Effects of Accessible Health Technology and Caregiver Support Posthospitalization on 30-Day Readmission Risk: A Randomized Trial

Abstract

Introduction:

Patients with chronic illness often require ongoing support post-discharge. We evaluated a simple to use, mobile health-based program designed to improve post-discharge follow-up via: (a) tailored communication to patients using automated calls, (b) structured feedback to informal caregivers, and (c) automated alerts to clinicians about urgent problems.

Methods:

A total of 283 patients with common medical diagnoses including chronic obstructive pulmonary disease, coronary artery disease, pneumonia, and diabetes were recruited from a university hospital, a community hospital, and a US Department of Veterans Affairs hospital. All patients identified an informal caregiver or “Care Partner” (CP) to participate in their post-discharge support. Patient-CP dyads were randomized to the intervention or usual care. Intervention patients received weekly automated assessment and behavior change calls. CPs received structured email feedback. Outpatient clinicians received fax alerts about serious problems. The primary outcomes were 30-day readmission rate and the combined outcome of readmission/emergency department (ED) use. Information about post-discharge outpatient visits, re-hospitalizations, and ED encounters was obtained from medical records.

Results:

Overall 11.4% of intervention patients and 17.9% of controls were rehospitalized within thirty days after discharge (hazard ratio [HR]: 0.59; CI: 0.31, 1.11; p=.102). Compared to intervention patients with other illnesses, those with pulmonary diagnoses generated the most clinical alerts (p=.004). Pulmonary patients in the intervention group showed significantly reduced 30-day risk of rehospitalization relative to controls (HR: 0.31; 0.11, 0.87; p=.026).

Discussion:

The CP intervention did not improve 30-day readmission rates overall although post-hoc analyses suggested that it may be promising among patients with pulmonary diagnoses.

Introduction

Between 20% and 30% of hospitalized medical patients are readmitted within 30 days post-discharge.^1–3 Adverse outcomes often result from patients’ difficulty managing their complex self-care regimens.^4,5 Nurse telephone calls post-discharge are labor intensive,^6,7 and evidence is mixed that post-discharge clinician follow-up alone improves outcomes.^8–12 Informal caregivers often play an important role in supporting chronically-ill patients’ efforts to follow self-management plans, identify early warning signs of acute illness,^13–15 and cope emotionally.^16,17 However, healthcare professionals often do not communicate with family caregivers in effective ways to engage them in the care of the patient, and this can contribute to preventable rehospitalization, emergency care visits, and higher healthcare costs.^18–20

In this trial, we used simple telephone based automated calls (interactive voice response or “IVR”) as a tool for increasing communication between patients, informal caregivers, and clinicians. IVR communication has low marginal cost, is accessible regardless of users’ computer literacy,^21,22 and can identify important health concerns arising post-discharge.^23,24 IVR-supported interventions can improve self-care in a variety of ways, including identifying patients with symptoms who need more follow-up,^25,26 providing evidence-based behavior change psychoeducation,^27–29 and reminding patients about key events, such as post-discharge visits. Most interventions using IVR to support self-management focus only on communication with the patient; however, studies have demonstrated that feedback to informal caregivers about patients’ status based on the patient’s IVR-reported symptoms and self-care challenges may increase the effectiveness of IVR-based interventions while decreasing caregiver distress.^30–32

We evaluated an intervention designed to improve post-hospitalization outcomes through tailored and timely self-care education to the patient via IVR, fax alerts to primary care providers to improve clinical follow-up, and activation / empowerment of informal caregivers. Feedback to caregivers was designed to: (a) increase their knowledge of patients’ health status and behavioral needs, (b) educate them about the goals of successful community transitions, and (c) provide them with targeted advice about how to address problems and communicate effectively. Our primary hypothesis was that the Care Partner intervention would improve 30-day readmission rates and the combined outcome of readmission/emergency department use.

