BACKGROUND

Patient-centered care has received increasing emphasis in recent years and can be defined as a partnership between patients and providers in which patients have the knowledge and opportunity to engage in shared decision-making.1,2,3 Patients’ comprehension of their disease, current therapies, and expected outcomes is fundamental to patient-centered care. Hospitalizations present not only opportunities for patient-centered care but also unique challenges given the complexity and rapid pace of clinical care. For some patients, the capacity to learn new information may be impaired by acute illness, medications, or prior knowledge gaps.

Perhaps not surprisingly, several studies have found that hospitalized patients have incomplete knowledge of important components of their care, such as admission diagnoses, planned tests and procedures, the names and purposes of their medications, and the names and roles of hospital team members.4,5,6,7,8 Furthermore, the literature describing patient knowledge differs in methods of assessment and domains measured.

We were unable to identify any prior systematic reviews on the state of patient knowledge in the inpatient setting. A summary of this issue is an important step in achieving patient-centeredness in the inpatient setting. Thus, we conducted a systematic review to characterize the current state of inpatients’ knowledge of their hospitalization, methods of knowledge assessment, and where applicable, effects of interventions on improving knowledge. Given the scope of this subject, we chose to focus on patient knowledge during hospitalization and shortly after discharge. When available, we sought to elucidate any association between patient knowledge and health behaviors (e.g., medication adherence) and clinical outcomes (e.g., hospital readmission), though this was not the primary aim of the review.

METHODS

Registration, Protocol, and Disclosures

This review was registered with Prospero (CRD42017059933). The authors declare that they do not have a conflict of interest.

Literature Search and Study Selection

A medical librarian (J.P.), with training in systematic review methodology, searched MEDLINE, EMBASE, and the Cochrane Library for articles published from January 1, 1995 through December 11, 2017. The following subject terms and keywords were used: inpatients, knowledge, comprehension, hospitalization, patient discharge, discharge planning, and goals of patient care (Appendixes 1, 2, and 3). We limited results to English language publications of randomized controlled trials, prospective analyses, retrospective analyses, case control, cohort, cross-sectional, and non-controlled before-and-after studies that were published in peer-reviewed journals. During review, eligible studies included patients under inpatient or observation status on internal medicine, family medicine, or neurology services. Studies on pediatric, surgical, obstetric, emergency room, or psychiatric patients were excluded, as were descriptions of patient perceptions, systematic reviews, and qualitative studies. We also excluded studies evaluating patient knowledge more than 7 days post-discharge.

Review Process

We used a three-step review process. One of the four authors (A.S., B.G., C.K., and L.O.) independently reviewed each title for inclusion using a web-based tool (Covidence systematic review software; Veritas Health Innovation). Early on, five authors (A.S., B.G., C.K., L.O., and K.O.) reviewed 80 studies as a group to ensure standardization of the review process. Subsequently, pairs of reviewers independently screened all remaining abstracts. When disagreements arose and primary reviewers could not reach a consensus, the other authors were consulted to resolve the disagreement. For abstracts identified as potentially relevant, two authors (A.S. and B.G.) reviewed the full-text articles, determining final inclusion of studies by consensus. When consensus could not be reached, a third reviewer (K.O.) resolved the disagreement.

Data Extraction and Synthesis

Two authors (A.S. and B.G.) extracted study characteristics (first author, year, country, study design, sample size, patient characteristics, methods, intervention, primary endpoints, results) from the final list of included studies. The authors independently abstracted the data and then reviewed and confirmed the accuracy of the others’ work. Studies were categorized by domain of knowledge assessed (e.g., medications, diagnosis) and method of assessment (e.g., self-report, comparison to primary source). The quality of each observational study was assessed by K.O. and A.S. or K.O. and B.G., using the NIH Quality Assessment Tool for Observation Cohort and Cross-Sectional Studies.9 The quality of each interventional study was assessed by both A.S. and B.G. using an adapted EPOC criteria from the Cochrane Collaboration tool for assessing risk of bias.10

Analysis

Whenever possible, we calculated Cohen’s d to estimate the effect size for continuous variables and odds ratios with confidence intervals for dichotomous variables. Due to the variation in domains of knowledge assessed and the heterogeneity of methods used, we were unable to pool results and conduct a meta-analysis.

