Computer-Based Interventionsto Improve Self-management inAdults With Type 2 Diabetes:A Systematic Review andMeta-analysisDiabetes Care 2014;37:1759–1766 | DOI: 10.2337/dc13-1386

OBJECTIVE

Structured patient education programs can reduce the risk of diabetes-relatedcomplications. However, people appear to have difficulties attending face-to-faceeducation and alternatives are needed. This review looked at the impact of computer-based diabetes self-management interventions on health status, cardiovascularrisk factors, and quality of life of adults with type 2 diabetes.

RESEARCH DESIGN AND METHODS

We searched The Cochrane Library, Medline, Embase, PsycINFO, Web of Science,and CINAHL for relevant trials from inception to November 2011. Reference listsfrom relevant published studies were screened and authors contacted for furtherinformation when required. Two authors independently extracted relevant datausing standard data extraction templates.

RESULTS

Sixteen randomized controlled trialswith 3,578 participantsmet the inclusion criteria.Interventions were delivered via clinics, the Internet, and mobile phones. Computer-based diabetes self-management interventions appear to have small benefits onglycemic control: the pooled effect on HbA1c was 20.2% (22.3 mmol/mol [95% CI20.4 to 20.1%]). A subgroup analysis on mobile phone–based interventionsshowed a larger effect: the pooled effect on HbA1c from three studies was20.50%(25.46 mmol/mol [95% CI20.7 to20.3%]). There was no evidence of improvementin depression, quality of life, blood pressure, serum lipids, or weight. There was noevidence of significant adverse effects.

CONCLUSIONS

Computer-based diabetes self-management interventions to manage type 2 di-abetes appear to have a small beneficial effect on blood glucose control, and thiseffect was larger in the mobile phone subgroup. There was no evidence of benefitfor other biological, cognitive, behavioral, or emotional outcomes.

The burden of diabetes is growing, with 347 million people currently affectedworldwide (1) and numbers projected to increase to 552 million by 2030 (2). TheInternational Diabetes Federation suggests that in the developed world, the cost ofcaring for patients with diabetes is double that of the background population, and

1UCL Research Department of Primary Care andPopulation Health, University College London,London, U.K.2International Centre for Circulatory Health, Im-perial College, London, U.K.3Department of Clinical, Educational and HealthPsychology,University College London, London,U.K.4Department of Primary Care Health Sciences,University of Oxford, Oxford, U.K.5Department of Diabetes, The Whittington Hos-pital NHS Trust, London, U.K.6Archway Healthcare Library, London, U.K.7Diabetes Self-Management Program (DSMP),Co-creating Health, London, U.K.8Department of Population Health, London Schoolof Hygiene & Tropical Medicine, London, U.K.

Corresponding author: Kingshuk Pal, k.pal@ucl.ac.uk.

Received 11 June 2013 and accepted 29 January2014.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc13-1386/-/DC1.

This article presents independent researchfunded by the National Institute for Health Re-search (NIHR). The views expressed are those ofthe author(s) and not necessarily those of theNHS, the NIHR, or the Department of Health.

© 2014 by the American Diabetes Association.See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.

Kingshuk Pal,1 Sophie V. Eastwood,2

Susan Michie,3 Andrew Farmer,4

Maria L. Barnard,5 Richard Peacock,6

Bindie Wood,7 Phil Edwards,8 and

Elizabeth Murray1

Diabetes Care Volume 37, June 2014 1759

META

-ANALYSIS

the estimated cost of managing peoplediagnosed with diabetes in the U.S. in2007 was $174 billion (3). Improvingself-management is an important compo-nent of improving cost-effective patient-centered care and dealing with thisgrowing health care challenge (4,5).The current gold standard for self-

management is face-to-face groupeducation (6,7). However, current at-tendance of diabetes education is poor(8). Digital interventions have thepotential to increase access to self-management training and improve out-comes if this can be done effectively.This review summarizes the effectsof computer-based self-managementinterventions on adults with type 2 di-abetes and uses a taxonomy of behavior-change techniques todescribe thepotentialactive components of these complexinterventions. Included interventionswere defined as any application thattook input from a patient and used com-munication or processing technologyto provide a tailored response, facilitat-ing one or more aspect of diabetes self-management. Interventions used mainlyfor communication between patientsand professionals were not included inthe review.

