Stroke Readmission Risk Factor Study Pam Roberts, PhD, CHPQ Maggie DiVita, MS Paulette Niewczyk, PhD Richard Riggs, MD UDSMR Annual Conference August 2012 2012 Uniform Data System for Medical Rehabilitation. FIM, UDSMR, and the UDSMR logo are trademarks of Uniform Data System for Medical Rehabilitation, a division of UB Foundation Activities, Inc.
Readmissions The Medicare Payment Advisory Committee (MedPAC) estimates that up to 80.4% of readmissions may be preventable, representing a potential savings to Medicare of over $12 billion in one year 17.6% of admissions result in readmissions within thirty days 6% result in readmissions within seven days Results in $15 billion in savings to Medicare in one year 2
Readmissions According to the most recent volume of the Post- Acute Care Demonstration Final Report, 17.4% of inpatient rehabilitation admissions result in readmissions within thirty days Reference: Post Acute Care Demonstration Report, Volume 4, March 2012 3
Readmissions According to CMS (2009), nearly one in five patients who is discharged from the hospital will be readmitted within the month (thirty days), and more than three-quarters of these readmissions are preventable Readmission rates have varied according to demographic, social, and disease-related characteristics 4
IRF Readmissions Year Readmission to Acute Hospital (National Average) Readmission to Acute Hospital Stroke Patients (National Average) 2008 10.7% 11.4% 2009 10.8% 11.3% 2010 10.5% 10.9% 2011 10.2% 10.4% Reference: Uniform Data System for Medical Rehabilitation National Average for Discharges to Acute Hospital 5
Readmissions CMS s focus is to reduce avoidable readmissions As readmission rates affect payment, and as PAC moves toward a bundled payment system, understanding the implications of discharge destinations as they influence outcomes and payment is imperative 6
Rehabilitation Readmissions Readmissions directly from the IRF represent revenue at risk Readmissions represent a lack of functional outcomes/optimization of rehabilitation outcomes Financial reductions Transfer reductions Loss of potential revenue due to an early transfer Payment on a per diem basis and fewer days than the expected case-mix group (CMG) length of stay (LOS) 7
Rehabilitation Readmissions Limited research has looked at the significance of specific stroke-related predictors for rehospitalization directly after inpatient rehabilitation Early identification of inpatient rehabilitation patients with associated conditions that place them at high risk for rehospitalization, combined with actionable risk reduction strategies, has the potential for a positive impact on rehabilitation healthcare 8
Study Aims Aims: 1. To identify medical risk factors, such as specific complication diagnoses, for readmitting IRF stroke patients to an acute care hospital 2. To identify functional health risk factors, such as average admission FIM motor and cognitive ratings, for readmitting IRF stroke patients to an acute care hospital 9
Study Design Design: Retrospective study design Participants: Stroke population at one academic medical center IRF from 2006 to 2010 (total population for study, n = 222) Seventy-four patients were discharged to acute care during this time frame Seen only once in the IRF Used a two-to-one match on gender for each patient discharged to acute care (148 gendermatched controls) with discharge to a community setting 10
What Is An Odds Ratio? An odds ratio is the measure of the odds of an event happening in one group compared to the odds of the same event happening in another group For example, the goal of our study, which examines cases discharged to acute care (cases) and cases discharged to the community (controls), is to determine how many people in each group were exposed to a certain substance or factor, such as a new infection within the IRF We calculate the odds of exposure in both groups and then compare the odds An odds ratio of 1.00 means that both groups had the same odds of exposure 11
