Care sequences leading to the diagnosis of Alzheimer's disease and related dementias: An analysis of electronic health records

Abstract

BACKGROUND

We examined the sequences of clinical care leading to diagnoses of Alzheimer's disease and related dementias (ADRD) using electronic health records from a large academic medical center.

METHODS

We included patients aged 65+ with their first ADRD diagnoses from January 1, 2014 to December 31, 2019. Using state sequence analysis, care sequences were defined by the ordering of healthcare utilizations occurred in the 2 years before ADRD diagnosis.

RESULTS

Of 3621 patients (median age 80), nearly half followed a care sequence of having one primary care visit close to their ADRD diagnosis. Additional care sequences included periodic (n = 322, 8.9%) and multiple (n = 416, 11.5%) outpatient visits to primary care and having one (n = 395, 10.9%), multiple (n = 469, 13.0%), or highly frequent (n = 357, 10.7%) outpatient visits to other specialties. Patients’ sociodemographic traits contributed to the variability in care sequences.

CONCLUSIONS

Several distinct patterns of care leading to ADRD diagnoses were identified. Integrated care models are needed to promote early identification of ADRD.

Highlights

Dementia patients followed distinct care pathways prior to their dementia diagnoses. Key sociodemographic traits contributed to the variation in the sequences of care. Racial differences in the sequencing of care were also found, but only in women.

BACKGROUND

Approximately one in 10 older Americans suffer from Alzheimer's disease and related dementias (ADRD). The prevalence of ADRD is even higher among older adults from racial and ethnic minority groups and those from disadvantaged socioeconomic backgrounds. The early diagnosis of ADRD has substantial medical, social, and financial benefits – such as access to new and evolving therapies to better support the management of symptoms through non-pharmacological measures, referrals to programs for family caregivers, and reduced costs of medical and long-term care. However, achieving early diagnosis of ADRD has been challenging. Despite having a higher prevalence of the disease, racial and ethnic minority groups and older adults with lower socioeconomic status often have delays in their ADRD diagnoses. These older adults may be less likely to have adequate knowledge about dementia and more likely to explain memory loss as part of the normal aging process, which in turn may lead to delayed care-seeking behaviors. Another factor potentially contributing to a delayed diagnosis may be that older adults from racial/ethnic minority backgrounds have less access to comprehensive services for a dementia diagnosis, such as neuroimaging or other tests. Therefore, identifying the care that patients received before their diagnosis is an essential first step in improving the early diagnosis of ADRD.

The life course perspective provides a valuable framework for understanding the care that patients received prior to their ADRD diagnoses. Drawing from this perspective, the care leading to an ADRD diagnosis (ie, pathway) can be conceptualized by several key components in life course research, including timing (eg, diagnosis date or patient's age at diagnosis), duration (eg, time interval between an initial assessment and diagnosis by a specialist), and sequencing (eg, primary care → specialist → diagnosis). From a sequencing standpoint, the guideline-recommended optimal pathway has been described in qualitative studies as a care process that begins with the recognition of symptoms, followed by a primary care visit, a subsequent referral for specialty care, and then a clinical diagnosis from a specialist. However, some qualitative studies have also suggested that older adults, especially those from racial/ethnic minority groups, may follow a crisis-event pathway that characterizes an ADRD diagnosis occurring at the time of a medical crisis or hospitalization for some other illness.

RESEARCH IN CONTEXT

• Systematic review: The authors reviewed the literature using PubMed. Prior studies that examined the sequencing of care leading to an ADRD diagnosis was largely limited to descriptive qualitative studies. We found no previous studies quantifying the care sequences of clinical care leading to ADRD diagnoses.

• Interpretation: To our knowledge, this is the first study applying a state sequence analysis to identify various care sequences leading to the diagnoses of ADRD. Using the electronic health records from a large academic medical center, we show that there are considerable variabilities in the care sequences that patients follow prior to their index ADRD diagnoses. Patients’ age, sex, race, and rurality partially contributed to the variability in the care sequences.