RESEARCH DESIGN AND METHODS

We searched the following electronicdatabases for trials or conference pro-ceedings from inception until November2011: The Cochrane Library, Medline,Embase, PsycINFO, Web of Science,and CINAHL. The search strategy forMedline can be found in SupplementaryTable 1. Three other databases weresearched for gray literature: Aslib Indexto Theses, Australasian Digital ThesesProgram, and ProQuest Digital Disserta-tions and Theses. Reference lists fromrelevant published studies were screenedand authors contacted for further infor-mation when required.Included studies were randomized,

controlled clinical trials involving pa-tients aged $18 years with type 2 dia-betes mellitus. Interventions met thefollowing criteria for inclusion: interactwith users to generate tailored contentthat aimed to improve one or more dia-betes self-management domains throughfeedback, advice, reinforcement, and re-wards, patient decision support, goal set-ting, or reminders. Studies published inany languagewere included.We excluded

studies of interventions that weretargeted only at patients with type 1 dia-betes, involved participants aged ,18years, used only for communicationbetween patients and professionals,or targeted at health professionals. Stud-ies of mixed populations of patients inwhich there was a majority of patientswith type 2 diabetes were included, butwhere possible, only data for patientswith type 2 diabetes were included inanalyses. Possible comparison groups in-cluded standard diabetes care, noninter-active computer-based programs, papereducationalmaterial, delayed start/waitinglist, or face-to-face self-managementeducation. Primary outcomes werehealth-related quality of life, HbA1c,or death from any cause. Secondaryoutcomes were changes in cognitions,behaviors, social support, cardiovascularrisk factors (blood pressure, serum lipids,and weight), complication rates, emo-tional outcomes, hypoglycemia, adverseeffects, cost-effectiveness, and economicdata.

Two authors independently scannedthe abstracts of retrieved reports andpotentially relevant articles were inves-tigated as full text. For studies that ful-filled the inclusion criteria, two authorsindependently extracted data usingstandard data extraction templateswith any disagreements resolved by dis-cussion with the study steering group.Any relevant missing information onthe trial was sought from the originalauthor(s) of the article. Quality assess-ment of the included studies used theCochrane Collaboration’s tool (9) forrisk of bias assessment.

Statistical analysis was performed ac-cording to the guidelines referenced inversion 5.1.0 of the Cochrane Handbookfor Systematic Reviews of Interventions(9). Where studies provided sufficientdata for meta-analysis, the CochraneCollaboration’s RevMan software wasused to perform ameta-analysis poolingthe mean difference or difference inmeans. Where outcomes were mea-sured on different scales, standardizedmean differences were combined. Instudies in which outcome data werenot suitable for meta-analysis, thedata were described narratively. Meta-analyses were based on a random-effects model. Heterogeneity wasidentified by visual inspection of theforest plots and examined with the I2

statistic to quantify inconsistency acrossstudies (9).

Specifying the Content of theIntervention and Control ConditionsThe intervention and control conditionswere coded in terms of their componentactive ingredients using a taxonomyof behavior-change techniques (10) (Sup-plementary Table 2). The two authorsrecorded if reports of a theoretical mech-anism were mentioned and whether itwas applied to the intervention. Therewere too few studies to conduct a meta-regression to investigate which behavior-change techniques were effective in theinterventions, so an exploratory exercisewas conducted involving comparison ofthe techniques that featured most com-monly in effective interventions to thosefeaturing most commonly in ineffectiveinterventions. Prespecified subgroupanalyses were performed on duration ofintervention and setting.

RESULTS

Descriptions of StudiesDatabase searches yielded 8,715 uniqueabstracts, of which 94 full-text articleswere assessed for eligibility. Twentyjournal articles describing 16 differentstudies with 3,578 participants fulfilledthe inclusion criteria and were selectedfor inclusion in the review (Fig. 1). Asummary of study characteristics canbe found in Table 1.

The numbers of participants rangedfrom 30 (11,12) to 886 (13) in eachstudy. One only included women (12).Participants in one study (14) were allLatin or Hispanic. Three studies (13,15,16)reported that.70% of participants wereCaucasian or non-Hispanic white. In 13studies, all participants had type 2 diabe-tes; 3 studies (12,17,18) included partic-ipants with both type 1 and type 2diabetes; the percentage of participantswith type 1 diabetes ranged from 19–26%. Six studies (11,17,19–22) reportedmean duration of diabetes with a rangeof between 6 and 13 years. The meanage of participants ranged from 46 (12)to 67 (19) years. Supplementary Table 3summarizes the interventions and con-trols used in the included studies.

