Advertisement

Apparent Changes in Inpatient Rehabilitation Facility Outcomes Due to a Change in the Definition of Program Interruption

      Abstract

      Deutsch A, Granger CV, Russell C, Heinemann AW, Ottenbacher KJ. Apparent changes in inpatient rehabilitation facility outcomes due to a change in the definition of program interruption.

      Objective

      To describe changes in inpatient rehabilitation facility (IRF) outcomes due to the program interruption definitional change, from 30 days to 3 days, in 2002.

      Design

      Secondary data analysis of the Uniform Data System for Medical Rehabilitation (UDSMR) database.

      Setting

      Four hundred eleven IRFs that submitted data to the UDSMR database in each of the years 1998 through 2003.

      Participants

      Patient assessment data for 772,584 Medicare fee-for-service beneficiaries.

      Interventions

      None.

      Main Outcome Measures

      The number of IRF patient discharges, percent of IRF patients discharged to the community, percent of IRF patients discharged to acute care, percent of IRF patients with program interruptions, percent of IRF inpatient deaths, and average IRF length of stay (LOS).

      Results

      IRF outcomes appeared to change because of the program interruption redefinition, with changes varying by impairment group. The largest changes due to the redefinition occurred for patients with traumatic spinal cord injury, including the largest percentage increase in patients (5.16%), the largest decrease in program interruptions (5.14%), the largest increase in acute care discharges (5.04%), and the largest mean decrease in LOS (1.27d). Community discharge showed the largest decrease for patients with Guillain-Barré syndrome (4.03%).

      Conclusion

      The change in the definition of program interruptions creates the appearance of changes in IRF performance and is important to consider when comparing the preprospective payment system (PPS) and PPS assessment data.

      Key Words

      List of Abbreviations:

      IRF (inpatient rehabilitation facility), IRF-PAI (Inpatient Rehabilitation Facility Patient Assessment Instrument), LOS (length of stay), PPS (prospective payment system), SCI (spinal cord injury), UDSMR (Uniform Data System for Medical Rehabilitation)
      THE PATIENT ASSESSMENT instrument used by IRFs to collect patient outcome data was modified when Medicare implemented the PPS for IRFs in 2002.
      Guide to the Uniform Data Set for Medical Rehabilitation (Version 5).
      Inpatient Rehabilitation Facility Patient Assessment Instrument (IRF-PAI) training manual.
      One of the changes that the Medicare program made to the IRF dataset was the definition of a program interruption. Before 2002,
      Guide to the Uniform Data Set for Medical Rehabilitation (Version 5).
      a program interruption was defined as a transfer from an IRF to an acute care unit for a period of 1 to 30 days and return to the same IRF to complete the rehabilitation program for the same impairment. In January 2002, with the introduction of the IRF-PAI,
      Inpatient Rehabilitation Facility Patient Assessment Instrument (IRF-PAI) training manual.
      the length of the interruption was limited to 3 calendar days (1 or 2 overnights). Therefore, any transfer and return of 3 nights or longer is a separate admission. Consequently, in the PPS data, patients with interruptions that last 4 to 30 days have 2 short IRF stays instead of 1 longer combined IRF stay. These additional stays in the PPS data result in facility and national data showing a lower percent of patients discharged to the community, a shorter mean LOS, and a higher percentage of patients discharged to acute care. Comparisons of pre-PPS and PPS data may show worse performance measures (ie, outcomes) in the PPS data; however, this difference may reflect, in part, changes in definitions rather than true changes in performance.
      This study quantifies the impact of the program interruption definitional change on IRF performance measures for patients by rehabilitation impairment category. We hypothesized that changing the program interruption length from 30 days to 3 days would create the appearance that (1) the number of patients treated increases, (2) the percent of patients discharged to the community decreases, (3) the percent of patients discharged to acute care increases, (4) the percent of patients with program interruptions decreases, (5) the percent of inpatients who died decreases slightly, and (6) the mean LOS decreases. We also hypothesized that the impact of the definition change would vary by impairment group because the percent of patients with program interruptions varies by impairment group.
      • Deutsch A.
      • Fiedler R.C.
      • Granger C.V.
      • Russell C.F.
      The Uniform Data System for Medical Rehabilitation report of patients discharged from comprehensive medical rehabilitation programs in 1999.

