NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Research Council (US) Committee on Future Directions for Behavioral and Social Sciences Research at the National Institutes of Health; Singer BH, Ryff CD, editors. New Horizons in Health: An Integrative Approach. Washington (DC): National Academies Press (US); 2001.

Cover of New Horizons in Health

New Horizons in Health: An Integrative Approach.

Show details

10Methodology Priorities

A central feature of each topic area in this report is a focus on complex dynamic systems. The conceptual formulation of such phenomena as predisease pathways, the influence of collective community properties on individual health, and resilience in the face of adversity involve integrating components of complex systems. Methodological innovation will be needed to achieve such integration. New measurement techniques and designs at both the animal and human levels are necessary to build bridges that link the social and psychological levels of description to biology. The specification of multiple methods to strengthen support for—or refutation of—proposed linkages among complex psychosocial and physiological systems (Kagan, 1999) is an important priority across the full range of phenomena discussed in this report. Finally, the objective of understanding complex systems at multiple levels of description poses new statistical challenges, beyond the reach of currently available techniques. These challenges extend to the need to design and evaluate multicomponent intervention studies.

We have intentionally refrained from putting forth a formal complex systems model incorporating the full range of phenomena discussed. This would be premature in topic areas that are in a state of flux, where new developments in the published literature appear weekly. The topics raised in this chapter are meant to be illustrative rather than comprehensive. The challenge of putting forth, analyzing, and defending (with empirical data) formal integrated systems models nonetheless represents an important priority in its own right.


A variety of measurement and data analytic questions need to be resolved before multiple topics discussed in this report can be characterized with precision. Prediseases pathways, life histories of resilience in the face of adversity, and delineation of the biology of flourishing are among phenomena that call for methodological innovation. Priority topics include the need for greater investment in longitudinal designs, the measurement of numerous components of predisease pathways (e.g., childhood and early life influences, work and unemployment, positive health and resilience, collective properties and inequalities), the need to advance understanding of biological mechanisms, and the need for innovative methods of data analysis.

Investment in Longitudinal Studies

Implementation of the integrative perspective requires longitudinal assessment at multiple levels (psychosocial and biological) on the same individuals. This raises the practical question of how comprehensive such measurement can be on single populations and where inferences about pathways must be derived from studies of multiple populations. Concerning psychosocial and biological measures on the same population, several longitudinal studies illustrate needed future directions. The MacArthur Study of Successful Aging (Seeman et al., 1997) has measures of social integration and social support together with assays of glucocorticoids, catecholamines, cholesterol, glycosylated hemoglobin, blood pressure, and measures of adipose tissue deposition (thereby representing the preliminary operationalization of the concept of allostatic load) on the same individuals. A subsample of the Wisconsin Longitudinal Study (WLS) contains extensive psychosocial life history data (over a span of 40 years), community-level information, and multiple neurophysiological assessments (e.g., measures in the current allostatic load inventory, immune system assays of antibody responses to influenza and hepatitis A vaccine, EEG assessments of brain asymmetry, and fMRI assessments; e.g., Singer and Ryff, 1999). The 1999 round of the National Long Term Care Survey contains an extensive array of biomarkers, including assessment of apolipoprotein E (apoE) markers of genetic susceptibility to onset of Alzheimer's. 1 The Whitehall II study of British civil servants (Marmot et al., 1991) and the 1946 (Wadsworth and Kuh, 1997) and 1958 (Power and Matthews, 1998) British birth cohorts each have new biomarker data collections simultaneous with the rich psychosocial and health measures that have been repeatedly collected in these studies. The ever more comprehensive biopsychosocial data available in the above studies provide far more informative understanding of health pathways than can be obtained by creating synthetic cohorts by working across multiple studies.

The above examples, by no means exhaustive, clarify that extensive data collection is in fact feasible on factors that may initially appear to pose excessive respondent burden. A counterpoint to the above surveys (originally oriented more toward psychosocial assessments), the Framingham Study (Dawber, 1980; Allaire et al., 1999) is rich in longitudinal health and biomarker assessments but weak in psychosocial information. Ideally, what is needed are both kinds of information assessed on the same individuals over time. Such data would greatly facilitate understanding the linkages between physiological, psychological, and sociological phenomena. Gene expression studies on large populations that, until a few years ago would not have been feasible, can also now be considered in future data collection, thanks to microarray chip technology (see Chapter 4). However, before delineating practical next steps regarding investment in longitudinal studies, it would be useful to consider what an optimal portfolio of such studies might be.

First, it would be useful to have several birth cohorts specifically designed for the integrative agenda outlined in this report. This would include DNA samples at birth with gene expression assessments over the life course. Measurement of successful functioning, or the lack thereof, of multiple physiological systems over the life course should be included. Collective properties of communities that influence these individuals (see Chapter 6) should be included together with a substantial array of psychological and social assessments, as delineated in Chapter 2, Chapter 3, and Chapter 5. Some individuals in these birth cohorts would be exposed to “natural” experiments (e.g., legislation prohibiting smoking in public places, welfare reform, regulation on the placement and protective features of toxic waste sites) having possible health consequences. Information about the impact of such policy-based, macro-level interventions on individuals should be assessed.

Second, beyond such comprehensive cross-time data collection, it would be desirable to have several explicit intervention studies in the portfolio, including some that are community based and involve multiple interventions acting simultaneously. The interventions should be of both the health-promoting and disease-preventing varieties (see Chapter 3). In addition, subsamples from such studies should be accessible for carrying out small-scale challenge studies, the purpose of which should be to link psychosocial challenges (e.g., marital conflict, performance evaluation) with biological antecedents and consequences. These purposive challenge studies should also be a part of protocols for the birth cohorts, where naturally occurring interventions predominate.

In-depth studies for some biomedical questions may not be practical within large longitudinal surveys or community-based intervention studies. Indeed, it is difficult to identify clear boundaries between the kinds of information usefully ascertained in large population studies and what must be left for investigations in smaller special populations. For this reason, we focus attention on the above broad categories of longitudinal studies as optimal targets and now turn to the practical processes of developing a useful portfolio.

