Interventions to Improve Communication at Hospital Discharge and Rates of Readmission

1Hospital discharge is a multidisciplinary process during which patients receive complex medical information and follow-up instructions. At discharge, health care practitioners need to explain critical information, such as patients’ diagnoses and their treatment, while integrating patients’ conditions, perceptions, and needs at the same time. However, patients may not understand or remember the information provided, resulting in confusion, misinterpretation and mismanagement of treatment regimen., Low health literacy, anxiety, cognitive impairment, or language barriers might further limit patients’ ability to understand medical information shared at discharge, resulting in treatment failures. Previous studies found that a clinically relevant proportion of patients being discharged from hospitals are not able to recall their diagnoses and discharge instructions., Shortcomings in the education of patients before hospital discharge have been associated with higher risk for hospital readmission and mortality.,

2Unplanned hospital readmissions may indicate poor quality of care. According to the US national health insurance program, Medicare, 15% of patients discharged from the hospital are readmitted within 30 days, and 1 in 4 of those readmissions is potentially preventable., Unplanned readmissions costs are estimated at $20 billion in the United States annually. This has led medical authorities to look for interventions to improve the transition of care and penalize hospitals for readmission. While several factors influence the risk of hospital readmissions, shortcomings in the education of patients at hospital discharge may be one of the main modifiable factors. Still, there is insufficient evidence that improving discharge communication results in lower readmission rates and other patient-relevant outcomes.

3In this study, we performed a systematic review and meta-analysis of randomized clinical trials (RCTs) that examined the effect of communication interventions in medical patients at hospital discharge on patient-relevant outcomes. We were especially interested in the association of communication interventions with readmission to hospital, adherence to treatment regimens, and patient knowledge 30 days after discharge.

4This systematic review and meta-analysis was registered with the International Prospective Register of Systematic Reviews (PROSPERO, CRD42020146415). We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.

5We included RCTs in which the effect of any communication intervention for readmission, medical adherence, mortality, satisfaction, medical knowledge, or reattendance to the emergency department (ED) was assessed. Studies were eligible if the communication intervention was performed shortly before or at hospital discharge, if they had a randomized clinical design, included medical patients, and the intervention was conducted at hospital discharge. Studies conducted in surgical wards, psychiatry hospitals, or outpatient settings were excluded. Studies with interventions that continued after hospital discharge (eg, continuous teaching sessions) were also excluded.

6We used a comprehensive search strategy consisting of a combination of subject headings and free words. We searched PubMed, as well as Embase, CINAHL, and PsycInfo via Ovid. To enhance quality, the electronic search strategy was reviewed according to Peer Review of Electronic Search Strategies by a librarian who specialized in systematic reviews. As we only focused on RCTs, we incorporated a sensitivity and precision-maximizing RCT filter in our search strategy. The final search strategy for PubMed, which was adapted according to each database’s syntax, is available in the eAppendix in the Supplement. To identify additional published, unpublished, and ongoing studies, we tracked relevant references through Web of Science’s and PubMed’s cited reference search, applied the similar articles search of PubMed, and screened all references for potentially eligible studies. The data search was performed between August 2020 and February 2021, last performed on February 28, 2021.

7Two review authors (C.B. and S.Z.) screened titles and abstracts, which were found through the systematic search strategy, independently. Studies were included or excluded according to the inclusion criteria. Later, C.B. and S.Z. reviewed the full texts of studies considered eligible for inclusion independently, and disagreement was resolved by discussion and consensus.

8Two authors (C.B. and S.Z.) independently extracted the data of the included studies. Two authors (C.B. and S.Z.) independently assessed the included RCTs for methodological quality using the Cochrane Risk of Bias Tool. In cases of disagreement, consensus was reached by discussion. If necessary, a third author (S.H.) was consulted.

9Readmission to the hospital was the primary end point of our meta-analysis because the aim of communication interventions at discharge was to prevent this adverse outcome. Secondary end points were defined as adherence to treatment regimen, satisfaction, mortality, and knowledge of medication or diagnoses assessed 30 days after hospital discharge.

