Every year, more than 20 million adult patients are discharged from hospitals across the United States.1 Typically, at least a third of them are referred for post acute care (PAC), including long-term care hospitals, inpatient rehabilitation facilities, skilled nursing facilities, and home health agencies.2 The number of patients referred to PAC is likely to increase in the near future because of recent legislative and reimbursement trends that incentivize care across the continuum rather than focus on inpatient settings.3,4
Appropriate and timely care in PAC settings results in better care outcomes, reduced costs, and higher patient satisfaction.5-8 Unfortunately, several barriers impair the potential effectiveness of PAC services. These barriers include variations in decision making by clinicians regarding who needs post acute care,9,10 stringent eligibility criteria, and misaligned financial incentives that favor inpatient settings.11 In addition, emerging evidence suggests that patients’ refusal of PAC services is a common, but underreported, barrier to timely and effective PAC. For example, several studies reported that between 6% and 27% of their participants refused PAC services.12-14 However, to our knowledge, no known studies have investigated the characteristics and outcomes of patients who refuse PAC services. To create alternative interventions and strategies when patients refuse, it is important for clinicians involved in discharge decision making (eg, physicians, nurses, social workers, physical therapists) to understand the characteristics of patients who are likely to refuse and to acknowledge the outcomes for patients who do not accept PAC services.
Recently, our team conducted studies at 2 academic medical centers using 2 evidence-based screening tools to provide decision support for discharge planners.15 The Early Screen for Discharge Planning (ESDP)16,17 is a 4-item assessment tool that identifies high-risk patients upon admission who need specialized discharge plans. The second tool, the Discharge Decision Support System (D2S2), is a 6-item assessment tool that identifies high-risk patients who should be referred to PAC.15,18
We found that of all patients offered PAC services, after being identified by the tools and/or clinicians (eg, physicians, discharge planners) as being at high risk for poor discharge outcomes, 28% refused them. Surprised by that high percentage, and recognizing that we understand very little about these patients and their outcomes, we conducted a secondary analysis to: 1) compare patients who accepted (“acceptors”) post acute care services (PAC) with those who were offered services and refused (“refusers”) in terms of their sociodemographic and clinical characteristics, quality of life, health-related problems, and unmet needs; and 2) examine the association between refusing PAC services and the risk for subsequent 30- and 60-day readmission.
This is a secondary analysis of data from a cross-sectional study designed to test 2 discharge planning risk screening tools: the ESDP and the D2S2.
Patients 55 years and older admitted to any of 8 medical units within 2 large academic medical centers in Philadelphia and New York City between March 2010 and February 2012 were eligible for the original study. Registered nurses and trained nursing and social work students served as research assistants; they approached patients and screened them for eligibility. Exclusion criteria were: patient did not speak English, had cognitive impairment, recalled 14 or fewer animals using the Animal Fluency Test,19 was on dialysis or in hospice care, died in the hospital, or had been admitted from a long-term care setting (because their post acute referral was predetermined). The study and secondary analysis were approved by the Institutional Review Boards of New York University Langone Medical Center and the Hospital of the University of Pennsylvania. In this analysis, we used all the patients from the original study who were referred by case managers to PAC services. The final sample included 495 patients for whom we had outcome data.
Sociodemographic and clinical information. At enrollment, the research team used standardized chart abstraction tools to collect sociodemographic (ie, age, race, gender, insurance) and clinical (ie, number of medications, previous healthcare utilization) information.
All Patient Refined–Diagnostic Related Group (APR-DRG). Severity of illness was measured using the APR-DRG, a valid and reliable system used for health severity adjustment by a variety of health organizations and federal/state authorities.20 APR-DRG is reported as 4 severity-of-illness subclasses (ie, minor, moderate, major, and extreme). The score is generated from the primary and secondary diagnoses and from procedure codes, age, gender, discharge date, status of discharge, and days on mechanical ventilator. These data were obtained from the hospitals’ databases after discharge.
Problems After Discharge Questionnaire–English Version (PADQ-E). The PADQ-E captures patient-reported health-related problems and unmet needs after hospital discharge. Problems are defined as troubles, worries, limitations, concerns, or difficulties experienced by patients after discharge from the hospital.21 The PADQ-E has 7 subscales—personal care, household activities, mobility, using equipment, following instructions, physical complaints, psychological complaints—that include 36 individual questions. Responders were given 5 response options for each question, ranging from “without any trouble” to “could not do it at all.” After assessing if a response indicated a change for the worse from their pre-hospitalization condition, unmet needs were assessed by noting the patient’s desire to have more assistance in performing the activity or more support or advice in dealing with physical or psychological issues. In addition, the information needs subscale of the PADQ-E includes 13 items asking the general question: “Did you feel you had enough information during the past week regarding…?” for several domains such as medications management, pain management, desirable levels of activity, etc. Each of those questions has 3 response options: “yes,” “no,” and “I don’t know.” The PADQ-E is reliable whether self-administered or completed by interview.21
Quality of Life
To assess participants’ health-related quality of life, we used the EuroQol-5 Dimensions (EQ-5D) tool, a standardized measure of health status developed by the EuroQol Group.22 By design, the EQ-5D is applicable to a wide range of health conditions, and it provides a simple descriptive profile and a single index value for health status. EQ-5D is a valid and reliable tool designed for either self-completion by respondents or for patient interviews,23 and includes these 5 dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. For each dimension, patients reported whether they had no problems, some problems, or extreme problems. EQ-5D responses were converted into a single summary index by applying a formula generated by the Agency for Healthcare Research and Quality that attaches US-specific weights to each response subscale.24
Referral to PAC Services, Acceptance or Refusal of PAC Services, Discharge Disposition, and 30/60-Day Readmissions
Patient recall by interview, chart review, and study site administrative databases provided information on discharge disposition, acceptance or refusal, reasons for refusal, and readmissions. Among the 495 patients for whom we had complete follow-up data, refusers were defined as those in the data set who were offered PAC and refused the service. Acceptors were defined as those who were offered and accepted PAC services.
Research assistants collected sociodemographic and clinical information within 24 to 48 hours of hospital admission. Quality of life and problems/unmet needs after discharge were assessed by telephone interview 7 to 14 days after each patient’s hospital discharge. Readmission data was obtained for up to 60 days after the index discharge date through patient or caregiver interviews to collect any readmissions occurring outside the study sites, and the study sites supplied reports of all readmissions occurring within.
Descriptive statistics were used to characterize the sample. We examined differences between refusers and acceptors using t tests for continuous variables, or χ2 or Fisher’s exact test for categorical variables. The alpha was set at P = .05. Then we conducted a logistic regression examining the association between refusing PAC services and 30/60-day readmission, adjusting for those factors shown significant at the .05 level in bivariate analysis. All analyses were performed using STATA, version 11 (StataCorp, College Station, Texas).