Introduction
Blood pressure variability (BPV) has emerged as a clinically meaningful cardiovascular risk factor independent of mean blood pressure [1, 2, 3, 4]. In large cohort studies and meta‐analyses, elevated visit‐to‐visit and day‐to‐day systolic BP variability predicts stroke, coronary events, and all‐cause mortality even after adjustment for average blood pressure levels [1, 3, 5, 6].
In addition to cardiovascular outcomes, BP variability has also been linked to cognitive decline and incident dementia, suggesting a broader neurovascular significance beyond traditional vascular risk assessment [6, 7, 8, 9]. High day‐to‐day systolic BPV (the top quartile of coefficient of variation) is associated with a more than two‐fold increased risk of all‐cause dementia (HR 2.27), vascular dementia (HR 2.79), and Alzheimer's disease (HR 2.22) [6]. Despite the growing clinical importance, the physiologic mechanisms underlying BP variability remain incompletely understood.
Traditional explanations attribute variability primarily to arterial stiffness, antihypertensive pharmacokinetics, medication adherence, or measurement error [1, 2, 3]. However, these mechanisms incompletely explain persistent within‐person variability observed across long observation periods [10, 11, 12]. Increasing evidence suggests blood pressure regulation depends heavily on central autonomic integration involving baroreflex pathways, circadian rhythmicity, and neurovascular coupling [13, 14].
Neurodegenerative diseases, particularly Alzheimer's disease, are characterized by progressive impairment in these regulatory systems, including reduced baroreflex sensitivity, autonomic imbalance, and disrupted diurnal blood pressure patterns [13, 14]. Autonomic dysfunction occurs in all common dementias, with orthostatic hypotension, a key marker of autonomic failure, showing a significantly higher prevalence in dementia patients than in healthy controls [14, 15]. Such dysfunction is associated with impaired cerebral perfusion regulation, and lower baseline cerebral perfusion has been shown to precede neurodegeneration, increasing the risk of subsequent dementia (HR 1.31) and accelerating cognitive decline [14, 15, 16].
In contrast, the mechanism for Parkinson's disease is degeneration of the cardiac sympathetic neurons and brainstem autonomic neurons due to abnormal accumulations of alpha‐synuclein proteins. Furthermore, levodopa or other dopamine receptor agonists may increase BPV through interactions with baroreflex and sympathoneural pathophysiological mechanisms that are part of the disease process [17].
Recent evidence further suggests BPV may not only accompany neurodegeneration but may also contribute to its development. Observational cohorts demonstrate that elevated visit‐to‐visit variability predicts worsening cognitive function, structural cerebrovascular injury, and increased risk of dementia independent of antihypertensive treatment [5, 6]. Mechanistically, Alzheimer's disease pathology contributes to autonomic dysfunction by affecting cortical and subcortical regions comprising the central autonomic network, with evidence of neurofibrillary degeneration and neuronal loss in the insular cortex, hypothalamus, and brainstem nuclei; regions responsible for cardiovascular homeostasis [13, 16].
Because these regions are affected early in the disease course, exaggerated BPV may represent a physiologic manifestation of preclinical neurodegeneration rather than a late consequence of cognitive decline [16].
Remote patient monitoring (RPM) provides a unique opportunity to evaluate these physiologic dynamics at scale. Unlike office measurements, which sample blood pressure intermittently and are susceptible to regression to the mean, home monitoring captures repeated measurements across weeks to months, allowing characterization of intrinsic day‐to‐day variability patterns [18, 19, 20]. Longitudinal home monitoring therefore enables separation of measurement noise and behavioral effects from persistent physiologic variability [4, 11].
Given the central role of autonomic regulation in neurodegenerative disease, we hypothesized that conditions involving central autonomic impairment would demonstrate greater day‐to‐day systolic BPV than in patients monitored for cardiometabolic disease alone, and that this effect would occur independent of overall blood pressure control.
