Content uploaded by Jessica S Ancker
Author content
All content in this area was uploaded by Jessica S Ancker on Jul 10, 2018
Content may be subject to copyright.
JAMIA
The Practice of Informatics
Review Paper 䡲
Rethinking Health Numeracy: A Multidisciplinary
Literature Review
JESSICA S. ANCKER, MPH, DAVID KAUFMAN,PHD
Abstract The purpose of this review is to organize various published conceptions of health numeracy and
to discuss how health numeracy contributes to the productive use of quantitative information for health. We
define health numeracy as the individual-level skills needed to understand and use quantitative health
information, including basic computation skills, ability to use information in documents and non-text formats such
as graphs, and ability to communicate orally. We also identify two other factors affecting whether a consumer can
use quantitative health information: design of documents and other information artifacts, and health-care
providers’ communication skills. We draw upon the distributed cognition perspective to argue that essential
ingredients for the productive use of quantitative health information include not only health numeracy but also
good provider communication skills, as well as documents and devices that are designed to enhance
comprehension and cognition.
䡲J Am Med Inform Assoc. 2007;14:713–721. DOI 10.1197/jamia.M2464.
Introduction
Much of the information provided to patients in written and
electronic health communication is quantitative—informa-
tion such as medication schedules, nutrition information,
laboratory values, and risks and benefits of therapies. A
growing literature attests to an awareness that many patients
get lost in numbers, unable to fully comprehend or use this
information. In this literature review, we examine the construct
of health numeracy as studied in several disciplines, including
consumer informatics and telemedicine,
1,2
medical decision-
making,
3,4
psychology,
5–7
health communication,
8,9
health
literacy,
10,11
adult literacy,
12
and adult education.
13,14
We
also examine how health numeracy contributes to the ap-
propriate use of quantitative information in health.
Background
Just as health literacy has come to describe the use of literacy
skills and health knowledge in health situations,
15–17
health
numeracy is emerging as an important concept describing
the use of quantitative skills in health contexts.
18
We adopt
the definition provided in a recent short review: “the degree
to which individuals have the capacity to access, process,
interpret, communicate, and act on numerical, quantitative,
graphical, biostatistical, and probabilistic health information
needed to make effective health decisions.”
18
Numeracy, as
assessed by different measures, has been associated with
outcomes such as poor anticoagulation control among pa-
tients taking anticoagulants
19
and history of hospitalization
in asthma.
20
However, our examination of the literature shows that the
term “numeracy” has been used to describe several different
specific skill sets. Some numeracy researchers have exam-
ined consumers’ ability to manipulate percentages and pro-
portions, while others have focused on ability to read
appointment slips, communicate orally about medication
schedules, or use tables and graphs. Furthermore, some
studies focus on individual skills, while others focus primar-
ily on effects of changing the numerical representation (for
example, the design of a graph or the format of numbers). In
addition, some numeracy-related articles discuss the skills of
the health-care provider, not of the patient.
In this review, we organize these published conceptions of
health numeracy and introduce the broader outcomes-ori-
ented concept of productive use of quantitative health informa-
tion, i.e., the effective use of quantitative information to
guide health behavior and make health decisions. Produc-
tive information use (for example, a patient’s successful
completion of a medication regimen) depends only in part
Affiliations of the authors: Department of Biomedical Informatics
(JSA, DK), Columbia University College of Physicians and Sur-
geons, New York, NY; New York State Psychiatric Institute (DK),
New York, NY.
Jessica Ancker is supported by Robert Wood Johnson/National
Library of Medicine predoctoral training grant 5T15-LM007079-15.
This work was partially supported by the IDEATel project and a
cooperative agreement from the Centers for Medicare and Medicaid
Services (95-C-90998). The authors would like to thank Suzanne
Bakken, RN, DNSc, Sara Czaja, PhD, Rita Kukafka, DrPH, MA, and
Harold P. Lehmann, MD, PhD, for valuable comments on earlier
versions of this manuscript.
Correspondence: Jessica S. Ancker, MPH, Department of Biomedi-
cal Informatics, Vanderbilt Clinic Room 534, 622 W. 168th Street,
Columbia University College of Physicians and Surgeons, New
York, NY 10032; e-mail: ⬍jsa2002@columbia.edu⬎.
Received for review: 04/02/07; accepted for publication: 08/07/07.
Journal of the American Medical Informatics Association Volume 14 Number 6 Nov / Dec 2007 713
on health numeracy, the individual-level skills to obtain,
interpret, and process quantitative information as defined
above. It also depends upon the ability of the expert or the
information artifact to provide appropriate and cognitively
manageable information. Even consumers with advanced
math skills may perform poorly when trying to use poorly
explained information; conversely, good design of informa-
tion artifacts can compensate for weak individual-level skills.
Thus, the productive use of the information can be consid-
ered a result of the entire system of health communication,
not solely of the individual patient’s skills.
We draw upon the distributed cognition approach, in which
cognition is viewed not as a solitary endeavor but as a process
of coordinating distributed internal representations (i.e.,
knowledge) and external representations (e.g., visual displays
or patient education materials).
21–26
In contrast to theories of
individual cognition, distributed cognition emphasizes both
the social nature of cognition (such as doctor-patient com-
munication) and the mediating effects of technology or other
artifacts (such as how written instructions or website design
guide completion of a task). Distributed cognition suggests
that productive health information use results from inter-
play between the quantitative competencies of the patient
(health numeracy), the properties of the artifacts that medi-
ate health cognition (information design), and the commu-
nication skills of the health-care provider.
Search Methods
We searched the MEDLINE, CINAHL, and PsycINFO data-
bases using the keywords “health literacy” and “numeracy,”
and drew upon a previously published systematic review of
risk graphics.
27
We then used a “pearl-growing” strategy to
seek additional publications through reference lists. We
focused on research and clinical assessment instruments of
quantitative skills among health consumers; studies of pa-
tient or consumer comprehension and decision-making; and
major literacy surveys and researchers in adult literacy.
