Caregiver Integration during Discharge Planning of Older Adults to Reduce Resource Utilization: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Abstract

OBJECTIVES

To determine the effect of integrating informal caregivers into discharge planning on post-discharge cost and resource utilization in the older adult population.

DESIGN

A systematic review and meta-analysis of randomized controlled trials that examine the effect of discharge planning with caregiver integration begun prior to patient discharge on healthcare cost and resource utilization outcomes. MEDLINE, EMBASE and the Cochrane Library databases were searched for all English language articles published between 1990 and April 2016.

SETTING

Hospital or skilled nursing facility.

PARTICIPANTS

Older adults with informal caregivers discharged to a community setting.

MEASUREMENTS

Readmission rates, length of and time to post-discharge rehospitalizations, costs of post-discharge care.

RESULTS

Of 10,715 abstracts identified, fifteen studies met the inclusion criteria. Eleven studies provided sufficient detail to calculate readmission rates for treatment and control. Compared to usual care, discharge planning interventions with caregiver integration were associated with a 25 percent reduction in readmissions at 90 days (Relative Risk [RR], .75 [95% CI, .62-.91]) and a 24 percent reduction in readmissions at 180 days (Relative Risk [RR], .76 [95% CI, .64–.90]). The majority of studies reported statistically significant reductions in time to readmission, length of rehospitalization, and costs of post-discharge care.

CONCLUSION

For older adult patients discharged to a community setting, the integration of caregivers into the discharge planning process, compared to non-systematic inclusion of caregivers, reduces the risk of hospital readmission.

INTRODUCTION

Discharge planning promotes safe and timely transfer of patients between levels of care and across care settings, especially during patient discharge from a hospital or skilled-nursing facility to a home or community setting.^1,2 Discharge planning includes determining the patient’s appropriate post-hospital discharge destination and identifying patient needs for a safe transition. Effective discharge planning is especially significant for aging patient populations, for whom informal caregivers, defined as unpaid individuals who provide support for medical tasks and daily activities, are critical to daily life and health. As hospital lengths of stay continue to decrease, informal caregivers of aging individuals are responsible for increasingly complex patient treatment including caring for wounds, managing medications, and operating specialized medical equipment.³ Caregivers often report unmet needs and dissatisfaction with the discharge planning process.¹

Recently adopted caregiver abuse, record and enable (CARE) legislation and proposed Medicare regulations require caregiver integration into the discharge planning process.^4,5 While the implementation of these requirements may impose additional time spent on caregiver education and training, the inclusion of caregivers in the discharge planning process may improve patient outcomes and help hospitals to avoid economic penalties for patient resource utilization and costs under programs like the Hospital Readmissions Reduction Program (HRRP).⁶ Under HRRP, for instance, the federal government reduces payments to hospitals that rank among the top quartile for high rates of readmissions. Limited evidence exists, however, on how caregiver integration in the discharge process could affect resource utilization and cost.^4,5,7 The aim of this systematic review and meta-analysis was to evaluate evidence on the effects of integrating informal caregivers of older adults into the discharge planning process on post-discharge cost and resource-utilization outcomes, including readmission rates, length of post-discharge rehospitalizations, and the cost of post-discharge care.

METHODS

Search strategy, screening and evaluation of studies were conducted using systematic review methods.⁸ The literature search strategy was developed by a Public Health Informationist for two principal concepts: “patient discharge planning” and “older adults.” The search was further refined using terms for randomized trials or intervention studies, and tested to determine that it was appropriate to limit studies to English language articles published from 1990 to April 2016. Databases searched include PubMed, EMBASE and the Cochrane Library. Reference lists of all review articles were screened and content experts were surveyed for additional article recommendations. The review protocol has been registered in PROSPERO, an international register of systematic reviews (ID# 37374).

