循环遍历 dataframe 中的列并根据 R 中的特定条件连接字符串

Question

The following df represents treatments that a single patient has received during the course of a study. They first received drug-v, followed by drug-w, followed by drug-x, and so on.

original <- tibble::tribble(
  ~treatment_administered,
                 "drug-v",
                 "drug-w",
                 "drug-x",
                 "drug-y",
                 "drug-z",
                 "drug-l"
  )
original

My aim is to keeping a cumulative record of prior treatment exposures that belong to a specific class of treatment - let's call this "class A". In this example, drug v, drug x and drug z belong to class A. Here is the final df I wish to create.

final <- tibble::tribble(
              ~prior_classA_details, ~treatment_administered,
                                 "",                "drug-v",
                           "drug-v",                "drug-w",
                           "drug-v",                "drug-x",
                   "drug-v,drug-x",                "drug-y",
                   "drug-v,drug-x",                "drug-z",
           "drug-v, drug x,drug-z",                "drug-l"
           )
final

As you can see, prior_classA_details is tracking treatment_administered on the previous row, and if it's a class A treatment, it adds the name to the following row. This is an iterative process as it goes down the list, concatenating prior_classA_details as class A treatments are administered.

There are multiple other data columns in this df that I have not included here (only relevant columns included). Ideally looking for a dplyr solution please.

Answer 1

Here's one way -

library(dplyr)
library(purrr)

classA <- c("drug-v", "drug-x", "drug-z")

original %>%
  mutate(prior_classA_details = lag(map_chr(row_number(), ~{
   toString(keep(treatment_administered[seq_len(.x)], function(y) y %in% classA))
    }), default = ''), .before = 1)

#  prior_classA_details     treatment_administered
#  <chr>                    <chr>                 
#1 ""                       drug-v                
#2 "drug-v"                 drug-w                
#3 "drug-v"                 drug-x                
#4 "drug-v, drug-x"         drug-y                
#5 "drug-v, drug-x"         drug-z                
#6 "drug-v, drug-x, drug-z" drug-l                

We create a vector for classA drugs and for each row keep only those values that are of type classA in a cumulative fashion and create one concatenated string. lag is used to get lagged records by step 1.