We were unable to clearly determine the net benefit of using multifactorial assessment and management interventions in older adults because of the heterogeneity in older adults studied, the broad spectrum and multifactorial nature of interventions evaluated, and the suboptimal and inconsistent use of outcomes measured. Our best attempt at synthesizing findings across this very heterogeneous body of evidence suggests a small, statistically significant benefit in functional ability (Table 6). This small effect on ADL and IADL within a subset of the included trials is difficult to interpret, given the 1) choice of ADL or IADL instruments that may not be responsive to detecting clinically significant changes in functioning in community-dwelling older adults, 2) likely heterogeneity of treatment effects for these interventions and inability to understand the heterogeneity of populations studied (due to inconsistent and inadequate reporting of risk factors or measures to assess patient risk for functional decline, and lack of important subgroup explorations at the individual trial level), and 3) inconsistent reporting of a set(s) of outcomes resulting in different bodies of literature (and therefore different interventions in different populations) being described with ADL and IADL outcomes versus other outcomes (e.g., HRQL, hospitalizations, or institutionalizations). Attempts to pool results by similar-risk populations and types of interventions significantly limited the number of studies in these analyses, without substantially affecting the magnitude or statistical heterogeneity of pooled results. However, variation in (and lack of) measurement or reporting of important population and intervention characteristics across this body of literature, as well as differences and inconsistencies in outcomes used across studies, limited truly meaningful subgroup analyses.

Table 6. Summary of Evidence.

Table 6

Summary of Evidence.

Our review has limited overlap in included studies as compared with other existing systematic reviews of multifactorial assessment and management interventions in geriatric populations 4,7,8,32,7577 (Appendix D). These existing reviews all had slightly different focuses (i.e., preventive home visits, primary care–relevant interventions, comprehensive geriatric assessments, complex geriatric interventions) and used different inclusion criteria, as well as differing methodological approaches. Even among reviews with more similar scope, inexact and inconsistent terminology describing complex interventions and lack of a unified theory or model describing interventions makes locating and applying inclusion criteria to identify cohesive bodies of literature challenging. The most similar, and most current, existing review by Beswick and colleagues included 89 trials focused on a broad set of complex geriatric interventions that evaluated “interdisciplinary teamwork for health and social problems.” 4 Despite differences in included studies and methodologies to pool results (e.g., our use of between-group change in score and Beswick’s use of measures at followup only, our more conservative pooling of results across different lengths of followup), both reviews found a similarly modest degree of benefit in preventing functional decline, and a reduction in hospitalizations in a different subset of articles reporting this outcome. Neither review found any evidence for mortality benefit. The Beswick review also concluded that this benefit in preventing functional decline was primarily accrued in a subset of interventions in general-risk older adults (as opposed to the frail elderly); however, our review cannot confirm or refute this finding due to differences in included studies for our pooled analyses and possibly more limited power to detect subgroup differences, because of more conservative pooling of outcomes. Another recent, more focused review on preventive home visits in community-dwelling older adults by Huss and colleagues had very limited overlap in included studies.7 Both reviews found a modest benefit for functional outcomes; however, the Huss review found benefit only for interventions that included clinical examination in the initial assessment. Again, our inability to detect this difference is likely due to the difference in included studies for our pooled analyses on preventive home visits, as the Huss reviewers were able to quantitatively combine more studies by obtaining nonpublished data on dichotomous outcomes from individual study authors and conducted less conservative pooling across a range of lengths of followup.

We understand that there is a natural tension between the goals of primary research that is interested in asking a specific clinical question and that of secondary research that is intended to inform health policy decisionmaking by synthesizing evidence broadly across primary research. However, we believe that the methodological challenges encountered provide insight into important considerations for future research to improve care for older adults to prevent functional decline (Table 7). First, consistently and completely ascertaining study population baseline risk is extremely important. The considerable variability in the natural history of functional decline in older adults introduces random error and reduces the likelihood of finding a consistent group effect.94,95 More complete and consistent ascertainment of population functioning and risk for decline in functioning would allow investigators to examine the effectiveness of interventions in subgroups that are at higher risk for functional decline and disability, as well as considering intervention effects on subgroups with differing functional status trajectories.

Table 7. Considerations for Future Research on the Health and Functional Decline of Older Adults.

Table 7

Considerations for Future Research on the Health and Functional Decline of Older Adults.

Second, complex interventions are hard to characterize, partly due to incomplete and inconsistent reporting. When possible, it is important to both enhance the consistency and reproducibility of interventions by improved reporting of important intervention details.96,97 Trials evaluating complex interventions should capture important details about, for example, conditions/targets, mode of delivery, frequency, contact time, duration, and personnel involved for both assessment and management. More research is needed to test consistent models, or intervention components, across a series of trials, in similar populations for reproducibility of effectiveness, as well as across different populations and settings.

Third, there is considerable variability in reported trial outcomes, as well as methodological challenges around outcome measurement. For measures of function, we focused on self-reported measures (i.e., ADL and IADL). However, trials used many different ADL and IADL measures that were often validated in very different populations and occasionally not clearly identified. There is a strong need for consensus and standardization in measuring global functioning and functional decline in community-dwelling older adults. Other evidence suggests this need applies to hospitalized older adults as well.98 Authors using ADL, IADL, and HRQL instruments need to report the name of the instrument, its intended purpose, and its appropriateness for intended use (e.g., document the instrument’s validity and sensitivity to change in the study population). In evaluating outcomes, it is important to report baseline and followup values, not just change in scores, to allow for best interpretability of trials. Selective reporting of subcomponents of HRQL measures should only be done if these subcomponents are specified a priori as primary or secondary outcomes. Dichotomous outcomes are perhaps more clinically relevant, certainly more clinically intuitive, than continuous outcomes, but this needs to be based on clinically meaningful and consistent standards to allow for comparison across trials.

Future research would greatly benefit from using a focused and consistent set of agreed-upon measures, or core clinical outcomes, within a given population that 1) adequately capture clinically meaningful change in functioning with respect to a certain population (e.g., valid and responsive measures for functional ability may differ for community-dwelling versus institutionalized older adults), 2) capture multiple dimensions of health (e.g., HRQL), and 3) include common health care utilization measures (e.g., emergency department visits, hospitalizations, institutionalizations) that may be proxies for health outcomes. Of course, the choice of individual trial outcomes must be guided in part by the trial’s population, intervention, and sample size. Some effort toward using a set of core clinical outcomes that are both responsive and multidimensional would greatly improve the ability of evidence synthesis to inform medical decisionmaking. Standards for these types of research should consider whether measurement of self-reported functioning should be enhanced by additional use of a set of well-validated performance-based measures. Although expert consensus on trial design aimed at preventing or slowing functional decline has recommended limiting outcome measures to “hard” measures of disability, such as measures of ADL,99 more recent evidence supports the use of global performance-based measures. Gait speed, for example, has been shown to be associated with mortality.100

Most clinicians and researchers who care for older adults believe that we can only truly optimize the care of all older adults by affecting multiple aspects of health, from multiple perspectives/disciplines, over a span of aging that includes many possible functional trajectories. It is imperative that valid, consistent, and targeted trials be performed to clarify and solidify the appropriate health interventions for this growing population.