Methods
Study Design and Data Source
We performed a retrospective observational analysis of de‐identified patient data derived from a cloud‐based healthcare analytics platform supporting a digital health remote monitoring and care management company operating in health systems across the United States. Data sources included structured diagnosis metadata captured within remote monitoring monthly patient records and longitudinal SBP and DBP summary measures produced during routine clinical monitoring. Longitudinal BP measurements transmitted across each patient's monitoring period were aggregated to produce patient‐level summary measures, including mean BP and day‐to‐day variability indices.
Institutional review board approval and informed consent were not required because this analysis used exclusively fully de‐identified data meeting the HIPAA de‐identification standard (45 CFR 164.514(b)). The study team had no access to identifiable information or any key or code permitting re‐identification. Accordingly, the analysis did not constitute human subjects research as defined under the U.S.
Common Rule (45 CFR 46.102(e)(1)).
Device and Measurement
Home blood pressure values were obtained using cellular‐enabled oscillometric blood pressure monitors capable of automatically transmitting measurements to the monitoring platform without manual patient data entry. Automated transmission minimizes transcription bias and preserves device‐generated timestamping of physiologic measurements [20, 21].
During program onboarding, patients received standardized education on home blood pressure measurement techniques consistent with American Heart Association recommendations for out‐of‐office blood pressure assessment [22]. Instruction included proper seated positioning, use of appropriately sized cuffs, avoidance of caffeine or recent physical activity prior to measurement, and the importance of resting quietly before measurement. Patients were also encouraged to obtain measurements under consistent conditions and at similar times of day whenever possible.
Education regarding measurement technique and adherence was reinforced during routine care navigation encounters.
Care Model Context
Patients participated in a structured remote blood pressure monitoring program integrated with routine care coordination workflows. Program activities included review of transmitted blood pressure readings, reinforcement of measurement adherence, patient education regarding lifestyle modification, and escalation according to clinician‐approved protocols when necessary. Care navigators documented patient‐reported symptoms, lifestyle counseling topics, and engagement barriers during routine care interactions.
Medication management decisions remained the responsibility of prescribing clinicians.
Cohort Construction
The analytic cohort was constructed using a predefined structured query executed in Databricks SQL against the longitudinal remote patient monitoring (RPM) measurement warehouse. Cohort inclusion criteria were designed to ensure reliable estimation of within‐person, day‐to‐day physiologic BPV over a sustained monitoring window. Patients were required to have transmitted systolic blood pressure (SBP) measurements on at least three distinct days in every week across a continuous monitoring stretch of at least 56 days (8 weeks).
Qualifying stretches were identified as runs of consecutive weeks in which each week contained at least three transmission days; a stretch was considered broken when more than seven days elapsed between qualifying weeks. For patients with more than one qualifying stretch, the longest stretch was selected for analysis. Stretch duration was defined as the number of days between the first and last transmitted reading within the selected stretch.
The three‐days‐per‐week, eight‐week minimum was implemented to ensure that day‐to‐day variability estimates reflected stable physiologic patterns sampled consistently over time rather than sparse or intermittent observations. After application of all inclusion criteria, the analytic cohort comprised 40,278 patients.
Patient Population
Patients were classified as having a neurodegenerative diagnosis based on the presence of ICD‐10 coded diagnoses beginning with G30 (Alzheimer's disease, including early and late onset), G20 (Parkinson's disease, with and without dyskinesia), or F02 and F03 (dementia in diseases classified elsewhere and unspecified dementia, respectively) recorded in the clinical record at the time of RPM enrollment. The comparator group comprised all remaining RPM patients without any of these diagnoses. The composition and overlap of diagnostic categories within the neurodegenerative cohort are summarized in Table 2.
BPV Measures
The primary outcome was day‐to‐day SBP variability, quantified using average real variability (ARV). For each patient, the absolute difference between consecutive blood pressure measurements was averaged across adjacent day pairs to yield ARV [9, 23, 24]. Unlike Standard Deviation (SD) and variability independent of the mean (VIM), ARV incorporates the temporal order of measurements and therefore reflects the magnitude of successive day‐to‐day changes [25, 26, 27].
