Low literacy defined as ≤sixth-grade level; higher literacy, as >sixth-grade
level. *Refused to sign consent and participate.†Due
to glycosylated hemoglobin level <8.07%, non–English speaking, or
life expectancy <6 months.
Adjusted differences at 12 months: A, –1.4 (95% confidence interval
[CI], –2.3 to –0.6), P<.001; B, –0.5
(95% CI, –1.4 to 0.3), P = .21; C, –7.9
(95% CI, –17.7 to 1.9), P = .11; and D, –7.1
(95% CI, –14.3 to 0.004), P = .05. Differences
are adjusted for baseline glycosylated hemoglobin (HbA1c) level
or systolic blood pressure (SBP) and for age, race, sex, income, insulin use,
and duration of diabetes. Error bars indicate interquartile range (for HbA1c) and standard deviation (for SBP).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rothman RL, DeWalt DA, Malone R, et al. Influence of Patient Literacy on the Effectiveness of a Primary Care–Based Diabetes Disease Management Program. JAMA. 2004;292(14):1711–1716. doi:10.1001/jama.292.14.1711
Context Low literacy is an important barrier for patients with diabetes, but
interventions to address low literacy have not been well examined.
Objective To examine the role of literacy on the effectiveness of a comprehensive
disease management program for patients with diabetes.
Design, Setting, and Participants Analysis of the influence of literacy on glycemic control and systolic
blood pressure using data from a randomized controlled trial (conducted from
February 2001 through April 2003) of a comprehensive diabetes management program.
Participants were 217 patients aged 18 years or older with type 2 diabetes
and poor glycemic control (glycosylated hemoglobin [HbA1c] levels
≥8.0%) and presenting to a US academic general internal medicine practice.
Interventions All communication to patients was individualized and delivered to enhance
comprehension among patients with low literacy. Intervention patients received
intensive disease management from a multidisciplinary team. Control patients
received an initial management session and continued with usual care.
Main Outcome Measures Achievement of goal HbA1c levels and systolic blood pressure
at 12-month follow-up for control and intervention patients stratified by
Results Complete 12-month data were available for 193 patients (89%). Among
patients with low literacy, intervention patients were more likely than control
patients to achieve goal HbA1c levels (≤7.0%) (42% vs 15%, respectively;
adjusted odds ratio [OR], 4.6; 95% confidence interval [CI], 1.3 to 17.2; P = .02). Patients with higher literacy had similar
odds of achieving goal HbA1c levels regardless of intervention
status (24% vs 23%; adjusted OR, 1.0; 95% CI, 0.4 to 2.5; P = .98). Improvements in systolic blood pressure were similar
by literacy status.
Conclusions Literacy may be an important factor for predicting who will benefit
from an intervention for diabetes management. A diabetes disease management
program that addresses literacy may be particularly beneficial for patients
with low literacy, and increasing access to such a program could help reduce
Approximately 90 million Americans have literacy skills insufficient
to function in today’s economy and health care settings.1-3 Patients
with low literacy can have trouble reading prescriptions and following medical
recommendations, poorer knowledge of their disease, and worse clinical outcomes.2,4-8 Low
literacy is common among patients with diabetes and is associated with poor
knowledge about diabetes.2,8-12 Although
1 recent cross-sectional study has found that low literacy is associated with
poor glycemic control,11 other studies have
No published studies have rigorously examined interventions that can
mitigate literacy-related disparities in patients with diabetes.14 We
previously conducted a pilot study that suggested that a comprehensive intervention
might improve glycemic control for patients with low literacy,15 but
that study lacked a control group. To better examine this issue, we recently
completed a randomized controlled trial of a comprehensive disease management
program that included strategies to overcome clinician deficits and patient
barriers, including low literacy, for patients with diabetes and poor glycemic
control. This program successfully improved blood pressure and glycemic control.16 This article examines how patient literacy influenced
the effectiveness of this program.
