R/bp_sleep_metrics.R

In bp: Blood Pressure Analysis in R

# small helper function to return tibble NA with the appropriate colnames
null_sleep_output <- function(){
  out = tibble::tibble(data.frame(matrix(nrow = 1, ncol = 16)))
  colnames(out) = c("sleep_DBP", "wake_DBP", "sleep_DBP_sd", "wake_DBP_sd",
                    "presleep_DBP", "prewake_DBP", "postwake_DBP", "lowest_DBP",
                    "sleep_SBP" , "wake_SBP", "sleep_SBP_sd", "wake_SBP_sd",
                    "presleep_SBP", "prewake_SBP", "postwake_SBP", "lowest_SBP"  )
  return(out)
}

# Calculate summaries for SBP and DBP to use with sleep metric calculations
# First does check on unusual sleep periods, or periods with unusually short naps
# For each subject, visit, group, want to do the following
# Give a time indicator that is more descriptive than sleep/wake by obeying the following
# pre-wake - 2 last hours in the sleep period before wake
# after-wake - 2 hours after waking up
# pre-sleep - 2 hours before falling asleep
# lowest SBP - average of 3 readings centered around lowest


# First, need to identify the sleep periods start and end. This is in general
# not easy because going to sleep time and BP measurement time are not going to
# be the same. If we use ToD as differentiation, and say measurements are 00:20,
# 00:50, 1:20, 1:50, 2:20, then technically the first 5 will be used rather than
# the first 4. Same is going to happen with sleep end.

# Approximation - the first time we observe 1 is going to be treated as sleep start,
# the last time we observe 1 is going to be treated as sleep end? or is it 0?
# something to discuss.

# In any case, for each subject/group/visit, we can create Diff column that
# takes consecutive measurements differences of Wake column, so that
# 1 - 0 = 1 sleep end
# 0 - 1 = -1 sleep start
# 0 - 0 = 1 - 1 = 0 no change

# Then for each 1, one can use lubridate to allocate pre-wake, post-wake
# For each -1, one can use lubridate to allocate pre-sleep
get_summaries_SBP_DBP <- function(data){

  # Initialize variables for dplyr
  SBP = DBP = WAKE = period = NULL
  rm(list = c("SBP", "DBP", "WAKE", "period"))

  # Assume the times are in chronological order
  # Figure out total number of measurements
  nread = nrow(data)

  # Create difference column (here I assume the time is already sorted,
  # should probably check) - time has been ordered in bp_sleep_metrics
  wake_binary = as.numeric(as.character(data$WAKE))
  # 1 - 0 = 1 sleep end
  # 0 - 1 = -1 sleep start
  # 0 - 0 = 1 - 1 = 0 no change
  data$diff = c(wake_binary[2:nread] - wake_binary[1:(nread - 1)], 0)

  # if no sleep measurements (all wake), return null and message
  if (!any(data$WAKE == 0)){
    message(paste("No sleep measurements found for subject", unique(data$ID),
                  ". Double check the WAKE column."))
    return(null_sleep_output())
  }
  # if no wake measurements (all sleep), return null and message
  if (!any(data$WAKE == 1)){
    message(paste("No wake measurements found for subject", unique(data$ID),
                  ". Double check the WAKE column."))
    return(null_sleep_output())
  }

  # if starts in a sleep period, then set first row as sleep start
  if (data$WAKE[1] == 0){
    message(paste0("Can not detect the beginning of sleep period for subject ", unique(data$ID),
                   ". First measurement automatically set as sleep start"))

    data <- data %>%
      # before first row, add row with diff = -1
      # row with -1 labeled as wake_end, actual first row of data will be sleep_start
      tibble::add_row(DATE_TIME = as.POSIXct(NA), diff = -1, .before = 1)
  }

  # if ends in a sleep period, then set last row as sleep end
  if (data$WAKE[nread] == 0){
    message(paste0("Can not detect the ending of sleep period for subject ", unique(data$ID),
                   ". Last measurement automatically set as sleep end"))

    data$diff[nread] = 1
  }

  # after checking/changing first and last measurements, calculate sleep starts and ends

  # Number of sleep period starts
  n_starts = sum(data$diff == -1)
  # Number of sleep period ends
  n_ends = sum(data$diff == 1)

  # in theory I think all sleep periods should now have a start and end
  # i.e. sleep starts equals sleep ends because incomplete sleeps have been
  # auto set to start or end at the first or last

