Risks and benefits of screening for dementia in primary care: the IU CHOICE trial

Abstract

Background/Objective:

The benefits and harms of screening of Alzheimer’ disease and related dementias (ADRD) are unknown. This study addressed the question of whether the benefits outweigh the harms of screening for ADRD among older adults in primary care.

Design, setting and participants:

Single-blinded, 2-arm, randomized controlled trial (October 2012-September 2016) in urban, suburban, and rural primary care settings in Indiana. 4,005 primary care patients (≥65 years) were randomized to ADRD screening (n=2,008) or control (n=1,997).

Intervention:

Patients were screened using the Memory Impairment Screen or the Mini-Cog™ and referred for a voluntary follow-up diagnostic assessment if they screened positive on either or both screening tests.

Measurements:

Primary measures were health-related quality of life (Health Utilities Index (HUI)) at 12 months, depressive symptoms (Patient Health Questionnaire-9 (PHQ-9)), and anxiety symptoms (Generalized Anxiety Disorder 7-item scale (GAD-7)) at 1 month.

Results:

The mean age was 74.2 years (standard deviation 6.9); 2,257 (66%) were female and 2,301 (67%) were white. At 12 months, we were unable to detect differences in health-related quality of life between the groups (effect Size (ES)= 0.009 [95% CI=−0.063, 0.080], p=0.81). At one month, differences in mean depressive symptoms (mean difference=−0.23 [90% CI= −0.42,−0.039]) and anxiety symptoms (mean difference=−0.087 [90% CI=−0.246, 0.072]) were within pre-specified equivalency range. Scores for depressive and anxiety symptoms were similar between the groups at all time points. No differences in healthcare utilization, advance care planning, and ADRD recognition by physicians were detected at 12 months.

Conclusion:

We were unable to detect a difference in health-related quality of life for screening for ADRD among older adults. We found no harm from screening measured by symptoms of depression or anxiety. Missing data, low rates of dementia detection, and high rate of refusal for follow-up diagnostic assessments after a positive screen may explain these findings.

Trial Registration:

ClinicalTrials.gov Identifier NCT01699503

Introduction

Primary care providers (PCPs) provide the vast majority of care to older adults in the United States. Nevertheless, at least 50% of older primary care patients living with Alzheimer’s disease and related dementias (ADRD) never receive a diagnosis.^1–3 For those patients who are diagnosed with ADRD, it often occurs 2–5 years after symptom onset.^4–7 Early detection of ADRD via screening in primary care has the potential to increase ADRD recognition⁸ and reduce individual, family, and societal burden.^{8, 9} The National Academy of Sciences, National Plan to Address Alzheimer’s Disease, and Affordable Care Act all describe earlier ADRD detection as a core aim for improving care quality for older adults.^{1, 4, 8} At the same time, patients and families perceive potential harms from screening,^{5, 10} including depression, anxiety, and stigma.^11–13 Furthermore, false positive results have the potential to lead to unnecessary treatment and high costs while false negatives can induce false reassurance.¹⁴ Given the lack of evidence concerning the benefits and harms of ADRD screening, the United States Preventive Services Task Force (USPSTF) does not recommend ADRD screening in primary care.¹⁵

Because the benefits and harms associated with screening asymptomatic, older adults for ADRD are unknown,^{16, 17} we conducted a randomized controlled trial, the Indiana University Cognitive Health Outcomes Investigation of the Comparative Effectiveness of Dementia Screening (IU CHOICE) with the following primary outcomes: health-related quality of life (HRQOL), depressive symptoms, and anxiety symptoms. We hypothesized that screened subjects will have a higher HRQOL at 12 months post-screening as compared with non-screened subjects; screened subjects will not have higher symptoms of depression or anxiety at 1-month post-screening.

