Factors predictive of stroke outcome in a rehabilitation setting

There are approximately 600,000 new cases of stroke in the United States every year. A sizable proportion of these cases will survive with significant residual deficits. Most of these survivors will automatically be offered early rehabilitation in an effort to improve their functional abilities. This practice has come under increasing fire as the escalation of health care costs has forced greater scrutiny on the entire range of health care delivery. Although the efficacy of stroke rehabilitation now seems clear, ^[1-4] the question has become whether or not the effort is worth the cost.

Attempts to clarify this question have raised several, still unresolved issues: What is an acceptable level of functional ability? Are all types of intervention equally effective in attaining this level of function? Is a given intervention equally effective at all levels of deficit severity? Which setting is optimal for a cost-effective delivery of the rehabilitation effort? How much functional improvement is worth how many health care dollars? Changes in health care economics seem to be driving rehabilitation care rapidly to less costly treatment settings without adequate answers to the above questions. Furthermore, none of these unresolved clinical or economic questions affect the basic assumptions of rehabilitation: Improved function leads to less disability for the individual patient and ultimately to lower costs to society through less long-term supportive care and fewer poststroke complications.

One step toward defining a system of rehabilitation programs that is both cost effective and amenable to rational quality--and cost--control would be to stratify stroke survivors by easily identifiable clinical factors and determine which subgroups show predictable and acceptable improvements at different levels of care within a spectrum of care delivery options, ranging from acute rehabilitation to custodial care. This would reduce both acute- and long-term costs and afford optimum individual care in a clinically principled manner. Such a system would require a reliable way of predicting functional outcome in stroke survivors. Numerous studies (reviewed by Alexander ^[5]) have utilized different approaches:

Single-variable predictors tend to be sensitive but nonspecific, while multivariate algorithms ^[6,7] are specific but insensitive, in addition to being difficult to use. ^[8] Alexander ^[5] suggested a multi-layered approach based on constellations of single variables in stratifying poststroke care delivery. Advanced age had a very high likelihood of nursing home placement regardless of admission Functional Independence Measure (AFIM) and FIM change, while the youngest subjects tended to go home regardless of functional outcome. In the remaining cases, those with low AFIM had a high likelihood of nursing home placement, and had lower FIM change and FIM efficiency (FIM change by length of stay), while cases in a ``middle band'' of AFIM showed good FIM change and FIM efficiency. The goal of the present paper is to study individual factors more closely, to help elaborate clinically applicable rules for optimal triage of poststroke care.

Subjects and methods.

Consecutive admissions to Braintree Hospital with a primary diagnosis of stroke (infarction, hemorrhage) in the calendar year 1993 were included. Subarachnoid hemorrhages and strokes requiring cerebral surgical interventions were excluded because of their complex constellation of comorbidities. Subjects discharged to acute care hospitals for intercurrent acute events during rehabilitation or subjects who died during the course of rehabilitation were not included in the outcome analyses. Patients with remote histories of stroke but admitted for other reasons were also not included.

There were 536 admissions with primary diagnosis of cerebrovascular accidents. Eighty-six cases were not evaluated for various logistical reasons. A post hoc review of this latter group found no systematic differences from the study group. Twenty-nine cases were properly excluded and 421 patients were evaluated. Thirty-eight patients died or were transferred back to acute care hospitals, 7 cases had ``other'' dispositions, and 376 cases were included in the outcome analysis. All subjects were provided rehabilitation on general rehabilitation units in the standard, multi-disciplinary fashion.

Independent measures suspected of influencing outcome were assessed within 2 or 3 days of admission to Braintree Hospital:

(1) Age (age groups: <55, 55-64, 65-74, 75-85, >85).

(2) Severity of deficit, measured either by (a) the Functional Independence Measure (FIM, an 18-item functional measure with maximum score of 126, ^[9] AFIM groups: <40, 40-60, 61-80, >80), or by (b) the severity of hemiparesis, determined on clinical evaluation by staff neurologists and rated on a scale of 0 to 9 (Modified Brunnstrom ^[10]: 0 = no spontaneous movement, 1 = proximal leg synergies only, 2 = full leg synergies, no or early arm synergies, 3 = isolated leg movements with arm synergies, 4 = proximal leg and arm weakness only, 5 = leg weakness only, 6 = isolated proximal arm motion with distal synergies, 7 = isolated distal arm motion, 8 = hand clumsiness only, 9 = no weakness. Synergistic movement across joints arises from release of brainstem or spinal integration with the loss of higher order motor control necessary for isolated joint motion in the extremities. See Gowland ^[10] and her references for details).

