years | 0.41 | |||
| Female, <1 year | ||||
| Female, 45–64 years | 0.69 | |||
| Female, 65–84 years | 0.55 | 0.41 | ||
| Female, >84 years | 0.54 | |||
| Diseases of the blood and blood‐forming organs | ||||
| Male, <1 year | ||||
| Male, 65–84 years | 0.44 | |||
| Male, >84 years | ||||
| Female, 65–84 years | 0.43 | |||
| Female, >84 years | ||||
| Mental disorders | ||||
| Male, 45–64 years | 0.53 | |||
| Male, 65–84 years | 0.65 | |||
| Male, >84 years | ||||
| Female, 65–84 years | 0.44 | |||
| Female, >84 years | ||||
| Diseases of the nervous system and sense organs | ||||
| Male, <1 year | 0.44 | |||
| Male, 45–64 years | 0.59 | |||
| Male, 65–84 years | 0.61 | 0.44 | ||
| Male, >84 years | ||||
| Female, <1 year | 0.43 | |||
| Female, 45–64 years | 0.56 | |||
| Female, 65–84 years | 0.70 | 0.40 | ||
| Female, >84 years | 0.47 | |||
| Diseases of the circulatory system | ||||
| Male, <1 year | ||||
| Male, 25–44 years | 0.55 | |||
| Male, 45–64 years | 0.89 | |||
| Male, 65–84 years | 0.87 | |||
| Male, >84 years | 0.69 | 0.48 | ||
| Female, <1 year | ||||
| Male, 25–44 years | 0.46 | |||
| Female, 45–64 years | 0.88 | |||
| Female, 65–84 years | 0.86 | |||
| Female, >84 years | 0.63 | 0.56 | ||
| Diseases of the respiratory system | ||||
| Male, <1 year | 0.80 | |||
| Male, 45–64 years | 0.73 | |||
| Male, 65–84 years | 0.73 | 0.41 | ||
| Male, >84 years | 0.70 | 0.55 | ||
| Female, <1 year | 0.78 | |||
| Female, 45–64 years | 0.70 | |||
| Female, 65–84 years | 0.79 | |||
| Female, >84 years | 0.74 | 0.49 | ||
| Diseases of the digestive system | ||||
| Male, <1 year | ||||
| Male, 25–44 years | ||||
| Male, 45–64 years | 0.63 | |||
| Male, 65–84 years | 0.55 | |||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 25–44 years | ||||
| Female, 45–64 years | 0.44 | |||
| Female, 65–84 years | ||||
| Female, >84 years | 0.47 | |||
| Diseases of the urogenital system | ||||
| Male, 45–64 years | 0.60 | |||
| Male, 65–84 years | 0.50 | |||
| Male, >84 years | ||||
| Female, 65–84 years | 0.54 | |||
| Female, >84 years | 0.62 | |||
| Diseases of the skin and subcutaneous tissue | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Diseases of the musculoskeletal system and connective tissue | ||||
| Female, 65–84 years | ||||
| Female, >84 years | ||||
| Congenital anomalies | ||||
| Male, <1 year | ||||
| Female, <1 year | 0.43 | |||
| Certain conditions originating in the perinatal period | ||||
| Male, <1 year | 0.64 | ‐0.60 | ||
| Female, <1 year | 0.63 | ‐0.59 | ||
| Symptoms, signs and ill‐defined conditions | ||||
| Male, <1 year | 0.58 | |||
| Male, 25–44 years | ||||
| Male, 45–64 years | 0.83 | |||
| Male, 65–84 years | 0.93 | |||
| Male, >84 years | 0.89 | |||
| Female, <1 year | 0.53 | |||
| Female, 45–64 years | 0.79 | |||
| Female, 65–84 years | 0.93 | |||
| Female, >84 years | 0.91 | |||
| Injury and poisoning | ||||
| Male, <1 year | ||||
| Male, 1–14 years | −0.54 | |||
| Male, 15–24 years. | 0.63 | |||
| Male, 25–44 years | −0.57 | 0.54 | ||
| Male, 45–64 years | 0.44 | −0.53 | ||
| Male, 65–84 years | 0.47 | |||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 1–14 years | −0.52 | |||
| Female, 15–24 years | 0.57 | |||
| Female, 25‐44 years | 0.42 | |||
| Female, 45–64 years | ||||
| Female, 65–84 years | 0.53 | |||
| Female, >84 years | 0.48 | 0.42 | ||
| “Residual” | ||||
| Male, <1 year | ||||
| Male, 1–14 years | 0.66 | |||
| Male, 15–24 years | ||||
| Male, 25–44 years | 0.78 | |||
| Male, 45–64 years | ||||
| Male, 65–84 years | ||||
| Male, >84 years | ||||
| Female, <1 year | ||||
| Female, 1–14 years | 0.68 | |||
| Female, 15–24 years | ||||
| Female, 25–44 years | 0.80 | |||
| Female, 45–64 years | 0.56 | |||
| Female, 65–84 years | 0.45 | |||
| Female, >84 years |
*Only rates that are substantially correlated with factors (loading>0.40), and help interpret these factors are indicated.