Review
As a first step to organizing this literature, we examined the
range of meanings of “numeracy” by categorizing the un-
derlying quantitative skills and characteristics captured in
each study. We identified three categories of individual-
level skills that have been studied among consumers: basic
quantitative skills;ability to use information artifacts (for exam-
ple, navigating documents); and ability to communicate orally
about quantitative health information. These three con-
structs can be considered components of “health numeracy.”
Many of the numeracy-related studies we found, however,
focused not on individual patients’ competencies but on the
format of the information presented to them. Thus, we created
a fourth category for this literature review: information design,
the systematic design of information (both symbolic repre-
sentations of information and information channels) to im-
prove comprehension and cognition.
In addition, some of the numeracy-related articles discussed
or focused on the contribution made by the health-care
provider or the information provider. Constructs included:
oral communication skills for health-care providers;basic quanti-
tative skills among health-care providers; and providers’ ability to
use information artifacts. As with patients, providers’ ability
to use information artifacts is inextricably linked to informa-
tion design.
In this review, we focus primarily on the factors of greatest
interest to consumer informatics: patients’ quantitative skills,
patients’ ability to use information artifacts, and information
design for patients. However, we believe that productive use of
quantitative information for health requires a health commu-
nication system that considers all eight of these factors (Figure
1,Table 1).
Patients’ Quantitative Skills
Much of the research on patients and quantitative informa-
tion has addressed individual-level competencies. We group
the individual-level competencies into three categories: 1)
basic computation, 2) estimation, and 3) statistical literacy.
Basic Computation
Basic computation encompasses number recognition and
comparisons, arithmetic, and the use of simple formulas.
The TOFHLA
10
screening test for comprehension of written
health information assesses these skills with questions such
as computing times for taking a pill, comparing a blood
sugar level to a standard to decide whether it is abnormal,
and interpreting an appointment slip. In the validation
sample, of whom 31% had education beyond high school,
only 25% answered all four numeracy questions correctly.
28
A four-item instrument validated among patients with
asthma includes asthma-specific computations of medica-
tion dosages and percentage of peak flow.
19
A broader view
of computational numeracy is provided by the national and
international adult literacy surveys.
12,29,30
These surveys
Figure 1. The successful use of quantitative health infor-
mation requires basic quantitative skills on the part of the
patient, but it also requires interactions with the person
providing the information (who may be a doctor, a nurse, a
counselor, or a friend or family member), and information
artifacts (such as a written document, a website, or a medical
device). Both the patient and the information provider must
be able to manipulate and interpret quantitative informa-
tion, as well as communicate about it. Both parties also need
skills for interacting with information artifacts, such as
document literacy and graphical literacy. Individual quan-
titative skills and artifact-interaction skills are both strongly
affected by the quality of information design; good design
can ease cognitive burden and improve comprehension.
714 ANCKER AND KAUFMAN, Health Numeracy
include a variety of quantitative problems, rated simple if
they are familiar and well-defined (e.g., adding numbers on
a bank deposit slip), or complex if they require using
abstract information to solve multi-step problems. An exam-
ple is a problem with cryptic price labels (“rich chnky pnt bt”
for “chunky peanut butter”) in dollars per pound; respon-
dents must calculate price per ounce.
12
One type of computational skill that has been widely stud-
ied in health contexts is the ability to manipulate percent-
ages and probabilities, because risks are often represented as
percentages (e.g., 15%), proportions (0.15), or frequencies (15
in 100). Many patients lack basic probability skills, according
to results from several studies using related screening mea-
sures.
3–5,19,31,32
These tools (ranging in length from 3 to 18
items) assess the ability to state the numerical probability of
heads in a coin toss and the ability to convert between
percentages and integers. In validation on a highly educated
sample, only 32% got all questions correct, and 16%-20%
were unable to answer questions such as “Which represents
the larger risk: 1%, 5%, or 10%?”
31
An understanding of
probabilities is required for the standard gamble method of
utility elicitation, in which subjects are asked to choose
between a health state and a specific probability of death.
33
Logical inconsistencies are more common with standard
gambles than with visual analog scales, and lower educa-
tional level is associated with errors with standard gam-
bles.
33
Subjects also tend to rate standard gambles as more
difficult to use than visual analog scales.
33
In two studies,
poor probability reasoning was associated with standard
gamble inconsistencies.
34,35
An entire body of work discusses the relationship between
the analytic use of quantitative probability information and
the affective feeling of risk.
5,7,36
In brief, a variety of circum-
stances including stress and time pressures can cause people
to base decisions on rapid and automatic affective responses
rather than the more effortful analytic assessment of quan-
titative information.
Estimation
Adult education theorists point out that people perform
precise calculations only in certain situations, such as com-
pleting financial forms, which have been termed generative
situations.
13
However, people solve many real-world prob-
lems not by calculating but by estimating or using number
sense (a basic feel for magnitudes and operations).
37
Apply-
ing an estimation heuristic is generally quicker and simpler
than calculating and is often sufficient for decision situations
(e.g., choosing groceries while meeting a budget).
13
Esti-
mates may also help in interpretive situations
13
such as
reading a news article, as well as in judging the probable
correctness of a calculation. As many health situations
involve interpreting quantitative information or making
decisions, these examples from the adult education litera-
ture suggest that the role of estimation in health contexts
warrants further study.
Statistical Literacy
Statistical literacy is an understanding of concepts such as
chance and uncertainty,
31
sampling variability, margins of
error, and randomization in clinical trials,
18
and the ability
to use such concepts to evaluate scientific information.
38
It is
thus strongly determined by domain knowledge as well as
by procedural skills. The measures of probability reasoning
discussed above, including the medical data interpretation
test, capture elements of statistical literacy such as the
concepts of uncertainty and chance.
3–5,19,31,32
Statistical un-
derstanding can help patients in specific decision situations
such as granting informed consent to be in a research study.
