Co-designing an Outpatient Clinical Decision Support Prototype for Managing Social Risks in Patients Living with Dementia

Abstract

Background

Patients living with dementia (PLWD) require attention to social determinants of health (SDoH), but social information is often unavailable or incomplete during care encounters, and clinicians report uncertainty about how to act on this information.

Objective

This study aimed to co-design the Intelligent Social Risk Management in Alzheimer's Disease and Alzheimer's Disease-Related Dementias Patients (iSMART) clinical decision support (CDS) prototype, which integrates an AI-driven individualized polysocial risk score (iPsRS) to identify PLWD at high social risk for hospitalization.

Methods

We conducted a qualitative user-centered design study with nine outpatient providers and clinical staff who care for PLWD at a large academic health system. Participants completed semi-structured interviews to explore perceived social risks in dementia care, gather feedback on the iSMART prototype and its features, and identify implementation considerations. Interview transcripts were coded and analyzed via rapid qualitative and thematic analysis.

Results

Nine participants (six providers, two nurses, and one social worker) were interviewed. Lack of caregiver or family support was identified as the most important social factor to consider for PLWD, followed by financial strain and transportation issues. Participants described the iSMART prototype as helpful for identifying social risks among PLWD, summarizing and prioritizing factors contributing to hospitalization risk, and connecting patients to relevant social resources. Participants suggested primary care providers and social workers should be the primary users. Recommended improvements included auto-populated (as opposed to manual entry) but editable SDoH fields for model input, incorporating caregiver presence as a model input, and providing clearer visual representations of risk.

Conclusion

Outpatient providers and staff viewed the iSMART prototype as a promising approach to support social risk screening, resource connection, and referral support in dementia care. These findings inform practical strategies for integrating iPsRS-informed CDS tools into routine workflows in dementia clinics.

Background and Significance

Alzheimer's disease and related dementias (ADRD) affect over 7.2 million Americans 65 and older, with prevalence projected to double by 2060. 1 Symptoms often progress from memory loss and mild confusion 2 to disorientation, communication difficulties, and mood and personality changes. 3 4 5 As cognition and physical functioning decline, many individuals need daily support, 6 while caregivers take on more responsibility for safety and daily tasks. 7

Social determinants of health (SDoH) contribute to ADRD progression and care quality. 8 9 Housing instability, food insecurity, and financial strain are associated with worse cognitive decline, 10 11 12 increased fall risk, 10 11 12 and more preventable hospitalizations in patients living with dementia (PLWD). 13 Despite these risks, SDoH often go unaddressed. Existing screening tools lack validation for dementia-specific outcomes, 14 15 16 17 18 19 and few digital tools support SDoH data in dementia care. 20 Further, clinicians often lack SDoH information during care, 21 and many report uncertainty about how to address social needs. 22

Digital tools support social risk screening and consolidate SDoH information for clinical decision-making. Building on the polysocial risk score paradigm, which emphasizes interconnected social risks, 23 24 25 26 we previously developed an AI-driven, individualized polysocial risk score (iPsRS) in diabetes. 27 For the current study, the iPsRS was developed and internally validated using historical electronic health record (EHR) data, specifically among patients living with ADRD (manuscript in progress). The model integrates individual-level SDoH from clinical documentation (e.g., housing stability, food security, financial strain) and contextual-level SDoH (e.g., neighborhood environment, local resource access). The model estimates 6-month risk of hospitalizations, falls/fractures, and mortality, identifies each factor's contribution using SHAP (SHapley Additive exPlanations) values, and highlights focal social risk drivers for clinician attention. Fairness was assessed during model development by comparing prediction errors across racial and sex subgroups. 28

In this paper, we describe the co-design of the I ntelligent S ocial Risk M anagement In A lzheimer ' s Disease And Alzheimer ' s Disease- R elated Demen t ias PATIENTS (iSMART) prototype, an early-stage clinical decision support (CDS) tool that integrates the iPsRS model into the EHR workflow. iSMART aggregates key SDoH in a single interface, estimates 1-year hospitalization risk in PLWD, highlights key SDoH drivers, and links patients to community resources. By bringing social risk information into the clinical workflow, iSMART helps providers identify PLWD at high risk for adverse outcomes and address priority social needs. A prior design phase used input from three physicians and two case managers to guide early conceptual development. 28 Building on this work, we engaged a broader sample to co-design the prototype.