The identical ARV procedure applied to diastolic values yielded day‐to‐day DBP variability, the principal secondary outcome.
The SD of BP measurements and VIM were calculated as secondary variability metrics for both SBP and DBP. VIM was considered secondary because the neurodegenerative and control groups were matched, in part, on mean SBP and DBP, thereby reducing between‐group differences in mean BP that VIM is intended to address. VIM was defined as VIM_i = SD_i × (mean BP_cohort / mean BP_i)^x, with the exponent x estimated separately for SBP and DBP by ordinary least squares regression of ln(SD) on ln(mean BP) across the complete analytic cohort (SBP: cohort mean = 133.11 mmHg, x = 0.7256; DBP: cohort mean = 78.63 mmHg, x = 0.3485); the same exponent and normalization mean were applied to both groups.
All variability metrics are reported in mmHg.
To provide additional context on the nature of blood pressure transmissions, we summarized the temporal distribution of readings. For each patient we computed the percentage of reading days with at least one measurement obtained during morning (3:00–10:00), midday (10:00–17:00), and evening (17:00–3:00) hours; because a single day may contain readings in more than one window, these percentages are not mutually exclusive and may sum to more than 100%. Average days per week with a transmitted reading and stretch duration were also retained as descriptors of monitoring intensity.
Propensity Score Matching
To address potential confounding, neurodegenerative patients were matched 1:3 to non‐neurodegenerative controls using propensity‐score matching on five pre‐specified continuous covariates: age, mean SBP during the stretch, mean DBP during the stretch, stretch duration, and average days per week with a transmitted reading. Propensity scores were estimated by logistic regression of neurodegenerative status on the five covariates. Matching was performed by nearest‐neighbor selection on the logit of the propensity score, without replacement, applying a caliper of 0.2 SDs of the logit.
Control patients with missing or implausible values on any matching covariate were excluded from the eligible pool. Average days per week and stretch duration were included as matching criteria specifically to ensure that differences in day‐to‐day variability could not be attributed to differences in measurement frequency or the length of the observation window, given that longer or denser monitoring may produce more stable variability estimates. The final matched dataset comprised 49 neurodegenerative patients and 147 matched controls (196 patients total).
Adequate balance across matching covariates was defined as absolute SMD <0.10 and variance ratios between 0.5 and 2.0.
Statistical Analysis
The BPV was evaluated using ARV as the primary metric, with SD and VIM as secondary metrics, separately for SBP and DBP. BPV measures were summarized as median (interquartile range) and compared between neurodegenerative and control groups using two‐sided Mann–Whitney U tests with tie and continuity corrections. Rank‐biserial correlations with 95% confidence intervals from 10,000 participant bootstrap resamples stratified by group were reported as effect sizes; VIM parameters were re‐estimated in each resample.
Spearman rank correlations between VIM and the corresponding mean BP were calculated across the analytic cohort as diagnostic checks for residual mean dependence. Exploratory subgroup analyses compared patients with Alzheimer's disease or dementia with patients with Parkinson's disease without a listed dementia diagnosis using the same metrics and tests, with rank‐biserial confidence intervals from 10,000 within‐subgroup bootstrap resamples; VIM in the subgroup analyses retained the transformation estimated in the complete analytic cohort. Day‐to‐day SBP BPV was pre‐specified as the primary outcome, whereas DBP BPV and subgroup analyses were considered secondary and exploratory.
Each patient was included once, and statistical significance was defined as p < 0.05 (two‐sided). Results for BPV measures are reported as median (interquartile range); other results are reported as mean ± SD unless otherwise indicated. Analyses were performed in Python 3 using SciPy, NumPy, pandas, and scikit‐learn.
Results
Study Population and Baseline Characteristics
Of the 40,278 patients meeting inclusion criteria, 49 carried a neurodegenerative diagnosis defined by ICD‐10 codes G30, G20, F02, or F03 and were matched 1:3 to 147 controls drawn from the 40,229 non‐neurodegenerative patients. Figure 1 depicts inclusion and matching criteria used to establish the analytic cohort. Baseline characteristics of the matched groups are presented in Table 1.