Methods

Study Design

This randomized, controlled, single-blinded study included 4,005 primary care patients, ≥65 years old, who receive primary care from one of three sites in Indiana: (1) Eskenazi Health, an urban safety net health care system with 11 primary care centers; (2) Indiana University Health (IUH) with more than twenty-five urban and suburban adult primary care practices in greater Indianapolis; and (3) IUH Arnett, a regional health care center with primary care clinics in rural Indiana affiliated with Indiana University Health. Recruitment started in October 2012 and ended in September 2016. Patients were randomized equally into two arms (ADRD screening or no ADRD screening)¹⁸ with 2,008 patients randomized to receive ADRD screening. The CHOICE trial was approved by the institutional review boards of Indiana University and the local board of Indiana University Health at Arnett Hospital (IRB# 1206009010), and is registered with clinicaltrials.gov, Clinical Trials.gov Identifier NCT01699503. Informed consent was obtained at enrollment and prior to any study procedures.

Population, setting and recruitment

Patients were selected based on the following inclusion criteria: ≥65 years old, at least one office visit to their PCP within the previous year, no previous diagnosis of ADRD or cognitive impairment as determined by ICD-10 codes, or the presence of prescriptions for cholinesterase inhibitors or memantine, ability to consent to participate in the study, and ability to communicate in English. Patients were excluded if they were a permanent resident of a nursing facility, had bipolar disorder or schizophrenia as determined by the presence of related ICD-10 codes, or had a pre-existing diagnosis of ADRD or cognitive impairment.

Randomization

A computer-generated randomization scheme implemented by study statisticians was used to assign patients to the screening or no screening groups and stratified by enrollment site with a block size of 4. The allocation of patients was concealed using sequentially numbered, sealed, opaque envelopes according to the randomization sequence.

Intervention and Procedures

Patients randomized to the screening (n=2,008) underwent the Memory Impairment Screen (MIS) assessment (0–8 points), which takes approximately 4 minutes to complete and has demonstrated excellent psychometric properties in primary care and community samples.^{19, 20} A cut-off score of <5 has 86% sensitivity and 91% specificity for dementia with a positive predictive value of 72% and negative predictive value of 96% in a setting with a dementia prevalence of 15%.^{21, 22} As a result of lower than expected positive screening results in the screening arm as compared to previous studies (3% vs. 11–15%,^23–26 respectively), the protocol was amended after approval from the Data Safety and Monitoring Board, in April 2016 to add the Mini-Cog™ as part of the screening. The Mini-Cog™, which has demonstrated validity and reliability in primary care, takes approximately 3 minutes to complete. With a score range of 0–5, a cut-off score of <3 has 76% sensitivity and 73% specificity of identifying individuals with dementia.^27,28,29 A total of 205 people were screened with both MIS and Mini-Cog™.

Patients who scored negative on both the MIS or Mini-Cog™ were not formally notified after testing about their negative screen, but had been informed during the informed consent process that only patients who screened positive would receive a follow-up call after screening. Patients who scored positive on either the MIS or Mini-Cog™ were first contacted by the study coordinator, by phone, who informed them that their score on the screening test was “lower than what we would expect” and informed them that they would be receiving a call from a care coordinator to schedule a follow-up diagnostic assessment at a local collaborative memory care program, the Aging Brain Care Medical Home (ABC MedHome). Patients were then contacted by a care coordinator from the ABC MedHome who answered any questions about the screening and scheduled a visit to conduct a full assessment. Additionally, the study notified the patient’s PCP, in writing, that the patient scored positive on a brief cognitive screen, the name of the screening test, the score range for the test, and the participant’s score. Patients who screened positive, but determined to not have ADRD by the subsequent ABC MedHome assessment, were referred for an annual cognitive assessment with the local memory care practice at EH or IUH. The ABC MedHome program provides interdisciplinary care to monitor the biopsychosocial needs of patients and their informal caregivers; coach caregivers on problem-solving strategies to cope with cognitive, functional, behavioral, and psychological symptoms of ADRD; support the caregivers in navigating access to community resources and healthcare services; and work with PCPs to optimize medication management by de-prescribing of medications with adverse cognitive effects, prescribing FDA approved medications, and enhancing adherence to appropriate medications.^{30, 31}

The remaining 1,997 subjects were randomized into the no ADRD screening group, which served as a usual care control group. These patients continued to receive their usual primary care, including a referral to a local memory care practice if their PCP suspected the presence of a cognitive impairment at any time during the study. Cognitive impairment in the non-screened group was defined by having a diagnosis of mild cognitive impairment or ADRD or starting a new prescription for a cholinesterase inhibitor or memantine at any point in the 12 months post study enrollment.

