4 Steps to 4Ms: A Navigation Guide for a Hospital-Based Composite Measure of 4Ms Care and the Implications for Outcomes Assessment

Abstract

The 4Ms Framework is the foundation of the Age-Friendly Health System (AFHS) movement. While the framework is based on standalone evidence for each M, there is limited evidence about the impact on outcomes when practiced as a set. A composite measure capturing adherence to the many care processes that comprise the 4Ms is a necessary but complex component of closing the evidence gap. We offer a navigation guide that addresses key considerations for developing a hospital-based composite measure of 4Ms care. The Institute for Healthcare Improvement operationalizes the 4Ms Framework as a minimum set of Assessment and Act On care processes. In developing a composite measure of inpatient 4Ms adherence, we offer a 4 step framework with associated discussion of considerations related to composite measure type (eg, continuous, dichotomous), and synchrony within and across the Ms containing these care processes. Using real-world electronic health record data capturing care process adherence in the 4Ms implementation at a large academic hospital, we illustrate the considerations, and report the implications for sample size and composite measure scoring. We also present our selected composite measure—a dichotomous measure delineating 4Ms care when all encounter-level processes (those needing to be done only once during the hospital encounter) are followed and all day- and shift-level processes are followed for at least 50% of hospital days. While there is no single, standard approach to create a 4Ms composite at this early stage of the AFHS movement, as organizations develop their measure(s), our guide and the considerations we suggest should serve to inform this process and support progress toward meaningful measurement.

Introduction

The 4Ms Framework is the foundation of the Age-Friendly Health System (AFHS) movement and is designed to ensure that systematically-delivered care provided to older adults is concordant with their unique needs and preferences. ¹ Despite broad adoption of the 4Ms Framework (What Matters, Medication, Mentation, and Mobility) across more than 3000 care settings, ² few studies have measured adherence to 4Ms care processes as a set during the hospital encounter and then assessed the relationship between adherence and patient outcomes.^{3 -5} This evidence gap is not due to lack of interest or opportunity; in fact, many health systems commit considerable resources to implementing the 4Ms Framework and have a vested interest in understanding the resulting impact. Instead, limited evidence is due, at least in part, to the challenge of defining a composite measure—that is, a single aggregate metric—of adherence to the many care processes that constitute the practice and delivery of all 4Ms as a set. While the principles of 4Ms care were drawn from an evidence-base associated with each M as a standalone process, such a composite measure is necessary to answer the foundational questions of whether, and to what degree, adherence to all 4Ms together results in improved outcomes for older adults.

This paper begins with the assertion that 4Ms stakeholders (eg, AFHS researchers, implementers, clinical operations leaders, policy makers, and patients and families) need a precise measure of adherence to 4Ms care in order to understand the impact of such care on patient outcomes, as described in detail in Burke et al’s ⁶ “What matters when it comes to measuring Age-Friendly Health System transformation.” Further, to be concordant with common hospital outcome measures (eg, length of stay, post-discharge healthcare utilization, patient satisfaction with hospital stay), a measure of 4Ms adherence in the hospital setting should also be at the encounter level (defined as the time spanning the entire course of the inpatient hospitalization for any given patient). Encounter-level measures also enable evaluations that adjust for patient and encounter characteristics (a necessary criterion, given that 4Ms implementations are rarely randomized). Encounter-level measures are not only conceptually appealing and clinically meaningful, but are also feasible with electronic health records (EHRs) that increasingly contain structured fields to capture many, if not all, of the care processes that constitute 4Ms care.^7,8 While these fields may not yet be configured and in use in all care settings that have implemented the 4Ms, a survey of a random sample of 797 acute care hospitals from 2018 to 2019 suggests that 64% have structured 4Ms documentation in at least 1 unit—a number that is almost certainly higher today. ⁹

Finally, beginning on January 1, 2025, hospitals participating in the Medicare Hospital Inpatient Quality Reporting Program are required to report on whether they meet all elements within 5 domains of an age-friendly hospital measure, including: (1) Eliciting patient healthcare goals, (2) Responsibly managing medications, (3) Implementing frailty screening and intervention, (4) Assessing social vulnerability, and (5) Designating leaders to ensure equitable delivery of age-friendly care. ¹⁰ While the self-attestation nature of the reporting, including Yes/No responses to each domain, does not specifically support assessment of the relationship between 4Ms care process adherence and outcomes, this is nevertheless an important step forward that will likely accelerate new measurement efforts.

