1 184 ORIGINAL ARTICLE Impact of Mental on Cost and Reimbursement for Patients in Inpatient Rehabilitation Facilities Deborah Dobrez, PhD, Allen W. Heinemann, PhD, Anne Deutsch, RN, PhD, CRRN, Elizabeth M. Durkin, PhD, Orit Almagor, MA ABSTRACT. Dobrez D, Heinemann AW, Deutsch A, Durkin EM, Almagor O. Impact of mental disorders on cost and reimbursement for patients in inpatient rehabilitation facilities. Arch Phys Med Rehabil 2010;91: Objective: To determine whether comorbid mental disorders affect inpatient rehabilitation facility (IRF) costs and to examine the extent to which Medicare s prospective payment system reimbursement sufficiently covers those costs. Design: Secondary analysis of Medicare IRF Patient Assessment Instrument files and Medicare Provider and Review files. Payment was compared with costs for patients with and without reported mood, major depression, substance use, or anxiety disorders. The relationships among payment group assignment, comorbidity-related adjustments in payment, and the presence of mental disorders were estimated. Setting: IRFs (N 1334) in the United States. Participants: Medicare fee-for-service beneficiaries (N 1,146,799) discharged from IRFs from 2002 to Interventions: Not applicable. Main Outcome Measure: IRF costs. Results: Mental disorders were reported for 13% of the Medicare fee-for-service beneficiaries. After controlling for payment group and comorbidity classifications, patients with mood, major depression, or anxiety disorders had significantly greater costs of $433, $1642, and $247 compared with patients without these disorders. The higher cost for patients with major depression (14.9% higher) is sufficient to justify a tier 2 comorbidity classification. Conclusions: A reimbursement adjustment for the presence of a major depressive disorder would bring Medicare reimbursement in line with facility costs. The failure to compensate facilities directly for providing care to patients with major depression may result in reduced access to care for these patients. It also may create a disincentive to meet mental health treatment needs during the rehabilitative episode. Further work is needed to compare costs between patients with and without confirmed mental health disorders, given concerns about the accurate reporting of mental health disorders. Key Words: Costs and cost analysis; Depression; Medicare; Mental disorders; Prospective payment system; Rehabilitation by the American Congress of Rehabilitation Medicine THE ACUTE CARE INPATIENT PPS implemented in 1983 created financial pressure for acute care hospitals that shortened lengths of stay and shifted more patient care to postacute care services. 1 As a result, Medicare spending for postacute care grew dramatically. For example, Medicare payments to inpatient rehabilitation facilities grew from $1.9 billion to $3.9 billion between 1990 and 1993, and the number of rehabilitation hospitals and units increased 11% and 7%, respectively. 2 This explosion of postacute care expenditures renewed interest in Medicare PPSs for postacute care settings, leading ultimately to the creation of a per-discharge payment system for inpatient rehabilitation facilities in Under the IRF PPS, Medicare reimburses facilities using the CMG classification system. Patients insured under the traditional Medicare program are assigned to a CMG based on the primary impairment (1 of 21 RICs, eg, stroke) and admission motor functional status. For some RICs, adjustments are also made for cognitive functional status and age. The presence of certain comorbidities is also considered when determining the reimbursement. Patients with comorbidities expected to increase resource use above the average CMG reimbursement level are placed in 1 of 3 tiers (tiers 1 3, with tier 1 receiving the greatest reimbursement), with patients with no qualifying comorbidity receiving no tier placement. 3 Within an RIC, placement into a CMG has the greatest impact on payment; for example, a nontiered payment for a patient with stroke is 5.7 times greater in the highest-paying CMG (114) relative to the lowest-paying CMG (101) for CMGs used between 2002 and However, tier classification for comorbidities also has a significant impact on payment, with payment for patients with stroke qualifying for the highest (most costly) tier placement, exceeding payment for patients with no qualifying comorbidity by as much as 24%. 