Only three interventions referencedpsychological theories (15,20,23). Theseincluded the transtheoretical model,social ecological theory, social cognitivetheory, and self-determination theory.

1760 Computer-Based Self-management for Adults With T2D Diabetes Care Volume 37, June 2014

The behavior-change techniques used ineach intervention are described in Sup-plementary Table 4. A summary of theoutcome data from all the studies is pro-vided in Supplementary Table 5.

Overall Study QualityDetails of the risk of bias assessment ofthe included studies can be found inSupplementary Table 6. All of the in-cluded studies were randomized con-trolled trials but none were blinded,with most reports citing study designchallenges as the main reason for this.Two studies appeared to be at high

risk of selection bias. One study usedan inadequate randomization proce-dure (21), and another had a high attri-tion rate of 39% in the interventiongroup (20). A summary of the strengthof the evidence can be found in Supple-mentary Table 7.

Study OutcomesMost studies measured a number ofoutcome measures to reflect clinical

and service-user priorities. These in-cluded biological markers, cognitiveoutcomes, behavioral outcomes, emo-tional outcomes, cost-effectiveness,and adverse event data.

Cardiovascular Risk Factors

All of the studies had HbA1c as an out-come measure and 11 studies (13–16,18–20,22–25) provided enoughdata to combine in a meta-analysis, asshown in Fig. 2. Pooled results indicatethat there is a small, statistically sig-nificant difference in the outcomesbetween intervention and comparatorgroups of 2.3 mmol/mol or 20.21%(95% CI 20.37 to 20.05%), favoringthe intervention group. However, therewas substantial heterogeneity in theeffects interventions (I2 = 58%). Possiblereasons for the heterogeneity wereexplored in a sensitivity analysis, de-scribed below.

Five studies (14,18,20,22,23) looked atchanges in blood pressure, and one studyshowed evidence of improvement.

Seven studies (14,15,17,19,22,23,25)reported changes in BMI or weight, andfive studies (14,15,19,22,23) were com-bined in a meta-analysis. The overallpooled effect did not reach statisticalsignificance (pooled standardized meandifference 20.07 [95% CI 20.20 to0.05]) (Supplementary Fig. 3).

Ten studies (13–15,17,19,20,22–25)measured serum lipids, and seven (13–15,19,20,22,23) were combined in ameta-analysis. The overall pooled effectdid not reach statistical significance(pooled standardized mean differ-ence 20.11 [95% CI, 20.28 to 0.05])(Supplementary Fig. 4). One study (14)attributed the difference in lipids to dif-ferences in the use of lipid-loweringmedication between the two studygroups.

Cognitive Outcomes

All four studies (11,16,21,26) thatlooked at change in knowledge and un-derstanding reported positive effects ofthe interventions on knowledge. Two

Figure 1—Preferred reporting items for systematic reviews and meta-analyses flowchart. RCT, randomized controlled trial.

care.diabetesjournals.org Pal and Associates 1761

Table

1—Su

mmary

ofstudych

aracteristics

Study

(referen

cenumber)

Setting

Sex

(%female)

Age

[meanyears

(SD)]

HbA1c

[mean%

(SD)]

BMI

[meankg/m

2

(SD)]

Durationof

disease

[meanyears

(SD)]

Ethnicgroups(%

)Country

Durationof

interven

tion

Christian20

08(14)

Clinic-based

I:65

I:53

.0(11.3)

I:8.1(2)

I:35

.4(6.6)

d.60

%Latin/H

ispanic

U.S.

One30

-min

exposure

C:68

C:53

.4(10.7)

C:8.3(1.9)

C:34

.8(7.1)

d

Glasgow19

97(17)

Clinic-based

I:63

I:61

.7(12.1)

I:7.9

I:30

.4I:13

.0(9.9)

dU.S.

3330

min

C:60

C:63

.1(10.5)

C:7.9

C:30

.2C:13

.7(12.2)

d

Glasgow20

03(24)

Internet-based

dd

I:7.5(1.7)

dd

dU.S.

10months

dd

C:7.4(1.6)

dd

d

Glasgow20

05(13)

Clinic-based

I:52

I:62

(1.4)SE

I:7.3(1.3)

dd

I:white,

83.5%;black,1

.7%;

Hispanic,1

1.3%

;other,3

.4%

U.S.

12months

C:50

C:64

(1.3)SE

C:7.3(1.2)

dd

C:white,77

.9%;black,2

.7%;

Hispanic,1

4.1%

;other,5

.4%

Glasgow20

06(15)

Clinic-based

I:50

I:62

.0(11.7)

I:7.4(1.6)

dI:at

least

6months

I:white,

74.1%;Hispan

ic,17

.5%

U.S.

