Exploring Goal‐Concordant Medication Use Among VA Community Living Center Residents With Dementia

ABSTRACT

Background

Documentation of patient goals and preferences within medical records has the potential to align medication use with goals of care (GoC) and individualize medication appropriateness criteria. We characterized patient and surrogate‐expressed GoC for older Veterans living with dementia and explored concordance with medication use during VA Community Living Center (CLC) (i.e., nursing home) stays.

Methods

We conducted a cross‐sectional analysis using the VA Residential History File, Minimum Data Set, Corporate Data Warehouse, and Medicare claims for Veterans with dementia admitted to VA CLCs from 4/2021 to 12/2021 for > 7 days. We extracted free text responses for “Veteran goals in own words” from a standardized GoC note. Two coders classified GoC topics using iterative coding. We examined bar code medication administration data for aspirin, benzodiazepines, opioids and antidementia medications within the 7 days following admission. We determined a schema for potential goal‐concordant medication use (e.g., opioids for GoC focused on comfort) and assessed concordance of medication use with GoC topics.

Results

Among 1000 VA CLC residents with dementia and GoC documented, 46.4% of responses were reported by the Veteran versus a surrogate. Common topics included comfort (44.6%), life‐sustaining treatments (31.8%), function (13.7%), care setting/transitions (12.9%), and life prolongation (11.2%). Medications were seldom discussed. Opioid and benzodiazepine use was classified as goal‐concordant for 56.7% and 72.2% of patients who used them. Aspirin and antidementia medication use was more commonly classified as goal‐discordant (54.7% and 38.7%, respectively).

Conclusions

Goals elicited via an open‐ended question provided only indirect information relevant to medication use, but in many cases could be used to refine judgments of appropriateness. Integration of patient goals into formal criteria evaluating medication appropriateness is a logical next step for medication optimization research. Future research should explore the utility of questions specific to medications in GoC conversations for individuals with dementia.

Summary.

• Key points

  • ○
    Existing tools for medication optimization rely on criteria such as advanced age, comorbid conditions, and prognosis to assign appropriateness, but there has been little exploration of how patients' goals of care could further enhance assessment of medication appropriateness.
  • ○
    We explored the utility of free‐text data capturing Veterans' goals of care for assigning medication appropriateness for four classes of medications among older Veteran nursing home residents living with dementia.
  • ○
    Patient‐stated goals from the electronic health record provided only indirect information relevant to medication use, but in many cases could reasonably be used to refine assessment of medication appropriateness.
• Why does this matter?

  • ○
    For older adults living with dementia, objective criteria for medication appropriateness may fail to adapt medication optimization as goals of care shift from preventive to comfort. This study lays the foundation for future research and interventions for medication optimization by establishing the feasibility of integrating data documenting patients' goals of care to enhance assessment of medication appropriateness.

1. Introduction

Medication optimization is a priority for older adults living in nursing homes [1]. Studies consistently demonstrate that a high proportion of older adults living with dementia, many of whom reside in nursing homes, receive medications that either carry high risk for adverse effects or are unlikely to provide substantial benefits in consideration of prognosis or life expectancy [2, 3]. However, existing tools that identify opportunities for medication optimization, such as the Beers Criteria [4] and the STOPP/START Criteria [5], define medication appropriateness based on advanced age, specific conditions, and medication‐specific risks, and do not account for variation in patient‐specific goals that may affect the appropriateness of medications on an individual basis. For older adults in nursing homes and those living with serious illnesses such as dementia, this approach may fail to adapt medication optimization as goals shift from favoring curative or preventive treatments to prioritizing short‐term symptom relief and comfort [6, 7]. For example, the Beers Criteria would objectively categorize antipsychotic use as inappropriate, even in situations in which managing behavioral symptoms of dementia would support patients' and caregivers' priorities for comfort. At the same time, the STOPP Criteria would categorize most preventive medications (e.g., statins, aspirin) as inappropriate for individuals with advanced age or advanced dementia, even if their goals were to prioritize life prolongation.