• Future directions: Integrated care models, educational resources, and other tools are needed particularly in primary care and non-ADRD specialties to promote early and equitable identification of dementia.

To date, most research examining the sequencing of care leading to an ADRD diagnosis has been largely limited to descriptive qualitative studies, as described earlier. A few recent studies have used sequence analyses to quantify patients’ utilization of care over time, but with a focus on conditions other than ADRD such as chronic obstructive pulmonary disease (COPD), multiple sclerosis, and heart failure. To our knowledge, no quantitative investigations have been conducted to characterize the sequences of care leading to the diagnosis of ADRD in older adults. In this study, we used electronic health records (EHRs) from a large academic medical center and its affiliated clinics to identify the major care sequences in the prior 2 years leading to the diagnosis of ADRD. We also assessed how key sociodemographic characteristics, at both the patient level (eg, sex, race) and context level (eg, rurality, neighborhood disadvantage), were associated with the different sequences of care.

METHODS

Participants

This is an observational study that used patients’ EHR data from the Duke University Health System (DUHS). The EHR data were extracted using Duke Enterprise Data Unified Content Explorer (DEDUCE), a data-extraction system that provides access to clinical data stored in an organizational data warehouse. Eligible patients were included if they (1) had a diagnosis of ADRD (ie, index diagnosis of ADRD) based on the International Classification of Diseases, Ninth Revision (ICD-9) or ICD-10 code for a final/primary discharge diagnosis between January 1, 2014 and December 31, 2019; (2) were aged 65 or older; and (3) self-identified as non-Hispanic Black or non-Hispanic White. To ensure that we captured all healthcare utilization that patients had leading up to their ADRD diagnosis, we limited our patient population to those who had had at least one clinical encounter in the DUHS system 2 years prior to their ADRD diagnoses, had a primary care provider from Duke documented in the EHR, and had their primary residence in North Carolina. Specifically, as in prior research, we excluded patients with an out-of-state address or with missing or incomplete address information. Preliminary analyses indicated that patients without a complete residential address in North Carolina had significantly fewer clinical encounters than those included in the study, suggesting that healthcare utilization may not be fully captured in patients who lack complete address information. Our final analytical file included 3621 older adults with ADRD covering 47,174 clinical encounters during our observational period (Figure ).

[IMAGE OMITTED. SEE PDF]

Measures

Care sequences

Drawing from a life course perspective, we operationalized the sequences of care as the series of healthcare utilizations that occurred over the 2-year period prior to the patients’ index diagnosis of ADRD. Using an approach similar to that of prior research, we categorized healthcare utilizations as follows: (1) ADRD specialty visits (ie, outpatient clinic visits to an ADRD specialist such as a neurologist, neuropsychiatrist, or geriatrician), (2) primary care visits (ie, outpatient clinic visits to primary care providers such as family medicine or general internal medicine), (3) in-patient hospitalizations, (4) visits to the emergency department (ED), and (5) other specialty visits (ie, outpatient clinic visits to specialists other than ADRD, such as cardiology or ophthalmology visits).

Covariates

We extracted patients’ sociodemographic characteristics, and comorbidities from the EHR, and linked Census data on neighborhood characteristics. Sociodemographic characteristics included age at index ADRD diagnosis, sex, race, and marital status. Comorbidities at the index diagnosis encounter were identified using the Elixhauser Comorbidity Index based on ICD 9/10 codes. Neighborhood characteristics were measured based on rurality and Area Deprivation Index (ADI), a composite measure that was constructed based on the American Community Survey 5-year estimates. In this study, we dichotomized areas into disadvantaged neighborhoods (yes/no) if the ADI values ranked in the bottom 15th percentile of the national level. For rurality, we used the federal rural–urban commuting area (RUCA) codes to categorize the rural–urban status of a neighborhood.

This study was approved by the DUHS Institutional Review Board. Due to the sensitive nature of the data, qualified researchers with appropriate human-subject training may send requests to the corresponding author to access the data used in this study.

