- Email: [email protected]
Utilization of Rehabilitation Services in Stroke: A Study Utilizing the Health and Retirement Study With Linked Medicare Claims Data
Archives of Physical Medicine and Rehabilitation journal homepage: www.archives-pmr.org Archives of Physical Medicine and Rehabilitation 2019;-:-------
ORIGINAL RESEARCH
Utilization of Rehabilitation Services in Stroke: A Study Utilizing the Health and Retirement Study With Linked Medicare Claims Data Karen M. Keptner, PhD, OTR/L, Kathleen Smyth, PhD, Siran Koroukian, PhD, Mark Schluchter, PhD, Anthony Furlan, MD From the School of Medicine, Case Western Reserve University, Cleveland, OH. Current affiliation for Keptner, Cleveland State University, Cleveland, Ohio
Abstract Objectives: To describe Medicare fee-for-service beneficiaries who used poststroke rehabilitation services and identified the strongest predictors of utilization after the initial stroke care episode. Design: Pooled, cross-sectional design using data from 1998 to 2010 from the Health and Retirement Study (HRS) with linked Medicare claims data. Setting: NA. Participants: Stroke survivors who were Medicare fee-for-service beneficiaries and participated in the HRS were included (NZ515). Main Outcome Measure: Utilization of rehabilitation services up to 10 years poststroke was the primary outcome with logistic regression used to predict utilization. Covariates included demographic factors, baseline functional status, health conditions, personal lifestyle factors, and social support. Results: Rehabilitation service utilization was 21.6%, 6.8%, 15.8%, 16.5%, and <16% in years 2, 4, 6, 8, and 10, respectively. Age was the primary factor predicting use of rehabilitation in the first 10 years poststroke (odds ratio: 1.14; PZ.001). Recurrent stroke (odds ratio: 1.64; PZ.051) was also significantly associated with utilization, whereas unspecified incident stroke at incident trended toward significance (odds ratio: 2.17; PZ.077). None of the other factors was a significant predictor of participation in rehabilitation services in this period. Conclusion: A small number of Medicare fee-for-service beneficiaries who are stroke survivors utilize rehabilitation services in the first 10 years poststroke. Of those who do, age is the primary driver of utilization. We analyzed a multitude of factors that might influence utilization, but other factors not available in these data also need to be explored. Archives of Physical Medicine and Rehabilitation 2019;-:------ª 2019 by the American Congress of Rehabilitation Medicine
In the United States, approximately 795,000 strokes occur every year.1 Although mortality rates have declined due to medical advances, more individuals have acquired stroke-related disability.2-4 To reduce disability, rehabilitation has been recommended.5-9 Most stroke survivors participate in inpatient rehabilitation, but many individuals do not continue to participate in rehabilitation after discharge.10-12 Whether or not stroke survivors utilize Supported in part by the Clinical and Translational Science Collaborative (CTSC) of Cleveland which is funded by the National Institutes of Health (NIH), National Center for Advancing Translational Science (NCATS), and Clinical and Translational Science Award (CTSA) grant, UL1TR002548. The content is solely the responsibility of the authors and do not necessarily represent the official views of the NIH. Disclosures: none.
therapy services depends on need, referral, and ability to follow-through with referrals.9,11 Most studies on utilization patterns in stroke focus on the period immediately poststroke11,13-15 and indicate that utilization can be influenced by demographic and personal characteristics (such as socioeconomic status and sex), comorbid conditions (such as number of comorbidities, severity of stroke, and presence of a new stroke), and lifestyle factors (such as physical activity behaviors, relation with a primary care provider, and social support network).16,17 Utilization of services over a longer timeframe has yet to be discussed in the literature. With that in mind, this study aimed to (1) examine utilization patterns among stroke survivors who are Medicare fee-for-service beneficiaries and (2) describe factors
0003-9993/19/$36 - see front matter ª 2019 by the American Congress of Rehabilitation Medicine https://doi.org/10.1016/j.apmr.2019.06.017
2
K.M. Keptner et al
associated with utilization of rehabilitation services up to 10 years poststroke. The researchers believed that determinants of utilization found in earlier phases of stroke recovery would also be found in later stages (>1y poststroke). In particular, the researchers believed that those with additional comorbidities, increased severity of initial stroke, presence of a new stroke, higher socioeconomic status, history of regular physician visits, higher baseline physical activity, and larger social support networks would be more likely to use rehabilitation services.
Methods This study used a pooled, cross-sectional approach to describe utilization patterns for rehabilitation services and identify the predictors of utilization by individuals who sustained a stroke. We used data from individuals who participated in the Health and Retirement Study (HRS)18 during the years 1998-2010 and also reported a stroke. The study was approved by the Case Western Reserve University Institutional Review Board (IRB-2012-215 and IRB-2013-55). A Data User Agreement with Centers for Medicare and Medicaid was also secured.
Databases The HRS is a longitudinal study developed at the University of Michigan.18 It was initiated in 1992 and focused on the health, economic status, and retirement of older adults in the United States. The HRS sample included more than 26,000 Americans older than 50 years who were surveyed biannually. The linked-Medicare file includes claims information for feefor-service beneficiaries including claims from inpatient hospital settings as well as outpatient institutional and noninstitutional settings. The linked-Medicare claims records carry a unique dummy identifier for each respondent that is linked to the HRS data through a file supplied by the HRS staff.
Participants The linked HRS-Medicare file from 1998 to 2010 was used to create the sample and define utilization. Stroke incidence between 2000 and 2008 was identified in the HRS with the participant stating that a doctor told them that they had a stroke. This was confirmed through the presence of diagnosis codes documented in Medicare claims data.19 Only those who were Medicare fee-forservice recipients and also participated in the HRS after the first incident stroke were included in the analyses. Those who reported a stroke prior to 2000 and those who had a proxy respondent at baseline were excluded. The sample was followed through the 2010 wave (fig 1). Power and sample size were determined by the number of eligible cases in the HRS Medicare file data. The 995 eligible cases provided 90% power for a 2-sided test to detect a difference in proportions seeking therapy services (of 0.15-0.24) between the 2 groups based on a binary factor with the proportion in 1 group
List of abbreviations: BMI FPL HRS SES
body mass index federal poverty level Health and Retirement Study socioeconomic status
ranging from 0.3 to 0.5. This assumed a significance level of 0.05, and a design effect of 1.1 assuming an average of 2 waves per participant and an intraclass correlation of 0.1.