It can also help consumers understand epidemiological
Table 1 yFactors Contributing to a Patient’s Ability to Use Quantitative Information for Health
Factor Definition Examples
Patients’ quantitative skills basic computational skills (such as
addition, multiplication, and use of
simple formulas), estimation, and
statistical literacy
computing calorie content; comparing computation
to estimate to determine whether it is correct;
understanding concept of randomization in a
clinical trial
Patients’ ability to use information
artifacts
ability to navigate documents, interpret
graphs, and translate between
different representations of the same
information
obtaining nutrient information from a nutrition
label; comparing personal health data as
displayed on different meters or devices
Patients’ oral communication skills ability to speak clearly about quantities
and understand spoken information
reporting a previous medication regimen
accurately to a new physician
Information design for patients arrangement of information media and
symbols to support comprehension
and cognition
designing a patient interface for an electronic
health record that provides graphics to illustrate
numerical information
Providers’ oral communication skills ability to communicate quantitative
concepts clearly to the patient
explaining a new medication regimen to a patient
in an understandable fashion
Providers’ quantitative skills basic computational skills, estimation,
and statistical literacy
converting between units of measure;
understanding the positive predictive power of a
diagnostic test
Providers’ ability to use information
artifacts
ability to navigate documents, interpret
graphs, translate between
representations of the same
information
interpreting a graph of patient lab values over
time; applying the numerical output of a
decision support system to an individual case
Information design for providers ability of a system or document to
support the provider’s cognition
designing a provider interface that provides
automated conversions between units of
measure
Journal of the American Medical Informatics Association Volume 14 Number 6 Nov / Dec 2007 715
information and how it is applicable to personal health or to
public policy. Statistical literacy might also promote trust in the
information. Lower educational level has been linked to poor
understanding of the concept of scientific uncertainty and
distrust of information presented as uncertain.
39,40
People
with low statistical literacy may be vulnerable to anti-
scientific messages such as exaggerated warnings about
vaccine risks. A comprehensive assessment of statistical
literacy and its relevance to health decisions would be
valuable.
Patients’ Ability to Use Information Artifacts
Before patients can compute, estimate, or interpret data,
they must acquire it, often from documents, scales, or other
information artifacts. Three interrelated skills (representa-
tional fluency, document literacy, and graphical literacy)
have been described as necessary for using information
artifacts.
Representational Fluency
Representational fluency is the ability to translate between
and recognize the identity of different representations of the
same quantity.
1,2
For example, the quantity “one-half” can
be represented as a fraction, a decimal, or a picture (Figure
2); some representations are more useful for problem-solv-
ing than others. Translating between representations is an
important mathematical problem-solving skill
41
and cogni-
tive milestone in children.
42,43
Mappings between represen-
tations and concepts are not inherent but are established
through learned cultural conventions.
43
Medicine, like other
complex domains, introduces new representations that can
challenge newcomers, such as patients or medical students.
Representational fluency may underlie several other com-
putational skills. Subjects with poor probability reasoning
were likely to make different decisions when numbers were
presented in different formats (10 in 100 versus 10%),
apparently because they were less able to translate between
representations of the same quantity.
5
Poor probability
reasoning was also associated with greater susceptibility to
framing effects, probably for the same reason.
5
Representational fluency is particularly relevant to the abil-
ity to use personal health care technologies and measure-
ment tools such as meters, which represent information in
unfamiliar formats. In a diabetes telehealth program for
elderly patients,
1,2
lack of representational fluency made it
difficult for some participants to review their glucose and
blood pressure values when they switched technologies.
These patients, who could readily read their blood pressure
from a meter, were not able to recognize the tabular repre-
sentations of the values. Some seemed unfamiliar with the
conventions of columns and rows (see ‘document literacy,’
below), and they may have been unable to equate blood
pressure shown as 120/90 with the same value shown as
(Figure 3). Some people who had kept glucose logs on paper
for years could not use the tables.
1,2
Similar issues are likely
to arise with electronic personal health records.
44
Document Literacy
Quantitative information is often provided in the form of a
complex print or electronic document that combines text
with lists, forms, or tables. Before applying any procedural
skills, such as computation, the patient must be able to
navigate and interpret these documents. Among adult liter-
acy researchers, the ability to use information embedded in
complex text and non-text formats is called document
literacy.
29
In the National Adult Literacy Survey (NALS),
23% of the public received the lowest score on document
literacy, indicating they would have difficulty performing
tasks such as completing forms or using information in
tables.
12
In the Health And Literacy Scale (HALS) analysis,
health literacy was worst among those with poor document
literacy.
16
Document literacy includes the abilities to com-
plete multi-step calculation problems and to infer the right
mathematical operation when it is not provided. One HALS
question
16
showed a medication label; although calculating
the correct dose was arithmetically simple, the task also
required deciding on the algorithm (not provided), locating
numbers in the table, and separating relevant from irrele-
vant information. Eisemon et al. provide an example of a
medication that had to be mixed with water; the instructions
explained how to calculate the dose, but a different calcula-
tion (not provided) was needed to determine how many
packets of powder to purchase.
45
In a study of nutrition label
interpretation, only about 28% of the errors were inaccurate
calculations; the remaining errors stemmed from an inability
to exclude extraneous or complex information, or wrong
application of serving size/servings per container informa-
tion.
46
Graphical Literacy
The ability to interpret quantitative graphics helps people
use educational materials, news reports, and electronic sys-
tems. Graphics are valuable in illustrating consumer infor-
matics applications,
47,48
and visual analog scales may be
more intuitive than standard gambles in utility elicitation.
33
However, many people have difficulty interpreting graphics
that are not familiar.
27,49–53
For example, in a sample drawn
from a jury pool, 23% could not use a survival curve to
determine the number of survivors at specific times, and
45% could not calculate the difference in survival between
two times.
51
Although these examples show that graphical
literacy has been studied in a variety of settings, it is not
assessed by any health literacy instrument.
