TABLE 2.
OASIS Validity Studies
| Citation | Type of validity | Method/Design | Sample | Items measured | Findings | Limitations/Gaps* |
|---|---|---|---|---|---|---|
| Bowles & Cater (2003) | Criterion-related Predictive |
Evaluated OASIS ability to accurately predict rehospitalization risk, compared to Probability of Readmission (Pra) instrument Secondary analysis of cross-sectional data |
147 patients age 65 and over with CHF 1 agency Convenience |
HHRG case mix weight, clinical, functional, and service scores Pra instrument Tested OASIS-B |
Pra is better at predicting rehospitalization than OASIS case mix weight (HHRG), clinical and service scores, function score was as good at predicting rehosp. as Pra |
Small & non- representative sample Limited items studied |
| Brown, Bruce, McAvay, Raue, Lachs, & Nassisi (2004) | Criterion-related Convergent |
Compared agency nurse’s ratings of OASIS depressive symptom items to SCID symptom & diagnostic status (blinded) administered by research associates Cross-sectional prospective |
220 patients 64 nurses 1 agency Random sample chosen from eligible patients |
OASIS “depressed mood” & “diminished interest in most activities” SCID symptom assess & psychologist diagnosis from symptom review Tested original OASIS |
OASIS ratings at Adm. did not accurately reflect depression in older home care pts. OASIS depression sensitivity 37.1% (13/35 cases); anhedonia 4.5% (1/22); dep.mood 33% (12/36) |
Data from 2/99–12/99 Small & non- representative sample SCID completed from 11 to 48 days after ADM. No discussion of assessor training. |
| Fortinsky, Garcia, Sheehan, Madigan, & Tullai-McGuinness (2003) | Construct validity Unidimen- sionality & item response categories |
Estimated item uni-dimensionality via principal components anal. & identified item response measurement challenges via Rasch modeling Secondary analysis, cross-sectional data |
583 patients 1 agency |
ADL & IADLs Tested original OASIS |
Response categories on bathing & telephone items should be revised; Some items may not accurately capture disability levels; Recommend Rasch modeling for summary disability scores. Unidimensionality supported by PCA |
Data collected 11/99–9/00 Non-representative sample (e.g., 96% White, 98% in traditional Medicare) Adm. only |
| Kinatukara, Rosati, & Huang (2005) | Criterion-related Convergent |
Compared OASIS to CMS485 certification & care plan form Used 1 RN hired from agency as research clinician to compare OASIS and 485 Secondary analysis |
141 patients 1 agency Random |
% of inconsistencies between tools on 13 categories of information found on both tools (e.g., functional, wound, medications, diagnoses, etc.) Tested OASIS-B |
Inconsistencies found: Functional, 47.5% Medications, 25.5% Prognosis, 17.7% Diagnosis, 14.2% Wound, 11.3% PT orders, 7.8% Pain, 7.1% Other, 5.7% Shortness of breath, 5.0% Psychosocial, 3.5% Incontinence, 2.1% Senses, 2.1% Enteral, 1.4% |
Data from 2002 Non-representative sample No psychometric testing done on CMS485 to confirm its status as a gold standard Different terms used on these tools No control for OASIS assessor Time lapse between completion of OASIS and CMS485 No weighting of inconsistencies within each category |
| Madigan & Fortinsky (2000) | Construct validity | Evaluated how closely related the items are on a specific domain area Secondary analysis from study on outcomes & resource consumption |
210 patients 10 Ohio agencies Convenience |
Functional, affect, behavioral items Principal axis factor analysis Tested original OASIS |
Functional domain items load cleanly & strongly onto 1 factor for adm. & DC Behavioral & Affect domains should be revised or treated as individual items |
Non-rep. sample Data from 1996 No control for discipline of the assessor No discussion of staff training on OASIS |
| Madigan, Tullai-McGuinness, & Fortinsky (2003) | Criterion-related Convergent |
Evaluated the accuracy of OASIS completion by agency nurses & therapists to expert derived answers Exploratory simulation |
436 clinicians 337 RNs 68 PTs 21 OTs 10 STs Sampling method not reported 29 Ohio agencies |
I/ADLs, clinical items, and behavioral items Percent accuracy to expert answers Tested original OASIS |
Clinician responses were similar to the “correct” answer for a majority of OASIS items, with more accuracy shown at discharge (76% with >80% accuracy) than at admission (58% with >80% accuracy) |
Data from 11/99–9/00 Non-rep. sample No validity tests on expert opinion Not in usual care context No controls for agency factors |
| Tullai-McGuinness, Madigan, & Fortinsky (2009) | Criterion-related Convergent |
Examined criterion validity of key OASIS items to “gold standard” tools Agency staff CRN, PT) completed the OASIS Research RNs completed the gold standard measures within five business days Prospective cross-sectional |
203 patients 188 RNs 14 PTs 5 Ohio agencies Convenience |
OASIS: I/ADLs, cognitive functioning, & depression Gold standards: I/ADLs of the OARS Instrument, SPMSQ, CES-D Scale, & BSI Pearson’s Correlation Tested OASIS-B |
ADL item correlations with OARS from .44 to .69; .71 overall IADL from .20 to .68; .49 overall Cognitive with SPMSQ .62 Depressive symptoms .36 (BSI) and .26 (CES-D) OASIS is valid for ADLs & cognition, but may not be sufficiently sensitive for depressive symptoms & IADL items |
Data from larger study from 12/99–3/02 but no specific dates reported for this analysis Non-rep. sample Gold standards conducted between 3 & 7 days post OASIS No interrater reliability check on gold standard measures Data collectors received structured orientation & reference manual. No controls for agency factors |
*
Data are listed as reported in original articles.