Assessing Family Caregiver Readiness for Hospital Discharge of Patients With Serious or Life‐Limiting Illness Using Electronic Health Record and Self‐Reported Data

ABSTRACT

Objective

To assess how patient and caregiver factors influence caregiver readiness for hospital discharge in palliative care patients.

Study Setting and Design

This transitional care study uses cross‐sectional data from a randomized controlled trial conducted from 2018 to 2023 testing an intervention for caregivers of hospitalized adult patients with a serious or life‐limiting illness who received a palliative care consult prior to transitioning out of the hospital.

Data Sources and Analytical Sample

Caregiver readiness was measured with the Family Readiness for Hospital Discharge Scale (n = 231). Caregiver demographic, intra‐ and interpersonal factors were self‐reported. Patient demographic, comorbidity score, and risk score for complicated discharge planning were extracted from electronic health records. Stepwise regression models estimated variance explained (r ²) in caregiver readiness for patient hospital discharge.

Principal Findings

Patient demographics and complexity were not statistically associated with caregiver readiness for discharge. Caregiver depressive symptoms, poor caregiver‐patient relationship quality, and fewer hours spent caregiving prior to hospitalization explained 29% of the variance in caregiver readiness.

Conclusions

Reliance on patient data may not be sufficient for explaining caregiver readiness for discharge. Assessing caregiver factors may be a better alternative for identifying caregivers at risk for low discharge readiness and those in need of additional support.

Trial Registration: ClinicalTrials.gov on November 13, 2017, (No. NCT03339271)

1. Introduction

For patients with serious or life‐limiting illnesses, transitioning out of the hospital is often fraught with risks for medication errors and symptom exacerbations, leading to readmissions or crucial delays in follow‐up care [1, 2]. Patients are often discharged with the assumption that family caregivers (hereafter called caregivers) can manage post‐discharge responsibilities, such as monitoring symptoms, assisting in adherence to new or modified medication or treatment regimens, and arranging follow‐up care [3].

The United States National Strategy to Support Family Caregivers [4, 5] has called for assessments of unmet needs to ensure caregivers are willing, able, and prepared for these responsibilities. Despite evidence linking caregiver readiness for discharge to reductions in subsequent readmissions and lengths of stay [6, 7, 8], common discharge planning practices do not include these assessments [9, 10, 11, 12]. Recent policies are dismantling barriers for identifying caregivers and integrating needs assessments into hospital practice: 42 U.S. states now require hospitals to identify and document caregivers in patient electronic health records (EHR) at admission and inform them of discharge plans [13, 14, 15], and Centers for Medicare and Medicaid (CMS)‐approved codes now allow for reimbursement of caregiver training and principal navigation services [16] when unmet needs are identified. Still needed, however, are the specific strategies to move from policy to practice and uniformly identify caregivers and assess their readiness for discharge [17].

A seemingly time‐ and cost‐effective approach for hospitals to consider is using readily available data from the patient's EHR to assess caregiver readiness. However, it is unclear if caregiver readiness is a function of the patient's characteristics, of caregiver characteristics, or a combination of the two. We tested whether variation in caregiver readiness is partly explained by patient demographic and complexity factors (accessible from the EHR), and further explained by caregiver factors requiring additional data collection.

2. Methods

We conducted a secondary analysis of data from a parent study conducted from 2018 to 2023—a virtual randomized controlled trial testing a video‐based transitional palliative care (TPC) intervention for rural‐dwelling caregivers. Detailed elsewhere [18], the trial was registered with ClinicalTrials.gov on November 13, 2017, and approved by the healthcare system's institutional review board (17‐005188).

Caregivers were broadly defined as those who self‐identify as the primary person who provides care and support to someone with unmet medical or care needs. Participants were caregivers of patients with serious or life‐limiting illnesses who received a palliative care consultation in one of four hospitals in the same health system in Minnesota and Wisconsin. Caregivers were enrolled pre‐discharge and after the patient's palliative care consultation. Study data, including caregiver readiness for discharge, were collected before patient discharge [19]. Data collection was not coordinated with timing for discharge planning instructions.