The data come from a randomized controlled trial that examined the effect
of an intensive diabetes management program (Figure 1).16 The study was conducted
in a university general internal medicine practice that serves a wide socioeconomic
range of patients. The study was initiated in February 2001 and completed
in April 2003. All patients were followed up for 1 year. The University of
North Carolina institutional review board approved the study.
Eligible patients included all adults (aged ≥18 years) with type
2 diabetes who were followed up for their diabetes care in the general internal
medicine practice, had poor glucose control (ie, glycosylated hemoglobin [HbA1c] levels ≥8.0%), spoke English, and had a life expectancy greater
than 6 months. Primary care clinicians referred eligible patients for possible
After written informed consent and baseline measures were obtained,
all patients attended a 1-hour educational session conducted by a clinical
pharmacist. The pharmacist provided treatment recommendations about glycemic
control and cardiovascular risk reduction to patients’ primary care
clinicians. After this session, patients were randomized with concealed allocation.
Patients in the control group received usual care from their primary
care clinician and had no further contact with the disease management team.
In the intervention group, usual care was supplemented by intensive diabetes
management from 3 clinical pharmacist practitioners and a diabetes care coordinator
(DCC). Pharmacists had training in outpatient disease management and 2 were
certified diabetes educators. The intervention included (1) one-to-one educational
sessions including counseling and medication management by the pharmacists
and the DCC; (2) application of evidence-based treatment algorithms (available
at http://www.med.unc.edu/medicine/edursrc/algor.htm) to help manage
glucose and cardiovascular risks by allowing pharmacists to both initiate
and titrate blood pressure– and glucose-lowering medications; and (3)
strategies to address patient barriers provided by the DCC, including telephone
reminders and, when needed, addressing difficulties with transportation, communication,
A pharmacist or the DCC contacted intervention patients by telephone
or in person every 2 to 4 weeks (more frequently if indicated). The pharmacist
and the DCC were aware of patients’ literacy status. Communication to
patients was individualized using techniques that enhance comprehension among
patients with low literacy,17-20 including
predominantly verbal education with concrete, simplified explanations of critical
behaviors and goals; “teach-back”21,22 to
assess patient comprehension; and picture-based materials. Main topics, revisited
throughout the follow-up period, included treatment goals, identification
of hypoglycemic and hyperglycemic symptoms, prevention of long-term complications,
Levels of HbA1c and systolic blood pressure (SBP) were collected
at baseline and at 6 and 12 months. Levels of HbA1c were measured
by staff at the University of North Carolina Hospital laboratories, who were
unaware of patients’ study status. Clinic nurses, unaware of study assignment,
recorded SBP from automated monitors. Literacy was assessed at enrollment
using the Rapid Estimate of Adult Literacy in Medicine.23,24 Race
was classified by patient self-report, and options were defined by the patient.
We measured race because of its role in disparities in diabetes care and in
During the intervention, the disease management team documented process
measures for all intervention patients, including time spent in direct contact
with the patients (including the initial management session), actions related
to patient care (eg, scheduling appointments, education), and medication changes
(initiating new medications, titrating current medications).
We performed all analyses using SAS version 8.02 (SAS Institute Inc,
Cary, NC). At baseline, we compared patients by intervention status and literacy
status using t tests or Wilcoxon rank-sum tests for
continuous variables and χ2 tests or the Fisher exact test
for categorical variables.
Our primary outcomes were improvement in HbA1c levels and
SBP from baseline to 12 months, stratified by literacy status. A priori, we
stratified literacy at the sixth-grade level based on research from the scale
developers and our own previous research.15,28 We
performed analysis of covariance, adjusting for baseline values (HbA1c level or SBP). We then conducted multivariable general linear models,
adjusting for baseline covariates if there were differences (P<.20) between groups or if we believed a priori that variables
were clinically relevant. This resulted in the inclusion of age, race, sex,
income, insulin status at enrollment, and duration of disease, in addition
to baseline HbA1c level or SBP in the multivariable models. We
did not include education or insurance status because they were highly correlated
with other covariates and were not associated with our outcome. Results were
very similar for unadjusted and adjusted analyses; adjusted analyses are presented.