Oneexposure

tointerven

tion

C:50

C:61

.0(11.0)

C:7.5(1.6)

dC:at

least

6months

C:white79

.6%;Hispan

ic,18

.3%

U.S.

2phonecalls

Glasgow20

10(23)

Internet-based

I:45

I:58

.7(9.3)

I:8.0(1.9)

I:34

.5(6.3)

dI:American

Indian/AlaskaNative,

4.9%

;Asian

,1.9%;black

or

African

American

,14.8%

;white,

74.1%;Latinoethnicity,25

.3%

U.S.

4months

C:52

C:58

.7(9.1)

C:8.1(1.8)

C:34

.8(6.6)

dC:American

Indian/AlaskaNative,

11.1%;Asian,1.6%;black

or

African

American

,12.7%

;white,

70.6%;Latinoethnicity,16

.8%

Leu20

05(18)

Pagers

dAverage

ageof

51years

I:8.5

dd

Pred

ominan

tlyCau

casian

U.S.

3–6months

dC:8.5

dd

Lim

2011

(19)

Mobile

phones

I:54

I:67

.2(4.1)

I:7.8(1.0)

I:24

.7(2.4)

dd

South

Korea

6months

Lo19

96(26)

Clinic-based

I:75

I:61

.6(11.6)

Nonstan

dard

units

dd

dAustralia

I:3–6sessions,

1heach

C:50

C:63

.4(8.9)

dd

dC:4weeklysessions

2.5–

3h

Lorig20

10(16)

Internet-based

I:64

I:54

.2(9.9)

dd

dI:78

%non-Hispanicwhite

U.S.

6–18

months

C:71

C:54

.4(10.6)

dd

dC:71

.1%non-Hispanicwhite

Quinn20

08(11)

Mobile

phones

I:69

I:8aged

20–54

,5aged

55–64

years

I:9.5

dI:7.6

I:10

of13

African

,77%

;3of13

non-Hispanicwhite,23

%U.S.

3months

C:62

C:6aged

20–54

,7aged

55–64

years

C:9.1

dC:11

C:6of13

African

,46%

;7of13

non-Hispanicwhite,54

%

Con

tinuedon

p.17

63

1762 Computer-Based Self-management for Adults With T2D Diabetes Care Volume 37, June 2014

studies (11,16) reported on changes inself-efficacy, and both studies showedpositive effects of the interventions.

Behavioral Outcomes

The effects of interventions on physicalactivity were mixed. One study (14)showed a statistically significant in-crease in the number of patients achiev-ing $150 metabolic equivalent of taskmin of physical activity a week, andanother (23) showed a statistically sig-nificant improvement in physical activitybased on a subgroup analysis. Threestudies (11,16,24) found no improvementin physical activity scores. The six stud-ies (11,13,14,15,23,24) that looked atchanges in diet reported statistically signif-icant improvements in interventiongroups.

Effect on Emotional Outcomes

None of the studies (11,13,15,16,20,24)that looked at depression showed evi-dence of improvements in the interven-tion groups.

Cost-effectiveness Data

Two studies provided data on cost-effectiveness. Glasgow et al. (17)looked at the cost per patient for atouch-screen dietary intervention. De-pending on the volume of patientsseen, the cost per patient in 1997 rangedfrom $115–139, with a cost per unit re-duction of cholesterol between $7 and$8.40 and a cost per 1% reduction in fatof $52 and $63. One study (16) investi-gated health behavior and resource uti-lization but found no difference betweenintervention or control groups.

Adverse Events

One study (21) reported a participantwithdrawing due to anxiety. A total ofthree participants died during the stud-ies, but no deaths were attributed to theinterventions. There were no reportedstatistically significant differences in hy-poglycemic episodes between groups inany of the studies.

Subgroup Analyses

A previous meta-analysis of diabetesself-management interventions (18 of20 were face to face) showed a greatereffect from shorter studies with short-term follow-up (27). Therefore, we did asubgroup analysis to see if this hypothesismight also be true for computer-basedself-management interventions. Thestudies were divided into those with fol-low up of ,6 months and those withfollow-up for $6 months. When out-comes at ,6 months were combined

Table

1—Continued

Study

(referen

cenumber)

Setting

Sex

(%female)

Age

[meanyears

(SD)]

HbA1c

[mean%

(SD)]

BMI

[meankg/m

2

(SD)]

Durationof

disease

[meanyears

(SD)]

Ethnicgroups(%

)Country

Durationof

interven

tion

Quinn20

11(20)

Mobile

phones

I:48

I:52

.8(8.0)

I:9.3(1.8)

I:36

.9(7.5)

I:7.7(5.6)

I:black

(non-Hispan

ic)43

.5%;white

(non-Hispan

ic)52

.2%;other

4.3%

U.S.