Results: Bivariate Table 1: Bivariate analysis of the characteristics of patients Discharge to Community (n = 148) Discharge to Acute Care (n = 74) Demographics N % N % P-value* Odds Ratio** 95% Confidence Interval Ethnicity 0.741 White 112 75.7% 55 74.3% 1.00 Black 23 15.5% 14 18.9% 1.24 (0.592 2.594) Other 13 8.8% 5 6.8% 0.78 (0.266 2.308) Primary Payer 0.830 Non-Medicare 40 27.0% 19 25.7% 1.00 Medicare 108 73.0% 55 74.3% 1.07 (0.568 2.024) Pre-Hospital Living With 0.697 Alone 44 29.9% 20 27.4% 1.00 With Others 103 70.1% 53 72.6% 1.13 (0.607 2.112) * The p-value comes from the chi-squared test; the odds ratio comes from Maental-Haenszel. 12
Results: Bivariate Table 1: Bivariate analysis of the characteristics of patients Discharge to Community (n = 148) Discharge to Acute Care (n = 74) N % N % P-value* Odds Ratio** 95% Confidence Interval Stroke Impairment Type 0.446 Right Body 70 47.3% 31 41.9% 1.00 Left Body/Other 78 52.7% 43 58.1% 1.25 (0.709 2.187) Type of Stroke 0.545 Ischemic 106 71.6% 58 78.4% 1.00 Hemorrhagic 40 27.0% 15 20.3% 0.69 (0.349 1.345) Other 2 1.4% 1 1.4% 0.91 (0.081 10.294) Type of Stroke 0.332 Ischemic 106 71.6% 58 78.4% 1.44 (0.743 2.776) Other 42 28.4% 16 21.6% 1.00 * The p-value comes from the chi-squared test; the odds ratio comes from Maental-Haenszel. 13
Results: Bivariate The study did not identify any differences between the two groups (discharged to acute care and discharged to community) in the following demographic factors: Ethnicity Payer source Pre-hospital living with The study also did not identify any differences between the two groups in the type of stroke 14
Results: Bivariate Table 1: Bivariate analysis of the characteristics of patients Discharge to Community (n = 148) Discharge to Acute Care (n = 74) N % N % P-value* Odds Ratio** 95% Confidence Interval New Infection While in the IRF 0.000 No 117 79.1% 41 55.4% 1.00 Yes 31 20.9% 33 44.6% 3.04 (1.658 5.567) New Infection Type 0.000 None 117 79.1% 41 55.4% 1.00 UTI 16 10.8% 9 12.2% 1.61 (0.659 3.912) Other 11 7.4% 16 21.6% 4.15 (1.781 9.673) Unknown 4 2.7% 8 10.8% 5.71 (1.632 19.957) Infection Identified Through Comorbidity 0.194 No 92 62.2% 39 52.7% 1.00 Yes 56 37.8% 35 47.3% 1.47 (0.838 2.593) * The p-value comes from the chi-squared test; the odds ratio comes from Maental-Haenszel. 15
Results: Bivariate The study identified differences between the two groups regarding the presence of a new infection while in the IRF Those discharged to acute care are three times more likely to have a new infection while in the IRF The increased odds exist for all infection types, except for UTI The study did not identify any difference between the two groups related to infections identified through comorbidity codes The difference exists only for new infections 16
Results: Bivariate Table 1: Bivariate analysis of the characteristics of patients Discharge to Community (n = 148) Discharge to Acute Care (n = 74) N % N % P-value* Odds Ratio** 95% Confidence Interval Case-Mix Group 0.000 0101 0107 108 73.0% 28 37.8% 1.00 0108 12 8.1% 12 16.2% 3.86 (1.566 9.503) 0109 10 6.8% 5 6.8% 1.93 (0.610 6.098) 0110 18 12.2% 29 39.2% 6.21 (3.024 12.769) Case-Mix Group 0.545 0101 0107 108 73.0% 28 37.8% 1.00 0108 0110 40 27.0% 46 62.2% 4.44 (2.450 8.030) * The p-value comes from the chi-squared test; the odds ratio comes from Maental-Haenszel. 17
Results: Bivariate Table 1: Bivariate analysis of the characteristics of patients Discharge to Community (n = 148) Discharge to Acute Care (n = 74) N % N % P-value* Odds Ratio** 95% Confidence Interval CMS Comorbidity Tier 0.041 0 (A) 106 71.6% 41 55.4% 1.00 1 (B) 6 4.1% 7 9.5% 3.02 (0.957 9.511) 2 (C) 4 2.7% 6 8.1% 3.88 (1.041 14.452) 3 (D) 32 21.6% 20 27.0% 1.62 (0.831 3.142) CMS Comorbidity Tier 0.024 None 106 71.6% 41 55.4% 1.00 Any 42 28.4% 33 44.6% 2.03 (1.136 3.632) * The p-value comes from the chi-squared test; the odds ratio comes from Maental-Haenszel. 18
Results: Bivariate The study identified a difference in CMG distribution between the two groups Cases discharged to acute care are more than four times more likely to be in CMGs 0108, 0109, and 0110 The study identified a difference in tier distribution between the two groups Cases discharged to acute care are more than two times more likely to have a tiered comorbidity 19