Four of the five matching covariates met the prespecified balance criterion of an absolute SMD below 0.10 (Table 1). A small residual difference remained in age (absolute SMD, 0.14). Mean age was 78.4 versus 79.7 years in the neurodegenerative and control groups, respectively; mean SBP during the stretch (133.8 vs.
133.2 mmHg) and mean DBP during the stretch (79.2 vs. 78.8 mmHg) were nearly identical, as were stretch duration (286.3 vs. 310.5 days) and average days per week with readings (5.3 vs.
5.3). The median monitoring‐period start date was December 18, 2023 in the neurodegenerative group versus May 8, 2023 in controls, with a corresponding median monitoring‐period midpoint of July 8, 2024 and October 23, 2023, respectively. All standardized mean differences were ≤ 0.14 in absolute value, and no covariate differed significantly between groups (all p > 0.35).
Sex was not included among the matching covariates and is reported descriptively; the neurodegenerative group was 54.4% female compared with 44.7% of controls. Sex was not specified in 9 patients, including 6 in the control group and 3 in the neurodegenerative cohort.
The composition of the neurodegenerative cohort is summarized in Table 2. Alzheimer's disease (G30) was present in 32 patients (65%), a dementia code (F02/F03) in 34 (69%), and Parkinson's disease (G20) in 13 (27%). The largest overlap was between Alzheimer's disease and dementia coding, present together in 30 patients (61%); 2 patients (4%) carried an Alzheimer's code only and 4 (8%) a dementia code only.
All 13 patients with Parkinson's disease (27%) had Parkinson's disease without a concurrent Alzheimer's or dementia code.
Primary Outcome: Day‐to‐Day SBP Variability
Patients with neurodegenerative disease demonstrated significantly greater day‐to‐day SBP ARV than matched controls (median, 12.92 mmHg [IQR, 10.74–15.35] vs 11.19 mmHg [IQR, 9.46–12.99]; Mann–Whitney (U = 4546); (p = 0.003). Median SBP ARV was therefore 1.73 mmHg, or 15.5%, higher in the neurodegenerative group (Figure 2).
Secondary Outcomes
Day‐to‐day DBP ARV was also significantly elevated in neurodegenerative patients (median 7.72 mmHg, IQR 6.34–8.87, vs. 6.76 mmHg, IQR 5.67–7.75; Mann–Whitney U = 4593.5, p = 0.002; rank‐biserial correlation = 0.293, 95% CI 0.096–0.478). Secondary SD and VIM metrics for SBP and DBP showed concordant results (all p ≤ 0.001; Table 3).
Spearman rank‐correlation analyses demonstrated no association between VIM and the corresponding mean BP across the analytic cohort (SBP: rs=−0.012, P=.866; DBP: rs=−0.021, P=.773), indicating that the VIM transformation effectively minimized residual dependence on mean BP. Primary and secondary outcomes are summarized in Table 3 and Figure 3.
Variability Within the Neurodegenerative Cohort
We next compared day‐to‐day variability between patients with Alzheimer's disease or dementia (n = 36) and those with Parkinson's disease only (n = 13) (Table 4, Figure 4). Day‐to‐day diastolic ARV was significantly higher in the Alzheimer's/dementia group (median 8.22 mmHg, IQR 6.77–9.10, vs. 6.53 mmHg, IQR 5.64–7.58; Mann–Whitney U = 343, p = 0.014; rank‐biserial correlation = 0.466, 95% CI 0.167–0.739); secondary SD and VIM metrics showed a concordant pattern (Table 4).
In contrast, day‐to‐day systolic ARV did not differ significantly between subgroups (median 13.55 mmHg, IQR 11.75–15.42, vs. 10.74 mmHg, IQR 9.27–13.54; U = 296, p = 0.164; rank‐biserial correlation = 0.265, 95% CI ‐0.141–0.650), with concordant non‐significant differences in the secondary SD and VIM metrics (Table 4). Interpret these non‐significant findings as no statistically detectable difference rather than equivalence, given the small Parkinson's‐only group.