3 The initial list of tiered comorbid conditions was developed based on the prevalence of comorbidities reported by IRFs and their calculated impact on costs before the final PPS rule in 2001, 4 a period during which financial incentives to code and report comorbidities accurately did not exist. Analyses of cost and comorbidity data collected since the PPS began showed that patients with depression and other affective disorders had List of Abbreviations From the University of Illinois at Chicago (Dobrez); Rehabilitation Institute of Chicago (Heinemann, Deutsch); and Northwestern University (Heinemann, Deutsch, Durkin, Almagor), Chicago, IL. Supported by the Health Services Research Disability and Rehabilitation Research Project on Medical Rehabilitation (grant no. H133A030807) awarded by the National Institute on Disability and Rehabilitation Research to the Rehabilitation Institute of Chicago. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated. Correspondence to Deborah Dobrez, PhD, 1603 W Taylor Street, Chicago, IL 60612, Reprints are not available from the author /10/ $36.00/0 doi: /j.apmr CMG ICD-9 ICD-9-CM IRF IRF-PAI MEDPAR PPS RIC case-mix group International Classification of Diseases 9th Revision International Classification of Diseases 9th Revision Clinical Modifications inpatient rehabilitation facility inpatient rehabilitation facility-patient assessment instrument Medicare provider and review prospective payment system rehabilitation impairment category
2 MENTAL DISORDERS AND REHABILITATION COST, Dobrez 185 costs 3% higher than patients without affective disorders, but although updates including additions to the comorbidity list have been made, these disorders were not added to the list of tiered comorbidities because of concerns about imprecise coding and the possibility that use of imprecise codes would increase if they were to affect payment. 5 If IRF costs for patients with mental disorders are greater than for patients without mental disorders, the failure to reimburse facilities for the care of patients with mental disorders creates several dilemmas: (1) IRFs that admit a disproportionate number of patients with comorbid mental disorders may face additional financial pressure; (2) IRFs may deny admission to a patient based in part on the presence of comorbid mental disorders; and (3) the care of IRF patients with comorbid mental disorders may be limited because of insufficient reimbursement. Among the most common and potentially debilitating mental disorders that can affect IRF patients are mood, 6 anxiety, 7 and substance use 8 disorders. This article focuses on the impact of mood, anxiety, and substance use disorders on costs and estimated IRF PPS payments for patients receiving inpatient rehabilitation between 2002 and This article additionally focuses on a subset of the patients with mood disorders, those with major depressive disorders. We hypothesized that for patients in the same CMG and tier, those with mental disorders would have greater costs than patients without mental disorders. Because no direct adjustment in reimbursement is made for a diagnosis of a mental disorder, PPS Medicare reimbursements were expected to be equivalent for patients with and without mental disorders. METHODS Data and Measures The study was approved by the institutional review board at the University of Illinois at Chicago and Northwestern University. The data used in this study are from the Centers for Medicare and Medicaid Services IRF-PAI files and MEDPAR files for patients discharged in calendar years 2002 through The MEDPAR file contains demographic, hospitalization, charge, and payment data for all fee-for-service Medicare beneficiaries receiving inpatient rehabilitation facility care. The IRF-PAI database includes demographic, hospitalization, impairment, diagnostic (ICD-9 codes), and functional status data. We linked the MEDPAR and IRF-PAI data for the same IRF stay using scrambled patient identifiers, provider identification codes, admission and discharge dates, and patient demographic data. We linked 92.6% of the MEDPAR records for 2002, 94.5% for 2003, and 94.6% for We obtained cost-tocharge ratios for IRFs from Centers for Medicare and Medicaid Services (http://www.cms.hhs.gov/inpatientrehabfacpps/). Medicare beneficiaries with a primary payer of Medicare fee-for-service who were discharged from an inpatient rehabilitation facility between 2002 and 2004 were included in the sample. Exclusion criteria were based on factors affecting payment under PPS, including incomplete or atypical rehabilitation stays: (1) discharge to acute care or another