Documented follow‐up to memory concerns reported at the Medicare Annual Wellness Visit  

Abstract

INTRODUCTION

The Medicare Annual Wellness Visit (AWV) may improve timely detection of Alzheimer's disease and related dementias (ADRD), yet little is known about the frequency of follow‐up on patient‐reported memory concerns during the AWV.

METHODS

We use electronic medical record (EMR) data from an academic health system to examine EMR‐documented follow‐up actions for patients with newly reported memory concerns on the AWV health risk assessment, including formal cognitive assessment or specialist referrals.

RESULTS

The 1411 patients with newly reported memory concerns were predominantly white (70%) and female (63%), with an average age of 78 (SD 7.5). At the AWV, few patients received a cognitive assessment (5.4%; n = 76), specialist referral (2.1%; n = 30), or both (0.4%; n = 6). In adjusted analyses, we did not observe statistically significant differences by sociodemographic characteristics.

DISCUSSION

This EMR‐based study highlights an opportunity to better leverage the AWV to improve ADRD detection and care.

Highlights

• The Medicare Annual Wellness Visit (AWV) offers opportunities to assess cognitive function and initiate appropriate follow‐up.

• About 8% of patients reporting memory concerns at the AWV received cognitive assessment/specialist referral.

• About 26% completed advanced care planning at the AWV.

• Findings highlight opportunities to enhance dementia care via the AWV.

1. BACKGROUND

The long course and devastating effects of Alzheimer's Disease and Related Dementias (ADRD) ¹ amplify the importance of early diagnosis and proactive care planning, however, just half of older adults with ADRD receive a formal diagnosis. ² The Medicare Annual Wellness Visit (AWV) is a reimbursed preventive care visit ³ that involves patient completion of a health risk assessment with questions regarding subjective memory concerns, and a requirement for “direct observation” of cognitive function. ³ , ⁴ Thus, the AWV presents an opportunity to initiate evaluation for ADRD at the point of care: when patients are actively engaged and have independently expressed concerns about their memory, they may be receptive to follow‐up, ³ , ⁴ including diagnosis, identification of caregivers, and engagement in care planning. ⁵ , ⁶ , ⁷

Prior work suggests receipt of an AWV is associated not only with increased rates of diagnosis of mild cognitive impairment (MCI) and ADRD ⁸ , ⁹ , ¹⁰ but also earlier ADRD diagnosis. ⁸ To facilitate early ADRD detection, an expert workgroup convened by the Alzheimer's Association recommended that indication of memory concerns during an AWV be followed by a brief structured cognitive assessment, triggering a subsequent full dementia evaluation when relevant. ¹¹ Other consensus bodies have emphasized the need for proactive advance care planning (ACP) in the presence of emerging cognitive concerns, while a patient is still able to actively participate. ⁵ , ⁶ , ⁷

A prior study found that endorsement of memory concerns on the health risk assessment is associated with a higher likelihood of referral to neurology or memory care, as well as increased odds of ACP completion. ¹² Yet, little is known about how frequently formal cognitive assessment occurs during the AWV as prior studies rely on self‐reported survey data ¹³ , ¹⁴ rather than real‐world clinical documentation. This study used electronic medical record (EMR) data from a large academic health system to (1) to determine the frequency with which recommended follow‐up actions (cognitive assessment and specialist referral) are undertaken after patients report new memory concerns as part of the health risk assessment during an AWV, and (2) to examine how patient characteristics differ between those who do and do not receive follow‐up actions.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed literature found in electronic databases (e.g., PubMed, Google Scholar) focused on the Medicare Annual Wellness Visit (AWV). Previous research finds that the AWV is associated with increased likelihood of diagnosis of Alzheimer's disease and related dementias (ADRD). Little is known about how memory concerns expressed by patients during the AWV are followed up on in everyday clinical practice.

2. Interpretation: This study is the first to use electronic medical record (EMR) data to examine follow‐up actions taken in response to memory concerns expressed on the health risk assessment during the AWV. We found follow‐up actions like cognitive assessment and specialist referrals are documented in the EMR infrequently.

3. Future directions: Health systems aiming to leverage the AWV to improve ADRD care may benefit from (1) training clinicians or other key care team members to administer and interpret cognitive assessments and (2) identifying workflows for appropriate follow‐up to memory concerns, such as future visits focused on ADRD care planning or referral pathways to memory‐focused specialist providers.

