Hospitalized Older Adults Anil Makam, MD, MAS 1 ; Oanh Nguyen, MD, - - PowerPoint PPT Presentation

1Assistant Professor

Division of GIM; Division of Outcomes & HSR UT Southwestern Medical Center

Predictors & Variation in LTAC Use among Non-Mechanically Ventilated Hospitalized Older Adults

anil.makam@utsouthwestern.edu | @anilmakam

Anil Makam, MD, MAS1; Oanh Nguyen, MD, MAS; Lei Xuan, PhD; Mike Miller, MS; Ethan Halm, MD, MPH

FUNDING DISCLOSURE

*No conflicts of interest to disclose @AnilMakam

• Long-term acute care hospitals (LTACs)

specialize in caring for patients with serious & complex illness who need inpatient care for an extended period of time (3-5 weeks)

• Most expensive post-acute care provider
• LTACs are most effective for ventilation weaning,

but fewer than 25% of LTAC patients receive prolonged mechanical ventilation

BACKGROUND

@AnilMakam

KNOWLEDGE GAP

• Despite providing an overlapping level of care

with acute care hospitals, unclear why non- mechanically ventilated older adults are transferred to LTACs

@AnilMakam

RESEARCH QUESTIONS

• 1. What are the patient, hospital, and region-level

predictors of LTAC transfer vs staying in the acute-care hospital?

• 2. How much of the variation in LTAC use is

explained by each of these 3 levels?

@AnilMakam

• Study Design: Observational cohort study
• Data: 5% national Medicare data

– Hospital: CMS Provider of Services file – Regions: Dartmouth Atlas

METHODS

@AnilMakam

PARTICIPANTS

Non-mechanically ventilated hospitalized older adults in FY2012 with continuous Medicare A&B (no part C) in prior 12 months Transferred to LTAC Stay in the Hospital

Prolonged hospitalization without a post-acute care transfer, defined as having LOS ≥ average hospital LOS among those transferred to LTAC

Restricted to the top 50 most common diagnoses leading to LTAC transfer (~2/3rds of LTAC population)

• Predictors: multilevel mixed effects model to

account for clustering of patients in hospitals, and hospitals in regions

• Variation: sequential multilevel models

– Variance partition coefficients (VPCs)

• Captures residual variation explained at each level

– Adjusted region and hospital LTAC transfer rates

• Heat map & variation profile plots

– Intraclass correlation coefficient (ICC)

• Assesses how similar hospitals are within regions

ANALYSIS

@AnilMakam

CHARACTERISTICS

LTAC Stay in Hospital N=1831 N=11,044 Age ≥85 years 24% 28% Female 55% 56% White 75% 81% Charlson index, median 5 4 Hospital LOS, median 10 12 ICU stay ≥ 3 days 50% 42% @AnilMakam

PREDICTORS

Adjusted OR (95% CI) Patient* Prior LTAC stay 1.77 (1.36-2.31) Wheelchair use 1.30 (1.10-1.53) Respiratory diagnosis 0.71 (0.60-0.85) Circulatory diagnosis 0.65 (0.54-0.78) Central line 1.57 (1.34-1.83) Excisional debridement 2.05 (1.31-3.22) Hospital For-profit hospital 1.58 (1.27-1.95) Urban location 0.54 (0.39-0.73) Region* Distance to LTAC, <1.4 vs >34 miles 6.18 (4.21-9.09) Medicare spending, per $3000 1.48 (1.06-2.06) Median income, per $10,000 0.88 (0.79-0.97) *selected list

RESULTS: VARIATION

• After accounting for differences in patient case-mix, differences

between hospitals (15%) & regions (28%) account for nearly half of why older adults are transferred to an LTAC

• Three-quarters of regional variation is explained by the region-

level predictors included in the model (28%7%)

VPCs Case-Mix Only Full Model Between patients 56.9% 74.5% Between hospitals 15.4% 18.4% Between regions 27.8% 7.1% @AnilMakam

RESULTS: VARIATION

• After accounting for differences in patient case-mix, differences

between hospitals (15%) & regions (28%) account for nearly half of why older adults are transferred to an LTAC

• Three-quarters of regional variation is explained by the region-

level predictors included in the full model (28%7%)

VPCs Case-Mix Only Full Model Between patients 56.9% 74.5% Between hospitals 15.4% 18.4% Between regions 27.8% 7.1% @AnilMakam

ADJUSTED LTAC TRANSFER RATE BY REGION

*Adjusted for differences in case-mix @AnilMakam

HOSPITAL VARIATION IN LTAC USE

*Adjusted for patient, hospital, and region-level predictors *n=1038 hospitals with ≥5 patients

% Transferred to LTAC

Median = 7% (IQR, 3%-18%)

Hospitals Sorted in Ascending Order of Adjusted LTAC Transfer Rate

@AnilMakam

• Hospitals within the same region are only modestly

correlated (ICC 0.25) with one another with respect to LTAC use

HOSPITAL CORRELATION

@AnilMakam

HOSPITAL VARIATION BY REGION

@AnilMakam

• May have omitted important patient-level predictors
• f LTAC use due to use of administrative data

– VPCs still robust since capture unobserved differences

LIMITATIONS

@AnilMakam

• Nearly half of variation in LTAC use is unrelated to

patients’ illness severity or preference

• Regional differences explain over a quarter of

variation, with far greater use in the South

• Considerable hospital variation, even among

hospitals in same regions with same LTAC access

CONSLUSIONS

@AnilMakam

• More than one place to recover from serious illness

among non-mechanically ventilated older adults

– LTAC transfer may improve care by more focus on interdisciplinary rehab, but could increase fragmentation – LTAC transfer generates a separate bundled payment, so more expensive for Medicare & ACOs

• Need for comparative effectiveness research of

LTAC vs hospital care for recovery & costs

IMPLICATIONS

@AnilMakam

anil.makam@utsouthwestern.edu @anilmakam

QUESTIONS?