IRF; (2) death during IRF stay; (3) length of stay less than 3 days; and (4) length of stay greater than 3 SDs above the mean (calculated separately for each year). In addition, we excluded patients age 16 years or younger. We identified reported mental disorder diagnoses using ICD- 9-CM codes recorded in the IRF-PAI files: (1) mood: , 296.xx, 300.4, and 311; (2) substance use: , , , , , and ; and (3) anxiety: , , , and Because patient needs may be especially affected by more severe mental disorders, we also analyzed separately patients with major depression, a particularly debilitating subset of mood disorders (ICD-9-CM codes 296.2x, 296.3x, 296.5x). Functional status data from the IRF-PAI database are from the FIM instrument, which includes 13 motor items and 5 cognitive items. For the IRF PPS in 2002 through 2004, CMG assignment was based on a FIM-12 motor score (the sum of the scores assigned to 12 of the items) and a cognitive score (the sum of the scores from the 5 cognitive items). Total cost was calculated by multiplying total charges by the facility-specific cost-to-charge ratios. Medicare PPS payments were estimated using yearly base rates of $11,838, 4 $12,193, 9 and $12, for fiscal years 2002, 2003, and 2004, respectively, and were adjusted for medical inflation to 2004 dollars. 11 The base rates were adjusted for CMG assignment and comorbidities (reflected by tier assignment) as reported in the IRF-PAI records. Medicare Part A facility payments were calculated without facility-specific adjustments, such as adjustments for wages, rural settings, and low-income populations. Statistical Analyses Two-sided hypothesis tests with an alpha level of.05 were conducted. Demographic characteristics of patients with and without reported mental disorders were compared using t tests and chi-square analyses. We compared costs for patients with and without mental disorders. Unadjusted comparisons for all patients were conducted first to determine whether payment differs from cost for patients with a mental disorder using t tests. We used ordinary least-squares regressions to estimate the impact of mental disorders on cost, controlling for patientlevel factors that adjust payment including CMG, tiered comorbidities, and unusually high costs that result in additional reimbursement (high-cost threshold). Correlation of costs for patients within a facility because of unobserved variables was expected. We controlled for these clustering effects in each of the regression models using the Huber clustering correction, which adjusts the coefficient SEs for within-facility correlation, but does not affect the parameter estimates. 12,13 Because inpatient costs are likely to be skewed, the analyses were conducted again excluding patients with the top 0.5% of all costs. Finally, we conducted individual regression analyses for each RIC to determine whether any differences in the full sample were observed within individual RICs. A Bonferroni correction was made for the multiple comparisons for each set of 21 regressions. RESULTS Characteristics of the Study Population A total of 1,146,799 Medicare fee-for-service beneficiaries from 1334 facilities met the selection criteria. From the full patient database of 1,621,491 patients from 2002 to 2004, 25% of the patients had incomplete stays (discharge to acute care or another IRF, death during the IRF stay, or length of stay of less than 3 days), 2% of the patients were admitted to IRFs with fewer than 50 annual admissions, and 4% of the patients did not have Medicare as a primary payer. After exclusion of these patients, an additional 0.1% were excluded for log lengths of stay greater than 3 SDs above the mean a year, and 0.3% were excluded because of missing facility cost-to-charge ratios. The patients were evenly distributed across years (32%, 34%, and 34% for 2002, 2003, and 2004, respectively). The mean SD age of patients was years, and 64% of the sample was female. The most populated RICs were lower extremity joint replacement (RIC 7; 28%), stroke (RIC 1;