Methods

We conducted a user-centered design (UCD) study, which engages end users to align tools with needs and workflows and improves usability and integration of digital tools into practice. 29 We describe this as co-design because participants provided feedback that will directly inform iSMART development, including features, data elements, potential users, workflow placement, and implementation strategies.

Participant Recruitment

Participants were recruited via flyers and listservs from a large academic health system in the southeastern United States. Interested individuals completed an eligibility screener survey with automated screening questions. The study team reviewed responses and confirmed eligibility. Inclusion criteria were: patient-facing healthcare professional (providers, nurses, social workers, or other patient care roles) in the health system; worked in primary care/family medicine, neurology, or other relevant specialty; age 18 to 89; and involved in the outpatient care of individuals with ADRD. Exclusion criteria included: not managing or treating patients with dementia or cognitive decline, or working exclusively in inpatient settings. Those confirmed eligible were contacted to confirm interest and schedule interviews.

Interview Procedures and iSMART Prototype

We conducted 45-minute virtual interviews that began with oral consent and used a semi-structured guide. Five team members trained in qualitative methods, implementation science, and clinical informatics conducted interviews. Participants discussed social risks in dementia care and reviewed the iSMART prototype, including the main iPsRS interface ( Fig. 1 ) and mockups of the iPsRS embedded into the EHR. The upper half displayed demographics and editable dropdown fields for social factors. The prototype displayed iPsRS-estimated hospitalization risk through bars and a donut chart summarizing clinical and social contributions. The lower portion displayed the top five social risk drivers, along with clickable links to relevant community resources through FindHelp, a national platform that links users to verified social service programs such as housing, food, and transportation. 30 These links were included within iSMART and directed users to the FindHelp page to identify resources relevant to the patient's specific needs. Participants commented on suggested end users, use cases, features, and implementation considerations. Interviews were recorded, transcribed, and de-identified. Participants received a $35 gift card.

Fig. 1.

Prototype of the main interface of the Intelligent Social Risk Management in Alzheimer's Disease and Alzheimer's Disease-Related Dementias Patients (iSMART) tool used in user-centered design (UCD) testing.

Data Analysis and Results Synthesis

We used rapid qualitative analysis to support initial content analysis across interviews. Transcripts were uploaded to Dedoose for thematic analysis. The same five research team members who conducted the interviews, along with one additional member, coded and analyzed the transcripts. Three transcripts were jointly coded, and the codebook was refined. Coding pairs then double-coded the remaining six transcripts and reached consensus on discrepancies.

Results

Participant Characteristics

We interviewed nine participants, comprising six providers, two nurses, and one social worker ( Table 1 ). Participants had between 6 months and 35 years of experience. Most reported 5 to 20% of their patient panel living with dementia, except one clinical neuropsychologist, who reported approximately 67%.

Table 1. Participant professional roles ( N  = 9) .

Category	Professional role	n (%)
Providers	Physician (Family Medicine)	3 (33.3)
	Physician (Neurology)	1 (11.1)
	Nurse practitioner (Family Medicine)	1 (11.1)
	Clinical neuropsychologist	1 (11.1)
Clinical staff	Nurse (Outpatient Internal Medicine)	2 (22.2)
	Social worker	1 (11.1)

Note: Categories are grouped for clarity. Percentages reflect the proportion of the total sample ( N  = 9).