  # so now need to check if there is more than one sleep period
  if (n_starts > 1){

    # Extract the times of start and end, should have the same length now
    wake_ends = data$DATE_TIME[which(data$diff == -1)] # This is the last time WAKE = 1
    sleep_starts = data$DATE_TIME[which(data$diff == -1) + 1] # This is the first time WAKE = 0
    sleep_ends = data$DATE_TIME[which(data$diff == 1)] # This is the last time WAKE = 0
    wake_starts = data$DATE_TIME[which(data$diff == 1) + 1] # This is the first time WAKE = 1

    # Calculate the durations
    durations = difftime(sleep_ends, sleep_starts, units = "hours")

    # If any duration is less than two hours, treat as nap and remove
    no_naps = which(as.numeric(durations) > 2)

    # if no periods are longer than 2 hours, return null and message
    if (length(no_naps) == 0){
      message(paste("Identified sleep period for subject", unique(data$ID), "is less than 2 hours. Double check the WAKE column."))
      return(null_sleep_output())
    }
    # if multiple periods and at least one longer than 2 hours, use first valid and message
    # have removed naps messaging because regardless, the first valid sleep period is always chosen
    if (length(no_naps) >= 1){
      message(paste("Current functionality only supports one sleep period per subject. Only the first sleep period longer than 2 hours is used for calculations."))
      wake_end = wake_ends[no_naps[1]]
      sleep_start = sleep_starts[no_naps[1]]
      sleep_end = sleep_ends[no_naps[1]]
      wake_start = wake_starts[no_naps[1]]
    }
    # if exactly one start and end, then simply extract start and end
  }  else{
    # Extract the times of start and end, should be exactly one
    wake_end = data$DATE_TIME[which(data$diff == -1)] # last time WAKE = 1
    sleep_start = data$DATE_TIME[which(data$diff == -1) + 1] # This is the first time WAKE = 0
    sleep_end = data$DATE_TIME[which(data$diff == 1)] # This is the last time WAKE = 0
    wake_start = data$DATE_TIME[which(data$diff == 1) + 1] # First time WAKE = 1
  }

  # Allocate period to pre-sleep, post-wake, pre-wake or just regular wake/sleep depending on the hours relative to sleep start/end
  data = data%>%
    dplyr::mutate(period = dplyr::case_when(
      (DATE_TIME < sleep_start) & (DATE_TIME >= sleep_start - lubridate::hours(2)) ~ "presleep",
      (DATE_TIME > sleep_end) & (DATE_TIME <= sleep_end + lubridate::hours(2)) ~ "postwake",
      (DATE_TIME <= sleep_end) & (DATE_TIME > sleep_end - lubridate::hours(2)) ~ "prewake",
      TRUE ~ ifelse(WAKE == 1, "wake", "sleep")
    ))

  # if first row is NA period, remove dummy row from coercing sleep_start
  if (is.na(data$period[1])){
    data <- data[2:nrow(data), ]
  }

  # Sometimes nothing is allocated to a particular period. Can adjust definition of sleep_start/sleep_end a little to see if extra measurements could be taken. This can only be done for presleep and postwake.
  if (sum(data$period == "presleep") == 0){
    # Presleep was calculated as 2 hours from sleep start, where sleep_start was the 1st measurement with WAKE = 0. Try to stretch out by counting 2 hours from wake_last instead
    data = data%>%
      dplyr::mutate(period = dplyr::case_when(
        (DATE_TIME <= wake_end) & (DATE_TIME > wake_end - lubridate::hours(2)) ~ "presleep",
        TRUE ~ period
      ))
  }

  if (sum(data$period == "postwake") == 0){
    # Postwake was calculated as 2 hours from sleep end. Try to stretch out by counting 2 hours from wake_start
    data = data%>%
      dplyr::mutate(period = dplyr::case_when(
        (DATE_TIME >= wake_start) & (DATE_TIME < wake_start + lubridate::hours(2)) ~ "postwake",
        TRUE ~ period
      ))
  }

  # Create a new tibble where only 3 periods are separated, and are factors, so that NA are formed if some are absent
  data_period = data %>%
    dplyr::filter(period %in% c("presleep", "prewake", "postwake"))%>%
    dplyr::mutate(period = factor(period, levels = c("presleep", "prewake", "postwake")))