3 186 MENTAL DISORDERS AND REHABILITATION COST, Dobrez Table 1: Patient Characteristics by the Presence of a Reported Mental Disorder Mood Major Depressive Anxiety Substance Use Variable (n 146,088) (n 1,000,711) (n 10,431) (n 1,136,368) (n 40,679) (n 1,106,120) (n 12,279) (n 1,134,520) Age (mean) Female (%) Admission motor FIM score (mean) * 41.8 Admission cognitive FIM score (mean) *P.01; P %), and lower extremity fracture (RIC 8; 12%). The mean SD FIM-12 motor admission and discharge scores (which range from 12 to 84, with higher numbers indicating greater function) 14,15 were and , respectively. Mean SD FIM cognitive admission and discharge scores (which range from 5 to 35, with higher scores indicating greater function) were and , respectively. The mean SD length of stay was days. Seventyfour percent of the patients did not have a qualifying comorbidity resulting in a tier classification. Fifteen percent had a tier 3 (low-cost) comorbidity, 9% had a tier 2 comorbidity, and 2% had a tier 1, the highest paying, comorbidity. The prevalence of comorbid mental disorders (based on ICD-9 codes) identified in the IRF-PAI records was as follows: all mood disorders, 13%; major depression only, 1% (7% of all patients with mood disorders); anxiety disorders, 4%; and substance use disorders, 1%. There was little overlap in mental disorders; only 1% of the sample had mental disorders in more than 1 category. Because there is little overlap in mental disorders, table 1 displays demographic and admission functional status data for patients with and without each of the 4 mental disorders. On average, patients with mental disorders were younger (especially those with substance use disorders, who were younger by 8 years) than those without mental disorders. Women were more likely to be diagnosed with an anxiety or mood disorder, while men were more likely to be diagnosed with substance use disorders. Impact of Mental on Cost and Payment Figure 1 shows the unadjusted difference in total cost IRF PPS payment for Medicare beneficiaries with and without a * 803* 2950* 1687* * 731* 1053* Mood Major Depr Anxiety Substance Use Differences in total cost Differences in IRF Payment Fig 1. Unadjusted differences in total cost ($) and IRF PPS payment ($) by the presence of a mental disorder. * P<.001. A positive value indicates that patients with this mental disorder had higher costs or payments compared with those without the disorder. Abbreviation: Depr, depression. recorded mental disorder. For all disorders but anxiety, both cost and payment were greater for patients with a mental disorder (P.001). Margins, calculated as the difference between payment and cost, divided by cost, were significantly lower (P.001) for patients with mood disorders (.093) compared with patients with no mood disorders (.118), for patients with major depressive disorders (.002) compared with patients with no major depressive disorder (.116), and for patients with anxiety disorders (.094) compared with patients with no anxiety disorder (.116). No differences in margins were found for patients with substance use disorders compared with patients without substance use disorders (.115 for both). Adjusted Differences in Cost IRF PPS payments (without facility-level adjustments) are identical within each CMG/tier-defined group of patients. To determine whether indirect payment adjustments from CMG/ tier-shifts (usually caused by differences in age or admission function related to the presence or absence of mental disorders) and tier-shifts (caused by an association between mental disorders and other tier-qualifying comorbidities) are sufficient to compensate for additional costs, the impact of mental disorders on cost was measured in multivariate analyses, controlling for CMG and tier. Table 2 shows the results of 4 regression analyses (1 each for mood, major depression, anxiety, and substance use disorders) predicting costs as a function of the presence of the mental disorder, tier classification, and CMG (tier and CMG coefficients not shown because of space limitations). Across the 4 categories of mental disorders, higherpaying tiers and higher-paying CMGs were associated with significantly greater cost. After controlling for CMG and for tier classifications that affect IRF payments, patients with mood, major depression, or anxiety disorders had significantly greater costs ($433, $1642, and $247, respectively) than patients without these disorders. These cost differences represent 3.9%, 14.9%, and 2.3% of the average total cost for each of the 3 groups. Because inpatient costs are likely to be skewed even within a sample of inpatients, the analyses were conducted again excluding patients with the top 0.5% of all costs (greater than $42,699). In the sample excluding the outlier patients, patients with mood, major depression, and anxiety costs had significantly greater costs of $414, $1593, and $301, respectively. Substance use remained statistically insignificant with lower costs of $77, compared with lower costs of $86 in the sample excluding the high-cost outlier patients. We completed 21 regression analyses, 1 for each RIC, to determine whether the effects of all mood disorders, major depressive disorders only, and anxiety disorders on costs were consistent across the 21 RICs (table 3). We observed statistically significant increases in cost for patients with 7 of the 21 RICs for mood disorders, 8 of the 21 RICS for major depressive disorders, and 1 of the 21 RICs for anxiety disorders. RICs