Information Design for Patients
The ability to apply basic computational skills is strongly
influenced by how the information is represented in a
questionnaire, document, graphic, or medical device. In-
cluding information design in a discussion of numeracy
shifts the focus, to some extent, from the patient to the
representations in the artifact. Even readers with strong
literacy skills can have difficulty using quantitative informa-
tion presented in complex formats such as nutrition labels,
and those with poorer literacy have more trouble.
46
For
nutrition labels, Rothman et al. recommend information
design improvements such as presenting nutrient totals for
the entire package, removing extraneous information, and
Figure 2. The quantity 1/2 may be represented in a
variety of logically equivalent but cognitively different
ways. Reprinted from
43
p. 101. With permission.
716 ANCKER AND KAUFMAN, Health Numeracy
adding visual cues (such as lines on an ice-cream container)
to indicate serving sizes.
46
The framing effect is another example of a representation
effect;
54
framing effects are nearly universal but they are
strongest among people with poor probability reasoning
skills.
5
Numerical format effects are also strong: in a study of
patients in a clinic waiting room, only 73% could tell that a
disease affecting “2.6 per 1000 women” was less common
than one affecting “8.9 per 1000 women,” and when the
same comparison was presented as “1 in 384 women” and “1
in 112 women,” only 56% answered correctly.
55
Evidence
conflicts about whether percentages such as 46% or frequen-
cies such as 46 in 100 are easier to understand.
56,57
Another
representation effect is the relative comparison effect: rela-
tive differences without absolute risks for context inflates the
apparent magnitude of the effect.
58–60
That is, an increase
from 1% to 2% seems larger when it is described as a 100%
(relative) increase than when it is described as a 1% (abso-
lute) increase. Adding baseline risk to either absolute risk
reductions or relative ones strongly improves the accuracy
of interpretation of risk reduction information.
61
Poor representational fluency can be addressed by provid-
ing quantitative information in multiple formats (that is, in
proportions as well as frequencies, or in text as well as
graphics). Supplementing text with graphics can reduce the
influence of less-relevant textual information to promote
better decision-making
62
and improve accuracy in compar-
ing probabilities.
56
By contrast, when people are asked to
think of themselves in unfamiliar health situations, supple-
menting dry textual and numeric information with anec-
dotes may improve decision-making by helping readers
imagine the situation more clearly.
63
Providing multiple
formats may also help reach a broader audience; highly
numerate lay people preferred risk information in numbers
instead of in words alone, while those with poor numeracy
skills did not.
64
There is some evidence that presenting
information in both gain and loss frames may reduce framing
effects.
50
Completing calculations for the consumer reduces the
cognitive effort required to process probability information
and improves accuracy in making risk tradeoffs.
56
For exam-
ple, instead of merely stating that a drug will triple the risk of
some bad outcome, the communication can present the
baseline risk of 4% and the end risk of 12%, and instead of
expecting consumers to add the risks of two separate
outcomes to decide about the overall usefulness of the drug,
the communication can provide the total summed risk for
both the outcomes.
56
Designing graphs to support patient cognition is challeng-
ing, in part because users often prefer design features
such as visual simplicity and familiarity that are not
necessarily associated with accurate judgments.
27
Patients
prefer graphics supplemented with text
65
probably be-
cause explanations help them draw conclusions. Risk
graphs such as bar charts can show the numerator only
(e.g., a bar 10 units high depicting a risk of 10%), or they
can display the part-to-whole relationship between the
Figure 3. Systolic and diastolic blood pressure displayed in a computer-generated table as part of the IDEATel
1,2
telehealth
program, and (inset) on the blood pressure meter. Some elderly participants who had no difficulty reading the values on the
meter were unable to understand the same information displayed in the table.
Journal of the American Medical Informatics Association Volume 14 Number 6 Nov / Dec 2007 717
numerator and denominator of the risk ratio (e.g., by
showing the risk of 10% as one-tenth of a larger bar
depicting 100%). Graphs that depict numerators alone
emphasize the risks and are more likely to promote
risk-related behavior changes, while graphs that depict
numerators in context of denominators promote accurate
judgments.
6,66,67
One informatics application that works in part by easing
document literacy burdens is tailoring information so it
contains only personally relevant facts.
68
Well-designed
electronic and print decision aids can also reduce document
literacy demands by presenting personally tailored informa-
tion in manageable chunks. Their measured benefits include
increased knowledge, more realistic expectations, and better
agreement between choices and expressed values,
69
all of
which might reflect better understanding of the quantitative
risk information.
Other applications that may help are those designed to
promote active information-processing. For example, writ-
ing answers to reflective questions and drawing graphs
improves comprehension of and satisfaction with risk infor-
mation.
70
It is possible that electronic health and medical
information systems could improve comprehension by in-
viting the patient to engage in active information processing
by completing self-tests or interacting with the system in
other ways. In an ongoing project, we are assessing a risk
communication module that allows users not only to input
personal risk characteristics and receive a graphic illustra-
tion of their risk, but also play with the graphic in a
game-like interaction; for example, for a disease risk of 10%,
the user can click on stick figure illustrations and see that
10% of them will contract the disease.
71
Technology can also
be designed to facilitate distributed cognition through col-
laborative learning and problem-solving: in the IDEATel
project,
1,2
the nurse and the patient could simultaneously
view tables and charts of the patient’s data during a video-
conference, with discussions to facilitate patients’ under-
standing of key concepts such as HbA1c level, an important
determinant of health in diabetes patients.
Oral Communication Skills among Providers
and Patients
The focus of this article is on patients’ quantitative skills,
their interaction with information artifacts, and the design of
these artifacts. However, it is important to acknowledge that
consumers and patients obtain much of their information,
particularly their personal clinical information, through oral
discussions. They listen to instructions, quantify their expe-
riences, express utilities and preferences, and ask questions.