Study Selection and Data Extraction

For inclusion, studies had to be a randomized controlled trial (RCT) published in English; have a study sample that included an average age of participants older than 65 years; examine the effectiveness of a discharge planning intervention from a hospital or skilled nursing facility on healthcare outcomes; and integrate an informal caregiver into at least one part of a discharge planning intervention. Exclusion criteria were discharge planning intervention that did not begin prior to discharge, and patient discharge to a non-community setting. Centers for Medicare & Medicaid definitions for settings were used and included home, retirement community and independent and assisted living facilities (community settings); and nursing facilities and inpatient care settings (non-community settings).⁹ No specific criteria were established to define an informal caregiver, except that the person providing care could not be present in a professional capacity. Four members of the project team, the coders as well as an investigator with content expertise, conducted a pilot abstract screening on a random sample of approximately 10 percent of total study abstracts to ensure consistency between coders and to refine the coding process.¹⁰ Three investigators then independently reviewed abstracts, and weekly meetings were held with the complete project team to resolve coding questions and to ensure continued fidelity to the initial training. An identical process was used for full-text review and data extraction. For all included studies, two independent reviewers coded interventions to evaluate the extent of caregiver integration and two team members assessed systematic errors or deviations from the truth using the Cochrane Collaboration’s risk of bias tool.¹¹ A statistician oversaw the extraction process and analysis of study outcomes. Discrepancies were resolved through team discussion.

The following data were extracted from all included RCTs: study setting; discharge location; patient and caregiver demographics; components of the discharge-planning intervention; who administered the intervention; and healthcare resource utilization and cost outcomes.

Qualitative Synthesis of Evidence

Study results and methodological limitations of included studies were summarized, as were patterns or inconsistencies, main themes, and potential explanations for patterns or inconsistencies.

Quantitative Synthesis of Evidence

Based on data availability, Relative Risks (RR) for 90-day and 180-day readmission were estimated. The RR is the ratio of intervention-group readmission rate to the control-group readmission rate. RRs less than 1 indicate a lower intervention-group readmission rate compared to the control-group rate. Study results were pooled in a random-effects (DerSimonian and Laird) model estimating the RR and 95% confidence interval (CI).¹² Potential statistical inconsistencies across studies despite methodological variability was assessed by calculating the I² statistic.¹³ Potential for publication bias was assessed using the Egger regression and the Begg rank tests.^14,15 All analyses were conducted in the statistical package Stata.¹⁶

Quality Rating

Two expert reviewers assessed the overall quality of the evidence for each included study based on the specific criteria outlined by the Cochrane Risk of Bias tool.¹⁷ For each included study, reviewers provided assessments of sequence generation, allocation concealment, blinding, completeness of outcome data, selectiveness of outcome reporting, as well as other sources of bias.

RESULTS

After duplicate removal, 10,546 abstracts remained and 169 were identified through reference lists and expert opinion. In total, 10,715 abstracts were reviewed. Of the 99 publications that met participation and intervention criteria, 27 studies were randomized controlled trials (RCTs) and 15 of these included outcomes on cost or health resource utilization (see supplementary Figure 1).

Study Characteristics

Table 1 summarizes the details of the 15 cost or resource utilization RCTs. Two included studies reported different results from the same trial.^18,19 Studies were published over a period of 19 years. Total study participant group size ranged from 49 to 930 patients, with control group size ranging from 24 to 478 patients and intervention group size ranging from 25 to 450 patients. Study locations were varied: seven studies were conducted in the United States and eight outside the United States. Study definitions of caregivers were not available. Studies indicated inclusion of caregivers or family members, but none specified methods for definition and identification.

Table 1.