Context, Motivation, and Setting

The 4Ms Framework was designed to be adaptable to different care settings, organizations, and populations. The Institute for Healthcare Improvement (IHI)’s Guide to Using the 4Ms in the Care of Older Adults^11,12 and the IHI’s 4Ms Care Description worksheet ¹³ define specific 4Ms care processes for inpatient and ambulatory settings. These care processes are divided into 2 categories: “Assessment” care processes (“Know about the 4Ms for each older adult in your care”) and “Act On” care processes (“Incorporating the 4Ms into care delivery”) for each M. The IHI’s 4Ms Care Description worksheet defines a minimum recommended set of 13 inpatient care processes against which 4Ms adherence can be measured. These include: (1) A minimum set of 4 Assessment processes, (2) A minimum frequency at which each Assessment should be done, and (3) A minimum set of 9 Act On processes. Notably, the Act Ons lack consistently specified minimum frequencies, but many can be inferred from their corresponding Assessment frequencies (eg, if a patient is assessed for delirium every 12 h, it would be clinically reasonable to also Act On those assessments every 12 h.) Table 1 presents the Assessment and Act On processes for the hospital setting as depicted by IHI, while Table 2 organizes the processes to illustrate the minimum set of 13.

Table 1.

IHI Specifications for 4Ms Care in the Hospital Setting.

Care Process	Medication	Mobility	Mentation	What matters
Assess	Check the medications you screen for regularly	Check the tool used to screen for mobility limitations	Check the tool used to screen for delirium	List the question(s) you ask to know and align care with each older adult’s specific outcome goals and care preferences.
	Minimum requirement: At least one of the boxes must be checked	Minimum requirement: One box must be checked	Minimum requirement: One box must be checked	(One or more What Matters question(s) must be listed. Questions cannot focus only on end-of-life forms.)
	☐ Benzodiazepines	☐ TUG	☐ UB-2
	☐ Opioids	☐ Get Up and Go	☐ CAM
	☐ Highly-anticholinergic medications (e.g. diphenhydramine)	☐ JH-HLM	☐ 3D-CAM
		☐ POMA	☐ CAM-ICU
	☐ All prescription and OTC sedatives and sleep medications	☐ Refer to physical therapy (PT)	☐ bCAM
		☐ Other	☐ Nu-DESC
	☐ Muscle relaxants		☐ Other
	☐ Tricyclic antidepressants
	☐ Antipsychotics
	☐ Mood stabilizers
	Minimum frequency: Once per encounter	Minimum frequency: Once per encounter	Minimum frequency: Every 12 h	Minimum frequency: Once per encounter
Act On	Minimum requirement: At least one box must be checked	Minimum requirement: Must check 1st box and at least one other box	Minimum requirement: 1st 5 boxes must be checked	Minimum requirement: 1st box must be checked
	☐ Deprescribe (includes both dose	☐ Ambulate 3 times a day	☐ Ensure sufficient oral hydration	☐ Align the care plan with the older
	reduction and medication	------------------------------------	☐ Orient older adult to time, place,	adult’s goals and preferences
	discontinuation)	☐ Out of bed or leave room for meals	and situation on every nursing shift	☐ Other
	☐ Pharmacy consult	☐ PT intervention (balance, gait,	☐ Ensure older adult has their
	☐ Other	strength, gait training, exercise	personal adaptive equipment (e.g.
		program)	glasses, hearing aids, dentures,
		☐ Avoid restraints	walkers)
		☐ Remove catheters and other	☐ Prevent sleep interruptions; use
		tethering devices	non-pharmacological
		☐ Avoid high-risk medications	interventions to support sleep
		☐ Other	☐ Avoid high-risk medications
	Minimum frequency: Not specified	Minimum frequency: Not specified	Minimum frequency: Orientation – every shift; Adaptive equipment – every shift; Sleep interruptions – every night. Others not specified.	Minimum frequency: Not specified

Table 2.

4Ms Care Processes, Frequencies and Number of Opportunities for 4Ms Care Process Adherence in a 5-Day/4-Night Hospital Encounter.

