Health Care Utilization in Veterans With Alzheimer Disease

ABSTRACT

Objective: To evaluate health care utilization in veterans with Alzheimer disease (AD) in the Veterans Affairs health system (VAHS).

Study Design: This retrospective analysis identified veterans with AD using clinical notes extracted from the VAHS electronic health record from fiscal years 2010 to 2019.

Methods: The first note identifying AD was the index date. Health care utilization in veterans with AD and a 1:1 matched comparison group without AD was evaluated at 2 years preindex, 1 year preindex, 1 year post index, and 2 years post index.

Results: From clinical notes, we identified 571,671 veterans with AD and 571,671 for the comparison group (overall: mean age, 74 years; 96% male; 75% White). In those with AD, outpatient visits per patient per year peaked 1 year post index at 67 and remained elevated 2 years post index at 57; without AD, the rate was approximately 19 at all time points. Hospitalization rates peaked at 1 year post index with AD but were lower and generally stable without AD. Nursing home utilization was relatively low overall. Veterans meeting the 2-code criteria (n = 56,305), defined as having 2 diagnostic codes for AD recorded at least 30 days apart, had consistently higher utilization than veterans without AD (especially post index).

Conclusion: Veterans with AD have higher health care utilization than veterans without AD, especially around the time of AD diagnosis.

Am J Manag Care. 2026;32(6):In Press

Takeaway Points

Identification of Alzheimer disease (AD) in the Veterans Affairs health system (VAHS) using clinical notes from the electronic health record may help address limitations associated with diagnostic code–based identification. Overall, veterans with AD had higher health care utilization than a matched comparison group without AD.

• An AD diagnosis is associated with increased health care utilization in the VAHS, especially around the time of diagnosis.
• In the AD cohort, the mean outpatient visits, inpatient hospitalization, and nursing home utilization rates notably increased from 2 years preindex to 1 year preindex and further increased at 1 year post index; by 2 years post index, rates had generally declined but remained comparable to or higher than the 1-year preindex rates. In contrast, outpatient visit rates in the comparison non-AD group were generally stable.

Alzheimer disease (AD) is a progressive neurodegenerative condition and the seventh leading cause of mortality in the US.¹ An estimated 6.7 million individuals in the US have AD, and the number is projected to rise to more than 13.8 million by 2060.² As AD progresses from early to later stages of disease, affected individuals become increasingly debilitated, requiring more assistance with activities of daily living and personal care, and they may eventually require adult day care or full-time institutional care.³

Compared with the general population, veterans may be at a higher risk for AD due to older age and/or higher prevalence of certain factors associated with the development of AD and related dementias, such as traumatic brain injury and posttraumatic stress disorder (PTSD).^4-7 A recent epidemiologic study of those enrolled in the Veterans Affairs health system (VAHS) estimated that in 2019 there were 31,864 veterans with 2 or more diagnostic codes for AD out of more than 5 million veterans in the system (prevalence of 0.41%)⁸; however, this figure is likely an underestimation of the actual prevalence of AD in the VAHS due to limitations of diagnostic code–based identification for AD, such as more common use of nonspecific dementia codes.⁹ Clinical notes in the VAHS electronic health record (EHR) provide another data source for identifying veterans with AD.

There are limited data regarding the impact of AD on health care utilization by veterans in the VAHS. In this study, we used clinical notes to identify veterans with AD and compared their health care utilization with that of a matched comparison group of veterans without AD in the VAHS.

METHODS

Data Source and Extraction

This retrospective analysis utilized the VA Informatics and Computing Infrastructure (VINCI) database¹⁰ and applied Text Integration Utilities (TIU) to query clinical notes. We extracted clinical notes of veterans with AD from the VAHS EHR using targeted keyword searches from fiscal year (FY) 2010 through FY 2019.¹¹ FYs began on October 1 and ended on September 30 of the next year, and they were designated by the calendar years in which they ended. This study was approved by the VA Bedford Healthcare System Institutional Review Board, and all data were fully deidentified before access.

Study Population

Veterans with at least 1 AD qualifying note that met iterative search criteria between FYs 2010 and 2019 were included in the study. Mini-Mental State Examination and Montreal Cognitive Assessment scores for these veterans were included for disease diagnosis in the clinical notes, and the cohort was created based on our prior published method.^11-14

Those in the comparison group, who did not have an AD qualifying note or a diagnostic code for AD, were directly matched 1:1 to veterans in the TIU-identified AD cohort based on age, sex, and race. For veterans with a date of death available, the comparison group was propensity score–matched based on age at death in addition to age, sex, and race. In the AD cohort, the index date was defined as the date of the first TIU-identified note with an AD diagnosis. The index date of each veteran with AD was assigned to a veteran in the matched comparison group without AD.