RESULTS

Study Selection and Description

A total of 18,017 unique records were identified, 11 of which were identified from citations during full-text reviews (Fig. 1). A total of 17,439 records were excluded due to lack of relevance during title reviews and 432 were excluded during abstract review. The remaining 146 articles underwent full-text review, yielding 28 articles for inclusion.

Figure 1
figure 1

PRISMA flow diagram.

Full size image

The majority (17) of studies were observational. Of these, 15 were cross-sectional or descriptive and two were prospective cohort studies (Table 1). Eleven studies were interventional, among which five were randomized controlled trials, five were non-randomized controlled trials, and one was a historically controlled trial (Table 2).

Table 1 Non-interventional Studies
Full size table
Table 2 Interventional Studies
Full size table

Most studies focused on general adult populations; however, some evaluated patients admitted to a specialty ward or with a particular diagnosis. The most common specific diagnoses (9 of 27 studies) were heart failure, acute coronary syndrome, and pneumonia.17, 19,20,21, 23, 24, 29, 30, 35 Other studies focused on patients with HIV infection, stroke, organ transplantation, or patients taking warfarin.16, 26, 33, 36

Observational Studies

Among the 17 observational studies, 15 were cross-sectional or descriptive and two were prospective cohort studies (Table 1). The quality was assessed using the NIH Quality Assessment Tool for Observation Cohort and Cross-Sectional Studies.9 One study was rated as good quality, 11 were rated as fair, and 5 were rated as poor (Table 3).

Table 3 NIH Quality Assessment Tool for Observation Cohort and Cross-Sectional Studies
Full size table

Medication-Focused Studies

Eleven out of 27 studies assessed knowledge of prescription medications and generally found it to be poor (Table 1).6, 13, 15, 16, 20, 22, 27, 31, 32, 36, 37

Holloway assessed knowledge of prescribed medications in 20 hospitalized patients at admission and discharge.37 Upon admission, 63% of patients did not know the name, 95% did not know the dosage, 26% did not know the frequency, and 47% did not know the side effects of at least one of their medications. At discharge, 50% of patients did not know the name, 75% did not know the dosage, 30% did not know the frequency, and 45% did not know the side effects of at least one of their medications.

Makaryus and Friedman found similar results from discharge interviews of 43 patients.6 Only 28% knew their medication names, 37% the purpose, and 14% the common side effects.

Vrhovac et al. evaluated 183 inpatients’ medication knowledge (name, dosage, indication), separating medications prescribed prior to the hospitalization from those initiated during the hospitalization.13 Responses were deemed as fully correct, partially correct, or incorrect. Patients showed significantly better knowledge of medications taken prior to their hospitalization compared to those started during it. Patients responded with the correct name for 38% of their pre-hospitalization medications compared to 17% of medications during hospitalization, with the correct dosage for 77% of pre-hospital medications compared to 49% of medications during hospitalization, and with the correct indication for 72% of pre-hospital medications compared to 42% of medications during hospitalization (p < 0.001 for all comparisons). Patients older than 70 (p < 0.001) and those with lower educational attainment (p < 0.001) had significantly worse overall knowledge. Other variables, such as gender, total number of medications, and duration of treatment were not statistically associated with knowledge.

Similar to Vrhovac, Eibergen et al. evaluated 124 patients’ knowledge of changes to chronic medications.38 One week after discharge, 42% could correctly recall all of the medication changes. Recall was better for dose and frequency changes (51%) than for switched medications (40%) and discontinued medication (38%). Patient characteristics, including age, sex, education level, length of stay, number of medications, and number of medication changes were not associated with knowledge.