Patient Characteristics and Current Practices

Social Risk Factors

Participants described several social challenges faced by PLWD that should inform iPsRS and iSMART ( Table 2 ). A prominent risk factor was lack of caregiver or family support, as one participant stated, “ knowing who provides care in the home and whether anyone does, family or private caregiver, and what kind of assistance they provide, feeds into almost any outcome you look at ” (Nurse 1). Financial strain and transportation issues also frequently emerged. Participants noted that risk factors often interact to influence outcomes for PLWD. For example, loss of driving ability had less effect when a reliable caregiver was able to arrange transportation.

Table 2. Common social risk factors for patients living with dementia identified by participants.

Domain	Social risk factor	n (%)
Social support	Lack of social support/caregiver strain	6 (66.7)
Economic needs	Financial strain	5 (55.6)
	Food insecurity	2 (22.2)
	Affording medications	1 (11.1)
	Housing instability	1 (11.1)
Access to care	Transportation limitations	5 (55.6)
	Insurance challenges	3 (33.3)
	Limited access to healthcare	2 (22.2)
	Limited access to long-term care	1 (11.1)
Health management	Medication adherence issues	3 (33.3)
	Comorbid conditions	2 (22.2)
	Feeding or nutrition difficulties	1 (11.1)
Sociodemographic context	Education level	2 (22.2)
Safety/environment	Safety or security concerns	1 (11.1)
Behavioral health	Mental health concerns	1 (11.1)
	Substance use or abuse	1 (11.1)
	Smoking	1 (11.1)

Note: Participants could indicate more than one factor. Counts and percentages reflect how many participants mentioned each risk.

A core theme was the need to include caregiver context in the tool interface to interpret social risks accurately. Respondents recommended structured fields for caregiver identity, stability, availability, level of support, and the home situation (e.g., lives alone or with family). Participants described caregiver support as central for PLWD, saying a “ social support system is able to catch things sooner and help protect them from falls or malnourishment, dehydration, and those other effects ” (Family Medicine Physician 1) and warned “ if they don't have the support, they might miss appointments or not take medications ” (Nurse 1).

Current Practices in Dementia Care

Participants identified social risks through patient and family conversations, chart review, patient-completed SDoH forms, and other EHR documentation. As one explained, “ we have our SDoH forms that we have patients fill out […] and then, sometimes I just ask ” (Family Medicine Physician 3). When social needs are identified, providers typically coordinate with families or refer patients to social workers, health coaches, or nurse coordinators. These team members often use FindHelp to locate community resources and then contact the referring clinician to authorize support, such as home care. A provider described how this process unfolds:

“ Assessing for the need for either additional help at home or getting someone into assisted living facility due to difficulties with managing medications… not being able to safely prepare food at home, wandering for sure. So kind of trying to gauge those things and counsel patient and care partners about the urgency of getting additional help and then connecting them with our social worker ” (Neurologist).

Clinician Perspectives on the iSMART Prototype

All participants were willing to use iSMART in practice. A nurse shared it “ would benefit every single aspect ” (Nurse 2), another noted it “ can really be applied for anybody ” (Nurse 1), and a neurologist emphasized its value for directing effort, stating it “ would help to focus the efforts of what's the first step or the highest priority thing to work on .”

Participants explained how the visuals could guide decision-making, stating “ having it in a quick visual summary ” made information easier to grasp (Neurologist). Participants viewed the clinical versus social risk chart as useful for prompting targeted follow-up questions. The bottom section, which lists social risk drivers, was considered valuable for triaging needs and tailoring discussions to dominant social risks. Some also valued comparing prior and current iPsRS scores over time. Hospitalization risk was viewed as the most important outcome (e.g., “ prediction of hospitalization is incredibly important ,” Family Medicine Physician 2). Participants noted that high social contributions to hospitalization risk would prompt additional questions about social support alongside clinical actions (e.g., “ if I see a massive social component of high risk of hospitalization […] I also might ask a few extra questions about their social support ,” Family Medicine Physician 1). Several participants proposed fall risk as a secondary output, such as when a patient is at “ 90% risk of a fall ” (Family Medicine Physician 1).