(3) Lesion types were classified based on review of history, clinical examination and head imaging studies (when available) as large vessel, small vessel (lacunes), intracerebral hemorrhage, or ``unknown.'' No attempt was made to distinguish etiologies (carotid vascular, cardiac, coagulopathies, etc.).

(4) Lesion site (left/right hemisphere, brainstem, cerebellum) was determined by clinical deficit profile and neuroimaging, when available. Impact of lesion side (left, right, bilateral, or unknown) was also evaluated.

(5) Existence of comorbidities such as neurologic, cardiac, pulmonary, peripheral vascular, or arthritic disease, hypertension, diabetes mellitus, and obesity were determined. Apart from hypertension and diabetes mellitus, only comorbidities that were previously documented and functionally limiting were scored to keep the assessment conservative. To account for possible interactions of comorbidities, a weighted, aggregate sum of comorbidities was computed. We arbitrarily weighted previous strokes 3 points, other neurologic comorbidities 2, and all other defined comorbidities 1 point each. Maximum score was twelve. Patients were grouped as having none (0), few (1 to 3), or many (>or=to4) comorbidities.

(6) Incontinence of bladder and bowel and presence of dysphagia on admission were assessed singly and as an aggregate sum, each deficit contributing 1 point. Maximum score was three.

(7) Socioeconomic constraints were estimated in two ways: (a) based on the presence or absence of a committed caregiver; (b) presence of significant financial limitations for home care (usually absence of detectable insurance coverage for home services)--both determined by the admission social service assessment.

Dependent measures were determined upon discharge: (a) FIM change (discharge FIM minus admission FIM); (b) FIM efficiency (FIM change divided by length of stay); (c) disposition (to home or skilled nursing facility).

Group averages were compared (df in parentheses) using analysis of variance for continuous variables and the chi² test for discontinuous variables. All analyses were performed on a Macintosh microcomputer using Statview and Superanova software from Abacus Concepts.

Results.

(Table 1) gives a summary description of our population. Of the 421 patients evaluated, the group of acute transfers and in-house deaths (combined n = 33) had a significantly lower AFIM than the group going home (48.4 vs 69.8, p < 0.05) and a higher cumulative incontinence/dysphagia score (1.6 vs 0.68, p < 0.05). They were not significantly different in age, severity of hemiparesis, or cumulative number of comorbidities. Seven patients had ``other'' dispositions (subacute units, other rehabilitation facilities, etc.), and outcome analyses were performed on the remaining 376 subjects (326 for lesion type, site, and side comparisons). Table 2 gives group means and standard deviations for the most important factors.

(1) Age had a strong influence on FIM change across the whole population (F[1,374] = 8.70, p < 0.003), while only the 55-64 vs >85 group comparison reached significance on post hoc testing. A similar influence was found for FIM efficiency across the whole population (F[1,374] = 10.05, p = 0.002). Only the 65-74 vs 75-85 group comparisons reached significance. Age also affected disposition (chi² [4] = 14.44, p = 0.006). Ninety percent of cases <55, 77% of cases 55-75, 68% of cases 75-85, and 57% of subjects older than 85 went home. The disposition comparisons for >85 and <55 reached significance.

(2) Severity of deficit had a strong influence on all outcome measures. AFIM influenced FIM change across the population (F[1,374] = 8.39, p = 0.004) and <40 vs 40-59, <40 vs 60-80 and <40 vs >80 group comparisons were significant. FIM efficiency was also strongly influenced (F[1,374] = 132.94, p = 0.0001), with all group comparisons significant. The same was found for disposition, (chi² [3] = 81.44, p = 0.0001). AFIM >80 had a 98% rate of home discharge; 60-80, 81%; 40-59, 64%; and <40, 38%, all comparisons significant. Severity of hemiparesis influenced FIM change less strongly (F[1,367] = 3.77, p = 0.053) but FIM efficiency quite strongly (F[1,367] = 71.10, p = 0.0001).

(3) Lesion type correlated with outcome. Large-vessel strokes (LGV) did significantly worse than small-vessel strokes (SMV) or hemorrhages (ICH) (F[3,372] = 5.56, p = 0.001). As can be seen from Figure 1, this effect is due to the most impaired LGV strokes performing much worse than ICH or SMV. While ICH were significantly less impaired on admission than the other two groups (AFIM 72 +/- 16 vs 59 +/- 22 for LGV and 57 +/- 25 for SMV, F[3,417] = 10.8, p < 0.0001), there was no difference in duration of acute stay prior to rehabilitation admission (F[3,418] = 0.50) nor was FIM efficiency during rehabilitation significantly different between groups (F[3,372] = 1.06). This suggests that although ICH were less impaired on admission and improved more than ischemic strokes, they experienced their recovery more slowly. There was no difference in disposition between the groups (chi² [3] = 5.77).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. FIM change by admission FIM.