Methods

The protocol was approved by the University of Michigan Institutional Review Board, the Ann Arbor VA Human Subjects Committee, and the MidMichigan Health Institutional Review Board. All patients provided written informed consent. The study protocol was previously published.³³

Recruitment

Participants were screened and enrolled between December 2012 and May 2016 at the time of an acute-care medical admission to a university healthcare system, community-based healthcare system, and VA healthcare system. Electronic admission data were reviewed daily to identify potential participants. Patients were eligible if they had at least one illness listed in their active problem list noted as a reason for hospitalization and frequently associated with an increased risk of rehospitalization,³⁴ (i.e., chronic heart failure, stroke, coronary artery disease, chronic obstructive pulmonary disease, peripheral vascular disease, pneumonia, type 2 diabetes, urinary tract infection, gastroenteritis, or asthma). If the admitting diagnosis was unclear but the patient presented symptoms matching one of the IVR modules, we followed the patient throughout their admission until final eligibility was determined. Additional eligibility / exclusionary criteria were determined by medical record review (e.g., psychiatric disorders, hearing loss, stable residence, language). Patients were excluded if they: were cognitively impaired, had a serious mental illness, did not speak English, were unable to communicate by telephone, did not have a health system-affiliated primary care provider, or were unable to nominate a potentially eligible “Care Partner” (CP, see criteria below). Recruiters met with patients in the hospital to obtain informed consent and complete baseline assessments.

Patients in both study arms were required to nominate an informal caregiver or CP to play a structured role in their transition care. We used the Norbeck Social Support Questionnaire (NSSQ)³⁵ to identify the person with whom the patient had at least monthly contact and from whom they received the most instrumental and emotional support. Potential CPs were ineligible if they had a serious mental illness, did not speak English, were less than 21 years old, or could not access email. Research staff telephoned CPs to explain the study, determine eligibility, and obtain verbal consent for participation.

Randomization

After baseline assessment, patient-CP dyads were randomized in a 1:1 ratio to the intervention or control group. Randomization was conducted in blocks of 10 within strata defined by recruitment site and whether or not the participant had an in-home caregiver. The randomization sequence was computer generated and concealed to recruiters by means of pre-sealed envelopes. Group assignment was not disclosed to patients until after they completed baseline surveys.

Usual Care

All participants received their usual inpatient care, discharge instructions, reconciled medication list, and follow-up care as directed by their attending hospitalist. CPs in the usual care group received printed guidance about supporting the patient’s transition from hospital to home, including information provided by the National Institute on Aging. They also were given the option to receive self-care information specific to their patient-partner’s discharge diagnoses.

Care Partner Program Intervention

A complete set of IVR call contents, clinical alert contents, and CP feedback messages is available from the authors on request.

IVR Calls.

Intervention patients indicated their preferred times and telephone numbers for receiving IVR assessment and self-care support calls. During the initial two weeks after discharge, patients received daily IVR calls, with up to three attempts per day. After the initial two weeks, patients received IVR calls three times per week for two weeks, and then weekly for nine weeks. Calls to patients consisted of queries and statements recorded in a human voice. Patients responded to requests for information using their touch-tone keypad.

IVR messages were designed to: reinforce the importance of medication adherence, remind patients to schedule and attend outpatient follow-up appointments, and teach patients how to identify and address “red-flags,” (i.e., indicators of worsening health specific to their discharge diagnoses). For example, for patients diagnosed with heart failure, the calling system asked about possible changes in weight and provided information about the connections between fluid retention and medication adherence, salt, and fluid intake.³⁶ At the end of each call, patients had the option of hearing the telephone number of their outpatient clinician.