Mean BP was descriptively similar between the diagnostic subgroups (SBP, 134.4 vs 131.9 mmHg; DBP, 78.9 vs 80.1 mmHg), although the subgroups were not matched directly to one another.
Measurement and Timing Characteristics
Measurement intensity and intraday timing were largely similar between matched groups (Table 5). Because average days per week and stretch duration were included as matching criteria, these characteristics were balanced by design. Patients in both groups transmitted readings on an average of 5.3 days per week and were monitored over comparable stretches of 286 and 311 days for neurodegenerative patients and controls, respectively (both p > 0.5).
The intraday distribution of readings was also similar. At least one measurement was recorded on 7.5% versus 8.4% of reading days during morning hours, 52.7% versus 61.7% during midday, and 47.9% versus 44.5% during evening hours for neurodegenerative patients and controls, respectively (all p > 0.12). Controls were more likely to record multiple readings on a given day (33.4% vs.
24.3% of reading days, p = 0.026). The overall similarity in measurement frequency and timing suggests that the observed differences in BPV are unlikely to be explained by differences in monitoring patterns.
Discussion
This study demonstrates that patients with neurodegenerative disease enrolled in a digital health RPM program exhibit significantly higher day‐to‐day SBP variability than propensity‐matched controls without a neurodegenerative diagnosis. In a matched cohort of 196 patients balanced on age, mean SBP and DBP during the monitoring stretch, stretch duration, and measurement frequency, neurodegenerative patients showed a median day‐to‐day SBP ARV of 12.92 mmHg (IQR 10.74–15.35) versus 11.19 mmHg (IQR 9.46–12.99) in controls (Mann–Whitney U = 4546, p = 0.003; rank‐biserial correlation = 0.280). Secondary SD and VIM metrics, and the diastolic ARV comparison, showed concordant results.
To our knowledge, this is among the first studies to characterize day‐to‐day BP variability in neurodegenerative disease using continuous, long‐term home monitoring data from a large real‐world RPM platform [28, 29, 30].
Interpretation of the Primary Finding
The magnitude and consistency of the observed difference support the hypothesis that neurodegenerative disease is associated with impaired long‐timescale blood pressure regulation. An elevation in day‐to‐day SBP ARV of this magnitude is clinically meaningful in the context of the existing literature, as visit‐to‐visit and day‐to‐day SBP variability have been independently associated with incident stroke, cardiovascular events, and accelerated cognitive decline in population‐based studies [31, 32]. The rank‐biserial correlation of 0.280 (95% CI, 0.083–0.470) reflects a small‐to‐moderate effect by conventional benchmarks, and the effect was concordant across the ARV, SD, and VIM metrics, reducing concern that the result is driven by the choice of variability metric or by distributional assumptions.
Importantly, the elevated variability was observed despite closely balanced mean SBP and DBP during the monitoring stretch between groups (133.8 vs. 133.2 mmHg and 79.2 vs. 78.8 mmHg, both non‐significant), indicating that the difference reflects dynamic BP dysregulation rather than a simple difference in average pressure.
Matching on mean SBP and DBP, and concordant results observed using VIM, reduced the likelihood that differences in average BP explained the observed variability difference. This is consistent with divergent mechanisms of autonomic dysfunction in Alzheimer's and Parkinson's disease, in which degeneration of central and peripheral autonomic circuits impairs baroreflex sensitivity and circadian BP patterning [33, 34].