Suggested Users

Participants identified primary care and social care as suggested primary users, but emphasized that “ anybody who touches this patient ” should have access to iSMART (Family Medicine Physician 3). Another participant recommended that caregivers access the tool to help communicate needs. Table 3 summarizes the suggested user groups.

Table 3. Suggested end users of the iSMART tool as specified by participants.

Category	Suggested role	n (%)
Provider	Primary care provider	4 (44.4)
	Neurologist	1 (11.1)
	All providers responsible for patient care	1 (11.1)
Nurse	Clinic nurse	3 (33.3)
	Nurse health coach	1 (11.1)
	Nurse coordinator	1 (11.1)
Other roles	Social worker	4 (44.4)
	Case manager	1 (11.1)
	Medical assistant	1 (11.1)
	Patient or caregiver	1 (11.1)
	Everyone who sees the patient	2 (22.2)

Abbreviation: iSMART, Intelligent Social Risk Management in Alzheimer's Disease and Alzheimer's Disease-Related Dementias Patients.

Note: Participants could suggest more than one end user. Counts reflect the number of participants who mentioned each role.

Implementation Facilitators

Participants mentioned potential barriers to iSMART implementation, including insufficient time in clinical workflows, concerns about workflow disruptions, and alert fatigue and cognitive load. To overcome these barriers, the following implementation facilitators were suggested.

Auto-population, Data Entry, and Accuracy

To address time constraints and workflow burden, participants emphasized that the future iSMART tool should auto-populate EHR data, rather than requiring manual entry, making it more “ widespread, usable, and a lot easier to implement, ” and “ for every extra click I have to do with this, the less chance that this gets a quality complete run through ” (Family Medicine Physician 1). Several favored the drop-down menus, as this reduces “ additional electronic charting ” (Family Medicine Nurse Practitioner). However, participants also noted concerns about outdated or changing social data. Most supported a hybrid approach combining auto-filled data with editable fields.

Preferred Tool Launch Location in the EHR

Recognizing workflow challenges, most participants preferred launching iSMART from the Epic Snapshot view ( Tables 4 and 5 ). A social worker noted, “ the Snapshot will be the first one that will show up when they open a patient's chart ,” describing it as the logical place to display social and clinical risk information (Social Worker).

Table 4. Participant preferences for iSMART tool launch location in the Epic electronic health record.

Location	n (%)
Snapshot	7 (77.8)
Assessment	4 (44.4)
Rooming	4 (44.4)
Plan	2 (22.2)

Abbreviation: iSMART, Intelligent Social Risk Management in Alzheimer's Disease and Alzheimer's Disease-Related Dementias Patients.

Note: Participants could suggest more than one preferred location. Counts reflect how many participants mentioned each option.

Table 5. iSMART tool launch location preferences by profession.

	Desired EHR launch location
Participant	Snapshot	Rooming	Assessment	Plan
Primary care provider a ( n  = 4)	n  = 3 (75%)	n  = 1 (25%)	n  = 2 (50%)	n  = 2 (50%)
Nurse ( n  = 2)	n  = 2 (100%)	n  = 1 (50%)	n  = 1 (50%)	n  = 0 (0%)
Social worker ( n  = 1) b	n  = 1 (100%)	n  = 0 (0%)	n  = 1 (100%)	n  = 0 (0%)
Specialty physician c ( n  = 2)	n  = 1 (50%)	n  = 2 (100%)	n  = 1 (50%)	n  = 0 (0%)

Abbreviations: EHR, electronic health record; iSMART, Intelligent Social Risk Management in Alzheimer's Disease and Alzheimer's Disease-Related Dementias Patients.