Variables of interest A modified version of the World Health Organization’s International Classification of Functioning, Disability and Health was used as the framework for this study because it allowed us to look at both stroke and those personal and environmental factors that influence utilization of services.20,21 A core set of variables specific to stroke, developed by a panel of stroke researchers, was used as a guide for the choice of variables in this study.20 Rehabilitation Respondents were asked whether or not they were participating in rehabilitation services at the time of the HRS survey (yes/no). Presence of rehabilitation services between waves was confirmed in Medicare claims through a search for CPT, Healthcare Common Procedure Coding System, and Revenue Center Codes (supplemental table 1, available online only at http://www. archives-pmr.org/). Demographic characteristics Demographic characteristics were taken from the HRS-RAND Tracker File.19 Age was a continuous variable and was calculated from date of interview and date of birth. Sex (male or female), race (white or black), and ethnicity (Hispanic or Latino) were selfreported in the survey. Ethnicity was dichotomized to identify those who were or were not of Hispanic or Latino origin. Socioeconomic status (SES) was measured as a percentage of selfreported assets and income as a ratio to the federal poverty level (FPL). The ratio of household income to FPL was dichotomized into 200% FPL and <200% FPL. Self-reported education was dichotomized to distinguish those who had completed at least high school or general equivalency diploma from those who had not. Comorbidity This study used 6 of the conditions that were self-reported by respondents in the HRS: high blood pressure, diabetes, lung disease, heart disease, stroke, and arthritis. Comorbidity was measured as a count of comorbidities and analyzed as a continuous variable. Use of self-reported count for comorbidity in this way has been validated in the literature22 and during the development of the HRS. Type of stroke Type of stroke was self-reported in the HRS and confirmed by Medicare claims data. Type of stroke was categorized as hemorrhagic, ischemic, or unspecified if the stroke was undefined in the claims or there was discrepancy in the coding within the claims data.19 Time elapsed since stroke Time since stroke was calculated as the time between each interview and the date of first reported stroke from the HRS. It was recalculated at each wave. Presence of a new stroke At each wave, each individual was asked whether they were diagnosed with a new “stroke” (yes/no). To confirm the presence of an incident stroke, Medicare claims data were searched for www.archives-pmr.org
Utilization of rehabilitation in stroke
3
300 250 200 150 100 50 0 1 wave (n=263) Fig 1
2 waves (n=135)
3 waves (n=53)
5 waves (n=47)
Number of survey respondents by number of waves included in the final regression analysis.
stroke International Classification of Diseasese9th Revision codes with a subsequent inpatient hospital admission, after accounting for a 90-day washout period after the initial stroke. Initial severity of stroke Number of days spent in inpatient ICU care was calculated from Medicare claims data as an additional measure of severity.23 Functional ability The RAND HRS data used factors identified by Wallace and Herzog to create indices of function for use by researchers.24 The RAND index score of functional ability chosen for this study was calculated from 4 subscales within the HRS: strength limitations, upper body mobility limitations, lower body mobility limitations, and activities of daily living. Function was measured on a continuous scale from 0 to 19. The index score ranged from independence (0) to dependence (19) in the above areas. For each wave, the function score was recalculated as a change in function scores from wave to wave so that functional ability could be described. Personal lifestyle factors Personal lifestyle factors that may influence outcomes in stroke include presence of a regular physician and smoking, drinking, and physical activity behaviors.17,25 Whether or not someone had a regular physician was determined by whether the respondent reported a visit to physician over the previous 12 months (yes/no). Smoking was a binary variable (currently smoking vs past smoker or nonsmoker).26 Drinking behavior was categorized as nondrinker, mild or moderate drinker (2 alcoholic drinks/day), or heavy drinker (3 alcoholic drinks/day).26 Physical activity was determined by whether the respondent self-reported vigorous physical activity at least 3 or more days per week (yes/no).26 Baseline working status Working status was used as a baseline characteristic. It was dichotomized to indicate respondents who were doing any work for pay versus those who were not at baseline. Body mass index Body mass index (BMI) was calculated from self-reported height and weight within the HRS and was used as a continuous variable in descriptive analysis. For multivariate analyses, BMI was a categorical variable divided into those who were underweight (BMI<18.5), normal (18.5BMI24.99), overweight (25BMI29.99), and obese (BMI30).
www.archives-pmr.org
4 waves (n=17)
Cognitive deficits The HRS contains an embedded cognitive assessment (the Telephone Interview Cognitive Score). The score was used as a continuous measure in analyses for those who were selfrespondents to the survey after their stroke. Cognition scores ranged from 0 to 35, with a higher score indicating better cognitive abilities. Presence of depressive symptoms The HRS uses a short form of the Center for Epidemiologic Studies Depression scale. The scores range from 0 (no depressive symptoms) to 8 (the highest number of depressive symptoms). The scores were used as a continuous variable. Social support Social support was derived as described by Muramatsu et al as family resources.27 It was taken from marital status, number of living siblings, child availability, and spouse’s self-reported health status. Marital status was coded as currently married or not married, including those who were never married. Child availability was categorical (coresidence, children living within 10 miles, child living more than 10 miles away, no living children or never had children). Spouse’s self-reported health status was dichotomized (fair or poor vs good or better). The categories were used to create an index of social support (0-4). A score of zero indicated lack of any family resources. A score of 4 indicated that the respondent was married with a spouse in good or better health, had children within 10 miles, and had at least 1 living sibling.
Analysis To describe the sample and utilization of services, descriptive analysis of all factors was conducted. Utilization of rehabilitation up to 10 years poststroke was calculated as a percentage of the sample. Univariable analyses were conducted on both the categorical and continuous variables. Constant factors in the analysis were sex, race, SES, stroke type, initial severity of stroke, having visited a physician in the previous 12 months, and baseline measures (employment status, comorbidity, BMI, smoking, drinking, and physical activity levels). Time-varying factors were age, comorbidity, cognitive score, depression score, social support, receipt of rehabilitation, and recurrent stroke. Comorbidity was used as both a static, baseline factor and as a time-varying factor. A multivariable logistic regression with stepwise selection was used to determine the effect of the constant and time-varying
4 factors on utilization of therapy services. Akaike Information Criterion was used to establish a P value cutoff for variable selection into the final model (P<.15) which allowed us to increase power and provide better predictions in the final results.28 Generalized estimating equations (GEEs) were used because we used a large panel database and generalized estimating equation adjusts for natural correlation within the data. In large databases, especially in survey data, it is likely that missing values influence the results of the final analysis. In our sample, 10%-15% of each wave was lost to follow-up. To address the effect missing values might have had on the outcomes, all analyses were done first with missing values and then with imputed missing values using the PROC MI function in SAS. After running the analysis in both manners, no change in statistical significance was noted; thus, the final multivariable logistic regression results are reported without imputation. SAS version 9.2 for UNIX was used for all analysis.a
Results Of the 23,695 records from the linked HRS-Medicare file, 1189 stroke survivors were identified between 2000 and 2008. Of those, 515 stroke survivors with 995 observations comprised the study sample. Utilization rates would have been influenced by loss to follow-up, which was 10%-15% at each wave.
Descriptive analysis of HRS-Medicare sample Most of the sample were women (56.7%), white (81%), had income more than 200% of the FPL (64.6%), and had at least a high school diploma (table 1). The average age for the sample was 73.7 years and most were retired (67.0%). The sample had characteristics consistent with other findings1,2: the average number of comorbidities was 2.4, most had an ischemic stroke (63.3%), and also reported a second stroke (66.4%). In addition, the average depression score SD was 2.22.2, indicating that the sample, on average, reported at least 1 or 2 depressive symptoms. The average cognitive score SD among the sample in the survey poststroke was 19.85.4. Social support varied among the sample with 18.7% of the sample having index scores of 0 or 1 (indicative of the least amount of social support) and 31.4% having scores of 2 (out of a maximum of 4 in social support). Most of the respondents (95.7%) had a physician visit during the preceding 12 months, 12.6% were current smokers, and 4.1% were heavy drinkers. Those who participated in regular physical activity comprised 29.5% of the sample.
Utilization of rehabilitation At the first interview post stroke, only 21.6% of stroke survivors reported receipt of rehabilitation services (table 2).
Logistic regression on participation in rehabilitation over time Univariable logistic regression analyses were conducted on all variables (table 3). Age, initial severity, functional ability, second stroke, cognitive scores, and personal lifestyle factors (physician visit and smoking status) predicted participation in therapy services.