      Methods

      Study Design and Data Source

      Secondary analysis of existing data that compared data collected before and after the IRF dataset was modified for IRF-PPS implementation. The patients represent a convenience sample of 772,584 Medicare fee-for-service beneficiaries who were discharged from an IRF that submitted data to the UDSMR in the years 1998 through 2003. Data were from 411 IRFs that submitted data to UDSMR for each of the 6 years (1998–2001 pre-PPS dataset and 2002–2003 PPS dataset). IRFs that contributed data each year were selected because this analysis is part of a larger study that examines the impact of the IRF-PPS.

      Dependent Variables

      The 6 dependent variables for this study were (1) number of discharges, (2) percent of patients discharged to the community, (3) percent of patients discharged to acute care, (4) percent of patients with 1 or more program interruptions, (5) percent of inpatients who died, and (6) the mean rehabilitation LOS.

      Data Analyses

      We calculated each dependent variable by calendar year for patients by rehabilitation impairment category by using the dataset in use that year (ie, the pre-PPS definition of program interruptions was used for data from 1998 through 2001, and the PPS definition was used for the years 2002 and 2003). The data from the 4 pre-PPS years, 1998 through 2001, were then recalculated after records of patients with a program interruption that lasted 4 to 30 days were redefined as 2 separate patient records. Splitting these records made the program interruption definitions equivalent across the 6 years and consistent with the PPS definition. Differences between the unadjusted and adjusted pre-PPS data were tested by using chi-square (categoric data) and t tests (continuous data), with P less than .01 considered statistically significant.

