Health Literacy and 30-Day Postdischarge Hospital Utilization

Abstract

Low health literacy is associated with higher mortality, higher rates of hospitalization, and poor self-management skills for chronic disease. Early, unplanned hospital reutilization after discharge is a common and costly occurrence in U.S. hospitals. Still, few studies have examined the relation between health literacy and 30-day hospital reutilization rates. The authors examined the association between health literacy and 30-day reutilization of hospital services (readmission or return to the emergency department) in an urban safety net hospital, and conducted a secondary analysis of data from the control arm subjects of the Project RED and the RED-LIT trials. Health literacy was measured using the REALM tool. The primary outcome was rate of 30-day reutilization. The authors used multivariate Poisson regression analysis to control for potential confounding. Of the 703 subjects, 20% had low health literacy, 29% had marginal health literacy, and 51% had adequate health literacy. Sixty-two percent of subjects had a 12th-grade education or less. Subjects with low health literacy were more likely to be insured by Medicaid (p < .001); Black non-Hispanic (p < .001); unemployed, disabled, or retired (p < .001); low income (p < .001); and less educated (high school education or less, p < .001). The fully adjusted incidence rate ratio for low health literacy compared with adequate health literacy was 1.46 (CI [1.04, 2.05]). Low health literacy is a significant, independent, and modifiable risk factor for 30-day hospital reutilization after discharge. Interventions designed to reduce early, unplanned, hospital utilization after discharge should include activities to mitigate the effect of patients’ low health literacy.

Nearly 20% of Medicare patients are readmitted to the hospital within 30 days of discharge (Jenks, Williams, & Coleman, 2009). Known predictors of early readmission include the following: lower socioeconomic status (Weissman, Stern, & Epstein, 1994), history of prior hospitalization (Van Walraven, Mamdani, Fang, & Austin, 2004) and advanced age (Cho, Lee, Arozullah, & Crittenden, 2008; Marcantonio et al., 1999), length of stay greater than 7 days (Krumholz et al., 1997), a high burden of comorbid illnesses (based on Charlson score; Librero, Peiro, & Ordinana, 1999), and specific diagnoses (e.g., depression, congestive heart failure, chronic obstructive pulmonary disease, myocardial infarction; Krumholz et al., 1997; Mitchell et al., 2010; Parashar et al., 2006; Wong et al., 2008). Meanwhile, a streamlined discharge process designed to enhance patient safety during care transitions can reduce 30-day postdischarge hospital utilization as much as 30% (Jack et al., 2009). The 30-day rehospitalization rate has emerged as an important indicator of hospital quality, particularly because the Centers for Medicare and Medicaid Services introduced a series of new reimbursement policies that include decreasing payments to hospitals with high rates of rehospitalization and bundled payment schemes in which accountable care organizations will be paid less for subsequent hospital utilization within 30 days. Thus, there is great interest in identifying modifiable risk factors for rehospitalization that could be used to refine intervention models and lead to improvements in quality of care, patient outcomes, and cost savings.

Low health literacy has been linked to poor health outcomes, particularly for patients with chronic conditions such as diabetes, asthma, cancer, depression (Lincoln et al., 2006), HIV/AIDS (Kalichman & Rompa, 2000), and heart failure (Peterson et al., 2011). Health literacy, defined as the degree to which individuals have the capacity to obtain, process, and understand health information, skills, and services needed to make informed health decisions and take informed actions, affects many Americans (Paasche-Orlow, 2011; U.S. Department of Health and Human Services, 2000). An estimated 26% of the U.S. population has low health literacy, and an additional 20% has marginal health literacy (Paasche-Orlow et al., 2005). Health literacy barriers are often associated with greater risk of hospitalization (Baker, Parker, Williams, & Clark, 1998), higher rates of self-reported poor health status (Baker, Parker, Williams, Clark, & Nurss, 1997) decreased knowledge of one's medical condition, poor medication recall, nonadherence to treatment plans, poor self-care behaviors (Evangelista et al., 2010), and increased all-cause mortality (Wolf, Feinglass, Thompson, & Baker, 2010). Patients with low health literacy also are more likely to report unsatisfactory patient–doctor communication at the time of discharge, suggesting that some physicians may be insensitive or unaware when their patients are having difficulties comprehending discharge instructions (Kripilani et al., 2010). Although several of these factors have been linked to an increased risk for 30-day rehospitalization, very limited evidence has been presented to demonstrate the independent association between low health literacy and hospital reutilization within 30 days of discharge.

