Individualized interventions and precision health: Lessons learned from a systematic review and implications for analytics-driven geriatric research

Abstract

Older adults are characterized by profound clinical heterogeneity. When designing and delivering interventions, there exist multiple approaches to account for heterogeneity. We present the results of a systematic review of data-driven, personalized interventions in older adults, which serves as a use case to distinguish the conceptual and methodologic differences between individualized intervention delivery and precision health-derived interventions. We define individualized interventions as those where all participants received the same parent intervention, modified on a case-by-case basis and using an evidence-based protocol, supplemented by clinical judgment as appropriate, while precision health-derived interventions are those that tailor care to individuals whereby the strategy for how to tailor care was determined through data-driven, precision health analytics. We discuss how their integration may offer new opportunities for analytics-based geriatric medicine that accommodates individual heterogeneity but allows for more flexible and resource-efficient population-level scaling.

“We recommend widespread implementation and dissemination of efficacious individualized intervention strategies, alongside efforts to design and execute precison medical trials with older adults to learn optimal interventions for individual and subgroups of older adult patients.”

INTRODUCTION

Compared to other demographic groups, older adults are characterized by profound clinical heterogeneity, including large interindividual differences in biopsychosocial needs and increased multimorbidity that accumulates over the lifespan.^1–3 As this population continues to grow,⁴ guidelines for care emphasize the importance of tailoring care to individual-level preferences and priorities, clinical needs, and health trajectories across multiple clinical contexts.^2,5

The significance of heterogeneity, from a clinical and health promotion perspective, is that the same intervention may not work well for everyone within a heterogeneous population. Unfortunately, older adults are often excluded from efficacy-based clinical trials, which makes it even more difficult to characterize the clinical implications of age-based heterogeneity, as well as the impacts of other inter-individual differences on the effects of interventions within this age group. Moreover, traditional study designs that compare singular or fixed interventions can preclude efforts to better address interindividual differences across older adults, particularly when different patients and their unique medical or social needs impact their response to the intervention.

The idea of personalized care is not novel. As a result, there exist multiple approaches to designing and delivering interventions that account for heterogeneity across a population of interest. One such approach is precision health, the goal of which is to leverage individual-level data to improve treatment decisions so that the right treatment is administered to the right patient at the right time.^6,7 Note that we use “precision health,” rather than the narrower term “precision medicine,” to capture a spectrum of clinical, preventative, and general health-promotion interventions. There has been growing recognition that precision health approaches are particularly well-suited to address the clinical and biopsychosocial heterogeneity of the growing population of older adults.^8–12 Formalized precision health approaches integrate methodological advances in adaptive trial designs, biostatistics, and machine learning analytics with rich patient databases to discover individual responses to interventions (including non-responses) and ultimately yield clinical decision support for personalized treatment decisions.⁶ Examples of precision health methodology are shown in Figure 1.

FIGURE 1.

Examples of precision health methodology.

Despite its potential, we noted that the evidence base for data-driven precision health interventions in older adults remains scattered. Moreover, there exists a great deal of vagueness surrounding the terminology of what defines “precision health.” Our goal in this article is to address both gaps, which is a critical step towards realizing the potential of precision medicine in geriatric research and care. We first describe a systematic review of the literature that our research team undertook to summarize the evidence base of data-driven precision health interventions among older adults. This review highlights a “use case” to illustrate the confusion surrounding what comprises data-driven precision health in the context of geriatrics research. To offer clarity, we propose a novel framework to distinguish the family of research we aimed to capture—precision health—with the family of studies we captured—individualized interventions. Our goal in providing this framework is to highlight existing evidence while also elucidating opportunities for future geriatrics research to incorporate precision health analytics for data-driven decision support for individual treatment decisions.

Definitions

Our team defines precision health-derived interventions as those that tailor care to individuals, whereby the strategy for how to tailor care was determined through data-driven, precision health analytics. As defined in Table 1, one example of such a statistically-oriented precision medicine strategy is the estimation of optimal dynamic treatment regimes. Learned from data via statistical and machine learning, these rules guide the selection of an intervention for each patient within a population based on their individual health status such that patient outcomes are optimized.^6,7,15,20,21 In contrast, our team defines individualized interventions as those where all participants received the same parent intervention, modified on a case-by-case basis and using an evidence-based protocol, supplemented by clinical judgment as appropriate. Both individualized interventions and precision health-derived interventions aim to deliver personalized treatment strategies to different individuals within the same population. However, they fundamentally differ in terms of intervention structure, the method and forms of personalization, and how individual level data are used. Table 1 outlines the similarities and differences. Note that we use the term personalization to describe the result of tailoring for either approach.

