如何让gtsummary基于ITime类型正确展示时间统计量？

我有POSIXct格式的医院入院时间（例如"1899-12-31 13:11:00"），无需关注日期，因此用lubridate包提取时分秒并格式化为%H:%M:%S，再转换为ITime类型（该类型以当日秒数整数存储），方便计算入院时间的均值、中位数、最值等统计量。

单独计算统计量时结果正常，比如执行mean(sample_data$HOSP_ADMSN_TIME.x)会输出预期的"14:11:18"。但将该变量传入gtsummary生成汇总表时，它把变量视为连续变量，均值显示为57078（秒数）而非预期的时间格式。

核心需求：让gtsummary基于ITime类型处理时间戳，计算并展示正确格式的统计量。

示例数据集 #

# A tibble: 10 × 1
   HOSP_ADMSN_TIME.x  
   <dttm>             
 1 2021-04-19 10:00:00
 2 2021-05-02 09:18:00
 3 2021-09-09 15:06:00
 4 2021-08-17 13:10:00
 5 2022-01-18 11:44:00
 6 2021-10-13 20:33:00
 7 2021-05-06 19:09:00
 8 2021-08-04 12:49:00
 9 2021-11-15 11:00:00
10 2021-09-01 19:04:00

数据处理代码 #

sample_data$HOSP_ADMSN_TIME.x <- format(ymd_hms(sample_data$HOSP_ADMSN_TIME.x), "%H:%M:%S")
sample_data$HOSP_ADMSN_TIME.x <- as.ITime(sample_data$HOSP_ADMSN_TIME.x)

验证统计量计算 #

mean(sample_data$HOSP_ADMSN_TIME.x)
[1] "14:11:18"

当前gtsummary汇总表代码 #

sample_data %>%
  tbl_summary(
    label = list(HOSP_ADMSN_TIME.x ~ "Hospital Admission Time"),
    missing = "ifany",
    missing_text = "Missing",
    type = HOSP_ADMSN_TIME.x ~ "continuous2",
    statistic = HOSP_ADMSN_TIME.x ~ c("{mean}",
                                        "{median} ({p25}, {p75})",
                                        "{min}, {max}")
  ) %>%
  bold_labels() %>%
  modify_header(label ~ "**Variable**")

（当前生成的表格中均值显示为57,078而非预期时间格式）

gtsummary默认会读取ITime类型变量的底层存储值（秒数整数）来生成统计量，因此需要通过自定义逻辑将计算结果转回时间格式，以下是三种可行方法：

方法1：自定义统计表达式 #

直接在statistic参数中使用自定义表达式，计算后将秒数结果转换为ITime格式再输出：

sample_data %>%
  tbl_summary(
    label = list(HOSP_ADMSN_TIME.x ~ "Hospital Admission Time"),
    missing = "ifany",
    missing_text = "Missing",
    type = HOSP_ADMSN_TIME.x ~ "continuous2",
    statistic = HOSP_ADMSN_TIME.x ~ list(
      ~ as.character(as.ITime(mean(.x))),
      ~ sprintf("%s (%s, %s)", 
                as.character(as.ITime(median(.x))),
                as.character(as.ITime(quantile(.x, 0.25))),
                as.character(as.ITime(quantile(.x, 0.75)))),
      ~ sprintf("%s, %s", 
                as.character(as.ITime(min(.x))),
                as.character(as.ITime(max(.x))))
    )
  ) %>%
  bold_labels() %>%
  modify_header(label ~ "**Variable**")

方法2：提前计算统计量再传入 #

手动计算所有需要的统计量，直接将格式化后的结果传入statistic参数，适合需要精细控制的场景：

# 预先计算并格式化统计量
time_stats <- list(
  mean = as.character(as.ITime(mean(sample_data$HOSP_ADMSN_TIME.x))),
  median_iqr = sprintf("%s (%s, %s)", 
                       as.character(as.ITime(median(sample_data$HOSP_ADMSN_TIME.x))),
                       as.character(as.ITime(quantile(sample_data$HOSP_ADMSN_TIME.x, 0.25))),
                       as.character(as.ITime(quantile(sample_data$HOSP_ADMSN_TIME.x, 0.75)))),
  min_max = sprintf("%s, %s", 
                    as.character(as.ITime(min(sample_data$HOSP_ADMSN_TIME.x))),
                    as.character(as.ITime(max(sample_data$HOSP_ADMSN_TIME.x))))
)

# 生成汇总表
sample_data %>%
  tbl_summary(
    label = list(HOSP_ADMSN_TIME.x ~ "Hospital Admission Time"),
    missing = "ifany",
    missing_text = "Missing",
    type = HOSP_ADMSN_TIME.x ~ "continuous2",
    statistic = HOSP_ADMSN_TIME.x ~ c(time_stats$mean, time_stats$median_iqr, time_stats$min_max)
  ) %>%
  bold_labels() %>%
  modify_header(label ~ "**Variable**")

方法3：注册ITime类型的全局处理规则 #

如果需要频繁处理ITime变量，可以注册一个gtsummary的全局统计函数，让它自动识别并格式化ITime类型的统计结果：

# 注册自定义统计函数
register_statistic(
  fn = function(data, variable, ...) {
    x <- data[[variable]]
    tibble(
      label = c("Mean", "Median (IQR)", "Min, Max"),
      stat = c(
        as.character(as.ITime(mean(x))),
        sprintf("%s (%s, %s)", 
                as.character(as.ITime(median(x))),
                as.character(as.ITime(quantile(x, 0.25))),
                as.character(as.ITime(quantile(x, 0.75)))),
        sprintf("%s, %s", 
                as.character(as.ITime(min(x))),
                as.character(as.ITime(max(x))))
      )
    )
  },
  type = "continuous",
  class = "ITime"
)

# 之后直接调用tbl_summary即可自动处理ITime变量
sample_data %>%
  tbl_summary(
    label = list(HOSP_ADMSN_TIME.x ~ "Hospital Admission Time"),
    missing = "ifany",
    missing_text = "Missing"
  ) %>%
  bold_labels() %>%
  modify_header(label ~ "**Variable**")

内容的提问来源于stack exchange，提问作者emvirgen