This study examined the relation between health literacy and hospital reutilization within 30 days of discharge at the Boston Medical Center. The Boston Medical Center is the largest safety net hospital in New England, providing a spectrum of medical services to an urban, socially and economically diverse population. We hypothesized that low health literacy would be an independent risk factor for early unplanned hospital reutilization after discharge for general medical patients (i.e., adult patients admitted for acute general medical conditions such as pneumonia, unstable angina, pancreatitis, acute renal failure).

Method

We conducted a secondary analysis of the Project RED (Re-Engineered Discharge) and RED-Lit clinical trial data sets (clinicaltrials.gov identifier: NCT00252057) to assess the association between health literacy and the rate of subsequent 30-day hospital reutilization. The original Project RED included 738 participants and RED-Lit included 802 participants. All Project RED studies enrolled English-speaking patients 18 years or older who were admitted to a general medical unit at the Boston Medical Center. Study subjects were required to have telephone access and be able to convey an understanding of study procedures and other consent elements in English. Participants were excluded if their admission was planned, they were on suicide watch, transferred from another health facility or were deaf or blind. Outcome data was not used in the secondary analysis if participants withdrew consent, died during the index admission, or were not discharged to the community. The combined sample included 1,540 patients from the control and intervention arms with complete information for the primary independent variable of interest—health literacy—and the outcome variable—30-day hospital utilization. We used a final sample of 703 patients identified from the control arms of each of these trials. Subjects from the intervention groups were excluded to eliminate effect modification introduced by exposure to the intervention.

Key Outcome Variables

The primary outcome variable for this analysis was a combined count measure of emergency department and hospital utilization events by a patient within a 30-day period after the index discharge. We also examined the emergency department revisit and hospital readmission count outcomes separately for the purposes of identifying factors associated with these distinct events. The number of utilizations ranged from 0 to 15; however, we top-coded to a count of 8 to avoid undue influence of outliers. We collected outcome data using the Boston Medical Center electronic medical record or participant self-report obtained by phone interview after 30 days. Both reutilizations of the Boston Medical Center and other hospitals and emergency departments were included.

Primary Independent Variable

We measured health Literacy using the Rapid Estimate of Adult Literacy in Medicine (REALM), which is a 66-item validated word recognition test (Davis et al., 1993). The REALM assigns a grade level of literacy, with scores of 0–18 corresponding to literacy of third grade or below, 19–44 to 4th–6th grade, 45–60 to 7th–8th grade, and 61–66 to 9th grade or above. For the purposes of our analysis, the two categories of lowest literacy were combined because of the distribution of scores. We administered the REALM in person to study participants.

Statistical Analysis

Descriptive Statistics

We performed bivariate analyses to assess the unadjusted relation between demographic and clinical characteristics and the three REALM categories of health literacy. The crosstabs reflect the results of chi-square tests. We then conducted Poisson regressions for all three outcome variables (combined reutilization, rehospitalization and return to the emergency department) to control for potential confounding. Predictors were included into the multivariable Poisson regressions on the basis of their effect on the association between health literacy and the outcomes, as well as the significance of their independent associations with the outcomes. Predictors were chosen from: age, gender, marital status, income, insurance, employment, education, race, having a primary care physician, being homeless in the 6 months before the index admission, evidence of depressive symptoms, frequent utilizer status, length of stay, and medical comorbidity.

Adjusted Charlson Score for Comorbidity

Because of differences in the methods used to calculate Charlson scores between the RED and RED lit trials, we calculated a correction factor using a linear transformation procedure to adjust for the inconsistency in the distributions of Charlson scores for the combined analytic dataset. Charlson scores for the RED Lit II sample were calculated in a manner that accounted for comorbid conditions using all available records. For Project RED and RED Lit I, this process did not include outpatient record review. To derive a comparable Charlson score for the complete dataset, we calculated a z-score for each RED-lit II participant's observed score using the mean and standard deviation for the RED-lit II sample. Adjusted Charlson scores were then calculated for RED and RED-lit I (using the mean and standard deviation of this sample) to correspond to the z scores from the more comprehensively calculated sample.