We also examined the effect of the intervention on the proportion of
patients who attained recommended goals for HbA1c level (≤7.0%)
and SBP (≤130 mm Hg)29 at 12 months. We
first performed χ2 analyses to obtain crude odds ratios (ORs)
of obtaining goal HbA1c levels and SBP at 12 months between control
and intervention groups, stratified by literacy status. We then used logistic
regression to adjust for the covariates used in our primary analysis. Interaction
between literacy status and treatment assignment was assessed by including
the interaction term in the multivariable logistic regression model.
Finally, in an intent-to-treat analysis in which we carried forward
baseline or 6-month HbA1c and SBP values for missing outcomes,
we found similar results for all analyses. We also examined differences in
process measures using Wilcoxon rank-sum tests to compare intervention patients
with low and high literacy levels.
Based on a 2-sided significance level of.05 and 80% power, we estimated
a sample size of 107 patients per group to detect a 1% difference in HbA1c level. To detect a 10–mm Hg between-group difference in SBP
required 93 patients per group. The study was not powered to detect differences
by literacy status.
Follow-up data were available for 193 of 217 enrolled patients (89%)
at 12 months (Figure 1). Baseline patient
characteristics were similar between the control and intervention groups and
reveal a population with low socioeconomic status and poor glycemic control
(Table 1). More than one third of patients
had low literacy (≤sixth grade). Patients with low literacy were more likely
to be older, be African American, report lower income, and have lower reported
educational attainment and less diabetes-specific knowledge. Other clinical
characteristics did not differ by literacy status.
Overall, patients in the intervention group had significantly greater
improvement in levels of HbA1c (–2.1%) compared with control
patients (–1.2%) (adjusted difference, –1.0%; 95% confidence interval
[CI], –1.5% to –0.4%; P = .001).
Among patients with higher literacy (Figure 2B), a small, nonsignificant difference between groups in improvement
in levels of HbA1c was observed (adjusted difference, –0.5%;
95% CI, –1.4% to 0.3%; P = .21).
However, among patients with low literacy (Figure
2A), patients in the intervention group had more improvement in
HbA1c levels than did the control patients (adjusted difference,
–1.4%; 95% CI, –2.3% to –0.6%; P<.001).
Similarly, patients in the intervention group were more likely than
were those in the control group to obtain goal HbA1c levels (≤7.0%)
at 12 months (adjusted OR, 1.9; 95% CI, 1.0 to 3.8; P = .05)
(Table 2). Again, there was no significant
treatment effect for patients with higher literacy (adjusted OR, 1.0; 95%
CI, 0.4 to 2.5; P = .98). However, among
patients with low literacy, intervention patients were significantly more
likely to obtain goal HbA1c levels than were the control patients
(adjusted OR, 4.6; 95% CI, 1.3 to 17.2; P = .02).
The P value for interaction between the ORs for the
patients with low and higher literacy was significant (P = .01), confirming that literacy is an effect modifier
for reaching goal levels of HbA1c. No other covariates showed significant
For SBP, overall, intervention patients improved more than control patients
(adjusted difference, –7.6 mm Hg; 95% CI, –13.0 to –2.2; P = .006). However, differences were comparable
for patients with low and higher literacy levels (Figure 2C and 2D). Similarly, patient literacy status did not modify
the effect of the intervention on obtaining goal SBP at 12 months (adjusted
OR for patients with low literacy, 1.5; 95% CI, 0.5 to 4.6; P = .44; adjusted OR for patients with higher literacy, 1.1;
95% CI, 0.5 to 2.3; P = .89).
Among intervention patients, there were no differences by literacy status
(higher vs low literacy) for frequency of completed actions (30 vs 31, P = .64), time spent with patients (364 vs 392
minutes, P = .28), or number of medications
added (3.6 vs 3.6, P = .81) or titrated
(5.2 vs 5.3, P = .53).
We found that a comprehensive diabetes disease management program benefited
patients with low literacy to a greater degree than it did patients with higher
literacy. Literacy appears to be an important factor for determining who benefited
from this program, even after adjusting for race, income, and clinical status.