4 MENTAL DISORDERS AND REHABILITATION COST, Dobrez 187 Table 2: Regressions of Cost on the Presence of Reported Mood, Major Depression, Anxiety, and Substance Use, Controlling for Tier and CMG Cost Coefficient (95% Confidence Interval) Mental Disorder Mood Major Depressive Anxiety Substance Use Mood disorders (321.94, ) Major depressive disorders 1, ( , ) Anxiety disorders * (111.98, ) Substance use disorders ( , ) R *P.01; P.001. with nonsignificant differences in costs for mood and major depressive disorders were among the least populated RICs, making up 16% and 13% of the analytic samples for mood and major depression. However, nonsignificant differences in costs for anxiety disorders were found for 58% of the analytic sample. DISCUSSION This study focused on fee-for-service Medicare beneficiaries receiving postacute rehabilitation in inpatient rehabilitation facilities in the 3 years after the implementation of Medicare s IRF PPS. The difference in payment for patients with reported mental disorders compared with patients without reported mental disorders was at least partially explained by concurrent differences in CMG assignment (determined mostly by admission functional status and age) and in tier adjustments (for Table 3: Impact of Reported Mood, Major Depression, and Anxiety on Cost for Each of 21 Rehabilitation Impairment Categories Rehabilitation Impairment Categories Regression Coefficients ($) Mood Major Depressive Anxiety Stroke 528* Traumatic brain injury Nontraumatic brain injury Traumatic spinal cord injury Nontraumatic spinal cord injury Neurologic * 84 Fracture of the lower extremity 341* 1915* 135 Replacement of the lower extremity joint 426* 2013* 400* Other orthopedic 453* 1712* 18 Amputation, lower extremity Amputation, nonlower extremity Osteoarthritis 658* Rheumatoid, other arthritis Cardiac * 219 Pulmonary 555* 2973* 457 Pain syndrome * 351 Major multiple trauma, nonbrain and spinal injury * 771 Major multiple trauma, brain and spinal injury Guillain-Barré Miscellaneous 388* Burns *P.002 (.05/21). comorbidities other than mental disorders). After adjustment for CMG and tier assignment, costs remained higher for patients with reported mood, major depression, and anxiety disorders, supporting the hypothesis that mental disorders are related to significantly greater cost. Tier classifications were designed to ensure sufficient coverage of increased costs because of comorbidities. A tier 1 classification is made for comorbidities associated with costs that are 15% or greater than average costs; tier 2 and 3 classifications are made for comorbidities with costs that range from 11% to 15% and from 4% to 10%, respectively, on average. 16 The unreimbursed increases in cost for patients with mood and anxiety disorders are 3.9% and 2.3%, respectively, of the average total cost both below the minimum 4% threshold for a tier 3 classification. However, the unreimbursed increase in cost for patients with major depression is 14.9% of average total cost, meeting the threshold for a tier 2 classification. This finding of significantly greater cost for patients with major depression was consistent across 8 of the 21 RICs, for 87% of the sample, supporting the use of a tier classification for all IRF Medicare beneficiaries with a diagnosis of major depression. Study Limitations Several limitations to this study should be noted. IRFs may code up to 10 comorbidities and 6 complications on the IRF PAI. However, facilities may be most apt to record comorbidities that affect tier classification and reimbursement, making the recording of mental disorders a low priority. Facilities without the resources to treat mental disorders may also be less likely to diagnose the disorders reliably, making accurate recording in the IRF PAI difficult. Mental disorders are likely to be inconsistently diagnosed and underreported across inpatient facilities. The true difference in cost caused by the presence of mental disorders is underestimated if patients