DP19 Classification Categories using Dplyr

Description

Having initial data with some fields about the characteristics of the record the task is to create a special classification field. For that purpose, an evaluation with conditional statements will be performed on each record to give them the final classification.

In this example, it is used the data of interactions with the patients and features about the contact and the outcome. Using the function mutate from the dplyr package helper columns will be added, then group_by to get a recap by categories. This method is similar to the Excel pivot tables.

Link to the Complete Script in Github

R Script - Classification Categories using mutate

First Classification: Patient-Centered Interaction (PCI)

From all the options for Primary.Participant it is required to classify a subgroup as “Patient-Centered Interaction” or PCI. So the field “Prim.Part.PCI” is added as a helper column:

# Adding PCI classification
dfINT <- dfINT %>% 
  dplyr::mutate(Prim.Part.PCI = if_else(Primary.Participant %in% c("Member",
          "Parent","Caregiver","Guardian","Power of Attorney"),"Yes", "")) # 
# Recap of fields: Primary.Participant and Prim.Part.PCI
data.frame(dfINT %>% group_by(Primary.Participant, Prim.Part.PCI) %>% 
             tally() %>% arrange(desc(Prim.Part.PCI), desc(n)))

This recap results shows the Primary Participants that were classified as PCI:

Categories classified as Patient-Centered Interaction

Second Classification: Mode PCI

Similarly, a classification field Mode.PCI is added depending on the Mode of contact:

# Mode.PCI field for interactions with Call, Visit, Videoconference
dfINT <- dfINT %>% mutate(Mode.PCI = if_else(Mode %in% 
                         c("Call", "Visit", "Videoconference"), "Yes", ""))
# Results
data.frame(dfINT %>% group_by(Mode, Mode.PCI) %>% 
             tally() %>% arrange(desc(Mode.PCI), desc(n)))

Mode.PCI classification

Final “Interaction Classification” field

Logic

The next is the logic to get the “Interaction Classification” field, using the previously defined helper fields.

Final Result	Conditions
“Unable to Reach (UTR)”	Outgoing.Contact.Result == “Unable to Reach (UTR)”
“Completed Pt. Centered Interaction”	Prim.Part.PCI == “Yes” & Mode.PCI == “Yes”
“Other Interactions”	if none of the above conditions were met

Code

# Int_Classification
# Without Other Completed Int. w/Patient
dfINT <- dfINT %>% 
  dplyr::mutate(Interaction_Classification = 
                  if_else(Outgoing.Contact.Result == "Unable to Reach (UTR)", 
                          "Unable to Reach (UTR)",
                          if_else(
                            Prim.Part.PCI == "Yes" &
                              Mode.PCI == "Yes",
                            "Completed Pt. Centered Interaction",  
                            "Other Interactions")))

Result of Classification using a Helper column

Summarizing the results:

dfINT %>% group_by(Interaction_Classification, 
                         Prim.Part.PCI, Mode.PCI, 
                         Mode) %>% tally() %>% 
        arrange(Interaction_Classification, desc(n))

Data classified by Interaction_Classification field

Exploring Specific groups. Data Mining

Using dplyr with specific filters, we can return to the interaction level and examine specific groups, such as patients who received more than 6 interactions.

dfINT %>% filter(Interaction_Classification == "Completed Pt. Centered Interaction") %>% 
  group_by(Client) %>% tally() %>% arrange(desc(n)) %>% filter(n >= 6)

Results in R console

Code of Patients that received more that 6 interacions.

__

End of Post