12months

C:50

C:53

.2(8.4)

C:9.2(1.7)

C:34

.3(6.3)

C:9.0(7.0)

C:black

(non-Hispan

ic)48

.2%;white

(non-Hispan

ic)46

.4%;other

5.4%

Smith20

00(12)

Internet-based

All:

100

Meanage46

.7years

dd

dd

U.S.

5months

dd

dd

Wise19

86(21)

Clinic-based

d55

(21)

SEI:8.7(0.7)

dI:8(5)SE

dU.K.

4–6months

dC:8.7

dC:7(4)SE

d

Yoo20

09(22)

Mobile

phones

I:47

I:57

.0(9.1)

I:7.6(0.9)

I:25

.6(3.5)

I:6.0(5.4)

dSouth

Korea

12weeks

C:35

C:59

.4(8.4)

C:7.4(0.9)

C:25

.5(3.3)

C:7.2(6.0)

d

Zhou20

03(25)

Internet-based

I:61

I:62

.4(8.3)

I:8.7(1.5)

I:24

.0(3.1)

dd

China

8weeks

C:56

C:59

.8(11.0)

C:9.0(1.8)

C:24

.5(2.8)

dd

C,control;I,interven

tion.

care.diabetesjournals.org Pal and Associates 1763

(15,18,22,23,25), heterogeneity was re-duced (I2 = 43%) with a larger effect sizefor HbA1c of 23.5 mmol/mol or 20.3%(95% CI20.6 to20.1). Combining stud-ies with outcomes measured at $6months (13,14,16,19,20,24), the overalleffect size for HbA1c was smaller and nolonger statistically significant: 21.5mmol/mol or 20.1% (95% CI 20.3 to0.1), with substantial heterogeneity(I2 = 61%).Pooling the results of the three mobile

phone–based interventions (19,20,22)identified a statistically and clinically sig-nificant reduction in HbA1c of 25.5mmol/mol or 20.5% (95% CI 20.7 to20.3) with no heterogeneity (I2 = 0%).Interventions delivered at home (16,23–25) appeared to have a smaller effect:22.7 mmol/mol or 20.3% (95% CI20.5to 20.04%), and the result was still as-sociated with moderate heterogeneity(I2 = 47%).As the interventions showed signifi-

cant heterogeneity in the overall pooledresult (I2 = 58%), a sensitivity analysiswas done to explore possible reasons.An analysis was performed excludingthree studies formethodological reasons:one study (14) noted that participantsin the control group had larger changesin hypoglycemic medication than the in-tervention group, and two other studies(13,20) were cluster randomized but an-alyzed as individually randomized trials.

Removing these studies increasedthe pooled effect sl ightly to 2.95mmol/mol or 20.27% (95% CI 20.42 to20.12%).

Behavior-Change Techniques Used by

Interventions

The two behavior-change techniquesused most commonly by effective inter-ventions were: prompt self-monitoringof behavioral outcome and providefeedback on performance. In contrast,the three techniques most commonlyassociated with interventions that hadno significant impact on HbA1c wereprovision of information on consequen-ces of behavior in general, goal setting(behavior), and barrier identification/problem solving.

CONCLUSIONS

Computer-based diabetes self-managementinterventions appear to have small ben-efits on glycemic control (pooled effecton HbA1c: 22.3 mmol/mol or 20.2%;95% CI20.4 to20.1%; I2 = 58%). A sub-group analysis on mobile phone–basedinterventions showed a larger effect(pooled effect on HbA1c from three stud-ies: 25.46 mmol/mol or 20.50%; 95%CI 20.7 to 20.3%; I2 = 0%). Current in-terventions do not appear to be effec-tive in improving depression, quality oflife, or weight. There was no evidence ofsignificant adverse effects.