Integration of patients' goals and preferences into criteria for evaluating medication appropriateness is a logical next step for research in the field of medication optimization, as this would more closely mirror real‐world clinical decision‐making about medications. However, goals of care (GoC) discussions seldom address medications directly. Relatedly, documentation of GoC within electronic health records is inconsistent and often not accessible or linkable to larger national or administrative healthcare datasets (e.g., insurance claims) for study. The VA national healthcare system presents a unique opportunity to overcome this challenge through a standardized note template and discrete data fields developed through the VA Life‐Sustaining Treatment Decisions Initiative [8]. One of the goals of this initiative is to encourage clinicians to engage in conversations with Veterans who have high risk for hospitalization or serious illnesses to discuss advance care planning and document their preferences for life‐sustaining treatments (LSTs) in a standardized note. The note is prompted during certain healthcare events (e.g., hospitalizations, nursing home admission) and was recently updated to include a free‐text response documenting GoC. Prior studies have examined the implementation of this note template [8] and associations with perceived quality of care [9], but no studies have leveraged these data to characterize the concordance of medication use with patients' GoC.

The objectives of this study were to characterize patient and surrogate expressed GoC for older Veterans living with dementia and to explore concordance of goals with medication use during VA Community Living Center (CLC) (i.e., nursing home) stays.

2. Methods

This study was conducted as part of a larger study using healthcare records to examine medication deprescribing for residents of VA CLCs living with dementia. It was overseen by the Institutional Review Boards of the VA Pittsburgh (exempt), the University of Pittsburgh (exempt), and the University of North Carolina at Chapel Hill (expedited approval).

2.1. Data Sources, Design, and Sample

We conducted cross‐sectional descriptive analysis using data from the VA Residential History File [10], VA Corporate Data Warehouse, Minimum Data Set (MDS), and Medicare claims. This analysis sought to examine patient‐stated GoC, which can be extracted from the VA Corporate Data Warehouse and is documented in a standardized note template designed to elicit and document preferences for GoC, such as LSTs. This note template is designed to support GoC discussions and is intended to be completed in the setting of major health events or healthcare transfers, including within 7 days after VA CLC admission. However, the template can also be completed during any other encounter. Specifically, a free‐text field documenting “Patient goal in own words” was implemented nationally in the second version of the note template in April 2021 and was the primary focus of our analysis.

Our target sample was Veterans with evidence of dementia who were newly admitted to VA CLCs. As part of the larger study, we initially identified all VA CLC admissions occurring between October 1, 2015 and December 31, 2021 (Figure 1). We required that Veterans had a full MDS admission assessment completed at admission (n = 167,565), were aged 65 or older (n = 130,257), and enrolled in VA and Medicare Parts A and B for 2 years prior to admission (n = 56,682). We identified Veterans with dementia (n = 34,375) based on meeting any of three criteria [11]: (1) Chronic Conditions Warehouse claims‐based criteria [12] using International Classification of Diseases‐10 (ICD‐10) codes from VA and Medicare utilization data, (2) MDS active diagnosis indicators for Alzheimer's disease (I4200) or non‐Alzheimer's dementia (I4800), (3) at least mild cognitive impairment on the MDS Cognitive Function Scale (CFS) [13]. The CFS and its individual components have been demonstrated to have high sensitivity and specificity for identifying any cognitive impairment [13]. The decision to use any of these three measures for possible dementia was intended to be inclusive, based on prior research that has demonstrated noticeable discordance and potential under‐reporting across measures in VA CLC residents [11]. We further limited our sample to VA CLC stays with an admission date between July 01, 2021 and December 31, 2021 (n = 1416) to align with the national roll‐out of the discrete data field documenting patient‐stated GoC.

FIGURE 1.

Sample construction. ICD, International Classification of Diseases; MDS, minimum dataset; VA, Veterans Affairs.

2.2. Measurement and Classification of Patient‐Stated GoC

To identify Veterans with documented GoC, we used VA Text Integration Utilities data files, which contain free‐text data from clinical notes and other documentation within the VA electronic health record. We searched the data for any notes titled “life‐sustaining treatment” or similar—the title of the standardized note template that documents responses for “Patient goal in own words”. The text data extracted solely contains responses to “patient goal in own words” as this is the only free text data in the note template. To characterize patient‐stated goals, we included responses dated prior to each Veteran's CLC admission date up until 30 days following admission, meaning these happened in previous care settings or during CLC admission. We ultimately selected the response closest to the CLC admission date (n = 1011 responses). If a Veteran had multiple unique responses with the same date, these were combined into one response for coding, yielding a final sample of n = 1000 responses.