2.1. Patient Data

Although the study targeted caregivers, data were also extracted from patients' EHR to account for patient factors associated with care transitions. These data included demographics captured in the EHR (i.e., age, sex, race, ethnicity, insurance status, discharge disposition) and patient “complexity” variables (i.e., the early screen for discharge planning [ESDP] assessment and the Charlson Comorbidity Index [CCI]). The ESDP is an evidence‐based decision tool calculated at hospital admission to determine the likelihood of having a complex discharge plan or unmet needs, with possible scores ranging from 0 to 23. Scores > 10 prompt early coordination of discharge management. Scores are based on age, previous living situation, availability of a caregiver, walking limitations, and functional disability [20, 21, 22]. CCI scores, based on ICD‐10‐CM codes, predict one‐year mortality risk [23]. Scores range from 0 to 39, with > 5 considered high risk for mortality [24].

2.2. Caregiver Data

Caregiver data were self‐reported using electronic surveys before patient discharge and before any intervention started. Variables included caregiver readiness for discharge, demographic factors, and intra‐ and interpersonal factors.

Caregiver readiness was assessed using the Family Caregiver version of the Readiness for Hospital Discharge Scale (FAM‐RHDS), a psychometrically stable 22‐item scale for assessing caregiver perceptions of their readiness to have the patient return home after an acute hospitalization [25], with scores < 7 considered “low caregiver readiness” [26, 27]. Used elsewhere with adult caregivers and parent caregivers of hospitalized children [28, 29, 30], the FAM‐RHDS has the same domains as the patient version [31] but with wording adjusted for caregiver respondents. Mean item scores are reported, with higher scores indicating greater caregiver readiness. Cronbach's alpha reliability for the scale in this study sample was 0.91.

Caregiver intra‐ and interpersonal factors included single questions about the relationship with the patient, co‐residing with the patient, age, sex, employment, and education. Caregiver health status (“Rate your general health status” from poor to excellent); available help from others pre‐hospital admission (“Have other people, such as family members or friends, helped you care for your loved one?”); other caregiving responsibilities (“Are you also a caregiver for someone else with medical problems or disabilities?”); and number of hours providing care prior to the hospitalization (“How many hours per week do YOU provide assistance, care, supervision, or companionship to your loved one?”) were also collected.

Caregiver depression symptoms were assessed using the 10‐item Center for Epidemiological Studies Depression Scale (CES‐D) [32, 33]. Mean scores range from 0 to 30. Scores ≥ 10 indicate the presence of depression symptoms. Quality of the relationship between the patient and the caregiver was assessed using the Mutuality Scale of the Family Care Inventory [34]. Scores ranged from 1 to 4, with higher scores indicating higher‐quality relationships.

2.3. Analysis

We developed an analytical model of potentially important patient and caregiver factors associated with caregiver readiness, drawing from other published models (see Figure S1) [12, 35], to inform our analysis.

Caregivers, regardless of intervention arm, were included. The analytic sample included 231 of 429 eligible caregivers. Included were those who completed the FAM‐RHDS and for whom patient data from the EHR was available. Missing item responses were managed by replacing the missing item with the average score from the items with responses.

Mean differences in FAM‐RHDS scores by covariates in the analytic model were examined first. Our analysis approach was intended to understand factors related to FAM‐RHDS scores using the most parsimonious model with the maximum amount of variance explained (r ²). Associations of FAM‐RHDS, a continuous variable, with patient and caregiver variables were described descriptively using univariable analysis and then with a four‐step, multivariable linear regression model (Figure S1). Step 1 fit patient demographic factors in the model (i.e., age, sex, insurance status). Step 2 included patient complexity scores (i.e., ESDP and CCI scores). Step 3 included caregiver demographic factors (i.e., age, sex, employment status, education, relationship to patient, living with patient). Step 4 included caregiver intra‐ and interpersonal factors (i.e., health status, depression, caring for multiple care recipients, availability of help, relationship quality, hours caregiving per week). All models were adjusted for time from caregiver data assessment to patient discharge. Regression estimates, standard errors and variance explained (r²) were calculated.

In the initial multivariable linear regression analysis, each step was treated as a separate model‐building exercise to identify the parameter estimate and variance explained for the FAM‐RHDS. For the final model, variables with p values < 0.05 in a step were carried forward in subsequent steps regardless of their statistical significance in future steps (e.g., if p < 0.01 for age in Step 1, age was carried forward to Steps 2 through 4, regardless of significance in those steps). Sensitivity analyses were also conducted to assess the relationship between discharge disposition (discharged home = 0; discharged elsewhere = 1) and FAM‐RHDS score. Discharge disposition was not included in the main analysis because, in practice, discharge decisions likely happen after caregiver assessment. All tests were two‐sided, and analyses were performed using SAS version 9.4 (SAS Inc. Cary, NC).