Baseline and Outcome assessments

The presence of nine common geriatric chronic conditions as well as the Charlson index score³² were obtained from the Indiana Network for Patient Care, a regional health information exchange.³³ Baseline assessments for participants included the Health Utilities Index (HUI), the Patient Health Questionnaire (PHQ-9), the Generalized Anxiety Disorder Scale (GAD-7), the Medical Outcomes Study (MOS) Social Support Survey Instrument, and seven questions inquiring about the presence of an Advance Directive and Power of Attorney for health care and financial affairs.

The first primary outcome measure, HRQOL at 12 months, was assessed at baseline, 1, 6, and 12 months in all of the patients using the HUI.³⁴ As a utility-based HRQOL instrument, the HUI has been applied in patients with a wide range of medical conditions, including ADRD.^{35, 36} The individual health domain scores range from 0 (maximum impairment) to 1.00 (no impairment) and the multi-attribute (HUI index) scores, a multiplicative function of individual attribute levels, range from 0.36 to 1.00 with anchors 0 = dead and 1.00 = perfect health. The test-retest reliability exceeded the standard for adequate reliability of 0.70 in those with mild dementia (ICC = 0.75).³⁶

The second primary outcomes were symptoms of depression and anxiety at 1 month and were assessed at baseline, 1, 6 and 12 months and measured by the PHQ-9^{37, 38} and GAD-7^{39, 40} respectively. Both scales have good internal consistency and test–retest reliability as well as convergent, construct, criterion, procedural, and factorial validity for the diagnosis of major depression and general anxiety disorder.^37–40 In the event that a positive response on the PHQ-9 question about suicidal thoughts was recorded, the CHOICE study coordinator immediately notified the subject’s primary care provider.

Health care utilization, including emergency department utilization and hospital admission, was determined for all subjects throughout the 12-month study period using the Indiana Network for Patient Care.³³

Statistical analysis

Sample Size

To achieve 80% power to detect a significant effect size of 0.09 between the ADRD screening group and the no screening group at α=0.05 level (two-sided), and allowing 10% of patients with missing follow-up outcomes at 12 months, we needed to enroll at least 3,951 patients into the study. The effect size of 0.09 between the screening group and the no screening group reflects a difference of 0.40 (SD) between patients with dementia in the collaborative care program and patients who are dementia but were not screened. Also, a difference of 0.06 (SD) in the majority of patients who do not have dementia in either group assuming 85% sensitivity of screening instrument and 15% dementia prevalence in this patient population. For the second primary aims, change in symptoms of depression and anxiety, given the sample size of 4,000, we had greater than 95% power to test the equivalence levels in PHQ-9 and GAD-7 at 1 month assuming equivalence differences of 0.6 (SD = 5.1) on PHQ-9 and 0.5 (SD = 3.2) on GAD-7 based on our previous studies of primary care patients ^41–44.

Analyses

Analysis of covariance (ANCOVA) models were used to compare the primary outcomes, HUI at 12 months, PHQ-9 and GAD-7 at 1 month, between the two randomization groups adjusting for baseline measures. Mixed effect models with repeated HUI, PHQ-9, and GAD-7 measurements at 1, 6, and 12 months were also used to examine differences between the two groups with random participant effect. Baseline score, group, time, and the interaction between group and time were entered as fixed effects into the mixed models. We used linear contrasts to test for difference between the two groups at 1, 6, and 12 months. The mixed effect models used all observed outcome data at any time point, even if they had missing data at a previous time point. Rates of emergency department utilization, hospital admission and advance directives between the two randomized groups were compared using Chi-square tests. For equivalence tests of PHQ-9 and GAD-7 scores, the two one-sided test procedure was used. ⁴⁵ The two-group means were determined to be equivalent if and only if the 90% confidence intervals for group mean differences were completely contained in the pre-specified equivalence interval, i.e. (−0.6, 0.6) for PHQ-9 and (−0.5, 0.5) for GAD-7.