(4) Lesion site influenced FIM change with bilateral and right-sided lesions doing worse than left-sided (F[3,371] = 4.36, p < 0.005). Right-sided lesions also had less FIM efficiency than left-sided (F[3,371] = 6.74, p < 0.0005), but lesion side did not influence disposition (chi² [3] = 7.59).

Since lesion site and lesion type may have confounding influences on each other, the population was broken down according to type and side of stroke for a more careful comparison. Lesion side plus type still had significant influence on FIM change (F[12,362] = 2.68, p < 0.005). In post hoc comparisons, left-sided ICH had the most FIM change (average = 33.1), bilateral large-vessel lesions the least (average = 19.5). Right-sided lesions did worse than left-sided whether they were large-vessel or small-vessel, but right-sided hemorrhages did no worse than left-sided. FIM efficiency was also influenced by lesion side plus type (F[12,363] = 2.86, p < 0.001), right-sided large-vessel strokes did the worst (average = 0.70) and cerebellar lesions the best (average = 2.09). Large-vessel strokes had better FIM efficiency if they were on the left. Outcome for small-vessel strokes and hemorrhages was the same whichever side was affected.

(5) No single comorbidity had an isolated effect on any dependent measure. This, surprisingly, was true also for existence of prior strokes (FIM change: F[1,368] = 2.73; FIM efficiency: F[1,368] = 1.23; and discharge: chi² [1] = 0.93). Taken as an aggregate sum, patients with a higher comorbidity score had less FIM change (F[1,374] = 6.36, p < 0.05) and worse FIM efficiency (F[1,374] = 6.15, p < 0.05) but were no more likely to be discharged to nursing homes than to home (chi² [2] = 0.42).

(6) Both incontinence and dysphagia had effects on all outcome measures. Urinary incontinence on admission was strongly associated with FIM change (t[374] = 3.81, p < 0.0005) and FIM efficiency (t[374] = 6.75, p < 0.0001). However, in a two-factor ANOVA with AFIM, this effect was not apparent (F[1,368] = 0.33, p = 0.56), suggesting that incontinence reflects severity of deficit. Incontinence on discharge had a strong correlation with FIM change (t[374] = 10.76, p < 0.0001) that persisted regardless of overall functional severity (F[1,369] = 94.51, p < 0.0001). Continence on admission was associated with an 84% rate of home discharge, while incontinence on admission reduced this rate to 55%, a significant difference (chi² [1] = 38.78, p < 0.0001). Persistence of incontinence on discharge reduced the rate of home discharge to 39% versus an 82% rate for continent patients (chi² [1] = 61.17, p < 0.0001). Dysphagia alone was associated with diminished FIM change (26 vs 20, t[367] = 3.62, p < 0.0005), less FIM efficiency (0.60 vs 1.3, t[367] = 4.80, p < 0.0001, and less likelihood of home discharge (50% vs 78%, chi² [1] = 21.26, p < 0.0001).

(7) Of the socioeconomic factors considered, the absence of a committed caregiver identified on admission to rehabilitation significantly reduced the rate of home discharge from 77% to 65% (chi² [1] = 5.75, p < 0.05). Admission assessment of financial status was not reliably reported, in part because financial difficulties identified on admission were addressed to case managers and subject to ongoing remediation efforts during the course of rehabilitation.

Discussion.

The results of the current study again emphasize the influence of age ^[11-13] and degree of impairment on stroke outcome. ^[14-16] The influence of age probably reflects the impact of many comorbid factors (such as senescent brain changes, diminished physical endurance, and various medical problems), while specific medical comorbidities (including prior strokes) individually did not have the same influence. Only when comorbidities were assessed in aggregate did an effect emerge for those patients with many comorbidities, probably in a manner that parallels the impact of age. Although patients with truly devastating medical comorbidities are often not considered acceptable candidates for rehabilitation, our inclusion criteria for functionally limiting comorbidities in the population under study were still quite conservative. Thus, the presence of prior, functionally limiting medical conditions does not preclude a good outcome from neurologic rehabilitation. The influence of initial severity, as measured by the AFIM (also a comprehensive measure), was more powerful in predicting outcome than a measure of motor function alone (hemiparesis severity). When FIM efficiency (rate of improvement) was the dependent measure, hemiparesis severity was a much stronger predictive factor.