Bold figures indicate a very high correlation between the rates and factors (loading>0.75).
Female, >84 years.
The correlations, patterns and trends for over 30 years observed with rates formed using “any” causes—that is, also using immediate and contributory causes (data not shown, available from the authors) —and those using a more detailed list of 46 diseases for the underlying cause of death were very similar (data not shown, available from the authors); this implies that the results are robust. Furthermore, analysis carried out with the detailed list indicated that the third pattern during the period 1988−99 was mainly shaped by HIV infection in men and that the fourth pattern was shaped by injuries and poisoning (other than suicide), falls and traffic accidents.
Discussion
The limitations of death certification are well known.18,19 There are various reasons for inaccuracies of death certificate information, including antemortem diagnosis errors, medical records being unavailable at the time of certification, misunderstanding of the certification process due to the inadequate training of doctors, and errors and inaccuracy when coding causes using the ICD. This classification reflects the prevailing medical knowledge (and caveats) and diagnostic habits, and also changes repeatedly. These phenomena impair long‐term comparisons. In particular, there were major ICD changes in France in 1968 (when ICD‐8 was implemented) and 2000 (when ICD‐10 was implemented), which obliged us to carry out simple longitudinal comparisons of restricted lists of diseases or groups of diseases little affected by nosological and diagnostic changes during the relatively short 30‐year period from 1968 to 1999 (unlike other successive ICD revisions, there were few differences between ICD‐8 and ICD‐9). Despite these limitations, our study nevertheless provides evidence for the coexistence of four main patterns of mortality in contemporary France and for the evolutionary nature of these patterns. Interestingly, patterns reflecting the “age of degenerative and man‐made diseases”; the “age of delayed degenerative diseases”; and the “age of emergent and re‐emergent infections” of Omran's classical presentation of the epidemiological transition theory were observed simultaneously. Also, we found that the recession of infectious (including epidemic) diseases was only partial, and even in more recent times these diseases remain causes of death in the elderly population. Although our findings are only from one country (France) and may be criticised for a lack of representativeness and generalisability, they generally confirm the conclusions of a previous analysis conducted in The Netherlands12 and suggest that the concept of epidemiological transition, the empirical foundations of which (shift of communicable diseases to non‐communicable diseases while populations are ageing) cannot be contested, at least for Western Europe and North America, must be used much less simplistically and deterministically than is sometimes the case.20 Rather than “succeeding ages” of the epidemiological transition, our results suggest “succeeding patterns of mortality” or, even more accurately, an “evolutionary structure of patterns” within which patterns strengthen or weaken or vary as causes of deaths in the population change. In the case of France, two main patterns of death clearly strengthened in this structure during the past 30 years: (a) death by neoplasms, cardiovascular and digestive diseases of those aged between 45 years and 84 years and (b) death of the oldest from all causes (replacing patterns of deaths from respiratory and cardiovascular diseases, and deaths from neoplasms, nervous diseases and mental disorders in the elderly people, respectively). Over the same period, mortality among young men (notably due to HIV infection) replaced that of young children as the third pattern, and mortality due to injuries remained stable as the fourth.
Our study also shows very strong and consistent correlations between several cause‐specific death rates. Although diagnostic and coding habit artefacts cannot be completely excluded, this correlation suggests an interdependency—association or competition—of several diseases as the leading causes of death. This result, not described previously, was predicted by the historian Mirko Grmek21 who elaborated the concept of pathocenosis in the 60s: the idea is that as for species of an ecosystem, frequencies of diseases of the “pathocenosis” (here, those diseases leading to death) depend on endogenous (eg, sex and age) and ecological factors (eg, diet, smoking and traffic), but also on the frequencies of other diseases (many diseases are competitive and some mutually exclusive in humans). Consequently, they tend towards a state of equilibrium, at least in the short term. We indeed observed this tendency towards equilibrium over the three decades of our study, with the mortality structure clearly strengthening along two main patterns (see earlier).
Finally, these findings may be relevant to mortality surveillance. Considering structures or patterns of mortality and changes among them rather than the trends for (many) individual causes of death may indeed give simpler and more robust predictions of effect of diseases leading to death, and the needs of populations in the future. Our findings suggest that cause‐specific mortality should be assessed in populations as a multidimensional phenomenon. Identifying its main patterns and components may help in the construction of composite measures of this phenomenon with higher validity than the groupings currently used (eg, by WHO22), which are often derived from oversimplified epidemiological transition theories.