2. METHODS

2.1. Study sample

We used EMR data from a large academic health system spanning a 5‐year period (October 2017 to October 2022). ¹⁵ , ¹⁶ , ¹⁷ Our sample included patients 66 years old or older who had at least two primary care encounters within a 2‐year period, to represent established patients of the health system. Because we were interested in identifying follow‐up actions for newly reported memory concerns, we excluded patients who had a previous diagnosis of MCI or ADRD based on International Classification of Diseases ‐ 10 Revision (ICD‐10) codes (Appendix Table A.1). We included only patients who had an AWV (Appendix Table A.2), based on Healthcare Common Procedure Coding System HCPCS codes G0438 or G0439, during which they indicated a subjective memory concern on the health risk assessment, based on an affirmative response (yes vs. no) to at least one of two questions: “Do you have new memory problems in the past year?” and “Do you have trouble thinking clearly or making decisions?” To identify new (versus ongoing) subjective memory concerns, we included only the first AWV with a recorded affirmative response to at least one of these questions. Inclusion in the sample required a “washout” period of “no” responses before identification of the first affirmative report. As such, patients whose first affirmative response occurred at their first AWV in our dataset were excluded (Figure 1).

FIGURE 1.

Study sample selection flow diagram. *Alzheimer's disease and related dementias (ADRD); mild cognitive impairment (MCI). †To identify new (versus ongoing) memory concerns, we included only the first Medicare Annual Wellness Visit (AWV) with a recorded affirmative response to at least one of these questions. Inclusion in the sample required a “washout” period of “no” responses before identification of the first affirmative report, given that if our first observed AWV had an affirmative response to memory concern related questions we would not be able to establish if these were new or ongoing memory concerns.

2.2. Primary outcomes

We examined documentation of two recommended actions at the AWV with newly reported subjective memory concern: a structured cognitive assessment or referral to a specialist, memory care‐relevant department (memory clinic, neurology, or psychiatry). We examined receipt of the two actions as a combined outcome and as separate outcomes, recognizing that primary care providers may not always have time to conduct an assessment during the visit or may not have a high level of comfort with administering and interpreting a cognitive assessment, in which case a referral is appropriate.

Documentation of a structured cognitive assessment during the AWV was based on documented results from the Mini‐Mental State Examination (MMSE), Mini‐Cog, or Montreal Cognitive Assessment (MoCA) in clinical notes. We selected these assessments based on consultation with clinicians who specialize in geriatric medicine and work within the health system (J.C., S.N., A.G.), to capture the commonly used cognitive assessments in the health system. Using a library for natural language processing in Python (the spaCy module), we extracted sentences from visit notes that referenced assessment names, abbreviations, or related terms (e.g., “clock draw”). Employing qualitative content analysis, one team member manually reviewed extracted sentences to confirm whether a cognitive assessment was administered based on the note. A second team member independently reviewed 15% of the same notes to verify the accuracy of the initial review. No disagreement between reviewers occurred.

Referrals to ADRD‐related care departments (i.e., memory care, geriatric psychiatry, or neurology) were based on manual identification of relevant department names. To improve the completeness of referral identification, we also analyzed all clinical notes for mentions of referrals to relevant departments. Using the Python spaCy module, we extracted sentences containing the terms “refer” or “referral”. One team member manually reviewed these sentences to determine whether the referral was directed to one of the three target departments, with a second team member independently verifying 15% of sentences to ensure accuracy, with one disagreement between reviewers which was reconciled via discussion.

2.3. Secondary outcomes

As a secondary outcome we examined ACP completion, as it is a required component of the AWV and best practice to initiate early after detection of memory concerns. ³ , ⁵ ACP completion was based on ACP Current Procedural Terminology (CPT) codes (Appendix Table A.3) documented during the AWV. Newly documented MCI or ADRD diagnosis during the AWV or in the year following the AWV was identified using ICD‐10 codes as previously described. We included MCI and ADRD diagnosis as a secondary outcome because we did not expect MCI or ADRD to be formally diagnosed during the AWV, but we anticipated that clinicians might add a diagnostic code for these conditions to initiate further diagnostic evaluation or to document a diagnosis made previously outside of the health system, based on patient or care partner report during the visit. We included the first instance of ICD‐10 codes from the visit diagnoses and the problem list, entered either on the date of the AWV or within the following year (365 days).