3. Results

3.1. Demographic and Caregiving Variables

Table 1 shows the distribution of caregiver and patient demographic variables and caregiver self‐reported health, caregiving demands, and pre‐admission support. Patients' mean age was 66.2 years (SD 14.87; range 21–98), less than half were women (44.6%), and most were white (97%), non‐Hispanic (95.2%), and had greater than high school education (61%). The majority had Medicare insurance (53.3%), and over 70% (72.7%) lived with their caregiver. Caregivers' mean age was 57.7 years (SD: 11.8; range, 27–87). Seventy percent were women, and nearly all were white (95.2%) and non‐Hispanic (91.3%). The majority completed college or university (71.3%). Over half were working full‐ or part‐time (55.7%). The majority rated their health as very good or excellent (52.2%), and nearly one‐third had additional caregiving responsibilities (33.2%). The majority of caregivers were married to the patient (63.5%) and nearly 75% received caregiving help from others (73.8%). The median caregiving hours/week prior to hospitalization were 40. Over 70% (72.3%) were discharged from the hospital to home.

TABLE 1.

Distribution of demographic and caregiving variables and available support by patient and family caregiver (FCG) (n = 231).

Variable	Response option	Patient	FCG
Age, mean (SD)		66.2 (14.87)	57.7 (11.8)
Gender, n (%)	Female	103 (44.6%)	162 (70.1%)
Race, n (%)	White	224 (97.0%)	220 (95.2%)
Ethnicity, n (%)	Not Hispanic	220 (95.2%)	211 (91.3%)
Education, n (%) a	<High school or high school only	75 (39.1%)	33 (14.4%)
	Vocational training	28 (14.6%)	33 (14.4%)
	College	89 (46.4%)	164 (71.3%)
Insurance, n (%) b	Medicaid	20 (8.8%)	9 (3.9%)
	Medicare	121 (53.3%)	44 (19.05%)
	Private	86 (37.9%)	102 (44.2%)
	None/unknown		76 (32.9%)
FCG co‐residing in same home, n (%) b	Yes	168 (72.7%)
Patient discharged home	Yes	167 (72.3%)
Relationship to patient, n (%)	Spouse		142 (61.5%)
	Child		41 (17.8%)
	Other		48 (20.8%)
Employment, n (%) c	Full time		99 (43.4%)
	Part time		28 (12.3%)
	Retired		69 (30.3%)
	Not working for pay		32 (14.0%)
Self‐rated health, n (%) d	Poor/Fair		12 (5.2%)
	Good		98 (42.6%)
	Very good/excellent		120 (52.2%)
Caregiving hours/week prior to hospitalization, median			40 (range: 0–168)
Cares for additional person, n (%) e	Yes		75 (33.2%)
Receives caregiving help from others, n (%) e	Yes		166 (73.8%)

^a Missing data on 39 for patient; 1 for FCG.

^b Missing data on 4; participants with multiple sources of insurance were coded into mutually exclusive groups (e.g., private and Medicare = Medicare).

^c Missing data on 3.

^d Missing data on 1.

^e Missing data on 5.

3.2. Patient Complexity and Caregiver Intra‐ and Interpersonal Factors

More than half (56.5%) of patients had ESDP scores > 10 (mean 11.1; SD 5.67), indicative of patients likely to benefit from early discharge planning. Over 80% had CCI scores > 5 (82.6%; mean = 9.4), indicative of a high risk for mortality. Caregiver CES‐D scores > 10 indicate depressive symptoms and the average score was 11.0 (SD: 5.8). The average relationship quality score was 3.4 (SD 0.51) out of 4, with higher scores indicating higher quality.

3.3. Univariate Associations Between Caregiver Readiness and Analytic Model Variables

The mean FAM‐RHDS score was 6.3, indicating low caregiver readiness for discharge. Table 2 includes the univariable linear regression parameter estimates and standard errors between variables from the analytic model and FAM‐RHDS score. Low FAM‐RHDS scores were associated with the patient and caregiver not co‐residing, caregiver depression, poor relationship quality and lack of available help. Higher FAM‐RHDS scores were associated with more caregiving hours/week prior to hospital admission.

TABLE 2.

Parameter estimates and standard errors from the univariable linear regression models assessing the relationship between FAM‐RHDS and explanatory variables.