      Results

      Table 1 presents the unadjusted and the adjusted data by impairment group. Results for patients with stroke, 1 of the largest groups of patients, are described first, as an example. For patients with stroke, the unadjusted number of records for patients decreased in the pre-PPS years, from 30,682 in 1998 to 28,800 in 2001, and then increased to 30,284 in 2002 and 30,531 in 2003. When the pre-PPS data were adjusted for the program interruption definitional change (to be consistent with the PPS definition), the number of pre-PPS records for patients with stroke increased by an average of 843 records, or 2.81%, each year.
      Table 1Number of Records, Community Discharge, Acute Care Discharge, Program Interruptions, Deaths, and LOS by Rehabilitation Impairment Category: Unadjusted and Program Interruption-Adjusted Data
      Dependent VariablesYearMean Change
      Mean change of pre-PPS data because of the PI change.
      199819992000200120022003
      Stroke
       Number of records
        Number of patient records30,68230,61729,84028,80030,28430,531
        PI-adjusted total number of patient records31,55531,45930,70529,591
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      873842865791843 (2.81%)
       Community discharge
        % of patients discharged to the community71.4971.6672.1171.9369.0467.82
        PI-adjusted % of patients discharged to the community69.51
      Change with PI adjustment = unadjusted – adjusted.
      69.74
      Change with PI adjustment = unadjusted – adjusted.
      70.08
      Change with PI adjustment = unadjusted – adjusted.
      70.01
      Change with PI adjustment = unadjusted – adjusted.
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −1.98−1.92−2.03−1.92−1.96
       Acute care discharge
        % of patients discharged to acute care5.685.675.736.109.3710.44
        PI-adjusted % of patients discharged to acute care8.29
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      8.20
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      8.39
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      8.61
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      2.612.532.662.512.57
       Program interruptions
        % of patients with program interruptions3.383.323.403.230.910.87
        PI-adjusted % with program interruptions0.70
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.72
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.65
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.63
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −2.68−2.60−2.75−2.60−2.65
       Inpatient deaths
        % of patients who died0.380.390.300.390.280.24
        PI-adjusted % who died0.370.380.300.38
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −0.01−0.01−0.01−0.01−0.01
       LOS
        LOS (d)19.1618.7218.2217.8416.6816.36
        PI-adjusted LOS (d)18.63
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      18.22
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      17.71
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      17.36
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −0.53−0.50−0.51−0.48−0.51
      Traumatic SCI
       Number of records
        Total number of patient records7647577988431,071995
        PI-adjusted total number of patient records809798847889
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      4541494645 (5.16%)
       Community discharge
        % of patients discharged to the community71.0772.1371.6869.7566.2064.02
        PI-adjusted % of patients discharged to the community67.1268.4267.5366.14
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −3.95−3.71−4.15−3.61−3.85
       Acute care discharge
        % of patients discharged to acute care6.816.875.898.1914.4714.87
        PI-adjusted % of patients discharged to acute care11.99
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      11.65
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      11.33
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      12.94
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      5.184.785.444.755.04
       Program interruptions
        % of patients with program interruptions6.546.347.026.291.121.11
        PI-adjusted % with program interruptions1.24
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      1.50
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      2.01
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.90
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −5.30−4.84−5.01−5.39−5.14
       Inpatient deaths
        % of patients who died0.391.060.630.360.470.10
        PI-adjusted % who died0.371.000.590.34
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −0.02−0.05−0.04−0.02−0.03
      Traumatic SCI (cont'd)
       LOS
        LOS (d)23.5722.9624.7722.5919.5118.22
        PI-adjusted LOS (d)22.24
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      21.78
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      23.37
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      21.42
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −1.33−1.18−1.40−1.17−1.27
      Replacement, lower extremity
       Number of records
        Total number of patient records26,74328,07331,92036,01941,79344,833
        PI-adjusted total number of patient records27,03228,38032,23736,353
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      289307317334312 (0.89%)
       Community discharge
        % of patients discharged to the community92.2792.5593.0593.3392.9092.94
        PI-adjusted % of patients discharged to the community91.28
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      91.55
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      92.14
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      92.47
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −0.99−1.00−0.91−0.86−0.94
       Acute care discharge
        % of patients discharged to acute care1.791.651.841.742.973.04
        PI-adjusted % of patients discharged to acute care2.84
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      2.71
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      2.80
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      2.65
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      1.051.060.960.911.00
       Program interruptions
        % of patients with program interruptions1.251.331.201.140.310.34
        PI-adjusted % with program interruptions0.21
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.27
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.26
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      0.25
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −1.04−1.06−0.94−0.89−0.98
       Inpatient deaths
        % of patients who died0.060.060.030.040.020.03
        PI-adjusted % who died0.060.060.030.04
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      0.000.000.000.000.00
       LOS
        LOS (d)10.179.919.709.348.668.49
        PI-adjusted LOS (d)10.06
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      9.80
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      9.60
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      9.25
      The unadjusted and PI-adjusted data are significantly different (P<.01).
        Change with PI adjustment
      Change with PI adjustment = unadjusted – adjusted.
      −0.11−0.11−0.10−0.090.10
      NOTE: The data results for patients in the other impairment groups are available from the primary author.
      Abbreviations: PI, program interruption.
      low asterisk Mean change of pre-PPS data because of the PI change.
      Change with PI adjustment = unadjusted – adjusted.
      The unadjusted and PI-adjusted data are significantly different (P<.01).
      The unadjusted data show that the percentage of patients with stroke discharged to the community was fairly constant in pre-PPS years, between 71.49 and 72.11, but decreased by 2.89% between 2001 and 2002. When the pre-PPS data were adjusted for the program interruption definitional change, the percent of patients discharged to the community decreased between 1.92% and 2.03%, so that instead of decreasing 2.89% from 2001 to 2002 (71.93–69.04=2.89), the difference was 0.97% (70.01–69.04=0.97). The difference between the unadjusted and adjusted pre-PPS data was statistically significant each year.
      The unadjusted percentage of discharges to acute care varied only slightly in the pre-PPS years, from a low of 5.67% in 1999 to a high of 6.10% in 2001, but increased to 9.37% in 2002 and 10.44% in 2003. When adjusted for the program interruption definitional change, the pre-PPS acute-care discharge percentages increased an average of 2.57%. Therefore, between 2001 and 2002, the acute-care discharge percent increased 0.76% (9.37–8.61) rather than 3.27% (9.37–6.10).
      Program interruptions occurred for 3.23% to 3.40% of patients with stroke in the pre-PPS years and decreased to 0.91 and 0.87 in 2002 and 2003, respectively. When the pre-PPS data were adjusted for the program interruption definitional change, the percentage decreased an average of 2.65%. The adjustment shows that instead of decreasing with PPS implementation, program interruptions actually increased slightly from 2001 to 2002 (.63–.91 rather than 3.23–.91).
      The percent of patients with stroke who died in the IRF was small. The addition of the new (readmission) records under the PPS dataset had a very small effect on the calculation of the percent of inpatients who died.
      The unadjusted LOS for patients with stroke decreased from 17.84 in 2001 to 16.68 in 2002, a difference of 1.16 days. The adjusted data show that once the program interruption redefinition is considered, LOS decreased by 0.68 days (17.36–16.68). The program interruption adjustment accounted for an average decrease of 0.51 days across the 4 pre-PPS years.
      The magnitude of the impact of the program interruption redefinition varied by impairment group. The largest percentage increase in the number of patients caused by the program interruption definitional change occurred with the records of patients with a traumatic SCI (n=45 or 5.16%), and the smallest increase occurred for patients with lower-extremity joint replacement (n=312 or 0.89%) and patients with burns (n=1 or 0.89%, data not shown). Community discharge showed the largest decrease for patients with Guillain-Barré syndrome (4.03%, data not shown in table), and the smallest decrease (change of <1%) occurred for patients with lower-extremity joint replacement(s) and patients with burns (data not shown). Patients with a traumatic SCI had the largest increase in discharges to acute care (5.04%), the largest decrease in program interruption percentages (5.14%), and the largest decrease in LOS (1.27d).