Care process number	UCSF process measure	4Ms care process	Minimum frequency (IHI Defined; Inferred)	Opportunities for 4Ms care for a 5 day (10 shift) encounter
Medication* Total: 2
1	1	Assess: Medication screening for high risk medications	Once per encounter	1
2	2	Act On: Deprescribe, or pharmacy consult	Once per encounter	1
Mobility Total: From 7 to 32
3	3	Assess: Mobility screening using an approved screening tool, or referral to physical therapy	Once per encounter	1
4	4	Act On: Ambulate three times a day PLUS one or more of the following 5 options:	Daily (all or none)	5
5	N/A in EHR	Out of bed or leave room for meals b	Daily	5
	5	Physical therapy	Variable	1 or more
	6	Avoid restraints	Daily	5
	7	Remove catheters and other tethering devices	Daily	5
	Concept measured below (UCSF Process Measure #12)	Avoid high risk medications a	Variable	1, 5, or 10
Mentation Total: From 45 to 54
6	8	Assess: Delirium screening using an approved delirium screening tool	Every shift	10
		Act On: All 5 of the following:
7	9	Oral hydration	Every shift	10
8	10	Orientation	Every shift	10
9	N/A in EHR	Access to personal adaptive equipment b	Every shift	10
10	11	Prevent sleep interruptions	Every night	4
11	12	Avoid high risk Medications a	Variable	1, 5 or 10
What Matters Total: 2
12	13	Assess: What Matters assessment	Once per encounter	1
13	14	Act On: Align What Matters with care plan	Once per encounter	1

a

“Avoid high risk medications” is implicit in the Medications M and is repeated in each of the Mobility and Mentation Ms.

b

Indicates that the care process was not documented in UCSF’s EHR and therefore no care process adherence measure was generated for the UCSF sample.

We sought to create a composite measure across the set of 4Ms care processes delivered for a given encounter that captures the intent of the framework. This required addressing many conceptual and practical considerations. Specifically, our team at the University of California San Francisco (UCSF)—a large urban academic medical center and an IHI Level 2 Recognized AFHS—examined these considerations in the context of assessing whether our 4Ms implementation was associated with improved outcomes for older adults. Our team included our AFHS lead (a geriatrician and implementer), a hospitalist and informatics researcher, a data scientist with substantial expertise in Epic Clarity (which served as the data source), and an informatics/health services researcher. To measure the 13 minimum recommended care processes, we used a set of 14 EHR-based measures of 4Ms care process adherence (Table 2). Prior work describes how we created operational definitions for these care process measures using EHR data. ⁷ These measures were captured for the 22 825 non-ICU inpatient encounters for those aged 65 and over that occurred between January 1, 2019 and December 31, 2021.

With this as our starting point, we created a 4 step navigation guide targeting health systems seeking to develop and implement a 4Ms adherence composite measure for use in assessing patient outcomes. Each step begins with a high-level description followed by a detailed description informed and illustrated by our approach at UCSF. Together, the guide accounts not only for the need to aggregate across heterogenous care processes that occur at different frequencies, but for the pragmatic considerations of frontline care delivery.

Navigation Guide

Step 1: Identify a Key Outcome of Interest and All Care Processes That Comprise 4Ms Care

High-level description

Each organization should identify a key outcome of interest and use the relationship of that outcome with measures of the 4Ms care processes to guide the remainder of their composite measure development. The IHI Guide suggests several hospital-encounter outcomes including readmissions, emergency department (ED) utilization, Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey questions, length of hospital stay, delirium rate, and collaboRATE (or similar tool to match goal concordant care). ¹¹ For facility of using EHR-based measures of adherence to 4Ms care, we recommend outcomes such as healthcare utilization including ED utilization and hospital readmission rates, but each organization should select based on their local needs. Many more candidate outcomes exist, and to explore them, a searchable and filterable collection of 4Ms outcomes—the 4Ms Evaluation Metrics Resource Library—has been made available on the AFHS Research Network website, https://tiny.ucsf.edu/4MsResourceLibrary.