To facilitate National Institutes of Health (NIH) engagement in support of future data collection for integrative studies, we recommend that a series of workshops be convened to identify the varieties of pathways (road maps) that are most in need of ongoing assessment. These workshops should also have the delineation of cost-effective optimal designs as a major objective. As indicated in other sections of this report (e.g., Chapter 2, Chapter 3, and Chapter 7), there are few extant longitudinal surveys (birth cohorts in particular) that could be the basis for implementing optimal designs. Thus, it would be important in such workshops to clarify the degree to which optimal designs could be achieved by piggybacking new measurements onto existing longitudinal studies. If resources were committed, for example, to separate surveys focused on children, midlife, and the elderly, which studies would represent the best opportunities for incorporating new instruments to build toward a picture of pathways? These workshops could also address what retrospective instruments (psychosocial and physiological signatures of past adversity and advantage) should be used in midlife and elderly population studies. The overall goal would be the specification of a long-term plan, with cost estimates, for implementing pathway studies. An important component of this recommendation would be the establishment of an ongoing advisory group that makes decisions adaptively, informed by accumulating evidence about the evolution of pathway studies. Measurement priorities related to components of pathways are identified below.

Measurement of Early Life and Childhood Influences

As documented in, prenatal experience plays a central role in interacting with the genome to influence brain development. These epigenetic influences in intrauterine life confer a set of predispositions that act across the life span to affect vulnerability for a host of chronic diseases. The quality of parenting behavior plays a central role in the development of stress regulatory systems. Children exposed to parenting characterized by conflict, aggression, and neglect showed disruption in the sympathetic adrenomedullary (SAM) and hypothalamic-pituitary-adrenal (HPA) regulatory systems. Conversely, positive maternal behavior protects against expression of genetic risk for serotonin dysregulation. Parenting also plays an important role in modifying genetically based temperamental differences in children. Such evidence indicates that empirical specification of pathways requires delineating best indicators of quality of parenting, temperaments, and family environments that can be utilized in large longitudinal studies. A companion set of physiological indicators of SAM and HPA system functioning is also required for use in large population studies. Protocols for salivary cortisol assessments, for example, on children are well developed (Gunnar, 1999). In addition, a broad array of biomarkers on children were collected as part of the NHANES III (U.S. DHHS, 1994). 2 What is required is the identification of subsets of biomarkers and protocols for their assessment that are most directly associated with psychosocial measures used to construct pathways.

Assessment of Personal Ties

Diverse forms of personal ties (e.g., mother-child attachment, spousal intimacy, close friendships, relationships with work colleagues) are central to pathway specifications relating to both negative and positive health outcomes (see Chapter 5). Recent literature suggests that the emotional aspects of personal ties are salient features linking the psychosocial level of description to biology and downstream health consequences (Ryff and Singer, in press). The study of emotion in personal ties requires further development at a conceptual and measurement level to relate emotions of both positive and negative valence to health. What is needed is a multi-method technology for assessing emotion in personal ties (e.g., survey instruments, focused interviews, writing tasks, experimental protocols) that together comprise comprehensive specification of pathways at the psychosocial level and their relationship to biology. A thorough investigation of the interplay between psychosocial assessments and neurophysiological measurements (e.g., left- versus right-side activation of the prefrontal cortex connected to the emotional aspects of personal ties; Ryff and Singer, in press) is an essential component of such a program of methodological development.

Assessment of Work and Unemployment Influences

There is an extensive literature associating adversity in the workplace and spells of unemployment with later-life chronic disease and mortality (Marmot et al., 1997). In addition there are observational studies over short time intervals relating the pace and character of repetitive work tasks with mental health and musculoskeletal disorders (Lundberg, 1999). What is needed is delineation of the best measuring instruments for relating workplace and unemployment experience to the cumulative physiological risk (Grossi et al., 1998), such as allostatic level. This should include measures of cumulative adversity and responses to short-term challenges that are reflected in shifts in allostatic load over the life course.

Assessment of Positive Health and Resilience

Pathways of resilience and of persistent flourishing require specification of the principal positive features of lives at the individual and community levels that are most strongly associated with the maintenance of allostasis (see Chapter 3). Few extant studies discuss this kind of question. Nevertheless, the growing literature linking, for example, physical exercise and a variety of enriched environments with neurogenesis in animals suggests that this is the primary area for NIH-sponsored research. At the level of elderly human populations, the documentation of subgroups who maintain or improve their mental and/or physical health with increasing age (Ryff et al., 1998) also points to the need for expanded assessment of behavioral, psychological, social, and biological factors implicated in the maintenance of positive health. Delineation of the best measuring instruments at the psychosocial and biological levels to understand optimal biopsychosocial functioning in the face of adversity is a high priority for methodological development.

Particularly important is the need for research on the neurobiological mechanisms underlying the health benefits ensuing from behavioral and psychosocial influences. What are the actual processes (e.g., neural circuitry, endocrine and immune functions) through which behavioral (e.g., nutrition, exercise, stress management) and psychosocial (optimism, coping quality social ties) factors convey their health-promoting effects? This is a call to advance what is known about the physiological substrates of flourishing (Ickovics and Park, 1998; Ryff and Singer, 1998). Such inquiry has begun, as for example, in studies linking social supports to physiological processes (Seeman and McEwen, 1996; Uchino et al., 1996). These agendas have tended to focus, however, on the endocrinological and immunological correlates of relational conflict or caregiving demands (Kiecolt-Glaser et al., 1996, 1997), not relational strengths. Thus, there is a major need for new studies linking positive aspects of social relationships (attachment, affection, intimacy) to the mechanisms that underlie good health (Ryff and Singer, 2000). Animal research provides extremely valuable models for such explication of the mechanisms that connect positive social relations to health (Carter, 1998; Uvnas-Moberg, 1997, 1998).