  # SBP
  ###########################################################################
  # Mean and sd by WAKE
  stat_SBP_wake = tidyr::pivot_wider(data %>%
                                       dplyr::group_by(WAKE) %>%
                                       dplyr::summarize(meanSBP = mean(SBP, na.rm = TRUE),
                                                        sdSBP = sd(SBP, na.rm = TRUE)),
                                     values_from = 2:3, names_from = 1)
  names(stat_SBP_wake) = c("sleep_SBP", "wake_SBP", "sleep_SBP_sd", "wake_SBP_sd")

  # Mean by period
  stat_SBP_period = tidyr::pivot_wider(data_period %>%
                                         dplyr::group_by(period) %>%
                                         dplyr::summarize(SBP = mean(SBP, na.rm = TRUE)) %>%
                                         tidyr::complete(period),
                                       values_from = 2, names_from = 1)

  names(stat_SBP_period) = c( "presleep_SBP", "prewake_SBP", "postwake_SBP")

  # Combine the two
  summary_SBP = dplyr::bind_cols(stat_SBP_wake, stat_SBP_period)

  # Mean over lowest BP reading during sleep, and two surrounding it
  lowest_id = which((data$SBP == min(data$SBP[data$WAKE == 0], na.rm = TRUE)) &
                      (data$WAKE == 0))
  summary_SBP$lowest_SBP = mean(data$SBP[c(lowest_id - 1, lowest_id, lowest_id + 1)],
                                na.rm = TRUE)

  # DBP
  ###########################################################################
  # Mean and sd by WAKE
  stat_DBP_wake = tidyr::pivot_wider(data %>%
                                       dplyr::group_by(WAKE) %>%
                                       dplyr::summarize(meanDBP = mean(DBP, na.rm = TRUE),
                                                        sdDBP = sd(DBP, na.rm = TRUE)),
                                     values_from = 2:3, names_from = 1)
  names(stat_DBP_wake) = c("sleep_DBP", "wake_DBP", "sleep_DBP_sd", "wake_DBP_sd")

  # Mean by period
  stat_DBP_period = tidyr::pivot_wider(data_period %>%
                                         dplyr::group_by(period) %>%
                                         dplyr::summarize(DBP = mean(DBP, na.rm = TRUE)) %>%
                                         tidyr::complete(period),
                                       values_from = 2, names_from = 1)
  names(stat_DBP_period) = c("presleep_DBP", "prewake_DBP", "postwake_DBP")

  # Combine the two
  summary_DBP = dplyr::bind_cols(stat_DBP_wake, stat_DBP_period)

  # Mean over lowest BP reading during sleep, and two surrounding it
  lowest_id = which((data$DBP == min(data$DBP[data$WAKE == 0], na.rm = TRUE)) &
                      (data$WAKE == 0))
  summary_DBP$lowest_DBP = mean(data$DBP[c(lowest_id - 1, lowest_id, lowest_id + 1)],
                                na.rm = TRUE)

