Validity evidence for the Patient Reported Outcome Measurement Information System (PROMIS)^® Cognitive Screener (PRO-CS) to detect risk for cognitive decline as part of the Medicare Annual Wellness Visit

Abstract

Objectives:

This study aims to provide evidence on the validity of the Patient Reported Outcome Measurement Information System (PROMIS)^® Cognitive Screener (PRO-CS) in a large sample of older adults who completed the screening as part of the Medicare Annual Wellness visit.

Methods:

We analyzed data of 19,037 patients aged 65+ who completed the cognitive screener during their Medicare Annual Wellness visit (Once=16,115; Twice= 2922) in a large healthcare system. Bivariate and multivariate ordinary least squares and logistic regression were used to examine key risk factors and their cross-sectional and predictive relationships with PRO-CS scores and cognitive decline.

Results:

We found cognitive impairment based on PRO-CS scores was associated with having a clinical diagnosis (unadjusted Odds Ratio = −3.22(0.18), p < .001). PRO-CS scores were also found to have cross-sectional relationships with a wide range of known risk factors for cognitive impairment, including age, emotional distress, social needs, frailty, and overall general health (all ps < .001). Group-level analyses indicate that PRO-CS scores were useful for detecting subjective cognitive decline occurring within one year, with declining scores predicted by indicators of frailty and general health (all ps < .001).

Conclusion:

Our findings support the validity of using the PRO-CS to detect risk for subjective cognitive decline among older adults during Medicare Annual Wellness visits. Specifically, the PRO-CS is useful in detecting potential cases of MCI and subjective cognitive decline who are at higher risk of transitioning into a confirmed diagnosis of MCI or dementia in the future.

Introduction

Cognitive impairment is often unrecognized by primary clinicians until it becomes severe, indicating a missed opportunity for initiating timely intervention to treat reversible causes, manage symptoms, and support care planning conversations [1–3]. The Medicare Annual Wellness Visit (AWV), a yearly primary care appointment focused on preventive care for Medicare enrollees in the United States, requires routine cognitive screening for early detection of cognitive impairment [2]. The cognitive assessment approach is not prescribed by the Centers for Medicare and Medicaid Services (CMS), thus there is significant variation in screening tools across health systems. However, to enable early detection, preferred AWV screening tools should demonstrate validity and utility in identifying impairment at preclinical stages of dementia.

Preclinical stages of dementia refer to intermediate stages between normal aging and dementia: mild cognitive impairment (MCI) is characterized by noticeable changes in cognition but not severe enough to interfere with everyday functioning [4], whereas subjective cognitive decline is a self-perceived decline in cognition, with typical onset at the pre-MCI stage when cognitive impairment is not yet detectable on standardized performance tests.[5,6] Approximately 36 % of individuals with subjective cognitive decline develop MCI or dementia within 4 years [5], and 30 % of individuals with MCI develop Alzheimer’s disease or related dementias within five years [7]. Among the wide variations in clinical approaches to assessing cognition during the AWV, the use of a standardized, patient-reported, and minimally burdensome cognitive screening tool is a promising approach to detect subjective decline over time and alert care providers to the earliest signs of cognitive impairment for follow-up and monitoring.

The Patient Reported Outcome Measurement Information System (PROMIS)^® Cognitive Screener, or PRO-CS, is a newly developed, patient-reported brief cognitive screener derived from the PROMIS Cognitive Function item bank [8–10]. The PROMIS Cognitive Function item bank, consisting of 63 items to assess patient-perceived cognition in everyday life, serves as an ideal source for deriving a brief cognitive screener because it is reliable in diverse populations, covering the full spectrum of functionality, and is accessible to low-literacy patients [8,9,11–14]. The development and implementation of the PRO-CS were reported in a prior publication [10]. The PRO-CS, a brief screener containing four items to assess self-perceived cognitive concerns, is optimized for electronic administration and integration in the electronic health record (EHR). So far, the PRO-CS has shown initial validity with clinical diagnoses and characteristics of cognitive impairment in a sample of AWV patients who completed the PRO-CS during the first 6 months of its implementation in a large healthcare system [10].