Assessment of Collective Properties and Inequalities

Pathways specified by measures at the individual level must be linked to more macro-level phenomena that have been associated with individual-level health. Neighborhood-, city-, district-, county-, state-, and even national-level properties can have impact on the health status of entire communities as well as on the individual person. The existence of inequality reflects degrees of collective adversity relative to advantage and requires establishing profiles of individual histories that specify cross-domain influences. The growing body of instruments for measuring such phenomena as social cohesion and social control (Sampson et al., 1997) must be augmented with improved measures of community physical environments that bear directly on health risks (e.g., garbage in the streets; unsafe housing; accessibility of parks, recreational sites, and open spaces to more densely populated urban tracts; Institute of Medicine, 1999b).

The measurement of collective social characteristics themselves, including assessment of social well-being (Keyes, 1998), also needs refinement regarding their relationship to a diversity of health outcomes. Environments that are less divisive, less undermining of self-confidence, less conducive to antagonism, and more supportive of developing effective life skills are important goals for future research and intervention. Overall, the science of ecological assessment of social and physical environments (i.e., “ecometrics,” see Chapter 6; Raudenbush and Sampson, 1999) relevant to health is underdeveloped and requires a concerted methodological effort bolstered by NIH support. Linking these collective measures to psychosocial and biological indicators at the individual level is a critical priority for characterizing the full range of pathways discussed in this report.


Going beyond statistical association to address the mechanisms by which behavioral, psychological, and social factors operate at various levels requires methodological development. In particular, ways of measuring the mechanisms that convey and moderate psychosocial influences on health are a top priority. The two most pressing needs are to understand the neural circuitry whereby environmental conditions and events affect emotions and downstream health outcomes, thereby extending the measurement of cumulative physiological burden and its consequences.

Neural Circuitry

Fundamental to developing pathway characterizations is an understanding of the chain of interrelationships from the macro-environmental level to the neural level. A central focus is on factors that influence psychological, social, and emotional processes, which when instantiated in the brain have downstream autonomic, endocrine, and immune consequences impinging on health. These psychosocial processes operate at an intermediate level that can be regarded as a necessary pathway through which external challenges influence health. In recent years there has been considerable progress in understanding the central neural circuitry that underlies emotion and affective style (Davidson, 1998). With this knowledge, it is possible to formulate mechanistic hypotheses about how life events and psychosocial factors, insofar as they influence brain circuitry, can have downstream effects on the periphery and thereby influence health.

Most extant research indicates variations in activation of the left prefrontal region with differences in approach-related positive affect (Sutton and Davidson, 1997). Decreased activation in this region is associated with increased vulnerability to depression, while increased activation is associated with dispositional positive affect and a coping style found to be protective against depressive symptomology (Tomarken and Davidson, 1994). Thus, given the same negative life events, an individual with baseline-left-anterior hypo-activation is hypothesized to show more intense depressive symptomology compared with a subject showing left-anterior hyper-activation.

From the perspective of brain function, resilience (see Chapter 3) may thus involve capacities to activate approach-related affective processes in the face of negative environmental stressors. Such capacities are likely products of enduring individual differences in dispositional mood and established patterns of emotional reactivity. Research has shown that subjects showing stable but extreme electrophysiological asymmetry differ in ratings of their own moods: those with extreme activation in the left prefrontal part of the brain showed both significantly more positive affect and significantly less negative affect than their right prefrontally activated counterparts (Tomarken et al., 1992).

To provide the necessary links between these aspects of affective neuroscience and pathway characterizations, it is essential to support the following research directions: (1) study profiles of brain circuitry over long time periods (prior work, described above, has focused on stability in asymmetries over relatively short intervals); (2) investigate whether greater left-sided prefrontal activation confers resilience in the face of cumulative negative life challenges; (3) map connections between hemispheric asymmetry and psychosocial variables (e.g., well-being, coping, quality of personal ties) and other biological indicators (allostatic load, measures of immune function). The development and implementation of protocols for longitudinal assessment on diverse populations (e.g., WLS, the 1946 and 1958 British birth cohorts) are a high priority. Resolution of these measurement issues would provide a critical link toward understanding the mechanisms by which proximate and distal social influences exert their health-transforming effects. An important advance would be instrumentation to facilitate EEG assessments on large populations under diverse environmental conditions. In particular, microminiaturization of the EEG recording and data storage units would make ambulatory measurement of brain activation in response to naturally occurring stimuli a reality.

Refinement in the Operationalization of Allostatic Load

Implicit in the concept of allostasis, “achieving stability through change” (see Chapter 3), are nature-nurture interactions in which genes are regulated by environmental factors, leading in the short run to adaptation and in the long run to increased risk for disease. A problem with the original conceptualization of allostatic load and its measurement, however, is that the components were not organized and categorized with regard to what each measure represents in the cascade of events that lead from allostasis to allostatic load. A step toward improving the formulation of culmulative physiological burden (McEwen and Seeman, 1999) has utilized the notion of primary mediators leading to primary effects and then to secondary outcomes, which lead finally to tertiary outcomes that represent actual diseases.

Primary mediators are chemical messengers released as part of allostasis. The present operationalization of allostatic load (Seeman et al., 1997Seeman et al., in press) has four such mediators: cortisol, epinephrine, norepinephrine, and DHEA-S. These mediators are accessible and relatively easy to collect and measure from body fluids, such as saliva, urine, and blood. Other mediators could include inflammatory cytokines and insulin-like growth factors. These and other future measures viewed as primary mediators have wide influences throughout the body and are useful in predicting a variety of secondary and tertiary outcomes.

Primary effects are cellular events, like enzymes, receptors, ion channels, or structural proteins induced genomically or phosphorylated via second messenger systems, that are regulated as part of allostasis by the primary mediators. These are not presently measured as part of an operationalization of allostatic load, although it may be desirable to include them in future formulations, as they are the basis for secondary and tertiary outcomes. Primary effects are organ and tissue specific, and as a result secondary and tertiary outcomes must be described at this level. The connections between primary effects and secondary and tertiary outcomes represent the current mechanistic research supported by NIH.