TABLE 1.

Key differences between individualized intervention delivery and precision health

	Individualized interventions	Precision health
Intervention structures	One “parent” intervention	One or more different interventions, combinations thereof, and adaptive sequences
How individual data are used to generate a personalized treatment strategy	Individual level data are matched against a protocol, on a case-by-case basis, with clinical judgment to supplement evidence-based thresholds, to dictate modifications to the parent intervention	Population level datasets used to discover optimal tailoring strategies or treatment rules to assign individuals to optimal treatments, which are then deployed at the bedside using an individual’s data to get an optimal, data-driven treatment recommendation.
Forms of personalization	Adaptations to intervention for personal relevance or feasibility	Selection of intervention, sequencing, or stopping of treatment
Clinical indications	Consistent clinical indications for the parent intervention across the population of interest	Potentially heterogeneous indications for one or more interventions across the population of interest
Needs for scaling and dissemination	More care providers to generate resources to adapt interventions to individuals	Validation of personalization strategies (e.g., algorithms) in different populations and datasets

THE SYSTEMATIC REVIEW OF THE LITERATURE: A USE CASE FOR THE NEED FOR CLARITY

Although the potential impact of the paradigm of precision health for older adult populations has been discussed,^10–12 to our knowledge, the evidence base for such interventions has not been synthesized. Our initial objective was thus to perform a systematic review to summarize studies of data-driven precision health interventions in older adults age ≥65 years, including the clinical contexts to which they had been applied, key outcomes and main effects, and the method of data-driven tailoring or individualization. We aimed to address how precision health interventions impacted mortality, morbidity, and quality of life, in addition to disease-specific outcomes, among older adult populations. By systematic identification and assessment, the nature and impact of these interventions, our aim was to reveal prime applications of and settings for precision health to improve key clinical and patient-oriented outcomes among this growing population.

We systematically evaluated the evidence base for randomized trials that evaluated the efficacy of data-driven precision health interventions among older adults. Specifically, we defined precision health as studies testing interventions to discover optimal dynamic treatment regimes⁶ or interventions with adaptive designs or tailored components delivered as a function of patient features or response patterns. The full study protocol is included as Text S1.

Study methods

Information criteria

A medical librarian developed a comprehensive search strategy to identify eligible studies; all searches were executed on May 28, 2021. The search strategy was developed in an iterative manner to refine search terms and results utilizing a combination of Medical Subject Headings (MeSH) and text words in PubMed (U.S. National Library of Medicine, National Institutes of Health). A report on terminology for personalized medicine developed by PACEOMICS²⁵ was consulted to develop the precision medicine portion of the search. In addition, a validated search hedge developed by the Canadian Journal of Health Technologies (CADTH) to identify randomized controlled trials was modified for this review.²⁶ The PubMed search was tested against a validation set of previously identified eligible studies and peer reviewed by another medical librarian. Once finalized, the PubMed search strategy was translated to Scopus (Elsevier), Web of Science Core Collection (Clarivate Analytics), EMBASE (Elsevier), Cumulative Index of Nursing and Allied Health Literature (CINAHL) Plus with Full-Text (EBSCOhost), PsycINFO (EBSCOhost), Cochrane Library (Wiley), and ACM Full-Text Collection (Association for Computing Machinery). Search results were limited by publication date (2010 to present) and language (English). In addition, we also completed forward and backward citation searching (using Scopus or Web of Science) of included studies identified from bibliographic databases to identify others potentially not retrieved by the search (November 2, 2021–January 2022). Due to our inclusion criteria specifying that publications must be peer-reviewed, we omitted searching gray literature, conference abstracts, and trial registry websites.

Selection criteria

We used the PICOS (population, intervention, comparator, outcome, study type) framework to develop and refine our eligibility criteria. Studies were included if they met all of the following inclusion criteria: (1) Population: studies with participants with a mean age 65+ years or had no participants younger than 60 years in inpatient and ambulatory settings; (2) Intervention: any therapeutic, technologic, or behavioral precision health interventions; (3) Comparator: a control (all types considered) must be present as part of the study design; (4) Outcome: primary outcomes included those generalizable across different diseases and interventions: morbidity, mortality, and quality of life. Secondary outcomes included disease-specific outcomes (i.e., study-specific endpoints relevant to the intervention); (5) Study type: peer-reviewed publications of randomized controlled trials (RCTs) or sequential multiple assignment randomized trials (SMARTs) in English language published since 2010.

