使用pivot_longer()，您可以指定".value" to names_to以成对堆叠date和n_symp。在这种情况下，必须提供names_sep或names_pattern之一以指定应如何拆分列名。然后，您可以轻松地将缺少n_symp的日期替换为NA。最后，将长数据旋转得更宽以获得原始格式。

library(dplyr)
library(tidyr)

df %>%
  mutate(id = 1:n()) %>% 
  pivot_longer(-id, names_to = c("fu", ".value"), names_sep = "(?<=\\d)_") %>%
  mutate(date = replace(date, is.na(n_symp), NA)) %>%
  pivot_wider(names_from = fu, values_from = c(date, n_symp),
              names_glue = "{fu}_{.value}", names_vary = "slowest")

# # A tibble: 3 × 7
#      id fu1_date   fu1_n_symp fu5_date   fu5_n_symp fu7_date   fu7_n_symp
#   <int> <date>          <int> <date>          <int> <date>          <int>
# 1     1 2012-03-05          1 NA                 NA 2016-03-05          1
# 2     2 2013-08-09          1 2015-10-09          2 NA                 NA
# 3     3 2019-05-05          1 2020-06-07          2 2021-07-09          2

pivot_wider()中的names_vary控制组合结果列名的顺序。

• “最快”（默认）

fu1_date   fu5_date   fu7_date   fu1_n_symp fu5_n_symp fu7_n_symp

• “最慢”

fu1_date   fu1_n_symp fu5_date   fu5_n_symp fu7_date   fu7_n_symp

更新：@Darren Tsai输入后调整代码：

df %>% 
  mutate(id = row_number()) %>% 
  pivot_longer(-id,
               names_to = "key",
               values_to = "val", 
               values_transform = list(val = as.character)) %>% # change all to character class
  mutate(val = ifelse(is.na(lead(val, default = val[1])), NA_character_, val)) %>% 
  pivot_wider(names_from= key, values_from = val) %>% 
  type_convert() %>% 
  select(-id)

下面是一种方法，使用透视：
x一个一个一个一个x一个一个二个x