Cheah and Martens examined 50 patients on warfarin 1 week after discharge using a ten-part open-ended questionnaire.36 When asked, 75% of patients were unable to describe the meaning of the term “blood thinner” and many believed they were not at risk for bleeding if their INR was within therapeutic range. Only one-third understood the need to monitor vitamin K intake and one-third were unable to name three vitamin K-rich foods. In one-way ANOVA analysis, correct responses were significantly higher for those under the age of 65 (F = 6.28, p < .05).

Knowledge of General Aspects of Hospitalization

Seven of 27 studies focused on knowledge of care teams, plans of care, and discharge planning (Table 1).8, 18, 19, 25, 28, 31, 34

O’Leary et al. assessed multiple domains of knowledge in 241 patients on their second inpatient day.8 In regard to care teams, only 32% of patients correctly named their hospital physician and 11% knew her/his role, yet 60% were able to correctly name their nurse. For diagnosis and plans of care, patients’ responses were rated as being in total agreement, partial agreement, or no agreement with the primary hospital physician. There was no agreement between patients and their physicians 36% of the time for the diagnosis, 38% of the time for planned tests, 10% of the time for planed procedures, and 54% of the time for medication changes. Interestingly, on self-assessment, the overwhelming majority (95–99%) of patients indicated they knew their diagnosis, planned tests, planed procedures, and medication changes. There was no significant difference in agreement scores based on age, ethnicity, sex, or education level.

In a similar study, Horwitz et al. asked 377 patients to describe their diagnosis, discharge instructions, and post-discharge follow-up plan and compared their answers to the medical record.19 All participants were over the age of 65 with admitting diagnoses of pneumonia, acute coronary syndrome, or heart failure. Though 95% of patients indicated they understood their hospital diagnosis, only 60% were accurate when compared with the medical record.

Ní Chróinín et al. rated 336 patients’ understanding of their discharge diagnosis using a score ranging from 0 to 6.18 Understanding was dichotomized as good (≥ 5) or poor (< 5). Overall, 72% (243/336) had a good understanding of their discharge diagnosis (CI 67.5–77.1%). In patients 65 and older, 36% (64/177) (CI 29.1–43.2%) had a poor understanding of their diagnosis compared with 18% (29/159) (CI 12.2–24.2%) younger than 65 (p < 0.001). Worse understanding was also associated with residing in a nursing home, not recalling the presenting complaint, male gender, and having an Abbreviated Mental Test Score (AMTS) less than 8 (p < 0.001).

Patient Variables Affecting Knowledge

Two of 27 studies focused on patient-specific variables affecting knowledge, emphasizing education level and health literacy.17, 23

Regalbuto et al. characterized demographic and clinical features that predicted patient understanding of heart failure discharge instructions.23 A binary score of 0 or 1 was given for each of 6 Joint Commission requirements for heart failure discharge instructions. Overall, mean understanding was 4.1 (SD 1.2). Symptoms of heart failure exacerbation and weight and management were the best understood, whereas medications were least. Mean understanding for English speakers was 4.2 (SD 1.1) compared to 3.4 (SD 1.0) for non-native English speakers (p < 0.001). Mean understanding for patients with a college education or greater was 4.6 (SD 0.9) compared to 3.8 (SD 1.2) for those with a high school education or less (p < 0.001). Age, sex, race, and recent hospitalization were not associated with understanding. Of the patients studied, 23.4% were readmitted within 30 days. No patients with complete understanding were admitted within 30 days compared to 24.1% of patients with imperfect understanding (p = 0.044); however, this finding was no longer significant after adjusting for level of education and English as a second language.

Similarly, Dennison et al. surveyed 95 heart failure patients in an urban teaching hospital.17 Only 39% of interviewed patients had “adequate” health literacy as assessed by the Short Test of Functional Health Literacy in Adults (S-TOFHLA). Age was inversely correlated with literacy (r = 0.352; p < 0.001), whereas education level was positively correlated with literacy (r = 0.245; p = 0.017). Health literacy was also positively correlated to both heart failure knowledge (measured via the Dutch Heart Failure Knowledge Scale) and self-care knowledge (measured via the Self-Care of Heart Failure Index) (r = 0.465; p < 0.001). Participants with marginal health literacy had higher 30-day readmission rates, though this did not reach statistical significance (p = 0.116).