2.4. Patient characteristics and other covariates

Patient demographic information originated from the EMR and included age (years) at the time of AWV, sex (female or male), race (White or Other: Black, Asian/Pacific Islander, or another race), ethnicity (non‐Hispanic or Hispanic), English as preferred language (yes or no), and marital status (married or not married/other: widowed, divorced, separated, single, unknown). We also identified if the patient was enrolled in traditional Medicare or Medicare Advantage at the time of the AWV. Patient addresses including nine‐digit zip codes were used to calculate the Area Deprivation Index (ADI) ¹⁸ , ¹⁹ as a proxy measure of socioeconomic disadvantage; ADI scores range from 0 to 100, with higher scores indicating higher disadvantage. We also relied on ICD‐10 codes to calculate a summary measure of health status (comorbidity), the Elixhauser Comorbidity Index. ²⁰

2.5. Analytic approach

We described patient characteristics for the full sample and reported the frequency of each follow‐up action. Next, we compared characteristics of patients who received cognitive assessment/specialist referral and those who did not. Unadjusted differences in characteristics were assessed using chi‐squared tests or t‐tests, as appropriate. Logistic regression was used to assess the odds of receiving a structured cognitive assessment or specialist referral as a combined outcome and separate outcomes. adjusting for patient characteristics. We conducted sensitivity analyses using Firth's penalized logistic regression, which is recommended for small sample sizes; ²¹ however, given similar results we report results from logistic regressions due to the broader familiarity and ease of interpretation.

3. RESULTS

Based on our inclusion criteria, the sample included 1411 unique patients who expressed new cognitive concerns on the health risk assessment as part of an AWV. As shown, in Table 1, patients were predominantly White (69.7%), female (63.3%), and not married (56.1%) with an average age of 78 (SD 7.48). The average Elixhauser Comorbidity Index score was 5.1 (SD 2.7) (moderate comorbidity), and the average ADI was 62.7 (SD 22.8) out of a maximum possible ADI of 100.

TABLE 1.

Patient characteristics, comparing patients who received a cognitive assessment or referral to patients who did not

Parameter	Total sample N (%)	Cognitive assessment/referral N (%)	No cognitive assessment/referral N (%)	p‐Value *
Total no. of patients	1411 (100)	112 † (7.9)	1299 (92.1)
Age at first AWV, mean (SD)	78.4 (7.5)	78.3 (7.1)	78.4 (7.5)	0.867
Sex
Male	518 (36.7)	38 (33.9)	480 (37.0)	0.524
Female	893 (63.3)	74 (66.1)	819 (63.1)
Race
White	983 (69.7)	83 (74.1)	900 (69.3)	0.287
Other ‡	428 (30.3)	29 (25.9)	399 (30.7)
Ethnicity
Non‐Hispanic	1341 (95.0)	108 (96.4)	1233 (94.9)	0.480
Marital status
Married	613 (43.4)	45 (40.2)	568 (43.7)	0.467
Other/not married §	798 (56.6)	67 (59.8)	731 (56.3)
Medicare coverage type at first AWV
Traditional Medicare	1151 (82.3)	87 (77.7)	1064 (82.7)	0.184
Medicare Advantage	248 (17.7)	25 (22.3)	223 (17.3)
Comorbidity score, mean (SD)	5.2 (2.8)	5.3 (3.2)	5.1 (2.7)	0.537
Area Deprivation Index, mean (SD)	62.6 (23.1)	62.1 (22.9)	62.6 (23.1)	0.816

Abbreviations: AWV, Annual Wellness Visit; SD, standard deviation

^*Test of statistical significance for continuous variables were calculated using t‐tests and for categorical variables using chi‐squared tests.

^† Seventy‐six patients (5.4%) received cognitive assessment only, 30 (2.1%) patients received referral only, and six patients (0.4) received both.

^‡ The “Other” category under race includes patients who identify as Black, Asian/Pacific Islander, or any other race.

^§ The “Other/Not married” category under Marital Status includes widowed, divorced, separated, single and unknown.