Results: Bivariate Table 1: Bivariate analysis of the characteristics of patients Discharge to Community (n = 148) Discharge to Acute Care (n = 74) N % N % P-value* Odds Ratio** 95% Confidence Interval Admission Swallowing Status 0.000 Regular Food 65 43.9% 29 39.2% 1.00 Modified Food/ Supervision 78 52.7% 30 40.5% 0.13 (0.043 0.384) Tube Feed 5 3.4% 15 20.3% 6.72 (2.232 20.255) Admission Swallowing Status 0.000 Other 143 96.6% 59 79.7% 1.00 Tube Feed 5 3.4% 15 20.3% 7.27 (2.528 20.916) Falls 0.023 No Falls 144 97.3% 66 89.2% 1.00 Falls 4 2.7% 8 10.8% 4.36 (1.269 15.005) * The p-value comes from the chi-squared test; the odds ratio comes from Maental-Haenszel. 20
Results: Bivariate The study identified a difference in swallowing status on admission between the two groups Cases discharged to acute care are more than seven times more likely to be tube fed The study identified a difference in fall status in the IRF between the two groups Cases discharged to acute care are more than four times more likely to fall in the IRF 21
Results: Multivariate Utilized backward conditional logistic regression with all significant variables from the bivariate analysis that are known prior to discharge Total admission motor FIM rating Total admission cognitive FIM rating CMG was not added because it is highly correlated with admission FIM ratings New infection in the IRF, type Tier Swallowing status Falls 22
How Logistic Regression Models Work You must sift through several factors to find the key significant predictors of unwanted healthcare outcomes How does a logistic model work? Identifies a significant factor/predictor which, if a given person has this factor, will render the person at an increased risk for the unwanted healthcare outcome, even after controlling for other factors Unlike the bivariate odds ratio, logistic regression quantifies the independent risk (odds ratio) associated with each significant factor in the model, if all other factors were equal 23
Results: Multivariate Final Model Variable Coefficient (β) Adjusted Logistic Regression Analysis Standard Error Wald x 2 P Odds Ratio 95% Confidence Interval Admission Motor FIM Rating 39 1.00 Admission Motor FIM Rating < 39 0.832 0.341 5.961 0.015 2.30 (1.118 4.481) No Falls 1.00 Falls 1.484 0.671 4.893 0.027 4.41 (1.184 16.427) No New Infection 1.00 UTI 0.303 0.483 0.393 0.531 1.35 (0.526 3.485) Other 1.397 0.468 8.916 0.003 4.04 (1.616 10.107) Unknown 1.241 0.686 3.273 0.070 3.46 (0.902 13.272) Swallowing Status at Admission, Other 1.00 Tube Fed 1.364 0.585 5.437 0.020 3.91 (1.243 12.322) C-statistic = 0.73 24
Conclusions Higher likelihood of not completing an inpatient rehabilitation stay Greater dependence on admission for motor FIM items (lower admission motor FIM total) A fall in the IRF stay New infection during the IRF stay Swallowing problem requiring tube feeding 25
Conclusions Utilizing the logistic regression model, we can determine the associated odds of being discharged to acute care in a patient with multiple risk factors For instance, if a patient has an admission motor FIM total of less than 39, a new infection (other), and is tube fed, the combined odds ratio is 36.3 This means that this person is 36.3 times more likely to be discharged to an acute care hospital The odds would be even higher (160.3) for a patient with all significant risk factors 26
Implications for Prevention of Readmissions Improve your processes to help you determine the following: The risk level of the population being discharged The impact on readmission Related to rehabilitation diagnosis severity level Prevention of secondary complications especially infections Management of comorbidities Fall reduction strategies 27
Key Operational Factors to Prevent Readmissions Inpatient Care Processes Quality of clinical care received across inpatient stay Appropriate medication reconciliation Compliance with evidence-based care Prevention of secondary complications Effective Discharge Planning Effectiveness of discharge instructions, patient education, and family/caregiver training Comprehensive patient/family/caregiver education on medications, diet/nutrition, exercise/activity, and care plan Proper assessment of patient needs postdischarge Community referrals Post-Discharge Follow-up Extent to which patient receives necessary follow-up care after hospital stay Scheduling of follow-up physician appointments Scheduling of follow-up therapy appointments Follow-up telephone calls Appropriate referrals and integration with community resources 28