Patient Education and Counseling xxx (2011) xxx–xxx

G Model

PEC-4078; No. of Pages 6

The effect of numeracy on the comprehension of information aboutmedicines in users of a patient information website

Peter H. Gardner a,*, Brian McMillan a, David K. Raynor b, Elizabeth Woolf c, Peter Knapp b

a Institute of Psychological Sciences, University of Leeds, Leeds, UKb School of Healthcare, University of Leeds, Leeds, UKc Cancer Research UK, London, UK

A R T I C L E I N F O

Article history:

Received 17 December 2010

Received in revised form 3 May 2011

Accepted 3 May 2011

Keywords:

Risk perception

Side effects

Numeracy

Patient information website

A B S T R A C T

Objective: To investigate the relationship between numeracy and the accuracy of side effect risk

estimation following the presentation of information about a medicine via the Cancer Research UK (CR-

UK) patient information website.

Methods: 591 website users were presented with information in different formats about the risks of side

effects from taking tamoxifen. Participants estimated the risk of each side effect, provided other

subjective ratings about the information and completed a numeracy task.

Results: Regardless of presentation format, numeracy was correlated with the accuracy of three side

effect risk estimates. People with cancer and tamoxifen users showed stronger correlations for all side

effect estimates. In addition, numeracy was positively related to the perceived influence of the

information on the decision to take the medicine and was negatively related to ratings of satisfaction

with the information.

Conclusion: People with a lower numeracy level make larger errors in interpreting medicines side effect

risk information.

Practice implications: Pharmacists, other health professionals and patient information websites should

ensure they provide clear explanations of risk, particularly to people with low numeracy, and assess their

understanding of those explanations. Future research into risk communication should take account of

numeracy level, to investigate the impact of different formats.

� 2011 Elsevier Ireland Ltd. All rights reserved.

Contents lists available at ScienceDirect

Patient Education and Counseling

jo ur n al h o mep ag e: w ww .e lsev ier . co m / loc ate /p ated u co u

1. Introduction

Research into written medicines information for patients (suchas European PILs or Australian CMI) has shown that some formatsfor presenting side effect information can adversely affectcomprehension of risk and may have an influence on patients’decisions to take the medicine in question [1–6]. The effect ofpresentation format has also been investigated with users of apatient information website [7–9] with the general conclusionthat, regardless of format, people tend to overestimate the risk thatthey will experience medicine side effects, although the effect ismore marked in some formats. For instance, Knapp et al. [9]presented side effect risk information about tamoxifen in afactorial design using either frequency bands (e.g., ‘‘Affects morethan 1 in 10 people’’) or absolute frequencies (e.g., ‘‘Occurs in about

* Corresponding author at: Institute of Psychological Sciences, University of

Leeds, Lifton Place, Leeds, West Yorkshire LS2 9JT, UK. Tel.: +44 113 343 6678;

fax: +44 113 343 5749.

E-mail address: p.h.gardner@leeds.ac.uk (P.H. Gardner).

Please cite this article in press as: Gardner PH, et al. The effect of numusers of a patient information website. Patient Educ Couns (2011),

0738-3991/$ – see front matter � 2011 Elsevier Ireland Ltd. All rights reserved.

doi:10.1016/j.pec.2011.05.006

48 people in 100’’), with or without verbal descriptors (such as‘‘common’’ or ‘‘rare’’). All but one of the mean risk estimatesdemonstrated an overestimation compared with the originalinformation. This effect was greatest for the ‘frequency bands’condition, and for the side effects with the lowest frequency.

The studies by Knapp et al. [7–9], also report the accuracy ofparticipants’ responses which is calculated in terms of the meanabsolute difference from the actual risk (low absolute differencestherefore equate to greater accuracy). Although the findings fromthese studies are mixed, there is evidence to suggest thatexpressing risk in terms of absolute frequencies can lead to moreaccurate estimations of risk [9].