What is already known on this topic
The popular and widely used epidemiological transition theory is based on a succession of specific “patterns” of causes of death in human societies.
However, the reality and consistency of patterns of causes of death in a population at a given moment has never been formally evaluated.
What this study adds
Our study provides evidence for an evolutionary structure of patterns of mortality in contemporary France and therefore suggests using the concept of epidemiological transition in a less simplistic way than is commonly the case.
Policy implications
Our study also shows much stronger interrelationships between diseases leading to death than is usually believed and suggests that current categorisations of cause‐specific mortality in populations need reconsideration.
Acknowledgements
We thank Mrs F Laurent (INSERM CépiDc) for help in data handling and Professor G Salem and Dr JC Salomon for helpful comments on a previous draft of the manuscript. We also thank the reviewers for their useful comments.
Abbreviations
ICD - International Classification of Diseases
PCA - principal component analysis
WHO - World Health Organization
Footnotes
Competing interests: None.
References
- 1.Seshamani M, Gray A M. A longitudinal study of the effects of age and time to death on hospital costs. J Health Econ 200423217–235. [DOI] [PubMed] [Google Scholar]
- 2.Tuljapurkar S, Li N, Boe C. A universal pattern of mortality decline in the G7 countries. Nature 2000405789–792. [DOI] [PubMed] [Google Scholar]
- 3.McMichael A J, McKee M, Shkolnikov V.et al Mortality trends and setbacks: global convergence or divergence? Lancet 20043631155–1159. [DOI] [PubMed] [Google Scholar]
- 4.Omran A R. The epidemiologic transition. A theory of the epidemiology of population change. Milbank Mem Fund Q 197149509–538. [PubMed] [Google Scholar]
- 5.Olshansky S J, Ault A B. The fourth stage of the epidemiologic transition: the age of delayed degenerative diseases. Milbank Q 198664355–391. [PubMed] [Google Scholar]
- 6.Rogers R G, Hackenberg R. Extending epidemiologic transition theory: a new stage. Soc Biol 198734234–243. [DOI] [PubMed] [Google Scholar]
- 7.Omran A R. The epidemiologic transition theory revisited thirty years later. World Health Stat Q 19985199–119. [Google Scholar]
- 8.Olshansky S J, Carnes B, Rogers R G.et al Emerging infectious diseases: the fifth stage of the epidemiologic transition? World Health Stat Q 199851207–217. [Google Scholar]
- 9.Olshansky S J, Carnes B, Rogers R G.et al Infectious diseases, new and ancient threats to world health. Popul Bull 1997521–52. [PubMed] [Google Scholar]
- 10.Carolina M S, Gustavo L F. Epidemiological transition: model or illusion? A look at the problem of health in Mexico. Soc Sci Med 200357539–550. [DOI] [PubMed] [Google Scholar]
- 11.Caselli G, Mesle F, Vallin J. Epidemiologic transition theory exceptions. Genus 2002589–52. [Google Scholar]
- 12.Wolleswinkel‐van den Bosch J H, Looman C W, Van Poppel F W.et al Cause‐specific mortality trends in The Netherlands, 1875–1992: a formal analysis of the epidemiologic transition. Int J Epidemiol 199726772–781. [DOI] [PubMed] [Google Scholar]
- 13.World Health Organization International classification of diseases. 8th revision. Geneva: WHO, 1967
- 14.World Health Organization International classification of diseases. 9th revision. Geneva: WHO, 1977
- 15.EUROSTAT Health statistics. Atlas on mortality in the European Union (data 1994–96). Luxembourg: European Communities, 2002
- 16.Jackson J E.A user's guide to principal components. New York: Wiley, 1991
- 17.Coste J, Bouee S, Ecosse E.et al Methodological issues in determining the dimensionality of composite health measures using principal component analysis: case illustration and suggestions for practice. Qual Life Res 200514641–654. [DOI] [PubMed] [Google Scholar]
- 18.Comstock G W, Markush R E. Further comments on problems in death certification. Am J Epidemiol 1986124180–181. [DOI] [PubMed] [Google Scholar]
- 19.Jougla E, Pavillon G, Rossollin F.et al Improvement of the quality and comparability of causes‐of‐death statistics inside the European Community. EUROSTAT Task Force on “uses of death statistics”. Rev Epidemiol Sante Publique 199846447–456. [PubMed] [Google Scholar]
- 20.Caldwell J C. Population health in transition. Bull World Health Organ 200179159–160. [PMC free article] [PubMed] [Google Scholar]
- 21.Grmek M D. Préliminaires d'une étude des maladies. Ann ESC 1969241437–1483. [Google Scholar]
- 22.Murray C J, Lopez A D. Mortality by cause for eight regions of the world: Global Burden of Disease Study. Lancet 19973491269–1276. [DOI] [PubMed] [Google Scholar]