Free text responses for “patient goal in own words” were entered into a long‐form spreadsheet for qualitative analysis. The study team identified several a priori categories of GoC based on expert opinion and review of the literature, which included prolonging life, supporting function, and comfort [1]. We also used an iterative coding approach with the option for open coding to allow for new categories of goals as they emerged. Two coders (J.D.N. and N.B.) initially examined 20% of responses to gain an understanding of the data and to identify the range of possible categories for coding. An initial codebook was established based on this review. The two coders then re‐reviewed and coded these responses to evaluate agreement. In addition to coding the topics discussed in each response, coders also determined the valence of each response—that is, whether the statement was in support of or opposed to that treatment or goal. For example, two separate responses may both discuss life support or heroic measures, with one respondent being in support of these measures and another being opposed to them. We allowed for multiple codes per response if several distinct goals or topics were discussed. Initial review of 20% of responses yielded 71% agreement and a primary coder applied codes and valences to the remaining responses. Throughout the process, the primary coder met with the study team to review any uncertainties in coding and new categories for GoC as they emerged.

2.3. Use of Medications

We used barcoded medication administration data to examine medications administered during each of the first 7 days of each Veteran's CLC stay. We leveraged the input of our study team, consisting of geriatricians, pharmacists, and health services researchers, to identify four medication classes that are commonly prescribed for older adults, yet might differ in appropriateness based on GoC. These included aspirin for primary or secondary prevention, antidementia medications (cholinesterase inhibitors and memantine), benzodiazepines, and opioids. Medications were identified based on generic drug names from the VA National Drug File [14]. Medication use was examined among the subgroup of Veterans with at least 7 days of follow‐up after CLC admission. To ensure correct temporality between goals and medication use, we also limited Veterans to those with goals documented by the 7th day of stay (medication sub‐sample, n = 766). We then used barcoded medication administration data to identify whether Veterans had any use versus no use of each medication class during the first 7 days of their CLC stay.

2.4. Other Descriptive Measures

For each free‐text response for “patient goals in own words”, we characterized the respondent type (patient or surrogate) and the time between CLC admission and documented response date in days. Other patient‐level descriptive characteristics were extracted from the first MDS assessment of the CLC stay. Demographic characteristics included age at admission, sex, race, and ethnicity. We also characterized the severity of cognitive impairment (intact, mild, moderate, and severe) based on the MDS CFS [13]. Due to our inclusive approach to identifying dementia [11], it was possible for patients who qualified for the cohort based on ICD code or MDS diagnoses for dementia to be rated as “intact” on the CFS. Finally, we characterized prognosis using MDS item J1400, which indicates physician‐perceived life expectancy of less than 6 months, and the MDS Mortality Risk Index‐v3, which suggests high risk of 6‐month mortality for scores ≥ 36 [15].

2.5. Analysis

We reported sample characteristics for the full cohort and the medication sub‐sample. We calculated the overall frequency of coded categories of GoC in the full cohort (n = 1000), along with the proportion of patient versus surrogate responses that fell within each category. Within each category of GoC, we reported the frequency of subthemes, as relevant. To understand how multiple GoC may co‐occur, we then examined potential overlap between individual categories of GoC in a matrix table.

To examine potential concordance of medications and GoC, we relied on the clinical expertise of our research team to identify clusters of goals that could support the use of each medication class. These discussions included a geriatric pharmacist, a VA geriatrician, a geriatrician with dual training in palliative care, an internal medicine physician, and researchers with expertise in geriatric pharmacoepidemiology and health services research. Within each medication class, the team discussed the relevance of each GoC category for decision‐making and whether medication use was aligned with a given GoC when thinking of “most people living with dementia”. Majority consensus was used to sort GoC into one of three designations for each medication class: Goals that potentially support medication use, goals that potentially do not support medication use, and goals with unclear relevance to medication use. Among users of each medication class, we reported the proportion of Veterans that fell into each of these designations based on their stated GoC.