Evidence on the use of new technologyin diabetes is still evolving, with mixedresults. However, there are trends emerg-ing that may highlight the aspects of self-management that might be effectivelysupported through computer-based inter-ventions and the areas that may requiremore intensive or face-to-face input. Theresults of this revieware supportedbyfind-ings from a previous qualitative reviewthat looked at 26 studies of interactivecomputer-assisted technology in diabetescare (28). It identified14studies that lookedat HbA1c levels and found that 6 of 14 dem-onstrated significant declines in HbA1c.Studies that looked at changes in bodyweight, blood pressure, microalbuminuria,and renal function found no significant dif-ferences postintervention, while effects onlipids and depression were mixed.

The Impact of Mobile PhoneInterventionsA recent review focused on the effect ofmobile phone interventions for diabeteson glycemic control (29) and carriedout a meta-analysis of 22 trials with1,657 participants. This showed thatmobile phone interventions for diabetesself-management reduced HbA1c valuesby a mean of 6 mmol/mol or 0.5% (95%CI 0.3 to 0.7) over a median follow-upduration of 6 months. This is similar tothe effect size seen in this review whenthe effects of the three mobile phoneinterventions (19,20,22) were pooled.

Figure 2—Forest plot of meta-analysis of HbA1c results.

1764 Computer-Based Self-management for Adults With T2D Diabetes Care Volume 37, June 2014

The benefit identified with mobilephone interventions may be related tofeedback on performance and promptsfor blood glucose self-monitoring. Thiswould be consistent with Control The-ory. Previous reviews of interventionsto increase physical activity and healthyeating (10,30,31) identified a cluster oftechniques consistent with Control The-ory (32) that were associated with effec-tive interventions. These includedreceiving information about one’s be-havior (via self-monitoring or feedback)and having a strategy for acting onthis information (action planning or in-formation on where and when to per-form the behavior). This also appearsto be true for interventions aimed athealth professionals. A recent reviewfound that computer decision supportsystems aimed at primary care clinicianswere only likely to improve patientoutcomes if they provided feedback onperformance, reminders, and case man-agement (33). It is potentially significantthat similar techniques are effective ininterventions targeted at different audi-ences. Providing feedback and prompt-ing behavior appear to be criticalelements of behavior change for bothprofessionals and patients, andmaximiz-ing improvements in patient outcomesmay require complementary interven-tions designed to change behavior inboth clinicians and patients: a meta-regression of computerized clinical deci-sion support systems found that provisionof advice to both practitioners and pa-tients improved the success of such sys-tems (34).

Limitations of the ReviewThere are a number of limitations thatcould affect the results of the review.Although all of the included studies wererandomized control trials, they werenot double blinded, which may have aff-ected the treatment fromhealth care pro-fessionals involved in the study and/or thestudy investigators. Some of the controlgroups had potentially active interven-tions that might reduce the apparent ef-fectiveness of the interventions (e.g.,increases in hypoglycemic medication,goal setting, or increased monitoring byhealth care providers). Finally, the litera-ture search was run from inception untilNovember 2011, and studies publishedafter that date are not included. To esti-mate the impact of the age of the search,

the Medline search was rerun from No-vember 2011 to January 2014, and 963more abstracts were screened. Thissearch identified two relevant publica-tions, of which one was a follow-upfrom a study that has been includedin the meta-analysis. There were 16 pro-tocols published for studies that arecurrently in progress. The results aretherefore likely to reflect the current evi-dence base at the time of publication, butthis picture is likely to evolve over the nextfew years.

Implications for Future ResearchThere were few published protocols forthe studies, and the theoretical bases ofthe interventions were not alwaysclearly described in the published re-ports. As these interventions are thera-peutic agents, it would be beneficial toexplicitly prescribe interventions for tri-als and formally state the active ingre-dients (behavior-change techniques),dose (frequency and intensity of inter-actions), route (mode of delivery, Inter-net, mobile phone, etc.), and duration oftreatment.

It is also not clear why interventionsdelivered over mobile phones appear tobe more effective; it could be due toconvenience (and therefore adherence),intensity of the interventions (mobilephone interventions were more likelyto have multiple daily contacts), or thebehavior-change techniques used bythe interventions (mobile phone inter-ventions were more likely to use cues toprompt behavior and provide rapidfeedback afterward).

A clear understanding of the mecha-nism of action of effective interventionsmay also facilitate systemic improve-ments through complementary inter-ventions that cover the whole systemof health care delivery. Health care pro-fessionals and patients both appear tobenefit from computerized prompts andfeedback, and the optimal solution tomaximize improved outcomes may besystems that are able to target bothpopulations.