Effects of lesion type have been less well studied. ^[17,18] In our study, small-vessel infarctions (lacunes) did better than other stroke types. Patients with hemorrhages experienced more FIM change than those with large ischemic strokes but improved at a slower rate despite coming to rehabilitation with less impairment (i.e., higher AFIM and FIM change with the same FIM efficiency). A recent study by Jorgensen et al. ^[18] found ICH to be related to more severe deficits, but found no difference in outcome or rate of recovery between ICH and ischemic strokes when severity was accounted for. The discrepancy between our findings and those of Jorgensen et al. ^[18] may be due to the different populations in the two studies or to methodology. Ours was a subacute rehabilitation population; theirs was community based. Thus, the more devastating ICHs may not have come to rehabilitation in our environment. As in the other study, ^[18] we accounted for severity when comparing ICH with ischemic stroke and still found ICHs to do better. The study of Jorgensen et al. ^[18] however, did not fractionate ischemic strokes into small- or large-vessel infarcts. Given the better prognosis for SMV infarcts when compared with LGV in our study, consolidating SMV and LGV strokes may have improved the outcome for the ischemic group in their study and brought the composite outcome up to that of ICHs.

Right hemisphere lesions did more poorly than left hemisphere lesions, even across degrees of severity, a finding similar to that of an earlier report. ^[19] While it may be easier for patients with left hemisphere lesions to gain rehabilitation admission due to their aphasia and thus bias rehabilitation admissions of right hemisphere damage toward more severely impaired patients (there was a nonsignificant trend in this direction in our data), the poorer outcomes in right hemisphere damaged patients may relate to specific cognitive abilities, such as neglect and agnosia. ^[20,21]

Incontinence was strongly associated with poor outcomes and nursing home discharge as noted by others, ^[8,19,22,23] perhaps as a reflection of impairment severity rather than isolated damage to CNS micturition control. Dysphagia on admission had a similar influence, probably for the same reasons. Discharge continence correlated more strongly with disposition than did continence status on admission, probably as much due to home caregiver concerns as to severity issues. While discharge continence itself is of no value in predicting outcome, these results suggest that there may be a point in time after admission, but prior to discharge, where persistent incontinence may be a valuable predictor of final disposition.

The absence of a committed caregiver did reduce the likelihood of home discharge significantly, but even without a clear potential caregiver most patients went home (65%).

There are plausible conclusions for systems of care. Patients with early hospital improvement (our AFIM >80 group) could be managed at home directly from acute care, if appropriate outpatient services are available. Others ^[24] have also proposed this. Patients with low FIM (<40) in the early subacute phase should be treated at a lower level of intensity (skilled nursing facility with basic therapies), but monitored closely for any delayed recovery, particularly patients with hemorrhages. All other stroke survivors will probably benefit from intensive, inpatient rehabilitation. The presence of medical comorbidities should not influence rehabilitation admissions beyond the need for a medically stable condition allowing for participation in therapies, unless they are multiple and seen in the setting of sphincter dyscontrol.

The form of stratification we are suggesting contributes to answering some of the unresolved questions regarding poststroke care. The acceptable level of functional ability, and hence the improvement desired through rehabilitation, is determined in part by the improvement and disposition expected. Affording a patient a home discharge with outpatient services represents large savings vis-a-vis nursing home care. By the same token, reducing a nursing home patient's need for care, for example from maximum assistance by three caretakers 8 hours a day to maximum assistance by two caretakers 2 hours a day, also represents a significant real-dollar savings over the life of that patient and should also be considered a successful intervention even though the absolute FIM change may be small. All treatments are probably not equally effective at all impairment levels, and thus functional stratification is important in determining not only which treatments are effective but also which treatment setting is most appropriate. Keith et al., ^[25] in one of the few studies available comparing levels of care, found a clear cost inefficiency in the acute rehabilitation setting. However, they did not stratify their analysis by clinical factors as we have done, nor did they factor in incidental expenses for intercurrent medical complications, known to be lower in specialized stroke rehabilitation units than in other settings. ^[26] Our results strongly suggest a role for acute rehabilitation while arguing for a principled triage of patients into such facilities. If this can be coupled with a relaxation in CARF mandated treatment levels for acute inpatient rehabilitation, so that one is freer to tailor the treatment program to each patient's specific need, the overall cost of acute rehabilitation will be dramatically reduced, thus improving the cost/efficacy ratio vis-a-vis other ``less costly'' settings. Better insight into stroke outcome and rehabilitation triage may thus place us in a better position to determine how much of our health care dollar we are willing to spend for functional improvements in stroke survivors.