  # Return the output (summary_SBP and summary_DBP) together
  return(dplyr::bind_cols(summary_DBP, summary_SBP))
}


#' Nocturnal Blood Pressure Metrics
#'
#' The \code{bp_sleep_metrics} function serves to calculate nocturnal metrics
#' from relevant medical literature.
#'
#' NOTE: Any reference to "sleep" in the bp package refers to an individual's nocturnal period;
#' "sleep" is used in an informal sense for intuitive purposes. Technically, from a clinical perspective,
#' indication of sleep is currently obtained through means of EEG or other highly specialized equipment.
#' For all intents and purposes, sleep in the context of this package refers to actigraphy-inferred
#' nocturnal periods of rest.
#'
#' @param data User-supplied data set containing blood pressure data. Must
#' contain a Systolic blood pressure (\code{SBP}), Diastolic blood pressure (\code{DBP})
#' as well as \code{ID}, \code{WAKE}, and \code{DATE_TIME} columns.
#'
#' @param subj Optional argument. Allows the user to specify and subset specific subjects
#' from the \code{ID} column of the supplied data set. The \code{subj} argument can be a single
#' value or a vector of elements. The input type should be character, but the function will
#' comply with integers so long as they are all present in the \code{ID} column of the data.
#'
#' @details
#' The calculation of BP metrics related to sleep is based on averages of BP readings from four periods as identified in Kairo et al. (2003):
#' presleep or evening (2 hours before sleep start),
#' prewake (2 hours before wake),
#' postwake or morning (2 hours after wake),
#' lowest (3 measurements centered at the minimal BP reading over sleep).
#' The function uses \code{WAKE} column to automatically allocate BP measurements to various periods. The following metrics are defined as a
#' function of the period averages (separately for SBP and DBP)
#'
#' \code{dip_calc} = 1 - mean_sleep_BP/mean_wake_BP (dip proportion)
#'
#' \code{noct_fall} = mean_presleep_BP - mean_lowest_BP (nocturnal fall)
#'
#' \code{ST_mbps} = mean_postwake_BP - mean_lowest_BP (sleep through morning blood pressure surge)
#'
#' \code{PW_mbps} = mean_postwake_BP - mean_prewake_BP (prewake morning blood pressure surge)
#'
#' \code{ME_avg} = (mean_presleep_BP + mean_postwake_BP)/2 (morning-evening average)
#'
#' \code{ME_diff} = mean_postwake_BP - mean_presleep_BP (morning-evening difference)
#'
#' \code{wSD} = ( (wake_SD x HRS_wake) + (sleep_SD x HRS_sleep) ) / (HRS_wake + HRS_sleep) (weighted standard deviation)
#'
#' @return The function outputs a list containing 4 tibble objects corresponding to the following tables:
#' \item{[[1]]}{Counts of how many BP measurements were observed overall (\code{N_total}),
#' total number of readings during sleep (\code{N_sleep}),
#' total number of readings during wake (\code{N_wake}),
#' number of unique hours recorded during the sleep period (\code{HRS_sleep}), and
#' number of unique hours recorded during the wake period (\code{HRS_wake}) for each subject ID and grouping variable}
#' \item{[[2]]}{Summary statistics for systolic BP measurements (SBP): mean SBP value over Sleep and Wake, sd SBP value
#' over Sleep and Wake, mean SBP value over presleep period (evening in Kario et al. (2003)), mean SBP value over prewake period,
#' mean SBP value over postwake period (morning in Kario et al. (2003)), mean SBP value over 3 reading centered at the lowest SBP
#' value during sleep}
#' \item{[[3]]}{Summary statistics for diastolic BP measurements (DBP), same as for SBP}
#' \item{[[4]]}{BP metrics associated with sleep as defined above, separately for SBP and DBP}
#'
#' @references
#' Kario, K., Pickering, T. G., Umeda, Y., Hoshide, S., Hoshide, Y., Morinari, M., ... & Shimada, K. (2003). Morning surge in blood
#' pressure as a predictor of silent and clinical cerebrovascular disease in elderly hypertensives: a prospective study. Circulation,
#' 107(10), 1401-1406.
#'
#' @export
#'
#' @examples
#' hypnos_proc <- process_data(bp_hypnos,
#'                            sbp = "syst",
#'                            dbp = "DIAST",
#'                            date_time = "date.time",
#'                            id = "id",
#'                            wake = "wake",
#'                            visit = "visit",
#'                            hr = "hr",
#'                            map = "map",
#'                            rpp = "rpp",
#'                            pp = "pp",
#'                            bp_type = "abpm")
#'
#' bp_sleep_metrics(hypnos_proc)
#'
bp_sleep_metrics <- function(data, subj = NULL){

  # Function requires: SBP, DBP, DATE_TIME, WAKE, ID

  # Initialize variables for dplyr
  ID = DATE_TIME = SBP = DBP = WAKE = HOUR = lowest_SBP = lowest_DBP = presleep_SBP = presleep_DBP = postwake_SBP = postwake_DBP = prewake_SBP = prewake_DBP = sleep_DBP = wake_DBP = sleep_SBP = wake_SBP = wSD_DBP = wake_SBP_sd = wake_DBP_sd = sleep_SBP_sd = sleep_DBP_sd = HRS_sleep = HRS_wake = NULL
  rm(list = c("ID", "DATE_TIME", "SBP", "DBP", "WAKE", "HOUR", "lowest_SBP", "lowest_DBP", "presleep_SBP", "presleep_DBP", "postwake_SBP", "postwake_DBP", "prewake_SBP", "prewake_DBP", "sleep_DBP", "wake_DBP", "sleep_SBP", "wake_SBP", "wSD_DBP", "wake_SBP_sd", "wake_DBP_sd", "sleep_SBP_sd", "sleep_DBP_sd", "HRS_sleep", "HRS_wake"))


  # Uppercase all column names
  colnames(data) <- toupper( colnames(data) )


  # Ensure SBP column is present
  if( ("SBP" %in% colnames(data)) == FALSE ){

    stop('Could not find SBP column in supplied data set. Make sure to run process_data() for data processing.')

