A Long-Term, Real-Life Parkinson Monitoring Database Combining Unscripted Objective and Subjective Recordings

Round 2

Reviewer 1 Report

I have no further concerns regarding the manuscript. 

Reviewer 2 Report

Dear Authors,

thank you for having considered the comments from my previous report. Although the amount of data you stored in your repository, I still believe the sample size is small and very heterogeneous to be used alone but it is also true that a researcher could collect this and other repositories to perform multi-center study.

Please 

I agree this data descriptor is worth publishing.

Best regards.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The authors presented a Data Descriptor manuscript to describe a database of raw data collected via accelerometers and gyroscopes placed on chest and both wrists, complemented with subjective data in people with Parkinson's Disease and in ecological settings. The paper needs extensive proof reading and formatting is not accurate. I also suggest using tables to report content of questionnaires rather than a so extensive bullet list (par 2.2.2).

We transformed the bullet point style presentation of the questionnaire content into a table (par 2.2.2) as suggested by reviewer 1.

Beside, I have one big concern on this paper. Data Descriptors published in Data, should fit with this description: "When a data descriptor links to a dataset published in a research article, it will have to provide additional content to merit a new paper". To be honest, it is not clear to me what is the difference between this Data Descriptor and the Authors' papers published at (and referenced here):
• https://www.nature.com/articles/s41531-019-0093-5
• https://mhealth.jmir.org/2020/5/e15628/
Figure 4 from the same first reference looks also the same as Figure 2A submitted here. Can the Authors strenghten what is the novelty of this new paper and why it is worth publishing?

We elaborated at the end of the Summary about how this data can merit new research, and elaborated on latest research and the BEAT-PD data challenge. We agree with reviewer 1 that this was too briefly explained before. However, this data set has more than sufficient additional value for new hypotheses. By merely making the database publicly available, we already got several requests to use the data. Therefore, our Data Descriptor will be of value for the scientific community and will attract attention of interested researchers (line 96-110). We also underlined the difference between the previous feasibility study, without duplicating the information which is shared in the Methods section. Figure 2A of the current manuscript is based on the same analysis as figure 4 of Heijmans et al (npj PD 2019) but is included here to a different point. We included the correct reference and elaborate how the figure is used here in a different analysis. We support significance and validity of this method and combined data by showing the same analysis is not successful using chest-sensor data (figure 2B). As hypothesized since the extracted features are designed for wrist-accelerometry rather than for chest-accelerometry, the AUC score is superior for wrist-data (2A). This paper is essentially different from earlier publications regarding this data set. In Heijmans et al (2019 npj PD), we demonstrate and discuss in detail the practical feasibility of this novel combination of methods. This should be regarded as a prerequisite to share this data. In Habets et al (JMIR 2020), we discuss the development of the novel PD-specific ESM questionnaire. We elaborate on how we came to the selected items, and how these items internally performed in this data set. Also, this can be regarded as a prerequisite to offer these data to other researchers to facilitate full use of its potential. This Data Descriptor contains the requested info by the Author Guidelines, detailed explanation of the collection process, potential new hypotheses, and how to process and combine these two data types for data analyses.

Maybe in the introduction, where they refer to their papers (ref n.10 and 37) without explaining what they have done previously and what is the novelty. Another big concern I have is regarding ethics. The Authors did not even change the Institutional Review Board Statement. Indeed, instruction from MDPI to complete that section are still there.

We elaborated on ref 10 and 37. We overlooked the ethical statement on line 372 (before 316) and now changed it.

Regarding the contents per se, I do believe the introduction will benefit from refering to more recent papers of the field. Here some examples (no need for the authors to consider and refer exactly to these, but take them as examples):
• https://www.mdpi.com/1424-8220/20/20/5920
• https://jneuroengrehab.biomedcentral.com/articles/10.1186/s12984-020-00774-3
• https://www.mdpi.com/1424-8220/21/1/128

We added recent literature to the summary/introduction, including their references. We replaced and corrected reference 29.