Study Characteristics, Intervention Components, and Interventionists

Study	Participants	Componentsa	Inter- ventionist
Naylor et al. (1994)20 US	Total N = 276 Medical N= 142 (CG=70; IG=72); Surgical N=134 4 (CG=66; IG=68)]. Patients 70 years and older.	ID; WI; LV; CNX	Nurse
Rich et al. (1995)21 US	N=282 (CG=140; IG=142). High-risk patients over 70 hospitalized for CHFb	MR; CNX	Teamc
Naylor et al. (1999) 1US	N=363 (CG=186; IG=177). Patients age 65 or older, hospitalized in last four years	LV; CA; WI; CNX	Nurse
Li et al. (2003)23 US	N=49 (CG=24; IG=25). Caregivers of hospitalized elders admitted to one of four units in an academic medical center	WI	Not specified
Laramee et al. (2003)27 US	N=287 (CG=146; IG=141). Patients hospitalized with cardiac conditions	ID; LV; CA; CNX; WI	Case manager
Lim et al. (2003)28 Australia	N=598 (CG=287; IG=311) Patients aged 65 and over, hospitalized and required community services after discharge	CA; CNX	Nurse or AHPd
Naylor et al. (2004)22 US	N=239 (CG=118; IG=121). Patients aged 65 and older and hospitalized with heart failure	CA; ID; LV; CNX;WI	APNe
Huang & Liang (2005)29 Taiwan	N=126 (CG=59; IG=63). Patients aged 65 and older hospitalized due to falling	CA; ID; LV; WI; CNX	GNf
Shyu et al. (2005)30 Taiwan	N=137 (CG=69; IG=68) Patients 60 and older hospitalized with hip fracture	CA; CNX	GN
Shyu et al. (2010)25 Taiwan	N=158 (CG=86; IG=72) Dyads of older patients with stroke and their family caregivers	ID; CA; CNX	Nurse
Legrain et al. (2011)19 France	N=665 (CG=348; IG=317) Patients admitted to 6 geriatric hospital units	MR; LV; WI; CNX	Geriatrician
Li et al. (2012)38 US	N=407 (CG=205; IG=202) Dyads of hospitalized older adults and family caregivers	CA; WI	Research assistant
Bonnet et al. (2013)18 France	N=665 (CG=348; IG=317) Patients admitted to 6 geriatric hospital units	MR; WI; CNX	Geriatrician
Lainscak et al. (2013)26 Slovenia	N=253 (CG=135; IG=118) Patients hospitalized for acute exacerbation of COPD	CA; CNX	Coordinator
Forster et al. (2013)24UK	N=930 (CG=478; IG=450) Dyads of medically stable stroke patients and caregivers helping with ADL.	LV; CA; CNX	Team

^aMR=Medication Reconciliation; ID= In-person Demonstration; LV = Teach Back or Learning Validation; CA=Caregiver Assessment; WI = Written Instructions; CNX= Connection to External/Community Resources

^bCongestive heart failure

^cMultidisciplinary team

^dAllied health professional

^eAdvanced practice nurse

^fGerontological nurse

Study Populations

The 13 unique study populations included a total of 4361 patients, 56% of whom were female. Patients in all studies had a mean age over 70 years. Six studies with 2137 patients included data on race and had largely white race populations (78%).^1,20–24 Patients in all studies were discharged from either hospital or skilled nursing facility settings; however, due to a lack of detail in reporting, it is not possible to determine the number of patients discharged from a hospital versus those discharged from a nursing facility. Demographic information for caregivers was only presented in 3 of the studies; in these, 34% of caregivers were male and their ages varied widely.^23–25 Two studies presented information on caregiver relationship to patient (caregiver n = 1086), in which 61% were a spouse or partner and 35% were adult children.^24,25

Intervention Components

Table 1 shows intervention components documented in the studies. Of the 15 studies, 13 had an intervention component that linked caregivers to external or community resources (such as sending hospitalization records to patient PCP) and nine included written care plans. Caregiver assessment was a component in eight studies, and three included medication reconciliation. Live or video demonstrations of care tasks was included in five studies, and seven included so-called “teach back” techniques in which caregivers and/or patients demonstrate care skills to the interventionist. Fourteen of the studies included more than one intervention component and nine included more than two components.

The majority of studies (11 of 15) had interventions that began in the hospital or nursing home and continued after discharge to the community.^{1,20–22,24–30} The length of the interventions that continued after discharge, when described, ranged from one week to three months and included follow-up phone calls (4 studies);^22,27–29 a phone call and a home visit (3 studies);^24,26,30 or multiple home visits and phone calls (3 studies).^1,21,25 One study did not specify the type of post-discharge intervention.²⁰

Interventionists

The interventionists that were most frequently involved in the RCTs were nurses (n=7), with two studies using gerontological nurses, and one employing an advanced practice nurse (APN) (Table 1). Geriatricians were involved in two studies and two examined multidisciplinary teams made up of multiple specialists. In two studies, a discharge coordinator or case manager was the interventionist. One study relied on research assistants to perform the intervention. Finally, one study did not specify the interventionist involved.