3. Results

3.1. Sample Characteristics

In the full cohort of VA CLC residents with dementia and documented GoC (n = 1000), most were male (97.0%) and aged older than 70 years old (70–79 years: 43.1%; 80–89 years: 29.6%; 90+ years: 19.6%) (Table 1). The full cohort was 79.2% White, 14.4% Black, and 3.2% Hispanic. The median length of CLC stay in our cohort was 20 days (IQR: 9–44 days). Most episodes in the full cohort were for stays shorter than 90 days (89.1%). Respondents to the GoC question were more often surrogates or others (53.6%) compared to patients (46.4%). Just under half of CLC residents had at least moderate cognitive impairment as measured by the CFS (moderate: 22.8%, severe: 22.6%). More than half of Veterans had high risk for 6‐month mortality based on the MDS Mortality Risk Index‐v3 (51.6%) and slightly fewer (42.8%) were designated as having prognosis < 6 months by a provider. The median time from CLC admission to goals documentation was −3.0 days (i.e., completed 3 days prior to CLC admission). The distribution of these characteristics was relatively comparable in the medication use sub‐sample. The only notable exceptions were a lower prevalence of severe cognitive impairment (16.3% vs. 22.6%) and lower prevalence of high risk for 6‐month mortality (41.5% vs. 51.6%).

TABLE 1.

Sample characteristics.

Characteristic	N (%) full sample, n = 1000	N (%) med sample, n = 766
Age
65 to < 70	77 (7.7)	58 (7.6)
70 to 79	431 (43.1)	341 (44.5)
80 to 89	296 (29.6)	219 (28.6)
90+	196 (19.6)	148 (19.3)
Sex
Male	970 (97.0)	743 (97.0)
Race/ethnicity
White (non‐Hispanic)	792 (79.2)	605 (78.9)
Black (non‐Hispanic)	144 (14.4)	110 (14.4)
All others	64 (6.4)	51 (6.6)
Respondent
Patient	464 (46.4)	386 (50.4)
Surrogate/other	536 (53.6)	380 (49.6)
Dementia severity
Intact	161 (16.1)	141 (18.4)
Mild	335 (33.5)	286 (37.3)
Moderate	228 (22.8)	194 (25.3)
Severe	226 (22.6)	125 (16.3)
Missing	50 (5.0)	20 (2.6)
MMRI‐v3 score ≥ 36	516 (51.6)	247 (41.5)
Limited prognosis (EndStage_MDS)	428 (42.8)	339 (44.2)
Hospice care	438 (43.8)	269 (35.1)
Length of stay
Median (IQR)	20 (9–44) days	27.5 (14–53) days
Length of stay > 90 days	109 (10.9)	99 (12.9)
Time from CLC admission to LST note completion
Mean ± SD	−13.5 ± 32.1 days	−15.8 ± 34.0 days
Median (IQR)	−3.0 (−11.0 to −1.0) days	−3.0 (−13.0 to −1.0) days

3.2. Patient‐Stated GoC

In the full cohort, categories of topics discussed during GoC conversations (Table 2) included comfort (44.6%), LSTs (31.8%), setting and burdensome transitions (13.7%), prolonging life (11.2%), person‐specific values of priorities (9.6%), quality of life (7.9%), specific medical treatment decisions (7.6%), acceptance of limited prognosis (4.7%), and symptom management (1.9%). Medications were discussed very infrequently. Overlap between categories of goals is presented as a matrix in Table S1. The two categories that were most often coded together were LSTs and comfort (7.8% of all responses). We observed several notable differences in the prevalence of categories of GoC between patient and surrogate responses. The most extreme differences were for comfort (patient: 29.5% vs. surrogate: 57.7%) and support function (patient: 21.6% vs. surrogate: 6.9%) goals.

TABLE 2.

Frequency of responses in full sample (TSR = cell sizes were too small to report per terms of data use agreement).