Interventional Studies

The 6 non-randomized studies ranged in size, enrolling 37 to 809 patients, of which 20 to 193 received interventions. The 5 randomized controlled studies enrolled 56 to 134 patients, of which 20 to 68 patients received interventions (Table 2). We used the seven EPOC domains to classify each interventional study as low, unclear, or high risk for bias, as it pertained to the primary outcome (Table 4). Overall, 3 studies were at low risk of bias, 1 was unclear, and 7 were at high risk of bias.

Table 4 Cochrane Effective Practice and Organization of Care (EPOCH): Summary of the Risk of Bias in Interventional Studies
Full size table

Some groups sought to improve health behaviors and outcomes through enhancing knowledge. Uniformly, interventions increased knowledge, but few studies evaluated the effect on health behaviors or outcomes.

Non-randomized Studies

Tan et al. aimed to increase knowledge of care teams, plans of care, and discharge planning using hospital room whiteboards.25 On a 1 to 5 ordinal scale, patients’ self-reported admission goal knowledge improved from 4.23 to 4.66 (p = 0.004).

Murphy et al. attempted a similar intervention using informational sheets completed during bedside rounds and left with the patient.28 Compared to controls, a greater proportion of patients in the intervention group knew their diagnosis (90% vs. 59%; p < 0.01), treatment plan (76% vs. 41%; p < 0.01), discharge criteria (76% vs. 25%; p < 0.01) and estimated date of discharge (83% vs. 52%; p < 0.01).

Louis-Simonet et al. studied the effect of structured patient-centered interviews on medication knowledge.27 The intervention group (193 patients) received individualized treatment cards and education emphasizing clarification of treatments and answering patient questions. The control group (616 patients) received usual instructions. The interviews significantly increased knowledge of medication indication (adjusted difference 6% (95% CI 3–8%); p < 0.001) and side effects (adjusted difference 19% (95% CI 9–29%); p < 0.001). Although there was no difference in the number of self-discontinued medications between the groups (p = 0.69), the study was not designed to rigorously assess compliance.

Randomized Controlled Trials (RCTs)

All but one of the RCTs increased knowledge. Although none resulted in improved health behaviors or outcomes, studies were either underpowered or did not report power calculations.

In Canada, Gwadry-Sridhar et al. randomized 134 heart failure patients to 2.5 h of individualized heart failure education in addition to standard care with an informational booklet and video (19). Using the Minnesota Living with Heart Failure questionnaire, the intervention group had significantly improved knowledge immediately after the intervention. Change in knowledge score was 2.24 ± 2.46 (95% CI 1.63–2.85) among intervention patients compared to 1.38 ± 2.16 (95% CI 0.85–1.91) in the control group (p = 0.02). Cohen’s d was 0.37, consistent with a small to moderate effect. There was no difference in 30-day utilization or medication adherence, although the authors comment that their sample size was likely too small to detect a significant difference in the composite outcome.

Marini et al. also used an educational video to affect knowledge about venous thromboembolism (VTE).32 The video increased VTE knowledge scores from 62 ± 17 to 83 ± 13% (p < 0.0001) with a Cohen’s d of 1.34 for VTE knowledge, consistent with a very large effect. However, there was no improvement in VTE prophylaxis adherence.

Domingues et al. evaluated the effect of a pre- and post-hospitalization nurse-led educational intervention on heart failure patients’ disease-specific knowledge, self-care knowledge, and post-discharge utilization.30 Both intervention and control groups had increased knowledge after discharge but there was no difference in knowledge scores or post-discharge utilization between groups.

