Table 1 Methods: Hypotheses, criteria and rationale used to test construct validity, criterion validity and responsiveness

Construct validity

1

At admission, there will be a large negative correlation between Self Care scores and FIM total scores, for the whole sample.

At admission there will be a moderate negative correlation between Domestic Life scores and FIM total scores for the whole sample.

Higher correlations expected between Self Care scale and FIM than between Domestic Life and FIM. Self Care scale contains more items with content directly related to the FIM than Domestic Life scale and appears to measure same construct at high levels of functioning (i.e. ‘OK by self’ on the PC-PART and scores of 6–7 on the FIM).

Magnitude of correlation coefficient (r _s )^a: r _s  ≥ .5 = large, r _s .3 to .49 = moderate, r _s .1 to .29 = small [44].

2

At admission, there will be a large negative correlation between Self Care scores and FIM total scores, irrespective of sex, age and major impairment groups

At admission, there will be a a moderate negative correlation between Domestic Life scores and FIM total scores, irrespective of sex, age and major impairment groups.

3

There will be a moderate positive correlation between admission Self Care and Charlson Comorbidity Index scores.

There will be a moderate positive correlation between admission Domestic Life scores and Charlson Comorbidity Index scores.

Patients with high co-morbidity expected to have more ADL activity limitations and more support needs than patients with low comorbidity. More support needs expected to be more difficult to satisfy, resulting in higher levels of ADL participation restriction than for those with low comorbidity.

4

On admission, there will be no observed differences in Self Care participation restriction scores between patient impairment groups.

On admission, there will be no observed differences in Domestic Life participation restriction scores between patient impairment groups.

Differences in scores between impairment groups not expected because PC-PART measurement records interactions between persons, tasks and environment. Scores not based on patients’ impairments or diagnoses.

Admission Self Care and Domestic Life Mean ± 95 % CI scores for each impairment group.

5

Self Care mean discharge scores will be lower for patients who attained their ADL goals than for patients who did not attain their ADL goals by at least one participation restriction on the Self Care scale.

Domestic Life scale mean discharge scores will be lower for patients who achieved their ADL goals than for patients who did not achieve their ADL goals by at least one participation restriction on the Domestic Life scale.

Patients’ inpatient rehabilitation ADL goals focused on optimising independence in self-care and domestic life activities of daily living and arranging zappropriate supports to enable discharge to the community. Achievement of ADL goals therefore expected to correspond to low Self Care and Domestic Life participation restriction (unmet needs) scores.

Mean difference in 1 Rasch-derived participation restriction scores: Self Care = 6.3 Domestic Life = 6.9. Differences assessed using 95 % CI mean estimates.

Criterion Validity

6

Self Care scales will discriminate between those patients discharged to ‘home or residential care’ versus patients discharged to ‘acute hospital or transitional care.’

Domestic Life scales will discriminate between those patients discharged to ‘home or residential care’ versus patients discharged to ‘acute hospital or transitional care’

‘Gold standard’ of ‘discharge destination’is the criterion for estimating the probability that Self Care and Domestic Life scale scores are an accurate reflection of discharge destination. Theoretical expectation is thatpatients discharged to community living situation (home, low- or high-level residential care) will have resolved ADL participation restrictions. Patients discharged to acute hospital or transitional care are likely to have unresolved ADL participation restrictions.

Area under the curve (AUC) range is 1.0 (perfect discrimination) to .5 (no discrimination): >.9 = high; .7 to .9 = moderate; >.5 to .69 = low; .5 = none [45]

7

Patients discharged home or to residential care will have mean scores on the discharge Self Care scale reflecting less than three ADL participation restrictions.

Patients discharged home or to residential care will have mean scores on the discharge Domestic Life scale reflecting less than three ADL participation restrictions.

Gold standard’ is ‘discharge destination’. Predicted cut-off scores reflecting three participation restrictions was a conservative, low estimate.

Rasch derived scores representing 3 ADL participation restrictions: Self Care = 25 Domestic Life = 33

8

Patients discharged to acute hospital care or transitional care will have mean scores on the discharge Self Care scale reflecting three or more ADL participation restrictions.

Patients discharged to acute hospital care or transitional care will have mean scores on the discharge Domestic Life scale reflecting three or more ADL participation restrictions.

Gold standard’ is ‘discharge destination’. Predicted cut-off scores reflecting three participation restrictions was a conservative, low estimate.

Responsive-ness

9

There will be a low to moderate negative correlation between change scores on the Self Care scale and the FIM change score across the whole sample.

There will be a low to moderate negative correlation between change scores on the Domestic Life scale and the FIM change score across the whole sample.

Self Care and Domestic Life scores expected to show greater reduction in scores than relative increase in FIM scores because patients’ ADL participation restrictions expected to be resolved at discharge to enable return to community living, reflecting PC-PART scale scores at/close to zero at discharge. Relatively small improvements in FIM scores between admission and discharge can be observed for patients discharged to community, provided adequate supports are provided.

Magnitude of correlation coefficient (r _s )^a: r _s  ≥ .5 = large, r _s .3 to .49 = moderate, r _s .1 to .29 = small [44].

10

There will be a low to moderate negative correlation between change scores on the Self Care scale and the FIM change score irrespective of sex, age and major impairment groups.

There will be a low to moderate negative correlation between change scores on the Domestic Life scale and the FIM change score irrespective of sex, age and major impairment groups.

11

The effect size observed on the Self Care and the FIM between admission and discharge will each be large, but the effect size observed on the FIM will be lower than that of the Self care scale.

The effect size observed on the Domestic Life scale and the FIM between admission and discharge will each be large, but the effect size observed on the FIM will be lower than that of the Domestic Life scale.

Effect size (ES) = (discharge mean – admission mean)/SD admission mean. Effect sizes: .2 = small; .5 = medium & .8 = large [44]

12

For patients discharged to ‘home or residential care’, there will be a large effect size on the Self Care scale.

For patients discharged to ‘home or residential care’, there will be a large effect size on the Domestic Life scale.

13

The effect size on the Self Care scale for those discharged to ‘acute hospital or transitional care’ will be small to medium.

The effect size on the Domestic Life scale for those discharged to ‘acute hospital or transitional care’ will be small to medium.