K.M. Keptner et al Table 1 Demographic characteristics of HRS-Medicare sample (1998-2008) (NZ515 individuals) Descriptive Variables
n (%)
Sex Male Age (y), mean SD Race White Black Ethnicity Hispanic/Latino Not Hispanic/Latino Socioeconomic status* <200% FPL 200% FPL Education
223 (43.30) 73.678.36 422 (81.94) 93 (18.05) 36 (6.99) 479 (93.01) 163 (31.65) 298 (64.64) 179 (34.76) 336 (65.24) 41 (7.96) 16 (3.11) y
42 (8.16) 345 (66.99) y
53 (10.29) 27.044.98 326 (63.30) 50 (9.71) 139 (26.99) 342 (66.4) 3.61.8 2.391.39 463 (95.66) 54 (12.59) 21 (4.08) 152 (29.51) 19.85.4 2.22.2 96 (18.7) 162 (31.4) 117 (22.7) 140 (27.2) 31.268.7
Abbreviations: HS, high school; MD, medical doctor. * Calculation influenced by missing values. y Category counts hidden, per Institutional Review Board and user data agreement with Centers for Medicare and Medicaid Services to limit reporting for categories that include <11 participants; additional data shielded to prevent identification of participants in the HRSMedicare database.
Table 4 presents the final logistic regression model. Of the factors that met the cutoff during stepwise selection, age (P<.001) and having a second stroke (PZ.051) were significant predictors, while having an undefined stroke (PZ.077) trended toward statistical significance as a predictor of participation in rehabilitation. www.archives-pmr.org
Utilization of rehabilitation in stroke Table 2
5
Self-reported participation in rehabilitation by HRS survey year (NZ515 individuals)
Response
0-2 y (NZ515)
2-4 y (nZ268)
4-6 y (nZ152)
6-8 y (nZ91)
8-10 y (nZ53)
Yes No
111 (21.6) 404 (78.6)
35 (6.8) 233 (86.9)
24 (15.8) 128 (84.2)
15 (16.5) 76 (83.5)
<11 (20.7) >42 (>79.2)
NOTE. Count hidden, per Institutional Review Board and user data agreement with Centers for Medicare and Medicaid Services to limit reporting for categories that include <11 participants; additional data shielded to prevent identification of participants in the HRS-Medicare database.
With each year of increased age, participants were 1.14 times more likely to participate in rehabilitation (PZ.001). Those who had a second stroke were 1.64 times more likely to participate (PZ.051).
Discussion Medicare claims data with linked HRS data allowed us to look at multiple demographic, health, and personal level factors that Medicare claims data alone do not provide. Linking the data allowed us to explore not only who used services, but if there was any predictive value in obtaining data on self-reported health conditions and lifestyle factors, which are in line with recent efforts to understand the utilization of rehabilitation poststroke.29 Many of our hypotheses were not supported by the data; having additional comorbidities, increased severity of initial stroke, higher SES, history of regular physician visits, regular physical activity, and a larger social support network were not significant predictors of rehabilitation utilization. Table 3 Univariable logistic regression of use of rehabilitation services on individual factors: demographics, health condition, and personal lifestyle factors (NZ995 observations) Descriptive Variables
OR
P Value
Sex, female Age Race (white vs black/other) Income, % FPL (200% FPL) Education (HS grad or GED) Employment at baseline (working) Comorbidity Stroke type ischemic Hemorrhagic Unspecified Initial severity Functional ability Second stroke BMI (continuous) Cognition score Depressive symptoms Presence of social support Personal lifestyle factors: MD visit within year (yes/no) Smoker (yes/no) Heavy drinker (yes/no) Physical activity (yes/no) Time elapsed
1.06 1.09 1.15 1.13 1.41 1.86 1.15 Ref 0.58 3.07 0.92 1.12 1.99 1.06 0.93 1.06 0.80 0.26
.85 .001* .71 .80 .29 .06 .23 —— .38 .01y .020z .001* .0007* .84 .015z .36 .11 .007y
2.74 2.29 1.14 1.00
.009y .16 .69 .96
Abbreviation: GED, general equivalency diploma; HS, high school; MD, medical doctor; OR, odds ratio. * P.001. y P.01. z P.05.
www.archives-pmr.org
Most of the sample did not report receiving therapy when asked during a given survey. This was well below previous estimates.11 The highest utilization period in this study was within 2 years poststroke, with just more than 20% of the sample stating that they received services in the previous 2 years. As time passed, utilization decreased before it increased again. The change in utilization patterns may be reflective of the recovery process and aging. After the initial episodes of care, survivors tend to adjust to disability3 and may not need additional rehabilitation services. The increase in services after a decrease might be indicative of increased morbidity, injury, or illness that necessitated services. It may be, however, that stroke survivors needed additional services but were not able to access them. Finances, including the ability to pay deductibles and copayments, may have limited access to service independent of functional status or need for services. Recent studies suggest that stroke survivors in the community do not receive the same level of formal care as individuals with other neurologic conditions,30 and a low level of therapy services as found in this study may confirm this conclusion. The day-to-day formal care, including therapy that is needed because of the stroke, may be limited by provider recognition of impairments that indicate a referral for therapy services. Providers may not refer for services if a patient does not request them or if a patient does not report a specific concern that would necessitate a referral. Further, therapists themselves may not be knowledgeable about reassessing stroke patients. Therapists may believe that patients have reached a plateau in progress, when in fact they have not. It may be difficult to find therapists in a local area who have the skill to determine the therapy needs of a stroke patient, further limiting access to appropriate services. The primary drivers of utilization of therapy in this study were increased age and incident stroke, which are confirmed by others.31,32 Older individuals and those with a recurrent stroke may have an increase in health care encounters due to functional Table 4 Logistic regression with stepwise selection of the presence of rehabilitation on significant univariable factors (age, second stroke, stroke type, functional ability, cognition) (NZ995 observations) Variables
OR
95% CI
P Value
Age Second stroke Stroke type Ischemic Hemorrhagic Unspecified Functional ability Cognition
1.14 1.64
(1.08-1.22) (1.00-2.70)
.001* .051y
Reference 0.76 2.17 1.07 0.99
— (0.10-5.70) (0.21-0.95) (0.98-1.18) (0.94-1.06)
— .788 .077 .101 .932
Abbreviations: 95% CI, 95% confidence interval; OR, odds ratio. * P.001. y P.05.
6 concerns that increase the likelihood that a referral to rehabilitation services is generated. This is consistent with other studies that find that utilization of services is determined by the health-related needs of patients.33
Study limitations Because this research was based on secondary data, it was not possible to explore all the issues that can effect utilization. Changes between 2 waves, not captured by these data, might have influenced the findings as would the use of self-reported data. The data may have been biased from participants under- or overreporting lifestyle factors that were used in the final analyses. In addition, because we chose to use self-report of comorbidity instead of using Medicare claims for comorbidity data, comorbidity measures may also have some bias. Finally, this sample was older than the original HRS sample after the addition of Medicare claims, because individuals older than the age of 65 are the primary recipients of Medicare benefits. Using Medicare fee-for-service claims also limited the conclusions because utilization in newer payment structures that have arisen over recent years might yield different results. However, because we used a large survey database in concurrence with Medicare claims, we believe that the results have some generalizability among an older, Medicare fee-for-service population. Despite these limitations, the methods allowed us to observe a multitude of personal factors that might influence utilization of rehabilitation up to 10 years poststroke, which is unique. Finally, the study looked at rehabilitation broadly, including both occupational therapy and physical therapy, with no distinction between the 2 services. This could have biased the conclusions because some patients may have only needed 1 service or the other, but this was not captured in this study.