However, closer examination of the data shows that partici-pants varied greatly in their ability to provide accurate riskestimates. A possible explanation for this variability is that somemay experience difficulty in comprehending the information as aresult of their level of numeracy. Indeed, the relationship betweennumeracy and the comprehension of health information hasreceived a good deal of attention in recent years and there are anumber of reviews that summarise the key issues [10–15]. Peterset al. [10] report that more numerate participants were better able

eracy on the comprehension of information about medicines indoi:10.1016/j.pec.2011.05.006

P.H. Gardner et al. / Patient Education and Counseling xxx (2011) xxx–xxx2

G Model

PEC-4078; No. of Pages 6

to use information to inform decision making and were lesssusceptible to the way the information had been ‘framed’. Theeffect of numeracy appeared to be independent of general level ofintelligence.

In essence, it is clear that a low level of numeracy can impair theperception of health risk information, although much depends onthe content of the information and the context within which it iscommunicated [13]. In particular there seems to be a distinctionbetween direct patient–clinician communication, when furtherexplanation can be provided more easily, and situations in which apatient is likely to be interpreting the information on their own(such as reading a leaflet or consulting a patient informationwebsite). In considering the latter, the possible interactionbetween the format of written or visual materials and numeracylevel is likely to be of key interest. It is in this situation that thegreatest consideration is required when developing potentialinterventions for improving health communication with people ofdiffering numeracy levels.

Keller and Siegrist [16] provided participants with high and lowrisk information about prenatal screening and colon cancerscreening using different presentation formats. The risk percep-tions of less numerate participants did not vary according to theformat used to convey level of risk. By contrast, participants withhigher numeracy were able to perceive differences in risk but onlywhen information was presented using the Paling PerspectiveScale. In addition, Wright et al. [17] report an experimental studyin which high, medium and low frequency risk information aboutcontracting Crohn’s disease was presented in vignettes using twographical formats (grouped dot vs. dispersed dot displays). Theyreport that more numerate participants were more likely to have amore accurate understanding of objective risks regardless of thetype of display used.

In sum, it would appear that, on the basis of these two studies,the effect of numeracy is likely to pervade the perception of riskinformation across different presentation formats. However, alimitation noted by the authors of both studies is that theirparticipants were presented with imaginary situations and as suchthe results may not readily generalise to clinical contexts. Incontrast, the current research presents data gathered from users ofthe Cancer Research UK (CR-UK) patient information website(which provides comprehensive information about cancer and itstreatment). The method of recruitment therefore resulted in a highproportion of participants who reported personal experience ofcancer and/or taking the cancer medicine tamoxifen. The aim of thepaper is to investigate the relationship between the level ofparticipants’ numeracy and the accuracy of their risk estimationfollowing the presentation of information about the risk of sideeffects from taking tamoxifen.

2. Methods

2.1. Design

This paper amalgamates data taken from four separate studiesconducted by the authors using the CR-UK patient informationwebsite; two of the studies have been published and have beendiscussed earlier, but do not report the results presented here [8,9].Each of the four studies used the same scenario in whichinformation about four side effects from the medicine tamoxifenwas presented (see Fig. 1 for an example taken from Knapp et al.[9]), but each study compared different formats for representingthese risks (including natural and absolute frequencies, percen-tages, verbal descriptors, frequency bands and some combinationsof these formats). However, the principal aim of the researchreported here was to investigate whether participants’ numeracywas related to their ability to provide accurate estimations of the

Please cite this article in press as: Gardner PH, et al. The effect of numusers of a patient information website. Patient Educ Couns (2011),

risk of side effects regardless of presentation format. This wasbecause we found that there was no interaction between format ofpresentation and the relationship between numeracy and riskestimates. The effect of participant numeracy was also investigatedin relation to participants’ subjective responses to questions aboutthe information that was presented.

