Individual-Level Analysis
At the individual level, we classify each participant based on whether their change score crosses the MID.
A Basic Example
We will use the claus_2020 dataset, which contains data
from a study on enhancing the placebo effect of antidepressants. The
primary outcome was the Beck Depression Inventory (BDI-II), where lower
scores indicate improvement. For the BDI-II, a change of 7
points is often considered the MID.
The claus_2020 dataset has four measurements per
patient. For a pre-post analysis, we must specify which time points to
compare using the pre and post arguments.
# Analyze individual change from the first to the fourth measurement
anchor_individual <- claus_2020 |>
cs_anchor(
id = id,
time = time,
outcome = bdi,
pre = 1,
post = 4,
mid_improvement = 7
)
# Print the summary table
anchor_individual
#>
#> ---- Clinical Significance Results ----
#>
#> Approach: Anchor-based
#> MID Improvement: 7
#> MID Deterioration: 7
#> Better is: Lower
#>
#> Category | N | Percent
#> ---------------------------
#> Improved | 25 | 62.50%
#> Unchanged | 11 | 27.50%
#> Deteriorated | 4 | 10.00%The output shows that 62.5% of patients were classified as “Improved” (their BDI-II score decreased by 7 points or more), 27.5% were “Unchanged”, and 10% “Deteriorated”.
Visualizing Individual Change
A clinical significance plot helps visualize these results. Each point represents a patient.
- Solid Diagonal Line: No change (
pre=post). Points below this line represent improvement for the BDI-II. - Shaded Area: The zone of “no meaningful change”.
Its boundaries are
post = pre ± MID. - Points below the shaded area: Clinically significant improvement.
- Points above the shaded area: Clinically significant deterioration.
Analyzing by Group
The claus_2020 study included a Treatment as Usual (TAU)
group and a Placebo-Amplification (PA) group. We can analyze the
clinical significance for each group by specifying the
group argument.
anchor_grouped <- claus_2020 |>
cs_anchor(
id = id,
time = time,
outcome = bdi,
pre = 1,
post = 4,
mid_improvement = 7,
group = treatment
)
# Print results by group
summary(anchor_grouped)
#>
#> ---- Clinical Significance Results ----
#>
#> Approach: Anchor-based
#> MID Improvement: 7
#> MID Deterioration: 7
#> N (original): 43
#> N (used): 40
#> Percent (used): 93.02%
#> Better is: Lower
#> Outcome: bdi
#>
#> Group | Category | N | Percent | Percent by Group
#> ------------------------------------------------------
#> TAU | Improved | 8 | 20.00% | 42.11%
#> TAU | Unchanged | 7 | 17.50% | 36.84%
#> TAU | Deteriorated | 4 | 10.00% | 21.05%
#> PA | Improved | 17 | 42.50% | 80.95%
#> PA | Unchanged | 4 | 10.00% | 19.05%
#> PA | Deteriorated | 0 | 0.00% | 0.00%
# Plot results by group
plot(anchor_grouped)
The plot clearly shows that more patients in the PA group (yellow
points) achieved a clinically significant improvement compared to the
TAU group (blue points).
Handling “Positive” Outcomes
Some instruments measure constructs where higher scores are better
(e.g., quality of life). For these, we must set
better_is = "higher". Let’s analyze the who
(WHO-5 Well-Being Index) outcome from the same dataset, assuming an MID
of 4.
anchor_positive <- claus_2020 |>
cs_anchor(
id = id,
time = time,
outcome = who,
pre = 1,
post = 4,
mid_improvement = 4,
better_is = "higher"
)
plot(anchor_positive)
#> Ignoring unknown labels:
#> • colour : "Group"
Now, points above the shaded area represent clinically
significant improvement.