Adjustment for Confounding

We constructed multivariable Poisson regression models using thorough backwards selection processes for each of the three outcome variables. We included certain variables (e.g., race, education, insurance) into the models because of their established relations with the variables of interest. We used SAS 9.1 to conduct the analysis with two-sided tests with p < .05 to judge significance.

Results

Of the 703 subjects, 138 (20%) patients had low health literacy (≤6th grade or REALM score of 0–18), 207 (29%) had marginal health literacy (7th to 8th grade or REALM score of 19–44) and 358 (51%) had adequate health literacy (≥9th grade or REALM score of 45–60). Study participants’ mean age was 49.2 years of age, which did not differ by REALM score (see Table 1). Patients with low health literacy were more likely to be insured by Medicaid (p < .001); Black (p < .01), unemployed, disabled, or retired (p < .001); low income (<$40,000/year, p < .001); and less educated (high school education or less, p < .001). Of subjects, 29% (n = 206) were frequent utilizers, defined as two or more admissions in 6 months before index admission. The mean length of stay for the index admission was 2.8 days and did not differ significantly by REALM score. The mean Charlson Comorbidity Index score was 0.6, with a majority of the sample reporting a score of 0. There was no relation between comorbidity and health literacy.

Table 1. Sociodemographic characteristics by health literacy status

	Health literacy status, by REALM* score
	1 ≤ 6th-grade level (n = 138)	2 7th–8th-grade level (n = 207)	3 ≥ 9th-grade level (n = 358)	Total	p
Gender
Female	63 45.7	113 54.6	196 54.9	372	.15
Male	75 54.4	94 45.4	161 45.1	330
Total	138	207	357	702
Marital status
Single, never been married	63 45.7	101 48.8	175 48.9	339	.07
Divorced, separated, or widowed	36 26.1	56 27.1	104 29.1	196
Unknown	7 5.1	1 0.5	5 1.4	13
Married	32 23.2	49 23.7	74 20.7	155
Total	138	207	358	703
Income
No income–$19,999	35 28.0	71 35.9	72 20.6	178	<.001
$20,000–$39,999	74 59.2	80 40.4	140 40.0	294
$40,000–$74,999	12 9.6	30 15.2	70 20.0	112
Unknown or refused	0 0.0	7 3.5	22 6.3	29
Total	125	198	350	673
Insurance
Free Care**	20 14.5	29 14.0	47 13.1	96	.02
Medicaid	70 50.7	97 46.9	133 37.2	300
Medicare	24 17.4	21 10.1	35 9.8	80
Other or unknown	8 5.8	10 4.8	30 8.4	48
Private	16 11.6	50 24.2	113 31.6	179
Total	138	207	358	703
Employment
Disabled or injured	46 35.1	53 25.7	74 21.0	173	.04
Part time	12 9.2	23 11.2	38 10.8	73
Retired	24 18.3	23 11.8	54 15.3	101
Student	2 1.5	4 1.9	9 2.6	15
Unemployed	29 22.1	56 27.2	61 17.3	146
Full time	18 13.7	47 22.8	116 33.0	181
Total	131	206	352	689
Education
Incomplete high school	60 43.5	62 30.0	44 12.3	166	<.01
Complete high school	55 39.9	91 44.0	120 33.5	266
Some college	9 6.5	33 15.9	106 29.6	148
Unknown	6 4.4	2 1.0	3 0.8	11
Complete college	8 5.8	19 9.2	85 23.7	112
Total	138	207	358	703
Race
Black	83 60.1	134 64.7	152 42.5	369	<.01
Hispanic	21 15.2	23 11.1	34 9.5	78
Other	7 5.1	9 4.4	16 4.5	32
Unknown	9 6.5	10 4.8	9 2.5	28
White	18 13.0	31 15.0	147 41.1	196
Total	138	207	358	703
Has primary care physician
No	28 20.3	38 18.4	57 15.9	123	.48
Yes	110 79.7	169 81.6	301 84.1	580
Total	138	207	358	703
Been homeless in the past 6 months
Yes	22 15.9	24 11.7	40 11.2	86	.33
No	116 84.1	182 88.4	318 88.8	616
Total	138	206	358	702
Depressive symptoms***
Depressive	40 29.0	45 21.7	78 21.8	163	.19
Not depressive	98 71.0	162 78.3	280 78.2	540
Total	138	207	358	703
Frequent utilizer status
Two or more admissions in the 6 months before index admission	41 29.7	54 26.1	111 31.0	206	.46
Fewer than two admissions in the 6 months before the index admission	97 70.3	153 73.9	247 69.0	497
Total	138	207	358	703
Age, M (SD)	51.23 (13.91)	49.42 (13.51)	48.14 (14.77)	702
Length of stay, M (SD)	2.69 (2.57)	2.86 (2.52)	2.72 (2.38)	702
Charlson Comorbidity Index, M (SD)	0.719 (1.592)	0.595 (1.382)	0.499 (1.341)	703	.03