Statistical analysis

We first examined the distributions of patient demographic and clinical characteristics at the index ADRD diagnoses overall and by sex using chi-squared or Mann-Whitney U tests. Next, we used state sequence analyses to holistically identify the typologies of care sequences that occurred in the 2-year period leading up to the diagnosis of ADRD. We first defined patients’ states – the type of healthcare utilization that patients had – in a given week throughout the period prior to the index ADRD diagnosis. Initially, we defined patients’ states on a given day and conducted state sequence analysis based on daily data. However, due to the complexity of the data (6^{^730} possible sequences for each patient) and the relatively large number of patients, no clear and meaningful typologies of care processes emerged. Therefore, as in prior research, we chose weeks as the time unit and assigned patients to one of the following states based on their healthcare utilization: (1) AD specialty (eg, visits that may be directly related to ADRD diagnoses/care), (2) hospitalizations (eg, inpatient stays for critical conditions that require comprehensive exams and may include cognitive assessments), (3) primary care visits (eg, first-line assessment for ADRD), (4) ED (eg, visits for urgent conditions), (5) other specialty visits (eg, visits for non-urgent conditions other than ADRD), and (6) no healthcare utilization. The ordering of the six states above reflects the priority of each state and was designed/refined based on utilization frequency and clinical relevance to ADRD diagnoses. Therefore, for patients who had multiple types of healthcare utilizations in a given week (< 5% of encounters), we assigned patients to the state with a higher priority. For example, if a patient had an ED visit and was subsequently hospitalized for the rest of the week, the patient state for that week would be hospitalization. This approach has been used by others, and our preliminary analyses assessed alternative prioritizations of states (eg, changing the ordering of primary care and ED visits), and the results were largely consistent.

Next, we used the optimal matching distance algorithm implemented in the R package TraMineR to compute the similarities between each patient's sequence of healthcare utilizations (ie, care sequences). The optimal matching algorithm is the most widely used approach to compare (dis)similarities. In brief, optimal matching measures the costs of changing from one sequence of states (ie, care sequences) to another. Only three types of changes are allowed in optional matching: adding a state, deleting a state, or substituting one state with another. We assigned the cost of adding/deleting a state to 1 and calculated the cost of substitution based on the transition probabilities from one state to another observed in our data.

We then used agglomerative hierarchical clustering analysis with Ward's criterion on the resulting dissimilarity matrix to cluster care processes in the period leading to an ADRD diagnosis. The optimal number of clusters (ie, typologies of care sequences) was chosen based on theoretical considerations and the clinical meaning of each care sequence. Although there is no established method to estimate the sample size needed for state sequence analysis, prior research using this analysis method included sample sizes ranging from 500 to 2600. Therefore, the number of patients we included in our analysis should be sufficient for state sequence analysis. Leveraging the data visualization options offered in TraMineR, we plotted the representative sequences to help interpret each identified typology of care sequences.

Finally, we compared patients’ sociodemographic and clinical characteristics across the identified typologies of care sequences using chi-squared and Kruskal-Wallis tests. We then conducted multivariate analyses using multinomial logistic regression models to assess the factors associated with the care sequences leading to an ADRD diagnosis. These statistical analyses were conducted in Stata 17 SE, and p < .05 was considered to be statistically significant.

RESULTS

Table  presents characteristics of the overall study participants and by sex. Overall, the median age at diagnosis was 80 years, and more than half the patients were female (61.8%). On average, patients utilized healthcare services 10 times (interquartile range [IQR] = 4 to 18) in the 2 years prior to receiving their diagnoses of ADRD. The diagnosis of ADRD occurred most often during encounters classified as AD specialty (48.4%), primary care (28.1%), and during a hospitalization (14.5%).

TABLE 1 Characteristics of study patients by sex (N = 3621).