Post-discharge primary care information was entered into the IVR system so that patients and their CPs could receive appointment reminders. In addition to IVR calls, patients also received printed materials based on the Society for Hospital Medicine’s Project BOOST, including: the tool to assess risk (8P’s), the patient PASS (preparation to address situations after discharge), and the Universal Patient Discharge Checklist.^37–39

During each IVR call, all patients received questions and feedback related to their general health, medication adherence, and follow-up in outpatient care. In addition, at the time of discharge, patients were assigned 1–3 additional clinical modules by the study nurse. Possible modules were designed to address: shortness of breath related to heart failure, shortness of breath related to COPD, shortness of breath with unspecified etiology, chest pain, fever, vomiting, diarrhea, and high/low blood sugar related to diabetes. Assignment of modules was based on a detailed, written protocol (available from the authors on request), taking into account a review of the inpatient record (chief complaint, important symptoms, comorbidities) and admission history over the prior six months. Using this structured-but-flexible guideline, the nurse assigned modules in consultation with the study hospitalist in order to focus on problems most likely to result in readmission while avoiding overwhelming multimorbid patients with IVR contacts that were unacceptably long.

Clinical Alerts.

A detailed protocol available from the authors on request was developed with the study hospitalist and nurse practitioner to determine which IVR responses (e.g., what severity of breathing problems) should generate alerts to patients’ clinician team. In conjunction with transition support teams in each site and depending on the urgency of the problem, alerts promoted patients to contact their transition support team in their usual source of care or 911. Fax alerts were accompanied by a cover page to outpatient teams within each health system. In the university health system, alerts were directed to “nurse navigators” responsible for post-discharge follow-up; in the community health system, they were sent to the patient’s primary care provider or (if available) outpatient nurse care manager. Example faxes are available on request. In the VA healthcare system, alerts were sent to the clinic’s centralized fax machine and an email was sent by the study coordinator to the team nurse to notify them about the alert. Faxes were used instead of electronic medical record (EMR) notifications at the recommendation of clinician leads in each site in order to assure rapid follow-up on alerts and to avoid data security barriers related to requesting write access to each site’s EMR system. The thresholds for clinician alerts were negotiated with representatives of the primary care leadership in the participating health systems, and in some cases, thresholds were tailored for individual patients. Alerts included the patient’s name, the date and time of the patient’s most recent IVR call, the concerning symptoms or behaviors, a note that the patient’s CP also was informed about the event, and the telephone number for the research team. A website that included data identifying patients with urgent and non-urgent problems and patients’ complete calling history was available to clinicians.

Care Partner Communication.

After each completed automated call, CPs received structured, automated email reports summarizing patients’ responses (content of these emails is available from the authors on request). If patients failed to complete four consecutive IVR calls, CPs received an email encouraging them to follow-up with their patient-partner. Email feedback about urgent issues was accompanied by a brief automated call that alerted the CP to check their email, and that gave them the option to hear the report over the telephone. CPs also could call into the system using a toll-free number and receive information about the patient’s most recent assessment.

CPs and any other in-home caregivers received materials designed to encourage effective communication regarding the patient’s post-discharge care based on principals of motivational interviewing.⁴⁰ The CP was asked to initiate 10–15 minute conversations at least weekly with the patient to review the patient’s upcoming post-discharge appointments, self-management goals, and most recent automated assessment report. During those conversations, CPs were asked to address the goals of effective transitions as defined by the Care Transitions Intervention model.³⁴ Patients, CPs, and other caregivers received a training DVD that extended the content provided in written materials regarding role expectations for patients and caregivers and a suggested structure for the CP-patient follow-up conversations. No other training for CPs was provided. Effective communication among all caregivers was promoted through communication of program principles including information sharing, strategies for assigning tasks, and resolving potential disagreements. Email summaries following patients’ assessment calls were sent to patients and other caregivers at their request.