For example, physicians rely upon patient report of fre-
quency, duration, and severity of symptoms, and, often, of
their medication regimens.
9
When cognition is seen as
distributed among social agents,
25
it is clear that quantitative
problems are often solved during interactions, as when a
physical therapist and a patient discuss scheduling exercise
sessions, or a married couple complete a health insurance
form together. Content analysis of audiotaped consultations
shows that physicians vary in oral communication style, and
may omit numerical information or fail to assess patients’
understanding of it.
72,73
Risk communication training has
been developed for physicians and public health profession-
als,
74
but not assessments of professionals’ ability to explain
numerical information in general. Patients cannot use quan-
titative information well if their providers do not explain it
clearly or are not attuned to the kinds of communication
most likely to be understood by a particular patient.
Quantitative Skills of Health Care Providers
The individual providing the information (physician, nurse,
public health worker, journalist, or other information pro-
vider) must also have skills in basic computation, estima-
tion, and statistical literacy. Brief reports have indicated that
many medical students
75
and health care professionals
32
perform poorly on numeracy assessments focusing on risk
and probability. Such skills are obviously important for
basic care issues such as calculating doses, converting units
of measure, and interpreting clinical data, but they are also
important for communicating with patients. An example
illustrates the role of professional numeracy in health com-
munication: in one report, physicians and social workers
giving HIV counseling at German clinics became confused
when asked about false positives, false negatives, and pre-
dictive power.
76
Patients are unlikely to use information
well if their providers do not understand it themselves.
Providers’ Ability to Use Information Artifacts
To provide patients with useful quantitative information,
the health care provider must be able to obtain it from
journal articles, information systems, medical devices, or
other information artifacts. An important focus of the evi-
dence-based medicine movement has been to train physi-
cians to use journal articles, appraise them with a critical
eye, and especially to extract and interpret relevant statisti-
cal and other numerical information.
77
Lack of training in
the use of statistical information has been identified as an
important barrier to use of decision-support systems: a
web-based tutorial improved adoption of a decision-support
system for acute cardiac ischemia by teaching clinicians how
to combine the population-based risk score output by the
system with patient-specific data.
78
Information Design for Health Care Providers
Information artifacts must be designed to support health
care providers’ cognition about numbers, particularly to
enable them to discuss information clearly with patients and
the public. A full discussion is beyond the scope of this
paper, but some examples demonstrate the importance of
this design issue. Adverse drug events associated with
incorrect doses and administration rates have been associ-
ated with poor design of human-computer interfaces in
patient-controlled analgesia devices.
79
Physicians are also
subject to judgment biases from framing
54
and the format of
numbers and graphs. In one study, physicians rated drug
effectiveness on an 11-point scale. Perceived effects were
larger when described as relative differences (e.g., a relative
decrease of 36%) than when described as absolute differ-
ences (e.g., an absolute decrease of 1.4%, from 3.9% to
2.5%).
59
In a second study, physicians were more likely to
answer computational problems about positive predictive
power correctly when probabilities were formatted as “nat-
ural frequencies” (ratios with the same denominator, such as
10 in 100 and 20 in 100) than as percentages.
57
In a simulated
patient-safety review experiment, Elting et al. tested the use
of different representations of data from a hypothetical
clinical trial showing a strong association between one of the
treatments and an adverse outcome.
80
One representation
718 ANCKER AND KAUFMAN, Health Numeracy
was a stacked bar chart, two were tables, and the last was a
group of squares, with each square representing an individ-
ual subject, colored to indicate his or her outcome. Physi-
cians were more likely to order a halt to the trial when
viewing the tables or the groups of colored squares than
when viewing the stacked bar charts.
80
Such information
design effects influence providers’ competency with quan-
titative information, and, indirectly, their communication of
quantitative information to patients.
Discussion
Consumers’ ability to use quantitative information to guide
their health behavior and health decisions requires more
than the ability to perform computations. Consider the
example of a woman who seeks to lose weight. She may
obtain information by reading articles in popular magazines,
doing Internet research, talking with friends or a support
group, or discussing her goals with a nurse. If she decides to
begin walking for exercise, a variety of external resources for
information seeking and processing are available: she can
use her car odometer to measure distances, look up calories
per mile walked in a table, and use a calculator to calculate
total calories expended. If she decides to adopt a healthier
diet, she can seek nutrition information on food labels and
use it for calculations, estimates, or comparison with a
threshold (she can, for example, decide to avoid products
with more than a certain percentage of calories from fat). She
can use a measuring cup to measure servings of breakfast
cereal, or learn to estimate the amount. She may also weigh
herself on a bathroom scale, use a diary to track weight or
exercise accomplishments over time, and share information
with friends or supporters.
As a second example, consider the case of a man seeking
treatment for prostate cancer. Initial information is likely to
be explained orally by a physician, who may or may not be
skillful at explaining laboratory and biopsy results and each
treatment option’s chance of success and risk of adverse
effects. The patient may work with his wife to use a decision
aid, find additional information online, or e-mail the physi-
cian. The couple may both attend the next appointment,
perhaps bringing simple memory aids such as a printout or
a list of handwritten questions, and seek help interpreting
information through additional conversations with the phy-
sician or nurse. The final decision is likely to be made in
collaboration with the doctor.
These examples suggest that the use of quantitative infor-
mation for health is often highly situated, distributed across
internal and external resources, and collaborative, as sug-
gested by theories of distributed cognition. Patients and health
consumers may obtain quantitative information orally as well
as through printed and electronic material, and from devices
such as bathroom scales. This information may be used to
make calculations, decisions, or estimates. Consumers extend
their memories by putting health information in writing or
computers. They reduce cognitive effort and augment per-
sonal skills by offloading computations onto external re-
sources such as calculators or paper. They frequently rely on
social interactions to help derive meaning from quantitative
information and make decisions from it. The representation
of the information strongly affects individual cognition;
better external resources can compensate for weak individ-
ual-level skill sets, while poor external resources require
more skilled individual cognition. As characterized by a
distributed cognition framework, productive use of quanti-
tative information in health requires individual skills (health
numeracy) but is also mediated by social interaction and
artifacts.
