Page 181 - Let them speak Towards a patient reported outcome measure set for

Evaluating the core PROM set
Reliability and efficiency
Reliability was assessed for the core PROM set (except for the single-item
PROMs) and subscales of the legacy questionnaires (T0). For PROMIS
CATs and SFs, mean standard error (SE) was calculated across the total
sample; a mean SE ≤ 3.2, corresponding to 90% reliability, was considered
sufficient for individual measurement. For legacy subscales, we assessed
reliability by calculating Cronbach’s alpha, with values > 0.70 indicating
sufficient reliability (51). Reliability was compared between the core PROM
set and the legacy subscales using descriptive analyses.
To assess efficiency, we reported the number of items completed by adults
and caregivers for the core PROM set (except for the single-item PROMs)
and legacy subscales; for CATs, the mean, minimum, and maximum number
of items were reported. Floor and ceiling effects were assessed for the
core PROM set (except for the single-item PROMs) and legacy subscales,
defined as the proportion of participants reporting the lowest or highest
possible score. Effects were considered problematic if they exceeded the
15% threshold (52). Floor and ceiling effects were compared between the
core PROM set and the legacy subscales using descriptive analyses.
Test-retest reliability of the core PROM set was assessed by calculating the
intraclass correlation coefficient (ICC) for participants who completed the
core PROM set at T0 and T1. For the PROMIS CATs and SFs, the ICC was
calculated using a two-way random-effects model for absolute agreement
of the T-scores. An ICC ≥ 0.70 was considered sufficient (53). For the single-
item PROMs, the ICC was calculated using a two-way random-effects model
for absolute agreement of the raw scores. An ICC ≥ 0.50 was considered
sufficient, given the use of a single item and raw scores rather than T-scores.
Construct validity
To assess correlation between the core PROM set and associated legacy
subscales, Pearson correlations were calculated. A priori hypotheses
regarding expected correlations were formulated based on the content of
the legacy subscales (Table 7 – 9). Correlations were interpreted as weak
(Pearson’s r = 0.10 – 0.39), moderate (Pearson’s r = 0.40 – 0.69), or strong
(Pearson’s r = 0.70 – 0.89) (54).
7
179