Missing Warfarin Discharge Communication and Risk for 30-Day Rehospitalization, Death: A Retrospective Cohort Study

To the Editor

Warfarin is a high-risk medication routinely utilized in stroke and hip fracture patients discharged from hospitals to sub-acute care facilities. Previous research has shown that key informational components for high risk medications frequently are missing in discharge communication between hospitals and sub-acute care facilities.^1,2 The impact of suboptimal discharge communication regarding warfarin monitoring instructions for this population is not known. We sought to determine how often warfarin discharge communication components are missing and how this absence impacted 30-day rehospitalization and/or death in this vulnerable population.

Methods

We analyzed data from Medicare beneficiaries with primary diagnosis of stroke or hip fracture discharged on warfarin therapy from one of two large Midwestern hospitals to a sub-acute facility (i.e., nursing home, rehabilitation, skilled-nursing facility) from 2004-2008 (N=376). We used the International Classification of Diseases, 9^th edition (ICD-9) diagnosis codes of 431, 432, 434, and 436 to identify stroke; and 805.6, 805.7, 806.6, 806.7, 808, and 820 to identify hip fractures. Discharge orders/summaries were abstracted from each patient's discharge summary by two trained medical abstractors utilizing a standardized abstraction protocol, manual, and form.³ A 10% random re-abstraction of abstracted records yielded a 95% inter-rater agreement. Expert-recommended warfarin discharge communication components were abstracted for each patient: a) instructions for timing of next international normalized ratio (INR) check, b) INR goal, c) specific timeframe for warfarin therapy duration, and d) responsible party for warfarin monitoring. The primary explanatory variable was an indicator variable for the presence of at least one of the four warfarin discharge communication components or absence of all communication components.

Outcomes (all cause 30-day rehospitalization and/or death) and control variables (sociodemographics, comorbidities, and disease severity measures) were created using Medicare claims data and linked to abstracted information at the patient-level. Multivariable logistic regression with robust variance estimates clustered by discharging hospital was used to estimate the association between missing warfarin discharge communication components and patient 30-day outcomes adjusting for patient characteristics including: patient age at index hospitalization, gender, Medicaid status, index hospital length of stay, presence of mechanical ventilation or gastrostomy tube placement during index hospitalization, Centers for Medicare and Medicaid Services Hierarchical Condition Category (CMS-HCC) score as a measure of utilization risk adjustment,⁴ and presence of comorbidities (with the exception of dementia) using methods proposed by Elixhauser et al.⁵ We used a 2-sided P < .05 to establish statistical significance and report risk ratios⁶ with 95% confidence intervals (CI). Statistical analyses were conducted using Stata version 13.1. The University of Wisconsin-Madison Institutional Review Board approved the study and waived informed consent.

Results

Overall, the mean age of beneficiaries included in the analyses was 80.8 years, 66% were women, 2.1% were non-white, 9.3% Medicaid enrolled, and mean length of index hospitalization was 7.0 days. Eighty-nine beneficiaries (23.7%) had all warfarin discharge communication components missing from their hospital discharge communication. Frequency of individual warfarin discharge communication components missing included: instructions for timing of next INR check (n=157/376 missing, 41.8%); INR goal (n=216, 57.4%); specific timeframe for warfarin therapy (n=300, 79.8%); and responsible party for warfarin monitoring (n=184, 48.9%). Patients with all warfarin discharge communication components missing were at greater risk of 30-day rehospitalization and/or death post-discharge than those with any component present (27.0% vs. 11.1%, p<0.001; unadjusted risk ratio [RR] 2.41, 95% CI 2.20-2.64; adjusted RR 2.61, 95% CI 2.37-2.85) (Figure 1). The majority of these events were rehospitalizations (n=45, 12.0%).

Figure 1. Missing Warfarin Discharge Communication Components^a and Risk of 30-Day Rehospitalization, Death.

CI = confidence interval; RR = risk ratio

^a Instructions for timing of next international normalized ratio (INR) check; INR goal; specific timeframe for warfarin therapy duration; and responsible party for warfarin monitoring

^b All models used logistic regression methods to assess the relationship between missing warfarin discharge communication components and 30-day rehospitalization and/or death as compared to patients with any warfarin discharge communication components present to produce risk ratios⁵

^c Adjusted for Hierarchical Condition Category (HCC) score; indicator variables denoting the presence of comorbid conditions created from Elixhauser (1998) methods (with the exception of dementia) included chronic pulmonary disease, congestive heart failure, dementia, deficiency anemia, depression, diabetes uncomplicated, diabetes complicated, fluid and electrolyte disorders, hypertension, hypothyroidism, neurodegenerative disorders, peripheral vascular disorders, psychoses, renal failure, rheumatoid arthritis/ collagen vascular diseases, solid tumor without metastasis, valvular disease, weight loss, and other comorbidities; discharge diagnosis (stroke or hip fracture); length of stay of the index hospitalization; patient demographic characteristics (including age, sex, and race [non-white]), Medicaid status; mechanical ventilation or G-tube during hospitalization.

Discussion

We found that warfarin discharge communication components frequently were missing and this omission greatly increased the risk of 30-day rehospitalization and/or death. For sub-acute care patients, written hospital discharge communication is the primary, and often only method used to communicate the patient's care plan.¹ Ongoing efforts focused on improving discharge communication need to specifically target information transfer regarding warfarin monitoring, particularly among high risk populations such as stroke and hip fracture patients. Study limitations were present. Two hospitals were included in the analyses, limiting generalizability. Research is needed in additional hospital settings to determine if these effects hold. Second, this study only included stroke and hip fracture patients and results may not apply to other diagnoses; however, these are the most common discharge diagnoses to sub-acute care. Third, verbal communications between hospitals and sub-acute care facilities were not captured; previous research suggests that sub-acute care nurses obtain vast majority of the care plan from written discharge communication.¹ Our findings underscore critical shortcomings in several areas of warfarin discharge communication, which may lead to suboptimal or inappropriate anticoagulation monitoring and adversely impact patient safety.