Next, each organization should define the details of their 4Ms care processes, including the expected care process frequencies, within the framework of the minimum set of IHI recommended care processes. This determines the set of measures that will comprise the composite measure. Specifically, this work involves specifying the standard expectations for each care process and its intended frequency (eg, once per encounter, once per day, once per shift; see Table 2 for minimum frequencies). Together, these specifications describe all possible opportunities to deliver 4Ms care within a given encounter. After this, an organization will have all the raw components needed to construct the composite measure.

Detailed description

At UCSF, 30-day unplanned hospital readmissions was chosen as the key outcome of interest, as reducing these readmissions was a major motivation for health system investment in 4Ms implementation. We then specified all opportunities to deliver 4Ms care to be in concordance with the IHI minimum set. We outline our process using a hypothetical, 5-day inpatient encounter (Table 2), detailing the expected 4Ms care processes from Table 1. A hospital encounter fully adherent to the 4Ms Framework would include adherence to a minimum of 13 care processes that span 4 Assessments and 9 Act Ons. Once care process frequency is also considered, the number of opportunities for 4Ms care adherence expands even further. For example, during this 5-day example encounter, which includes ten 12-hour shifts, there are at least 3 encounter-level Assessment opportunities for What Matters, Medication, and Mobility (ie, those 4Ms care processes that should be done once per hospital admission), at least 10 shift-level Mentation Assessment opportunities, and between 35 and 44 Mentation Act On opportunities (depending on inferred frequencies for care processes, where frequency is not explicitly specified).

In addition to making decisions about the measured care processes and their frequencies for those that are not explicitly defined by IHI (eg, physical therapy), several of the IHI care processes are open-ended in how they should be implemented and, therefore, measured in the context of a composite measure. Notable examples include the number of hours of continuous sleep that constitute uninterrupted sleep, the definition of a tethering device, or what alignment of the care plan with What Matters means in practice. While these measurement decisions have considerable impact on an ultimate composite measure, they can be set at an initial level (reflecting the current state of practice in the given organization), and then modified as care processes are optimized. UCSF specifications are reported in the prior work described above ⁷ and can be used as a model.

Step 2: Select a Composite Measure Type

High-level description

To begin constructing the composite, the next decision is the choice of measure-type. There are many types of composite measures, along with tradeoffs when deciding which to use. A composite measure, for example, could be dichotomous, continuous, or ordinal based on thresholds. We selected a dichotomous measure such that we could classify encounters as having received or not received 4Ms care. However, we did not require perfect adherence for care delivery to qualify as 4Ms care because we found this was almost never achieved in our real-world data. Instead, we identified an adherence threshold at which an effect on the outcome was achieved by a real, albeit small subset of encounters, and used this threshold to define a realistic target for future care delivery.

Detailed description

Dichotomous measures are defined in a binary way: adherence to 4Ms care processes for a given encounter either did or did not meet a threshold that constitutes 4Ms care. A dichotomous composite measure has the benefit of simplicity, but it requires determination of a cutoff value. One hundred percent adherence across all possible opportunities for 4Ms care in the IHI model would represent the most stringent definition. In practice however, as evidenced by only 84 of 22 825 (0.37%) of encounters in the UCSF sample that achieved this level, this stringent definition may collide with the challenges of real-world practice and result in a diminishingly small sample of encounters such that determining the impact on outcomes is infeasible. One alternative is to lower the stringency of the adherence threshold, but the choice of an alternative threshold is arbitrary unless informed by data showing a meaningful inflection point in an outcome at that threshold. An alternative dichotomous composite measure which avoids determining a threshold could instead require that at least 1 care practice within each of the Assessment and Act On groups of the 4Ms be performed no less than once during the encounter. However, this approach neglects not only the recommended shift and daily frequency of many care processes, but the recommendation that several care processes within each Act On for Mentation and Mobility be carried out. It also runs the risk of adherence inflation by allowing the selection of only those care processes that might be easiest to achieve.

Continuous measures, on the other hand, do not use a threshold to define 4Ms care and instead capture the extent to which all 4Ms care processes are followed—from 0% to 100%. They offer flexibility in the denominator that could be used for such a measure, and support assessment of a dose-response relationship with outcomes. However, continuous measures come with their own complexity because of the heterogeneity of measurement units across 4Ms care processes. For example, some processes occur at the hospital encounter level while others occur at the day or shift level, giving some processes potentially disproportionate influence on a composite measure. Furthermore, with a continuous composite measure, it would be problematic to treat as equivalent moderate adherence to all 4Ms versus more complete adherence to 3Ms and not the fourth M at all. Additionally, interpretation of a continuous measure in the context of outcomes assessment is less straightforward than a dichotomous measure.