To expand the scope of the previous study, we aimed to provide evidence on the validity of the PRO-CS in a larger sample of older adults who completed the PRO-CS as part of the Medicare AWV in the same healthcare system, and to examine change in PRO-CS scores among patients who completed the PRO-CS in two consecutive AWVs. Specifically, in this current study, we evaluated 1) the associations of PRO-CS scores with a broad range of risk factors for cognitive impairment, including demographic and clinical characteristics, and 2) predictors of decline in PRO-CS scores over the course of two AWVs. We hypothesized the presence of significant cross-sectional and predictive correlates between PRO-CS scores and commonly known risk factors of MCI and subjective cognitive decline.

Methods

Sample

The overall sample for these analyses includes a total of 19,037 patients aged 65 or older who completed electronically administered AWV questionnaires during the primary study period (between 6/06/2023 and 1/19/2024) at one of 206 University of Pittsburgh Medical Center (UPMC) health system ambulatory care clinics. The majority of these patients (N = 16,115, Primary Sample) completed their first electronically administered PRO-CS during the primary study period (time 1) and are used for the cross-sectional analyses. The other subset of the total consists of patients (N = 2922, Change Sample) who also completed the PRO-CS as part of an AWV a year earlier during the initial rollout period (from June 2022-January 2023, Time 0), and are used in the analysis of change over time.

Measures

Cognitive function was measured with the PROMIS Cognitive Screener (PRO-CS), a 4-item scale from the PROMIS Cognitive Function item bank that assesses memory/general function, multi-tasking, working memory, and verbal fluency [15]. The PRO-CS is scored on a T-score metric (mean = 50 and SD = 10 in the U.S. general population), with higher scores representing better functioning. Two versions of the PRO-CS, Abilities and Concerns, were fielded in the initial rollout. The two versions displayed similar psychometric properties and yielded comparable scores; however, the Concerns version was preferred by patients and providers [10], thus the Abilities version was discontinued by the start of 2023. Notably, a subset of patients in the Change Sample completed the PRO-CS abilities version at time 0.

Emotional distress measures included Anxiety, Depression, and Stress. Anxiety was measured with the Generalized Anxiety Disorder – 2 (GAD-2), which queries the extent of experiencing two anxiety symptoms (Feeling nervous, anxious or on edge; Not being able to stop or control worrying) over the past two weeks on a 4-point scale from 0 =Not at all to 3 =Nearly every day, yielding a total score range from 0–6 with scores of 3 or greater indicating a positive screen. The 0/1 indicator for a positive screen of anxiety was used in analyses [16]. Depression was measured with the Patient Health Questionnaire – 2 (PHQ-2), which queries the extent of experiencing the two cardinal depression symptoms (Little interest or pleasure in doing things; Feeling down, depressed, or hopeless) over the past two weeks on a 4-point scale from 0 =Not at all to 3 =Nearly every day, yielding a total score range from 0–6 with scores of 3 or greater indicating a positive screen. The 0/1 indicator for a positive screen of depression was used in analyses [16]. Stress was measured with a single item (Are you under a lot of stress?) with three response options (Never=0, Sometimes=1, Always/Often=2). A dichotomous indicator of stress, which took the value 1 for item responses “Sometimes” or “Always/Often” and 0 for “Never”, was used in analyses.

Instrumental Activities of Daily Living (IADLs) were measured with three items. The first (Who administers your medications?) had eight response options (Self, Spouse/Partner, Adult Child(ren), Minor Child (ren), Friend, Family Member, Caregiver, Other). The derived indicator, Need help with medications, took the value 1 if any non-self option was selected and 0 if “Self” was selected. A second item (I need others to help me with) listed five IADLs as response options (Shopping, Household Chores, Doing Laundry, Managing Medications, Finances). The derived indicator, Need help with 1 or more IADLs, took the value 1 if any IADL was selected and 0 otherwise. Finally, the time 1 survey (completed by the primary sample) included the question Are you uncomfortable or nervous about driving? with four response options (Never, Sometimes, Always/Often and I don’t drive). The derived indicator, Nervous driving, took the value 1 if “Sometimes” or “Always/Often” was selected and 0 otherwise.