	Overall (N = 3621)	Female (N = 2238)	Male (N = 1383)	p value
Age at diagnosis	80 (74–86)	81 (75–87)	79 (73–84)	<0.001
Non-Hispanic Black	927 (25.60)	622 (27.79)	305 (22.05)	<0.001
Marital status				<0.001
Currently married	1731 (47.80)	768 (34.32)	963 (69.63)
Widowed	1288 (35.57)	1053 (47.05)	235 (16.99)
Divorced	284 (7.84)	206 (9.20)	78 (5.64)
Others	318 (8.78)	211 (9.43)	107 (7.74)
Rurality	515 (14.22)	307 (13.72)	208 (15.04)	0.048
Disadvantaged neighborhood	353 (9.75)	220 (9.83)	133 (9.62)	0.860
Diagnosis encounter type				<0.001
AD specialty	1751 (48.36)	1049 (46.87)	702 (50.76)
ED	99 (2.73)	64 (2.86)	35 (2.53)
Hospitalization	524 (14.47)	293 (13.09)	231 (16.70)
Other specialty	230 (6.35)	141 (6.30)	89 (6.44)
Primary care	1017 (28.09)	691 (30.88)	326 (23.57)
No. prior encounters	10 (4–18)	9 (4–18)	10 (4–19)	0.580
Elixhauser Comorbidity Index	1 (0–2)	1 (0–2)	0 (0–2)	0.996

Seven types of care sequences were identified by the state sequence analysis. This number of clusters (ie, types of care sequences) shared the maximum within-cluster homogeneity and maximum between-cluster heterogeneity. Figure  presents the typical care sequencing in each type, and Figure presents all of the care sequences among study patients. Preliminary analyses showed very similar characteristics for patients with frequent use of other specialty services prior to their ADRD diagnoses, regardless of the intensity of the utilization (ie, Types 4 and 7, Table ). Therefore, as shown in prior research, we combined these two types in our final analyses. This resulted in six distinct and clinically meaningful care sequences in the 2-year period leading to a diagnosis of ADRD: (1) one primary care visit about 1 month prior to ADRD diagnosis (45.9%, n = 1662); (2) one other specialty visit approximately 3 months prior to ADRD diagnosis (10.9%, n = 395); (3) multiple visits to primary care within a half-year period prior to ADRD diagnosis (11.5%, n = 416); (4) highly frequent visits to other specialties throughout the 2-year period (10.7%, n = 357); (5) periodic monthly visits to primary care (8.9%, n = 322); and (6) multiple visits to other specialties within 1 month prior to ADRD diagnosis (13.0%, n = 469).

[IMAGE OMITTED. SEE PDF]

Table  presents the characteristics of study patients categorized by the major types of care sequences. The care sequence that included a single primary care visit (Type 1) had the largest number of patients, nearly half of all patients. Patients with periodic primary care visits (Type 5) were among the smallest group of patients and were among the oldest age on average. Older adults with multiple or highly frequent visits to other specialties before their ADRD diagnoses were slightly younger. Patients who were female and living in rural areas were more likely to have more frequent primary care visits (Types 1, 3, and 5). Conversely, those who were currently married were more likely to utilize other specialties (Types 2, 4, 6). No significant differences were found across typologies with regard to racial composition, disadvantaged neighborhoods, and/or comorbidities.

TABLE 2 Characteristics of study participants by type of care sequence (N = 3621).