Secondary outcomes are integrated processes that reflect the cumulative outcome of the primary effects in a tissue/organ-specific manner in response to the primary mediators. Current operationalizations of allostatic load include the following secondary mediators, all of which are related to abnormal metabolism and risk for cardiovascular disease: waist/hip ratio, systolic and diastolic blood pressure, glycosylated hemoglobin, total/HDL cholesterol, and HDL cholesterol. In the future the secondary outcomes should be expanded in two directions. First, there is a need for more specific outcomes related to damage along the pathway of cardiovascular risk in relation to job stress and socioeconomic status (Markowe et al., 1985). Second, outcomes of cumulative burden in other systems, such as the brain and the immune system, are needed. For the brain, assessments of declarative and spatial memory have been employed to identify individual differences in brain aging, reflecting atrophy of the hippocampus and progressive elevation of cortisol (Lupien et al., 1998). For the immune system, integrated measures of the immune response such as delayed-type hypersensitivity (Dhabhar and McEwen, 1999) and immunization challenge (Dhabhar and McEwen, 1996) could reveal the impact of allostatic load on cellular and humoral immune function and help distinguish between the immuno-enhancing effects of acute stress and the immunosuppressive effects of chronic stress.

Tertiary outcomes are the actual diseases or disorders that are the result of the allostatic load predicted from the elevated/extreme values of the secondary outcomes and primary mediators. Thus far, the outcomes associated with high allostatic load have been cardiovascular disease, decreased physical capacity, severe cognitive decline, and mortality. Some redefinition of outcomes is needed. A stricter criterion based upon the definitions of primary, secondary, and tertiary outcomes would assign cognitive decline as a secondary outcome. Alzheimer's disease or vascular dementia would be included as tertiary outcomes, as these are cases where there is serious and permanent disease. Cancer would also be a tertiary outcome, clearly reflecting a compromised immune system as well as other systemic changes in endocrine and metabolic responses.

The above classification system should serve as a template for new studies relating progression from primary mediators to secondary outcomes and then to tertiary outcomes. New research should identify clusters of secondary outcomes that are relevant to particular diseases. Measurement of secondary outcomes at younger ages (as part of new operationalizations of allostatic load) must be implemented in longitudinal studies with tertiary outcomes at later ages. A more refined statistical methodology for combining information across primary mediators, primary effects, and secondary outcomes to establish a new index of allostatic load is a high priority for future research.

Beyond the above classification system and its attendant measurement problems are other methodological issues that warrant attention. First, the measures incorporated in the current operationalization of allostatic load are all based on resting level, or static, assessments. As a collectivity, they are supposed to represent the physiological signature of cumulative psychosocial adversity. This is only one aspect of allostatic load (McEwen and Seeman, 1999). Equally important is abnormal transient response of primary mediators to specific challenges because the repeated overactivity of a physiological system is a source of allostatic load (McEwen, 1998). Impaired transient response to acute challenge should be part of the physiological signature of cumulative adversity. Protocols in animals and humans are needed to document the phenomena and lead to an index of allostatic load incorporating both resting level and transient response measures.

A second missing feature of current operationalization of allostatic load is the specification and implementation of a defensible set of indicators for children. The extant indicators were developed for and calibrated on a longitudinal study of the elderly (Seeman et al., 1997). The same indicators have been used on a midlife sample (ages 59-60; Singer and Ryff, 1999), and a slightly modified operationalization has been utilized in the Normative Aging Study (Kubzansky et al., 1999). A comparison of the distribution of responses across these samples suggests that an age-specific specification of allostatic load would be appropriate. Since a characterization must involve physiological processes that evolve over the life course, the development of a whole-life animal model of allostatic load is likely to yield major payoff in developing the human analog. This would entail setting up experimental protocols that vary the frequency and types of psychosocial challenges to which the animals are exposed over their lives. It would also require age-dependent specification of the various components of the classification system described above.

A third aspect of the further operationalization of allostatic load that warrants further investigation is gender differences in resting levels and in transient response to challenges of particular primary mediators and secondary outcomes. For example, a higher proportion of elderly women show elevated levels of cortisol, epinephrine, and norepinephrine and low levels of DHEA-S. Elderly men tend to have elevated levels of syndrome X markers with greater frequency than women. In response to a 30-minute cognitive challenge, younger men showed greater cortisol reactivity than women. However, in elderly populations the pattern is reversed, with elderly women showing greater transient responses (Seeman et al., in press.) There is a pressing need to specify protocols that would clarify the scope and character of age-specific gender differences in the components of a more refined specification of allostatic load.

The fourth challenge is the integration of knowledge about environmental influences on primary, secondary, and tertiary outcomes with the wealth of information coming from the human genome project. The ultimate goal of mapping the human genome is to enable medical science to unravel the genes that are involved in disease processes as well as the genes that help protect the body from disease. Such genes may be involved at any or all stages of the life course. Regulation of these genes is one of the greatest challenges facing biomedical science in this century.

A few words are in order about statistical analyses pertaining to the current formulation of allostatic load and its further development and operationalization. The primary and secondary mediators linked to secondary and tertiary outcomes represent a system of multiple indicators reaching high risk levels over time and subsequently influencing diverse kinds of outcomes. Organizing these dynamics in a comprehensive statistical framework could involve the use of structural equation modeling, including the incorporation of latent variables to represent unmeasured biological parameters. Development and implementation of such a framework are important aspects of operationalizing the notion of cumulative physiological risk, for which allostatic load is an illustrative beginning.


The statistical challenges associated with pathway characterizations center around the necessity of developing new person-centered methods for representing the complex multidimensional and multilevel dynamics described throughout this report. Priorities in this area focus on three interrelated topics. First is the need to further develop both agglomerative and partition-based strategies for specifying categories of pathways. Second, new modeling, estimation, and testing protocols are required to carry out multilevel analyses that incorporate many of the nonlinearities within and between collective and individual-level descriptions of pathway components. Third, we require effective strategies for integrating numerical and narrative information as part of pathway characterizations. This integration includes the use of narratives in two forms: the construction of narratives by an investigator as part of a data analytic strategy and the analysis of narratives as data. These priorities are described in detail in the following three subsections.