We excluded studies with participants younger than 60 years old or with an age range over 60 years old with a mean age of less than 65 years. In addition, we excluded studies that did not meet a pre-determined definition of precision health as studies designed to either (1) discover optimal treatment regimens; or (2) interventions with data-driven, adaptive, or tailored components delivered as a function of patient features or response patterns to assess the efficacy of treatment regimes. Data-driven was defined as the use of data and research methods to derive reproducible, generalizable, pre-specified treatment protocols (i.e., rules, regimes), but did not include other personalization methods, which may be data-informed but are ultimately driven by clinical judgment or other methods that may vary from provider-to-provider or across clinical settings. Thus, the remaining exclusion criteria included interventions in which the criteria by which personalization or individualization is determined is not derived in a systematic, data-driven way (qualitative data were included); pharmacogenetic or pharmacogenomic studies (defined studies which aim to identify genetic variants that influence drug effects, typically through alterations in pharmacokinetics or pharmacodynamics²⁷), all study types other than RCTs or sequential multiple assignment randomized trials, including response-adaptive clinical trial designs and post-hoc analyses; abstracts and clinical trial registries with no associated full-text publication; and articles published prior to 2010 or in languages other than English. From a scientific perspective, pharmacogenetics/genomics studies comprise a distinct field of biomedical research focused on characterizing the fixed nature of pharmacogenetic (variant–drug) associations; evidence has been summarized previously.^27,28 While pharmacogenomics and precision health are related in their overarching goals to individualize therapy, the latter extends beyond molecular markers and deterministic associations to study how the delivery of pharmacologic and non-pharmacologic interventions can be adapted in response to broad patient features, as well as changes in those features or responses over time.

Selection process

All citations were exported from the individual bibliographic databases, imported into Endnote, deduplicated, and imported into Covidence for screening. Each reviewer (5 total) conducted pilot screening for quality assurance purposes by manually conducting a title/abstract review of 200 citations, for which concordance was required to exceed 80%. Each set of reviewers conducted a test review for quality assurance purposes by manually conducting a title/abstract review of 200 citations, for which concordance was required to exceed 80%. Discrepancies between reviewers were adjudicated by the authors ARK and NBLF, an approach previously used.^29,30 In the first phase of screening, a total of five reviewers, in sets of two, independently screened titles and abstracts of citations for inclusion based on the pre-defined eligibility criteria. Then, a second round of screening was performed by two independent reviewers assessing the full-text articles. All conflicts in the title/abstract and full-text screening stages were resolved by a third reviewer.

Data extraction

A total of three blinded, independent reviewers extracted data in sets of two from included studies using Microsoft Word^®. Data from both files were collated and conflicts were resolved by authors ARK and NBLF.

Study risk of bias assessment

The Cochrane Risk of Bias 2.0 (RoB 2) or RoB 2 for Cluster-Randomized Trials tools were used to evaluate all included studies for bias. For each trial, we assessed the risk of bias (“low risk,” “some concerns,” or “high risk” of bias) in the overall effect of assignment to the intervention primary and secondary outcomes. Risk of bias assessments were based on the trial protocols and flowcharts following the Consolidated Standards of Reporting Trials together with information supplied by the investigators of each trial on the methods used to randomize participants and conceal randomized allocation, the methods used to ensure that patients received their allocated intervention and the extent of deviations from the assigned intervention, the methods used to measure outcomes, the extent of missing outcome data, and how the reportable results were selected. A total of three independent, blinded reviewers, in sets of two, assessed bias using the appropriate tool for the study, and conflicts were resolved through discussion by authors ARK and NBLF.

Outcomes of the systematic review

A total of 11,517 citations were identified. After duplicates (n = 4695) were removed, 6822 unique citations underwent title and abstract screening. There were 6441 determined to not meet our pre-defined eligibility criteria on title and abstract screening. 381 studies were screened for the full-text review, from which 353 were excluded. In total, 27 studies were identified for inclusion based on the systematic search, with one article identified by hand-searching. The PRISMA flow diagram of the study selection process, including exclusion reasons identified in the full-text stage, can be found in Figure 2.

FIGURE 2.

PRISMA 2020 flow diagram for new systematic reviews which included searches of databases, registers, and other sources.