	Type 1 (N = 1662)	Type 2 (N = 395)	Type 3 (N = 416)	Type 4 (N = 357)	Type 5 (N = 322)	Type 6 (N = 469)	p value
Age at diagnosis	80 (12)	80 (12)	82 (12)	78 (10)	82 (10)	78 (13)	<.001
Female	1048 (63.1)	224 (56.7)	251 (60.3)	200 (56.0)	234 (72.7)	281 (59.9)	<.001
Non-Hispanic Black	413 (24.9)	97 (24.6)	122 (29.3)	94 (26.3)	88 (27.3)	113 (24.1)	.426
Marital status							<.001
Currently married	767 (46.2)	207 (52.4)	178 (42.8)	202 (56.6)	134 (41.6)	243 (51.8)
Widowed	603 (36.3)	124 (31.4)	172 (41.4)	101 (28.3)	133 (41.3)	155 (33.1)
Divorced	120 (7.2)	28 (7.1)	31 (7.5)	37 (10.4)	28 (8.7)	40 (8.5)
Others	172 (10.4)	36 (9.1)	35 (8.4)	17 (4.8)	27 (8.4)	31 (6.6)
Rurality	262 (15.8)	39 (9.9)	79 (19.0)	28 (7.8)	47 (14.6)	60 (12.8)	<.001
Disadvantaged neighborhood	170 (10.2)	35 (8.9)	46 (11.1)	26 (7.3)	23 (7.1)	53 (11.3)	.166
No. prior encounters	5 (7)	12 (9)	12 (8)	35 (17)	22 (12)	11 (10)	<.001
Elixhauser Comorbidity Index	1 (2)	0 (1)	0 (1)	1 (2)	1 (2)	1 (2)	.097

Table  shows the results from the multivariate analyses. Although the proportions of non-Hispanic Black patients were similar across all care sequences, we found a significant sex–race interaction. Specifically, non-Hispanic Black women were more likely to have multiple primary care visits than a single primary care visit prior to their diagnosis (relative risk ratio [RRR] = 1.72, 95% confidence interval [CI]: 1.03 to 2.86); whereas non-Hispanic White women were less likely to have multiple primary care visits (RRR = 0.69, 95% CI: 0.53 to 0.91). Non-Hispanic Black women were also more likely to have multiple other specialty visits in the period right before an ADRD diagnosis (RRR = 1.95, 95% CI: 1.16 to 3.29).

TABLE 3 Multivariate analysis of factors associated with type of care sequences (N = 3621).

	Type 2	Type 3	Type 4	Type 5	Type 6
	RRR (95% CI)	p value	RRR (95% CI)	p value	RRR (95% CI)	p value	RRR (95% CI)	p value	RRR (95% CI)	p value
Age at diagnosis	0.99 (0.98 to 1.01)	.241	1.01 (1.00 to 1.03)	.072	0.97 (0.95 to 0.99)	<.001	1.02 (1.01 to 1.04)	.005	0.97 (0.96 to 0.99)	<.001
Female	0.77 (0.59 to 1.01)	.062	0.69 (0.53 to 0.91)	.008	0.80 (0.60 to 1.06)	.120	1.43 (1.04 to 1.97)	.028	0.80 (0.62 to 1.03)	.088
Non-Hispanic Black	0.92 (0.60 to 1.39)	.684	0.90 (0.59 to 1.37)	.617	1.03 (0.68 to 1.58)	.876	0.98 (0.57 to 1.68)	.952	0.66 (0.42 to 1.01)	.056
Female × Non-Hispanic Black	1.26 (0.74 to 2.13)	.398	1.72 (1.03 to 2.86)	.039	1.29 (0.75 to 2.22)	.355	1.25 (0.67 to 2.33)	.473	1.95 (1.16 to 3.29)	.012
Marital status (Ref: currently married)
Widowed	0.84 (0.64 to 1.12)	.237	1.22 (0.93 to 1.61)	.151	0.76 (0.57 to 1.03)	.078	0.92 (0.68 to 1.25)	.604	0.97 (0.75 to 1.26)	.831
Divorced	0.87 (0.56 to 1.36)	.547	1.19 (0.77 to 1.85)	.433	1.08 (0.71 to 1.63)	.72	1.17 (0.73 to 1.85)	.517	1.03 (0.69 to 1.53)	.888
Others	0.78 (0.52 to 1.17)	.226	0.87 (0.58 to 1.32)	.526	0.35 (0.21 to 0.60)	<.001	0.76 (0.48 to 1.20)	.241	0.57 (0.37 to 0.86)	.008
Rurality	0.57 (0.40 to 0.82)	.002	1.29 (0.97 to 1.71)	.077	0.43 (0.28 to 0.65)	<.001	0.95 (0.68 to 1.34)	.784	0.76 (0.56 to 1.03)	.076
Disadvantaged neighborhood	0.87 (0.59 to 1.27)	.462	1.10 (0.77 to 1.55)	.606	0.71 (0.46 to 1.09)	.114	0.68 (0.43 to 1.07)	.094	1.11 (0.80 to 1.55)	.521
Elixhauser Comorbidity Index	0.99 (0.94 to 1.05)	.778	0.89 (0.84 to 0.96)	.001	1.05 (1.00 to 1.11)	.065	1.00 (0.93 to 1.06)	.912	1.02 (0.97 to 1.07)	.540