Categories of Pathways

For an individual, a pathway specification at the psychosocial level consists of a chronology of the person's experience (where adversity and advantage interdigitate and sometimes run in parallel) across multiple life domains. There are feedback effects in which deteriorating health influences the extent and nature of subsequent adversity, and sustained positive experience can promote improvements in health. Similarly chronic adversity can promote deterioration in health, while high levels of mental and physical health can facilitate advantageous experience. The details of these individual histories are highly idiosyncratic. Thus, in specifying pathways to given health outcomes, strategies are required for abstracting and aggregating multiple histories so that empirically defined categories of histories emerge, in each of which individuals are regarded as approximate matches to one another. Statistical methodologies for the formation of coherent categories of pathways based on complex life histories are in their infancy. Such strategies are broadly referred to as person-centered methods, and they differ in both detail and objective from the widely available variable-centered techniques (Abbott and Hrycak, 1990; Cairns et al., 1998; Giele and Elder, 1998; Singer et al., 1998). The extant corpus of methods represents two distinct perspectives with the same basic objective. One set of methods starts with fine-grained individual case histories and aggregates up to categories of pathways. These are exemplified by the optimal matching (Abbott and Hrycak, 1990; Abbott and Barman, 1997), event-structure analysis (Griffin, 1993), and hybrid narrative/number-analytic techniques (Singer et al., 1998). The full individual history is the basic unit of analysis for all these methods. In optimal matching, for example, a metric is imposed on the set of possible histories. Those histories that fall within a prescribed small distance of one another are then identified as members of a common equivalence class (i.e., small idiosyncratic details that differ from one history to the next are disregarded and a more aggregate set of patterns is identified). There is an inevitable tension between the desire to preserve the nuances of an individual history and the necessity of disregarding some if aggregate categories are to be generated. The choice of metric reflects the features of histories to be emphasized in an aggregate category.

A second set of methods partitions the space of individual histories into successively more homogeneous categories, the final product being the desired pathway representations (Breiman et al., 1984; Zhang and Singer, 1999; Manton et al., 1994). Both of these classes of person-centered methods are substantially underdeveloped relative to the complexity of histories that would be the raw ingredients for longitudinal data analysis as outlined above. Furthermore, there are virtually no studies comparing and contrasting the scope and limitations of each perspective when applied to the same data set. Advancing person-centered statistical methodologies is essential for carrying out the empirical programs that appear throughout this report.


There is a growing set of applications of multilevel linear statistical models to quantify the impact of community-level indicators on individual experience and health (Malmstrom et al., 1999; Sampson et al., 1997; Yen and Kaplan, 1999). Several investigators (Manton et al., 1994; Breiman et al., 1984; Zhang and Singer, 1999) have called attention to the fact that the effect of substantial nonlinearities in the actual phenomena could be masked by the use of such over simplified models. Thus, it is imperative to expand the person-centered methodologies to incorporate collective properties of communities as they impact on individual-level experience. The multiple measurement strategies in the work on “ecometrics” as recently put forth by Raudenbush and Sampson (1999; see Chapter 6) must be linked systematically to the individual-level pathway construction methods. Another possible resource is the growing methodology of integrated assessment, drawing upon Bayesian approaches (e.g., influence diagrams) to combine research results from diverse disciplines (e.g., Clemen, 1991; Dowlatabadi and Morgan, in press; Howard, 1989). Applied most widely to complex environmental problems, with their nonlinearities, integrated assessments are increasingly being conducted for social-psychological domains, and their interaction with natural systems (e.g., Casman et al., 2000; Fischhoff, 2000; Fischhoff et al., 1998).

Person-Centered Narratives

Some of the extant person-centered methods use narrative information in two ways: (1) narratives based on interviews comprise part of the raw data that are to be combined with quantitative information in the process of pathway construction and (2) narratives describing whole lives, or major segments of them, are constructed by the analyst from long sequences of numerical information representing the longitudinal survey responses of a single individual. Under the latter the construction of narratives is part of an analytical strategy designed to formulate a coherent picture of a life history as it relates to physiology and health outcomes. Effectively integrating narratives as data, narratives constructed as part of an analytical strategy, and numerical information is a major challenge in pathway specification urgently in need of research support. Some movement in this direction is already apparent from recent (1999) planning workshops convened by the National Institute of Mental Health, but the general area is largely uncharted and is in need of much broader support.


Many of the most effective health promotion and disease prevention programs in the past have involved the use of multiple interventions acting simultaneously and/or in sequence (Watson, 1953; Warner, 2000). Voluntary self-selection into the programs is frequently a criterion for participation, thereby making controlled randomized trials impossible as a strategy for evaluation. Rehabilitation programs for chronic alcoholics, methadone maintenance programs for rehabilitation of heroin addicts, and family planning programs are among the many examples of interventions of this character (Singer, 1986). Self-selection is viewed in many quarters (Lipsey and Cordray, 2000) as a source of bias in assessing the impact of interventions. For programs of the type mentioned above, voluntary self-selection is part of the overall process to be studied. What is needed are defensible strategies for designing and evaluating complex intervention programs where self-selection is a criterion for entry into the program. A quite sophisticated labor economics literature on this topic has developed in the context of evaluating manpower training programs (Heckman and Smith, 1995). There is very little of similar character focused on health-based interventions where the details and level of complexity differ substantially from the labor economics context. Among the many issues that require attention is determining when one can defensibly ascertain the relative impact of a given intervention that is part of a package of interventions. A second important issue is that many successful multicomponent interventions in the past involved adaptive designs where the interventions were tuned experimentally, often over periods of several years, before optimal responses were obtained (Watson, 1921, 1953). In an environment where funding is linked to having all of the intervention structure specified a priori, packages of interventions that might be effectively tuned to optimal settings on given populations are seriously undermined. To broaden the base of design and analysis of multiple-component intervention programs, a series of methodological workshops should be convened to delineate funding priorities that would advance this subject on a broad scale. Many of the intervention programs implied by the discussion throughout this report are of the primary prevention variety. The methodology for effective evaluation of such programs is very underdeveloped and represents a pressing need.