2.2. Procedure

Participants were recruited from the CR-UK patient informationwebsite, Cancer Help UK (http://www.cancerhelp.org.uk/), using a‘pop-up’ request linked to the webpage containing informationabout the medicine tamoxifen. After agreeing to take part in a study,participants were randomly allocated to a presentation formatcondition (although this is of no relevance to this paper). Followingthis, a scenario was presented which included risk informationabout four side effects that people taking tamoxifen may have (anexample is shown in Fig. 1). The four side effects with associatedincidence rates were ‘hot flushes’ (48%), ‘cataracts’ (3%), ‘deep veinthrombosis’ (0.2%) and ‘pulmonary embolism’ (0.06%). At this pointparticipants completed the measures as described in Section 2.4below. Participants were thanked and the ‘pop up’ window closed toreturn them to the tamoxifen page on the CR-UK website.

2.3. Participants

591 participants completed the studies (568 females, 21 maleand 2 unspecified) with a mean age of 46.5 years (sd = 10.8;range = 15–66 years). However, one study did not includequestions regarding the side effect ‘hot flushes’, therefore only448 (432 females; 16 males; 2 unspecified) participants wereasked to provide data relating to this side effect. The characteristicsof the sample are shown in Table 1. It can be seen that 461 (78%)indicated that they had cancer and 262 (44.3%) were eithercurrently taking tamoxifen or had taken it previously. In addition97 people (16.4%) had a close friend or relative who had cancer. Ofthose people who were taking or had taken tamoxifen, 5.4% statedthat they had experienced cataracts, 0.4% pulmonary embolism,83.1% hot flushes, and 1.9% deep vein thrombosis. A chi-square testof association demonstrated that the distribution of numeracylevels was approximately equal across the different presentationformats (chi-square = 4.253; df = 9; p = 0.894).

2.4. Measures

Initially, participants were asked to provide subjective ratingsabout the information they had received using a 6-point ratingscale (with verbal labels for points 1 and 6 but no other labels forintermediate points) for each of the questions as shown in Table 2.Participants were also asked to make estimates of the probabilitythat they would experience each of the side effects when takingtamoxifen. For example participants were asked ‘What do youthink is the chance that you will have hot flushes from takingtamoxifen? Please state as a percentage in the box below ___ %’. Inaddition, participants were asked ‘What do you think is the chancethat you will have ANY side effect from taking tamoxifen?’ and‘What do you think is the chance that the average person takingtamoxifen will have ANY side effects?’ (to look at any differencescaused by perceived personal susceptibility). Participants wereasked to answer these questions by providing a percentage. Thus,participants provided twelve answers relating to their perceptionof the information that they received. Following this, participantswere asked to complete a 4-item numeracy task (see Fig. 2) usingitems adapted from Lipkus et al. [18] and Gigerenzer [19].Participants scored one mark for each correct answer, thus thepossible scores ranged from 0 to 4.

eracy on the comprehension of information about medicines indoi:10.1016/j.pec.2011.05.006

In this im aginary s ituation your doctor has told yo u that yo u need to ta ke the hormone therapy, ta moxife n.

• Ple ase read the inform ation below about tamoxif en

• Then an swer th e questio ns that follo w

• Yo u can look at th e info rmation again when an swering the questions.

We are interest ed in your first though ts - plea se don't spend too lon g thi nking abou t your answers.

Tam oxifen ha s some side eff ects ov er a five year period of ta king th e drug. The se in clud e:-

• Hot flushes

- In abou t 48 peopl e in 100 (provide d in th e Absolute Frequen cy co ndition)

• Hot flushes

- Af fect s more than 1 in 10 people (provide d in th e Frequency Band con dition)

• Hot flushes

- Very common (occurs in about 48 peop le in 100) (provide d in th e Verbal + Absolu te Frequen cy conditi on)

• Hot flushes - Ve ry comm on (affects more than 1 in 10 patient s) (provide d in th e Ve rbal + Frequ ency Band condition)

Fig. 1. Scenario given to participants and examples of four presentation formats.

