1. Introduction
Parkinson’s disease (PD) has become the second most common neurodegenerative condition following Alzheimer’s disease. Worldwide evidence highlights the rising prevalence of the disease, especially after the sixth decade [
1], with an almost 10-fold increase in disease incidence from the sixth to the ninth decades of life [
2]. Accounting for the predicted steep increase in PD cases until 2030 [
3], the stress of upsizing on healthcare systems and the augmented burden on healthcare providers around the globe may result in system overburdens and poor patient care.
The pathogenetic cascade of PD finally results in a decrease in neurotransmitter dopamine in the caudate nucleus and putamen. This mainly involves the accumulation of Lewy bodies containing aggregates of α-synuclein with a prion-like propagation motif, leading to the loss of dopaminergic neurons in the substantia nigra pars compacta and the striatal dopaminergic denervation [
4]. A clinical diagnosis is mainly based on three predominant motor symptoms as delineated by the Movement Disorder Society: bradykinesia, rigidity, and resting tremor [
5]. Other motor symptoms that commonly manifest as the disease progresses are hypophonia, a decline in facial expressions, and gait impairment with freezing and falling, as well as the wearing off of responses to treatments, leading to fluctuations of the symptoms along with dyskinesia [
6]. Although motor symptoms are the principal identifiers of PD, a variety of non-motor symptoms are also implicated in PD development and progression. However, these symptoms are widely under-reported by patients or are overlooked during clinical assessments.
Currently, there is no proven neuroprotective or disease-modifying therapy that can stop or delay the progress of the degeneration in PD. However, the most effective substance in terms of improving motor symptom complications with the fewest short-term adverse effects is levodopa. The systematic administration of levodopa has remained the greatest weapon in a physician’s arsenal for more than 50 years [
7]. Along with other dopaminergic pharmacological targets comprising the dopaminergic therapy regime, dopaminergic medications aim to restore dopamine homeostasis at the synaptic level by transiently acting on elements implicating dopamine metabolism and neuronal excitability [
8]. Over time, patients’ therapeutic window is pruned, losing their long-duration response to dopaminergic medication due to the fact of disease-related pathophysiological changes in the brain. The time without good control of symptoms (OFF) and therapy complications, such as dyskinesias, become more frequent as the disease progresses, and the effort to keep patients well controlled turns out to be really challenging [
9]. Thus, questions such as “What is going to happen next?” are a common riddle that physicians are called upon to solve.
Recent care strategies include brief appointments of random frequency at outpatient clinics where, in the best case scenario, neurologists evaluate disease progression using established rating scales and/or patient-reported questionnaires, illustrating patients’ state at home. The Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) is a rater-based scale globally applied in routine clinical practice and commonly accepted as a reference standard for assessing patients’ motor and non-motor symptoms and complications during a physical examination visit. Although MDS-UPDRS, as a benchmark for PD assessment, possesses comprehensive clinimetric properties, the low rater consistency and modest within-subject reliability due to the fact of psychometric issues, especially for patients’ longitudinal follow-up, create doubt in symptom evaluation [
10,
11]. Furthermore, the proper application and interpretation of such a scale relies on physician expertise and intuition, and these are scientific skills not widely mastered. In addition, this brief examination not only provides a snapshot of patients’ actual condition but is also subject to the Hawthorne effect, meaning that the patient performs their best during an evaluation [
12].
Diaries, a useful tool for collecting self-reported feedback on the disease situation and daily activities longitudinally, are extensively applied in PD as a qualitative input to support physicians’ decision making. As a medical source of information, however, they are subject to many types of bias, creating the potential for fallacy behind physicians’ line of thinking. Mental recession is a commonly encountered situation in PD patients leading to recall bias [
13]. Moreover, individuals are not duly qualified or self-aware to identify—all the more so discriminate—disease symptomatology [
14]. These shortcomings in the current management of PD in combination with the relatively high number of patients (approximately 40%) that do not have access to consultation with a PD specialist or neurologist result in a higher risk of disease-related complications and mortality [
15].
To address this issue, sensor-based systems have been developed over the last decade to monitor patients in their own environment and evaluate disease symptomatology [
16,
17,
18,
19]. The results of the monitoring are analyzed remotely, while a summary is provided to the treating physician and/or patient. The specialist can use this summary to assess motor and non-motor symptoms and how they are affected by the use and timing of medication. The data should be used to determine whether any changes to the treatment regimen are desirable, in consultation with the patient. These outcomes are intended to complement existing assessment methods and are not intended as substitutes. In this way, the physician has a more comprehensive picture of the patient, as well as a clear argumentation of the outcome of the treatment and the need for any modification. Of interest are the practical recommendations issued by a movement disorder specialist panel, pinpointing the value of an objective continuous monitoring strategy acting ambulatory to routine clinical practice to promote evidence-based decision making [
20]. However, while many monitoring systems have been tested for their usability and accuracy in motor symptom detection [
21,
22,
23,
24,
25], none of them can actually record the wide range of PD symptomatology.