Enrolled patients cover a wide range of ages and disease severity. Is 20 patients enough to define a so heterogenues sample? Authors also mentioned that neurosurgeon have recruited participants: are the Authors considering also patients with a DBS implant? If yes, do they expect this to be a confounding factor for the database and subsequent analyses?
Reviewer 1 is correct that 6 out of 20 patients were treated with DBS. We did not make these individual data publicly available but will do on reasonable request. This is due to local ethical privacy regulations. We added this information to par 3.1.

The link to the GitHub page does not allow downloading the codes to check the what is actually stored in the *edf files. Based on these considerations, it is difficult for me to evaluate the paper as suitable for publication.
Regarding reviewer 1’s comment on our code available on our github page, we want to emphasize that the whole code is made available to replicate our results. This code can be used as a building block to explore all EMA and sensor data parallel.

Reviewer 2 Report

Very nicely design of experiment for free-living condition data collection. I think the manuscript should be accepted in its current form.

Reviewer 3 Report

In this manuscript, a dataset of real-life recording of inertial signals and subjective experiences of motor and non-motor PD symptoms is described. 20 subjects with PD participated in the study, and the data recording was continuous at home for 2 weeks using three sensors mounted on both wrists and chest. This dataset is significant to validate PD monitoring systems on unscripted daily-living activities. The data collection methodology is described with sufficient details. It was used in previous publications [10, 37], but in this manuscript, more details are provided included the link to the dataset and validation code. I have the following comments:
• The copyright license of the dataset is not described in the manuscript.

The copyright statement is added to par 2.

• For section 3.1, a detailed table showing the subjects’ demographics should be provided. The history of motor fluctuation for the subjects should be reported if available.

We added a demographic table to section 3.1. we reported the individual patient which reported to suffer from motor fluctuations. For individual disease specifications were not shared publicly due to privacy regulations. As stated in the article, we will share upon reasonable request.

• In line 36 and 57, you mentioned limitations of current monitor tool. What are the major limitations?

We elaborated on limitations of current monitor tools in the Summary, lines 36, 59 (previous line 36, 57).

• In line 83, you mentioned some public PD datasets such as BEAT-PD dataset (https://www.synapse.org /#!Synapse:syn20825169/wiki/600903). A brief review of them should be included in the introduction that show how this dataset is deferent or what it is adding.

We elaborated on the BEAT-PD data challenge.

• In line 136, one item of EMA is 'I am with..' and 'I am doing ..', but in EMA_data.cvs, there are six columns ('soc_who', 'soc_who02', ...). Would you explain why it is different, what are the numbers represent, and why some of the cells are empty?

Minor revisions/ other comments addressing review:
Regarding reviewer 3’s question on multiple columns for soc_who.. : patient were able to select multiple answers for with whom they were and what they were doing. We added this to the table.

• In line 151 regarding item 'I took Parkinson medications ...', is it 'sanpar_medic' in EMA_data.cvs? and what do the numbers represents (1 means yes ...)?

Regarding reviewer 3’s question on the sanpar_medic item: that is correct. Answer coding: [1=OFF; 2 = changing from OFF towards ON; 3 = ON; 4 = changing from ON to OFF]. This information is available in the ‘EMA_data_coding.xlsx’ file, which is available on the dataverse.

• In Section 3.4.2, it is not mentioned the orientation of the sensors' axes in respect to the body. Adding a figure would be helpful. Did all the subjects have the sensors placed in the same orientation?

We added the axial orientation of the sensors in 3.4.2.

• Were the recorded signals been checked for noises or been filtered?

Sensor signals were band pass filtered before feature extraction. The specific band widths differed between features. Details per feature are available in the code on our github.

Some minor comments:
• Figure 2-A was reported before in reference [10]. Proper citation is required.

We cited properly, and elaborated on the repeated use of the analysis.

• In line 316, the institutional review board statement needs to be corrected.

Statement is corrected.

• Correct reference [29]

Adjusted.