In this geographically diverse sample of over 30,000 working age acute stroke survivors, insurance status was associated with both utilization and intensity of institutional PAC. Compared with privately insured, stroke survivors with Medicaid were more likely to utilize any institutional PAC. On the contrary, uninsured stroke survivors were less likely to utilize any institutional PAC than those with private insurance. If the uninsured or stroke survivors with Medicaid utilized institutional PAC, they were more likely to utilize an SNF than an IRF compared with the privately insured. These findings are particularly important today for 2 reasons. First, an estimated 7 million Americans have had a stroke and the prevalence of stroke is increasing, resulting in a growing need for PAC services.2,20
Second, as a result of the Patient Protection and Affordable Care Act, the United States is preparing to add 16 million more Medicaid beneficiaries by 2019,the vast majority of whom will be uninsured adults without dependents.21
Our results suggest a mixed outlook for these new Medicaid enrollees. On one hand, this health care reform will likely result in increased utilization of institutional PAC among stroke survivors, but on the other hand, the rehabilitation may be less intense than if they were privately insured.
Among stroke survivors with Medicaid who utilized institutional PAC, 51% were discharged to an SNF compared to 27% of the privately insured. Reasons for the greater utilization of SNFs among stroke survivors with Medicaid are unknown. One possibility is that Medicaid reimbursement policies incentivize utilization of an SNF rather than the more expensive IRF. Based on Medicare payments, utilization of an IRF leads to an average of $24,219 more spending than if discharged home and an average of $10,121 more spending than if discharged to an SNF.13
In general, Medicaid reimbursement is lower than Medicare and private insurance, potentially creating a disincentive for IRFs to admit these stroke survivors.22
The association between reimbursement and intensity of institutional PAC has been shown among Medicare beneficiaries.23,24
A second potential explanation for the increased utilization of an SNF compared to an IRF among stroke survivors with Medicaid is the possibility that they have greater preexisting disability and are thus better candidates for utilization of an SNF.25
In our study population, stroke survivors with Medicaid had greater comorbidity as measured by the Charlson comorbidity index, but comorbidity and stroke severity measures were not confounders of the association between Medicaid and intensity of institutional PAC (data not shown). However, given our absence of a measurement of prestroke disability and crude measures of stroke severity and comorbidities due to the data available in the NIS, we cannot exclude that residual confounding by disability or severity may explain some of the association between Medicaid and intensity of institutional PAC.
Differences in utilization and intensity of PAC may partially explain the greater decline in quality of life and functional status experienced by uninsured and Medicaid stroke survivors compared with privately insured or Medicare recipients.26,27
Although no definitive trials address the impact of rehabilitation intensity on poststroke outcomes, the available evidence suggests that IRF utilization is associated with better outcomes than SNF utilization.13,28,29
A systematic review showed that organized and coordinated postacute inpatient rehabilitation care following an acute stroke is associated with decreased mortality and dependence compared to alternative forms of rehabilitation.7
Furthermore, another study found that among stroke survivors who were candidates for either an SNF or an IRF, survivors discharged to an SNF were more likely to be dead or institutionalized at 120 days.13
In our study, IRF utilization was assumed to be more intense rehabilitation than SNF based on the duration of daily rehabilitation.30
However, there was no information available about the length of stay, which may be longer in an SNF than IRF, or quality of rehabilitation that the stroke survivors received in SNFs or IRFs. More research is needed to understand the role of utilization and intensity of PAC on poststroke outcomes and on identified disparities in stroke outcomes by insurance status. Additionally, we did not explore variation in home health care or outpatient rehabilitation because the NIS dataset does not distinguish home rehabilitation from other types of home health care services or contain information about outpatient services. Further study is needed to explore the utilization of PAC among the over 75% of stroke survivors who are discharged home.
Several limitations of this study warrant discussion. The accuracy of NIS discharge destination field has not been determined. We expect that there is some misclassification of discharge location but we do not expect that this misclassification is related to insurance status.31
If true, the results would tend to be biased toward the null. Despite the high sensitivity and positive predicative value of the ICD-9-CM codes used to identify the stroke discharge, it is unknown if insurance status influences ICD-9-CM coding of ischemic stroke. If this is the case, the association between insurance status and PAC may be biased. Racial and ethnic information is not available for every patient in the NIS due to selective reporting by states. It is possible that race-ethnicity may confound the observed associations. However, we found no evidence of confounding in a sensitivity analysis limited to stroke survivors with reported race and ethnicity. Markers of stroke severity such as the NIH Stroke Scale that are important predictors of institutional PAC utilization were not available in the NIS dataset. Length of stay and tissue plasminogen activator administration were used as proxies for stroke severity although they likely account for only a moderate proportion of the variation in severity. The Charlson comorbidity index was used as a measure of comorbid illness and while it accounts for a large number of comorbidities, not all potentially relevant comorbidities are included and the severity of comorbidities is not accounted for. In addition to insurance status, other factors have been shown to influence utilization of PAC services, including proximity to PAC, family preferences and family support, discharge planning, and referral systems and practice pattern of providers, which were not available in our dataset.32,33
Finally, this study was limited to those younger than 65; PAC utilization may be different for Medicare recipients and requires further study.
Nearly 40% of working age stroke survivors are uninsured or have Medicaid. The results of this study suggest differences in utilization and intensity of institutional PAC for working age stroke survivors based on their insurance status. Compared to the privately insured, stroke survivors with Medicaid were more likely while uninsured stroke survivors were less likely to utilize any institutional PAC. Stroke survivors with Medicaid and without insurance who received institutional PAC received less intense PAC. In light of the impending significant increase in Medicaid enrollees, more research regarding the impact of utilization and intensity of PAC on identified poststroke outcome disparities, particularly among working age stroke survivors, is needed.