Frailty indicators included: 1. Fear of falling, which was derived from responses to the question Are you afraid of falling? Taking the value 0 if the patient responded “Never”, and 1 if the patient responded “Sometimes” or “Always/Often”; 2. Two or more past year falls, which took the value 1 if the patient responded “Yes” to the question Have you fallen more than once in the past year? And 0 if “No”; 3. Fall injury, which took the value 1 if the patient responded “Yes” to the question If you have fallen at all, were you injured? And 0 if “No”; 4. Slow to stand, which took the value 1 if the patient responded “Sometimes” or “Always/Often” to the question Does it take you longer that 30 seconds to get up and out of a chair and get moving? And 0 if “Never”; and 5. Balance problems, which took the value 1 if the patient responded “Sometimes” or “Always/Often” to the question Do you have problems with balance or walking? And 0 if “Never”.

Hearing was assessed with a single item Do you have to strain or struggle to hear/understand conversations? With three response options (Never, Sometimes, Always/Often). The derived indicator, Trouble hearing, took the value 1 for patients who selected “Sometimes”, or “Always/Often”, and 0 for those who responded “Never”.

Social Determinants of Health Needs Indicators included: 1. Financial strain, which was derived from responses to the question (How hard is it for you to pay for the very basics like food, housing, medical care, and heating?), taking the value 0 if the patient responded “Not hard at all”, and 1 for responses “Not very hard” “Somewhat hard” “Hard” or “Very hard”; 2. Food insecurity, derived from responses to the two questions (Within the past 12 months, you worried that your food would run out before you got money to buy more; Within the past 12 months, the food you bought just didn’t last and you didn’t have money to get more.), taking the value 1 if the patient responded “Sometimes true” or “Often true” to either statement and 0 if the response to both statements was “Never true”; and 3. Transportation barriers, based on responses to two questions (Within the past 12 months, has the lack of transportation kept you from medical appointments or from getting medication? Within the past 12 months, has the lack of transportation kept you from meetings, work or getting things needed for daily living?), taking the value 1 if the patient responded “Sometimes true” or “Often true” to either statement and 0 if the response to both statements was “Never true”.

General Health was measured with the single question from the PROMIS (Global03) [17]: In general, how would you rate your health?, which had 5 response options (Excellent=5, Very Good, Good, Fair, Poor = 1) with higher scores indicating better general health.

Demographics included Age, grouped into three categories, 65–74, 75–89, and 90 or older; Gender, categorized as Male or Female; and Race/ethnicity which was characterized as Non-Hispanic White vs. Other due to the large majority of Non-Hispanic White patients in the sample.

Diagnoses based on categorized International Classification of Diseases (ICD)−10 codes for diagnoses occurring in the 180 days prior to the date of the AWV (pre period) and on the date of or within 30 days after the AWV (post period) using the Clinical Classifications Software Refined (CCS-R) were used to calculate the Charlson Comorbidity Index and to classify patients as having no, possible, or definite Cognitive Impairment Diagnosis. For analyses, the Charlson Comorbidity Index was grouped into 5 categories (0, 1, 2, 3, 4 or more), and the Cognitive Impairment Diagnosis indicator took the value 1 for possible or definite impairment, and 0 otherwise.

Human subjects protection

All procedures were reviewed and approved by the research team’s Institutional Review Board.

Analysis plan

Analysis 1: Primary Sample (N = 16,115). Association of PRO-CS scores with demographic characteristics and items assessed in the AWV.

We first conducted bivariate ordinary least squares (OLS) and logistic regression analyses using data from the primary sample to examine the association of PRO-CS scores with demographic characteristics and questions on the AWV. The OLS regression predicted the continuous PRO-CS T-score, whereas the dependent variable for the logistic regressions took the value of 1 if PR-CS T-score < 45 or .5 SD below the population mean, to denote cognitive impairment concern and 0 otherwise. Variables with significant associations in the bivariate regressions were included in a multivariable logistic regression model to ascertain the unique contribution of each predictor. Because bivariate results were similar for OLS and logistic regression models, only a logistic model was estimated for the multivariable analysis.

Analysis 2: Change Sample (N = 2922). Predictors of declining change scores.