Description of Measures

The study’s primary outcomes were 30-day readmission rates and the combined outcome of readmission/emergency department visits in the first 30 days post-discharge. We also report on secondary outcomes, including differences in the rates of outpatient follow-up post-discharge. Information on patients’ inpatient admissions and outpatient visits to primary care and specialty care was obtained from electronic clinical and billing records. At enrollment, patients were asked to approve retrieval of discharge information from non-affiliated hospitals. The IVR calling system automatically captured information on patients’ responses and whether a call attempt was successful. A completed IVR call was defined as a call in which patients identified themselves using their phone number and date of birth, and responded to the first two questions asking about their general health and changes in general health. In practice, almost no patients who completed those two questions hung-up before the end of the calls. Patients also completed telephone interviews with a trained research assistant at baseline, 30 days and 90 days post-discharge (results to be reported elsewhere).

Protocol Modifications and Power Calculation

There were three modifications from the original study protocol. First, power for the trial was calculated assuming 22% of usual care patients would experience a rehospitalization within 30 days post-discharge, a rate consistent with prior studies and our hospital tracking systems.^7,41,42 We calculated that 760 patients would be needed to provide 80% power to detect a 35% reduction in this rate, assuming a two-tailed alpha of 0.05. For administrative reasons, an enrollment site that was expected to provide more than half of the study sample never was able to enroll patients, and we did not reach this target sample. Second, in order to increase the number of eligible patients in the remaining sites, the original requirement that patients identify out-of-home CP’s was relaxed to allow in-home CPs (mainly spouses). Preliminary analyses showed no difference in intervention effects according to whether the CP lived in the household or not. Third, the original intervention proposal included a web-based platform through which CP’s could communicate directly with post-discharge care managers and physicians about their patient-partner. Due to clinicians’ concerns regarding time burden and medico-legal confidentiality issues, this feature was not developed.

Statistical Analysis

We examined differences across treatment groups in baseline measures of potential prognostic indicators, such as patients’ age, race, gender, index length of stay, previous acute events, and discharge diagnoses. Outcome analyses were conducted based on intention to treat. Here, we present the trial’s primary outcomes (i.e., 30-day readmission and readmission/emergency department use) plus rates of outpatient follow-up attendance. We used Kaplan-Meier survival curves to illustrate differences in event-free survival during the 30 days post-discharge, and we used Cox proportional hazards models to provide a summary estimate (with 95% confidence interval and p-value) of the differences between arms.

Post-hoc subgroup analyses stratified by IVR module type suggested that patients who were assigned the module focused on shortness of breath (SOB; i.e., patients with COPD, asthma, or pneumonia) generated proportionately more alerts than patients who were not assigned the SOB module. Based on this, we examined in further post-hoc subgroup analyses rehospitalization rates and ED encounters in the subset of patients with pulmonary diagnoses.

Results

Patient Characteristics

Of the 2,570 patients screened during their hospitalization, 342 were unreachable, 1,102 declined participation, and 851 were ineligible (Figure 1). A total of 283 patients were randomized. Medical records were reviewed on all randomized patients until they either completed the study (n=234), died (n=6), were lost to follow-up (n=30), or were withdrawn (n=14). Fifty-four percent of participants were women, the mean age was 60.7 years (SD=13.1), and 80.2% were White (Table 1). The majority of patients (72.4%) had one or more ED encounters in the year preceding their index admission, and 59.0% had been hospitalized in the prior year. Forty-five percent of patients had a pulmonary diagnosis at discharge, 39.9% had cardiovascular diseases, 23.7% had endocrine disorders, 13.8% had gastroenteritis, and 13.4% had deep vein thrombosis or a pulmonary embolism.

Figure 1.

CONSORT Diagram

Table 1.