In total, 28 studies were included in this review and are summarized in Text S1. Briefly, with regards to bias, the subjective methodological quality of all included studies was considered low to some concerns.

Methodological problems in the included studies consisted of limited information on randomization, self-reported outcome measures, early termination, low adherence to intervention components, and other biases secondary to dietary tracking and loss of follow-up. A wide array of international locations was represented, with most studies included conducted in European countries. Two studies were cluster randomized trials, and the rest were randomized control trials. The studies were a mix of hospital-based, clinic-based, and home-based interventions. The number of older adult participants ranged from 42 to 5451. The mean age spanned approximately 65–84 years. The aims of the included studies spanned a range of clinical domains and disease contexts, with findings that in aggregate suggested efficacy for individualized interventions over control conditions.

As described above, we set out to capture studies with protocols that tailor treatments to individuals, where the tailoring strategy was derived from the application of precision health analytics to individual-level data.^7,31 We expected to identify study interventions based on SMART designs with adaptive sequences modified based on interim response patterns, studies focused on the discovery or evaluation of individualized treatment rules such as optimal dynamic treatment regimes, and interventions where components were modified according to predicted treatment response or baseline phenotypes.⁶ The review instead produced a group of studies evaluating the effect interventions in which all participants received largely the same parent intervention, with minor modifications or permutations of the intervention to individualize it for participants. For example, included studies took individual characteristics into account for personalization or individualization of interventions using a range of methods, including assessing and addressing individual risk factors,³² computational algorithms,³³ operationalization of clinical cut-points,³⁴ and integration of qualitative data.³⁵

IMPLICATIONS OF OUR FINDINGS

In sum, we found that the studies in our systematic review largely represented individualized interventions, rather than precision health. We used the discrepancy between our targeted studies and those we captured as a basis for shaping key definitions; as above, we define individualized interventions as those where all participants received the same parent intervention, modified on a case-by-case basis and using an evidence-based protocol, supplemented by clinical judgment as appropriate, while precision health-derived interventions are those that tailor care to individuals, whereby the strategy for how to tailor care was determined through data-driven, precision health analytics, such as optimal dynamic treatment regimes, or other relevant precision health statistical methodology (Table 1). Both individualized interventions and precision health-derived interventions hinge upon the fundamental assumption that patients with heterogeneity warrant personalized treatment strategies. Where individualized interventions and precision health-derived interventions differ is in the intervention structure, the method and forms of personalization, and how individual level data are used.

While the distinction between what we define as precision health interventions, and what we refer to as individualized intervention delivery may seem subtle, we believe these approaches to tailoring geriatric care are more conceptually and methodologically distinct than they appear These terms are used interchangeably, and as a result, this only perpetuates the existing definitional challenges in the literature. Furthermore, their integration may offer new opportunities to leverage the strengths of each approach, in parallel or in series, to accommodate individual heterogeneity among older adult patients, and with great potential for scaling and dissemination. We discuss each approach in detail, below.

Individualized intervention delivery: Parent interventions with adaptations to ensure feasibility and relevance for individuals

The studies captured in our review reflected well-conducted, evidence-grounded interventions based on the underlying premise that older adults are a heterogeneous and complex group. We identified the following categories of interventions within the review: (1) personalized self-management and educational interventions, spanning nutrition, physical activity, and action or care plans; (2) studies exploring computational or computer guided tailoring of interventions (including feedback from computer games or tailored m-health messaging); (3) risk assessment and personalized action plans (including geriatric assessments and multi-risk fall assessments); (4) multidisciplinary care coordination (inpatient and outpatient providers); and (5) biomarker-guided dosefinding methods. What defines these individualized intervention studies, as a group, is that all participants received the same parent intervention, modified on a case-by-case basis and using an evidence-based protocol, supplemented by clinical judgment as appropriate. This approach largely mirrors existing best practices for clinical care, whereby treatment strategies known to be effective at a population level are adapted to individual circumstances to ensure their relevance and pragmatism for that person; meanwhile, the indications for and intended effects of the intervention remain consistent across individuals and over time.

Precision health interventions: Addressing interindividual needs through empirical learning and data-driven treatment selection

In precision health-derived interventions, data from individuals is used in two different ways: first, individual level data from the population is used to learn how to optimally tailor treatments through the use of precision health data analytics; and second, once the strategy for personalizing treatments is learned, it can be deployed in practice so that when a patient presents in the clinic, their individual characteristics (data) can be put into the precision health-derived treatment strategy to get an optimal, data-driven treatment recommendation.^6,7,15 Moreover, precision health provides tools to investigate sequences of care and design trials for optimal learning and confirming the benefit of precision health-derived interventions.