Taken from Knapp et al. [9].

P.H. Gardner et al. / Patient Education and Counseling xxx (2011) xxx–xxx 3

G Model

PEC-4078; No. of Pages 6

2.5. Statistical analyses

The data were analysed using the PASW statistical package(version 17). The accuracy of the estimate for each individual sideeffect was calculated using the absolute difference between theestimate and the actual incidence rate that was relevant for eachside effect (see Section 2.2). In the case of presentation formats thatused frequency bands or verbal descriptors, the actual incidencerate was consequently not known to the participant but was usedin the calculation of accuracy. It is therefore important to note thatbecause accuracy is measured in terms of the deviation of theestimate from the correct value, lower values actually indicatehigher accuracy.

With regard to the question about the risk that any side effectwould be experienced, the actual likelihood was not known, and sothe maximal chance that any one of the four side effects wouldoccur was calculated to be 49.7% using the formula shown in Fig. 3.This formula assumes that the occurrence of each side effect isindependent of the others, which results in a maximal rate.

Following these computations, descriptive measures werecalculated and then a number of Spearman’s rank correlationswere conducted to investigate firstly the relationship between

Table 1Characteristics of the sample (n = 591).

Gender 568 female; 21 male; 2 missing

Age Mean 46.5 (sd 10.8). Range 15–66

Location 428 UK; 93 USA; 70 Other

English as first language 573 yes; 17 no; 1 missing

Reason for visiting the

webpage

230 (38.9%) currently taking tamoxifen

50 (8.5%) have cancer but not taking tamoxifen

32 (5.4%) have previously taken tamoxifen

110 (18.6%) were about to take tamoxifen

97 (16.4%) have a close relative or friend

with cancer

31 (5.2%) health professionals

41 (6.9%) none of the above, just looking

Please cite this article in press as: Gardner PH, et al. The effect of numusers of a patient information website. Patient Educ Couns (2011),

numeracy scores and each of the subjective rating scale measures,and then secondly, the relationship between numeracy scores andthe accuracy of participants’ risk estimates. Finally, furthercorrelational analyses and two multivariate analyses of variancetested whether responses from participants who had cancer orprior experience of taking tamoxifen differed from those who didnot.

3. Results

3.1. Subjective ratings about the information

The mean and standard deviation for each of the questionsrelating to participants’ perceptions about the information areshown in Table 2. The variables of particular interest relate toparticipants’ perceptions about whether the information wouldaffect their decision to take tamoxifen (mean rating = 4.10) andhow beneficial they felt that taking tamoxifen would be (meanrating = 4.80). In both cases, the mean ratings were on the positiveside indicating that, overall, participants felt that the informationwould affect their decision and that tamoxifen was perceived asbeing beneficial.

3.2. Numeracy scores

With regard to the spread of numeracy scores amongstparticipants, it was found that 12 (2.0%) participants failed toget any numeracy questions correct, 53 (9.0%) participants scoredone correct, 133 (22.5%) scored two correct, 169 (28.6%) scoredthree and 224 (37.9%) got all four questions correct. This indicateda variation in the numeracy level across the sample.

3.3. Relationship between numeracy and other measures

Table 3 shows the results of the correlation analyses of therelationship between numeracy and the subjective ratings of the

eracy on the comprehension of information about medicines indoi:10.1016/j.pec.2011.05.006

Table 2Mean (sd) for participants’ subjective ratings of information about drug side effects.

