Utilization Frailty and Association with 30-day Readmission and Patient Mortality Under the Hospital Readmissions Reduction Program - Part I
Part I | Part II
Editor's Note: Commentary based on Kundi H, Wadhera RK, Strom JB, et al. Association of Frailty With 30-Day Outcomes for Acute Myocardial Infarction, Heart Failure, and Pneumonia Among Elderly Adults. JAMA Cardiol 2019;4:1084-91.
Rationale for the study: The Centers for Medicare and Medicaid Services (CMS) financially penalize hospitals with excess short-term readmissions for acute myocardial infarction (AMI), heart failure (HF), and pneumonia (PNA). Current risk adjustment algorithms do not incorporate frailty metrics and the impact of adding a claims-based frailty metric to traditional comorbidity-based risk adjustment is unknown.
Study design: Retrospective cohort analysis of 785,127 Medicare Fee-for-Service beneficiaries.
Inclusion Criteria: Older aged adults (≥65 years old), enrolled in a Medicare Fee-for-Service plan for 3 months prior to index hospitalization and 1 month after discharge for alive patients, hospitalized at an acute care hospital from January 1, 2016 to December 1, 2016 with a principal discharge diagnosis of AMI, HF, or PNA.
- Patients not meeting the above criteria.
- Patients with missing admission or discharge dates.
- Patients discharged against medical advice (AMA) or admitted and discharged for AMI on the same day.
Exposure: Frailty was identified using the validated Hospital Frailty Risk Score (HFRS), based on present-on-admission flags and discharge diagnoses for hospitalizations within 3 months of the index hospital admission. Individuals were categorized into low (<5), intermediate (5-15), and high (>15) risk frailty groups based on the calculated HFRS using previously validated cut points.
Primary outcome: All-cause mortality within 30-days of index hospital date of admission, obtained from linkage to the 2016 Medicare Master Beneficiary Summary File.
Secondary outcomes: Long length of stay (>10 days in the hospital), all-cause mortality within 30-days of the index hospital date of discharge, and readmission within 30 days.
Statistics: Multivariable logistic regression models (adjusted for age, race, and comorbidities) were constructed to model the relationship of categorical frailty levels and primary and secondary outcomes. Model concordance statistics (C-statistics) were compared between models including and not including the HFRS using the DeLong test, and the integrated discrimination improvement assessed. Restricted cubic spline regression was used to model the relationship between the HFRS and 30-day outcomes. Sensitivity analyses were conducted using the HFRS as a continuous measure.
Results: Of 785,127 hospitalizations (83.6% non-Hispanic Caucasian):
- 21.2% (166,200) hospitalized with AMI (mean [SD] age = 77.4 [8.7] years; 44.5% female).
- 44.4% (348,619) hospitalized with HF (mean [SD] age = 80.1 [9.0] years; 52.7% female).
- 34.4% (270,308) hospitalized with PNA (mean [SD] age = 79.2 (8.9) years; 53.6% female)
- Mean (SD) HFRS = 7.3 (7.4) for AMI; 10.8 (8.3) for HF; 8.2 (5.7) for PNA.
- 13.9% of AMI, 25.0% of HF, and 11.5% of PNA hospitalizations were for those with HFRS > 15.
Patients with higher frailty scores had worse 30-day outcomes and longer length of stay. Compared to those with the lowest level of frailty (HFRS < 5), those with the highest level of frailty (HFRS >15) had 3.6x the risk of 30-day post-admission mortality, 4x the risk of post-discharge mortality, and 3x the risk of 30-day readmission (p < 0.001 for all).
Addition of the HFRS to traditional comorbidities significantly enhanced discrimination of all 3 post-discharge outcomes:
30-day post-admission mortality
- AMI - C-statistic without vs. with HFRS: 0.73 vs. 0.76; p < 0.001
- HF - C-statistic without vs. with HFRS: 0.67 vs. 0.70; p < 0.001
- PNA C-statistic without vs. with HFRS: 0.70 vs. 0.73; p < 0.001
30-day post-discharge mortality
- AMI - C-statistic without vs. with HFRS: 0.76 vs. 0.78; p < 0.001
- HF - C-statistic without vs. with HFRS: 0.68 vs. 0.70; p < 0.001
- PNA - C-statistic without vs. with HFRS: 0.69 vs. 0.71; p < 0.001
- AMI - C-statistic without vs. with HFRS: 0.65 vs. 0.68; p < 0.001
- HF - C-statistic without vs. with HFRS: 0.61 vs. 0.64; p < 0.001
- PNA - C-statistic without vs. with HFRS: 0.60 vs. 0.63; p <0.001
- Limited to Medicare Fee-for-Service beneficiaries
- Administrative codes don't provide granularity on severity of conditions
- As the HFRS was developed based on clusters of increased resource utilization, it only correlates moderately with phenotypic definitions of frailty (e.g. the Fried and Rockwood indices).
Conclusion: The Hospital Frailty Risk Score was strongly associated with short-term mortality and readmissions for AMI, HF, and PNA among Medicare Fee-for-Service beneficiaries. The addition of this score to standard risk prediction algorithms improved identification of individuals at high risk of adverse short-term outcomes. Hospitals taking care of patients with larger numbers of frail individuals may be unduly penalized if frailty is not considered in risk adjustment models.
Please see accompanying Expert Analysis by Drs. Martin and Strom.
- Kundi H, Wadhera RK, Strom JB, et al. Association of Frailty With 30-Day Outcomes for Acute Myocardial Infarction, Heart Failure, and Pneumonia Among Elderly Adults. JAMA Cardiol 2019;4:1084-91.
Keywords: Geriatrics, Patient Readmission, Fee-for-Service Plans, Risk Adjustment, Centers for Medicare and Medicaid Services (U.S.), Length of Stay, Retrospective Studies, Hemorrhagic Fever with Renal Syndrome, Medicare, Patient Discharge, Hospitalization, Magnetite Nanoparticles, Heart Failure, Pneumonia, Myocardial Infarction, Comorbidity, Algorithms
< Back to Listings