Conclusions This study demonstrated that age and a new, incident stroke are predictors of utilization of rehabilitation services beyond 1-year poststroke, but that utilization of rehabilitation was lower than previous estimates. Future research should explore the reasons for such low rates of utilization of rehabilitation, whether rehabilitation services are needed by most of the individuals who sustain a stroke, and what, if any, barriers are present for individuals to participate in rehabilitation services.
Supplier a. SAS version 9.2 for UNIX; SAS Institute Inc.
Keywords Health services; Occupational therapists; Physical therapists; Rehabilitation; Stroke
Corresponding author Karen M. Keptner, PhD, OTR/L, Cleveland State University, 2121 Euclid Avenue HS 101, Cleveland, OH 44115. E-mail address: [email protected].
K.M. Keptner et al
References 1. Benjamin EJ, Blaha MJ, Chiuve SE, et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statisticsd2017 update: a report from the American Heart Association. Circulation 2017;135:e229-445. 2. Schwamm L, Fayad P, Acker J, et al. Translating evidence into practice: a decade of efforts by the American Heart Association/ American Stroke Association to reduce death and disability due to stroke: a presidential advisory from the American Heart Association/American Stroke Association. Stroke 2010;41:1051. 3. World Health Organization. Global health observatory. Available at: http://www.who.int/gho/ncd/mortality_morbidity/cvd_text/en/. Accessed April 1, 2014. 4. Desrosiers J, Noreau L, Rochette A, Bravo G, Boutin C. Predictors of handicap situations following post-stroke rehabilitation. Disabil Rehabil 2002;24:774. 5. National Stroke Foundation. Organization of services. In: Clinical guidelines for stroke management 2010. Available at: http://www.guideline.gov/ content.aspx?idZ24168&searchZcva. Accessed July 27, 2014. 6. Outpatient Service Trialists. Rehabilitation therapy services for stroke patients living at home: systematic review of randomised trials. Lancet 2004;363:352-6. 7. Winstein CJ, Stein J, Arena R, et al. Guidelines for adult stroke rehabilitation and recovery: a guideline for health care professionals from the American Heart Association/American Stroke Association. Stroke 2016;47:e98-169. 8. Miller E, Murray L, Richards L, et al. Comprehensive overview of nursing and interdisciplinary rehabilitation care of the stroke patient: a scientific statement from the American Heart Association. Stroke 2010;41:2402. 9. Freburger JK, Li D, Johnson AM, Fraher EP. Physical and occupational therapy from the acute care to the community setting after stroke: predictors of use, continuity of care, and timeliness of care. Arch Phys Med Rehabil 2017;99:1077-89. 10. Xie J, George M, Ayala C, McGruder H, Denny C, Croft J. Outpatient rehabilitation among stroke survivors e 21 states and the District of Columbia. MMWR Morb Mortal Wkly Rep 2007;56:504. 11. Kuo Y, Ostir G, Granger C, Ottenbacher K. Examination of follow-up therapy in patients with stroke. Am J Phys Med Rehabil 2006;85: 192-200. 12. Freburger JK, Konrad TR. The use of federal and state databases to conduct health services research related to physical and occupational therapy. Arch Phys Med Rehabil 2002;83:837-45. 13. Gadidi V, Katz-Leurer M, Carmeli E, Bornstein N. Long-term outcome poststroke: predictors of activity limitation and participation restriction. Arch Phys Med Rehabil 2011;92:1802-8. 14. Andrews AW, Li D, Freburger JK. Association of rehabilitation intensity for hospital readmission. Phys Ther 2015;95:1660-7. 15. Cruz-Flores S, Rabinstein A, Biller J, et al. Racial-ethnic disparities in stroke care: the American experience: a statement for health professionals from the American Heart Association/American Stroke Association. Stroke 2011;42:2091-116. 16. Lee Y, Kyung-Park M. Health practices that predict recovery from functional limitations in older adults. Am J Prev Med 2006;31:25. 17. Mold F, McKevitt C, Wolfe C. A review and commentary of the social factors which influence stroke care: issues of inequality in qualitative literature. Health Soc Care Community 2003;11:405-14. 18. Health and Retirement Study, (RAND enhanced Fat Files, 1998-2010, RAND v. M, and Tracker File v.1.0) public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009740) Ann Arbor, MI, (2013). 19. Andrade SE, Harrold LR, Tija J, et al. A systematic review of methods identifying cerebrovascular accident or transient ischemic attack using administrative data. Pharmacoepidemiol Drug Saf 2012;21(Suppl 1): 100-28.
www.archives-pmr.org
Utilization of rehabilitation in stroke 20. Geyh S, Cieza A, Schouten J, et al. ICF core sets for stroke. J Rehabil Med 2004;(44 Suppl):135-41. 21. World Health Organization. International classification of functioning, disability and health (ICF). Geneva, Switzerland: World Health Organization; 2001. 22. Huntley AL, Johnson R, Purdy S, Valderas JM, Salisbury C. Measures of multimorbidity and morbidity burden for use in primary care and community settings: a systematic review and guide. Ann Fam Med 2012;10:134-41. 23. Williams TA, Ho KM, Dobb GJ, Finn JC, Knuiman M, Webb SA. Effect of length of stay in intensive care unit on hospital and long-term mortality of critically ill adult patients. Br J Anaesth 2010;104:459-64. 24. Wallace RB, Herzog AR. Overview of the health measures in the Health and Retirement Study. J Hum Resour 1995:S84-107. 25. Wray L, Alwin D, McCammon R. Social status and risky health behaviors: results from the Health and Retirement Study. J Gerontol B Psychol Sci Soc Sci 2005;60(S2):85-92. 26. Rahrig Jenkins K, Ofstedal MB, Weir D. Documentation of health behaviors and risk factors measured in the Health and Retirement Study (HRS/AHEAD). Ann Arbor, MI: Survey Research Center, University of Michigan; 2008.
www.archives-pmr.org
7 27. Muramatsu N, Yin H, Campbell R, Hoyem R, Jacob M, Ross C. Risk of nursing home admission among older Americans: does states’ spending on home- and community-based services matter? J Gerontol B Psychol Sci Soc Sci 2007;62:S169-78. 28. Steyerberg E. Clinical prediction models. New York: Springer-Verlag; 2009. 29. Nelson ML, McKellar KA, Munce S, et al. Addressing the evidence gap in stroke rehabilitation for complex patients: a preliminary research agenda. Arch Phys Med Rehabil 2018;99:1232-41. 30. Obembe AO, Goldsmith CH, Simpson LA, Sakakibara BM, Eng JJ. Support service utilization and out-of-pocket payments for health services in a population-based sample of adults with neurological conditions. PLoS One 2018;13:e0192911. 31. Cook C, Stickley L, Ramey K, Knotts V. Variables associated with occupational and physical therapy stroke rehabilitation utilization and outcomes. J Allied Health 2005;34:3-10. 32. Pettersen R, Dahl T, Wyller T. Prediction of long-term functional outcome after stroke rehabilitation. Clin Rehabil 2002;16:149-59. 33. Jourdan C, Bayen E, Darnoux E, et al. Patterns of post-acute health care utilization after a severe traumatic brain injury: results from the PariS-TBI cohort. Brain Inj 2015;29:701-8.