Variable	FAM‐RHDS a , b
Patient age	−0.01 (0.01)
Patient sex	0.12 (0.24)
Patient insurance status (Ref = Private)	**
Medicaid	−0.98 (0.44)
Medicare	−0.49 (0.25)
Co‐residing with FCG	−0.59 (0.26)*
ESDP c	−0.03 (0.02)
CCI d	0.01 (0.03)
FCG relationship to Patient (Ref = Spouse)	**
Child	−0.47 (0.31)
Other	−0.85 (0.29)
FCG age	0.01 (0.01)
FCG sex (Ref = female)	−0.04 (0.26)
FCG employment (Ref = full time)
Part time	−0.16 (0.38)
Not working for pay	−0.77 (0.36)
Retired	0.08 (0.28)
FCG health status (Ref = very good /excellent)
Poor/fair	−0.61 (0.54)
Good	−0.39 (0.24)
Depression	−0.10 (0.02)*
Other caregiving responsibilities (Ref = yes)	−0.04 (0.25)
Availability of help from others (Ref = yes)	−0.55 (0.27)*
Relationship quality	1.01 (0.23)*
Caregiving hours/week prior to hospitalization	0.005 (0.002)*
FCG Education (Ref = College)	**
< High school, high school only	0.99 (0.34)
Vocational training	0.17 (0.34)

^a FAM‐RHDS = Family Readiness for Hospital Discharge Scale.

^b Parameter estimate (standard error).

^c ESDP = Early Screen for Discharge Planning.

^d CCI = Charlson Comorbidity Index.

^* p‐value < 0.05 for this binary or continuous variable.

^**Overall p‐value < 0.05 for the categorical variable with more than two categories.

3.4. Multivariable Regression Models Explaining Caregiver Readiness

Table 3 provides parameter estimates for FAM‐RHDS scores when only the significant variables from each step were included in the final model.

TABLE 3.

Parameter estimates and standard errors for patient and FCG variables with significant associations to FAM‐RHDS score at each sequential step in regression model (model is adjusted for time from caregiver readiness assessment to patient discharge).

	Parameter estimate (standard error)	p	R‐square
Step 1: Patient demographic factors	NA	NA	NA
Step 2: Patient demographic + complexity factors	NA	NA	NA
Step 3: Patient demographic + complexity factors + FCG demographic factors			0.106
FCG education
Ref = College
< High school, high school only	0.91 (0.32)	0.005
Vocational Training	0.41 (0.33)	0.22
Step 4: Patient demographic + complexity factors + FCG demographic factors + FCG intra‐ and interpersonal factors			0.292
FCG education
Ref = College
< High school, high school only	0.13 (0.35)	0.72
Vocational training	0.45 (0.33)	0.18
Depression	−0.10 (0.02)	< 0.0001
Relationship quality	0.52 (0.23)	0.03
Caregiving hours/week prior to hospitalization	0.005 (0.002)	0.003

Note: NA indicates not applicable, or no variable in the univariable or multivariable model with statistical significance was identified in this step.

Abbreviations: FAM‐RHDS, family readiness for hospital discharge scale; FCG, family caregivers.

Neither patient demographic nor patient complexity factors were independently associated with FAM‐RHDS (Steps 1 and 2), but caregiver factors were (Steps 3 and 4). Caregivers with a high school education reported lower FAM‐RHDS scores than those with a college education. Greater caregiver depressive symptoms were associated with lower FAM‐RHDS scores. Better relationship quality and more hours of caregiving prior to the patient's hospitalization were associated with higher FAM‐RHDS scores. This final model explained 29% of the variance in FAM‐RHDS scores.

3.5. Sensitivity Analysis

Table S1 provides parameter estimates for the sensitivity analysis that includes discharge disposition in the models. Patient discharge to somewhere other than home was statistically associated with lower FAM‐RHDS scores and the variance explained increased from 29% to 32%.

4. Discussion

This study aimed to identify a parsimonious set of patient and caregiver variables associated with caregiver readiness. As a pragmatic approach intended for clinical use, we examined variables that could be extracted from the EHR or computed from caregiver self‐report for integration into routine discharge planning. Our study advances research on caregiver readiness [6, 13, 36] by showing average caregiver readiness is low among caregivers of palliative care patients and that caregiver intra‐ and interpersonal factors are more strongly related to readiness than patient demographics and complexity. Unlike previous research [12], patient factors explained little variance in caregiver readiness, even for patients at high risk for complex discharge and with severe comorbidities. Instead, caregiver factors, like depression, poor relationship quality with the patient, and caregiving hours/week prior to hospital admission consistently explained the greatest amount of variance.

