Devices for remote monitoring of Parkinson’s disease

The EAG said that there is no clearly established reference standard for measuring Parkinson's disease symptoms, beyond clinician and patient assessment, that could be used to establish test accuracy. A potential benefit of the technologies is that they may more accurately evaluate symptoms than patient recall or clinical opinion, so a reference standard based on patient recall or clinical opinion (as used in accuracy studies identified by the EAG in its systematic review) could underestimate technology performance. The committee noted that accuracy estimates may not be the best outcome for assessing the performance of these technologies.

Only 3 studies had data comparing clinical outcomes with the technologies against not using the technologies. Two of these (which assessed the Kinesia 360) were small randomised controlled trials. The largest study (which assessed the PKG and was by Woodrow et al.) reported clinical improvements in terms of statistically significant reductions in the Unified Parkinson's Disease Rating Scale (UPDRS) 3 (motor examination), UPDRS 4 (complications of therapy), total UPDRS score and PDQ‑39 (Parkinson's Disease Questionnaire 39). The committee noted that the trial was not randomised. It understood that there could have been a systematic difference between the centres included in the trial and their catchment areas because people were allocated to PKG based on the centre they attended. The clinical experts also pointed out that the PKG was used every 5 weeks in the Woodrow et al. study, which would not be realistic in the NHS. Standard care in the study's comparator arms also may not represent NHS care. Because the length of follow up from the available studies was relatively short, there was also a lack of data on how long any benefit of the devices lasted once they were not used any more. Also, the EAG did not identify any data specifically on the populations who were identified in the scope as likely to have particular benefit from the technologies. These included people with communication barriers and people from black, Asian and minority ethnic family backgrounds, because symptoms can vary by family background. The committee concluded that, while the identified studies gave some indication of how the technologies could benefit people with Parkinson's disease, there was considerable uncertainty about the likely size of this benefit if the technologies were adopted in the NHS.

The clinical and patient experts suggested that the technologies may be particularly useful for people who are eligible for more advanced therapies such as deep brain stimulation. But the EAG said the only evidence on the devices' comparative effectiveness was in the maintenance stage of Parkinson's disease. The committee understood that the devices may perform differently in these different populations.

The clinical experts said that, as well as changes to medication, symptoms can be managed with other changes to care like physiotherapy and exercise. Only the PKG had data on how the technologies can change decisions about care. The proportion of people who had a change in clinical management as a result of the PKG varied considerably (between 31.8% and 79%). A UK study reported that the PKG provided additional information in 45.5% of cases. In studies of clinician opinion, between 4% and 41% agreed that the PKG provided enough additional information to consider making treatment adjustments. The committee concluded that there was uncertainty about how much using remote monitoring devices would change care for people in the NHS.

In the EAG's model, the intervention arm resulted in a higher cost than standard care. Device cost had the biggest impact on this. The EAG said that the device cost depended on how the devices were modelled as being used (see section 3.3) and the cost per use. The technologies have differing payment mechanisms: pay per use, a subscription model or outright purchase of the device. The committee noted that the payment options differed in terms of upfront investment and how reversible a decision to use a technology would be. The committee recalled that the technologies could be used in various ways in the NHS and noted that the EAG had modelled use at varying frequencies. This included one-time use (at baseline) and routine use (every 6 months) in its base cases, and a recurrent use scenario analysis (at 6 and 18 months in place of clinic appointments). How frequently the technologies were used substantially affected device-related costs and the cost-effectiveness estimates.

The committee recalled that there was uncertainty about the size of impact that device use would have on symptoms if used in the NHS (see section 3.6). The EAG used the Woodrow et al. trial to inform the estimates of health-related quality of life in its model. It also had to make assumptions about how long any benefit would last after using the devices. Sensitivity analyses showed that the cost-effectiveness estimates were very sensitive to this assumption. The committee concluded that the size and longevity of the benefits of device-guided decisions about care was very uncertain, and consequently so were the cost-effectiveness estimates produced by the EAG's model.

Cost-effectiveness modelling reported by Chaudhuri et al. (2022) estimated that the PKG would be cost saving by £17,362, whereas in the EAG's model the PKG incurred costs. Both models compared the PKG with standard care. The committee questioned the approach used in Chaudhuri et al., which took scores with PKG use on the UPDRS from Woodrow et al. and converted them to the Hoehn and Yahr scale. They then used this to calculate resource costs. The committee noted that the EAG had been unable to verify or validate the approach used by Chaudhuri et al. It was also not clear if the symptom improvements observed from using the technologies could realistically translate into the large-scale changes in healthcare use predicted by this model. A clinical expert said that it was very unlikely that remote monitoring devices would have the effect shown on the Hoehn and Yahr scale. The EAG added that in Woodrow et al., device use did not have a statistically significant effect on the Hoehn and Yahr scores (-0.044, standard error 0.097 for adjusted data). It also said that in the Chaudhuri et al. model, the PKG's effect on symptoms was assumed to last for 5 years with only limited waning of effect after that, although there was no evidence for this. Clinical experts noted that there are no disease-modifying treatments available for Parkinson's disease that can stop progression. The committee concluded that the Chaudhuri et al. model was likely to have overestimated how much the PKG can reduce healthcare-associated resource costs, and how long the benefit from PKG-guided care on symptoms would last for.