ORIGINAL RESEARCH
Utilization of Rehabilitation Services in Stroke: A Study Utilizing the Health and Retirement Study With Linked Medicare Claims Data Karen M. Keptner, PhD, OTR/L, Kathleen Smyth, PhD, Siran Koroukian, PhD, Mark Schluchter, PhD, Anthony Furlan, MD From the School of Medicine, Case Western Reserve University, Cleveland, OH. Current affiliation for Keptner, Cleveland State University, Cleveland, Ohio
Abstract Objectives: To describe Medicare fee-for-service beneficiaries who used poststroke rehabilitation services and identified the strongest predictors of utilization after the initial stroke care episode. Design: Pooled, cross-sectional design using data from 1998 to 2010 from the Health and Retirement Study (HRS) with linked Medicare claims data. Setting: NA. Participants: Stroke survivors who were Medicare fee-for-service beneficiaries and participated in the HRS were included (NZ515). Main Outcome Measure: Utilization of rehabilitation services up to 10 years poststroke was the primary outcome with logistic regression used to predict utilization. Covariates included demographic factors, baseline functional status, health conditions, personal lifestyle factors, and social support. Results: Rehabilitation service utilization was 21.6%, 6.8%, 15.8%, 16.5%, and <16% in years 2, 4, 6, 8, and 10, respectively. Age was the primary factor predicting use of rehabilitation in the first 10 years poststroke (odds ratio: 1.14; PZ.001). Recurrent stroke (odds ratio: 1.64; PZ.051) was also significantly associated with utilization, whereas unspecified incident stroke at incident trended toward significance (odds ratio: 2.17; PZ.077). None of the other factors was a significant predictor of participation in rehabilitation services in this period. Conclusion: A small number of Medicare fee-for-service beneficiaries who are stroke survivors utilize rehabilitation services in the first 10 years poststroke. Of those who do, age is the primary driver of utilization. We analyzed a multitude of factors that might influence utilization, but other factors not available in these data also need to be explored. Archives of Physical Medicine and Rehabilitation 2019;-:------ª 2019 by the American Congress of Rehabilitation Medicine
In the United States, approximately 795,000 strokes occur every year.1 Although mortality rates have declined due to medical advances, more individuals have acquired stroke-related disability.2-4 To reduce disability, rehabilitation has been recommended.5-9 Most stroke survivors participate in inpatient rehabilitation, but many individuals do not continue to participate in rehabilitation after discharge.10-12 Whether or not stroke survivors utilize Supported in part by the Clinical and Translational Science Collaborative (CTSC) of Cleveland which is funded by the National Institutes of Health (NIH), National Center for Advancing Translational Science (NCATS), and Clinical and Translational Science Award (CTSA) grant, UL1TR002548. The content is solely the responsibility of the authors and do not necessarily represent the official views of the NIH. Disclosures: none.
therapy services depends on need, referral, and ability to follow-through with referrals.9,11 Most studies on utilization patterns in stroke focus on the period immediately poststroke11,13-15 and indicate that utilization can be influenced by demographic and personal characteristics (such as socioeconomic status and sex), comorbid conditions (such as number of comorbidities, severity of stroke, and presence of a new stroke), and lifestyle factors (such as physical activity behaviors, relation with a primary care provider, and social support network).16,17 Utilization of services over a longer timeframe has yet to be discussed in the literature. With that in mind, this study aimed to (1) examine utilization patterns among stroke survivors who are Medicare fee-for-service beneficiaries and (2) describe factors
0003-9993/19/$36 - see front matter ª 2019 by the American Congress of Rehabilitation Medicine https://doi.org/10.1016/j.apmr.2019.06.017
2
K.M. Keptner et al
associated with utilization of rehabilitation services up to 10 years poststroke. The researchers believed that determinants of utilization found in earlier phases of stroke recovery would also be found in later stages (>1y poststroke). In particular, the researchers believed that those with additional comorbidities, increased severity of initial stroke, presence of a new stroke, higher socioeconomic status, history of regular physician visits, higher baseline physical activity, and larger social support networks would be more likely to use rehabilitation services.
Methods This study used a pooled, cross-sectional approach to describe utilization patterns for rehabilitation services and identify the predictors of utilization by individuals who sustained a stroke. We used data from individuals who participated in the Health and Retirement Study (HRS)18 during the years 1998-2010 and also reported a stroke. The study was approved by the Case Western Reserve University Institutional Review Board (IRB-2012-215 and IRB-2013-55). A Data User Agreement with Centers for Medicare and Medicaid was also secured.
Databases The HRS is a longitudinal study developed at the University of Michigan.18 It was initiated in 1992 and focused on the health, economic status, and retirement of older adults in the United States. The HRS sample included more than 26,000 Americans older than 50 years who were surveyed biannually. The linked-Medicare file includes claims information for feefor-service beneficiaries including claims from inpatient hospital settings as well as outpatient institutional and noninstitutional settings. The linked-Medicare claims records carry a unique dummy identifier for each respondent that is linked to the HRS data through a file supplied by the HRS staff.
Participants The linked HRS-Medicare file from 1998 to 2010 was used to create the sample and define utilization. Stroke incidence between 2000 and 2008 was identified in the HRS with the participant stating that a doctor told them that they had a stroke. This was confirmed through the presence of diagnosis codes documented in Medicare claims data.19 Only those who were Medicare fee-forservice recipients and also participated in the HRS after the first incident stroke were included in the analyses. Those who reported a stroke prior to 2000 and those who had a proxy respondent at baseline were excluded. The sample was followed through the 2010 wave (fig 1). Power and sample size were determined by the number of eligible cases in the HRS Medicare file data. The 995 eligible cases provided 90% power for a 2-sided test to detect a difference in proportions seeking therapy services (of 0.15-0.24) between the 2 groups based on a binary factor with the proportion in 1 group
List of abbreviations: BMI FPL HRS SES
body mass index federal poverty level Health and Retirement Study socioeconomic status
ranging from 0.3 to 0.5. This assumed a significance level of 0.05, and a design effect of 1.1 assuming an average of 2 waves per participant and an intraclass correlation of 0.1.