DISCUSSION

To the best of our knowledge, this is the first study to quantify the care sequences leading up to ADRD diagnoses using data from the EHR. Results from state sequence analysis suggested six distinct care sequences in over 3500 older adults with ADRD. Age, rurality, and marital status contributed to the variation in the sequence(s) of care prior to an ADRD diagnosis. Racial differences in the sequencing of care were also found, but only in women. In addition, no significant differences were found with regard to patients’ neighborhood socioeconomic status across the sequences of care.

In this study, most patients increased their healthcare utilization within the 6-month period before the index diagnosis of ADRD. We found that four of the six types of care sequences showed an increase in either primary care or other specialty visits prior to ADRD diagnoses, representing over 80% of the care patterns that patients followed. These findings reflect the optimal care pathways suggested in prior qualitative studies and are similar to previous research showing that older adults had increased use of healthcare services before their ADRD was diagnosed. The increased use of primary care services before ADRD diagnoses may be due to the critical role that primary care providers play in caring for persons who develop dementia. Specifically, prior research found that most older adults chose their primary care providers when first discussing their concerns about cognitive function. Insurance policies could require that older adults obtain a referral from their primary care providers to be able to seek specialty care. A national survey also has shown a strong belief among primary care providers that they are on the front lines for providing care for older adults with memory concerns and/or ADRD. However, prior research has also suggested that not all primary care providers were prepared and/or well-equipped with the tools and referral systems to diagnose ADRD. Therefore, additional training opportunities on dementia care and integrated clinical care models are needed to promote high-quality dementia care in primary care settings. For example, a recent study among primary care providers found a strong need for additional training in the most cutting-edge evidence in ADRD and the social/cultural aspects of dementia care. In addition, although federal agencies and professional societies have developed several sources, such as the KAER Toolkit, to support early detection of ADRD in primary care settings, implementing these models will require additional training in building the clinical workflow and streamlining referrals for social and support services.

Our study also showed that about one in three older adults followed care sequences that involved at least one visit to other specialty care approximately 1 to 3 months prior to their ADRD diagnoses. The findings on the use of other specialty care are consistent with previous qualitative research indicating that some older adults are diagnosed with ADRD during a medical crisis due to other conditions. It also may be that the diagnosis of ADRD often involves a comprehensive workup to rule out other possible factors that may impact cognition. For example, patients may need to visit providers in other specialty care for further assessment, such as seeing an endocrinologist for thyroid issues. In addition, we found that older adults with more frequent use of other specialty care were slightly younger. It is possible that these older adults had more complex health conditions and/or comorbidities that required care from other specialties such as cardiology, orthopedics, and ophthalmology to help treat and manage their disease(s). Furthermore, due to these frequent healthcare utilizations, these older adults may be more likely to receive additional assessments and evaluations that may have led to a diagnosis of ADRD at a younger age. As the proportion of the older adult population continues to increase, it is likely that healthcare providers in non-ADRD-related specialties will also play a critical role in the care processes for ADRD diagnoses. Thus, more research is needed to assess the knowledge and awareness of ADRD in other specialties to optimize the care transitions for older adults with cognitive concerns.