Thesmall treatmenteffect (2.3mmol/molor 0.2%) on HbA1c with computer-basedself-management interventions wouldbe important if it could be achieved andsustained across the population via theInternet (at very low cost), but farfrom cost-effective if it required signifi-cant health professional support and/or

additional drugs. More studies with longerfollow-up are needed to determine thecost-effectiveness of different types ofcomputer-based interventions, the long-term impact on health outcomes, and tolook for evidence of harm.

There also needs to be more researchdone to determine which populationgroups will benefit the most from theseinterventions (e.g., HbA1c.53mmol/molor 7%) and the impact of these interven-tions on older patients. There are alsoquestions surrounding the “digital divide”and whether access to such interventionsor their effectivenessmight be influencedby age, education, computer literacy, cul-ture, and affluence.

ConclusionComputer-based diabetes self-managementinterventions to manage type 2 diabetesappear to have a small beneficial effecton blood glucose control, and the effectwas largest in themobile phone subgroup.Better designed and more targeted inter-ventions are needed to improve other as-pects of diabetes self-management.

Acknowledgments. The authors thank theCochrane Metabolic and Endocrine DisordersGroup, Dusseldorf, Germany, for peer reviewand support with literature searches.Funding. This study was funded by the Nation-al Institute for Health Research (NIHR) Schoolfor Primary Care Research (SPCR).Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. K.P. and S.V.E. acquiredthe data and contributed to analysis and in-terpretation of the data, drafting the manuscript,and editing the manuscript. S.M., A.F., M.L.B.,R.P., B.W., P.E., and E.M. contributed to analysisand interpretation of the data and reviewing andediting the manuscript.

References1. Danaei G, Finucane MM, Lu Y, et al.; GlobalBurden of Metabolic Risk Factors of Chronic Dis-eases Collaborating Group (Blood Glucose). Na-tional, regional, and global trends in fastingplasma glucose and diabetes prevalence since1980: systematic analysis of health examinationsurveys and epidemiological studies with 370country-years and 2z7 million participants. Lan-cet 2011;378:31–402. International Diabetes Federation. IDF Dia-betes Atlas 5th Edition: The Global Burden, [In-ternet], 2011. Available fromhttp://www.idf.org/diabetesatlas/5e/the-global-burden. Accessed 11June 20123. American Diabetes Association. Economiccosts of diabetes in the U.S. In 2007. DiabetesCare 2008;31:596–6154. Norris SL, Engelgau MM, Narayan KM. Effec-tiveness of self-management training in type 2

care.diabetesjournals.org Pal and Associates 1765

diabetes: a systematic review of randomizedcontrolled trials. Diabetes Care 2001;24:561–5875. Norris SL, Lau J, Smith SJ, Schmid CH,Engelgau MM. Self-management education foradults with type 2 diabetes: a meta-analysis ofthe effect on glycemic control. Diabetes Care2002;25:1159–11716. Duke SA, Colagiuri S, Colagiuri R. Individualpatient education for people with type 2 diabe-tes mellitus. Cochrane Database Syst Rev 2009;1:CD0052687. Deakin T, McShane CE, Cade JE, Williams RD.Group based training for self-managementstrategies in people with type 2 diabetes melli-tus. Cochrane Database Syst Rev 2005;2:CD0034178. Duncan I, Birkmeyer C, Coughlin S, Li QE,Sherr D, Boren S. Assessing the value of diabeteseducation. Diabetes Educ 2009;35:752–7609. Higgins JPT, Green S. Cochrane Handbook forSystematic Reviews of Interventions Version5.1.0 [Internet], 2011. Available from http://www.cochrane-handbook.org. Accessed 5 Feb-ruary 201410. Michie S, Ashford S, Sniehotta FF, DombrowskiSU, Bishop A, French DP. A refined taxonomy ofbehaviour change techniques to help peoplechange their physical activity and healthy eatingbehaviours: the CALO-RE taxonomy. Psychol Health2011;26:1479–149811. Quinn CC, Clough SS, Minor JM, Lender D,Okafor MC, Gruber-Baldini A. WellDoc mobilediabetes management randomized controlledtrial: change in clinical and behavioral outcomesand patient and physician satisfaction. DiabetesTechnol Ther 2008;10:160–16812. Smith L, Weinert C. Telecommunicationsupport for rural women with diabetes. Diabe-tes Educ 2000;26:645–65513. Glasgow RE, Nutting PA, King DK, et al. Ran-domized effectiveness trial of a computer-assistedintervention to improve diabetes care. DiabetesCare 2005;28:33–3914. Christian JG, Bessesen DH, Byers TE,Christian KK, GoldsteinMG, Bock BC. Clinic-basedsupport to help overweight patients with type 2