Notes: In all other instances where >1 locations were indicated by a participant, there was no clear stated preference for one location over the other.

Colors correspond with percentage of participants by role who indicated preference for each EHR launch location; Dark green = 100%, light green = 75%, yellow = 50%, orange = 25%, red = 0%.

^aTwo primary care providers explicitly stated they would NOT want Rooming to be the location.

^bSocial workers indicated they preferred Snapshot, with Assessments as a second option.

^cOne specialist suggested Assessment for specialists, but Rooming for primary care providers.

Alert Preferences

To mitigate alert fatigue and cognitive overload, all participants supported the use of a Best Practice Alert (BPA; Epic's proprietary decision support system that triggers automated pop-up notifications based on pre-programmed criteria), linked to iSMART, but with careful calibration. Several emphasized appropriate alert frequency, ensuring “ the threshold is set at the right level for the alert ” (Neurologist). However, others also raised concerns about alert fatigue, explaining that too many and overly frequent prompts would lead them to “ stop paying attention to it ” (Family Medicine Physician 3) or become “ less likely to be read […] and click past things ” (Family Medicine Physician 1). Another stressed verifying content before acting on recommendations. As a result, many participants viewed the BPA as a “ recommendation ” (Nurse 1) or “ additional piece of information ” (Family Medicine Physician 1) rather than a directive.

Clear Visuals

Participants emphasized risk visualization. Some participants asked to include percentages on the pie chart for quick interpretation, for example, “ in a percentage style, because it is easier to gauge ” (Nurse 2). Another found the two pie charts confusing and suggested simplifying.

Implementation Barriers

Despite strong interest, participants described potential limits to using iSMART in routine care. The clinical neuropsychologist highlighted limited capacity to address identified social needs appropriately, stating, “ I would certainly like to help this person, but my capacity to impact housing instability is probably not as great as, say, social work resources, ” reinforcing the importance of team-based follow-up. Another emphasized time constraints in daily practice, stating, “ I'm seeing patients the whole 40-hour work week. So, to try and find out about resources in my spare time, I don't have that spare time ” (Family Medicine Physician 3). Participants also believed iSMART may be less effective during initial visits without sufficient EHR data to calculate the iPsRS, in specialty consultations with limited long-term responsibility, or during encounters focused on mild problems.

Perspectives on Workflow Integration

The following use cases reflect how participants envisioned incorporating iSMART into workflows across clinical roles. Of note, the specialist in our sample (neuropsychologist) indicated the tool would be less useful to them than the feedback received from the other participants; therefore, workflows primarily focus on primary care and social work/care coordinators, with participants in these areas better envisioning uses for iSMART.

Primary Care Provider Workflow

Participants highlighted the importance of pre-populated, up-to-date data and the iPsRS available in the EHR before an encounter. A physician mentioned that seeing it during pre-charting would “ help paint a picture of what I should be thinking of when I walk into the room ” (Family Medicine Physician 1). Information could be collected from the EHR (auto-filled), from caregivers prior to the visit (e.g., via MyChart or phone), or by clinical staff (e.g., nurse or medical assistant) during rooming. The neurologist noted that “ if the information is gathered by the medical assistant […] it would potentially save me time if I'm able to see it quickly as I'm previewing the chart .”

Providers may review iSMART during pre-charting or the encounter, use a BPA for high-risk patients, and use identified risks to guide conversations, coordinate services, or refer to social workers, nurse health coaches, or nurse coordinators. Some would use iSMART for any dementia visit, while others would use it only for high-risk or recently discharged patients.

Social Work and Care Coordination Workflow

For social workers and nurses, the typical workflow would begin with a referral from the primary provider. Before or during the patient interaction, they indicated that they would access iSMART in the EHR to review the iPsRS, identify the most pressing social factors, and plan resources. A physician described a scenario in which care coordinators could add “ this risk calculator into their note ” so that primary care sees the score later and uses it to “ follow up on some of those especially social concerns ” after hospitalization (Family Medicine Physician 3). These team members could then help connect the patient to appropriate services, often via FindHelp. If the plan involves medical equipment or home health, the provider would be consulted to order.