In total, 7.9% of patients reporting new subjective memory concerns in the health risk assessment were documented as having received a structured cognitive assessment or being referred to a relevant specialist, with 5.4% receiving a structured cognitive assessment, 2.1% receiving a referral, and 0.4% receiving both (Table 1). In unadjusted analyses, patients who did (versus did not) receive a cognitive assessment/specialist referral did not differ across sociodemographic characteristics. More than one in four patients (26.0%) reporting new subjective memory concerns were found to have ACP as part of the AWV (Appendix Table A.4).  Few patients had an MCI or ADRD diagnosis documented at the AWV (3.2%; 45/1411) or in the year following the AWV (5.0%; 70/1411) (Appendix Figure A.1).

In adjusted analyses, while non‐White patients had lower odds of receiving a cognitive assessment (adjusted odds ratio [aOR] = 0.66, 95% confidence interval [CI[: 0.39–1.12) and slightly higher odds of receiving a specialist referral (aOR = 1.20, 95% CI: 0.59–2.41), group differences were not statistically significant (Table 2). Patients enrolled in Medicare Advantage had higher odds of both cognitive assessment (aOR = 1.31, 95% CI: 0.75–2.29) and specialist referral (aOR = 1.51, 95% CI: 0.70–3.27), though estimates were imprecise. We observed similar findings when examining cognitive assessment/specialist referral as a combined outcome (Table 3).

TABLE 2.

Adjusted odds ratio from models assessing receipt of cognitive assessment (model 1) and specialist referral (model 2)

	Model 1: Adjusted odds ratios for receipt of cognitive assessment			Model 2: Adjusted odds ratios for receipt of specialist referral
Patient characteristics	Odds ratio	p‐Value	95% CI	Odds ratio	p‐Value	95% CI
Age (years)	1.01	0.57	0.98–1.04	0.98	0.37	0.93–1.03
Sex
Female (reference: male)	0.82	0.42	0.51–1.32	0.78	0.50	0.38–1.61
Race
Other * (reference: White)	0.66	0.12	0.39–1.12	1.20	0.62	0.59–2.42
Medicare type
Medicare Advantage (reference: traditional Medicare)	1.31	0.34	0.75–2.29	1.51	0.30	0.70–3.27
Comorbidity score (Elixhauser)	1.02	0.59	0.94–1.11	0.99	0.82	0.87–1.11
Area Deprivation Index	1.00	1.00	0.99–1.01	1.00	0.49	0.98–1.01

Abbreviation: CI, confidence interval.

^*The “Other” category under race includes patients who identify as Black, Asian/Pacific Islander, or any other non‐White race.

TABLE 3.

Adjusted odds ratio from model assessing receipt of cognitive assessment/specialist referral (combined outcome).

Patient characteristics	Odds ratio	p‐Value	95% CI
Age (years)	0.99	0.77	0.97–1.02
Sex
Female (reference: male)	0.89	0.62	0.58–1.38
Race
Other * (reference: white)	0.74	0.17	0.47–1.16
Marital status
Other/not married (reference: married)	1.13	0.56	0.74–1.72
Medicare type
Medicare advantage (reference: traditional Medicare)	1.42	0.15	0.89–2.28
Comorbidity score (Elixhauser)	1.03	0.47	0.96–1.10
Area deprivation Index	1.00	0.80	0.99–1.01

Abbreviation: CI, confidence interval.

^*The “Other” category under race includes patients who identify as Black, Asian/Pacific Islander, or any other non‐White race.

4. DISCUSSION

Using EMR data from a large academic health system, we found that structured cognitive assessments or referrals to specialists were documented in the EMR infrequently (∼8% of AWVs) following newly reported memory concerns during the AWV and did not statistically significantly vary by patient characteristics. Roughly 1 in 4 patients with newly reported memory concerns received ACP Approximately 3% of patients had a new diagnosis of MCI or ADRD added at the time of the AWV and ∼5% had a new diagnosis added in the year following the AWV.