Conclusion
In this review, we organize research on health numeracy,
defined as the individual-level skills to obtain, interpret, and
process quantitative information for health behavior and
decisions. We find that the term health numeracy has been
used to describe a number of different types of individual-
level competencies, including basic arithmetic operations on
health-related data, facility with probabilities, and statistical
literacy. Other individual-level skills sometimes discussed
in connection with health numeracy include oral fluency
and ability to navigate documents, tables, and graphs, but
no available instrument for health numeracy assesses all
these component skills. We also find that the ability to use
quantitative information is strongly influenced by two other
factors: design of information artifacts, and providers’ com-
munication skills. Although tools are available for assessing
the ‘readability’ level of documents,
15
there are no similar
tools that characterize their quantitative demands. Similarly,
instruments are available to assess aspects of numeracy in
patients, but no assessments have been developed to evalu-
ate providers’ ability to provide comprehensible informa-
tion about numbers.
In 2000, health literacy was described as the skills needed to
“obtain, process, and understand basic health information
and services for appropriate health decisions.”
81
Subsequent
commentaries pointed out that this definition focused on the
individual patient’s skills without recognizing the comple-
mentary importance of good communication skills by health
experts or the complexity of written information.
17,82
We
similarly argue that discussions of quantitative information
in health should be expanded beyond a focus on the skills of
the patient. Poor use of quantitative information may stem
from poor individual-level skills, but it may also be the
result of mismatch between the patient’s skills and the
provider’s communication skills or the artifact’s information
design (Figure 1). Baker has made a similar argument about
health literacy.
82
This topic is important because the amount of quantitative
information for patients is growing rapidly, largely through
the internet and health information technologies. Although
this information explosion promises to empower many
patients, it also has the potential to exacerbatthe literacy
divide for those who lack numeracy, literacy, or computer
skills. Our framework suggests that the divide can be
narrowed by educating not only patients but also informa-
tion providers. Furthermore, by enhancing the design of
health-related systems and documents, the informatics com-
munity can help improve the fit between task demands and
individual competencies, helping consumers use quantita-
tive information to make genuinely informed decisions
about health.
Journal of the American Medical Informatics Association Volume 14 Number 6 Nov / Dec 2007 719
References y
1. Kaufman DR, Patel VL, Hilliman C, et al. Usability in the real
world: assessing medical information technologies in patients’
homes. J Biomed Inform. 2003;36:45–60.
2. Kaufman DR, Pevzner J, Hilliman C, et al. Redesigning a
telehealth diabetes management program for a digital divide
seniors population. Home Health Care Management & Practice.
2006;18(3):223–34.
3. Schwartz L, Woloshin S, Black W, Welch H. The role of
numeracy in understanding the benefit of screening mammog-
raphy. Ann Intern Med. 1997;127(11):966–72.
4. Schwartz LM, Woloshin S, Welch HG. Can patients interpret
health information? An assessment of the medical data inter-
pretation test. Med Decis Making. 2005;25:290–300.
5. Peters E, Vastfjall D, Slovic P, Mertz CK, Mazzocco K, Dickert
S. Numeracy and decision making. Psychol Sci. 2006;17(5):
407–13.
6. Stone ER, Sieck WR, Bull BE, Yates JF, Parks SC, Rush CJ.
Foreground: background salience: Explaining the effects of
graphical displays on risk avoidance. Organ Behav Hum Decis
Process. 2003;90(1):19–36.
7. Slovic P, Peters E, Finucane M, MacGregor DG. Affect, risk, and
decision making. Health Psychol. 2005;24(4(Suppl.)):S35–S40.
8. Schillinger D, Piette J, Grumbach K, et al. Closing the loop:
Physician communication with diabetic patients who have low
health literacy. Arch Intern Med. 2003;163:83–90.
9. Schillinger D, Wang F, Palacios J, Rodriguez M. Language,
literacy, and communication regarding medication in an anti-
coagulation cliic: A comparison of verbal vs. visual assessment.
J Health Comm. 2006;11:651–64.
10. Nurss J, Parker R, Williams M, Baker D. TOFHLA test of
functional health literacy in adults. Show Camp, NC: Pepper-
corn Books and Press; 2001.
11. Kalichman SC, Benotsch E, Suarez T, Catz S, Miller J, Rompa
D. Health literacy and health-related knowledge among
persons living with HIV/AIDS. Am J Prev Med. 2000;18(4):
325–31.
12. Kirsch IS, Jungeblut A, Jenkins L, Kolstad A. Adult Literacy
in America: A First Look at the Findings of the National
Adult Literacy Survey. U.S. Department of Education Of-
fice of Educational Research and Improvement; 2002. NCES
1993–275.
13. Gal I, (ed). Adult Numeracy Development: Theory, Research,
and Practice. Series on Literacy: Research, Policy and Practice.
Cresskill, NJ: Hampton Press, Inc.; 2000.
14. Coben D. Adult numeracy: review of research and related
literature. London: National Research and Development Centre
for Adult Literacy and Numeracy; 2003.
15. McCray AT. Promoting health literacy. Jamia. 2005;12:152– 63.
16. Rudd R, Kirsch I, Yamamoto K. Literacy and Health in
America. Center for Global Assessment Policy Information
Center Research and Development Educational Testing Ser-
vice; 2004.
17. Nielsen-Bohlman L, Panzer AM, Kindig DA, (eds). Committee
on Health Literacy. Health Literacy: A Prescription to End
Confusion. Washington, DC: Institute of Medicine. The Na-
tional Academies Press; 2004.
18. Golbeck AL, Ahlers-Schmidt, Carolyn R., Paschal AM, Dismuke
SE. A definition and operational framework for health numeracy.
Am J Prev Med. 2005;29(4):375–6.