PRO-CS scores at each time point were transformed into 4 severity groups based on half standard deviation units: normal (PRO-CS ≥ 45), mild (PRO-CS = 40–44), moderate (PRO-CS= 35–39), and severe (PRO-CS < 35). Category shifts/changes were then computed as time0-time1 and classified into three categories: no change, improve, decline. For analytic purposes, no change and improvement categories were combined. A multivariable logistic regression model was estimated to predict cognitive decline. Predictor variables included in this model were those found to be significant in the final model from Analysis 1.

Results

Sample description

The primary sample included 16,115 patients aged 65 and older who completed an AWV during the predefined period. Nearly 40 % of this sample was aged 75 or older, 58 % were female, and the vast majority were non-Hispanic White. Thirty-five percent of the sample had at least one comorbidity, and 13 % had probable cognitive impairment based on ICD-10 diagnosis data (see Table 1). The overall mean PRO-CS score was 51.64 (SD=7.70), with 81.8 % (n = 13,189) scoring in the normal range, 13.6 % (n = 2192) scoring in the mild range, 2.6 % (n = 421) in the moderate range, and 1.9 % (n = 313) in the severe range. Supplementary Tables 1 and 2 display mean PRO-CS scores according to demographic groupings and several relevant questions included in the AWV, for the primary and change samples, respectively. The PRO-CS means differed as expected according to demographic and clinical groups in both samples. PRO-CS scores declined with age, were lower for those with anxiety or depression, tended to be lower according to frailty items such as fear of falling, balance problems, and requiring help with one or more IADLs, were lower for patients with food, housing and transportation needs, and for those with lower general health scores.

Table 1.

Sample Characteristics for primary (N = 16,115) and change (N = 2922) samples.

	Levels	Primary Sample N (%)	Change Sample N (%)
Age group	65–74	9914 (61.5)	2047 (70.1)
	75–89	5898 (36.6)	848 (29.0)
	90 +	303 (1.9)	27 (0.90)
Female	Yes	9470 (58.8)	1698 (58.1)
Non-Hispanic White	Yes	15,037(96.2)	2758 (94.4)
Charlson Comorbidity Index	0	10,333 (64.1)	1993 (68.2)
	1	2467 (15.3)	412 (14.1)
	2	1635 (10.2)	253 (8.7)
	3	774(4.8)	136 (4.7)
	4+	906 (5.6)	128 (4.4)
Cognitive Impairment diagnosis	Yes	2101 (13.1)	337 (11.5)
Anxiety positive screen	Yes	1196 (7.5)	139 (4.8)
Depression positive screen	Yes	835 (5.3)	100 (3.4)
Stress	Yes	9367 (58.1)	1829 (62.6)
Need help with medications	Yes	768 (4.8)	102 (3.5)
Need help with 1 or more IADLsa	Yes	1808(11.2)	218 (7.5)
Nervous driving	Yes	3835 (23.9)	NA
Fear of falling	Yes	7124 (44.2)	1093 (37.4)
Two or more past year falls	Yes	2144 (13.3)	256 (8.8)
Fall injury	Yes	1678(10.4)	280 (9.6)
Slow to stand	Yes	3382 (21.0)	442 (15.1)
Balance problems	Yes	6823(42.4)	1029 (35.2)
Trouble hearing	Yes	7730(52.0)	1376 (47.1)
Financial strain	Yes	3412 (21.9)	545 (18.7)
Food insecurity	Yes	608 (3.9)	58 (2.0)
Transportation barriers	Yes	246 (1.5)	24 (0.80)
General health rating	Mean (SD)	3.4 (0.82)	3.4 (0.80)

^aIADLs=Instrumental Activities of Daily Living

Analysis 1. Primary Sample Association of PRO-CS scores with demographic characteristics and items assessed in the AWV.

Results of univariate ordinary least squares (OLS) and logistic regressions showed that all variables were significantly associated with PRO-CS scores in the expected direction (Table 2). Results from the multivariable logistic regression model indicated that most independent predictors remained significant with the exception of gender, injury from a fall, and transportation barriers (Table 3).

Table 2.

Primary Sample Bivariate associations of demographic and AWV variables with PRO-CS T-Scores (OLS regression) and PRO-CS Impairment (logistic regression) (N = 16,115)^a.