Question (the labels for each 6-point rating scale are shown in brackets) Mean (sd)

Overall, how satisfied are you with the information you have just read about the risk of side effects

from tamoxifen? (1, not at all satisfied; 6, very satisfied)

3.6 (1.4)

Thinking about the information you have just read, how bad overall are the side effects of

tamoxifen? (1, not at all severe; 6, very severe)

3.4 (1.1)

From the information you have just read, how likely is it that you would have a side effect from

tamoxifen? (1, not at all likely; 6, very likely)

3.8 (1.4)

Thinking about the information you have just read, what do you think is the general risk to health

from tamoxifen? (1, no risk to all; 6, very high risk)

3.2 (1.1)

How much would the side effect information you have just read affect your decision to continue

with tamoxifen? (1, definitely would not affect; 6, definitely would affect)

4.1 (1.7)

How likely do you think it is that you will benefit from taking tamoxifen? (1, not at all likely; 6, very

likely)

4.8 (1.3)

P.H. Gardner et al. / Patient Education and Counseling xxx (2011) xxx–xxx4

G Model

PEC-4078; No. of Pages 6

information, and also of the relationship between numeracy andthe accuracy of participants’ side effect risk estimations. Table 3also shows the results of these analyses conducted on subsets ofthe sample. We looked to see whether the results were differentfirstly for participants with and without cancer, and secondly forthose who had, or had not, previously taken, tamoxifen.

The correlations demonstrate that, both in the whole sampleand in each of the subsets, there is a strong relationship betweenlevel of numeracy and the accuracy of side effect risk estimates(note that the correlations are negative because accuracy ismeasured in terms of the deviation from the correct value and solower scores indicate higher accuracy). The main exception relatesto the estimates of the risk of hot flushes for which the relationshipis not significant in the whole sample. However, analysis of thesubsets reveals that this relationship is significant for participantswith cancer and for those who have taken tamoxifen, whereas forthose without cancer and no experience of tamoxifen, it is not.

In addition, numeracy scores were positively related to howmuch participants felt that the information about side effectswould influence their decision to take the medicine (r = .10,p � .05). The analysis within the sub-groups revealed that thisrelationship was only significant for those who have not had cancer(r = .24, p � .01) or who have not previously taken tamoxifen(r = .13, p � .05).

Finally, within the whole sample, numeracy was found to benegatively correlated with ratings of satisfaction with the

Table 3Correlations between numeracy and subjective ratings of the information and accurac

(Spearman’s rank correlation).

Whole sample

(n = 591)

Correlations between numeracy and subjective

ratings of the information

Satisfaction with information about tamoxifen �.10*

How bad overall are side effects from taking tamoxifen �.04

Likelihood of side effects for taking tamoxifen �.01

General risk to health from taking tamoxifen �.06

How much would this affect your decision to take tamoxifen .10*

Likely benefit from taking tamoxifen .07

Correlations between numeracy and accuracy of risk

estimate (excluding participants with missing data)

Hot flushes y �.09

Cataracts �.24**

Deep vein thrombosis �.41**

Pulmonary embolism �.48**

Risk of self getting any side effect �.02

Risk of average person getting any side effect .03

* Correlation is significant at the 0.05 level (2-tailed).** Correlation is significant at the 0.01 level (2-tailed).y One study did not include questions about hot flushes, therefore the number of par

n = 348; those who have not had cancer, n = 100; those who have taken/are taking tam

Please cite this article in press as: Gardner PH, et al. The effect of numusers of a patient information website. Patient Educ Couns (2011),

information (r = �.10; p � .05). However, further analysis withinthe sub-groups revealed that this relationship was only significantfor those who were currently taking or had previously takentamoxifen. This indicates that more numerate tamoxifen usersindicated lower satisfaction with the information that had beenpresented.

The mean accuracy combined across all side effect riskestimations was also calculated for each participant dependingon their level of numeracy and is shown in Table 4. In addition,Table 4 also shows the proportion of participants within eachnumeracy level that provided exactly correct risk estimations. Itcan be seen that, on average, participant accuracy improved acrossall levels of numeracy but that even within the highest numeracylevel only around 1 in 5 participants provided correct values for therisk of side effects. There also appears to be some evidence tosuggest that participants scoring 2 or less on the numeracy testwere less accurate in their estimations but unequal numbers ineach level did not permit further analysis.