19. Estrada CA, Martin-Hryniewicz M, Peek BT, Collins C, Byrd JC.
Literacy and numeracy skills and anticoagulation control. Am J
Med Sci. 2004;328(2):88–93.
20. Apter AJ, Cheng J, Small D, et al. Asthma numeracy skill and
health literacy. J Asthma. 2006;43:705–10.
21. Perry M. Distributed Cognition. In: Carroll JM (ed) HCI
Models, Theories, and Frameworks: Toward an Interdiscipli-
nary Science. San Francisco, CA: Morgan Kaufmann;
2003:193–223.
22. Patel VL, Kaufman DR. Cognitive Science and Biomedical
Informatics. In (Cimino JJ, Shortliffe EH, eds.) Biomedical Infor-
matics: Computer Applications in Health Care and Biomedicine
3rd ed. New York: Springer; 2006.
23. Zhang J, Norman DA. Representations in distributed cognitive
tasks. Cognitive Science. 1994;18:87–122.
24. Cohen T, Blatter B, Almeida C, Shortliffe E, Patel V. A cognitive
blueprint of collaboration in context: Distributed cognition in
the psychiatric emergency department. Artif Intell Med. 2006;
37(2):73–83.
25. Patel VL, Arocha JF, Kaufman DR. A Primer on Aspects of
Cognition for Medical Informatics. J Am Med Inform Assoc.
2001;8(4):324–43.
26. Xiao Y. Artifacts and collaborative work in healthcare: method-
ological, theoretical, and technological implications of the tan-
gible. J Biomed Inform. 2005;38(1):26–33.
27. Ancker JS, Senathirajah Y, Kukafka R, Starren JB. Design fea-
tures of graphs in health risk communication: A systematic
review. J Am Med Inform Assoc. 2006;13(6):608–18.
28. Baker D, Williams M, Parker R, Gazmararian J, Nurss J. Devel-
opment of a brief test to measure functional health literacy.
Patient Educ Couns. 1999;38(1):33–42.
29. Kutner M, Greenberg E, Baer J. A first look at the literacy of
America’s adults in the 21st century: National Assessment of
Adult Literacy. Washington, DC: National Center for Education
Statistics; 2005.
30. Anonymous. Literacy in the Information Age: Final Report of
the International Adult Literacy Survey. Paris, France: Organi-
sation for Economic Cooperation and Development and Statis-
tics Canada; 2000.
31. Lipkus IM, Samsa G, Rimer BK. General performance on a
numeracy scale among highly educated samples. Med Decis
Making. 2001;21:37–44.
32. Estrada C, Barnes V, Collins C, Byrd JC. Health literacy and
numeracy [letter]. JAMA. 1999;282(6):527.
33. Lenert L, Sturley A. Acceptability of visual analog scale, time
trade-off and standard gamble rating methods in patients and
the public. J Med Inform Assoc (Symp Suppl). 2001:3948.
34. Woloshin S, Schwartz LM, Moncur M, Gabriel S, Tosteson
ANA. Assessing values for health: Numeracy matters. Med
Decis Making. 2001;(Sept-Oct):382–90.
35. Schwartz SR, McDowell J, Yueh B. Numeracy and the shortcom-
ings of utility assessment in head and neck cancer patients.
Head Neck. 2004(May):401–7.
36. Loewenstein G, Weber EU, Hsee C, Welch N. Risk as feelings.
Psychol Bull. 2001;127(2):267–86.
37. Curry D, Schmitt MJ, Waldron S. A Framework for Adult
Numeracy Standards: The Mathematical Skills and Abilities
Adults Need To Be Equipped for the Future. The Adult Nu-
meracy Practitioners Network System Reform Planning Project;
1996.
38. Zarcadoolas C, Pleasant A, Greer DS. Understanding health
literacy: an expanded model. Health Promo Intern. 2005;20(2):
195–203.
39. Johnson BB, Slovic P. Presenting uncertainty in health risk
assessment: Initial studies of its effects on risk perception and
trust. Risk Anal. 1995;15(4):485–94.
40. Schapira MM, Nattinger AB, McHorney CA. Frequency or
probability? A qualitative study of risk communication formats
used in health care. Med Decis Making. 2001;21:459–67.
41. Gagatsis A, Shiakalli M. Ability to translate from one represen-
tation of the concept of function to another and mathematical
problem solving. Edu Psychol. 2004;24(5):646–57.
42. Kilpatrick J. Understanding mathematical literacy: The contri-
bution of research. Edu Studies Math. 2001;47:101–16.
720 ANCKER AND KAUFMAN, Health Numeracy
43. Kilpatrick J, Swafford J, Findell B, for the Mathematics
Learning Study Committee (eds.) Adding it up: Helping
children learn mathematics. Washington, DC: National Acad-
emies Press; 2001.
44. Baorto D, Cimino J. An “infobutton” for enabling patients to
interpret on-line Pap smear reports. Proc AMIA Annu Symp.
2000:47–50.
45. Eisemon TO, Ratzlaff J, Patel VL. Reading instructions for using
commercial medicines. Ann Am Acad Pol Soc Sci. 1992;520:76–90.
46. Rothman RL, Housam R, Weiss H, et al. Patient understanding
of food labels: The role of literacy and numeracy. Am J Prev
Med. 2006;31(5):391–8.
47. Tang P, Newcomb C. Informing patients: a guide for providing
patient health information. J Am Med Inform Assoc. 1998;5(6):
563–70.
48. Lenert LA, Cher DJ. Use of meta-analytic results to facilitate
shared decision making. J Am Med Inform Assoc. 1999;6(5):
412–9.
49. Lipkus IM, Hollands JG. The visual communication of risk.
JNCI Monogr. 1999;25:149–63.
50. Armstrong K, Schwartz JS, Fitzgerald G, Putt M, Ubel PA. Effect
of framing as gain versus loss on understanding and hypothet-
ical treatment choices: survival and mortality curves. Med Decis
Making. 2002;22:76–83.