The unadjusted 30-day postdischarge hospital reutilization incidence rate ratio for subjects with low health literacy compared to subjects with adequate health literacy was 1.76 (95% CI [1.21, 2.55]). After adjusting for potential confounding using a multivariate Poisson regression analysis—which included education, gender, marital status, income, race, affiliation with primary care provider, homelessness, depression, frequent utilizer status, age, length of stay, and the Charlson Comorbidity Index—we found that patients with low health literacy are 1.46 times (95% CI [1.04, 2.05]) more likely than patients with adequate health literacy to return to the hospital or emergency department within 30 days. Frequent utilizers were 2.04 times more likely than nonutilizers (incidence rate ratio = 2.04 [1.58, 2.62] p < .0001) and those who were homeless in the 6 months before their index admission were 1.54 times more likely than those who were not homeless (incidence rate ratio = 1.54 [1.14,2.08] p = .0053; see Table 2) to return to the hospital or emergency department. Female gender (incidence rate ratio = 0.75 [0.58, 0.96] p = .02) and Black race (incidence rate ratio = 0.73 [0.55, 0.98] p = .04) were both associated with lower risk for early hospital reutilization. In subanalyses of the outcome variable, we used multivariate Poisson regression to examine incidence rate ratios for emergency department visits and hospital readmissions separately. We found that, compared with patients with adequate health literacy, low health literacy patients were 1.71 times more likely to be readmitted to the emergency department (p < .05) and 1.67 (95% CI [0.98, 2.83], p < .06, times more likely to be readmitted into the hospital within 30 days of index admission.

Table 2. Adjusted incident rate ratio of hospital utilization within 30 days of discharge

	Multivariate Poisson regression with reutilization
	Estimate	Wald 95% confidence limits		p
Intercept	0.41	0.20	0.82	.01
REALM
≤6th grade	1.46	1.04	2.05	.03
7th and 8th grade	1.36	1.00	1.85	.05
≥9th grade	1.00	1.00	1.00
Education
Incomplete high school	0.96	0.62	1.48	.85
High school or equivalent	1.18	0.80	1.72	.40
Incomplete college	0.87	0.55	1.36	.54
Unknown	0.65	0.14	3.13	.59
Complete college	1.00	1.00	1.00
Gender
Female	0.75	0.58	0.96	.02
Male	1.00	1.00	1.00
Marital status
Single, never married	0.86	0.63	1.19	.37
Divorced, separated, or widowed	1.05	0.74	1.49	.78
Unknown	0.76	0.21	2.72	.68
Married	1.00	1.00	1.00
Insurance
Free Care*	1.15	0.73	1.80	.55
Medicaid	1.24	0.89	1.74	.21
Medicare	1.04	0.64	1.68	.88
Other or unknown	0.96	0.49	1.91	.92
Private	1.00	1.00	1.00
Race
Black	0.73	0.55	0.98	.04
Hispanic	0.73	0.47	1.12	.14
Other	0.76	0.46	1.25	.28
White	1.00	1.00	1.00
Has primary care physician
No	0.96	0.69	1.34	.82
Yes	1.00	1.00	1.00
Been homeless
Yes	1.54	1.14	2.08	.005
No	1.00	1.00	1.00
Depressive symptoms**
Depressive	1.24	0.96	1.60	.09
Not depressive	1.00	1.00	1.00
Frequent user
Two or more admissions	2.04	1.58	2.62	<.0001
Fewer than two admissions	1.00	1.00	1.00
Age, continuous, increment of 1 year	0.99	0.98	1.00	.14
Length of stay, continuous, increment of 1 day	0.99	0.94	1.04	.61
Charlson Comorbidity Index, continuous, increment of 1 unit	1.00	0.91	1.09	.92

Discussion

Concern about 30-day unplanned hospital reutilization is warranted as these events correlate with higher morbidity, mortality, and costs and is being used as a marker of the quality of hospital care (Ashton, Del Junco, Souchek, Wray, & Mansyur, 1997). Our study suggests that patients with low health literacy are more likely than patients with adequate health literacy to return to the hospital or emergency department within 30 days of discharge. These results were robust and health literacy remained an independent predictor of hospital utilization within 30 days of discharge after adjusting for a range of potential confounding phenomena, including education level. This study adds to growing evidence demonstrating the negative consequences of limited health literacy on patients’ health and well-being (Davis & Wolf, 2004). It also establishes a link between health literacy and the process of care transitions, which has emerged as a centerpiece of efforts to improve quality and decrease cost.

