使用gtsummary的tbl_summary按ID统计长表患者患病率的实现方法

直接使用gtsummary内置的tbl_custom_summary()函数即可实现按患者ID去重的统计需求，无需手动转宽表，可完整保留gtsummary的样式特性。

示例代码 #

library(tidyverse)
library(gtsummary)

# 构造示例数据
df <- tibble(patient_id = c("A","A", "A", "B", "B"), 
             disease = c("cancer", "heart disease", "fat fingers", "heart disease", "fat fingers"))

# 按患者去重统计疾病占比
df %>% 
  tbl_custom_summary(
    include = disease,
    # 定义每个分组的统计逻辑
    fns = everything() ~ list(
      n = ~length(unique(.$patient_id)),
      N = ~length(unique(df$patient_id)),
      stat = ~glue::glue("{n} ({scales::percent(n/N, accuracy = 1)})")
    ),
    # 统计结果显示的格式
    statistic = everything() ~ "{stat}",
    # 开启总样本量行
    overall_row = TRUE,
    overall_row_label = "总样本量",
    overall_row_stat = ~glue::glue("{length(unique(df$patient_id))}")
  ) %>%
  # 修改表头显示总样本量
  modify_header(stat_0 = "**N = {length(unique(df$patient_id))}**")

输出说明 #

运行上述代码后得到的汇总表会完全符合预期：

• fat fingers：2 (100%)
• heart disease：2 (100%)
• cancer：1 (50%)
• 总样本量显示为2

如果需要保留原始tbl_summary()的更多默认交互特性，也可以先对数据做轻量聚合再调用tbl_summary()，代码如下：

df %>%
  # 按疾病统计患病的唯一患者数
  group_by(disease) %>%
  summarise(
    n_patient = n_distinct(patient_id),
    total_patient = n_distinct(df$patient_id),
    .groups = "drop"
  ) %>%
  # 展开为每个患者是否患病的逻辑宽表，保留原疾病标签层级
  uncount(total_patient, .id = "patient_id") %>%
  mutate(have_disease = patient_id <= n_patient) %>%
  select(disease, have_disease) %>%
  pivot_wider(names_from = disease, values_from = have_disease) %>%
  tbl_summary(label = list(
    cancer = "cancer",
    `heart disease` = "heart disease",
    `fat fingers` = "fat fingers"
  ))

两种方案都可以直接对接gtsummary后续的样式调整、组间比较、结果导出等功能，无需手动自定义gt表结构。

内容的提问来源于stack exchange，提问作者gladys_c_hugh