      Discussion

      In the 411 IRFs, changing the program interruption length from 30 days to 3 days gives the appearance that when analyzed by impairment group (1) the number of patients treated at the hospital increased by between 0.89% and 5.16%, (2) the percent of patients discharged to the community decreased by between 0.71% and 4.03%, (3) the percent of patients discharged to acute care increased by between 1% and 5%, (4) the percent of patients with program interruptions decreased by between 0.89% and 5.14%, (5) the percent of inpatients who died decreased only slightly, and (6) the mean LOS decreased by between 0.10 and 1.27 days.
      Studies examining the impact of the IRF-PPS by comparing pre-PPS and PPS data must adjust for modifications made to the IRF assessment instrument, including the change made to the program interruption definition. This study documents 1 change to the IRF dataset when it became the IRF-PAI. Other modifications
      • Granger C.V.
      • Deutsch A.
      • Russell C.
      • Black T.
      • Ottenbacher K.J.
      Modifications to the FIM(TM) instrument under the inpatient rehabilitation prospective payment system.
      as well as documentation changes must be considered to estimate the true effect of the IRF-PPS on admission practices (ie, access to IRF care) and patient outcomes.

      Study Limitations

      Limitations of this study include the use of data from only 411 IRFs, about 33% of the IRFs in the United States in 1998. Study IRFs, when compared with all IRFs, included slightly fewer freestanding facilities (14.6% vs 18.3%) and a higher percentage of not-for-profit facilities (75.2% vs 59.6%). The geographic distribution showed a slightly higher percentage in the East North Central and South Atlantic regions and slightly lower percentages in the West North Central, East North Central, and West South Central regions. In addition, only the records of Medicare fee-for-service patients were analyzed.
      The pre-PPS data were adjusted by diagnosis for the impact of the program interruption redefinition; data were not adjusted for other factors, such as age, functional status and comorbidities. Some differences in outcomes across years may be explained by these factors.

      Conclusions

      The program interruption definition change gave the appearance of increasing the number of discharges, decreasing the percent of patients discharged to the community, increasing the percent of patients discharged to acute care, decreasing the percent of patients with program interruptions, and decreasing the average LOS. The impact varied by impairment group. The definition change is important to consider when comparing pre-PPS and PPS assessment data.

      Acknowledgments

      The authors thank Paulette Niewczyk, MPH, PhD, and James Graham, PhD, DC, who provided technical assistance with this work. Thanks also to the many clinicians who collected the FIM data used in this study.

      References

      1. Guide to the Uniform Data Set for Medical Rehabilitation (Version 5).
        State University of New York at Buffalo, Buffalo1997
      2. Inpatient Rehabilitation Facility Patient Assessment Instrument (IRF-PAI) training manual.
        Uniform Data System for Medical Rehabilitation, Buffalo2001
        • Deutsch A.
        • Fiedler R.C.
        • Granger C.V.
        • Russell C.F.
        The Uniform Data System for Medical Rehabilitation report of patients discharged from comprehensive medical rehabilitation programs in 1999.
        Am J Phys Med Rehabil. 2002; 81: 133-142
        • Granger C.V.
        • Deutsch A.
        • Russell C.
        • Black T.
        • Ottenbacher K.J.
        Modifications to the FIM(TM) instrument under the inpatient rehabilitation prospective payment system.
        Am J Phys Med Rehabil. 2007; 86: 883-892