Variables of interest A modified version of the World Health Organization’s International Classification of Functioning, Disability and Health was used as the framework for this study because it allowed us to look at both stroke and those personal and environmental factors that influence utilization of services.20,21 A core set of variables specific to stroke, developed by a panel of stroke researchers, was used as a guide for the choice of variables in this study.20 Rehabilitation Respondents were asked whether or not they were participating in rehabilitation services at the time of the HRS survey (yes/no). Presence of rehabilitation services between waves was confirmed in Medicare claims through a search for CPT, Healthcare Common Procedure Coding System, and Revenue Center Codes (supplemental table 1, available online only at http://www. archives-pmr.org/). Demographic characteristics Demographic characteristics were taken from the HRS-RAND Tracker File.19 Age was a continuous variable and was calculated from date of interview and date of birth. Sex (male or female), race (white or black), and ethnicity (Hispanic or Latino) were selfreported in the survey. Ethnicity was dichotomized to identify those who were or were not of Hispanic or Latino origin. Socioeconomic status (SES) was measured as a percentage of selfreported assets and income as a ratio to the federal poverty level (FPL). The ratio of household income to FPL was dichotomized into 200% FPL and <200% FPL. Self-reported education was dichotomized to distinguish those who had completed at least high school or general equivalency diploma from those who had not. Comorbidity This study used 6 of the conditions that were self-reported by respondents in the HRS: high blood pressure, diabetes, lung disease, heart disease, stroke, and arthritis. Comorbidity was measured as a count of comorbidities and analyzed as a continuous variable. Use of self-reported count for comorbidity in this way has been validated in the literature22 and during the development of the HRS. Type of stroke Type of stroke was self-reported in the HRS and confirmed by Medicare claims data. Type of stroke was categorized as hemorrhagic, ischemic, or unspecified if the stroke was undefined in the claims or there was discrepancy in the coding within the claims data.19 Time elapsed since stroke Time since stroke was calculated as the time between each interview and the date of first reported stroke from the HRS. It was recalculated at each wave. Presence of a new stroke At each wave, each individual was asked whether they were diagnosed with a new “stroke” (yes/no). To confirm the presence of an incident stroke, Medicare claims data were searched for www.archives-pmr.org
Utilization of rehabilitation in stroke
3
300 250 200 150 100 50 0 1 wave (n=263) Fig 1
2 waves (n=135)
3 waves (n=53)
5 waves (n=47)
Number of survey respondents by number of waves included in the final regression analysis.
stroke International Classification of Diseasese9th Revision codes with a subsequent inpatient hospital admission, after accounting for a 90-day washout period after the initial stroke. Initial severity of stroke Number of days spent in inpatient ICU care was calculated from Medicare claims data as an additional measure of severity.23 Functional ability The RAND HRS data used factors identified by Wallace and Herzog to create indices of function for use by researchers.24 The RAND index score of functional ability chosen for this study was calculated from 4 subscales within the HRS: strength limitations, upper body mobility limitations, lower body mobility limitations, and activities of daily living. Function was measured on a continuous scale from 0 to 19. The index score ranged from independence (0) to dependence (19) in the above areas. For each wave, the function score was recalculated as a change in function scores from wave to wave so that functional ability could be described. Personal lifestyle factors Personal lifestyle factors that may influence outcomes in stroke include presence of a regular physician and smoking, drinking, and physical activity behaviors.17,25 Whether or not someone had a regular physician was determined by whether the respondent reported a visit to physician over the previous 12 months (yes/no). Smoking was a binary variable (currently smoking vs past smoker or nonsmoker).26 Drinking behavior was categorized as nondrinker, mild or moderate drinker (2 alcoholic drinks/day), or heavy drinker (3 alcoholic drinks/day).26 Physical activity was determined by whether the respondent self-reported vigorous physical activity at least 3 or more days per week (yes/no).26 Baseline working status Working status was used as a baseline characteristic. It was dichotomized to indicate respondents who were doing any work for pay versus those who were not at baseline. Body mass index Body mass index (BMI) was calculated from self-reported height and weight within the HRS and was used as a continuous variable in descriptive analysis. For multivariate analyses, BMI was a categorical variable divided into those who were underweight (BMI<18.5), normal (18.5BMI24.99), overweight (25BMI29.99), and obese (BMI30).
www.archives-pmr.org
4 waves (n=17)
Cognitive deficits The HRS contains an embedded cognitive assessment (the Telephone Interview Cognitive Score). The score was used as a continuous measure in analyses for those who were selfrespondents to the survey after their stroke. Cognition scores ranged from 0 to 35, with a higher score indicating better cognitive abilities. Presence of depressive symptoms The HRS uses a short form of the Center for Epidemiologic Studies Depression scale. The scores range from 0 (no depressive symptoms) to 8 (the highest number of depressive symptoms). The scores were used as a continuous variable. Social support Social support was derived as described by Muramatsu et al as family resources.27 It was taken from marital status, number of living siblings, child availability, and spouse’s self-reported health status. Marital status was coded as currently married or not married, including those who were never married. Child availability was categorical (coresidence, children living within 10 miles, child living more than 10 miles away, no living children or never had children). Spouse’s self-reported health status was dichotomized (fair or poor vs good or better). The categories were used to create an index of social support (0-4). A score of zero indicated lack of any family resources. A score of 4 indicated that the respondent was married with a spouse in good or better health, had children within 10 miles, and had at least 1 living sibling.
Analysis To describe the sample and utilization of services, descriptive analysis of all factors was conducted. Utilization of rehabilitation up to 10 years poststroke was calculated as a percentage of the sample. Univariable analyses were conducted on both the categorical and continuous variables. Constant factors in the analysis were sex, race, SES, stroke type, initial severity of stroke, having visited a physician in the previous 12 months, and baseline measures (employment status, comorbidity, BMI, smoking, drinking, and physical activity levels). Time-varying factors were age, comorbidity, cognitive score, depression score, social support, receipt of rehabilitation, and recurrent stroke. Comorbidity was used as both a static, baseline factor and as a time-varying factor. A multivariable logistic regression with stepwise selection was used to determine the effect of the constant and time-varying
4 factors on utilization of therapy services. Akaike Information Criterion was used to establish a P value cutoff for variable selection into the final model (P<.15) which allowed us to increase power and provide better predictions in the final results.28 Generalized estimating equations (GEEs) were used because we used a large panel database and generalized estimating equation adjusts for natural correlation within the data. In large databases, especially in survey data, it is likely that missing values influence the results of the final analysis. In our sample, 10%-15% of each wave was lost to follow-up. To address the effect missing values might have had on the outcomes, all analyses were done first with missing values and then with imputed missing values using the PROC MI function in SAS. After running the analysis in both manners, no change in statistical significance was noted; thus, the final multivariable logistic regression results are reported without imputation. SAS version 9.2 for UNIX was used for all analysis.a
Results Of the 23,695 records from the linked HRS-Medicare file, 1189 stroke survivors were identified between 2000 and 2008. Of those, 515 stroke survivors with 995 observations comprised the study sample. Utilization rates would have been influenced by loss to follow-up, which was 10%-15% at each wave.
Descriptive analysis of HRS-Medicare sample Most of the sample were women (56.7%), white (81%), had income more than 200% of the FPL (64.6%), and had at least a high school diploma (table 1). The average age for the sample was 73.7 years and most were retired (67.0%). The sample had characteristics consistent with other findings1,2: the average number of comorbidities was 2.4, most had an ischemic stroke (63.3%), and also reported a second stroke (66.4%). In addition, the average depression score SD was 2.22.2, indicating that the sample, on average, reported at least 1 or 2 depressive symptoms. The average cognitive score SD among the sample in the survey poststroke was 19.85.4. Social support varied among the sample with 18.7% of the sample having index scores of 0 or 1 (indicative of the least amount of social support) and 31.4% having scores of 2 (out of a maximum of 4 in social support). Most of the respondents (95.7%) had a physician visit during the preceding 12 months, 12.6% were current smokers, and 4.1% were heavy drinkers. Those who participated in regular physical activity comprised 29.5% of the sample.
Utilization of rehabilitation At the first interview post stroke, only 21.6% of stroke survivors reported receipt of rehabilitation services (table 2).
Logistic regression on participation in rehabilitation over time Univariable logistic regression analyses were conducted on all variables (table 3). Age, initial severity, functional ability, second stroke, cognitive scores, and personal lifestyle factors (physician visit and smoking status) predicted participation in therapy services.