Day‐to‐Day Variability and Autonomic Regulation
Day‐to‐day BPV reflects longer‐timescale regulatory processes, including circadian rhythm integrity, autonomic tone, and neurohormonal modulation, rather than the acute hemodynamic responses to posture and activity that dominate moment‐to‐moment fluctuation [11]. The elevation observed here is therefore consistent with disruption of the hypothalamic and brainstem autonomic nuclei known to be affected in Alzheimer's and Parkinson's disease, and with the more severe peripheral sympathetic denervation characteristic of Parkinson's disease, all of which would be expected to impair baroreflex‐mediated buffering of blood pressure across days [34, 35, 36]. Although this pattern aligns with known features of autonomic dysfunction in neurodegenerative disease, including impaired baroreflex sensitivity and disrupted circadian regulation, direct measures of autonomic function were not available in the present study.
Accordingly, these findings should be interpreted as physiologically suggestive rather than mechanistically definitive.
Heterogeneity within the Neurodegenerative Cohort
Within the neurodegenerative cohort, day‐to‐day systolic ARV was similar across diagnostic subgroups, whereas diastolic ARV was significantly higher in patients with Alzheimer's disease or dementia than in those with Parkinson's disease alone (median 8.22 vs. 6.53 mmHg; Mann–Whitney U = 343, p = 0.014; rank‐biserial correlation = 0.466). The comparable systolic ARV suggests that elevated day‐to‐day systolic fluctuation is a shared feature across these conditions, consistent with overlapping degeneration of brainstem and hypothalamic structures involved in cardiovascular regulation [34, 35].
The divergence in diastolic variability may reflect differences in the predominant autonomic abnormalities across neurodegenerative disorders. Alzheimer's disease and related dementias are characterized largely by central autonomic network dysfunction, whereas Parkinson's disease is notable for peripheral sympathetic denervation in addition to central autonomic involvement [16, 35]. These differing patterns of autonomic impairment could contribute to distinct effects on blood pressure regulation, although the present study cannot directly address underlying mechanisms.
The sample sizes within each subgroup are modest, however, particularly the Parkinson's‐only group (n = 13), and these analyses should be interpreted cautiously; adequately powered subgroup comparisons remain an important direction for future research. These subgroup comparisons are unadjusted and do not establish causal differences between the underlying diseases.
Measurement Behavior
Additional analyses of measurement behavior indicate that the observed variability differences are not explained by differences in measurement frequency or timing. Neurodegenerative patients and controls transmitted readings on a nearly identical number of days per week (5.3 vs. 5.3) over comparable monitoring stretches, and the intraday distribution of readings across morning, midday, and evening windows was similar between groups.
Because measurement frequency and stretch duration were matching criteria, and intraday timing did not differ, these findings strengthen the interpretation that elevated day‐to‐day BP variability reflects true physiologic dysregulation rather than behavioral or measurement artifacts.
Clinical Implications
These findings have several potential clinical implications. First, elevated day‐to‐day SBP variability in neurodegenerative patients may represent an underappreciated cardiovascular risk factor in a population that already exhibits elevated stroke risk. A standard clinic‐based BP assessment captures a single time point and is poorly suited to detect increased variability; RPM offers a scalable approach to quantify this risk continuously in the outpatient setting.
Second, day‐to‐day BP variability may serve as a remotely accessible biomarker of autonomic dysfunction in neurodegenerative disease. As RPM programs expand across health systems, longitudinal BP variability data are generated passively as a byproduct of routine monitoring; incorporating variability metrics into clinical dashboards alongside mean BP could flag patients with neurodegeneration who may benefit from additional autonomic workup, medication review (particularly agents affecting BP regulation such as cholinesterase inhibitors and dopaminergic medications), or falls‐risk assessment.
Strengths and Limitations
Strengths
This study benefits from several methodological strengths. The RPM platform provides longitudinal BP measurements obtained under standardized conditions using a cellular‐enabled device with automatic transmission, avoiding common biases associated with self‐reported or manually entered home BP data. The inclusion criterion of at least three reading days per week (resulting in a mean of 5.3 days/week) over an eight‐week stretch ensures dense, temporally consistent sampling well suited to estimating day‐to‐day variability.