	Bivariate OLS Regression (T-Score)		Bivariate Logistic Regression (Impairment vs No Impairment)
	Beta (SE)	p value	OR (95 %CI)	p value
Age group	−2.29 (0.11)	< .001	1.78 (1.65,1.90)	< .001
Female	−0.85(0.12)	< .001	1.19 (1.09,1.29)	< .001
Non-Hispanic White	1.29 (0.32)	< .001	0.65 (0.54,0.79)	< .001
Charlson Comorbidity Index	−0.77(0.05)	< .001	1.24 (1.20,1.28)	< .001
Cognitive Impairment diagnosis	−3.22 (0.18)	< .001	2.18 (1.97,2.42)	< .001
Anxiety positive screen	−6.25 (0.23)	< .001	4.33 (3.84,4.90)	< .001
Depression positive screen	−8.53 (0.27)	< .001	7.35 (6.36,8.49)	< .001
Stress	−3.84 (0.12)	< .001	2.68 (2.44,2.93)	< .001
Need help with medications	−8.18 (0.28)	< .001	6.11 (5.27,7.09)	< .001
Need help with 1 or more IADLsb	−7.85(0.18)	< .001	6.44 (5.81,7.15)	< .001
Nervous driving	−5.50 (0.14)	< .001	4.02 (3.69,4.38)	< .001
Fear of falling	−5.03(0.12)	< .001	3.62 (3.32,3.95)	< .001
Two or more past year falls	−4.70(0.17)	< .001	3.21 (2.90,3.54)	< .001
Fall injury	−3.48(0.20)	< .001	2.27 (2.03,2.53)	< .001
Slow to stand	−5.23(0.14)	< .001	3.65 (3.35,3.98)	< .001
Balance problems	−5.46 (0.11)	< .001	4.39 (4.02,4.79)	< .001
Trouble hearing	−4.61(0.12)	< .001	3.23 (2.97,3.53)	< .001
Financial strain	−3.09 (0.15)	< .001	2.13 (1.95,2.33)	< .001
Food insecurity	−4.16 (0.32)	< .001	3.28 (2.77,3.87)	< .001
Transportation barriers	−4.90 (0.49)	< .001	4.02 (3.12,5.18)	< .001
General health rating	2.92 (0.07)	< .001	0.43 (0.41,0.45)	< .001

^aSample sizes for each analysis vary slightly due to item-level missing data;

^bIADLs=Instrumental Activities of Daily Living

Table 3.

Multivariable logistic regression model results of demographic and AWV variables predicting PRO-CS impairment (primary sample, N = 16,115) and PRO-CS decline from time 0 to time 1 (Change sample, N = 2922).

	Impairment vs No Impairment (N = 14,723)b		Decline vs No Declinea (N = 2703)b
Predictorsc	OR (95 %CI)	p value	OR (95 %CI)	p value
Age group	1.20 (1.09,1.31)	< .001	1.22 (0.94,1.57)	0.13
Female	0.91(0.82,1.02)	.11	——
Non-Hispanic White	0.78 (0.62,0.99)	< .05	0.65 (0.36, 1.16)	0.14
Charlson Comorbidity Index	0.96 (0.92, 0.99)	< .05	1.01 (0.90, 1.13)	0.91
Cognitive Impairment diagnosis	1.30(1.14, 1.48)	< .001	1.06 (0.73, 1.53)	0.76
Anxiety positive screen	1.30 (1.10, 1.55)	< .01	1.21 (0.67, 2.18)	0.52
Depression positive screen	2.30 (1.89, 2.81)	< .001	0.57 (0.28, 1.18)	0.13
Stress	1.82 (1.62,2.03)	< .001	1.63 (1.21, 2.20)	< 0.002
Need help with medications	2.95 (2.41,3.61)	< .001	1.66 (0.95, 2.90)	0.28
Need help with 1 or more IADLsd	1.66 (1.44,1.92)	< .001	2.09 (1.38, 3.15)	< 0.001
Nervous drivinge	1.87 (1.68, 2.08)	< .001	NA
Fear of falling	1.26 (1.11, 1.43)	< .001	1.19 (0.87, 1.63)	0.28
Two or more past year falls	1.31 (1.15, 1.50)	< .001	1.37 (0.93, 2.01)	0.11
Fall injury	1.02 (0.88, 1.19)	.80	——
Slow to stand	1.25 (1.11, 1.41)	< .001	0.72 (0.50, 1.04)	0.08
Balance problems	1.58 (1.39, 1.79)	< .001	1.31 (0.95, 1.81)	0.10
Trouble hearing	2.47 (2.23,2.74)	< .001	1.81 (1.39, 2.36)	< 0.0001
Financial strain	1.18 (1.05,1.32)	< .01	1.02 (0.74, 1.40)	0.92
Food insecurity	1.33 (1.06, 1.67)	< .05	1.16 (0.54, 2.50)	0.70
Transportation barriers	1.13 (0.81, 1.56)	.47	——
General health rating	0.78 (0.73,0.84)	< .001	0.71 (0.59, 0.86)	< 0.001