Sensitivity analyses were conducted comparing baseline characteristics between participants who completed a follow-up assessment and those who missed a follow-up assessment. We also used multiple imputations to examine the robustness of our results to missing data. Details on the methods of these additional analyses are described in Supplementary Text S1.

All analyses were conducted using SAS 9.4 (SAS Institute, Carey, NC).

Results

Enrollment

As shown in Figure 1, we evaluated 7,725 potential patients for eligibility; 792 patients were not eligible to participate and 2,928 refused to participate, leaving 4,005 who enrolled. Patients who enrolled in the study were more likely to be female (67% vs. 57%; p<.001), African-American (28% vs. 4%; p<.001) and younger (74.5 vs. 78.3; p<.001) than those who were eligible but refused to participate. In total, 2,008 individuals were randomized to the screening arm while the remaining 1,997 patients were in the control arm.

Figure 1.

Flow chart of patient enrollment, randomization, and follow-up in the CHOICE trial.

While preparing for final analyses, study statisticians discovered 32 outcome observations that were questionable due to date of collection occurring after a recorded date of death for study participants. This finding initiated a full review of all data and an investigation of the data source, staff collecting the data, and the data management system. Results from the investigation uncovered that all questionable data were collected by one study staff member and were unable to be verified through phone records or other data auditing systems. As a result, study data were comparatively re-examined for potential errors or inconsistencies including survey completion rates across all study staff, at each time point, and the mean variance of summary scores of the outcomes, by study staff. Both study statisticians and an independent biostatistician who was not involved in the study, reviewed the results of these analyses. After consultation with the Indiana University IRB and the study’s Data Safety and Monitoring Board, it was recommended that the study exclude all data points collected by this staff person to ensure that data of only the highest quality was used for final analyses. Thus, for all analyses presented, we have excluded all observations for 589 participants after randomization and an additional 2,233 follow-up observations, in total, across each of the time points for the remaining patients.

Baseline Characteristics

The mean (SD) age of the overall study population was 74.1 (6.9) years (range, 65–100 years); 2,256 (66%) were female and 2,301 (67%) white (Table 1). 134 (7.7%) participants in the screening arm screened positive on either the MIS or the Mini-Cog™. Among those who screened positive 88 (66%) refused a follow-up diagnostic assessment at the ABC MedHome.

Table 1.

Demographic and clinical characteristics of the study participants at baseline^*

Characteristic	Overall (n=3416)	ADRD screen (n=1723)	No ADRD screen (n =1693)
Age in years, mean±SD	74.1±6.9	74.2±7.0	74.1±6.8
Female sex, no (%)	2256 (66)	1167 (68)	1089 (64)
Race, no (%)
Black	1056 (31)	526 (31)	530 (31)
White	2301 (67)	1164 (68)	1137 (67)
Other	55 (2)	31 (2)	24 (1)
Hispanic ethnicity, no (%)	33 (1)	14 (1)	19 (1)
Co-morbidity status, no (%)
Cancer	771 (23)	379 (22)	392 (23)
Chronic Lung Disease/Emphysema/COPD	1353 (40)	698 (41)	655 (39)
Congestive Heart Failure	612 (18)	301 (18)	311 (18)
Peripheral Vascular Disease	509 (15)	253 (15)	257 (15)
Coronary Artery Disease	1152 (34)	567 (33)	585 (35)
Diabetes	1416 (42)	731 (42)	685 (41)
Hypertension	2811 (82)	1428 (83)	1383 (82)
Stroke	783 (23)	390 (23)	393 (23)
Charlson co-morbidity score, mean±SD	2.8±2.9	2.7±2.8	2.8±3.0
Education level, no (%)
Less than high school	720 (21)	363 (21)	357 (21)
High school	1127 (33)	571 (33)	556 (33)
Some college or college degree	1547 (46)	778 (45)	769 (46)
Self-reported social supporta, mean±SD	20.9 ±4.6	21.0 ±4.5	20.8 ±4.6
Study Site, no (%)
Urban, safety net health system	1434 (42)	725 (42)	709 (42)
Rural health system	1277 (37)	653 (38)	624 (37)
Suburban and urban academic health	705 (21)	345 (20)	360 (21)