Complementary to the need for population-level program evaluation strategies is the need to advance methodology focused on the information required by clinicians to guide patient management. Controlled randomized trials of pharmacological therapies usually focus on evaluating efficacy of a single drug as it performs “on average” in a population vastly more heterogeneous than the single patient of interest to the clinician. The standard clinical question is to ascertain how the contemplated therapy performed on a population of approximate matches, in terms of clinically relevant history, to a given patient (Horwitz et al., 1996, 1997, 1998). A single pharmacological agent is rarely the only intervention at issue. What the physician—and patient—need to know is how various packages of cotherapies perform on approximate matches to the given patient for whom an intervention program is being contemplated. Strategies for the assessment of packages of co-therapies that are tuned to the information needs of clinicians are in serious need of development. They represent the clinical counterpart to the evaluation methodology for multicomponent primary prevention programs described above. Serious attention should be given to this much neglected but ever more important aspect of providing effective curative medicine for the U.S. population.

Various NIH institutes support research designed to improve health-related decision making. We support the priority that NIH has set on developing better ways of presenting health care communications to lay people, especially the presentation of unfamiliar outcomes, processes, and units or comparisons (Fischhoff, 1999; Marcus, 1999; Woloshin and Schwartz, 1999). Attention also needs to be given to appropriate strategies in situation where causes, mechanisms, and outcomes are not viewed with broad consensus. In such situations the success of interventions cannot be evaluated simply in terms of how many individuals followed a common path. Rather, we need ways to evaluate our success in helping people to reach personally appropriate health decisions, ones that they understand, can live with, and are able to implement (e.g., Croyle and Lerman, 1999; Fischhoff, 1992; Rimer, 1995). Although health communications can be viewed in purely cognitive terms, they also touch the full human experience: not just the development of competency but the relationship between intellectual competence and self-efficacy needed for action, affective responses, and their impact on well-being (e.g., Baum et al., 1997; Institute of Medicine, 1999a). We recommend that research on these issues be encouraged and expanded.


New measurement techniques and designs for both animal and human studies are necessary to build bridges that will link behavioral, psychological, and social levels of analysis to multiple levels of biology (organ systems, cellular, molecular). This broad purview underscores the need for methodologies that are responsive to the functioning of complex dynamic systems through time. The emphasis on pathways in this report also implies a need for statistical methodologies that can specify pathway trajectories, address nonlinearities in diverse indicators, and incorporate narratives as sources of data. NIH should support methodological initiatives in three broad areas:

  1. refined operationalization of allostatic load that takes explicit account of the cascade of events from allostasis to cumulating load; ambulatory assessments and responses to transient challenges should be given consideration as potential components of improved measures of allostatic load.
  2. development of person-centered statistical methodologies to facilitate characterizations of predisease and positive health pathways that link behavioral, psychosocial, environmental, and biological levels of analysis;
  3. development of design, implementation, and analysis strategies for multiple-component interventions where, to obtain optimal outcomes, allowance is made for adaptive, dynamic tuning of the interventions.


