Skip to contents

Dual MDC-vs-MCID responder classification

Source: R/responder-classify.R
classifyResponder.Rd

Classifies each subject's pre-to-post change with the two-criterion responder framework (Beaton 2001; de Vet 2006): a change is cross-tabulated on whether it exceeds the Minimal Detectable Change (is it real, above measurement error?) and the Minimal Clinically Important Difference (is it clinically meaningful?), giving four categories — true_responder (exceeds both), subclinical_change (real but below MCID), measurement_error (claims MCID but within noise, only when MCID < MDC) and non_responder. Vectorized over subjects and direction-aware.

Usage

classifyResponder(
  baseline,
  followup,
  instrument = NULL,
  population = NULL,
  mdc = NULL,
  mcid = NULL,
  direction = c("increase", "decrease")
)

# S3 method for class 'responder_classification'
summary(object, ...)

Arguments

baseline, followup

Numeric vectors of pre / post scores (one per subject).

instrument, population

Optional instrument / population used to look up mdc/mcid from the clinimetric store when they are not given.

mdc, mcid

Optional explicit MDC / MCID thresholds (positive); override the store lookup.

direction

"increase" (default) if higher is better, or "decrease" if lower is better.

object

A "responder_classification".

...

Unused.

Value

A "responder_classification" data.frame with the change, the direction-aware improvement, the MDC/MCID crossing flags and the four-level classification factor. A subject with a missing baseline/followup gets NA flags and classification (missing, not assumed a non-responder); summary() counts it.

summary returns the MDC x MCID contingency table.

References

Beaton DE et al. (2001); de Vet HCW et al. (2006).

See also

estimateMDC(), getClinimetric(), mdcResponder()

Examples

classifyResponder(c(20, 25, 30), c(31, 26, 45), mdc = 5.2, mcid = 9)
#> <responder_classification> 3 subject(s), direction: increase
#>   MDC = 5.2, MCID = 9
#> 
#>     true_responder subclinical_change  measurement_error      non_responder 
#>                  2                  0                  0                  1