Threshold-based ordinal measures (eg, Poor, Fair, Good, Excellent) have the advantage of assessing dose-response relationships with outcomes without the assumption of equal impact of an increase from 5% to 10% adherence versus 90% to 95%. This is helpful when there is not necessarily a strong linear relationship between performance and outcomes (a scenario that is likely). However, as with dichotomous measures, implementing ordinal measures requires deciding where to set meaningful cutoffs for each category. In addition, because an ordinal approach requires more than 1 such decision about category thresholds, it may be subject to greater measurement error than a dichotomous approach alone.

Illustrating the Conceptual Considerations of Composite Measure Types

With these conceptual considerations in mind, Figure 1 presents a hypothetical 5-day hospital encounter. As illustrated in the left panel of Figure 1A, every opportunity to deliver 4Ms care except 1—assessing for delirium on the second shift of day 3—was completed. By missing this single opportunity for delirium assessment, a dichotomous composite measure requiring 100% adherence to all 4M care process opportunities would categorize this encounter as not constituting 4Ms care (right panel, composite value 0), but would constitute 4Ms care, for example, if an arbitrarily selected 50% adherence threshold had been set as in Figure 1A, (right panel, composite value 1). In contrast, a different dichotomous composite measure (also right panel) that only required evidence of adherence to at least one of the measured care processes within each of the 8 Assessment and Act On domains would categorize Panel A as constituting 4Ms care. A continuous composite measure, on the other hand, would capture that 29 out of 30 (95%) care process opportunities were performed during the encounter. Variations on this continuous composite measure could reflect weighting by M or adjustment for encounter-level versus day-level versus shift-level process frequencies. Finally, while an ordinal composite would characterize the care of Panel A in a top category such as “Excellent,” it would need to determine an appropriate combination of care processes and adherence that meaningfully associate with the outcome to define the thresholds between the other ordinal groups (eg, Poor, Fair, and Good).

Figure 1.

Illustrations of decreasing levels of adherence to 4Ms care delivery against one example of IHI’s specification for inpatient care. For visual simplicity, we have reduced the Assessment and Act Ons for each M in Table 1 to a single row each. (A) Left panel—Just short of 100% adherence; Right panel—4Ms composite measure values with different composite metric approaches. (B) Adherence in which all care processes within each M and the full set of Ms are completed synchronously (ie, in the same day or shift) for >50% of all days. (C) Adherence in which within-M synchrony is achieved for each M at least once throughout the encounter, but in which across-M synchrony is not. (D) Adherence for which a minimum set of care processes have been performed at least once during the hospital encounter, but not necessarily synchronous within or across Ms (eg, Mentation and Mobility are completed on different days). Sample sizes of patients meeting or exceeding each adherence level, from a starting n of 22 825 older adult inpatients at UCSF (upper left of each panel) are approximate for illustrative purposes. * and red box surrounding Panel B indicate our final approach.

We illustrate this example using the real-world data from the UCSF sample of 22 825 older adult inpatient encounters. As presented earlier, the care during only 84 (0.37%) of these hospital encounters met the perfect level of adherence (and only 100 met the near-perfect level of adherence illustrated in Figure 1A). In addition, nearly perfect adherence targets become increasingly unrealistic as length of stay increases. Because there are many more opportunities for delivering 4Ms care in longer encounters, failure to deliver any one opportunity becomes increasingly likely over time. (Figure 2) Therefore, if the care illustrated in Figure 1A was represented using a dichotomous composite measure with a high threshold for positive adherence (eg, 100%), it would potentially neglect the positive effects on outcomes of care that was largely excellent during a long encounter.

Figure 2.

Percent of hospital encounters, from a UCSF sample of 22 825, whose care throughout hospitalization achieved 100% adherence to 4Ms Assessments + Act Ons (n = 84; 0.37%) as a function of length of stay.