Examples of precision health interventions are more limited than those of individualized interventions, which in part reflects the relatively recent advancements in complex trial designs and statistical methodology to make empirically based optimal treatment recommendations. Examples of SMART designs that have yielded or are intended to yield insights into treatment approaches for heterogeneous population have been applied in behavioral health,³⁶ plastic surgery,³⁷ and weight management,^38,39 among others. A review of clinicaltrials.gov shows a series of other SMARTs underway, including those focused on psychiatric diseases (NCT01588457), back pain (NCT05396014), and cancer pain (NCT02791646). Precision health-based platforms have also been developed to characterize subtypes of asthma and their responses to various interventions.^40,41 Particularly relevant to older adult patients, individualized treatment rules have been applied to determine optimal exercise therapies⁴² and multicomponent treatment regimes for patients with osteoarthritis.^43,44 At this time, our team is also actively developing new studies, which will leverage SMART designs and precision health analytics to individualize and optimize health promotion among older adults; the goal of these studies is to generate data-driven, robust algorithms that can help providers to empirical approaches to tailor obesity and diabetes management within geriatric populations for their best possible clinical and patient-oriented outcomes, using information from similar patients and their responses to different interventions.

Reconciling individualized interventions with precision health

As part of their daily clinical practice, geriatric care providers engage in complex decision-making that requires the integration of preference-sensitive care with evidence-based care on a case-by-case basis.^7,31 Tailoring clinical interventions to individual patients is not a new paradigm, and our review further generated evidence that older adults require and ultimately benefit from individualized approaches that account for the heterogeneity between patients and their biopsychosocial needs.

Where precision health, as a field, may augment this finding, is by offering trial designs and statistical tools to learn how to tailor care beyond feasibility and relevance. This includes new algorithms that provide optimal treatment recommendations when multiple different and guidelines-based treatments are available, yet it is unclear which approach might be optimal for an individual patient. Precision health algorithms can also be used to learn how to sequence one or more treatments over time based on response and non-response patterns.⁶ Together, these recommendations, or treatment algorithms, can be integrated into EHR-based clinical decision support tools, augmenting provider judgment as to what, from a series of treatments or options, may be optimal for a given patient, based on their health status at that time.⁶ Yet, selecting an intervention is just the first step, and individualization of an optimal therapy, once selected, will still be necessary to ensure that it remains feasible and acceptable to individual patients.

Further opportunities for integration may become clear when new evidence for personalized care is translated and scaled to routine care settings. Individualized interventions and precision health offer different considerations for scalability, particularly with respect to the resources and time required to scale individualized interventions compared to the requisite infrastructure to scale efficacious precision health interventions. Assessing individual patients for tailored interventions within the individualization framework may necessitate added provider time, additional education, clinical support, and so forth, and thus represent a significant barrier to dissemination, particularly in busy clinical care settings where operational and cost constraints are greatest. In contrast, precision medicine analytics are designed to be deployed at scale (e.g., a precision health algorithm for optimal treatment recommendations can be re-trained with new data and implemented in diverse settings as clinical decision support), easing the burden of personalization and making personalization much more accessible even in resource limited settings. Further, if they can be successfully scaled, optimal treatment assignments aided by DTRs could reduce mismatches between patients and health-related treatments or interventions, preventing poorer outcomes or medical waste related to excessive, suboptimal, or ineffective therapies. These improvements may actually free up provider time and resources for individual-level tailoring to maximize the feasibility, relevance, and acceptability of optimal interventions or other modifications for more nuanced optimization, providing the opportunity for geriatric patients to benefit from empirical, data-driven insights in addition to clinical expertise and personalized action plans.

We envision several other future opportunities for precision health in geriatric research and care. Precision health decision support may complement other and more conventional decision support in the form of cost-effectiveness models by offering specific insight into new patterns by which to consider prioritizing individuals and subgroups for receipt of limited resources. Algorithms may also be extended to optimize adherence to evidence-based therapies in the context of unique medical and social needs. For instance, a major gap exists in integrating social determinants of health into therapeutic decision making. With such data, an optimal dynamic treatment regime has considerable potential to allocate critical system-level resources to bolster individual care, such as added provider time or education, clinical or peer support, or accessibility features for patients, which may lead to decreased healthcare utilization as a result of suboptimal therapy assignments or adherence.