Abbreviations: SD=standard deviation

^*Data are presented as mean (percentage) unless otherwise indicated.

^aMeasured with the MOS-5 Social Support Survey Instrument

Primary Outcomes

At baseline, the mean HUI scores for the screening and no screening groups were similar (0.67 and 0.67, respectively; p=0.47). The mean HUI score at 12 months was 0.68 in both the screening group and no screening group ((Effect Size (ES)=0.009 [95% CI= −0.063, 0.080], p=0.81); Table 2). Comparisons of HUI scores at 1 and 6 months were unable to detect differences between the study arms (Table 2).

Table 2.

Health-related quality of life, depressive symptoms, and anxiety at all study time points, by group

	ADRD screen (n=1723)		No ADRD screen (n=1693)		Predicted Differences ANCOVA*		Predicted Difference Mixed Model*
	N	Mean Score (95%CI)	N	Mean Score (95%CI)	Mean Difference (95%CI)	P-value	Mean Difference (95%CI)	P-value
Health-related quality of lifea
Baseline	1662	0.67 (0.65, 0.68)	1633	0.67 (0.66, 0.69)
1 month	974	0.71 (0.69, 0.72)	1000	0.69 (0.68, 0.71)			0.014 (−0.004, 0.031)	0.13
6 months	885	0.69 (0.67, 0.71)	875	0.70 (0.68, 0.72)			−0.003 (−0.022, 0.015)	0.73
12 months¥	993	0.68 (0.66, 0.69)	976	0.68 (0.66, 0.70)	0.002 (−0.017, 0.021)	0.81	0.003 (−0.014, 0.021)	0.71
Depressive symptomsb
Baseline	1698	3.23 (3.04, 3.43)	1665	3.06 (2.88, 3.24)
1 month¥	996	2.33 (2.13, 2.53)	1022	2.46 (2.26, 2.66)	−0.230 (−0.421, −0.039)		−0.233 (−0.441, −0.026)
6 months	903	2.50 (2.27, 2.72)	893	2.45 (2.23, 2.67)			−0.043 (−0.260, 0.174)
12 months	1006	2.74 (2.51, 2.97)	992	2.82 (2.59, 3.05)			−0.184 (−0.393, 0.025)
Anxiety symptomsc
Baseline	1698	1.79 (1.63, 1.95)	1664	1.77 (1.62, 1.92)
1 month¥	997	1.16 (1.00, 1.32)	1022	1.24 (1.09, 1.40)	−0.087 (−0.246, 0.072)		−0.084 (−0.249, 0.080)
6 months	903	1.24 (1.07, 1.41)	893	1.13 (0.98, 1.29)			0.155 (−0.018, 0.328)
12 months	1005	1.31 (1.14, 1.48)	992	1.45 (1.27, 1.63)			−0.164 (−0.329, 0.002)

Abbreviations: ADRD= Alzheimer’s disease and related dementias; CI=Confidence Interval

^*Adjusted for baseline score

^¥Primary outcomes

^aMeasured with the Health Utilities Index (HUI);

^bMeasured with the Patient Health Questionaire-9 (PHQ-9);

^cMeasured with the Generalized Anxiety Disorder Scale-7 (GAD-7).