Discussion

To our knowledge, design requirements for translating AI social risk prediction into dementia-specific EHR workflows have not previously been established. General SDoH CDS tools often emphasize universal screenings without dementia-specific risk prediction. In contrast, iSMART targets dementia care and links an individualized iPsRS to 1-year hospitalization risk. The prototype connects risk outputs with social factor summaries and FindHelp resources to help providers and staff move from risk recognition to resource referral. These findings advance clinical informatics by showing how an AI-driven social risk model can be translated into a clinician-facing CDS tool that supports interpretation, prioritization, and response in dementia care.

Participants showed strong interest in using iSMART, reflecting the need for automated SDoH tools for PLWD. Participants identified social risks mainly through conversations with patients and families, SDoH forms, and EHR checks, which is consistent with reports of limited routine SDoH screening 16 31 32 and incomplete SDoH EHR documentation. 16 31 iSMART centralizes social factors in one view and can potentially rely on existing EHR data to populate social risk information. Participants valued iSMART because they lack guidance on how to respond once social risks are identified. 15 33 iSMART addresses this by linking social risks to community resources. Because many risks are modifiable, 11 34 35 systematic identification and response could influence outcomes.

Participants described important design priorities and features, such as data fields that update regularly, as social risks can change. 36 37 38 This may only be feasible if social data are routinely and consistently documented. Yet, only 2% of patients are typically screened for SDoH. 31 Participants emphasized the need to correct inaccurate entries, consistent with studies that have shown incomplete and inaccurate SDoH EHR histories. 38 39 40 41 This indicates that iSMART should support a hybrid design with EHR auto-population and editable fields to support accuracy and timeliness. This may be especially important for PLWD, whose health, needs, caregiver support, and living situation may change between encounters and require frequent updates.

Primary care providers and social workers were identified as the most appropriate users, followed by nurses. This aligns with prior studies, which note that primary care is central to addressing medical and social needs in dementia care. 42 43 44 Social workers, geriatric nurse practitioners, or nurses typically serve as coordinators for PLWD, linking to medical and community resources. 45 46 iSMART should therefore primarily be tailored to the needs of social workers and primary care providers.

Preference for the Epic Snapshot view aligns with recommendations to place CDS in common EHR summary views, rather than on separate screens or in messages, to improve efficiency. 47 48 This matters, as over 85% of physicians report having insufficient time or incompatible workflows to address SDoH. 49 At the same time, summary views often already contain many data elements, so adding content requires careful prioritization. Altogether, this indicates that a concise iPsRS section (e.g., most recent risk score) should be displayed on the Snapshot view, linking to the full iSMART application if more information is desired.

Participants supported a BPA but cautioned about alert fatigue. Prior studies have linked alert workload to burnout and lower alert acceptance. 50 51 52 53 Additionally, irrelevant alerts and those with low specificity were linked to high override rates. 54 55 These findings support a conservative BPA design in iSMART, targeting high-risk patients and providing actionable prompts.

Participants also noted limited usefulness of iSMART during initial visits without prior EHR data and in specialty encounters, where there is less capacity to address social needs. 32 56 57 Thus, iSMART may require modified workflows or additional support in some settings.

Findings align with the broader CDS and SDoH screening literature, 14 19 58 but also highlight the need for specialized workflows and PLWD-specific risk models. Lack of caregiver support was identified as the most frequent social risk factor in PLWD, with prior work showing that caregiver stress is associated with higher fall risk for PLWD. 59 60 61 Hospitalization was viewed as the most meaningful outcome. In a preliminary study, we developed an iPsRS using contextual and person-level social risks to predict 1-year hospitalization, falls/fractures, and mortality. Unlike prior polysocial scores developed for mortality or cardiovascular events in general populations, 23 24 25 62 63 our model specifically targets PLWD. Prior iPsRS model development identified housing instability as the strongest predictor of 1-year hospitalization risk. In another review, hospitalization due to dementia was associated with functional decline, increased mortality, and higher readmission rates. 64 65 66 Focusing iSMART on hospitalization risk aligns with clinician priorities and outcomes affecting patients and caregivers.