Outcomes

All studies reported at least some results on readmissions, with 14 reporting on readmissions for any cause (Table 2). Six studies reported time to readmission and seven studies reported length of rehospitalization. Other utilization outcomes included unscheduled acute care visits after discharge, skilled nursing facility admission, emergency department visits, as well as caregiver and patient utilization of a range of services.^19,23–25 Three studies reported on the cost of initial hospitalization and seven reported on the cost of rehospitalization.

Table 2.

Evidence Table of Included Studies- Readmissions and Costs

Study	Readmissions, %g	Length of re- hospitalizationh	Mean cost, initial hospitalization	Mean cost, rehospitalization
Naylor (1994)	12-week, Medical: 33(CGi), 22(IGj) 12-week, Surgical: 32(CG), 27(IG)	12-weekk: Medical: 222 (CG); 131 (IG) Surgical: 110 (CG), 149 (IG)	Medical: $23,810 (CG), $24,352 (IG) Surgical: $96,640 (CG), $105,936 (IG)	6–12 week: Medical: $340,496 (CG), $471,456 (IG) Surgical: $85,124 (CG), $170,248 (IG)
Rich (1995)	90-day: 46(CG); 34(IG) (p<.1) 90 Days, >1 readmission: 16(CG), 6(IG) (p<.01)	90-day: 6.2 (CG), 3.9 (IG) (p=.04)	N/A	90-day, total: $5,275 (CG) $4,815 (IG) (p<.05); 90-day, readmissions: $3,236 (CG), $2,178 (IG) (p<.05);
Naylor (1999)	Up to 24 weeks: 37.1(CG), 20.3(IG) (p<.01) Over 24 weeks: 14.5(CG), 6.2(IG) (p<.01)	24-week: 1.53 (IG), 4.09 days (CG) (p<.001) 24-week, readmitted patients: 10.1 (CG), 7.50 (IG) (p<.001)	N/A	24-weeks, aggregate costs: $1,024,218 (CG), $427,217(IG) (p<.001)
Li (2003)	60-dayl: .21 (CG), .04 (IG) (p<.1)	N/A	N/A	N/A
Laramee (2003)	90-day: 37(CG), 37 (IG)	9.5(CG), 6.9 (IG) (p=.15)	$19,081 (CG), $16,119 (IG) (p=.18)	90-day readmission: $5,163 (CG), $5,253 (IG) 90-day readmission, readmitted patients: $16,395 (CG), $15,417 (IG)
Lim (2003)	180 Days: 28(CG), 25(IG)	5.2(CG), 3.0(IG) (p=.01)	N/A	6-month hospital: $10,161 (CG), $8,390 (IG) (p=.02) 6-month total: $10,687 (CG), $9,142 (IG) (p=.05)
Naylor (2004)	52-week: 61.2(CG); 51.2(IG) (p=.01)	52-week: 8 (CG), 5 (IG) (p<.07)		52-week, total adjusted per-patient: $12,481 (CG), $7,636 (IG) (p=.002)
Huang (2005)	90-day: 20.63(CG), 6.35(IG) (p=.02)	N/A	N/A	N/A
Shyu (2005)	30-day: 7.6(CG), 4.5(IG); 90-day: 14.1(CG); 7.9 (IG)	N/A	N/A	N/A
Shyu (2010)	6-month: 19.5(CG); 13 (IG) 6–12 month: 7.2(CG); 0(IG)	N/A	N/A	N/A
Legrain (2011)	90-day: 28.4 (CG), 20.2(IG) (p<.05) 180-day: 38.2(CG); 32.5(IG)	N/A	N/A	N/A
Li (2012)	60-day, mean number per patient: .06 (CG), .11 (IG)	N/A	N/A	N/A
Bonnet (2013)	6-month: 28.7 (CG), 17.3(IG) (p=.12)	N/A	N/A	N/A
Lainscak (2013)	180-day: 44(CG), 31(IG) (p<.05)	N/A	N/A	N/A
Forster (2013)	6-month: 19(CG), 18(IG) 6–12 month: 18(CG), 15(IG)	6-month: 8 (CG), 12 (IG) 6–12 months: 9 (CG); 9 (IG)	6 –month: £12, 471 (CG), £13,127 (IG)	6-month: £26,381 (CG), £26,894 (IG) (p=.432) 12-month: £37,884 (CG), £37,453 (IG) (p=.159)

