Format

Send to

Choose Destination
Qual Life Res. 2019 Feb 13. doi: 10.1007/s11136-019-02130-y. [Epub ahead of print]

Patient versus proxy response on global health scales: no meaningful DIFference.

Author information

1
Quantitative Health Sciences, Lerner Research Institute, Cleveland Clinic, 9500 Euclid Avenue, JJ3-603, Cleveland, OH, 44195, USA. LapinB@ccf.org.
2
Center for Outcomes Research & Evaluation, Cerebrovascular Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA. LapinB@ccf.org.
3
Quantitative Health Sciences, Lerner Research Institute, Cleveland Clinic, 9500 Euclid Avenue, JJ3-603, Cleveland, OH, 44195, USA.
4
Center for Outcomes Research & Evaluation, Cerebrovascular Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA.

Abstract

PURPOSE:

Assessment of outcomes from a proxy is often substituted for the patient's self-report when the patient is unable or unwilling to report their status. Research has indicated that proxies over-report symptoms on the patient's behalf. This study aimed to quantify the extent of proxy-introduced bias on the Patient-Reported Outcomes Measurement Information System Global Health (PROMIS GH) scale for mental (GMH) and physical (GPH) scores.

METHODS:

This retrospective cohort study included incident stroke patients seen in a cerebrovascular clinic who completed PROMIS GH between 10/12/15 and 6/6/18. Differential item functioning (DIF) evaluated measurement invariance of patient versus proxy responses. DIF impact was assessed by comparing the initial score to the DIF-adjusted score. Subgroup analyses evaluated DIF within strata of stroke severity, measured by modified Rankin Scale (≤ 1, 2, 3+), and time since stroke (≤ 30, 31-90, > 90 days).

RESULTS:

Of 1351 stroke patients (age 60.5 ± 14.9, 45.1% female), proxy help completing PROMIS GH was required by 406 patients (30.1%). Proxies indicated significantly worse response to all items. No items for GMH or GPH were identified as having meaningful DIF. In subgroup analyses, no DIF was found by severity or 31-90 days post-stroke. In patients within 30 and > 90 days of stroke, DIF was detected for 2 items. Accounting for DIF had negligible effects on scores.

CONCLUSIONS:

Our findings revealed the overestimation of symptoms by proxies is a real difference and not the result of measurement non-invariance. PROMIS GH items do not perform differently or have spuriously inflated severity estimates when administered to proxies instead of patients.

KEYWORDS:

Differential item functioning; PROMIS GH; Patient-reported outcomes; Proxy

PMID:
30759300
DOI:
10.1007/s11136-019-02130-y

Supplemental Content

Full text links

Icon for Springer
Loading ...
Support Center