diabetes increase physical activity and loseweight. Arch Intern Med 2008;168:141–14615. Glasgow RE, Nutting PA, Toobert DJ, et al.Effects of a brief computer-assisted diabetesself-management intervention on dietary, bio-logical and quality-of-life outcomes. Chronic Illn2006;2:27–3816. Lorig K, Ritter PL, Laurent DD, et al. Onlinediabetes self-management program: a random-ized study. Diabetes Care 2010;33:1275–128117. Glasgow RE, La Chance PA, Toobert DJ,Brown J, Hampson SE, Riddle MC. Long-termeffects and costs of brief behavioural dietaryintervention for patients with diabetes deliv-ered from the medical office. Patient EducCouns 1997;32:175–18418. Leu MG, Norris TE, Hummel J, Isaac M,Brogan MW. A randomized, controlled trial ofan automated wireless messaging system fordiabetes. Diabetes Technol Ther 2005;7:710–718; discussion 719–72019. Lim S, Kang SM, Shin H, et al. Improvedglycemic control without hypoglycemia in el-derly diabetic patients using the ubiquitoushealthcare service, a new medical informationsystem. Diabetes Care 2011;34:308–31320. Quinn CC, Shardell MD, Terrin ML, Barr EA,Ballew SH, Gruber-Baldini AL. Cluster-randomizedtrial of a mobile phone personalized behavioralintervention for blood glucose control. DiabetesCare 2011;34:1934–194221. Wise PH, Dowlatshahi DC, Farrant S,Fromson S, Meadows KA. Effect of computer-based learning on diabetes knowledge and con-trol. Diabetes Care 1986;9:504–50822. Yoo HJ, Park MS, Kim TN, et al. A UbiquitousChronic Disease Care system using cellular phonesand the internet. Diabet Med 2009;26:628–63523. GlasgowRE, Kurz D, King D, et al. Outcomes ofminimal and moderate support versions of aninternet-based diabetes self-management supportprogram. J Gen Intern Med 2010;25:1315–132224. Glasgow RE, Boles SM, McKay HG, Feil EG,BarreraM Jr. TheD-Netdiabetes self-managementprogram: long-term implementation, outcomes,and generalization results. Prev Med 2003;36:410–419

25. Zhou YD, Gu W. [Computer assisted nutri-tion therapy for patients with type 2 diabetes].Zhejiang Da Xue Xue Bao Yi Xue Ban 2003;32:244–24826. Lo R, Lo B, Wells E, Chard M, Hathaway J.The development and evaluation of a computer-aided diabetes education program. Aust J AdvNurs 1996;13:19–2727. Minet L, Møller S, Vach W, Wagner L,Henriksen JE. Mediating the effect of self-caremanagement intervention in type 2 diabetes:a meta-analysis of 47 randomised controlled tri-als. Patient Educ Couns 2010;80:29–4128. Jackson CL, Bolen S, Brancati FL, Batts-Turner ML, Gary TL. A systematic review of inter-active computer-assisted technology in diabetescare. Interactive information technology in diabe-tes care. J Gen Intern Med 2006;21:105–11029. Liang X, Wang Q, Yang X, et al. Effect ofmobile phone intervention for diabetes on gly-caemic control: a meta-analysis. Diabet Med2011;28:455–46330. Michie S, Abraham C, Whittington C,McAteer J, Gupta S. Effective techniques inhealthy eating and physical activity interven-tions: a meta-regression. Health Psychol 2009;28:690–70131. Michie S, Fixsen D, Grimshaw JM, EcclesMP. Specifying and reporting complex behav-iour change interventions: the need for a scien-tific method. Implement Sci 2009;4:4032. Carver CS, Scheier MF. Control theory:a useful conceptual framework for personality-social, clinical, and health psychology. PsycholBull 1982;92:111–13533. Cleveringa FG, Gorter KJ, van den Donk M,van Gijsel J, Rutten GE. Computerized decisionsupport systems in primary care for type 2 di-abetes patients only improve patients’ out-comes when combined with feedback onperformance and case management: a system-atic review. Diabetes Technol Ther 2013;15:180–19234. Roshanov PS, Fernandes N, Wilczynski JM,et al. Features of effective computerised clinicaldecision support systems: meta-regression of162 randomised trials. BMJ 2013;346:f657

1766 Computer-Based Self-management for Adults With T2D Diabetes Care Volume 37, June 2014