  }

  # Ensure DBP column is present
  if( ("DBP" %in% colnames(data)) == FALSE ){

    stop('Could not find DBP column in supplied data set. Make sure to run process_data() for data processing.')

  }

  # Ensure that both DATE_TIME column is present
  if( ("DATE_TIME" %in% colnames(data)) == FALSE){

    stop('No DATE_TIME column found. Cannot compute sleep stage metrics.')

  }

  # Ensure WAKE column is present
  if( ("WAKE" %in% colnames(data)) == FALSE ){

    stop('No WAKE column found. Cannot compute sleep stage metrics.')

  }

  # Ensure that both DATE_TIME column is present
  if( ("ID" %in% colnames(data)) == FALSE){

    stop('No ID column found. Cannot compute sleep stage metrics.')

  }

  # Check the data type
  if ( ("BP_TYPE" %in% colnames(data)) == FALSE){
    stop('No BP_TYPE column found. Cannot automatically identify the type of BP data. Make sure to process the data using process_data function.')
  }

  if (unique(data$BP_TYPE) != "ABPM"){
    if (unique(data$BP_TYPE) == "HBPM"){
      warning("The supplied data has HBPM type, for which calculation of sleep BP metrics is not recommended. If the supplied data should be ABPM type, please rerun process_data function with correct BP type specfication.")
    }
  }

  # Check function for whether or not DATE_TIME values are proper date/time format
  is.POSIXct <- function(x) inherits(x, "POSIXct")

  if(is.POSIXct(data$DATE_TIME) == FALSE){

    stop('DATE_TIME column not proper time format. Make sure to run process_data() for data processing.')

  }

  # If user supplies a vector corresponding to a subset of multiple subjects (multi-subject only)
  if(!is.null(subj)){

    # check to ensure that supplied subject vector is compatible
    subject_subset_check(data, subj)

    if(length(unique(data$ID)) > 1){

      # Filter data based on subset of subjects
      data <- data %>%
        dplyr::filter(ID %in% subj)
    }
  }


  #++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++#

  # Grouping variables

  grps = c("ID", "VISIT", "GROUP")
  grps = grps[which( grps %in% colnames(data) == TRUE)]

  #++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++#
  # Calculation of sleep counts table (table 1) based on WAKE column

  sleep_counts = data %>%
    dplyr::group_by_at(dplyr::vars(grps)) %>%
    dplyr::summarize(N_total = dplyr::n(), # Total number of readings
                     N_wake = sum(WAKE == 1), # Number of total readings while awake
                     N_sleep = sum(WAKE == 0), # Number of total readings while asleep
                     HRS_wake = dplyr::n_distinct(HOUR[WAKE == 1]), # Number of unique hours recorded while awake (for wSD calc)
                     HRS_sleep = dplyr::n_distinct(HOUR[WAKE == 0]), # Number of unique hours recorded during sleep (for wSD calc)
                     .groups = 'drop')



  # ******************************************************************************************************************************* #
  #                                               BP Sleep Periods
  # ******************************************************************************************************************************* #

  # Order time from oldest to newest
  data = data[order(data$DATE_TIME), ]

  # Calculate all summaries on SBP and DBP using get_summaries_SBP_DBP
  output = data %>%
    dplyr::group_by_at(dplyr::vars(grps)) %>%
    dplyr::summarise(get_summaries_SBP_DBP(data.frame(ID, DATE_TIME, SBP, DBP, WAKE)))

  # Extract ids of grouping variables in output
  idgrps = which(names(output) %in% grps)

  #########################
  ##    Systolic (SBP)   ##
  #########################

  # Extract positions of SBP summaries in output
  idSBP = grep("SBP", names(output))

  # Create SBP table
  sleep_summary_SBP = output[ , c(idgrps, idSBP)]

  ##########################
  ##    Diastolic (DBP)   ##
  ##########################