^aSample sizes for analyses are smaller than total sample sizes due to item-level missing data.

^bDecline vs. No Decline models included a control for PRO-CS version (Abilities vs. Concerns) at time 0. That coefficient was significant at p < .01 with those completing PRO-CS Abilities being more likely to be in the decline group.

^cPredictor variables for Impairment vs. No Impairment are measured at time 1, concurrent with primary sample PRO-CS score; Predictor variables for Decline vs. No Decline are measured at time 0, concurrent with change sample first PRO-CS score, predicting decline in PRO-CS from time 0 to time 1 and exclude those found non-significant in the primary sample analyses.

^dIADLs=Instrumental Activities of Daily Living;

^eNervous Driving had 60 % missing at time 0 thus was not included in change sample analysis.

Analysis 2: Change Sample (N = 2922). Predictors of declining change scores.

Cognitive decline was observed in 11 % of patients (n = 321). Most patients did not change between severity categories (86 %) with only a small number of patients reporting improvement (3 %). Results shown in Table 3 indicated that cognitive decline from time 0 to time 1 was significantly predicted by time 0 measures of stress (p = .002), requiring IADL assistance (p < .001), trouble hearing conversations (p < .0001), and general health rating (p < .001).

Discussion

In a large cohort of older adults who completed the PRO-CS as part of the AWV from June 2023 to January 2024 in a large health system, our analyses provide evidence to support the strong correlation between PRO-CS scores and cognitive impairment diagnosis ascertained in medical records. Being identified as having possible cognitive impairment by PRO-CS was related, as expected, to common risk factors of cognitive impairment, such as age, emotional distress, social needs, frailty, and overall general health. Further, the group-level change score analyses indicate that PRO-CS scores were useful for detecting subjective cognitive decline occurring within one year, with declining scores predicted by indicators of frailty and general health. These results suggest that the PRO-CS may be a useful tool for detecting subjective cognitive decline, providing an early warning sign to providers, patients and caregivers.

It is within the expectation that PRO-CS scores were strongly associated with having an ascertained diagnosis of cognitive impairment in medical records. But even after controlling for clinical diagnoses of cognitive impairment, positive PRO-CS screening results still maintained significant correlations with a wide range of known risk factors for MCI. These risk factors included age [18], negative emotional experiences such as depression, anxiety, and stress [19,20,21], subtle deficits in Instrumental activities of daily living (IADL) [22], fall and balance problems [23,24], trouble hearing [25], self-rated general health [26], and social determinants of health needs such as financial strain and food insecurity [27]. Our study suggests that the PRO-CS is sensitive to risk factors of MCI and has potential utility in screening people with low resources and a high risk of transition to confirmed cases of MCI or dementia.

There has been increasing evidence that subjective cognitive decline assessed with two or more time points is related to biomarkers indicative of Alzheimer’s disease [28,29]. Based on PRO-CS scores, we found that 11 % of our sample experienced subjective cognitive decline over the past year, consistent with a prior study revealing that 12 % of older adults in primary care had self-perceived cognitive decline within 12 months, as defined by established criteria [30,31]. While a meta-analysis reported that 25 % of older adults had subjective cognitive decline, there is high variability across primary studies as the timeframe during which cognitive decline occurred was not specified [32]. Interestingly, our results indicate that a decline in PRO-CS scores is not associated with depression or anxiety when other known risk factors are controlled, such as stress [33], IADLs [34], trouble hearing [35], and general health [36]. This indicates that using the PRO-CS to assess cognitive decline at two-time points can, to some extent, help parse out the confounding effect of depression and anxiety when interpreting screening results, a major concern associated with using non-standardized measures to identify subjective cognitive decline cross-sectionally [31,37].