Category	Coded overall %	Respondent %	Substantive interpretation(s)
	N = 1000	n = 464 patients, n = 536 surrogates	N = 1000
Comfort	44.6%	Patient: 29.5% Surrogate: 57.7%	Patient being comfortable was discussed as a top priority, either by pursuing treatments that support comfort or avoiding treatments that may cause discomfort. (44.6%)
Life sustaining treatments	31.8%	Patient: 32.8% Surrogate: 30.9%	Expressed a preference in support of pursuing heroic or life‐sustaining measures like CPR, ventilation, and so forth. (6.4%)
			Expressed a preference opposed to pursuing heroic or life‐sustaining measures or in support of a natural death with minimal interventions. (25.4%)
Support function	13.7%	Patient: 21.6% Surrogate: 6.9%	Expressed a desire to regain mobility, independence, or pursue short‐term rehabilitation to regain function. (13.7%)
Setting and burdensome transitions	12.9%	Patient: 14.9% Surrogate: 11.2%	Discussed a goal of being discharged or transferred to a specific setting (e.g., home, CLC, rehabilitation). (10.4%)
			Or discussed a preference to not be admitted or transferred to a specific setting (e.g., hospital, ICU). (2.5%)
Prolong life	11.2%	Patient: 14.4% Surrogate: 8.4%	Expressed a preference for pursuing treatments that would extend life. (4.1%)
			Expressed a preference against pursuing treatments that would extend life. (7.1%)
Person‐specific value or priorities	9.6%	Patient: 11.4% Surrogate: 8.2%	Mentioned a person‐specific value or goal as a priority that involved family, faith, or other topic not directly related to healthcare treatments. (9.6%)
Quality of life	7.9%	Patient: 6.5% Surrogate: 9.1%	Expressed preference for prioritizing quality of life as a goal. (7.9%)
Specific medical treatment decisions	7.6%	Patient: 5.6% Surrogate: 9.3%	Expressed a preference in support of a specific treatment with the goal of resolving an acute illness. (1.8%)
			Expressed a preference for a specific treatment other than heroic measures or life‐sustaining treatments (e.g., chemotherapy). (1.4%)
			Expressed a preference against pursuing a specific treatment other than heroic measures or life‐sustaining treatments (e.g., chemotherapy). (4.3%)
Acceptance of limited prognosis	4.7%	Patient: 5.4% Surrogate: 4.1%	Patient or respondent acknowledged limited life expectancy. (4.7%)
Symptom Management	1.9%	TSR	Patient discussed prioritizing management of specific symptoms or condition (i.e., other than generalized pain or discomfort). (1.9%)
Medications	TSR	TSR	Discussed preference for or against medications (either specifically or generally).

Note: 4.7% of the 1000 responses were unusable—that is, irrelevant to goals of care discussions, uninterpretable, or lacked substantive content.

The substantive interpretation of each category is also presented in Table 2. For some categories, there was variation in whether the Veteran discussed being in support of versus opposed to the particular aspect of treatment. For example, respondents more often discussed their opposition (25.4%) to receiving heroic or life‐sustaining measures rather than their support for receiving them (6.4%). Similarly, respondents more often discussed a preference against (7.1%) versus in support of (4.1%) pursuing treatments that would extend life.

3.3. Goal‐Concordance of Medication Use

In the medication sub‐sample (n = 766), 41.9% used opioids, 30.6% used aspirin, 20.6% used benzodiazepines, and 12.1% used antidementia medications (Figure 2). Detailed information on the categories of goals that were designated as goal‐concordant versus not, including several examples, is included in Table 3. Among opioid users, most CLC residents had documented goals that were potentially goal‐concordant (56.7%; e.g., comfort, quality of life, acceptance of prognosis, symptom management) compared to goals that were potentially not concordant (4.4%; e.g., in support of function or prolonging life) or that had unclear relevance to medication use (38.9%). A similar pattern was observed for benzodiazepine users, with most having documented goals that were potentially goal‐concordant (72.2%) compared to goals that were potentially not concordant or unclear (27.8% combined; subcategories not reported due to small cell sizes per the data use agreement with VA and the Centers for Medicare and Medicaid Services). Among aspirin users, most CLC residents had documented goals that were potentially discordant (54.7%) compared to goals that were potentially concordant (11.1%) or that had unclear relevance to medication use (34.1%). For residents with potentially discordant use of aspirin, fewer than half had evidence of peripheral vascular disease based on diagnosis codes or the MDS, which may represent a compelling indication for use independent of goals. Among users of antidementia medications, 38.7% of CLC residents had documented goals that were potentially discordant and 61.2% with unclear relevance to medication use. None had documented goals that were potentially aligned with antidementia medication use.