10We express dichotomous data risk ratios (RRs) with 95% CIs. Data were pooled using a random-effects model. Heterogeneity of studies was identified through visual inspection of the forest plots. We used the I2 statistic, which quantifies inconsistency across studies, to assess the consequences of heterogeneity on the meta-analysis. An I2 of 50% or greater indicates a high level of heterogeneity. If data were not suitable for direct comparison, we applied narrative synthesis. Also, we assessed for evidence of publication bias (small-study effects). This was assessed visually using the funnel plot and quantitatively using some statistical tests (eg, Egger test).

11For the primary and secondary end points, we determined the associations within each predefined subgroup and compared the summary associations across subgroups using random-effect models. Results were stratified based on type of intervention (medication counseling vs education on patient’s disease and its management vs specific communication techniques [ie, shared decision-making, motivational interviewing, and teach-back]), patient handout (additional written material vs no written material), age (>65 years vs ≤65 years), the proportion of women in the trial (≤50% women vs >50% women), location of study (US vs Europe vs other), risk of bias according to the Cochrane Risk of Bias Tool (poor or fair vs good), study setting (hospitalized patients vs ED patients) and primary disease (cardiology vs respiratory vs other). Cutoffs for stratification were chosen post hoc based on the distribution among trials to achieve a balanced number of patients per group.

12We performed statistical analyses in Stata MP version 15.1 (StataCorp) using the METAN package. Two-sided P <.05 was considered statistically significant.

13A total 15 778 of records were identified through our database search, and 3 additional records,, were found through other sources, such as cited references and similar-article search. We removed 5879 duplicates, discarded 7801 studies after examining titles, and discarded an additional 1843 studies after screening abstracts. Of the remaining 258 full-text articles, 60 studies,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, were eligible for inclusion, with 19 trials,,,,,,,,,,,,,,,,,, included in quantitative analysis (eFigure in the Supplement).

14Characteristics of studies are shown in Table 1 (included for quantitative analysis regarding the primary end point) and eTable 1 in the Supplement (all studies). Publication dates of the included studies range from 1981 to 2021. The studies were from 18 different countries, with 26 studies (43%) from the United States,,,,,,,,,,,,,,,,,,,,,,,,,, 8 studies (13%) from the United Kingdom,,,,,,, and 5 studies (8%) from Spain,,,,, and the remaining 21 studies (35%) from other countries. Across all 60 studies, a total of 16 070 participants were included, with a study sample size ranging from 25 to 3386 participants.

15A total of 24 studies (35%) recruited medical inpatients,,,,,,,,,,,,,,,,,,,,,,,, of which 5 studies (24%) focused on patients aged 65 years or older,,,, and 5 studies (24%) focused on polymedicated patients,,,,, whereas 13 studies (22%) recruited ED patients.,,,,,,,,,,,, Thirteen studies (22%),,,,,,,,,,,, recruited cardiology patients and 10 studies (17%),,,,,,,,, focused on patients with respiratory diseases.

16In 28 studies (47%),,,,,,,,,,,,,,,,,,,,,,,,,,,, the intervention focused on medication counseling, which was often conducted with the involvement of a pharmacist and consisted of education regarding dosage, the importance of medication adherence, and possible side effects. In 27 studies (45%),,,,,,,,,,,,,,,,,,,,,,,,,,, patients in the intervention groups were educated regarding their disease and its management. Study personnel focused on patient knowledge regarding their disease, prognosis, and complications, as well as self-management, such as fluid intake or lifestyle modification. There were 5 studies (8%),,,, that used specific communication strategies, such as motivational interviewing, teach-back, or shared decision-making, as their interventions to encourage treatment adherence or patient involvement in decision-making. In total, we determined 19 studies (32%) were at low risk of bias, 32 studies (53%) were at high risk of bias, and 9 studies (15%) were at unclear risk of bias (eTable 2 in the Supplement).

17Of 60 studies, 19 studies (32%),,,,,,,,,,,,,,,,,, reported data regarding readmission within 30 days after hospital discharge and were thus included in the quantitative analysis. Regarding bias, 11 trials (58%) had low risk of bias, 6 trials (32%) had high risk, and 2 trials (11%) had unclear risk. There was no evidence for publication bias (Egger test: P =.21). Compared with usual care, the pooled results showed a significant association between communication interventions and fewer readmissions to hospital (179 of 1959 patients [9.1%] in intervention groups vs 270 of 1994 patients [13.5%] in control groups; RR, 0.69; 95% CI, 0.56-0.84). There was low heterogeneity among trials (I2 = 9.4%; P =.34) (Figure 1).