Baseline PHQ-9 and GAD-7 scores were also similar between the screening and no screening groups (3.23 vs. 3.06; p=0.19) and (1.79 vs. 1.77; p=0.86), respectively. The mean PHQ-9 scores at 1 month between the groups were within our pre-specified equivalence interval (mean difference: −0.23 [90% CI=−0.42, −0.04]). Mean GAD-7 scores at 1 month were also within equivalence interval (mean difference= −0.087 [90% CI= −0.246, 0.072]; Table 2). Furthermore, difference in mean PHQ-9 and GAD-7 scores between the groups were also within equivalence intervals at 6 and 12 months (Table 2 and Figure 2).

Figure 2.

Mean scores at baseline and at 1, 6, and 12 months’ follow-up. Top: HUI health-related quality of life scores. Middle: PHQ-9 depressive symptoms score. Bottom: GAD-7 anxiety symptoms score.

Secondary outcomes- health care utilization

We analyzed the number of patients with emergency department visits and hospital admissions at 12 months. The number of patients with at least one emergency department visit or one hospitalization were similar between the screening and no screening groups (emergency department visits, 30% vs. 30%, respectively; p=0.71 and hospitalizations, 20% vs. 20%, respectively; p=0.93). Furthermore, no differences in advance care planning at 12 months were observed between the groups (results not shown).

We also analyzed the impact of screening on recognition of new ADRD cases. There was no statistical difference in new cases of ADRD, as determined by an incident diagnosis of mild cognitive impairment or ADRD or starting a new prescription for a cholinesterase inhibitor or memantine within 12 months from the screening event (3.4% no screening vs. 3.1% screened; p=0.70). (results not shown).

Sensitivity Analyses

We compared baseline characteristics of participants who completed the 1 and 12-month follow-up assessments to those who did not complete the 1 and 12-month assessments. Importantly, the comparison for 1-month completion status revealed that within both groups, those missing 1-month outcomes were younger, more likely to be African American, had less education, and had higher depressive symptoms at baseline than those who completed the 1-month assessment. Despite of these differences, the missing data patterns were mostly similar between the two groups (Supplementary Tables S1 and S2).

We also conducted a variety of multiple imputation analyses for the primary outcomes to assess the robustness of our results to missing data. Detailed description of these analyses are presented in the Supplementary Text S1 and the results of all sensitivity analyses are presented in Supplementary Table S3. Multiple imputation results for HUI at 12 months and PHQ-9, GAD-7 at 1 month were similar to main results from observed data presented in Table 2.

Discussion

In response to the lack of evidence for ADRD screening by the USPSTF,¹⁵ this is the first randomized controlled trial to evaluate the benefits and harms of population screening for ADRD among asymptomatic patients attending primary care. Our trial demonstrated no harm from screening, as measured by depressive and anxiety symptoms. We were unable to demonstrate benefits from screening, as measured by and health-related quality of life. Screening did not influence health care utilization as measured by emergency department visits and hospitalizations, advance care planning, or new ADRD recognition by physicians (primary care or specialists).

The finding of statistical equivalence of screening on patients’ symptoms of depression and anxiety is important given that previous studies measuring the publics’ perceived attitude of ADRD screening showed that patients were concerned that screening would make them feel depressed or anxious.¹¹ In that prior study, ADRD caregivers were more likely to rate both the benefits and suffering from ADRD identification as higher than non-caregivers;¹¹ therefore, it is possible that caregivers’ quality of life may be uniquely impacted by screening. In recognition of the need to support family caregivers, we are currently undertaking the Caregiver Outcomes of Alzheimer’s Disease Screening (COADS) trial (clinicaltrials.gov NCT03300180) to examine the impact of ADRD screening on improving outcomes for family members of older adults.

Although our study was the first clinical trial investigating the benefits and harms of ADRD screening, we recently published the results of an observational study of a national sample of Medicare patients demonstrating no benefit of the Annual Wellness visit on ADRD care.⁴⁶ As well as results of a clinical trial showing no health outcome benefit of sharing the results of neuropsychological testing with primary care clinicians,⁴⁷ which is consistent with the findings of the present study.