Thus, in the present study we implemented a device that can accurately record almost all PD motor symptoms [
26]. The question that remains is how informative the in-person clinical evaluation is compared to patients’ state at home during activities of daily living. To answer this question, we performed an observational clinical study comparing the results of the in-person clinical evaluation to the recording of symptoms in a home environment with the AI enabled wearable system PDMonitor
®.
4. Discussion
The routine clinical practice currently followed in Greece is streamlined with most countries worldwide, comprised of infrequent, brief appointments at outpatient clinics. Oftentimes, non-PD expert physicians try to apprehend motor symptoms along with complications and intuitively track changes outside of the clinic with some ecological validity. However, the tools used in clinical practice to monitor temporal disease patterns in symptom fluctuations have various limitations. Symptom diaries, validated clinical scales, such as the MDS-UPDRS, and patient-reported questionnaires, such as QoL questionnaires, are often retrospective, being subject to recall bias and lack objectivity, or suffer from inter- and intra-rater variability [
10,
13]. Symptoms are very often overlooked by raters or under-reported by patients. Therefore, objective measurements providing an accurate evaluation of symptoms can pave the way for evidence-based clinical decision making [
32].
Up-to-date AI-driven sensor-based technologies still lack concrete evidence of their clinical benefit. However, favorable for their adoption into routine practice are the results of the diagnostic accuracy for most of the devices on the market [
21,
22,
23,
25]. To the best of our knowledge, only a few studies using remote monitoring devices reported association outcomes, correlating home monitoring with visit assessment using expert-response clinical scales. Bradykinesia, dyskinesia, and tremor scores in a home environment were generally moderately correlated with MDS-UPDRS scores for in-office clinical evaluations [
33,
34,
35].
In the present study, we aimed to evaluate how comprehensive the information provided by routine visits to outpatient clinics for motor symptoms is in relation to the actual situation of patients at home, as reflected by a wearable remote monitoring medical device. The comparative analysis of the results focused on two axes. The first investigated the correlation of the motor symptoms detected in the clinical evaluation with the MDS-UPDRS to the objective measurements of the PDMonitor® system. In the second, correlations between the system’s recordings and the patients’ quality of life were explored.
All motor symptoms of the patients with PD participating in the study were evaluated by means of in-person clinical examination with the MDS-UPDRS, as well as in-house monitoring with the PDMonitor
®. Comparing the results, we appreciate that there were various degrees of agreement between the two methods for all symptoms, ranging from moderate to strong correlations (
Table 6). This shows that the clinical examination of the patients with PD, even when performed with validated scales, can only partially reveal the situation of the patients in their home environment during their daily activities.
For bradykinesia, rigidity, gait impairment, and instability, there was a moderate correlation between the clinical evaluation and the recordings in the home environment. The differences observed could possibly be attributed to the effort that patients usually exert during their visit at the physician’s office. These specific symptoms can be partly masked during the patient’s “best performance” in front of the treating physician, while they are quite differently expressed during “usual performance” in activities of daily living. On the other hand, for freezing of gait, tremor, and dyskinesia, a high correlation was revealed, except rest tremor in the lower limbs. The latter can be explained by the fact that the symptom does not recur as often in the lower as in the upper extremities. Not all patients with tremor in the lower limbs experienced the symptom and, in particular, the limited duration of the in-person evaluation reduced the likelihood of detecting the symptom; however, the information was provided to the physician by the home recordings. These results should not come as a surprise, since even the best performance cannot overcome the sudden inability to step forward or involuntary movements.
Intriguingly, although the device has not yet demonstrated any surrogate marker of rigidity, a moderate correlation was observed for bradykinesia, gait, and %OFF as measured by the device. In addition, the PDM-TREMOR was strongly associated with the average rigidity score. This does not necessarily imply the establishment of a composite measure of rigidity, as this cardinal symptom occurs and aggravates at a similar rate to bradykinesia and gait and balance disturbances but not at the same rate as tremor [
36]. Nevertheless, the strong correlation with the PDMonitor
® tremor offers a chance to study this association in the future, especially considering that the majority of patients participating in the study were tremor-dominant (14/20).
Thus, the strong correlation that was observed in the identification of freezing of gait, resting tremor, and dyskinesia means that when these symptoms are detected in the clinical evaluation, they are also present in the home environment, as expected. On the other hand, for gait impairment, instability, and bradykinesia, the moderate correlation between the results of the clinical examination and the home recordings is indicative of the symptoms’ wide fluctuations throughout each day, leading to the overestimation or underestimation of a patient’s overall situation by the clinical assessment. It is interesting, however, that a symptom that cannot be captured by inertial sensors (i.e., rigidity) is also moderately correlated to the bradykinesia of the home recordings. This is another finding highlighting the value of monitoring devices for diverse motor symptom detection.