Baseline Characteristics

	Intervention (n=143)	Control (n=140)	Total (N=283)
Female gender (%)	58.7 (84)	48.6 (68)	53.7 (152)
Age	60.9, 13.1	60.5, 12.8	60.7, 12.9
White race (%)	78.6 (110)	81.9 (113)	80.2 (223)
Married (%)	62.4 (89)	59.3 (83)	60.8 (172)
Education (%)
< HS Degree	4.9 (7)	5.7 (8)	5.3 (15)
HS - Some College	62.9 (90)	60.0 (84)	61.5 (174)
College Degree+	32.2 (46)	34.3 (48)	33.2 (94)
ED visits in prior year (%)
0	28.7 (41)	26.4 (37)	27.6 (78)
1	27.3 (39)	26.4 (37)	26.9 (76)
2 or more	44.1 (63)	47.1 (66)	45.6 (129)
Admissions in prior year (%)
0	46.2 (66)	35.7 (50)	41.0 (126)
1	25.2 (36)	30.0 (42)	27.6 (78)
2 or more	28.7 (41)	34.3 (48)	31.5 (89)
Index length of stay	3.25, 2.16	3.49, 2.95	3.37, 2.58
Discharge diagnoses (%)
Pulmonary	48.3 (69)	42.1 (59)	45.2 (128)
Cardiovascular	37.8 (54)	42.1 (59)	39.9 (113)
Hematology	16.8 (24)	10.0 (14)	13.4 (38)
Endocrine	25.2 (36)	22.1 (31)	23.7 (67)
Infection	21.7 (31)	20.7 (29)	21.2 (60)

Notes: Unless indicated, cell entries are mean, SD. ED: emergency department. Diagnoses represent either primary or secondary discharge diagnoses recorded in the medical record. Most patients had multiple diagnoses. Pulmonary: asthma, COPD, or pneumonia. Cardiovascular: arrhythmias, coronary artery disease, stroke, angina, or heart failure. Hematology: deep-vein thrombosis or pulmonary embolism. Endocrine: diabetes or high blood sugar. Infection: C difficile, urinary tract infection, gastroenteritis, or ‘infection’.

Care Partner (CP) characteristics were similar across arms (data not shown). Patients selected a wide range of people to serve as their CP, including adult children (32.1%), friends (21.4%), siblings (16.8%), spouses/partners (14.9%), or some other person (14.8%). CPs on average were younger than their patient-partner (mean age 50.3 years, SD=14.7), and most were women (79.9%).

IVR Call Completion and Clinician Alerts

Intervention participants completed 77.2% of their attempted IVR calls (2,920 out of 3,831). Call completion was similar across groups of patients assigned different modules, although there was some variation. Patients were more likely to complete calls if they were assigned the module for shortness of breath (SOB) versus only one or more of the other modules (78.6% complete versus 74.9%, p=.011). Patients assigned the vomiting module also were more likely to complete calls (88.4% versus 74.2%, p<.001), and patients were less likely to complete calls when assigned the module for chest pain (74.7% versus 81.1%, p<.001) and diarrhea (73.7% versus 77.0%, p=.041). Clinician alerts were generated for 18.0% of completed calls (525 alerts). Patients receiving IVR calls focused on SOB were more likely to generate alerts than patients who did not receive IVR content focused on SOB (20.5% versus 16.3%, p=.004).

Post-Discharge Encounters

Most patients in both arms were seen in outpatient care within the first two weeks post discharge, with no significant difference between intervention (72.5%) and control (68.9%) groups (p=.516) over the 30 days post-discharge (Figure 2). Overall 11.4% of intervention patients and 17.9% of controls were rehospitalized within thirty days after discharge (hazard ratio [HR]: 0.59; CI: 0.31, 1.11; p=.102; Figure 3). There was no significant difference in time to the first acute event of either rehospitalizations or ED encounters. In the subset of patients with an index pulmonary diagnosis, intervention-group patients had significantly lower risk of rehospitalization within 30 days than control patients (HR: 0.31; 0.11, 0.87; p=.026), and there was a trend toward decreased acute events in this subgroup when both rehospitalization and ED visits were considered (HR: 0.46; CI: 0.20, 1.03; p=.064).

Figure 2.

Time to first outpatient encounter post-discharge. HR: Hazard Ratio. Numbers in parentheses represent the 95% confidence interval for the HR. Blue line: Intervention. Red line: Controls.