Results from this study suggest several sociodemographic factors contribute to the different care sequences. At a context level, patients who resided in rural areas were more likely to follow sequences of care that included more primary care visits. It might be that primary care is more accessible than specialty care in rural areas. Unlike prior research, we did not find significant differences in the care sequences between older adults who lived in disadvantaged neighborhoods and those who did not. The reasons are potentially twofold. First, this study only included older adults with a diagnosis of ADRD documented in the EHR. Older adults who lived in disadvantaged areas may have under or undiagnosed ADRD due to their difficulties in access to care and/or inadequate documentation of their symptoms in the EHR, therefore were less likely to be included in the analysis. Second, the non-significant finding regarding neighborhood characteristics may be because less than 10% of study participants lived in a disadvantaged neighborhood. These numbers were consistent with prior work based on EHR data from this health system, which may partially reflect the catchment areas of DUHS. Future research with more socioeconomically diverse patient samples is needed to further assess how neighborhood characteristics are associated with the care sequences prior to ADRD diagnoses.

At the patient level, we found racial differences in the care sequences leading to the diagnoses of ADRD among women. In particular, women who are non-Hispanic Black were more like to follow care sequences that included multiple primary care or other specialty visits than their non-Hispanic White counterparts. These findings remained significant after accounting for other sociodemographic and clinical factors. It may be that women played a leading role in help-seeking in dementia care and Black older adults were more likely to seek treatment at a later stage of ADRD when the condition was more complex and required multiple visits for various exams and assessments. It is also possible that these multiple-visit care sequences reflect the difficulty Black women may encounter in accessing high-quality care and, thus, in having multiple visits to be able to fulfill their care needs. We encourage future studies to include more comprehensive psychosocial measures to further examine the interplay between sex and race in the care sequences leading to ADRD diagnoses.

With several new treatments for mild dementia now approved by the Food and Drug Administration, identifying patients at the early stages of ADRD has become more critical. Our study has identified significant variability in the care sequences leading to the diagnosis of ADRD. These findings have shed new light on potential key areas and critical time points for optimizing the care processes in diagnosing ADRD and, ultimately, addressing disparities in dementia care.

This study has several limitations. First, the care sequences were identified based on EHR data from a single academic medical center and its affiliated clinics in North Carolina, and only two racial groups were included in the analyses. Therefore, these findings may not be generalizable to patients from other geographic locations and/or from other racial/ethnic minority groups. Relatedly, although we limited our patients to those who reside in North Carolina, whether these patients utilized healthcare services outside of DUHS remained unknown. Third, we lacked information on the stage of ADRD at the time of diagnosis, which limited our ability to assess whether certain care sequences were associated with more (or less) timely diagnoses of ADRD. In addition, we did not assess whether the diagnoses of mild cognitive impairment (MCI) had an impact on the care sequences due to the inadequate documentation of MCI in the EHR. Lastly, we were not able to obtain information on individual-level socioeconomic background and/or the psychosocial support available to patients. These factors may have had an impact on the care sequences that patients followed.

CONCLUSION

In this EHR-based analysis, close to half the patients had one primary care visit a few months before their ADRD diagnosis. Other care sequences varied with regard to the type and frequency of the visits. We also found that age, sex, race, and rurality partially contributed to the variability in the care sequences that occurred prior to the diagnosis of ADRD. These findings have clinical implications that underscore the effort needed in primary care and non-ADRD specialties to promote high-quality dementia care through early detection and intervention.

ACKNOWLEDGMENTS

Results from this study were presented at the Gerontological Society of America's 73rd Annual Scientific Meeting. This research was funded in part by the National Institute on Aging (R01AG075210 for HX, MED) and the National Institute on Minority Health and Health Disparities (U54MD012530 for HX). The funding agency had no role in the design or conduct of the study, collection, analysis, or interpretation of the data or preparation, review, or approval of the manuscript.

CONFLICT OF INTEREST STATEMENT

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

CONSENT STATEMENT

This study used existing data from the electronic health records and the census, so consent was not necessary. This study was exempt by the DUHS Institutional Review Board.

References

2023 Alzheimer's disease facts and figures. Alzheimers Dement. 2023;19(4):1598‐1695. doi: [DOI: https://dx.doi.org/10.1002/alz.13016]