Performing a structured cognitive assessment is recommended by the Alzheimer's Association as a best practice for responding to patient‐reported memory concerns during an AWV. ¹¹ To our knowledge, this is the first study to use real‐world EMR data to examine how often such assessments occur in practice. Consistent with anecdotal evidence ²² and prior survey‐based research, ¹³ , ¹⁴ we find that structured cognitive assessment is rarely performed during the AWV. Several factors may contribute to our findings. First, the health risk assessment includes questions about many risk factors (e.g., mental health, diet and nutrition, physical activity, and mobility ¹⁷ ), which may take priority over subjective memory concerns. Primary care clinicians also report gaps in training related to administering cognitive assessments and interpreting results or developing an ADRD care plan following an ADRD diagnosis. ²³ , ²⁴ The lack of a standardized field in the EMR to record the completion and results of cognitive assessments makes it difficult to determine whether an assessment was performed previously, to track changes over time, and to ensure appropriate follow‐up was taken. Regarding the similarly low rate of referrals, long waitlists for memory clinics or neurology ²⁵ may prevent patients from accessing specialists when referred. The low frequency of follow‐up actions identified in our analysis reinforces the potential benefit of supporting primary care with training, education, and development of new workflows to be able to effectively act on indications of subjective memory concerns during the AWV. ²⁴ , ²⁶ , ²⁷ , ²⁸

Our study reinforces the importance of dementia‐capable initiatives, ²⁷ such as designating trained staff who are equipped to administer and interpret results from cognitive assessment and to provide needed support and care planning, under clinician supervision. Pharmacist‐ or nurse‐led AWVs are one approach to increase AWV uptake and have demonstrated benefits for patient satisfaction and care experience, ²⁹ , ³⁰ , ³¹ , ³² suggesting promise for adapting them to enhance ADRD care. The Cognitive Assessment and Care Planning (CACP) visit offers dedicated time for assessment of cognitive function and relevant care planning after cognitive issues are identified. ³³ An ADRD nurse champion could facilitate scheduling a CACP visit as follow‐up to memory concerns identified during the AWV, serving as a key step in a comprehensive workflow to improve ADRD care within primary care. ³⁴ To be effective, this approach should be coupled with training and education in the administration and interpretation of cognitive assessments and ADRD management.

Enhancements to the EMR could improve AWV workflows. For example, electronic completion of the health risk assessment prior to the AWV could be used to flag patients with cognitive or memory concerns who may especially benefit from structured cognitive assessment. A structured EMR field to document the completion of cognitive assessments would enable tracking of scores over time and allow health systems to monitor assessment rates, supporting efforts to increase appropriate follow‐up.

ACP was documented in about a quarter of AWVs. This rate of ACP is consistent with other findings focused on ACP at the AWV, ³⁵ but notably higher than the ACP rates reported in previous studies examining a broader range of visits among Medicare beneficiaries. ³⁶ , ³⁷ , ³⁸ The care planning focused nature of the AWV coupled with indication of memory related concerns at the AWV may prompt ACP discussion to a greater extent than other types of visits. Yet, in part due to stringent time‐based billing requirements, even if ACP conversations are occurring during a visit, ACP is often underbilled. ³⁹ , ⁴⁰ , ⁴¹ This presents a valuable opportunity for future research to understand whether and how ACP conversations occur during the AWV but may not be formally documented or billed. Prioritizing the integration of ACP conversations into the dementia care planning process, especially early in the disease trajectory while patients can still actively participate in ACP conversations is a best practice ⁴² and would reduce missed opportunities for shared decision‐making.

With regard to limitations, this analysis draws on data from a single health system and may not be representative of other health systems in the United States. Although we sought to comprehensively capture follow‐up actions, we could not determine whether informal cognitive assessments were conducted during the AWV through clinician observation. Even when subjective memory concerns are reported on the health risk assessment, clinicians may judge that formal screening is unnecessary at the time. Clinician‐level factors likely influence the content of AWV conversations and follow‐up actions, but clinician‐level information was not available in our dataset; therefore, we are unable to comment on this important point. Finally, changes in care following indication of memory concerns may occur gradually; because our analysis focused only on the first AWV with documented memory concerns, follow‐up actions at later visits were not captured.

5. CONCLUSIONS

Our study identifies opportunities to enhance AWV workflows and appropriate follow‐up to facilitate high quality ADRD care. Health systems can leverage EMR data to track cognitive assessment completion and identify opportunities for reimbursable follow‐up visits focused on ADRD‐focused care planning and coordination. ²⁷ , ⁴³ Achieving these improvements will require institutional buy‐in, as well as adequate training and support for clinicians and care teams in ADRD diagnosis and management.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

This study involved analysis of secondary data and did not require informed consent. The study was approved by the Johns Hopkins Medicine Institutional Review Board (IRB00303190).

Supporting information

Supporting information: alz71576‐sup‐0002‐Appendix

Supporting information: alz71576‐sup‐0001‐ICMJE