Step 3: Decide on an Approach to Specifying Synchrony of Processes Within Each M and Across Ms

High-level description

The measurement considerations presented thus far have largely been agnostic to the nature of care delivered under each M and their interplay within and across Ms. However, there are important considerations for a composite measure that arise from the relationships within and across M-specific care processes. Specifically, at issue is whether within each M, multiple care processes should occur on the same day or shift, and whether across Ms, daily or shift-level care processes should occur on the same day or shift as other Ms, so that together, the Ms are delivered in synchrony. To differentiate 4Ms as a set from each M as a standalone process, we decided to require that all care processes within each M and across all Ms occur in synchrony at least some of the time during the encounter, and used our key outcome (unplanned 30-day hospital readmissions) to guide where to set the cutoff.

Detailed Description

Inherent in the 4Ms Framework is that each of the Ms is important and that some level of adherence to every M is essential to define 4Ms care in a composite measure. Adhering fully to 3 of the Ms, but not to the fourth would therefore be inconsistent with delivering 4Ms care. At the same time, it should also be recognized that organizations often build out their 4Ms deployment in stages, with some but not necessarily all Ms deployed at once. In this scenario, adherence to those Ms that are available is essential, and establishing this framework as each M is added lays a strong composite measurement foundation. Nevertheless, because there may be synergies when all 4Ms are deployed, and because the spirit of 4Ms care is only achieved once all 4Ms are practiced, organizations must recognize that achieving a specific outcome, or failure to do so, before all 4Ms are deployed, does not tell a full 4Ms story.

Because the 4Ms are a set of practices intended to be carried out together, synchronous adherence within each M to daily and shift-level Assessment and Act On criteria is clinically reasonable and appropriate. If, for example, a patient is to be assessed for delirium every 12 h, then it would be clinically indicated to perform Act Ons to either mitigate or to correct potential causes of delirium at a similar, synchronous frequency. By analogy, another consideration is whether a composite measure should require synchrony across Ms (eg, the day-level care processes of Mentation are addressed on the same day(s) as the day-level care processes of Mobility and of Medication).

These concepts are illustrated in Figure 1 and can be described as follows.

•  Do Everything All of the Time—In the adherence model of Figure 1A, there is largely full synchrony within and across Ms (by definition, as virtually all 4Ms care is delivered). This model of adherence strives to accomplish all care processes at every opportunity as defined by the IHI Guide.

•  Do Everything (Within and Across Ms) Some of the Time—In Figure 1B synchrony occurs both within each M (eg, both the single Assessment and all of the Act Ons for Mobility are done within the same day or shift), as well as across Ms (eg, a complete Mobility M occurs on the same day as complete Mentation and Medication Ms, while the What Matters M, as an encounter-level M, can happen any time during the encounter). This model suggests an adherence approach in which outcomes stand to be improved not only when there is synchrony within each M, but when there is synchronous adherence across Ms on the same day. Although a level of adherence of >50% of days for this model is what is illustrated in Figure 1B, the adherence depicted by this model of within-M and across-M synchrony could occur anywhere on a spectrum from >0% of days to <100% of days.

•  Do Everything (Within Ms) Some of the Time—Figure 1C, on the other hand, shows a more minimalistic adherence model in which synchrony occurs within each M at least once over the course of the encounter, but not necessarily synchrony across Ms. For example, while all care processes might be complete for Mentation on day 2, and all care processes might be complete for Mobility on Day 1, there is no same-day synchrony achieved between complete Mentation and complete Mobility Ms. While this model of care illustrates that when a given Assessment screening is performed, there is appropriate synchronous (same day) action taken for that M, it does not require (and therefore assess the impact of) across-M synergy. In this sense, a composite measure that reflects this approach to adherence may serve more to replicate the standalone M-specific evidence that served as the original foundation for the 4Ms Framework ¹¹ rather than advance evidence on the 4Ms as a set.

•  Do (At Least) Something All of the Time—Finally, Figure 1D illustrates an adherence model in which every individual care process is completed at least once during the encounter, and at least one care process is done every day, but in which there is no regard for synchrony within Ms (see eg, Mentation in which the Assessment is done on one day, and the Act On—or the various components of the Act On—are done on one or more different days) or across Ms (eg, Mentation and Mobility are done on different days). The lack of synchrony within Mentation in this example would likely be viewed as clinically inappropriate care for many reasons, notable among them because a positive delirium screening was not synchronously mitigated or addressed by any Act Ons. This adherence model would check the box that every care process was done but requires neither within-M nor across-M synchrony, serving overall as a weaker approach to assess the impact of the 4Ms Framework.