51. Armstrong K, Fitzgerald G, Schwartz JS, Ubel PA. Using sur-
vival curve comparisons to inform patient decision making: Can
a practice exercise improve understanding? J Gen Intern Med.
2001(16):482–5.
52. Mazur DJ, Hickam DH. Patients’ and physicians’ interpretations
of graphic data displays. Med Decis Making. 1993;13:59–63.
53. Parrott R, Silk M, Dorgan K, Condit C, Harris T. Risk compre-
hension and judgments of statistical evidentiary appeals: when
the picture is not worth a thousand words. Human Comm Res.
2005;31(3):423–52.
54. McNeil B, Pauker S, Sox H, Tversky A. On the elicitation of
preferences for alternative therapies. N Engl J Med. 1982;306:
1259–62.
55. Grimes DA, Snively GR. Patients’ understanding of medical
risks: Implications for genetic counseling. Obstet Gynecol. 1999;
93:910–4.
56. Waters EA, Weinstein ND, Colditz GA, Emmons K. Formats for
improving risk communication in medical tradeoff decisions.
J Health Comm. 2006;11:167–82.
57. Hoffrage U, Lindsey S, Hertwig R, Gigerenzer G. Communicat-
ing statistical information. Science. 2000;290:2261–2.
58. Nuovo J, Melnikow J, Chang D. Reporting number needed to
treat and absolute risk reduction in randomized controlled
trials. JAMA. 2002;287:2813–4.
59. Naylor C, Chen E, Strauss B. Measured enthusiasm: does the
method of reporting trial results alter perceptions of therapeutic
effectiveness? Ann Intern Med. 1992;117:916–21.
60. Epstein RM, Alper BS, Quill TE. Communicating evidence for
participatory decision-making. JAMA. 2006;291:2359–66.
61. Natter HM, Berry DC. Effects of presenting the baseline risk
when communicating absolute and relative risk reductions.
Psychol Health Med. 2006;10(4):326–34.
62. Fagerlin A, Wang C, Ubel PA. Reducing the influence of
anecdotal reasoning on people’s health care decisions: Is a
picture worth a thousand statistics? Med Decis Making. 2005;
25:398–405.
63. Hibbard JH, Peters E. Supporting informed consumer health
care decisions: Data presentation approaches that facilitate the
use of information in choice. Annu Rev Public Health. 2003;24:
413–33.
64. Gurmankin AD, Baron J, Armstrong K. The effect of numerical
statements of risk on trust and comfort with hypothetical
physician risk communication Med Decis Making. 2004;24(3):
265–71.
65. Fortin J, Hirota L, Bond B, O’Connor A, Col N. Identifying
patient preferences for communicating risk estimates: A de-
scriptive pilot study. BMC Medical Informatics and Decision
Making. 2001;1:2.
66. Schirillo JA, Stone ER. The greater ability of graphical versus
numerical displays to increase risk avoidance involves a com-
mon mechanism. Risk Anal. 2005;25(3):555–66.
67. Stone ER, Yates JF, Parker AM. Effects of numerical and
graphical displays on professed risk-taking behavior. J Exper
Psychol: Appl. 1997;3(4):243–56.
68. Kukafka R, Lussier YA, Eng P, Patel VL, JJ C. Web-based
tailoring and its effect on self-efficacy: results from the MI-
HEART randomized controlled trial. Proc AMIA Symp. 2002:
410–4.
69. O’Connor A, Stacey D, Entwistle V, et al. Decision aids for
people facing health treatment or screening decisions. Cochrane
Database of Systematic Reviews. 2007;2.
70. Natter HM, Berry DC. Effects of active information processing
on the understanding of risk information. Appl Cognit Psychol.
2005;19:123–35.
71. Ancker JS, Kukafka R. A combined qualitative method for
testing an interactive risk communication tool. Proc AMIA
Annu Symp 2007;(in press).
72. Siminoff LA, Fetting J, Abeloff M. Doctor-patient communica-
tion about breast cancer adjuvant therapy. J Clin Oncol. 1989;
7(9):1192–200.
73. Leighl N, Gattellari M, Butow P, Brown R, Tattersall MHN.
Discussing adjuvant cancer therapy. J Clin Oncol. 2001;19(6):
1768–78.
74. Edwards A, Elwyn G, Wood F, Atwell C, Prior L, Houston H.
Shared decision making and risk communication in practice: a
qualitative study of GPs’ experiences. Br J Gen Pract. 2005;
55(510):6–13.
75. Sheridan SL, Pignone M. Numeracy and the medical student’s
ability to interpret data. Eff Clin Pract. 2002;5:35–40.
76. Gigerenzer G, Hoffrage U, Ebert A. AIDS counselling for
low-risk clients. AIDS Care. 1998;10(2):197–211.
77. Guyatt G, Rennie D. User’s Guides to the Medical Literature. A
Manual for Evidence-based Clinical Practice. Chicago: AMA
Press; 2002.
78. Lai F, Macmillan J, Daudelin DH. The potential of training to
increase acceptance and use of computerized decision sup-
port systems for medical diagnosis. Hum Factors. 2006;48(1):
95–108.
79. Lin L, Vicente KJ, Doyle DJ. Patient safety, potential ad-
verse drug events, and medical device design: a human
factors engineering approach. J Biomed Inform. 2001;34(4):
274–84.
80. Elting L, Martin C, Cantor S, Rubenstein E. Influence of data
display formats on physician investigators’ decisions to stop
clinical trials. Br Med J. 1999;318:1527–31.
81. U.S.D.H.H.S. U.S. Department of Health and Human Services.
Tracking Healthy People 2010. Washington, DC: U.S. Govern-
ment Printing Office; November 2000.
82. Baker DW. The meaning and measure of health literacy. J Gen
Intern Med. 2006;21:878–83.
Journal of the American Medical Informatics Association Volume 14 Number 6 Nov / Dec 2007 721