A UK survey reported that people with Parkinson's disease interact with 18 different healthcare professions. The EAG had suggested that more data on this could identify further areas in which remote monitoring device-aided care could reduce costs. The patient experts also explained that the cost of providing care to people with Parkinson's disease can differ significantly according to whether they have a live-in carer, a paid carer or no carer at all. Costs related to social care were not included in the EAG's model because of a lack of data. Any impact of remote monitoring devices on these costs, for example delaying when someone with Parkinson's disease goes into a care home, could have been missed. The companies said that the EAG's model did not include costs related to falls and hip fractures prevented, which they said could be an uncaptured benefit if the devices did improve symptom management. The EAG explained that there was a lack of data to inform this. Also, because of a lack of data, its analysis was constrained to the management phase of Parkinson's disease. It understood that falls, hip fractures and social care costs are largely confined to the advanced stages of the disease. The committee concluded that the EAG's model may have underestimated the impact of using remote monitoring devices on some resources and associated costs. However, other costs related to implementing the devices, for example interconnectivity, may be higher in practice.

The committee recalled that remote monitoring devices could benefit carers (see section 3.2). But, because of a lack of data, the EAG's model did not include costs or health-related quality of life for them. The committee noted that relatively small improvements in quality-adjusted life years (QALYs) resulting from device use (perhaps related to carer benefits) could have a large impact on the cost-effectiveness results, depending on the analysis used. The committee concluded that there was considerable uncertainty about how much the remote monitoring devices could affect carers, and noted that this had not been captured in the EAG's model.

There was not enough evidence for the EAG to be able to do subgroup analyses for people who have communication barriers, people from black, Asian and minority ethnic family backgrounds, or people from different socio-economic backgrounds. These populations were identified at scoping as being potentially likely to gain additional benefits from the technologies. The patient experts noted that remote monitoring may also benefit people who are having difficulty attending consultations, or getting care because services are at full capacity. The EAG had also not been able to model how remote monitoring devices might work for people being considered for advanced therapies such as deep brain stimulation, because there was not enough evidence.

The committee recognised the promise that remote monitoring devices offer to people with Parkinson's disease and their carers (see section 3.1 and section 3.2). These devices could also help with increasing capacity pressures on the NHS. But their cost effectiveness is very uncertain, and to estimate it more accurately, there are several areas of uncertainty for which more data is needed (see section 4). The PKG has the most evidence, but the committee recalled that there was uncertainty about how well data from the main trial for this technology (Woodrow et al.) represents how well the device would work in the NHS (see section 3.6). The committee noted that the technologies differed in how sensors are worn (see sections 2.6 to 2.21) and the algorithms they use. Because of this, the EAG stated that clinical benefits observed for 1 technology could not be assumed for the other technologies. No identified studies compared the performance of 1 technology against another. A lot of identified data was on test accuracy. The committee recalled its concern about the suitability of accuracy estimates for assessing the performance of these technologies (see section 3.5). It also recalled the uncertainty in how much remote monitoring devices would change decisions about care in the NHS (see section 3.8). The committee concluded that, to assess performance, further data is needed for all technologies on the size of impact of using all of the technologies on symptoms or health-related quality of life (see section 4.2).

The committee considered the risks associated with using the technologies in the NHS while further data is collected. The clinical and patient experts said that their main concern about potentially worse patient outcomes was if use of the devices was poorly implemented. For example, if they were used to entirely replace face-to-face appointments, and were not integrated into care pathways. They also said that remote monitoring devices were already being used in some NHS centres. The committee noted that the overall cost impact of using the devices was uncertain and was largely dictated by how much the companies charged, and the different payment mechanisms (see section 3.9). The committee concluded that this was important for commissioners to consider, because the devices were likely to differ in terms of ongoing or irrecoverable costs if a later decision was made to stop using the technologies, for example if further data collection showed they did not work as well as anticipated.

There was no or limited data for several populations who could particularly benefit from the remote monitoring technologies (see section 3.14), including people who might be helped by advanced therapies.