In Australia, Lin et al. used a Patient-Directed Discharge Letter (PADDLE) to increase knowledge of diagnoses, tests, treatments, and discharge recommendations among 67 inpatients.34 They used a 1–5 rating scale for a maximum knowledge score of 20. Median total knowledge increased from 11.7 ± 2.6 to 15.5 ± 3.2 (p < 0.001) in the intervention group. Cohen’s d was 1.30, but change in knowledge for the control group was not reported. In a single domain, the percentage of patients attaining a 5 on the rating scale increased from 71 to 100% (p = 0.09) for diagnoses, 50 to 88% (p < 0.001) for tests, 50 to 100% (p = 0.001) for treatments, and from 27 to 80% (p < 0.001) for discharge recommendations. Readmission rates did not differ between the groups; however, there was insufficient power to detect a small difference in readmissions.

O’Leary et al. aimed to increase patient knowledge and activation, as measured by the Patient Activation Measure (PAM), by randomizing 202 medical patients to usual care or to an interactive patient portal.31 The portal provided information on the care team, medication list, and plans of care. More patients in the intervention group could name their physician (56% versus 29.6%; p < 0.001) and her/his role (47% versus 15.7%; p < 0.001); however, there was no difference in medication knowledge, knowledge of planned tests and procedures, or patient activation.

DISCUSSION

In summary of the available literature, we found that the current state of inpatients’ knowledge of their hospitalization is poor, especially when it comes to knowledge of medications, diagnoses, and plans of care. Domains of knowledge assessed varied across studies and methods of evaluating knowledge were often author-derived and study specific. Interventions aimed at improving knowledge generally worked, but evidence to support an effect on behaviors and outcomes is limited by the fact that most studies were either underpowered or did not report power estimates for these measures. Several themes and important inferences arose.

Patient characteristics were variably associated with comprehension. Older individuals (usually defined as 65 years and older) and those with lower levels of education (usually defined as less than a college degree) appeared more likely to have significant knowledge gaps.6, 13, 18, 36 Although less frequently assessed, lower health literacy and cognition were also associated with knowledge deficits.10, 38 Other demographic factors, such as gender, marital status, length of hospital stay, and total number of medications prescribed were not associated with patient understanding.6, 8, 13, 16 Given these findings, providers could consider targeting educational interventions to older patients and those with barriers to learning about their health.

The inpatient setting could provide a particularly valuable opportunity to focus on improving medication knowledge, as this was consistently poor across studies. Though patients’ capacity to learn may be adversely affected by acute illness, educational interventions were able to increase knowledge and patients may be uniquely motivated to learn while recovering from an acute episode of illness. A variety of interventions appeared to be successful. Further research should compare strategies to identify an optimal approach.

We also found that patients are generally poor estimators of their own knowledge. The literature suggests that patients overestimate their knowledge when compared against objective measures.8, 18 The discrepancy between patient perception of their knowledge and objective assessments of this knowledge is an important reminder that physicians should routinely confirm patient comprehension during clinical discussions.

Though many interventions increased patient knowledge, the existing evidence is not robust enough to make conclusions regarding improvement in health behaviors or outcomes. Relatively few studies assessed health behaviors or outcomes and those that did include these measures, either did not report power calculations or reported that they were underpowered. Importantly, health behaviors such as medication adherence are influenced not only by knowledge but also the patient’s agreement with the medication, activation (self-efficacy, motivation), and access (coverage, expense, etc.).39,40,41 Clinical outcomes, such as hospital readmissions are influenced by many factors as well, including underlying illness, social support, home health resources, access to outpatient care, and patient and location-specific factors.

Our systematic review has several limitations. First, we found relatively few high-quality studies. Second, studies varied with regard to methodology and domains of knowledge assessed. Hence, we were unable to conduct a meta-analysis. Third, while most interventional studies were adequately powered to detect improvements in knowledge, few were powered to detect significant differences in health behaviors or outcomes. Finally, few studies in English-speaking countries included non-English-speaking patients, a group at especially high risk for deficits in comprehension.

In summation, we found that studies do not use a standard method of knowledge assessment and that patient comprehension of hospital care is overall poor, especially in certain demographics. Given that patient comprehension is a fundamental principle of patient-centered care, we are reassured that interventions generally improve knowledge. Thus, we recommend targeted interventions at hospitalized patients, using standardized methods of assessment, that aim to improve knowledge while also addressing factors influencing health behaviors and outcomes.