K.M. Keptner et al Table 1 Demographic characteristics of HRS-Medicare sample (1998-2008) (NZ515 individuals) Descriptive Variables
n (%)
Sex Male Age (y), mean SD Race White Black Ethnicity Hispanic/Latino Not Hispanic/Latino Socioeconomic status* <200% FPL 200% FPL Education
223 (43.30) 73.678.36 422 (81.94) 93 (18.05) 36 (6.99) 479 (93.01) 163 (31.65) 298 (64.64) 179 (34.76) 336 (65.24) 41 (7.96) 16 (3.11) y
42 (8.16) 345 (66.99) y
53 (10.29) 27.044.98 326 (63.30) 50 (9.71) 139 (26.99) 342 (66.4) 3.61.8 2.391.39 463 (95.66) 54 (12.59) 21 (4.08) 152 (29.51) 19.85.4 2.22.2 96 (18.7) 162 (31.4) 117 (22.7) 140 (27.2) 31.268.7
Abbreviations: HS, high school; MD, medical doctor. * Calculation influenced by missing values. y Category counts hidden, per Institutional Review Board and user data agreement with Centers for Medicare and Medicaid Services to limit reporting for categories that include <11 participants; additional data shielded to prevent identification of participants in the HRSMedicare database.
Table 4 presents the final logistic regression model. Of the factors that met the cutoff during stepwise selection, age (P<.001) and having a second stroke (PZ.051) were significant predictors, while having an undefined stroke (PZ.077) trended toward statistical significance as a predictor of participation in rehabilitation. www.archives-pmr.org
Utilization of rehabilitation in stroke Table 2
5
Self-reported participation in rehabilitation by HRS survey year (NZ515 individuals)
Response
0-2 y (NZ515)
2-4 y (nZ268)
4-6 y (nZ152)
6-8 y (nZ91)
8-10 y (nZ53)
Yes No
111 (21.6) 404 (78.6)
35 (6.8) 233 (86.9)
24 (15.8) 128 (84.2)
15 (16.5) 76 (83.5)
<11 (20.7) >42 (>79.2)
NOTE. Count hidden, per Institutional Review Board and user data agreement with Centers for Medicare and Medicaid Services to limit reporting for categories that include <11 participants; additional data shielded to prevent identification of participants in the HRS-Medicare database.
With each year of increased age, participants were 1.14 times more likely to participate in rehabilitation (PZ.001). Those who had a second stroke were 1.64 times more likely to participate (PZ.051).
Discussion Medicare claims data with linked HRS data allowed us to look at multiple demographic, health, and personal level factors that Medicare claims data alone do not provide. Linking the data allowed us to explore not only who used services, but if there was any predictive value in obtaining data on self-reported health conditions and lifestyle factors, which are in line with recent efforts to understand the utilization of rehabilitation poststroke.29 Many of our hypotheses were not supported by the data; having additional comorbidities, increased severity of initial stroke, higher SES, history of regular physician visits, regular physical activity, and a larger social support network were not significant predictors of rehabilitation utilization. Table 3 Univariable logistic regression of use of rehabilitation services on individual factors: demographics, health condition, and personal lifestyle factors (NZ995 observations) Descriptive Variables
OR
P Value
Sex, female Age Race (white vs black/other) Income, % FPL (200% FPL) Education (HS grad or GED) Employment at baseline (working) Comorbidity Stroke type ischemic Hemorrhagic Unspecified Initial severity Functional ability Second stroke BMI (continuous) Cognition score Depressive symptoms Presence of social support Personal lifestyle factors: MD visit within year (yes/no) Smoker (yes/no) Heavy drinker (yes/no) Physical activity (yes/no) Time elapsed
1.06 1.09 1.15 1.13 1.41 1.86 1.15 Ref 0.58 3.07 0.92 1.12 1.99 1.06 0.93 1.06 0.80 0.26
.85 .001* .71 .80 .29 .06 .23 —— .38 .01y .020z .001* .0007* .84 .015z .36 .11 .007y
2.74 2.29 1.14 1.00
.009y .16 .69 .96
Abbreviation: GED, general equivalency diploma; HS, high school; MD, medical doctor; OR, odds ratio. * P.001. y P.01. z P.05.
www.archives-pmr.org
Most of the sample did not report receiving therapy when asked during a given survey. This was well below previous estimates.11 The highest utilization period in this study was within 2 years poststroke, with just more than 20% of the sample stating that they received services in the previous 2 years. As time passed, utilization decreased before it increased again. The change in utilization patterns may be reflective of the recovery process and aging. After the initial episodes of care, survivors tend to adjust to disability3 and may not need additional rehabilitation services. The increase in services after a decrease might be indicative of increased morbidity, injury, or illness that necessitated services. It may be, however, that stroke survivors needed additional services but were not able to access them. Finances, including the ability to pay deductibles and copayments, may have limited access to service independent of functional status or need for services. Recent studies suggest that stroke survivors in the community do not receive the same level of formal care as individuals with other neurologic conditions,30 and a low level of therapy services as found in this study may confirm this conclusion. The day-to-day formal care, including therapy that is needed because of the stroke, may be limited by provider recognition of impairments that indicate a referral for therapy services. Providers may not refer for services if a patient does not request them or if a patient does not report a specific concern that would necessitate a referral. Further, therapists themselves may not be knowledgeable about reassessing stroke patients. Therapists may believe that patients have reached a plateau in progress, when in fact they have not. It may be difficult to find therapists in a local area who have the skill to determine the therapy needs of a stroke patient, further limiting access to appropriate services. The primary drivers of utilization of therapy in this study were increased age and incident stroke, which are confirmed by others.31,32 Older individuals and those with a recurrent stroke may have an increase in health care encounters due to functional Table 4 Logistic regression with stepwise selection of the presence of rehabilitation on significant univariable factors (age, second stroke, stroke type, functional ability, cognition) (NZ995 observations) Variables
OR
95% CI
P Value
Age Second stroke Stroke type Ischemic Hemorrhagic Unspecified Functional ability Cognition
1.14 1.64
(1.08-1.22) (1.00-2.70)
.001* .051y
Reference 0.76 2.17 1.07 0.99
— (0.10-5.70) (0.21-0.95) (0.98-1.18) (0.94-1.06)
— .788 .077 .101 .932
Abbreviations: 95% CI, 95% confidence interval; OR, odds ratio. * P.001. y P.05.
6 concerns that increase the likelihood that a referral to rehabilitation services is generated. This is consistent with other studies that find that utilization of services is determined by the health-related needs of patients.33
Study limitations Because this research was based on secondary data, it was not possible to explore all the issues that can effect utilization. Changes between 2 waves, not captured by these data, might have influenced the findings as would the use of self-reported data. The data may have been biased from participants under- or overreporting lifestyle factors that were used in the final analyses. In addition, because we chose to use self-report of comorbidity instead of using Medicare claims for comorbidity data, comorbidity measures may also have some bias. Finally, this sample was older than the original HRS sample after the addition of Medicare claims, because individuals older than the age of 65 are the primary recipients of Medicare benefits. Using Medicare fee-for-service claims also limited the conclusions because utilization in newer payment structures that have arisen over recent years might yield different results. However, because we used a large survey database in concurrence with Medicare claims, we believe that the results have some generalizability among an older, Medicare fee-for-service population. Despite these limitations, the methods allowed us to observe a multitude of personal factors that might influence utilization of rehabilitation up to 10 years poststroke, which is unique. Finally, the study looked at rehabilitation broadly, including both occupational therapy and physical therapy, with no distinction between the 2 services. This could have biased the conclusions because some patients may have only needed 1 service or the other, but this was not captured in this study.