The screening described in the present study may also be useful once disease-modifying ADRD treatment does become available. However, screening must be coupled with follow-up diagnostic assessment and care, if cognitive impairment is diagnosed. Coordinated care delivered through dementia collaborative care programs has been shown to reduce behavioral and psychological symptoms in patients living with ADRD and the depressive symptoms of their informal caregivers.^30,31,48 In the present study, screening did not contribute to health care utilization as measured by emergency department visits and hospital admissions. However, patients who screened positive who received collaborative care had significantly decreased hospital admissions as compared with those not screened but with evidence of cognitive impairment (data not shown). This is consistent with our previous studies showing that that dementia collaborative care reduced healthcare utilization, resulting in annual cost savings ranging from $908 to $2,856 per patient.^49,50

In the present study, 7.7% of the patients in the screening arm screened positive. In a multivariate logistic regression model, age, sex, education level, and history of chronic obstructive pulmonary disease were significantly (p=0.02) associated with screening results.⁵¹ Although the positive screening rate is lower than previous studies describing a range up to 15% and may be a result of the screening instruments used in this study^23–27 more recent epidemiological studies have reported a reduction in ADRD incidence.^{33, 52–55} Specifically, Langa et al.³³ reported that the prevalence of ADRD decreased to 8.8% in 2012 from 11.6% in 2000. Furthermore, the declining ADRD incidence was especially observed in younger age groups,⁵⁵ which may be attributed to increased education and lifetime cognitive stimulation,³³ reduced vascular risk factors,⁵⁵ or less tobacco use.⁵³

The present study has some limitations. First, we anticipated a 10% missing rate for the 12-month follow-up assessment. We experienced a 42% missing rate at 12 months for the health-related quality of life outcome, hence, our study had lower power than anticipated due to the data that we excluded for quality and the less the follow-up data at each time point. However, mean differences between groups for health-related quality of life was very small, even in sensitivity analyses adjusting for missing data, thus it is unlikely that more complete follow-up data would lead to a different conclusion.

Second, approximately 38% of the potential participants assessed for eligibility refused to participate, which may have biased the sample. Of those included in the study, participants were more likely to be female, African-American, and younger. Moreover, 66% of the participants who screened positive refused diagnostic assessment and subsequent follow-up care which may have impacted our results. The high refusal rate is similar to previous ADRD screening studies.⁵⁶ In our previous study, 67% of patients who screened positive refused a diagnostic assessment. We found that patients who lived alone and had a perceived stigma of ADRD were more likely to refuse diagnostic assessment after screening positive²⁴ and that barriers to screening, in general, included perceived emotional suffering and fear of losing of indepdence.^11,12

A third limitation is that this study did not measure other, long-term implications of screening such as misdiagnosis and the opportunity costs for the participating health systems for increased resources used to provide diagnostic follow-up care for false positive screening tests. Furthermore, the design of the present study and its limitations and challenges demonstrate just how challenging it would be to get people through the entire screening to treatment cascade –perhaps even if effective drugs become available in the future.

Conclusions

Among a sample of older adults who were screened for ADRD in primary care, we were unable to detect an effect on measures of health- related quality of life and found screening does not lead to harm as measured by symptoms of depression and anxiety. Studies assessing the effects of ADRD screening on family member outcomes are ongoing⁵⁷ and will complement this study in fully evaluating the potential harms and benefits of ADRD screening in primary care.

Supplementary Material

Supp info

Supplementary Text S1: Description of additional sensitivity analyses are described to examine the impact of missing follow-up assessment data on the primary outcomes.

Supplementary Table S1: Comparisons of participants’ baseline characteristics for those who completed a 12-month assessment and those who did not complete a 12-month assessment, by randomized group

Supplementary Table S2: Comparisons of study participants’ baseline characteristics for those who completed 1-month assessment and those who did not complete 1-month assessment within and between groups.

Supplementary Table S3: Comparison of primary and secondary outcomes using multiple imputation techniques for missing data.