This study has limitations. First, our sample was predominantly non-Hispanic White. Given that disadvantaged populations, including those from low socioeconomic status and racial and ethnic minorities, are at higher risk of having unrecognized cognitive impairment during AWVs, future research is needed to expand the understanding of the validity and clinical utility of the PRO-CS among these populations. Second, we did not collect data on performance-based cognitive tests. Thus, although we were able to establish the potential validity of PRO-CS in relation to clinically confirmed cases of cognitive impairment, we were unable to examine the hypothetically predictive relationship between subjective decline and early signs of objective impairment in the aging process, nor were we able to establish specificity and sensitivity with thresholds of performance-based tests [38]. Future research to incorporate performance tests of cognitive function is needed to establish the concurrent and predictive validity of PRO-CS against early signs of objective cognitive impairment. Additionally, our data were collected at only two time points over one year. This design prevented us from observing trajectories of cognitive change at the individual level to identify persistent cognitive decline over multiple years, a stronger predictor for the risk of developing dementia [28]. While the timeframe for observing cognitive impairment or decline varies and depends on the underlying cause and individual progression, routine screening and longitudinal tracking are crucial for understanding individual differences in the onset, progression, and potential causes for cognitive decline observed.

Lastly, the clinical implications of our results should be interpreted with caution. As subjective cognitive symptoms are prone to confounding factors such as psychological distress, anosognosia, and the use of medications among older adults, prior literature often observed a discordance when comparing concurrent screen results of self-reported versus performance-based measures of cognitive function [38]. It is important to acknowledge that cross-sectional use of PRO-CS as a cognitive screener might lead to an underestimation (e.g., false negative due to anosognosia) or overestimation (e.g., depression, anxiety, and medications) of objective cognitive impairment in certain populations [38–40]. Thus, our results do not suggest PRO-CS can replace traditional performance-based cognitive tests during AWVs, but serve as an opportunity to standardize and incorporate a patient’s subjective perspectives and experiences as part of the evaluation of cognitive function during AWVs.

Despite these limitations, the potential utility of PRO-CS in detecting subjective cognitive decline during AWVs should not be undermined. Many confounding effects that typically interfere with the interpretation of positive screening results can be, to a great extent, parsed out by administering PRO-CS longitudinally at two or more time points. Moreover, PRO-CS can be easily adapted for use as an informant-reported measure among caregivers, which helps address anosognosia in self-reporting as a major source of bias, and improve reliability when used in combination with self-reported measures. Additionally, identifying subjective cognitive symptoms, in addition to assessing cognitive performance, can provide nuanced insights into determining appropriate interventions for a targeted subgroup. For instance, identifying individuals who screened positive in PRO-CS but with no objective cognitive impairment is important for initiating interventions to proactively monitor and improve their emotional and cognitive health, as they are at a high risk of transitioning into confirmed cases [41,42]. In consideration of the implementation challenges for standardized cognitive tests (e.g., burdensome, stressful, requiring a standard test environment, relying on secondary conditions such as physical ability to complete tasks), PRO-CS is suitable for large-scale implementation in a clinical setting with minimal burden. This serves as an advantage for using PRO-CS in settings where a performance test is too burdensome, or among those who may have difficulty participating in performance testing due to their complex medical conditions and physical limitations, for instance, those with stroke and Parkinson’s.

Conclusion

Our findings supported the validity of using the PRO-CS to detect subjective cognitive decline, which is an early indicator of objective cognitive impairment among older adults, during AWVs [5]. With strong cross-sectional and predictive associations with clinical diagnosis of cognitive impairment and a wide range of risk factors, the PRO-CS showed utility in detecting potential cases of MCI and subjective cognitive decline who are at higher risk of transitioning into a confirmed diagnosis of MCI or dementia in the future.

Impact statement.

Our results supported the potential clinical utility of the Patient Reported Outcome Measurement Information System (PROMIS)^® Cognitive Screener in detecting risk for subjective cognitive decline during the Centers for Medicare and Medicaid Services Annual Wellness visit, providing an early warning sign to providers, patients and caregivers.

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.apro.2025.100190.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available because they are protected health information from the electronic health record.