4. Discussion and conclusion

4.1. Discussion

It would appear that errors in interpreting side effect riskinformation are partly related to individuals’ level of numeracy.The relationship between numeracy and accuracy of side effect risk

y, for the whole sample and according to experience of cancer and tamoxifen use

Those who

have cancer

(n = 461)

Those who have

not had cancer

(n = 130)

Those who have

taken/are taking

tamoxifen (n = 262)

Those who have

not taken tamoxifen

(n = 329)

�.09 �.13 �.12* �.07

�.08 .13 �.08 �.01

�.03 �.03 �.02 .04

�.08 .03 �.13* �.01

.08 .24** .09 .13*

.10* �.02 .07 .10

�.15** �.08 �.25** �.03

�.27** �.17 �.31** �.20**

�.44** �.29** �.44** �.39**

�.50** �.38** �.50** �.46**

�.04 .002 �.08 .04

�.003 .03 �.04 .07

ticipants were reduced as follows: whole sample, n = 448; those who have cancer,

oxifen, n = 201; those who have not taken tamoxifen, n = 247.

eracy on the comprehension of information about medicines indoi:10.1016/j.pec.2011.05.006

Table 4Mean accuracy (sd) and proportion correct (combined across all side effect estimations) within each numeracy level (N = 591).

Numeracy level (score on numeracy test) N Mean accuracy (sd) % Correct estimates (across all side effects)

Zero 12 36.7 (22.4) 0

1 53 23.9 (21.3) 1.6

2 133 23.3 (20.8) 6.9

3 169 14.9 (16.1) 10.4

4 224 10.7 (12.6) 20.2

1. Im agine th at we fli p a fair coi n 1,00 0 times. Wh at is your be st guess about ho w many times the coin would come up heads?

2. In an im agi nary lott ery, the ch anc e of win nin g a prize is 1%. If 1,000 people each buy a single ticket, how man y wo uld win a prize?

3. In an imagin ary sweepstake , the chance of winning a car is 1 in 1,000. Wha t perce nt of tickets in the swee psta ke win s a car?

4. Wh at do es 40 perce nt mean? [3 op tion s: on e qu arter; 4 out of 10 ; eve ry 40th person]

Fig. 2. Items used to measure numeracy.

Adapted from [18,19].

P.H. Gardner et al. / Patient Education and Counseling xxx (2011) xxx–xxx 5

G Model

PEC-4078; No. of Pages 6

estimation was stronger for side effects with a relatively smallactual incidence rate. For instance, the actual incidence rates fordeep vein thrombosis and pulmonary embolism were quoted asfractions of 1% and it would appear that participants with lowernumeracy levels struggled to comprehend the level of risk in thesecases. These findings are in line with those of previous studies thattested graphical methods of representing risk information, inparticular the possibility that less numerate individuals may find itmore difficult to discriminate between different levels of riskparticularly when the actual risks are small [16,17]. Although therelationship between numeracy and accuracy appears to be similarregardless of participants’ experience of cancer and tamoxifen use,it is possible that personal experience of side effects amongst thoseparticipants who have taken tamoxifen may have influenced theresults. In this sample, the difference in the reported frequency ofside effects means that such analysis was not possible. Forinstance, hot flushes were experienced by almost all those who hadtaken tamoxifen, whereas Pulmonary Embolism has been experi-enced by very few. Nevertheless, this could be an important factorand may be an area worthy of further research.

There was also some evidence to suggest that participants withlower levels of numeracy were marginally more satisfied with theinformation they received but conversely, they tended to makelarger errors in interpreting the risk information. It is not certainwhy this would be the case and given the relatively weak

p (o ne or more s ide eff ects) = 1 - p(no s ide eff ects )

= 1 - [p(n o hot fl ushe s) * p(no cata ract ) * p(no DVT) * p( no PE)]

= 1 - [(1-0.48 ) * (1 - 0.03 ) * (1-0.002 ) * (1-0.0006)]

= 1 - [0.52 * 0.97 * 0.998 * 0.9994 ]

= 1 - 0.503089

= 0.49 7 = 49.7%

Fig. 3. Calculation for the maximal chance of having at least one side effect from

tamoxifen (p). Note: DVT, deep vein thrombosis; PE, pulmonary embolism.