  • Abbott A, Barman E. “Sequence comparison via alignment and Gibbs sampling: A formal analysis of the emergence of the modern sociological article” In: Raftery A, editor. Sociological Methodology. Washington DC: American Sociological Association; 1997. pp. 47–87.
  • Abbott A, Hrycak A. “Measuring resemblance in sequence data: An optimal matching analysis of musicians' careers” American Journal of Sociology. 1990;96:144–185.
  • Allaire SH, LaValley MP, Evans SR, O'Connor GT, Kelly-Hayes M, Meenan RF, Levy D, Felson DT. “Evidence for decline in disability and improved health among persons aged 55 to 70 years: The Framingham Heart Study” American Journal of Public Health. 1999;89(11):1678–1683. [PMC free article: PMC1508997] [PubMed: 10553388]
  • Baum A, Friedman AL, Zakowski SG. “Stress and genetic testing for disease risk” Health Psychology. 1997;16(1):8–19. [PubMed: 9028812]
  • Breiman L, Friedman J, Olshen R, Stone C. Classification and Regression Trees. Belmont CA: Wadsworth Publishing Co.; 1984.
  • Carter CS. “Neuroendocrine perspectives on social attachment and love” Psychoneuroendocrinology. 1998;23(8):779–818. [PubMed: 9924738]
  • Cairns R, Xie H, Leung MC. “The popularity of friendship and the neglect of social networks: Toward a new balance” New Directions for Child Development. 1998;(Summer)(80):25–53. [PubMed: 9641075]
  • Casman E, Fischhoff B, Palmgren C, Small M, Wu F. “Integrated risk model of a drinking waterborne cryptosporidiosis outbreak” Risk Analysis. 2000;20:493–509.
  • Clemen RT. Making Hard Decisions: An Introduction to Decision Analysis. Boston: PWS-Kent Publishing Co.; 1991.
  • Croyle RT, Lerman C. “Risk communication in genetic testing for cancer susceptibility” Journal of the National Cancer Institute Monographs. 1999;25:59–66. [PubMed: 10854459]
  • Dawber T. The Framingham Study. New York: Oxford University Press; 1980.
  • Davidson RJ. “Affective style and affective disorders: Perspectives from affective neuroscience” Cognition and Emotion. 1998;12:307–330.
  • Dhabhar F, McEwen BS. “Moderate stress enhances and chronic stress suppresses, cell-mediated immunity in vivo” Society of Neuroscience. 1996;22(536.5):1350.
  • Dhabhar F, McEwen BS. “Enhancing versus suppressive effects of stress hormones on skin immune function” Proceedings of the National Academy of Sciences of the United States of America. 1999;96:1059–1064. [PMC free article: PMC15350] [PubMed: 9927693]
  • Dowlatabadi H, Morgan MG. Integrated Assessment. New York: Cambridge University Press; In press.
  • Fischhoff B. “Giving advice: Decision theory perspectives on sexual assault” American Psychologist. 1992;47(4):577–588. [PubMed: 1595985]
  • Fischhoff B. “Why (cancer) risk communication can be hard” Journal of the National Cancer Institute Monographs. 1999;25:7–13. [PubMed: 10854449]
  • Fischhoff B. “Scientific management of science?” Policy Sciences. 2000;33:73–87.
  • Fischhoff B, Downs JS, Bruine-de-Bruin W. “Adolescent vulnerability: A framework for behavioral interventions” Applied and Preventive Psychology. 1998;7(2):77–94.
  • Giele J, Elder G. Methods of Life Course Research: Qualitative and Quantitative Approaches. Thousand Oaks CA: Sage Publications; 1998.
  • Griffin L. “Narrative, event structure analysis, and causal interpretation in historical sociology” American Journal of Sociology. 1993;98:1094–1133.
  • Grossi A, Ahs A, Lundberg U. “Psychological correlates of salivary cortisol excretion among unemployed men and women” Integrative Physiological and Behavioral Science. 1998;33:249–263. [PubMed: 9829437]
  • Gunnar M. “Early adversity and the development of stress reactivity and regulation” In: Nelson CA, editor. The Effects of Adversity on Neurobehavioral Development: Minnesota Symposia on Child Psychology. Vol. 31. Mahwah NJ: Lawrence Erlbaum; 1999.
  • Heckman J, Smith JA. “Assessing the case for social experiments” Journal of Economic Perspectives. 1995;9(2):85–110.
  • Horwitz RI, Singer BH, Makuch RW, Viscoli CM. “Can treatment that is helpful on average be harmful to some patients? A study of the conflicting information needs of clinical inquiry and drug regulation” Journal of Clinical Epidemiology. 1996;49(4):395–400. [PubMed: 8621989]
  • Horwitz RI, Singer BH, Makuch RW, Viscoli CW. “On reaching the tunnel at the end of the light” Journal of Clinical Epidemiology. 1997;50(5):753–755. [PubMed: 9253384]
  • Horwitz RI, Singer BH, Makuch RW, Viscoli CM. “Clinical versus statistical considerations in the design and analysis of clinical research” Journal of Clinical Epidemiology. 1998;51(4):305–307. [PubMed: 9539886]
  • Howard RA. “Knowledge maps” Management Science. 1989;35:903–922.
  • Ickovics JR, Park CL. “Paradigm shift: Why a focus on health is important” Journal of Social Issues. 1998;54:237–244.
  • Institute of Medicine. Adolescent Decision Making. Washington DC: National Academy Press; 1999a.
  • Institute of Medicine. Toward Environmental Justice. Washington DC: National Academy Press; 1999b.
  • Kagan J. Three Seductive Ideas. Cambridge MA: Harvard University Press; 1999.
  • Keyes CLM. “Social well-being” Social Psychology Quarterly. 1998;61:121–140.
  • Kiecolt-Glaser JK, Glaser R, Gravenstein S, Malarkey WB, Sheridan J. “Chronic stress alters the immune response to influenza virus vaccine in older adults.” Proceedings of the National Academy of Sciences of the United States of America. 1996;93:3043–3047. [PMC free article: PMC39758] [PubMed: 8610165]
  • Kiecolt-Glaser JK, Glaser R, Cacioppo JT, MacCallum RC, Snydersmith M, Cheongtag K, Malarkey WB. “Marital conflict in older adults: Endocrinological and immunological correlates” Psychosomatic Medicine. 1997;59:339–349. [PubMed: 9251151]
  • Kubzansky L, Kawachi I, Sparrow D. “Socioeconomic status and risk factor clustering in the Normative Aging Study: Any help from the concept of allostatic load?” Annals of Behavioral Medicine. 1999;21(4):330–338. [PubMed: 10721441]
  • Lipsey MW, Cordray DS. “Evaluation methods for social intervention” Annual Review of Psychology. 2000;51:345–375. [PubMed: 10751975]
  • Lundberg U. “Stress responses in low-status jobs and their relationship to health risks: Musculoskeletal disorders” Annals of the New York Academy of Sciences. 1999;896:162–172. [PubMed: 10681896]
  • Lupien SJ, DeLeon MJ, De Santi S, Convit A, Tarshish C, Nair NPV, Thakur M, McEwen BS, Hauger RL, Meaney MJ. “Cortisol levels during human aging predict hippocampal atrophy and memory deficits” Nature Neuroscience. 1998;1:69–73. [PubMed: 10195112]
  • Malmstrom M, Sundquist J, Johansson SE. “Neighborhood environment and self-reported health status: A multi-level analysis” American Journal of Public Health. 1999;89(8):1181–1186. [PMC free article: PMC1508675] [PubMed: 10432903]
  • Manton KG, Woodbury M, Tolley D. Statistical Applications Using Fuzzy Sets. New York: John Wiley and Sons; 1994.