Study End Points

The primary study end point was health care utilization 2 years before and after the first qualifying AD note (index date), excluding utilization less than 30 days before or after the index date. Health care utilization outcomes included outpatient visits, hospitalizations, hospitalization-associated length of stay (LOS), nursing home (NH) visits, and NH visit–associated LOS. If a veteran had multiple visits (ie, distinct clinical encounters) in a single day, each one was counted.

Health care utilization was also assessed in a subgroup of veterans who had at least 1 qualifying clinical note for AD plus at least 2 International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) or ICD, Tenth Edition, Clinical Modification (ICD-10-CM) diagnostic codes for AD (ICD-9-CM: 331.0; ICD-10-CM: G30.X) that were 30 or more days apart. In this subgroup, the earliest date between the note and ICD code was referenced as the index date, and veterans in the matched comparison group were assigned the same index date. This subgroup has increased confidence in the AD diagnosis due to veterans having both a note and 2 codes for AD.

Statistical Analysis

We used descriptive statistics to present baseline demographic and comorbid conditions for both the veterans with and without AD. Baseline comorbidities were identified by the presence of a single ICD-9-CM or ICD-10-CM diagnostic code (eAppendix Table 1) (eAppendix available at ajmc.com). Categorical variables were described using counts and percentages, whereas continuous variables were described using means and SDs. Health care utilization between the 2 cohorts was compared at the following time points: 2 years preindex, 1 year preindex, 1 year post index, and 2 years post index, using t tests for mean comparison and z tests for proportion comparison. An α of .05 was considered significant.

RESULTS

Overall Study Cohorts

The overall study sample included 571,671 veterans who met the inclusion and exclusion criteria for AD notes and 571,671 veterans in the matched comparison group without AD notes. The mean (SD) age was 74 (11) years, and the majority (96%) of veterans were men in both the AD and matched comparison groups (Table 1 and eAppendix Figure). In both cohorts, approximately 75% of the veterans were White and 14% were Black/African American. Hispanic ethnicity was documented in 5% and 3% of veterans with and without AD, respectively.

The overall comorbidity burden was higher in veterans with AD vs without AD. The most common comorbidities were hypertension (76.7% with AD vs 55.7% without AD) and hyperlipidemia (72.4% vs 51.7%, respectively). The burden of behavioral and psychiatric comorbidities was markedly higher in the AD vs comparison group: depression (50.0% vs 21.9%), anxiety (27.8% vs 11.5%), bipolar/mania (22.4% vs 7.8%), and PTSD (19.1% vs 7.6%).

Health Care Utilization Associated With AD: Overall Cohort

The health care utilization of the TIU-identified AD cohort and comparison group without AD is summarized in Table 2. At almost all time points assessed, veterans with AD had markedly greater health care utilization than those without AD (P < .01 for all comparisons excluding 1 pair; ie, hospitalization-
associated LOS days per hospitalized patient at 2 years preindex).

Outpatient utilization. Most veterans in both groups had an outpatient visit in each assessment period; however, the proportion of veterans with outpatient visits was consistently at least 30% higher in the AD vs non-AD groups (P < .01 at all time points): 86.7% vs 60.4% at 2 years preindex, 94.8% vs 60.8% at 1 year preindex, 95.9% vs 59.0% at 1 year post index, and 82.0% vs 54.4% at 2 years post index. The mean (SD) outpatient visits per patient per year in the AD cohort increased from 45.9 (76.5) at 2 years preindex to 55.6 (89.8) at 1 year preindex and further increased to 67.0 (108.4) at 1 year post index, followed by a decrease to 56.3 (107.2) at 2 years post index (P < .01 for all comparisons between time points). The rate of outpatient visits in the comparison group was generally stable (approximately 19 visits per patient per year at each time point).

Hospital utilization. The proportion of veterans with a hospitalization was consistently higher in the AD vs comparison group (P < .01 at all time points): 11.2% vs 4.2% at 2 years preindex, 14.4% vs 4.5% at 1 year preindex, 16.4% vs 4.8% at 1 year post index, and 12.6% vs 4.4% at 2 years post index (Table 2). The mean (SD) hospitalizations per 1000 patients per year in veterans with AD increased from 196.6 (719.7) at 2 years preindex to 265.2 (851.4) at 1 year preindex and further increased to 300.9 (894.0) at 1 year post index, followed by a decrease to 232.4 (804.7) at 2 years post index; at all time points, the hospitalization rate with AD was more than double the rate without AD (P < .01 for all comparisons between time points) (eAppendix Table 2). The mean LOS associated with hospitalization in the AD cohort ranged from approximately 5 to 7 days and was slightly longer (1 or 2 days) than that in the comparison group at the 1-year preindex and postindex time points (P < .01).