Please cite this article in press as: Gardner PH, et al. The effect of numusers of a patient information website. Patient Educ Couns (2011),

relationship between numeracy and satisfaction this is an area thatwould benefit from further work.

However, a potential limitation of this research concerns therequirement for participants to provide side effect risk estimates interms of a percentage figure. Given that three of the questions usedto measure numeracy involve percentages, it could be argued thatit is not surprising that those with lower numeracy levels are morelikely to find difficulty in expressing risk in terms of a percentagefigure. Although the numeracy questions used are well-known (see[15] for a discussion), one suggestion would be to attemptreplication of this work using a different assessment of numeracy.The most promising alternative would be the Newest Vital Sign(NVS) [20] which has been proposed as a measure of health literacybased on numeracy that can be administered very quickly (lessthan 3 min).

The evidence also demonstrated that those with highernumeracy scores thought that the information they receivedwould influence their decision to take the medicine more thanthose with lower numeracy and, for those who have cancer, thatthe likelihood of benefit from taking the medicine would be higher.Taking these two results together, it seems reasonable to concludethat participants with higher levels of numeracy felt moreconfident in their interpretation of side effect risks and that,given their perception of benefit from taking the medicine, thosewith cancer would be more likely continue with their treatment.Alternatively, one might interpret these findings on the basis thatpeople with lower levels of numeracy are more confused by therepresentations of side effect risk and that this might diminishtheir willingness to continue with treatment.

4.2. Conclusion

The findings reported in this paper are important because theyare based on data gathered from a sample that is particularlyrelevant for this research area. Access to participants was achievedvia a patient information website dedicated to the provision ofinformation about cancer and its treatment, and as a result, a veryhigh proportion of respondents had personal experience of cancerand of taking the medicine tamoxifen. The data suggested thatwithin this sample, people with a lower numeracy level makelarger errors in interpreting medicines side effect risk information.

eracy on the comprehension of information about medicines indoi:10.1016/j.pec.2011.05.006

P.H. Gardner et al. / Patient Education and Counseling xxx (2011) xxx–xxx6

G Model

PEC-4078; No. of Pages 6

Such errors may also be related to perceptions about the benefit oftaking the medicine and decisions about either starting orcontinuing to take the medicine.

4.3. Practice implications

Pharmacists, other health professionals and patient informationwebsites should pay particular attention to providing clearexplanations about levels of risk of medicines (e.g., using absoluteor ‘natural’ frequencies), particularly where frequency is small.This is especially important for people with low levels ofnumeracy. Future research into the format of side effect riskinformation should therefore take account of numeracy level sothat the benefits and drawbacks of different formats for peoplewith different numeracy levels can be investigated, e.g., bettervisual depiction of risk, and particularly very small levels of risk.Pharmacists should also give consideration to the possibility ofmeasuring numeracy before dispensing medicines for which thecomprehension of side effect risk is considered to be particularlyimportant. This might be achieved by using the numeracyquestions detailed in this paper, although some clients may findsuch a ‘test’ to be off-putting. Alternatively, pharmacists could usea measure of health literacy, such as the Newest Vital Sign [20],which has more relevance to everyday tasks that clients are likelyto encounter on a regular basis. However there may be practicalbarriers to undertaking such an assessment in practice.

Conflict of interest

DK Raynor is a Director of Luto Research Ltd. which providespatient information testing services.

Funding

The study was funded by the internal support of the Universityof Leeds, UK. No external sponsor provided funds.

Acknowledgments

We would like to thank Dave Horton and Robin Waxman fortheir help with data collection.