FIGURE 2.

Use of select drug classes by Veteran goals and preferences (Days 1–7 of VA nursing home stay).

TABLE 3.

Medication use and proposed goal‐aligned preferences.

Category	Alignment of goals with medication use
	Opioids	Benzodiazepines	Aspirin	Antidementia
In support of comfort
Opposed to life‐sustaining treatments
In support of function
Setting and burdensome transitions
In support of prolonging life
Person‐specific value or priorities
In support of quality of life
Specific medical treatment decisions
Acceptance of limited prognosis
In support of symptom management
Opposed to medications

Note: Blue = potentially goal‐concordant; orange = potentially not goal‐concordant; gray = unclear. Example interpretation. Row 1: Use of opioids and benzodiazepines can provide symptom management to improve quality of life for someone with goals of care that prioritize comfort. Whereas aspirin, which is most often used for long‐term prevention, and antidementia medications, which are used to slow disease progression, are unlikely to provide substantial short‐term comfort. Row 3: Conversely, antidementia medications may be more appropriate for someone with goals of care that prioritize maintaining function and independence by slowing cognitive decline. Other medications with serious adverse effects that can contribute to somnolence or falls (opioids and benzodiazepines) may outweigh the potential benefits, particularly when other effective options for symptom management are available. Aspirin may support function in cases where patients have compelling risk factors such as peripheral artery disease. But for most patients, the risk of bleeding or hemorrhagic stroke may outweigh the potential benefits of prevention.

Surrogates' responses were more often aligned with medication use for opioids and benzodiazepines. For residents receiving opioids, 69.7% of surrogate responses were aligned with medication use compared to 44.2% of patient responses. The same trend held true for benzodiazepine use, though the difference was less distinct (73.8% surrogate responses vs. 68.0% patient responses). We were unable to report these differences for the use of aspirin and antidementia medications due to small cell sizes.

4. Discussion

We found that medications were rarely discussed in response to questions about Veterans' GoC in their own words. Although the note template that was used to collect “patient goals in own words” was designed to focus on preferences for LSTs, the majority of responses to this prompt addressed topics with broad relevance to healthcare decision‐making. Our study supports the feasibility of using these data to evaluate the concordance of medication use with goals, though indirectly. Although our results are only exploratory, they suggest that opioids and benzodiazepines are more often used in the context of goals prioritizing comfort and symptom control, while preventive medications (aspirin and antidementia) are used even when goals are not focused on life prolongation.

A recent investigation examined the association of having any documented LST preferences with family members' perceptions of care quality for VA CLC residents in the final month of life [9]. Using data from the same VA note templates, investigators linked responses from the Bereaved Family Survey to evaluate whether LST note completion was associated with the receipt of unwanted medications and medical treatment along with overall quality of care in the last month of life. There were no significant differences in perceived quality of care between Veterans with and without documented LST preferences, including the receipt of unwanted medications, suggesting that discussions centered solely around LSTs may not lead to meaningful integration of patient and family preferences into care delivery. A recent report from the NIA IMPACT Collaboratory [16] corroborates these findings. A series of targeted conversations with patients and caregivers and proxies for older adults living with dementia highlighted the potential discordance between patients' and caregivers' perceptions of GoC and approaches for evaluating goal‐concordance of healthcare treatments and decision‐making for research. Overall, caregivers reported that identifying broad overarching goals was challenging and was not reflective of their experiences for which healthcare decision‐making was more often driven by immediate priorities and unmet needs. Thus, further study is needed to understand how current communication and documentation tools in the electronic health record can be modified to be more useful for aligning healthcare decisions with patients' needs.