^gPercent readmitted unless otherwise noted

^hMean days unless otherwise noted

ⁱCG = control group

^jIG = intervention group

^kTotal days

^lMean per patient

Table 2 provides greater detail on several common outcomes: readmissions, length of rehospitalization and costs of initial hospitalization and rehospitalization. Of the 14 studies that reported on readmissions for any cause, nine reported statistically significant reductions. Five of six studies reporting on time to readmission reported statistically significant reductions in the intervention group. Five of seven studies with outcomes on length of rehospitalization also reported statistically significant reductions. Of the seven studies reporting outcomes on cost of post-discharge care, four reported significant reductions. One study reported significant reductions in the cost of initial hospitalizations (pre-discharge).

Eleven studies provided sufficient detail to calculate readmission rates for treatment and control, six at 90 days and five at 180 days. Naylor et al. (1994) reported 90-day readmission rates for two separate groups observed within the same study.²⁰ Legrain et al. (2011) provided detail for both 180-day and 90-day readmission rates.¹⁹ As Figure 1 shows, across all studies that reported 90-day readmission rates, the pooled intervention effect (RR) was 0.75 (95% CI: 0.62–0.91; p=.004). There was limited evidence of heterogeneity in 90-day readmission rates (I²=27.6%; p=0.218). As shown in Figure 2, across all studies that reported 180-day readmission rates, the pooled intervention effect (RR) was 0.76 (95% CI: 0.64–0.90; p=.001). Limited statistical inconsistency was found with the test for heterogeneity across the studies assessing 180-day readmission rates (I² = 30.8%; p=0.22).

Figure 1.

Relative Risk ratio (RR) status of intervention compared to control, 90-day readmissions (CI=Confidence Interval)

Figure 2.

Relative Risk ratio (RR) status of intervention compared to control, 180-day readmissions (CI=Confidence Interval)

Publication Bias

For studies analyzing readmissions within 180 days, no evidence of publication bias was identified (Egger test P = .342; Begg test P > .99). There were similar results for readmissions at 90 days (Egger test P =.379; Begg test P > .462).

Quality of Included RCTs

Several methodological limitations were identified for the 15 RCTs that were included (See Supplementary Figure 2). Six studies provided no information on sequence generation. Eight studies provided no information on allocation concealment. One study did not provide adequate blinding of participants or outcome assessors and seven provided inadequate information on blinding. Seven studies provided inadequate information on outcome reporting and one study selectively reported outcomes.

DISCUSSION

This study demonstrates that integration of informal caregivers in the discharge planning process for older adults in hospitals or nursing facilities reduces hospital readmissions. Integrating caregivers in discharge planning yielded a 25 percent reduction in risk of 90-day readmission and a 24 percent reduction in 180-day readmission compared to those who received usual care practices. One of the strengths of these findings is that included studies varied in how they included caregivers, yet the interventions did not treat patients and caregivers in isolation from one another. Thus, these findings represent the real-world and older adults in hospital or a nursing facility where caregivers could be included when appropriate. Adding to the credibility of these findings, the studies had low variability in the estimates due to statistical heterogeneity rather than sampling or methodological variability (I²=27.6% and 30.8%, respectively).

The potential impact of incorporating caregivers in patient discharge planning could be significant. Potentially preventable 30-day readmissions have been estimated to cost $12 billion dollars annually in Medicare spending alone.³¹ As the result of programs like the HRRP, hospitals that have a high proportion of patients readmitted within a short time frame are looking for methods to reduce readmissions. Furthermore, prior studies that examined discharge-planning interventions and readmission risk have focused on disease-specific interventions, and therefore may not be generalizable. For example, research has demonstrated that tailored discharge interventions for patients with congestive heart failure (CHF) can reduce readmissions.^32,33 An effective discharge-planning intervention for these patients includes an emphasis on nutrition due to the link between diet and severity of CHF; an emphasis that may not be beneficial for all patients. The inclusion of caregivers in the discharge planning process, however, may be generalizable outside of disease-specific interventions. Due to medical advances, shorter hospital stays, and the expansion of home care technology, caregivers are taking on considerable care responsibilities for patients.³⁴ This study demonstrates that the systematic inclusion of caregivers in the discharge-planning process may help hospitals to avert readmissions in light of these complex care responsibilities.