The large control pool (n = 40,229) enabled robust 3:1 propensity matching, yielding groups well balanced across five covariates, including mean SBP and DBP during the stretch and measurement frequency. The primary result was replicated across secondary SD and VIM metrics and was concordant with the diastolic ARV finding.
Limitations
Several limitations should be noted. The neurodegenerative cohort is modest in size (n = 49 after matching), limiting statistical power for subgroup analyses and generalizability. Diagnoses were ascertained from administrative coding within the RPM platform and may not reflect clinical severity, disease duration, or the full diagnostic picture; misclassification of neurodegenerative status cannot be excluded.
The retrospective observational design precludes causal inference, and unmeasured confounders, including medication use (particularly antihypertensives, cholinesterase inhibitors, and dopaminergic agents), physical activity, and cognitive severity, may influence the observed variability differences. Sex was not included among the matching covariates and differed modestly between groups (54.4% vs. 44.7% female), representing a potential residual confounder; age balance was also marginally outside the conventional threshold (SMD 0.14).
Control patients with missing or implausible matching data were excluded, which may introduce selection effects, and ethnicity data were incomplete and could not be incorporated into matching. The Mann–Whitney comparisons were unadjusted and did not account for individual matched‐set structure; VIM was calibrated and evaluated in the same cohort, and minimal cohort‐wide correlation with mean BP does not establish independence in all subgroups. Finally, the study does not include direct measures of autonomic function, such as heart rate variability, baroreflex sensitivity, or tilt‐table testing, limiting the ability to definitively attribute the observed differences to autonomic dysfunction.
Future Directions
Future studies should seek to replicate these findings in larger, independently ascertained neurodegenerative cohorts with prospective design and detailed clinical characterization, including disease severity, medication reconciliation, and cognitive testing. Examining whether day‐to‐day SBP and DBP variability predict longitudinal outcomes, including cognitive decline trajectory, falls, or cardiovascular events, in RPM‐enrolled patients with neurodegeneration would substantially advance the clinical utility of these findings. Incorporating time‐of‐day BP patterns and nocturnal dipping data, where available, could further elucidate the autonomic mechanisms underlying the observed high variability.
While propensity‐score matching strengthened internal validity by balancing groups on age, mean SBP and DBP, monitoring duration, and measurement frequency, it cannot account for unmeasured or unavailable confounders. Medication classes and comorbidities that influence autonomic function may contribute to residual variability differences despite matching. Accordingly, future analyses incorporating multivariable adjustment for antihypertensive classes, autonomically active medications, and cardiometabolic comorbidities will be critical to further isolate disease‑specific effects.
Conclusions
Patients with neurodegenerative disease enrolled in a real‐world RPM program demonstrate significantly elevated day‐to‐day SBP and DBP variability, quantified using ARV with concordant secondary SD and VIM metrics, compared with matched controls, independent of mean blood pressure, age, monitoring duration, and measurement frequency. Within the cohort, diastolic ARV was higher in Alzheimer's disease and dementia than in Parkinson's disease alone, while systolic ARV was shared across subgroups. These findings, consistent with impaired autonomic blood pressure regulation in neurodegeneration, highlight the potential of RPM‐derived BP variability metrics as clinically accessible markers of autonomic dysfunction and cardiovascular risk in this vulnerable population.
Author Contributions
W.S. and C.F. contributed to the conceptualization and design of the study. W.S. and C.F. conducted the formal analysis and prepared the original manuscript draft. J.F.S., A.A.S., D.P.C., and B.L. provided critical review and revisions for important intellectual content.
C.F. coordinated manuscript preparation.
Funding
No external funding was received for this study.
Ethics Statement
This study used fully de‐identified data and was conducted under a data use agreement permitting analysis and reporting of aggregated results. Institutional review board oversight was not required due to the use of de‐identified data.
Consent
The authors have nothing to report.
Conflicts of Interest
Authors WS and CF are employees of HealthSnap, Inc. and hold equity in the company.
Clinical Trial Registration
The authors have nothing to report.
Permission to Reproduce Material from Other Sources
The authors have nothing to report.