There are several potential reasons as to why patients with low health literacy may return to the hospital soon after their discharge relative to those with higher health literacy levels. Health literacy may prevent patients from understanding their discharge instructions—including proper comprehension of their diagnosis and the treatment regimen (Anonymous, 2009; Williams et al., 1995). Williams and colleagues who showed that patients with low health literacy misread medication dosing (23.6% incorrect) and appointment slips (39.6% incorrect) in a survey of hospitalized patients while those with adequate literacy did well on these tasks. Patients may also have difficulty managing self-care instructions after hospital discharge, such as understanding symptoms of medication side effects and how to mitigate such phenomena, appreciating early signs of disease exacerbation and responding appropriately, and accessing and utilizing routine and urgent outpatient services. These issues may all contribute to how health literacy is related to returning to the hospital.

Our results bring attention to the need for interventions designed to decrease the rate of unplanned hospital reutilization to focus on health literacy. Such interventions can augment the level of patient and family education and guidance that is provided. Agency for Healthcare Research and Quality offers an evidence-based toolkit to address health literacy titled, “Health Literacy Universal Precautions Toolkit.” The toolkit notably addresses four strategies for overcoming health literacy barriers. These include the following: (a) using easily understood spoken communication, (b) modifying written communication, (c) teaching self-management and empowerment, and (d) bolstering patient's support systems. Agency for Healthcare Research and Quality's approach advocates for creating an environment in which patients of all health literacy levels can thrive. Some of these methods include the following: drawing pictures, using plain (nonmedical) language, and using the teachback approach (Brach et al., 2012; Koh et al., 2012; Weiss, 2003).

Limitations

Major strengths of the present study are the reasonably large dataset and the broad range of covariates available for analyses. However, several limitations should be noted. First, this study was conducted using data from clinical trials implemented at a single safety net hospital; therefore, results may not be generalizable to other patient populations. Further, reutilization events outside of the Boston Medical Center were collected by subject self-report but were not independently confirmed. However, we were able to confirm 91% of all events by medical record review. Last, although we attempted to account for known confounders, other factors may also exist and could remain unaddressed.

Conclusion

Our study suggests that low literacy is significantly associated with a higher rate of 30-day postdischarge hospital utilization. Patient health literacy plays an influential role in health outcomes and low health literacy can be a significant barrier to patients’ safe transitions from hospital to home. Future directions include interventions to improve patient education for care transitions, reduce the health literacy burden of the discharge process, and remove unnecessary complexity from critical self-care tasks.

Acknowledgments

This work was supported by the Agency for Health Research and Policy 1K08HS019771-01 (Dr. Mitchell). The authors acknowledge Ms. Jessica Martin for her support in data preparation.

Notes

Note. Some columns may not add up to 100% because of omission of the “Other” categories. REALM: Rapid Estimate of Adult Literacy in Medicine.

*REALM scores are typically categorized into four categories; however, because of the distribution of the scores, the lowest two categories were combined into one category, resulting in three REALM categories.

**Free Care was a program in Massachusetts that aimed to fund medical care for uninsured individuals.

***Positive depression symptom screen determined by the Patient Health Questionnaire-9 screen tool, a nine-item 4-point Likert scale, standard scoring algorithm to screen for major and minor depression. A score of 5 or higher indicates a positive depression symptom screen.

Note. Some columns may not add up to 100% because of omission of the “Other” categories. REALM: Rapid Estimate of Adult Literacy in Medicine.

**Free Care was a program in Massachusetts that aimed to fund medical care for uninsured individuals.

**Positive depressive symptom screen determined by the Patient Health Questionnaire-9 screen tool, a nine-item 4-point Likert scale, standard scoring algorithm to screen for major and minor depression. A score of 5 or higher indicates a positive depression symptom screen.