18With regard to the primary end point, we performed a subgroup analysis stratified for the type of intervention, age, primary disease, use of handouts, study quality, setting of the study, sex, and location of study (Table 2). There was a significant subgroup difference regarding the primary disease of patients with trials including patients with respiratory conditions (RR, 0.32; 95% CI, 0.18-0.57) and patients with other illnesses (RR, 0.78; 95% CI, 0.64-0.96), whereas trials including patients with cardiac conditions showed no significant difference (RR, 0.62; 95% CI, 0.38-1.02; between-group heterogeneity: P =.01). We also found that trials with less than 50% of patients being women, compared to studies with more than 50% women patients, had better outcomes in hospital admission (RR, 0.55; 95% CI, 0.39-0.77 vs 0.82; 95% CI, 0.64-1.06; between-group heterogeneity: P =.08). Stratification by other subgroups did not show any differences in associations.

19Adherence to treatment regimen was assessed in 20 studies, with 15 studies (75%),,,,,,,,,,,,,, reporting adherence in a dichotomous format (adherent vs not adherent), which we therefore pooled for a meta-analysis including 4033 patients. Regarding bias, 5 trials (33%) had a low risk of bias, 9 trials (60%) had a high risk, and 1 trial (7%) had an unclear risk. There was evidence for significant publication bias (Egger test: P =.006). The pooled analysis showed a significant association between a communication intervention at discharge and higher patient adherence to treatment 30 days after discharge (RR, 1.24; 95% CI, 1.13-1.37). There was substantial heterogeneity among trials (I2 = 85.3%; P <.001) (Figure 2A).

20In a subgroup analysis, trials including older patients (ie, mean age >65 years) found associations of the intervention with adherence, whereas trials including younger patients (ie, ≤65 years) did not (RR, 1.58; 95% CI, 1.08-2.29 vs RR, 1.05; 95% CI, 0.97-1.14; between-group heterogeneity: P =.03). Also, in hospitalized patients, there was an association of the intervention with adherence, whereas in ED patients, there was no association (RR, 1.42; 95% CI, 1.13-1.78 vs 1.04; 95% CI, 0.97-1.11; between-group heterogeneity: P =.04). Of the 5 studies not included in the meta-analysis, 3 studies,, showed a significant increase in treatment adherence in patients who received a communication intervention.

21A total of 15 studies (25%) evaluated the association of communication interventions with patients’ satisfaction within 30 days after hospital discharge. Of these, 11 studies (73%),,,,,,,,,, reported satisfaction in a dichotomous format (satisfied vs not satisfied) and were thus pooled for a meta-analysis (including 3951 patients). Regarding bias, 5 trials (45%) had a low risk of bias, 5 trials (45%) had a high risk, and 1 trial (9%) had an unclear risk. There was evidence for significant publication bias (Egger test: P =.02).

22Compared with usual care, the pooled analysis showed a significant association between communicational interventions and higher patient satisfaction (RR, 1.41; 95% CI, 1.20-1.66). There was substantial heterogeneity among trials (I2 = 91.1%; P <.001) (Figure 2B). A subgroup analysis found an association of the intervention with patient satisfaction in trials using medication counseling (RR, 2.30; 95% CI, 0.45-11.85; between-group heterogeneity: P =.02). Of the 4 studies not included in the meta-analysis, 1 study found a significant improvement of satisfaction in patients who received a communication intervention.

23We found 14 trials (23%) that assessed mortality of patients within 30 days after discharge. Of these, 3 studies,, did not report any deaths. The remaining 11 studies (23%),,,,,,,,,,, including 1787 patients. Of these, 4 trials (36%) had a low risk of bias, 5 trials (45%) had a high risk, and 2 trials (18%) had an unclear risk. In the quantitative analysis, there were no significant associations between communication interventions at discharge and mortality within 30 days (RR, 0.70; 95% CI, 0.38-1.29). There was no evidence for publication bias (Egger test: P =.10). There was no heterogeneity among trials (I2 = 0.0%; P =.79), and there were no differences in any subgroup analyses.