with mental disorders that are not reported have higher costs than other patients. However, the true difference in cost is overestimated if patients with mental disorders that are not reported have similar or lower costs than other patients. Our study focuses on Part A hospital reimbursement. Mental health care may be reimbursed through Part B of Medicare when provided through outside contractors, increasing the likelihood of receipt of appropriate mental health care. Finally, this study was not sufficiently powered to detect cost differences for RICs with relatively small sample sizes. PPS reimbursement is not designed to cover perfectly the costs associated with individual patients, but rather to cover the average expected costs with patient-level adjustments for factors including admission function and comorbidities. Failure to reimburse for comorbidities that substantially raise treatment costs for a group of patients can have significant effects on access to care and adequacy of care for those patients.
5 188 MENTAL DISORDERS AND REHABILITATION COST, Dobrez CONCLUSIONS Mental disorders have been excluded entirely from the tier classifications in the IRF PPS. Our study suggests that this exclusion is reasonable given the available data for substance use, anxiety disorders, and a broadly defined group of mood disorders. However, major depression substantially raises cost beyond what is indirectly reimbursed through other comorbidities and case-mix group assignment. A tier 2 classification would more adequately cover treatment costs, encouraging access to care and quality of care for patients with major depression. Efforts are needed to assess the accuracy of diagnoses of major depression in particular, and mental disorders more generally, in inpatient rehabilitation facilities. Further comparisons of cost and payment for patients with and without mental disorders with confirmed mental health diagnoses would strengthen the case for tier adjustments for specific mental disorders. The inclusion of any mental disorder in the tier coding is likely to increase the prevalence of coding, possibly altering the relationship between payment and costs for this group. Further re-analysis of billing data will be required to confirm a tier adjustment or alter the adjustment as appropriate. References 1. Medicare Payment Advisory Committee. Medicare payment policy: report to the Congress. Washington (DC): U.S. Government Printing Office; Prospective Payment Assessment Commission. Post-acute providers, report and recommendations to the Congress. Washington (DC): U.S. Government Printing Office; Inpatient Rehabilitation Facility Patient Assessment Instrument (IRF-PAI) training manual Buffalo: Foundation Activities, Inc/Uniform Data System for Medical Rehabilitation; Medicare program; prospective payment system for inpatient rehabilitation facilities: final rule. Federal Register 2001;66: Carter GM, Totten ME. Preliminary analyses of the refinement of the tier comorbidities in the inpatient rehabilitation facility prospective payment system. Santa Monica: RAND Corp; Turner-Stokes L, Hassan N. Depression after stroke: a review of the evidence base to inform the development of an integrated care pathway: part 1: diagnosis, frequency and impact. Clin Rehabil 2002;16: Morrison V, Pollard B, Johnston M, MacWalter R. Anxiety and depression 3 years following stroke: demographic, clinical and psychological predictors. J Psychosom Res 2002;59: Hinkle JL, Smith R, Revere K. A comparison of stroke risk factors between men and women with disabilities. Rehabil Nurs 2006;31: Notice. Federal Register 2002;67: Final rule. Federal Register 2003;68: Bureau of Labor Statistics. Archived Consumer Price Index news releases. Available at: Accessed January, Moulton BR. An illustration of a pitfall in estimating the effects of aggregate variables on micro units. Rev Econ Stat 1990;72: Deaton A. The analysis of household surveys. Baltimore: Johns Hopkins Pr; Stineman MG, Shea JA, Jette A, et al. The Functional Independence Measure: tests of scaling assumptions, structure, and reliability across 20 diverse impairment categories. Arch Phys Med Rehabil 1996;77: Kidd D, Stewart G, Baldry J, et al. The Functional Independence Measure: a comparative validity and reliability study. Disabil Rehabil 1995;17: Carter GM, Buchanan JL, Buntin MB, et al. Executive summary of analyses for the initial implementation of the inpatient rehabilitation facility prospective payment system. Santa Monica: RAND Health; 2002.