Figure 3.

Time to first re-hospitalization or emergency department (ED) visit. HR: Hazard Ratio. Numbers in parentheses represent the 95% confidence interval for the HR. Blue line: Intervention. Red line: Controls.

Discussion

Patients with chronic conditions experience frequent hospitalizations,¹ and many have unsuccessful transitions back to the community post-discharge.⁴³ Post-discharge transition care interventions can reduce patients’ rehospitalization risk,^44,45 but most health systems lack the infrastructure to provide these services effectively. In this trial, post-discharge IVR support for patients with feedback to clinical teams and informal caregivers did not significantly decrease 30-day readmission rates, although post-hoc subgroup analyses suggested that it may be impactful for patients with pulmonary diagnoses.

Despite ongoing efforts to reduce patients’ rehospitalization risk, identifying effective interventions to reduce 30 day rehospitalization rates has proved to be challenging.⁴⁶ Other studies have reported that most hospital readmission cases may not be preventable,⁴ and factors beyond patients’ medical conditions have a significant impact on their risk of being readmitted.⁴⁷ Efforts to reduce hospital readmission rates may benefit from beyond hospital based interventions to include primary care based practice changes.⁴⁸ Informal caregivers are often those closest to the patient, and can be invaluable in helping patients connect to their clinical care providers in a timely manner. To our knowledge, this is the first study to explicitly provide informal caregivers with a structured role and the tools they need to be integral members of the transition-support team.

We observed a non-significant trend suggesting decreased 30-day readmissions in the intervention group (p=.102). The relative hazard observed across arms was consistent with the study’s original hypothesis and power calculation. Unfortunately, a sample of 760 patients would have been required for an effect of this size to reach statistical significance with a power of 0.80 and a type-1 error rate of 5%. Given the importance of 30-day hospital readmission rates under current reimbursement policies, this trial merits replication to determine whether the results can be confirmed in a larger sample.

The post-hoc subgroup analyses focused on patients with pulmonary-related conditions are promising, albeit exploratory. These analyses were conducted based on findings showing that such patients generated more fax alerts per completed IVR call than other patients. It is encouraging to note that these increased alerts may have led to fewer, rather than more acute-care encounters. This suggests that outpatient providers may have intervened effectively to prevent those events. Unfortunately, due to resource constraints we were unable to systematically track clinician responses to clinician alerts in this study.

It may be that some out-of-system events were not recorded, and therefore our estimate of the number of post-discharge rehospitalizations may be low. However, we have no reason to expect that out-of-system acute care use varied by randomization arm, therefore we do not expect that any unobserved encounters would bias the relative risk of events across groups. Strengths of the medical record review include that it was done by an experienced nurse practitioner who reviewed in detail all encounter notes in inpatient and outpatient charts. These records included information on out-of-system acute encounters, when those events were reported back to the patient’s usual provider.

Other limitations of the current trial include the fact that we were unable to capture individual actions taken by CPs based on email feedback. As such, the relative impact of the IVR patient education, CP feedback, and clinician follow-up in this multi-dimensional intervention remain undetermined. Also, in future studies it may be useful to know additional detail about CP characteristics, such as their experience providing informal care, and whether they had any formal healthcare training. The current sample of patients was 80% White – that and other characteristics of the sample may limit the generalizability of the findings to different populations. Finally, among patients with the target diagnoses who did not refuse screening and were reached, (75%) were ineligible, due to issues including comorbid dementia, delirium, substance abuse, or psychiatric disorders. This may limit adoption of interventions such as this one in practice.

In summary, the studied combination of accessible health technologies and coordinated communication for patients, caregivers, and clinical teams did not reduce 30 day readmissions or ED visit rates. However, the study was underpowered, and future studies are needed to confirm the impact of the intervention on outcomes, and improve enrollment of patients into care transitions research. In addition, the impact of education and support on caregiver burdens requires further study.