NH utilization. The proportion of veterans with NH admissions was low and generally stable across all time points, but consistently higher (P < .01) in the AD cohort (1.1%-3.0%) than in the comparison group (0.3%-0.6%) (Table 2). The mean (SD) number of NH admissions per 1000 patients per year in the AD cohort increased from 13.7 (142.2) at 2 years preindex to 22.9 (178.7) at 1 year preindex and further increased to 35.8 (222.7) at 1 year post index, followed by a decrease to 26.4 (199.3) at 2 years post index (P < .01 for all comparisons between time points) (eAppendix Table 2). The rate of NH visits in the comparison group was relatively low, ranging from 3 to 6 visits per 1000 patients per year across all 4 time points, with a slight increase from the preindex to the postindex period. The mean LOS associated with NH admission in the AD cohort ranged from approximately 48 to 82 days and was at least 6 days longer than that in the comparison group from the 2-year preindex time point onward (P < .01).

Health Care Utilization Associated With AD: Subgroup With AD Note Plus 2 AD Diagnostic Codes

A total of 56,305 veterans with an AD note in FYs 2010-2019 met the 2-code requirement for AD (Table 3). For all utilization parameters assessed, the AD cohort with note plus 2 diagnostic codes had markedly higher utilization rates than the matched comparison group (P < .01 at all time points). Outpatient visits per patient per year in the AD cohort ranged from approximately 32.4 to 38.0 in the preindex years to approximately 61.8 to 62.4 in the postindex years. For the comparison group, outpatient visits per patient per year ranged from 16.6 to 20.9 across all time points. Hospitalizations per 1000 patients per year in the overall AD cohort rose from 110.3 to 145.4 during the preindex years to 234.7 to 241.9 during the postindex years. In the comparison group, hospitalizations ranged from 48.5 to 69.2 per 1000 patients per year throughout. NH admissions per patient also rose markedly from approximately 7.4 to 10.9 visits per 1000 patients per year during the preindex years to approximately 41 visits per 1000 patients per year during the postindex years. In the comparison group, NH utilization ranged from 2.0 to 4.1 visits per 1000 patients per year throughout. LOS associated with hospitalization and with NH admission was generally longer in veterans with AD than without AD, especially at the postindex time points. Hospitalization-associated LOS with AD slightly rose from the range of 5 to 6 days in the preindex years to 7 days in the postindex years; similarly, NH LOS with AD rose from 37 to 56 days in the preindex years to 60 to 85 days in the postindex years.

DISCUSSION

Given the projected rise in the number of people with AD,² it is imperative that health care systems understand the implications for appropriate allocation of resources and effective health care planning. Our retrospective analysis of the VINCI database found that the TIU-identified AD cohort had significantly greater health care utilization than the age-, sex-, and race-matched comparison group without AD. The proportion of veterans with hospitalizations, NH admissions, and outpatient visits was consistently higher in the AD cohort in all assessment periods. In addition, veterans with AD had longer LOS associated with hospitalization or NH admission than those without AD. This finding of increased health care utilization with AD is consistent with the literature.¹⁵ 

Interestingly, the utilization trend was evident even before the AD diagnoses were initially identified in the EHR (index). Among patients with AD, all utilization types rose markedly from the initial study time point (2 years preindex) to the 1-year preindex time point, and continued to rise post index, whereas utilization in the matched comparison group did not change markedly over time. Notably, the AD group had a substantial baseline comorbidity burden, particularly in terms of psychiatric comorbidities such as depression and anxiety, which were more than double that of the comparison group without AD. It is possible that some veterans had AD-related signs or symptoms preceding their diagnosis that were not recognized, suggesting an opportunity to identify patients earlier and potentially implement interventions that may lessen resource requirements.

Our findings in the subgroup with a clinical note for AD plus 2 diagnostic codes for AD were consistent with our findings in the overall AD cohort, with the largest rise in utilization observed between the 1-year pre- and postindex time points. Collectively, these findings suggest that a new AD diagnosis may be associated with greater disease burden and management needs.

