Skip to contents

Interview Monitoring, Paradata, and Quality Control

Source: vignettes/interview_monitoring.Rmd
interview_monitoring.Rmd

Introduction

Continuous quality control (QC) and high-frequency checks (HFC) are vital to preventing non-sampling errors during fieldwork.

SurveySolutionsAPIv2 equips survey teams with real-time tools to: 1. Monitor interview workflow statuses. 2. Take supervisory actions (commenting, approving, rejecting, reassigning). 3. Query real-time response distributions. 4. Process granular event paradata (timings, pauses, and spatial traces).

1. Tracking Interview Progress

Every interview in Survey Solutions moves through a defined lifecycle:

[Created / Assigned]
        │
        ▼
   [Completed] ──► [Supervisor Review]
                          │
             ┌────────────┴────────────┐
             ▼                         ▼
   [RejectedBySupervisor]    [ApprovedBySupervisor]
             │                         │
             ▼                         ▼
   (Back to enumerator)       [Headquarters Review]
                                       │
                          ┌────────────┴────────────┐
                          ▼                         ▼
               [RejectedByHeadquarters]  [ApprovedByHeadquarters]

Querying Interviews by Status

suso_getAllInterviewQuestionnaire() retrieves all interviews matching target criteria:

# Get all completed interviews awaiting review
completed_ints <- suso_getAllInterviewQuestionnaire(
  questID = "17a9fa22-1ef6-46d2-8c30-426aa876f273",
  version = 1,
  workStatus = "Completed"
)

The resulting table provides status, error counts, enumerator assignment, and last modification timestamp:

# Sample interview statuses
demo_ints[, .(InterviewId = substr(InterviewId, 1, 13), ResponsibleName, Status, ErrorsCount)]
#>      InterviewId ResponsibleName                 Status ErrorsCount
#>           <char>          <char>                 <char>       <int>
#> 1: a1b2c3d4-0001   interviewer01              Completed           0
#> 2: a1b2c3d4-0002   interviewer01   ApprovedBySupervisor           0
#> 3: a1b2c3d4-0003   interviewer01   RejectedBySupervisor           2
#> 4: a1b2c3d4-0004   interviewer02    InterviewerAssigned           1
#> 5: a1b2c3d4-0005   interviewer03 ApprovedByHeadquarters           0
# Summary breakdown of interview statuses
demo_ints[, .N, by = Status]
#>                    Status     N
#>                    <char> <int>
#> 1:              Completed     1
#> 2:   ApprovedBySupervisor     1
#> 3:   RejectedBySupervisor     1
#> 4:    InterviewerAssigned     1
#> 5: ApprovedByHeadquarters     1

2. Supervisory Actions and Audit

When high-frequency checks flag inconsistencies, supervisors or HQ monitors can interact directly with the interview.

Leaving Comments on Question Responses

Comments can be attached directly to a specific variable with optional feedback:

# Leave a comment on the respondent age variable
suso_commentInterview(
  intID    = "a1b2c3d4-0003-4000-8000-000000000003",
  variable = "age",
  comment  = "Reported age (14) is inconsistent with marital status (Married). Please verify with respondent."
)

Rejecting and Approving Interviews 

# Reject back to supervisor with explanation
suso_patchRejectInterview(
  intID   = "a1b2c3d4-0003-4000-8000-000000000003",
  comment = "Multiple validation errors require re-visiting the household."
)

# Approve interview at Headquarters
suso_patchApproveInterview(
  intID   = "a1b2c3d4-0001-4000-8000-000000000001",
  comment = "Passed all automated consistency audits."
)

Downloading Interview PDF Transcripts

For formal documentation or audit trails, download the full interview transcript as a formatted PDF:

# Download completed transcript
pdf_file <- suso_getInterviewPDF(
  intID = "a1b2c3d4-0001-4000-8000-000000000001",
  path  = "audits/interview_0001_transcript.pdf"
)

3. Real-Time Question Statistics

To detect enumerator fabrication or question misinterpretation early, suso_get_stats() queries live response distributions directly from the server:

# Query response distribution for a categorical question
q_stats <- suso_get_stats(
  questID = "17a9fa22-1ef6-46d2-8c30-426aa876f273",
  version = 1,
  qQuest  = "household_head_gender"
)

4. Paradata Processing

Paradata captures timestamped events triggered during data collection: question focus, response edits, validation messages, and pauses.

Exporting Paradata with suso_export_paradata()

suso_export_paradata() downloads and structures paradata events into an enriched exportClass table:

# Export paradata for a questionnaire
para <- suso_export_paradata(
  questID           = "17a9fa22-1ef6-46d2-8c30-426aa876f273",
  version           = 1,
  onlyActiveEvents  = TRUE,
  reloadTimeDiff    = 0
)

Key QC Metrics from Paradata

  1. Active Interview Duration: Calculates total active time excluding breaks (default break threshold is 120 seconds).
  2. Response Latency: Measures elapsed seconds per question to identify speeders or difficult modules.
  3. GPS Coordinates: Automatically extracts geographic coordinates associated with paradata events to verify that interviews occurred within assigned sample clusters.