Our program, with frequent one-to-one patient contact and interventions oriented
toward patients with low literacy, may have helped such patients overcome
barriers and fully participate in their diabetes management.
Patients with diabetes and low literacy have poor knowledge of their
may have difficulties learning the advanced self-care skills needed to improve
glycemic control, particularly in our fast-paced and complex health care system.
We believe that the success of our program was at least in part due to our
using effective strategies for communicating with patients with low literacy,
eg, focusing on selected critical behaviors, decreasing the complexity of
information, using concrete examples, limiting the number of topics covered
in one session, avoiding jargon, and using “teach back” to ensure
comprehension. Previous studies suggest that these strategies may improve
patient self-care and outcomes,4,17,21,30-32 but
to our knowledge ours is the first study to examine differences in effectiveness
of diabetes disease management by literacy levels in a randomized trial. Interestingly,
we did not observe differential improvement in blood pressure by literacy
level. Compared with glucose control, improving blood pressure may be more
dependent on providers’ actions than on strong patient self-care skills.
This was a small, single-site study, which may limit generalizability.
Our study was not powered to detect differences when stratified by literacy
status. The comprehensive nature of our intervention makes it difficult to
discern which aspects of our intervention were the most beneficial. One possible
explanation for the greater benefit for patients with lower literacy in our
program would be that these patients received more time and attention. However,
our analysis of process measures suggests otherwise—patients with low
and higher literacy had similar amounts of contact (approximately 30 min/patient
per month). Furthermore, by enrolling only patients with elevated levels of
HbA1c, it is possible that patients with higher and low literacy
have poor glycemic control for differing reasons that we did not measure.
For example, perhaps patients with higher literacy were more likely to have
poor control because of nonadherence. This could contribute to patients with
higher literacy being less responsive to our intervention for lowering levels
of HbA1c. Arguing against this possibility is the fact that patients
with higher literacy still had significant improvements in blood pressure.
Although many diabetes disease management programs have been able to
reduce levels of HbA1c by 1 to 2 percentage points, programs that
focus on socially disadvantaged populations have often been less successful.33-36 It
is possible that this is partly because these programs did not directly address
the problem of low literacy. Our study suggests that literacy is an important
factor for influencing who will benefit from a diabetes management intervention.
Our disease management program, sensitive to literacy, was able to improve
outcomes for patients with both low and higher literacy—and particularly
for patients with low literacy. Future studies will need to examine the optimal
role of disease management for improving outcomes and addressing disparities
for patients with higher and low literacy.
Corresponding Author: Russell L. Rothman,
MD, MPP, Center for Health Services Research, Vanderbilt University Medical
Center, Suite 6000, Medical Center East, Nashville, TN 37232-8300 (firstname.lastname@example.org).
Author Contributions: Dr Rothman had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analyses.
Study concept and design: Rothman, DeWalt,
Malone, Bryant, Pignone.
Acquisition of data: Rothman, DeWalt, Malone, Bryant, Crigler, Pignone.
Analysis and interpretation of data:
Rothman, DeWalt, Shintani, Weinberger, Pignone.
Drafting of the manuscript:
Rothman, DeWalt, Shintani, Weinberger.
Critical revision of the manuscript for important
intellectual content: DeWalt, Malone, Bryant, Crigler, Weinberger,
Statistical analysis: Rothman, DeWalt, Shintani, Weinberger.
Obtained funding: Malone, Bryant, Pignone.
Administrative, technical, or material support:
DeWalt, Malone, Bryant, Crigler.
Study supervision: Rothman, Weinberger, Pignone.
Funding/Support: This study was completed with
support from the Robert Wood Johnson Clinical Scholars Program, the University
of North Carolina (UNC) Program on Health Outcomes, the UNC Division of General
Internal Medicine, the Vanderbilt Center for Health Services Research, and
the Vanderbilt Diabetes Research and Training Center.
Role of the Sponsors: The funding sources had
no role in the design and conduct of the study; the collection, management,
analysis, and interpretation of the data; or the preparation, review, or approval
of the manuscript.