  # Extract positions of DBP summaries in output
  idDBP = grep("DBP", names(output))

  # Create DBP table
  sleep_summary_DBP = output[ , c(idgrps, idDBP)]


  #++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++#
  ### Calculation of Sleep Metrics (Table 4) based on SBP/DBP sleep summaries (Tables 2-3)
  ####################################

  # Join the sleep_summary_SBP/DBP with the sleep counts to include the HRS_wake/sleep for the wSD calculation
  sleep_summary_SBP_all <- suppressMessages(dplyr::left_join(sleep_summary_SBP, sleep_counts)) #suppressMessages ignores the "joining by: ..." message
  sleep_summary_DBP_all <- suppressMessages(dplyr::left_join(sleep_summary_DBP, sleep_counts))

  # Metrics based on SBP
  sleep_metrics_SBP <- sleep_summary_SBP_all %>%
    dplyr::group_by_at(dplyr::vars(grps)) %>%
    dplyr::transmute(
      # dip calculation (proportion)
      dip_calc_SBP = 1 - sleep_SBP / wake_SBP,
      # "Nocturnal BP Fall" (Kario et al 2002)
      noct_fall_SBP = presleep_SBP / lowest_SBP,
      # Sleep-Trough MBPS (Kario et al 2002)
      ST_mbps_SBP = postwake_SBP - lowest_SBP,
      # Prewake MBPS (Kario et al 2002)
      PW_mbps_SBP = postwake_SBP - prewake_SBP,
      # Morningness-Eveningness Average (Kario 2005)
      ME_SBP_avg = (postwake_SBP + presleep_SBP) / 2,
      # Morningness-Eveningness Difference (Kario 2005)
      ME_SBP_diff = postwake_SBP - presleep_SBP,
      # Weighted Standard Deviation (Bilo et al 2007)
      wSD_SBP = ( ( (wake_SBP_sd * HRS_wake) + (sleep_SBP_sd * HRS_sleep) ) / (HRS_wake + HRS_sleep) )
    )

  # Metrics based on DBP
  sleep_metrics_DBP <- sleep_summary_DBP_all %>%
    dplyr::group_by_at(dplyr::vars(grps)) %>%
    dplyr::transmute(
      # dip calculation (proportion)
      dip_calc_DBP = 1 - sleep_DBP / wake_DBP,
      # "Nocturnal BP Fall" (Kario et al 2002)
      noct_fall_DBP = presleep_DBP / lowest_DBP,
      # Sleep-Trough MBPS (Kario et al 2002)
      ST_mbps_DBP = postwake_DBP - lowest_DBP,
      # Prewake MBPS (Kario et al 2002)
      PW_mbps_DBP = postwake_DBP - prewake_DBP,
      # Morningness-Eveningness Average (Kario 2005)
      ME_DBP_avg = (postwake_DBP + presleep_DBP) / 2,
      # Morningness-Eveningness Difference (Kario 2005)
      ME_DBP_diff = postwake_DBP - presleep_DBP,
      # Weighted Standard Deviation (Bilo et al 2007)
      wSD_DBP = ( ( (wake_DBP_sd * HRS_wake) + (sleep_DBP_sd * HRS_sleep) ) / (HRS_wake + HRS_sleep) )
    )

  # Combine the two together
  sleep_metrics <- dplyr::left_join(sleep_metrics_SBP, sleep_metrics_DBP, by = c(grps))

  #
  # # Weighted Standard Deviation (wSD)
  # tmp1 <- sleep_summary_SBP %>%
  #               dplyr::mutate( wSD_SBP = ( ( awake_SBP_sd * N_awake ) + ( sleep_SBP_sd * N_sleep ) ) / ( N_awake + N_sleep ) ) %>%
  #               dplyr::select("wSD_SBP")
  # tmp2 <- sleep_summary_DBP %>%
  #               dplyr::mutate( wSD_DBP = ( ( awake_DBP_sd * N_awake ) + ( sleep_DBP_sd * N_sleep ) ) / ( N_awake + N_sleep ) ) %>%
  #               dplyr::select("wSD_SBP")
  #
  # sleep_metrics <- dplyr::left_join(sleep_metrics, tmp1)
  # sleep_metrics <- dplyr::left_join(sleep_metrics, tmp2)
  #

  return(list("Sleep_Counts" = sleep_counts,
              "SBP_Sleep_Summary" = sleep_summary_SBP,
              "DBP_Sleep_Summary" = sleep_summary_DBP,
              "Sleep_Metrics" = sleep_metrics))
}