基于特定条件合并R语言患者随访数据的技术问询

原始数据集 #

df <- data.frame(
  Patient = c('Dave', 'Dave', 'Dave','Dave','Dave','Dave','Dave', "Angel", "Angel", "Angel", "Cara", "Cara"),
  V1 = c(1, 150, 240,430,530,650,800, 1, 150, 375, 1, 150),
  V2 = c(150, 240,430,530,650,800,900, 150, 375, 568, 150,375),
  R1 = c("Disease","Disease","Disease","Response","Response","Disease","Response", "Disease","Response", "Response", "Disease", "Response"),
  R2 = c("Disease", "Disease", "Response","Response","Disease","Response","Response", "Response", "Response", "Response", "Response", "Death")
)

需求说明 #

按患者分组，合并状态连续相同的行直至状态发生变化，最终数据集保留R1≠R2的行（最后一条记录除外）。逻辑伪代码如下：

if (r1 = r2 & not last record) { 
  record v1 and r1 
  walkthrough data until r1=r2 or last record 
  record v2 and r2 
  print 
}

预期输出 #

解决方案（基于dplyr） #

library(dplyr)

df_processed <- df %>%
  group_by(Patient) %>%
  # 生成状态分组：当相邻行的状态衔接中断时，创建新分组
  mutate(status_group = cumsum(c(TRUE, R2[-n()] != R1[-1]))) %>%
  group_by(Patient, status_group) %>%
  # 聚合每组的起始/结束时间、起始/结束状态
  summarise(
    V1 = first(V1),
    V2 = last(V2),
    R1 = first(R1),
    R2 = last(R2),
    .groups = "drop"
  ) %>%
  group_by(Patient) %>%
  # 标记每组最后一条记录，筛选符合条件的行
  mutate(is_last = row_number() == n()) %>%
  filter(R1 != R2 | is_last) %>%
  select(-status_group, -is_last)

# 查看结果
print(df_processed)

代码逻辑说明 #

1. 状态分组：按患者分组后，通过cumsum识别状态变化节点——每当前一行的结束状态（R2）和当前行的起始状态（R1）不一致时，就生成新的分组编号，将连续相同状态的行归为一组。
2. 聚合数据：对每个状态分组，提取组内最早的随访起始时间（V1）、最晚的随访结束时间（V2）、起始状态（R1）和结束状态（R2）。
3. 筛选结果：再次按患者分组，标记每组的最后一条记录，保留R1≠R2的行或每组的最后一条记录，满足需求。

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