FUTURE DIRECTIONS FOR GERIATRIC RESEARCH

At this time, the vision for analytics-based geriatric medicine is just emerging; its infancy is evidenced by the dearth of studies that have applied statistical precision medicine tools for geriatric patients or in clinical settings that are highly relevant to older adulthood. As a result of the efficacy of individualized interventions and the lack of true, data-driven, precision health interventions involving older adults, we offer the following recommendations moving forward:

1. Implementation and dissemination of individualized intervention delivery shown to be efficacious. It is critical that evidence-based interventions are made accessible to patients in routine care, as these patients are often highly complex and may stand to benefit even more from tailored interventions than their research participant counterparts.

2. Design and execution of pragmatic, precision medicine trial designs in geriatric research. While most traditional randomized controlled studies exclude older adults and limit participants with complex multimorbidity, precision health interventions can leverage this heterogeneity to discover and evaluate optimal, individualized treatment options, recommendations that are critically needed in geriatric medicine. Future research to utilize SMART designs with older adults can both design high-quality, adaptive interventions and learn optimal intervention recommendations for individuals, enriching geriatric science with maximally informative and scientifically valid data for estimating optimal treatments.⁶ SMARTs that are strategically designed as pragmatic trials (using the PRECIS-2 framework,⁴⁵ for example) may accelerate the process to translate new data into routine care settings.

3. Efforts to bridge personalized care with population health. When expanded from individual patients to the population level, tools based on precision health may offer a data-driven way to stratify populations according to estimated optimal treatment recommendations. Future research can use these analytics, with large and representative datasets, to understand and characterize geriatric subpopulations who are likely to benefit from a given therapy or intervention. These insights may help healthcare systems efficiently allocate and target interventions, particularly those that may be associated with increased cost or other investments at the provider- or healthcare system-level. Integrating data on social and medical needs may advance health equity and the accessibility of evidence-based interventions across the lifespan.

4. Assembly of transdisciplinary teams to develop and deploy rigorous and clinically-relevant analytics in geriatric care. Precision health is, inherently, a team science that requires deep technical expertise (e.g., study design, biostatistics, machine learning) alongside expertise in geriatric medicine, population health, implementation science, and health economics to generate sustainable interventions that improve quality of care. We advocate for forming collaborations across disciplines early in the research lifecycle to ensure that data collection, analysis, and implementation efforts may be efficiently iterated, accelerating the translation from analytic models to their applications in clinical and community systems. Translation of precision health into routine care is a relatively unexplored area and may include facilitating the training required for providers to effectively use AI-driven decision support tools in daily practice. Moreover, commitment and champions of AI-driven decision support will be imperative for cultivating the cultural shift for providers and patients engaging in the new paradigm of precision health decision support and ensuring high fidelity implementation to the tools.

The findings of our review process are of notable importance to the geriatrics research community to not only differentiate between individualized interventions and precision health interventions, but also to underscore how their integration has the potential to advance the science of aging and improve are for older adults through a new, analytic-driven approach to geriatrics care.

Supplementary Material

supinfo

Text S1. Supporting information.

Table S1. Reference number and key study information.

Table S2. Methodological Quality of the Included Studies–Cochrane Risk-of-Bias Tool.

Table S3. Design Characteristics of the Included Studies.

Table S4. Participant Characteristics.

Table S5. Study Aims and Results.

Table S6. Method of precision medicine or individualization.

Key points

• Both individualized interventions and precision health-derived interventions hinge upon the fundamental assumption that heterogeneous patients warrant personalized treatment strategies, but they fundamentally differ in terms of intervention structure, the method and forms of personalization, and how individual level data are used.

• We define individualized interventions as those where all participants received the same parent intervention, modified on a case-by-case basis and using an evidence-based protocol, supplemented by clinical judgment as appropriate, while precision health-derived interventions are those that tailor care to individuals whereby the strategy for how to tailor care was determined through data-driven, precision health analytics.

• We recommend widespread implementation and dissemination of efficacious individualized intervention strategies, alongside efforts to design and execute precision medicine trials with older adults to learn optimal interventions for individuals and subgroups of older adult patients.

Why does this paper matter?

Our systematic review underscores the value of individualized approaches to account for interindividual heterogeneity and biopsychosocial needs among older adult patients and highlights opportunities for precision health analytics to build upon these findings and augment the evidence base for personalized delivery of geriatric care.