Limitations

The current study provides insight into assessments of AD severity¹⁶ in one of the largest US managed care settings; however, several limitations must be considered. Retrospective analyses are more susceptible to confounding variables than prospective studies.¹⁷ We extracted data from clinical notes to identify veterans with AD; the data included disease stage based on cognitive evaluations when available. This method allowed us to efficiently examine a large volume of clinical note data and overcome issues with diagnostic code–based cohort identification (ie, undercoding or miscoding, as well as lack of disease stage information in codes), but text- or keyword-based data extraction may also misidentify information. Comorbidities were still identified using diagnostic codes, which are subject to inaccuracies and/or undercoding.¹⁸ Future studies may examine extracting comorbidity information from clinical notes as well. Among veterans identified via a clinical note for AD, 10.6% had a comorbid diagnostic code for unspecified dementia at baseline (Table 1). This overlap may represent variations in use of diagnostic codes across clinician types or cases in which an unspecified dementia code was assigned prior to the clinical AD diagnosis. Another limitation is that our current analysis did not assess measures of health care utilization, such as medication use, imaging studies, and laboratory use, nor could we capture any health care utilization that may have occurred outside the VAHS, such as NH admissions at an external facility. It is possible that a portion of veterans with later-stage AD or older age may have left the VAHS or died. Because this was not a prospective longitudinal analysis, we could not evaluate how utilization may change over time with disease progression. Additionally, we did not explore the impact of death on utilization specifically; AD and mortality will be investigated in a separate analysis. Finally, these findings from a population of veterans in the VAHS may not be generalizable to the overall US population and health care system. Future studies will be carried out using different databases, such as CMS’ databases, which will allow us to include similar numbers of male and female participants and understand the resource utilization for patients with AD.

CONCLUSIONS

Our study provides a real-world assessment of health care utilization associated with AD in a veteran population. We found higher health care utilization in veterans with AD compared with those without AD, with particularly high utilization burden in the years closest to the AD diagnosis. Our results highlight some of the challenges and opportunities that will need to be considered. Follow-up studies may identify patterns of health care utilization that will be useful for early identification of AD and mitigation of the burden on health care systems.

Acknowledgments

The views expressed in this article are those of the authors and do not represent the views of the US Department of Veterans Affairs or the US government. The authors thank Kulvinder Katie Singh, PharmD, for providing a scientific edit of the manuscript, and John Wells, PhD, for helpful discussion.

Author Affiliations: Geriatric Research Education and Clinical Center (YW, BJA, WX) and Center for Health Optimization and Implementation Research (ML), VA Bedford Healthcare System, Bedford, MA; Department of Applied Mathematics, Wentworth Institute of Technology (YW), Boston, MA; Department of Mathematical Sciences, Bentley University (ML), Waltham, MA; Department of Neurology (PM) and Department of Pharmacology, Physiology & Biophysics (WX), Boston University Chobanian & Avedisian School of Medicine, Boston, MA; Department of Biological Sciences, Kennedy College of Sciences, University of Massachusetts Lowell (JR, WH, WX), Lowell, MA; Department of Public Health, Zuckerberg College of Health Sciences, University of Massachusetts Lowell (JSL, DB), Lowell, MA; Department of Mathematics, University of Edinburgh (QL), Edinburgh, UK; Alzheimer’s Disease and Brain Health, Eisai Inc (RZ, AATM, QZ), Nutley, NJ; Epidemiology, Biostatistics and Occupational Health, McGill University (AATM), Montreal, Quebec, Canada.

Source of Funding: Eisai Inc and National Institutes of Health grant R01AG063913.

Author Disclosures: Dr Tahami Monfared, Dr Raymond Zhang, and Dr Quanwu Zhang report being employees of Eisai Inc. Dr Xia reports receiving a research grant from Eisai Inc. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (ML, PM, DB, RZ, AATM, QZ, WX); acquisition of data (BJA, ML, JR, WH, QZ); analysis and interpretation of data (YW, BJA, ML, PM, JR, WH, JSL, DB, QL, AATM, QZ, WX); drafting of the manuscript (BJA, JR); critical revision of the manuscript for important intellectual content (YW, BJA, PM, JSL, DB, RZ, AATM, QZ); statistical analysis (YW, WH, JSL, QL, QZ); obtaining funding (WX); administrative, technical, or logistic support (YW, BJA, JR, WH, QL, WX); and supervision (RZ, QZ, WX).

Address Correspondence to: Weiming Xia, PhD, VA Bedford Healthcare System, 200 Springs Rd, Bedford, MA 01730. Email: weiming.xia@va.gov.