Limitations include small sample size. Role-specific findings, especially for social workers and nurses, are preliminary and should be validated in larger samples. The wide range of professional experience in our sample (6 months to 35 years) is another potential limitation because early-career and experienced clinicians may differ in perceived utility, feasibility, barriers, and receptivity to social risk screening tools. Additionally, participants were from a single academic health system; therefore, findings may differ in other settings. Because we used a low-fidelity prototype, findings represent anticipated use, not actual use.

In future work, we will extend iSMART beyond a prototype by using the feedback gathered in this study to refine the design and features. Although the primary objective was to inform implementation and workflow integration, findings also identified clinician-prioritized social factors that may inform future iPsRS updates. In particular, participants emphasized caregiver availability, stability, and level of support as important considerations for PLWD, suggesting these factors should be examined in future model refinement. Another direction will be to explore caregivers' perspectives on how to obtain accurate and up-to-date SDoH information, including through patient- or caregiver-completed pre-visit surveys. These could supplement EHR information to enable iSMART functionality at initial patient encounters.

Next, we will build a production-ready iSMART application that integrates with the EHR through FHIR and CDS Hooks, links to FindHelp, and uses dementia- and ADRD-related ICD diagnosis codes to identify eligible patients and trigger iPsRS calculation. The iPsRS will then estimate 1-year hospitalization risk among eligible PLWD. For initial implementation, high risk will be defined as a predicted score in the highest 10% of the model score distribution, corresponding to the 90th percentile or higher. Future work should further examine how providers and clinical staff interpret and trust the AI-driven iPsRS, including how they use model outputs in clinical decision-making. Additional research should test whether iSMART improves outcomes.

Conclusion

Outpatient providers and clinical staff viewed the iSMART prototype as a promising approach to support social risk screening and referral in dementia care. Practical implications include integration into common EHR views, promoting shared use among primary care and social care teams, and including caregiver context, auto-populated SDoH data, and direct links to community resources. These insights provide a foundation for real-world integration of iSMART and adaptation of social risk tools in dementia care. Continued development has the potential to reduce social risks among PLWD.

Funding Statement

Funding Information The project is supported by NIH/NIA (grant no.: R01AG089445).

Clinical Relevance Statement

Our study shows how an EHR-integrated clinical decision support tool can assist healthcare providers in better recognizing and prioritizing social risks that contribute to hospitalization among patients living with dementia. The iSMART tool offers a practical way to translate social risk information into timely referrals and coordinated care that may support improved healthcare outcomes and quality of life.

Multiple-Choice Questions

1. Where did most participants prefer the iSMART tool to be launched within the electronic health record?

   1. Inbox messaging

   2. Best Practice Alert pop-up only

   3. Snapshot view

   4. Discharge summary module

   Correct Answer: The correct answer is option c. The majority of participants preferred the Snapshot view because it aligns with routine chart review workflows and allows clinicians to quickly assess social and clinical risk information at the point of care.

2. Which outcome did participants consider most important for the iSMART prototype to predict among patients living with dementia?

   1. Long-term mortality

   2. Cognitive decline severity

   3. 1-year hospitalization risk

   4. Medication adherence

   Correct Answer: The correct answer is option c. Interviewees viewed prediction of hospitalization risk as the most important outcome because it informs real-time clinical decision-making. Participants described that seeing elevated hospitalization risk, particularly when driven by social factors, would prompt additional questions about social support and guide referrals alongside clinical care.