Our findings point to the need to implement care pathways with actionable plans to identify caregivers with low readiness and implement anticipatory interventions to reduce risks of post‐discharge adverse outcomes that result from poor caregiver readiness to take on post‐discharge care challenges. Our findings support the importance of others' efforts to establish efficient ways to identify caregivers' needs prior to discharge and develop methods to collect, code, and score their data [37, 38, 39]. Caregiver data may be accessible in some hospitals as some state laws require hospitals identify whether a patient has a caregiver [13]. Even so, evidence‐based processes for data collection are not standard practice [14]. Research is needed to further evaluate data collection methods and test the most effective ways of addressing unmet caregiver needs so that caregivers can benefit from interventions.

Our finding that the number of hours spent providing care prior to hospital admission is associated with greater readiness suggests caregivers with more hours may have been caregiving longer and providing diverse caregiving tasks, making them better prepared for care transitions. Future research could examine whether immediate needs at hospital discharge differ between new and experienced caregivers (variables not available here). Differences between perceptions of short‐term capacity to provide care and longer‐term may also be important for contextualizing caregiver readiness. Caregiver depression, for example, has been associated with general capacity to provide care [40]; but to our knowledge, how caregiver depression affects immediate capacity to provide care after hospitalization has not been evaluated. Likewise, relationship quality between the caregiver and patient has been associated with preparedness to provide care among dementia [41] and cancer caregivers [42] but not readiness to provide care after hospitalization.

Strengths of our study include a patient sample with multiple serious conditions, allowing for more generalizable conclusions than samples with a single diagnosis, and the use of tested caregiver assessments to reduce measurement error. However, limitations also exist. Our cross‐sectional design was intended to simulate available data that hospital staff would have prior to discharge, but this limits causal interpretations. Over one‐quarter of patients were discharged somewhere other than home, and sensitivity analyses showed this was associated with lower caregiver readiness. It is possible that care teams may have informally assessed other indicators of caregiver readiness not captured in our analysis, such personal health conditions or self‐efficacy, when making decisions on discharge disposition. This finding highlights a need to understand how best to address unmet needs when a patient is not discharged home. Additionally, our predominately white, educated sample of women living in rural areas limits generalizability to demographically and clinically diverse caregivers and patients. Finally, a longer and more comprehensive set of questions or a shorter and specific set of patient variables could improve the variance explained. For example, Lutz has developed a comprehensive caregiver assessment for readiness to care for stroke survivors [43] and Hetland has used a smaller set of variables on very specific treatments for intensive care unit patients that were associated with caregiver readiness [12]. Although both options may increase the variance explained, our aim was to assess if easily retrievable patient data is sufficient to explain caregiver readiness or if short, caregiver‐reported measures were necessary. Because effort required at hospital discharge for long or more thorough assessments is significant, they may not be seen as feasible or aligning with time pressures that staff face at discharge. Shorter assessments that are not condition specific, such as those identified in our study, may be more amenable for implementation.

5. Conclusion

Study results suggest caregiver readiness for patient discharge hinges less on patient factors and more on intra‐ and inter‐personal caregiver factors. Reliance on information about a patient's condition and their risk for a complicated care transition may not provide adequate information to determine if a caregiver is prepared, willing, and able to provide adequate care after discharge. Hospitals will need to collect caregiver reported data to screen for potential modifiable gaps in readiness and identify priority pathways for intervention.

Ethics Statement

The Mayo Clinic Institutional Review Board (IRB) approved this study (# 17–005188) and data have been lawfully acquired. Dr. Jay Mandrekar (mandrekar.jay@mayo.edu) is the study's statistician. He has reviewed the statistics and tables, along with the manuscript, prior to submission.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1. Analytical model of patient and caregiver variables related to caregiver readiness for hospital discharge.

Table S1. Sensitivity analyses with discharge disposition added to the final model: parameter estimates and standard errors for patient and FCG variables and significant associations to FAM‐RHDS.

Griffin J. M., Holland D. E., Vanderboom C. E., et al., “Assessing Family Caregiver Readiness for Hospital Discharge of Patients With Serious or Life‐Limiting Illness Using Electronic Health Record and Self‐Reported Data,” Health Services Research 60, no. 4 (2025): e14441, 10.1111/1475-6773.14441.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.