Please cite this article in press as: Gardner PH, et al. The effect of numusers of a patient information website. Patient Educ Couns (2011),

References

[1] Berry DC, Knapp P, Raynor DK. Provision of information about drug side effectsto patients. Lancet 2002;359:853–4.

[2] Berry DC, Knapp P, Raynor DK. Is 15% very common? Informing people aboutthe risks of medicine side effects. Int J Pharm Pract 2002;10:145–51.

[3] Berry DC, Raynor DK, Knapp P. Communicating risk of medication side effects:an empirical evaluation of EU recommended terminology. Psychol Health Med2003;8:251–63.

[4] Berry DC, Holden W, Bersellini E. Interpretation of recommended risk terms:differences between doctors and lay people. Int J Pharm Pract 2004;12:117–24.

[5] Berry DC, Hochhauser M. Verbal labels can triple perceived risk in clinicaltrials. Drug Inf J 2006;40:249–58.

[6] Knapp P, Raynor DK, Berry DC. Comparison of two methods of presenting riskinformation to patients about the side effects of medicines. Qual Saf HealthCare 2004;13:176–80.

[7] Knapp P, Gardner PH, Carrigan N, Raynor DK, Woolf E. Perceived risk ofmedicine side effects in users of a patient information website: a study ofthe use of verbal descriptors, percentages and natural frequencies. Brit J HealthPsychol 2009;14:579–94.

[8] Knapp P, Raynor DK, Woolf E, Gardner PH, Carrigan N, McMillan B. Communi-cating the risk of side effects to patients: an evaluation of UK regulatoryrecommendations. Drug Saf 2009;32:837–49.

[9] Knapp P, Gardner PH, Raynor DK, Woolf E, McMillan B. Perceived risk oftamoxifen side effects: a study of the use of absolute frequencies or frequencybands, with or without verbal descriptors. Patient Educ Couns 2010;79:267–71.

[10] Peters E, Vastfjall D, Slovic P, Mertz CK, Mazzocco K, Dickert S. Numeracy anddecision making. Psychol Sci 2006;17:407–14.

[11] Ancker JS, Kaufman D. Rethinking health numeracy: a multidisciplinaryliterature review. J Am Med Inform Assoc 2007;14:713–21.

[12] Schwartz LM, Woloshin S, Welch HG. Can patients interpret health informa-tion? An assessment of the medical data interpretation test. Med DecisMaking 2005;25:290–300.

[13] Apter AJ, Paasche-Orlow MK, Remillard JT, Bennett IM, Ben-Joseph EP, BatistaRM, et al. Numeracy and communication with patients: they are counting onus. J Gen Intern Med 2008;23:2117–24.

[14] Visschers VHM, Meertens RM, Passchier WWF, de Vries NNK. Probabilityinformation in risk communication: a review of the research literature. RiskAnal 2009;29:267–87.

[15] Reyna VF, Nelson WL, Han PK, Dieckmann NF. How numeracy influences riskcomprehension and medical decision making. Psychol Bull 2009;135:943–73.

[16] Keller C, Siegrist M. Effect of risk communication formats on risk perceptiondepending on numeracy. Med Decis Making 2009;29:482–90.

[17] Wright AJ, Whitwell SCL, Takeichi C, Hankins M, Marteau TM. The impact ofnumeracy on reactions to different graphic risk perception formats: an ex-perimental analogue study. Brit J Health Psychol 2009;14:107–25.

[18] Lipkus IM, Samsa G, Rimer BK. General performance on a numeracy scaleamong highly educated samples. Med Decis Making 2001;21:37–44.

[19] Gigerenzer G. Reckoning with risk. London: Penguin; 2002.[20] Weiss BD, Mays MZ, Martz W, Castro KM, DeWalt DA, Pignone MP, et al. Quick

assessment of literacy in primary care: the newest vital sign. Ann Fam Med2005;3:514–22.

eracy on the comprehension of information about medicines indoi:10.1016/j.pec.2011.05.006