Conclusions This study demonstrated that age and a new, incident stroke are predictors of utilization of rehabilitation services beyond 1-year poststroke, but that utilization of rehabilitation was lower than previous estimates. Future research should explore the reasons for such low rates of utilization of rehabilitation, whether rehabilitation services are needed by most of the individuals who sustain a stroke, and what, if any, barriers are present for individuals to participate in rehabilitation services.
Supplier a. SAS version 9.2 for UNIX; SAS Institute Inc.
Keywords Health services; Occupational therapists; Physical therapists; Rehabilitation; Stroke
Corresponding author Karen M. Keptner, PhD, OTR/L, Cleveland State University, 2121 Euclid Avenue HS 101, Cleveland, OH 44115. E-mail address: [email protected].
K.M. Keptner et al
References 1. Benjamin EJ, Blaha MJ, Chiuve SE, et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statisticsd2017 update: a report from the American Heart Association. Circulation 2017;135:e229-445. 2. Schwamm L, Fayad P, Acker J, et al. Translating evidence into practice: a decade of efforts by the American Heart Association/ American Stroke Association to reduce death and disability due to stroke: a presidential advisory from the American Heart Association/American Stroke Association. Stroke 2010;41:1051. 3. World Health Organization. Global health observatory. Available at: http://www.who.int/gho/ncd/mortality_morbidity/cvd_text/en/. Accessed April 1, 2014. 4. Desrosiers J, Noreau L, Rochette A, Bravo G, Boutin C. Predictors of handicap situations following post-stroke rehabilitation. Disabil Rehabil 2002;24:774. 5. National Stroke Foundation. Organization of services. In: Clinical guidelines for stroke management 2010. Available at: http://www.guideline.gov/ content.aspx?idZ24168&searchZcva. Accessed July 27, 2014. 6. Outpatient Service Trialists. Rehabilitation therapy services for stroke patients living at home: systematic review of randomised trials. Lancet 2004;363:352-6. 7. Winstein CJ, Stein J, Arena R, et al. Guidelines for adult stroke rehabilitation and recovery: a guideline for health care professionals from the American Heart Association/American Stroke Association. Stroke 2016;47:e98-169. 8. Miller E, Murray L, Richards L, et al. Comprehensive overview of nursing and interdisciplinary rehabilitation care of the stroke patient: a scientific statement from the American Heart Association. Stroke 2010;41:2402. 9. Freburger JK, Li D, Johnson AM, Fraher EP. Physical and occupational therapy from the acute care to the community setting after stroke: predictors of use, continuity of care, and timeliness of care. Arch Phys Med Rehabil 2017;99:1077-89. 10. Xie J, George M, Ayala C, McGruder H, Denny C, Croft J. Outpatient rehabilitation among stroke survivors e 21 states and the District of Columbia. MMWR Morb Mortal Wkly Rep 2007;56:504. 11. Kuo Y, Ostir G, Granger C, Ottenbacher K. Examination of follow-up therapy in patients with stroke. Am J Phys Med Rehabil 2006;85: 192-200. 12. Freburger JK, Konrad TR. The use of federal and state databases to conduct health services research related to physical and occupational therapy. Arch Phys Med Rehabil 2002;83:837-45. 13. Gadidi V, Katz-Leurer M, Carmeli E, Bornstein N. Long-term outcome poststroke: predictors of activity limitation and participation restriction. Arch Phys Med Rehabil 2011;92:1802-8. 14. Andrews AW, Li D, Freburger JK. Association of rehabilitation intensity for hospital readmission. Phys Ther 2015;95:1660-7. 15. Cruz-Flores S, Rabinstein A, Biller J, et al. Racial-ethnic disparities in stroke care: the American experience: a statement for health professionals from the American Heart Association/American Stroke Association. Stroke 2011;42:2091-116. 16. Lee Y, Kyung-Park M. Health practices that predict recovery from functional limitations in older adults. Am J Prev Med 2006;31:25. 17. Mold F, McKevitt C, Wolfe C. A review and commentary of the social factors which influence stroke care: issues of inequality in qualitative literature. Health Soc Care Community 2003;11:405-14. 18. Health and Retirement Study, (RAND enhanced Fat Files, 1998-2010, RAND v. M, and Tracker File v.1.0) public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009740) Ann Arbor, MI, (2013). 19. Andrade SE, Harrold LR, Tija J, et al. A systematic review of methods identifying cerebrovascular accident or transient ischemic attack using administrative data. Pharmacoepidemiol Drug Saf 2012;21(Suppl 1): 100-28.
www.archives-pmr.org
Utilization of rehabilitation in stroke 20. Geyh S, Cieza A, Schouten J, et al. ICF core sets for stroke. J Rehabil Med 2004;(44 Suppl):135-41. 21. World Health Organization. International classification of functioning, disability and health (ICF). Geneva, Switzerland: World Health Organization; 2001. 22. Huntley AL, Johnson R, Purdy S, Valderas JM, Salisbury C. Measures of multimorbidity and morbidity burden for use in primary care and community settings: a systematic review and guide. Ann Fam Med 2012;10:134-41. 23. Williams TA, Ho KM, Dobb GJ, Finn JC, Knuiman M, Webb SA. Effect of length of stay in intensive care unit on hospital and long-term mortality of critically ill adult patients. Br J Anaesth 2010;104:459-64. 24. Wallace RB, Herzog AR. Overview of the health measures in the Health and Retirement Study. J Hum Resour 1995:S84-107. 25. Wray L, Alwin D, McCammon R. Social status and risky health behaviors: results from the Health and Retirement Study. J Gerontol B Psychol Sci Soc Sci 2005;60(S2):85-92. 26. Rahrig Jenkins K, Ofstedal MB, Weir D. Documentation of health behaviors and risk factors measured in the Health and Retirement Study (HRS/AHEAD). Ann Arbor, MI: Survey Research Center, University of Michigan; 2008.
www.archives-pmr.org
7 27. Muramatsu N, Yin H, Campbell R, Hoyem R, Jacob M, Ross C. Risk of nursing home admission among older Americans: does states’ spending on home- and community-based services matter? J Gerontol B Psychol Sci Soc Sci 2007;62:S169-78. 28. Steyerberg E. Clinical prediction models. New York: Springer-Verlag; 2009. 29. Nelson ML, McKellar KA, Munce S, et al. Addressing the evidence gap in stroke rehabilitation for complex patients: a preliminary research agenda. Arch Phys Med Rehabil 2018;99:1232-41. 30. Obembe AO, Goldsmith CH, Simpson LA, Sakakibara BM, Eng JJ. Support service utilization and out-of-pocket payments for health services in a population-based sample of adults with neurological conditions. PLoS One 2018;13:e0192911. 31. Cook C, Stickley L, Ramey K, Knotts V. Variables associated with occupational and physical therapy stroke rehabilitation utilization and outcomes. J Allied Health 2005;34:3-10. 32. Pettersen R, Dahl T, Wyller T. Prediction of long-term functional outcome after stroke rehabilitation. Clin Rehabil 2002;16:149-59. 33. Jourdan C, Bayen E, Darnoux E, et al. Patterns of post-acute health care utilization after a severe traumatic brain injury: results from the PariS-TBI cohort. Brain Inj 2015;29:701-8.