Our study has several implications for future research and clinical practice. First, our results suggest that using explicit criteria (e.g., Beers, STOPP/START) in research to determine medication appropriateness may overestimate rates of inappropriate medication use for many medications. This is particularly true for situations in which patients, caregivers, and prescribers view the short‐term benefits of medications to be more aligned with comfort and symptom management goals and to outweigh the potential harms, for example, the use of psychotropic medications to manage behavioral symptoms of dementia [17, 18]. Further research should explore ways to integrate patient goals with explicit prescribing criteria to improve decision‐making for medications in clinical practice. Second, our results highlight the potential utility of expanding GoC prompts and documentation used in practice to explicitly address preferences for medications to guide decisions for medication optimization. Medications were seldom addressed directly, which may be attributable to a lack of perceived relevance to GoC discussions due to clinician training gaps or limitations of the electronic health record to capture detailed information on medications. Our research demonstrates that prompts designed to capture broad GoC provide clinicians with useful information in many cases to classify medications as concordant or discordant with patient goals. However, these classifications rely on subjective interpretations and require substantial clinical expertise to apply this information appropriately to medication decisions as there are no explicit criteria for classifying the appropriateness of medications according to patients' goals. Future research should seek to extend our work from this study to develop a comprehensive set of criteria for operationalizing goal‐concordant medication use based on the consensus of clinical providers, patients, and caregivers. Relatedly, future research could explore the utility of structured prompts in the electronic health record to elicit and document medication‐related goals directly. This could lay the foundation for future studies to examine the concordance of patient goals with medication use longitudinally and in larger national datasets.

Several limitations should be considered in the interpretation of our results. As this study was conducted within the VA Healthcare System, and our sample was predominantly male and for short stays, findings may not be generalizable to other populations. We also acknowledge the subjective nature of our classification scheme for goal concordance, which did not directly incorporate specific patient characteristics or comorbidities. Future studies should explore more rigorous approaches (e.g., Delphi panel) to gain consensus on the alignment of medications with GoC. Nonetheless, the strengths of our study include the use of national data from the VA healthcare system and demonstrating the potential utility of an innovative data source with the potential to advance the evaluation of goal‐concordant care for older adults.

5. Conclusions

In a sample of Veteran CLC residents with dementia, we found that goals and preferences elicited via an open‐ended question and documented in the electronic health record were easily categorized and addressed a broad range of GoC. Although these data only provided indirect evidence for the concordance of medications with goals, the present study lays the groundwork for future research to understand how GoC may change over time for patients with serious illnesses, such as dementia, and how this affects healthcare delivery.

Author Contributions

Study concept and design: Joshua D. Niznik, Lena K. Makaroun, and Carolyn T. Thorpe. Data management and analysis: Joshua D. Niznik, Florentia E. Sileanu, Nicole Beyer, Xinhua Zhao, Kelvin Tran, Keri L. Rodriguez, Loren J. Schleiden, and Carolyn T. Thorpe. Clinical perspective: Joshua D. Niznik, Lena K. Makaroun, Laura C. Hanson, Thomas R. Radomski. Preparation of manuscript: Joshua D. Niznik, Lena K. Makaroun, Keri L. Rodriguez, and Carolyn T. Thorpe. Critical revision and feedback: Joshua D. Niznik, Lena K. Makaroun, Florentia E. Sileanu, Nicole Beyer, Xinhua Zhao, Kelvin Tran, Keri L. Rodriguez, Laura C. Hanson, Thomas R. Radomski, Loren J. Schleiden, Alexa Ehlert, and Carolyn T. Thorpe.

Disclosure

Sponsor's role: The content is solely the responsibility of the authors and does not necessarily represent the official views of the Department of Veterans Affairs or the National Institutes of Health. Support for VA/CMS data provided by the Department of Veterans Affairs, Office of Research and Development, VA Information Resource Center (project numbers: SDR 02‐237 and 98‐004).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Matrix of overlap across goals of care categories (N = 1000).

Niznik J. D., Makaroun L. K., Sileanu F. E., et al., “Exploring Goal‐Concordant Medication Use Among VA Community Living Center Residents With Dementia,” Journal of the American Geriatrics Society 73, no. 12 (2025): 3738–3746, 10.1111/jgs.70158.