Given these varied patterns of synchrony, there are options for how they might be factored into a composite measure. First, there could be a requirement for synchrony within each M. From a clinical perspective, this level of synchrony makes sense (as previously described). A second option is to require synchrony across Ms, but not necessarily within Ms. This approach offers cross-M synergy but suffers from the same concern that a positive Assessment screen may have no specific Act Ons performed. Furthermore, if within-M synchrony is incomplete, then across-M synchrony might have little meaning (eg, a partially completed Act On from Mentation—which has multiple Act On components—synergizing with another partially completed Act On from Mobility). A third option is to require both synchrony within and across Ms. Namely, 4Ms care is only considered to be 4Ms care when all care processes within each M are conducted synchronously, and when all day and shift-level Ms are also conducted synchronously (Encounter-level care processes—those that need to be done only once during the hospital encounter—need not be done synchronously with the other Ms, but alignment with other Ms would be enhanced when encounter-level care processes such as What Matters, are done early in the encounter. This requirement could also be built in to a composite measure.).

The implications of different approaches to constructing a composite based on within- and across-M synchrony are illustrated by comparing Figure 1B and C. Figure 1B features both within- and across-M synchrony while Figure 1C features only within-M synchrony. While Figure 1B illustrates within- and across-M synchrony occurring on >50% of days, one could imagine a similar version in which within- and across-M synchrony occurs only 1 out of 4 days (“25% scenario”). Compare such a “25% scenario” to Figure 1C in which the same proportion (25%) of care processes are done, but for which there is no across-M synchrony. Even though a “25% scenario” in Figure 1B would have the same number of care processes performed as in Figure 1C, only the former would enjoy the benefits of cross-M synergy that distinguishes 4Ms care from single M care.

So, if Figure 1B is advantageous relative to Figure 1C because of cross-M synchrony, it would then be important to determine at what percent of days—whether 20%, 50%, 75%, etc.—such adherence would positively impact outcomes. In our own UCSF data, with 30-day unplanned hospital readmissions as the outcome, this occurred at 50% of days. However, because this cutoff value is likely to vary by setting or organization, it is essential that each organization identify the adherence level at which the outcome’s inflection point occurs.

Overall, this combination of within- and across-M synchrony for some proportion of hospital days (as illustrated in Figure 1B) is likely the best approach to reflect 4Ms inpatient care as a set, thereby distinguishing the 4Ms Framework from standalone care processes that address each M in isolation.

Step 4: Assess the Relationship Between the Composite Measure and the Key Outcome in Real-World Setting

High-level description

As a final step, it is necessary to confirm a clinically meaningful and statistically significant relationship between the composite measure and the selected outcome. It is equally important to affirm that the level of performance on the composite measure at which an inflection point in the outcome occurs is achievable in an organization’s real-world clinical settings by a reasonable proportion of clinicians or within a reasonable proportion of clinical encounters—even if relatively small. If, for example, it is only being achieved in a fraction of a percentage of encounters, then targeting a level of performance improvement for more encounters to achieve a threshold that is already largely unattainable, may be setting the bar too high. In addition, if there are too few observations in a given group to establish a meaningful relationship between the composite measure and the outcome, then adjustments to thresholds (as described above) may be warranted. Once the relationship between the composite measure and the outcome has been confirmed in a reasonable sample size, that fixed composite measure specification should then be used to assess the relationship with all other outcomes.

Detailed description

We started with a composite measure that required perfect adherence (ie, following all 4Ms care processes during the entire encounter), but this proved to be a high bar. In our UCSF sample, we found that this ideal level of 4Ms care only occurred in 84 (0.37%) encounters, which offered too few cases for a meaningful outcomes analysis. After several iterations, we selected an alternative composite measure that achieved a balance between ensuring fidelity to the 4Ms Framework and a sufficient sample size of encounters classified as having received 4Ms care. In our case, we found significantly lower 30-day unplanned hospital readmissions for patients when: (1) all encounter-level care processes were adhered to and (2) 50% or more of hospital days had full adherence to day- and shift-level care processes.