24There were 11 studies (18%),,,,,,,,,, that assessed ED reattendance that were included in the quantitative analysis, including 9 trials (27%) with a low risk of bias, 1 trial (9%) with a high risk, and 1 trial (9%) with an unclear risk. There was no evidence for publication bias (Egger test: P =.28). Overall, there was no significant association between communication interventions at discharge and ED reattendance (RR, 0.86; 95% CI, 0.67-1.10). There was a moderate heterogeneity among trials (I2 = 48.3%; P =.04).

25There were 22 trials (37%) that evaluated the effect of communication interventions on patient knowledge within 30 days after hospital discharge, including 6 trials (27%) with low risk of bias, 13 trials (22%) with high risk, and 3 trials (14%) with unclear risk. Of these. 11 studies (50%),,,,,,,,,, evaluated patient knowledge of medication, 8 studies (36%),,,,,,, analyzed knowledge of diagnosis, and 3 studies (14%),, evaluated both.

26These studies used various interventions, such as medication counseling,,,,,,,, with reminder handouts, face-to-face counseling, or videos to educate about various aspects of the disease,,,,,,,, that were adapted, for example, to patients’ age, language, learning styles, or health literacy, or used specific communication strategies.,, Studies used either disease-specific knowledge questionnaires,,,,, and asked patients about their understanding of their diagnosis,, or their risk of adverse outcomes, to assess patient knowledge. Studies that assessed knowledge of medication primarily counted medication errors,, questioned patients about their treatment or its purpose,,,,,,, or used different structured questionnaires,, as their way of assessment. In summary, 13 studies (59%),,,,,,,,,,,, found an association of communication interventions with an increase in knowledge in patients.

27This systematic review and meta-analysis from 60 trials and 16 070 patients from 18 countries, including 19 RCTs and 3953 patients from 7 countries for analysis of the main outcome, found communication interventions at discharge to be associated with fewer hospital readmissions and improvement of treatment adherence and patient satisfaction. A subgroup analysis found associations in patients with respiratory illnesses regarding readmission, in older and hospitalized patients regarding adherence, and in trials using educational interventions regarding satisfaction.

28This study confirms previous individual RCTs suggesting that communication interventions are highly effective in reducing hospital readmissions. Several reasons for preventable hospital readmissions have been proposed, including low adherence to following instructions or treatment regimens., We found that educational interventions, such as medication counseling or disease-specific education, were associated with lower readmission rates. This finding suggests that educating patients at discharge regarding their medication, diagnoses, or therapeutic regimen might partly explain the lower risk of readmissions. According to our results, patients with chronic conditions, such as respiratory illnesses, experienced the most benefit from communication interventions with regard to readmission rates. Patients with chronic conditions (eg, chronic obstructive pulmonary disease) rely on stringent treatment plans and are encouraged to actively participate in their care. Communication interventions with an educational or counseling approach might have highlighted the importance of adherence and thus explain the difference between the subgroups.

29Previous research has highlighted the importance of family members in the discharge process of older patients, particularly those with frailty, delirium, or other cognitive deficits. Educational efforts in family members might also help to facilitate the discharge process in this cohort of patients with uniquely high risk. Compared with educational interventions, specific communication techniques, such as shared decision-making or motivational interviewing, were not associated with reduced readmission rates. In fact, 2 studies, using shared decision-making in patients with low-risk chest pain reported an association with fewer hospitalization days. In consequence, this might have resulted in higher rates of readmission to hospital and reattendance to the ED.

30Our data show that communication interventions were associated with an increase of adherence to treatment regimen. Adherence is known to be an independent factor associated with health-related outcomes, such as hospital readmission, mortality, morbidity, or quality of life. Especially in patients with chronic diseases, careful adherence to therapeutic regimens is of major importance, and approximately 40% of readmissions in these patients can be traced back to a lack of adherence. In a subgroup analysis, we found that communication interventions had were associated with better adherence in older and hospitalized patients from medical wards but not in younger patients and ED patients. This finding might be explained by the fact that most of these studies assessed interventions with an educational approach, as knowledge of disease and the purpose of the medication is known to be associated with adherence to therapeutic regimens in older adults.,

31Although adherence is considered a requirement for a successful treatment and several of its barriers are potentially modifiable, the topic is commonly not addressed during physician-patient encounters. Furthermore, physicians may not be able to estimate whether their patient adheres to the prescribed treatment. Thus, communication strategies at discharge addressing patient knowledge might improve patient adherence.