Current health policy activity around the engagement of caregivers in discharge planning is trending in the direction of recognizing the value of including caregivers on patient outcomes and health services utilization. More than thirty states and the District of Columbia have passed CARE legislation that requires hospitals to designate and provide instruction and training to informal caregivers.^4,35,36 CARE legislation in most states requires that providers demonstrate, or at least offer caregivers the opportunity to ask questions about, the performance of post-discharge activities such as wound care or administering medications. In addition, recently proposed Medicare regulations would require caregiver integration in the patient discharge planning process.⁵ Under these new regulations, hospitals would be required to consider the availability of informal caregivers and community-based support when engaged in discharge planning.

Several of the intervention components identified in this study are commonly used in current practice, such as connecting patients and caregivers to community resources by recommending out-patient rehabilitation or home-health services. Likewise, the provision of written care plans and medication reconciliation are pervasive components in current practice that are intended to streamline medical service and information delivery. Less commonly reported in the studies we reviewed were the assessment of caregiver needs and the use of teach-back, or learning validation, methods.

Variability existed in the length of the interventions, although the majority of them continued after discharge. The continuation of the intervention after discharged allowed for new or ongoing patient and caregiver needs to be addressed. The effect of the continuation of services in the community versus services received only in a hospital or nursing facility could not be ascertained, but warrants further investigation.

There was variability in the type of health professional who delivered the interventions across studies. It may be that the specific professional background of the interventionist is less important than the systematic inclusion of a caregiver in the discharge-planning process, but this warrants further investigation.

Limitations

Our results must be interpreted in light of several limitations. First, though we assessed publication bias, the small number of studies we had available may mean we did not have sufficient power to detect such bias if it does exist. Second, this was a study-level meta-analysis constrained to RCTs only. While we may have been able to more thoroughly assess the impact of incorporating caregivers in discharge planning had we included non-randomized studies, we did not include such studies due to potential bias. Furthermore, we did not have individual patient-level data. Thus, we could not determine if patients with certain characteristics or specific diseases are better served by caregiver inclusion. Also, it may be argued that limitations existed in the potential bias present in some of the RCTs. Several of the RCTs were unclear in how they handled blinding, allocation concealment, and outcome reporting. These limitations are commonly noted in the caregiving literature.³⁷ Additionally, health outcomes were variably reported across studies; thus, we could not determine how the inclusion of caregivers in discharge planning influences patient health or quality of life.

While all interventions included caregivers in the discharge-planning process, the methods for their inclusion varied across studies. It is therefore not possible to determine, from the current literature, what is the most effective method of caregiver integration during discharge planning of older adult patients, in order to reduce hospital readmissions. In addition, the included studies were predominately multimodal interventions. They relied on several intervention components to create their specific discharge-planning intervention. Future studies may identify which intervention components are the most effective in reducing hospital readmissions.

The studies we identified provided little information about their caregiver populations or the extent of caregiver participation in the discharge planning process, individual discharge planning intervention components, implementation factors, contextual factors affecting the success of the intervention or costs of implementation. Attempts were made to identify study protocols and/or additional publications of the studies included, but it was not possible to find further material for them all. Those that were found generally provided little additional information that was helpful for this investigation. Given the small number of studies, it was also difficult to isolate the effects of caregiver-centered intervention components. Future research should consider addressing the amount of caregiver participation necessary, specific effects of intervention components, as well as implementation barriers and solutions for caregivers. This type of information will be needed to allow health care system leaders and policymakers to plan strategically as they consider implementing programs to prevent readmissions and other harms associated with hospital discharge.

Conclusion

For older adult patients, the systematic inclusion of caregivers during discharge planning leads to greater than a 20% reduction in hospital readmissions. These benefits were observed in older adult populations across disease states. Given the potential for improved patient care and reduced costs, hospitals and nursing facilities should develop care delivery systems that integrate informal caregivers into discharge planning.