  • Markowe HLJ, Marmot MG, Shipley MJ, Bulpitt CJ, Meade TW, Stirling Y, Vickers MV, Sernmence A. “Fibrinogen: A possible link between social class and coronary heart disease” British Medical Journal. 1985;291:1312–1314. [PMC free article: PMC1417417] [PubMed: 3933646]
  • Marcus AC. “New directions for risk communication research” Journal of the National Cancer Institute Monographs. 1999;25:35–42. [PubMed: 10854455]
  • Marmot MG, Smith GD, Stansfield S, Patel C, North F, Head J, White I, Brunner E, Feeney A. “Health inequalities among British civil servants: The Whitehall II study” Lancet. 1991;337(8754):1387–1393. [PubMed: 1674771]
  • Marmot MG, Bosma H, Hemingway H, Brunner E, Stansfield S. “Contribution of job control and other risk factors to social variations in coronary heart disease incidence” Lancet. 1997;350:235–239. [PubMed: 9242799]
  • McEwen BS. “Protective and damaging effects of stress mediators” New England Journal of Medicine. 1998;338(3):171–179. [PubMed: 9428819]
  • McEwen BS, Seeman T. “Protective and damaging effects of mediators of stress: Elaborating and testing the concepts of allostasis and allostatic load” Annals of New York Academy of Sciences. 1999;896:30–47. [PubMed: 10681886]
  • Power C, Matthews S. “Accumulation of health risks across social groups in a national longitudinal study” In: Strickland SS, Shetty PS, editors. Human Biology and Social Inequality. Cambridge U.K.: Cambridge University Press; 1998. pp. 33–56.
  • Raudenbush SW, Sampson RJ. “Ecometrics: Toward a science of assessing ecological settings, with application to the systematic social observation of neighborhoods” Sociological Methodology. 1999;29:1–41.
  • Rimer BK. “Putting the ‘informed' in informed consent about mammography” Journal of the National Cancer Institute. 1995;87(10):703–704. [PubMed: 7563141]
  • Ryff CD, Singer BH. “The contours of positive human health” Psychological Inquiry. 1998;9:1–28.
  • Ryff CD, Singer BH. “Interpersonal flourishing: A positive health agenda for the new millennium” Personality and Social Psychology Review. 2000;4:30–44.
  • Ryff CD, Singer BH. “The role of emotion on pathways to positive health” In: Davidson RJ, Goldsmith HH, Scherer K, editors. Handbook of Affective Science. New York: Oxford University Press; In press.
  • Ryff CD, Singer BH, Love GD, Essex MJ. “Resilience in adulthood and later life: Defining features and dynamic processes” In: Lomranz J, editor. Handbook of Aging and Mental Health: An Integrative Approach. New York: Plenum; 1998. pp. 69–96.
  • Sampson RJ, Raudenbush JD, Earls F. “Neighborhoods and violent crime: A multilevel study of collective efficacy” Science. 1997;227(5328):918–924. [PubMed: 9252316]
  • Sampson RJ, Morenoff JD, Earls F. “Beyond social capital: Spatial dynamics of collective efficacy for children” American Sociological Review. 1999;64:633–660.
  • Seeman TE, McEwen BS. “Impact of social environment characteristics on neuroendocrine regulation” Psychosomatic Medicine. 1996;58:459–471. [PubMed: 8902897]
  • Seeman TE, Singer BH, Rowe JW, Horwitz RI, McEwen BS. “The price of adaptation—allostatic load and its health consequences” Archives of Internal Medicine. 1997;157:2259–2268. [PubMed: 9343003]
  • Seeman TE, Singer BH, Wilkinson RW, McEwen BS. “Gender differences in age-related changes in HPA axis reactivity” Psychoneuroendocrinology. In press. [PubMed: 11166486]
  • Singer B. “Self-selection and performance-based ratings: A case study in program evaluation” In: Wainer H, editor. Drawing Inferences from Self-Selected Samples. New York: Springer-Verlag; 1986. pp. 29–49.
  • Singer B, Ryff CD. “Hierarchies of life histories and associated health risks” Annals of the New York Academy of Sciences. 1999;896:96–115. [PubMed: 10681891]
  • Singer B, Ryff CD, Carr D, Magee W. “Linking life histories and mental health: A person-centered strategy” In: Raftery A, editor. Sociological Methodology. Washington DC: American Sociological Association; 1998. pp. 1–51.
  • Sutton SK, Davidson RJ. “A biological substrate of the behavioral approach and inhibitor systems” Psychological Science. 1997;8:204–210.
  • Tomarken AJ, Davidson RJ. “Frontal brain activation in repressors” Journal of Abnormal Psychology. 1994;103:339–349. [PubMed: 8040503]
  • Tomarken AJ, Davidson RJ, Wheeler RE, Doss RC. “Individual differences in anterior brain asymmetry and fundamental dimensions of emotion” Journal of Personality and Social Psychology. 1992;76:830–838. [PubMed: 1583591]
  • Uchino BN, Cacioppo JT, Kiecolt-Glaser JK. “The relationship between social support and physiological processes: A review with emphasis on underlying mechanisms and implications for health” Psychological Bulletin. 1996;119:488–531. [PubMed: 8668748]
  • U.S. DHHS (U.S. Department of Health and Human Services) Vital and Health Statistics: Plan and Operation of the Third National Health and Nutrition Examination Survey, 1988-1994. Washington DC: U.S. Department of Health and Human Services; 1994.
  • Uvnas-Moberg K. “Physiological and endocrine effects of social contact” Annals of the New York Academy of Sciences. 1997;807:146–163. [PubMed: 9071348]
  • Uvnas-Moberg K. “Oxytocin may mediate the benefits of positive social interaction and emotions.” Psychoneuroendocrinology. 1998;23:819–835. [PubMed: 9924739]
  • Wadsworth MEJ, Kuh DJL. “Childhood influences on adult health: A review of recent work from the 1946 National Birth Cohort Study, the MRC national survey of health and development” Pediatric Perinatal Epidemiology. 1997;11:2–20. [PubMed: 9018723]
  • Warner KE. “The need for, and value of a multi-level approach to disease prevention: The case of tobacco control” In: Smedley BD, Syme SL, editors. Promoting Health: Intervention Strategies from Social and Behavioral Research. Washington DC: National Academy Press; 2000. pp. 326–352.
  • Watson M. The Prevention and Control of Malaria in the Federated Malay States: A Record of Twenty Years of Progress. London: John Murray; 1921.
  • Watson M. African Highway: The Battle for Health in Central Africa. London: John Murray; 1953.
  • Woloshin S, Schwartz LM. “How can we help people make sense of medical data?” Effective Clinical Practice. 1999;2(4):176–83. [PubMed: 10539543]
  • Yen IH, Kaplan GA. “Neighborhood social environment and risk of death: Multilevel evidence from the Alameda County study” American Journal of Epidemiology. 1999;149(10):898–907. [PubMed: 10342798]
  • Zhang H, Singer B. Recursive Partitioning in the Health Sciences. New York: SpringerVerlag; 1999.

The data are available electronically: Alzheimer's: http://cds​



The data are available electronically: Alzheimer's: http://cds​

Copyright © 2001, National Academy of Sciences.
Bookshelf ID: NBK43789
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.1M)

Related information

  • PMC
    PubMed Central citations
  • PubMed
    Links to pubmed

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...