32Communication interventions at discharge were associated with improved patient satisfaction. Previous research has shown that patient satisfaction may not only affect patient outcomes but may also prevent complaints or even malpractice claims. Today, patient satisfaction is considered an important quality indicator in health care. Regulations in several countries, including the US, increasingly require that patients’ hospital experience be assessed through patient-related experience measurements, such as the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS). HCAHPS assesses patients’ perception of care delivered, and high patient satisfaction is linked to favorable HCAHPS metrics. Hospitals’ HCAHPS results are publicly accessible on the internet and not only influence a hospital’s reputation but also their reimbursements for care provided, which is why improving patient satisfaction has moved into the spotlight of health care systems. Our analysis suggests that communication interventions at discharge may improve quality of care and be considered a cost intervention. Still, cost-effectiveness studies are needed to understand costs associated with resource use and cost-savings resulting from the improvements in outcomes.

33The qualitative results of our systematic review suggest that communication interventions at discharge may help to increase patient knowledge regarding disease, its therapy, and further therapeutic regimen. It is known that patient knowledge is an independent factor associated with adherence to treatment regimens, which is of major importance for a patient’s recovery. Furthermore, with regard to patient-centered care, knowledge empowers patients to understand complex medical information and instructions and act accordingly. Hence, communication interventions at discharge might increase patient knowledge, helping patients to participate in clinical decision-making more effectively.

34Finally, several discharge pathways include patient handouts as supplementary information during the discharge process. Two systematic reviews, investigated the associations of discharge interventions that were facilitated by written information or information technology, such as videos, with patient comprehension and satisfaction. Newnham and colleagues investigated which hospital discharge communication practices were preferred by patients and health care practitioners and were associated with improved patient and practitioner satisfaction and increased patient understanding of their medical condition. Based on review of 30 trials (3489 patients), Newnham et al concluded that well-designed information technology solutions may improve communication, coordination, and retention of information. Hoek and colleagues aimed to provide an overview of the different manners of providing discharge instructions in the ED and to assess their associations with comprehension. Based on 51 included articles, Hoek et al concluded that communicating discharge instructions verbally to patients may not be sufficient and adding video or written information is needed. Our analysis focused on associations of communication interventions at hospital discharge with patient-relevant outcomes. Interestingly, in our subgroup analysis, interventions that had included written patient handouts did not show better patient outcomes in readmission or adherence compared with interventions with oral information only. Clearly, further research is needed to investigate optimal ways to combined oral and written information at discharge.

35This study has several limitations. First, as we were interested in interventions to improve patient-centered hospital-based care, we only included studies that focused on adult medical inpatients and excluded studies with outpatients, pediatric studies, and studies conducted in an outpatient setting. Also, we focused on communication interventions in isolation from each other and did not assess the complexity of a multidisciplinary discharge process. In clinical practice, interventions are often combined, and different members of the interprofessional team may deliver distinct discharge education. This approach might limit the generalizability of our results. Second, the included studies are very heterogeneous concerning the assessment of some end points, such as knowledge, which only allowed us to conduct a qualitative assessment and thus limits our ability to draw a systematic conclusion. Furthermore, we found evidence of publication bias for some secondary end points. Therefore, prospective validation is warranted. Also, some of the variables we selected for the subgroup analyses (eg, age, sex) may suffer from aggregation bias. Furthermore, we focused on studies in which communication interventions had finished at hospital discharge and disregarded studies with ongoing interventions, such as follow-up appointments or reminder messages or phone calls, which might also influence the patient-relevant outcomes that we assessed in our meta-analysis. Additionally, social determinants of health, such as race and ethnicity, educational level, and economic status, have an important role in the discharge process and might influence hospital readmission, treatment adherence, or medical knowledge. However, most of the studies we included in our meta-analysis only provided limited sociodemographic information, which made it impossible to stratify our results for these determinants.

36The findings of this systematic review and meta-analysis suggest that communication interventions at discharge are associated with reducing hospital readmissions and improving treatment adherence and patient satisfaction. Communication interventions at hospital discharge are important to facilitate the transition of care. Thus, health care systems should implement such communication strategies at discharge to facilitate the transition of care.