Evidence-based findings and device performance — summary of clinical evidence on actigraphy and accelerometry performance, strengths, limitations, and implications for clinical use.
Actigraphy provides objective estimates of sleep versus wake and has high sensitivity (>90%) for detecting sleep and good concordance with polysomnography (PSG) for total sleep time (TST) in healthy subjects, but its specificity for detecting wakefulness is low, especially in populations with fragmented sleep or challenged sleep-wake cycles (e.g., jet lag, shift work) (Paquet et al., 2007; Martin & Hakim, 2011).
Actigraphy tends to over-estimate sleep (TST) and sleep efficiency when wakefulness is characterized by immobility (e.g., insomnia, hospitalized or bed-bound patients), leading to potential under-estimation of sleep disturbance severity and limited validity for sleep-onset latency and daytime sleep measurements (Martin & Hakim, 2011).
Device- and algorithm-dependent variability is substantial: different actigraphs and scoring algorithms yield differing estimates (e.g., threshold vs regression algorithms), affecting wake detection and number of awakenings (Paquet et al., 2007; Gschliesser et al., 2009; Plante, 2014).
For detection of nocturnal awakenings in young children, actigraphy shows high overall agreement and sensitivity but very low specificity compared with videosomnography, indicating poor detection of awakenings and potential misclassification (Sitnick et al., 2008).
Actigraphy is not validated to measure sleep stages or to evaluate arousals without concurrent EEG; therefore it cannot replace PSG when sleep staging or arousal detection is required (Martin & Hakim, 2011).
Regarding periodic limb movements in sleep (PLMS), actigraphy devices show variable performance: some devices under-estimate PLM indices (Actiwatch) while others over-estimate (PAM-RL); although correlations with PSG may be high, mean values differ and actigraphy cannot be used interchangeably with PSG/EMG for diagnostic decision-making (Gschliesser et al., 2009; Plante, 2014; Athavale et al., 2017).
Advanced signal processing and machine learning (e.g., Naïve-Bayes classifiers) applied to high-frequency bilateral ankle actigraphy can improve detection/classification of PLMS (reported sensitivity ~80%, specificity ~74% in one study), but these approaches require further validation before routine clinical use (Athavale et al., 2017).
Clinical practice guidelines conditionally support actigraphy for estimating sleep parameters in insomnia and circadian rhythm disorders and as an adjunct in several settings (e.g., prior to MSLT, integrated with home sleep apnea testing to estimate TST), but strongly recommend against using actigraphy in place of electromyography for diagnosis of periodic limb movement disorder (AASM; summarized in Martin & Hakim, 2011 and guideline statements).
In mood disorders and bipolar disorder research, pooled and individual studies indicate actigraphy-derived metrics (daily activity, wake after sleep onset, sleep efficiency) differ between patients and controls and may change with treatment; however, heterogeneity of devices, small sample sizes, and inconsistent clinical linkage limit current clinical utility and require more research (Tazawa et al., 2019; Difrancesco et al., 2021; Etain/Panchal studies).
For epilepsy/seizure monitoring, multimodal systems combining accelerometry with heart rate and other signals show promise for detecting convulsive nocturnal motor seizures, but sensitivity and false detection rates vary widely across devices and seizure types; non-generalized and non-rhythmic seizures are frequently missed (Van de Vel et al., 2014; Van de Vel pilot 2016; Milosevic et al., 2016).
Accelerometry in movement disorders (e.g., Parkinson disease, essential tremor) can provide objective measures of tremor frequency, amplitude, bradykinesia and dyskinesia and can correlate with clinical scales; some systems (e.g., Kinesia, PKG) produce quantitative scores showing fair-to-good association with clinical/PSG measures and potential utility for longitudinal monitoring, but evidence that accelerometry use improves patient outcomes is insufficient (Godfrey et al., 2008; Cleveland Medical Devices/Kinesia clearance; McGregor et al., 2018; Bove et al., 2018).
In Parkinson disease, specific accelerometry-derived scores (e.g., PKG bradykinesia-based sleep metrics) demonstrated reasonable sensitivity and specificity to distinguish normal vs abnormal PSG and correlated with patient-reported assessments, suggesting utility for screening/monitoring but not replacement of standard testing (McGregor et al., 2018).
In stroke and rehabilitation research, accelerometry yields valid and reliable data on activity, steps, gait asymmetry, and upper-extremity use; however, responsiveness, predictive value, and minimal clinically important differences remain incompletely defined, limiting direct application as a standalone clinical outcome measure (accelerometry stroke literature).
In trials assessing daily life physical activity (DLPA) outcomes (e.g., pulmonary arterial hypertension/TRACE trial), accelerometry-derived endpoints were feasible with high compliance but changes were small, highly variable, and sensitive to environmental factors (season, lifestyle), indicating challenges for endpoint selection, patient selection, and study duration (Howard et al., 2023).
In critically ill and pediatric populations, accelerometry correlates with direct observation for activity timing but cannot reliably distinguish intensity or voluntary versus involuntary movements, limiting its utility as a sole measure of functional activity in these settings (Verceles & Hager, 2015; eczema trial findings).
Device tolerability and data completeness can be issues (e.g., incomplete datasets, user usability in children); convergent validity with disease-specific severity measures has been low in some trials (e.g., eczema), and responsiveness to change was poor, indicating limited usefulness as an outcome without further methodological refinement (eczema trial; trials cited).
Overall implication: actigraphy and accelerometry can provide objective, ecologically valid measurements of movement and sleep-related parameters and are useful adjuncts in specific clinical and research contexts (circadian disorders, pre-MSLT monitoring, activity monitoring). However, performance is device- and algorithm-dependent, limited for certain measurements (sleep staging, arousal detection, PLMD diagnosis, non-convulsive seizure detection), and current evidence does not uniformly demonstrate improved clinical outcomes. Careful selection of device, algorithm, population, and endpoints — and in many cases concurrent gold-standard testing (PSG/EMG/EEG) — is necessary.