REFERENCES

1. Ahmad FB, Cisewski JA, Xu J, Anderson RN. Provisional mortality data - United States, 2022. MMWR Morb Mortal Wkly Rep. 2023;72(18):488-492. doi:10.15585/MMWR.MM7218A3

2. 2023 Alzheimer’s disease facts and figures. Alzheimers Dement. 2023;19(4):1598-1695. doi:10.1002/ALZ.13016

3. Deb A, Thornton JD, Sambamoorthi U, Innes K. Direct and indirect cost of managing Alzheimer’s disease and related dementias in the United States. Expert Rev Pharmacoecon Outcomes Res. 2017;17(2):189-202.
doi:10.1080/14737167.2017.1313118

4. Zhu CW, Sano M. Demographic, health, and exposure risks associated with cognitive loss, Alzheimer’s disease and other dementias in US military veterans. Front Psychiatry. 2021;12:610334. doi:10.3389/FPSYT.2021.610334

5. Weiner MW, Friedl KE, Pacifico A, et al. Military risk factors for Alzheimer’s disease. Alzheimers Dement. 2013;9(4):445-451. doi:10.1016/J.JALZ.2013.03.005

6. Sibener L, Zaganjor I, Snyder HM, Bain LJ, Egge R, Carrillo MC. Alzheimer’s disease prevalence, costs, and prevention for military personnel and veterans. Alzheimers Dement. 2014;10(suppl 3):S105-S110. doi:10.1016/J.JALZ.2014.04.011

7. VA Utilization Profile: FY 2016. US Department of Veterans Affairs. November 2017. Accessed May 11, 2023. https://web.archive.org/web/20230507072814/https://www.va.gov/vetdata/docs/Quickfacts/VA_Utilization_Profile.pdf

8. Dinesh D, Shao Q, Palnati M, et al. The epidemiology of mild cognitive impairment, Alzheimer’s disease and related dementia in U.S. veterans. Alzheimers Dement. 2023;19(9):3977-3984. doi:10.1002/ALZ.13071

9. Butler D, Kowall NW, Lawler E, Gaziano JM, Driver JA. Underuse of diagnostic codes for specific dementias in the Veterans Affairs New England healthcare system. J Am Geriatr Soc. 2012;60(5):910-915. doi:10.1111/J.1532-5415.2012.03933.X

10. VA Informatics and Computing Infrastructure (VINCI). US Department of Veterans Affairs. Accessed May 7, 2026. https://www.research.va.gov/programs/vinci/

11. Aguilar BJ, Miller D, Jasuja G, et al. Rule-based identification of individuals with mild cognitive impairment or Alzheimer’s disease using clinical notes from the United States Veterans Affairs Healthcare System. Neurol Ther. 2023;12(6):2067-2078. doi:10.1007/s40120-023-00540-2

12. Morin P, Li M, Wang Y, et al. Clinical staging of Alzheimer’s disease: concordance of subjective and objective assessments in the Veteran’s Affairs Healthcare System. Neurol Ther. 2022;11(3):1341-1352. doi:10.1007/S40120-022-00379-Z

13. Nasreddine ZS, Phillips NA, Bédirian V, et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc. 2005;53(4):695-699. doi:10.1111/J.1532-5415.2005.53221.X

14. Perneczky R, Wagenpfeil S, Komossa K, Grimmer T, Diehl J, Kurz A. Mapping scores onto stages: mini-mental state examination and clinical dementia rating. Am J Geriatr Psychiatry. 2006;14(2):139-144. doi:10.1097/01.JGP.0000192478.82189.a8

15. Park S, White L, Fishman P, Larson EB, Coe NB. Health care utilization, care satisfaction, and health status for Medicare Advantage and traditional Medicare beneficiaries with and without Alzheimer disease and related dementias. JAMA Netw Open. 2020;3(3):e201809. doi:10.1001/jamanetworkopen.2020.1809

16. Chen YH, Lai YC, Wu YC, Sasaki J, Tsai KT, Ho CH. Healthcare utilization in different stages among patients with dementia: a nationwide population-based study. Int J Environ Res Public Health. 2021;18(11):5705. doi:10.3390/IJERPH18115705

17. Mann CJ. Observational research methods: research design II: cohort, cross sectional, and case-control studies. Emerg Med J. 2003;20(1):54-60. doi:10.1136/EMJ.20.1.54

18. Wilchesky M, Tamblyn RM, Huang A. Validation of diagnostic codes within medical services claims. J Clin Epidemiol. 2004;57(2):131-141. doi:10.1016/S0895-4356(03)00246-4