Furthermore, the clinical assessment and the results of the home recordings appear to have a moderate correlation in identifying the percentage of time in which the patients were OFF. This finding, although expected, adds great value to the home monitoring, as it is the only reliable data that the treating physician can use to adjust the treatment more effectively. By exploiting not only the temporal spread of the OFF intervals within the day but also the severity of the OFF as provided by PDMonitor
®, physicians can accurately capture the actual state of a patient throughout a day. In this way, the physician gains a comprehensive picture of a symptom’s daily fluctuations, pinpointing the value of continuous monitoring in the management of patients with Parkinson’s disease. A summary of the correlation analysis between the PDMonitor
® and the MDS-UPDRS scores is presented in
Table 6.
The question that arises from all of the above findings is whether there is any measure that mirrors the actual state of the patient at home and, furthermore, their quality of life. Part II of the MDS-UPDRS shows the impact of the disease on the patient’s daily activities, and it is considered a reliable measure of the patient’s quality of life. Similarly, the length of time the patient is in the OFF state is an indirect indicator of quality of life, as in this state, the patient’s overall activity and mobility are limited. The remarkable correlation of the percentage of time patients are in the OFF state, as measured by the PDMonitor
®, with the PDQ8 scale reflects the value that artificial-intelligence-enabled recording systems have in the overall evaluation of a patient’s actual state at home. This value is further highlighted by the strong correlation between MDS-UPDRS part II and PDQ8, a result in line with previous studies [
37], showing that even a detailed history taken from a patient oftentimes does not reveal their quality of life in the way that the monitoring of symptoms in their home environment can.
As of now, few studies have been undertaken comparing similar devices worn in an uncontrolled environment with accredited scales and questionnaires, e.g., UPDRS and patient diary [
33,
34,
35,
38]. The correlation between the dyskinesia score derived from a wrist-worn monitoring device (Parkinson’s Kinetrigraph
TM) and the UPDRS part IV dyskinesias was proven to be low (r = 0.38,
p = 0.09), while no significant correlation was found for the device bradykinesia score with the UPDRS part IV fluctuations (r = 0.25,
p = 0.28). However, a moderate association was found between patients’ perception of disability due to the presence of motor fluctuations and the device’s fluctuation score (r = 0.52,
p = 0.018) [
35]. Recently, L. Chen et al. [
34] demonstrated a moderate association between the device’s bradykinesia score and the UPDRS items indexing arms bradykinesia (items 23–25), with the coefficients (r) ranging from 0.456 (for finger taps) to 0.557 (for hand movement). Moreover, the device metrics were moderately correlated to the average score of all bradykinesia items (0.588). The percent of time with tremor, as measured by the device, showed a slender correlation with patient-reported tremor (0.269) and a moderate correlation with the expert-rated items of UPDRS part III (0.434). Of interest, the rigidity score was moderately correlated to the bradykinesia score (0.479). Notwithstanding the moderate agreement of the above measurement strategies, no correlation was present between dyskinesia, as measured by the device, and the OFF time, as reported by patients.
Furthermore, a moderate concordance was found between a waist-worn device (STAT-ON
TM) and patient diary (Hauser diary) on the percentage of daily time in the OFF state (r = 0.57), for the percentage of daily time in the ON state (r = 0.48), and for daily time with dyskinesias (r = 0.65) [
33]. The gait and balance characteristics measured using another wearable device (OPAL), composed of three wearable sensors worn at the lumbar and on both lower extremities, demonstrated a moderate association with total UPDRS part III, postural instability and gait difficulty subscores of UPDRS part III, and rigidity subscore with coefficients of 0.48, 0.61, and 0.49, respectively [
38]. This benchmarking is not, of course, a compendium, as the context in which each technology is used, the methods of comparison, and the scales employed vary from one study to another. However, each study shares the same goal, namely, to demonstrate the clinical validity and degree of association between the remote objective monitoring of a patient’s symptoms in a free-living setting and the golden standard of in-person clinical examination using validated clinical scales.
The present study confirms the results of previous studies regarding the degree of correlation in identifying the main symptoms of the disease (rigidity and bradykinesia) and disease complications (dyskinesia and ON/OFF fluctuations) with improved rates, however. Notably, for resting tremor we found a strong correlation both in terms of severity and constancy, not reported in other studies to such a high degree. Moreover, the moderate and high correlation in the detection of OFF and dyskinesia, respectively, are variables of particular clinical relevance in terms of their impact on medication adjustment, underlining at the same time the reliability of clinical assessment and the added value that remote monitoring can have in this context. Interestingly, a strong correlation was found between the OFF rate detected by the device and the PDQ8 on the patients’ quality of life. The existence of an index capable of remotely measuring patients’ quality of life based on the objective recording of their motor status and symptoms is an innovation that may contribute to a more effective management of the disease.