After confirming that a sufficient number of encounters would be classified as having received 4Ms care using this definition (1075 encounters or 4.7% of the sample), we selected this as our final composite specification and regressed this dichotomous measure against other outcomes of interest to us (eg, inpatient length of stay, 30-day ED revisits and discharge disposition to home). These secondary outcome relationships were assessed using generalized linear regression models, clustered at the patient level, controlling for a number of additional patient and encounter level covariates.

This approach to specifying the composite measure identified an inflection point in 30-day unplanned readmissions at which a small yet reasonable level of adherence is already attained in real-world care at our institution that still captures the intent of the 4Ms Framework. While only 4.7% of patient hospital encounters in our sample met this final level of adherence, it yielded a sufficient sample size to identify an adherence threshold at which a clinically meaningful impact on a key outcome was detected. With this composite metric and baseline performance rate in hand, future performance targets can then be set to incrementally boost the proportion of encounters meeting this level of adherence.

It is possible that such an approach at other organizations will not establish an obvious performance threshold at which the outcome changes. This may be a function of the care practices of that organization, how many of the Ms are implemented, the sample size, whether there are too few cases of real-world performance for the composite metric to be pragmatic, or a function of the selected outcome itself (ie, if it truly has no relationship to the composite measure). This framework provides a structure for defining a candidate composite measure, but it does not guarantee that such a composite measure is necessarily feasible at all sites. Finally, while a composite specification determined in this manner does imply a relationship between 4Ms care and the selected outcome of interest, it does not necessarily imply a relationship with other outcomes.

Conclusion

To support rigorous evaluation of the 4Ms Framework, it is essential to develop a composite measure of 4Ms care delivery. A health system that assesses the impact of the 4Ms on outcomes using a measure that fails to capture the true nature of the framework risks coming to incomplete or inaccurate conclusions. In this piece, we describe a breadth of considerations, summarized in Table 3, for developing such a measure at the hospital encounter level and suggest a recommended approach based on our real-world experience. Many of the considerations are equally applicable to developing a composite measure of 4Ms care in other settings, which is a key future direction for this work. While it is unlikely that a single, standardized composite measure specification is feasible at this early stage of 4Ms evidence generation, the approach that we describe and illustrate offers a starting point for similar approaches. Only with shared concepts related to developing composite measures can we collectively understand current 4Ms adherence and its relationship to outcomes that will help advance the field toward the broader goal of evidence-based age-friendly care.

Table 3.

Recommended approach to 4Ms composite measure construction.

Step of composite measure development	Dimension of UCSF implementation	Specific UCSF implementation
1. Identify key outcome of interest and care process frequencies relevant to the organizational context, making sure that all of the Ms are addressed.	30 Day unplanned hospital readmissions, with minimally required care processes defined by IHI measured for all 4Ms.	All numbered care processes in Table 2, Column 2 were measured for every encounter.
2. Decide on Measure Type: Dichotomous vs. continuous vs. ordinal measure.	Assess candidate dichotomous composite measures to identify a “final” dichotomous composite measure.	We assessed increasingly stringent levels of adherence on a dichotomous measure.
3. Consider within-M synchrony and across-M synchrony	All minimally required care processes within an M are performed synchronously.	All day/shift-level processes (Mobility Act Ons, Mentation Assessments, and Mentation Act Ons) are all performed on the same day and during the same encounter as the encounter-level processes (Mobility Assessment, What Matters Assessment & Act Ons, and Medication Assessment & Act Ons.)
	All care processes within an M must occur synchronously across all 4Ms.	All day/shift level care processes of one M are performed on the same day as all day/shift level care processes of all other Ms.
4. Assess the relationship between the composite measure and the key outcome in the real-world setting.	Clinically meaningful and statistically significant reduction in 30-day unplanned hospital readmissions achieved in a reasonable volume of real-world encounters.	We found significant and clinically meaningful reduction in 30-day unplanned hospital readmissions once adherence reached 100% of encounter level processes and full adherence to all other processes (meeting synchrony requirements above) on at least 50% of hospital days. This was achieved in 4.7% of real-world encounters.