Table 2

Data pipeline stages and features for cluster analysis

Sensor
type of data		Data pipeline step
CleaningLabelling (of)Featurisation of variables for cluster analysis
DescriptiveDescribing adherence*
ACT Sensor
Pressure–Time

  • Removal of blank, duplicate and non-physiological values.

  • Any non-linear baseline drift corrected, using a sparsity based de-noising approach.19

  • Treatment sessions

  • Pressure peaks

  • Breaths

  • Breaks between breaths

  • Sets of breaths

  • If any breaths recorded on a day Y/N

  • Breath count†

  • Breath length†

  • Breath peak pressure†

  • Treatment duration†

  • Number of treatments per day

  • Number of sets in a treatment†

  • Number of breaths per set†

  • Adherence score (proportion of days with any breaths recorded per total number of days)

  • Breath count adherence (proportion of completed breaths against prescribed breaths per treatment)

  • Set adherence (proportion of sets against prescribed sets per treatment)

  • Treatment session adherence (proportion of completed treatments against prescribed treatments)

  • Pressure adherence (proportion of ideal expiratory pressure breaths per treatment)

  • Breath length adherence (proportion of expiratory breaths at ideal length per treatment)

Fitbit
Heart rate and step count

  • Removal of erroneous or non-physiological data (caused by improper wearing, depleted battery, full memory capacity due to infrequent syncing).

  • Heart rate sampling frequency made consistent (per min) using a rolling average

  • Gaps in data

  • Wear time (from heart rate data)

  • Awake wear time

  • Time in MVPA using personalised heart rate cut-off value17

  • Points crossing MVPA cut-off value

  • Heart rate

    • Resting20

    • Peak

    • Density and variability

    • MVPA threshold switches

  • Step count

    • Daily step count

    • Density and variability

    • Active minutes (greater than a threshold value)

  • Combined

    • Active minutes both heart rate and step count.

    • Step count during periods of MVPA

  • Daily time in MVPA compared with 60 daily minutes recommended.

  • Weekly time in MVPA, 7 day window compared with 420 weekly minutes recommended.

  • Not all features shown.

  • *Features quantified as adherence per prescribed treatment session, per day, per rolling 7 day week, or other time point as required for analysis. ACT prescription is taken from clinical records and physiotherapy questionnaires.

  • †Total and/or average per treatment, SD, min, max values.

  • ACT, airway clearance technique; MPVA, moderate to vigorous physical activity.