- Public Health and the Risk Factor: A History of an Uneven Medical Revolution
Risk factors can be reduced or removed by interventions in populations, individuals, or both. In this study, William Rothstein first traces the use of the concept of the risk factor (not the name) to improve population health from the seventeenth century through the first half of the twentieth; surprisingly, he does not cite the work of Thomas McKeown or his recent critics.1 Then he deals with the extrapolation of risk factors in populations to risk factors in individuals. Quoting Claude Bernard, he points out one problem: "'an average description [of the risk factor for a group] will never be matched in nature,' inasmuch as each case differs from the average. . . .' 'In my opinion statistics [risk factor data] can never yield scientific truth'" (p. 19).
Another reason why risk factor data may fail is that despite a risk factor's association with disease, it may not be causally related. In this regard, Rothstein cogently notes that the decline of premature deaths due to coronary heart disease (CHD) began before campaigns to reduce putative personal risk factors (as well as changes in the medical care of CHD) became widespread. He does not discuss the relation of socio-economic status (SES) to the decline, although he documents low SES in the second half of the twentieth century as a risk factor for CHD. Readers are left wondering what was going on. One possibility, not discussed by Rothstein, is the much poorer SES of the cohort born during the Great Depression and World War II than of previous or subsequent cohorts: they comprised the middle-aged population (45-60 years) between 1975 and 1990, and it was in this cohort that the decline in CHD mortality occurred.
According to Rothstein, the risk factor was invented by life insurance companies when they realized that instead of denying insurance to anyone who was not healthy, "they should have graded the risks of the applicants and adjusted the premiums to the level of risk" (p. 61). With the sale of inexpensive "industrial" life insurance (which paid for little more than burials) to millions of workers, the major life insurance companies had sufficiently large numbers of policyholders in the first decades of the twentieth century to grade risk fairly accurately. As Rothstein describes, census and vital statistics data go back much further, but their lack of completeness and accuracy were drawbacks. Moreover, not until the second half of the nineteenth century did census data include occupation and other social factors.
Rothstein nicely describes the different meanings of "statistics" over the past centuries. He does not describe how statisticians estimate whether associations [End Page 927] are due to chance by techniques such as chi-square or t-test; this will be a relief to some readers, but a shortcoming to others. He discusses problems of small and large sample sizes, pointing out: "Statistical tests are designed to attach great weight to the size of the samples, so that very large samples make trivial differences statistically significant" (p. 366). Surprisingly, he seldom considers magnitude of risk in terms of odds ratios. Instead, he emphasizes correlations between mortality rates and putative risk factors, a much cruder measure. Moreover, society's interest has shifted from the relation of risk factors to mortality to their relation to disease onset. The emphasis throughout the book is on mortality.
Despite its weaknesses, this is a highly readable and informative book.
Bloomberg School of Public Health
1. Thomas McKeown, The Role of Medicine: Dream, Mirage, or Nemesis? (Princeton: Princeton University Press, 2003), p. 207; James Colgrove, "The McKeown Thesis: A Historical Controversy and Its Enduring Influence," Amer. J. Publ. Health, 2002, 92: 725-29; Simon Szreter, "Rethinking McKeown: The Relationship between Public Health and Social Change," ibid., pp. 722-25.