Objective: To address provider, payer, and patient concernsabout the use of online communication in healthcare settings byperforming a randomized controlled trial of a Web-based patient-providercommunication tool in primary care.
Subjects and Methods: Forty-one staff physicians and 91 residentsin 4 primary care centers were randomized to a Web-basedonline communication system. Patients of intervention physicianswere encouraged to communicate via the system about healthissues, scheduling, prescription renewals, referrals, and billing.Data collected included patient Web use, e-mail use, telephonecalls, visit distribution, and physician and patient attitudes towardand satisfaction with communication.
Results: One thousand thirty-eight patients sent 2238 messagesduring the 40-week study. Half of the messages were directly relatedto a patient's health; half were administrative. Patient Web usepeaked at 8.5 weekly messages per 100 scheduled visits. Patient emailand telephone volume remained similar across groups.Intervention physicians reported more positive attitudes towardWeb-based communication than control physicians (mean Webbenefits scale score, 4.0 vs 1.1; = .008), but there were nobetween-group differences in attitudes toward communication ingeneral. Patients and physicians reported differential preferencesfor the use of online communication based on problem complexityand sensitivity.
Conclusions: Web-based messaging was lower than expectedbecause of patient-related factors and limitations of the technology.Patients, physicians, and staff had positive attitudes toward onlinecommunication. There was no detectable difference in communicationvolume between study groups, but more sensitive measuresof work burden need to be developed and evaluated.
(Am J Manag Care. 2004;10:593-598)
Advances in technology have motivated interestin the role of online patient-provider communicationin clinical settings. Communicatingonline could reduce barriers associated with traditionalmodes of communication,1-4 yet it has diffused slowly inclinical practice.5 Lack of reimbursement for onlinecommunication has motivated questions about whetherit will reduce use of other clinic resources or will createadditional work for staff.6 Providers have voiced concernsabout appropriate uses of online communicationin clinical practice.7-9 Patients have been reluctant tocommunicate online because of lack of experience anduncertainty about provider responses.10 In this issue ofthe , Houston et al11 report as a barrier to use ofonline communication that adults who used the Web forhealth information frequently voiced concerns aboutbothering physicians. Finally, there is uncertaintyabout how to build and integrate Web-based communicationtools into existing operations and informationsystems.12-14
We built a Web-based patient-provider communicationsystem and performed a randomized controlledtrial of its use in an academic primary care setting. Thestudy addressed several questions: Does Web communicationbetween patients and the clinic team substitutefor personal e-mail or telephone calls? Does Web-basedcommunication affect patient or provider perceptions ofcommunication? We hypothesized that a Web-basedcommunication tool would reduce e-mail and telephonecalls and improve communication between patients andtheir providers.
A randomized controlled trial was performed in 4university-affiliated primary care clinics for a 40-weekperiod from September 2001 through June 2002. Threefaculty physicians and 2 residents declined to participate.Forty-one faculty physicians (averaging 37 scheduledvisits per week) and 91 resident physicians(averaging 6 scheduled visits per week) in general medicineand family practice were randomized into 2groups, intervention (n = 65) and control (n = 67). Thepatients of intervention physicians were encouraged touse a secure Web-based tool to communicate with theirclinic staff about health and administrative issues andprescription renewals. The Web site had educationalcontent addressing appropriate message content,expected response times, and message handling by clinic staff. Incoming messages were routed to appropriateclinic staff, who entered the site through a secure log-in.Patients were prompted through regular e-mail to enterthe Web site to read responses from the staff. Physiciansdid not have access to the Web system, and staff contactedphysicians for pertinent messages through theusual means. Patients of intervention physicians wereencouraged to use the Web system through promotion,including cards distributed by intervention physiciansand brochures mailed to patient homes.
Information about the number and types of Web messageswas collected from the Web site. E-mail volumewas measured based on physician recall of the numberof e-mail messages received directly from patients duringthe previous week. On average, 91.7% of staff physiciansresponded vs 67.4% of residents. Missing estimates forresidents were imputed to zero because feedback suggestedthat these residents had low e-mail use withpatients. Staff logs were used to collect data on telephonecall volume by type of call and physician.
Self-administered patient and physician surveys conductedat the end of the study assessed attitudes towardWeb and e-mail communication, preferences for differentmodes of communication, and satisfaction withcommunication.9,10 Because there was no valid physician-patient roster, we selected patients in 2 equallydivided groups: 425 patients who had seen an interventionphysician 1 or more times and a control physicianno more than 1 time during the study period, and 425patients who had seen a control physician 1 or moretimes and an intervention physician no more than 1time during the study period. The Dillman method wasused to maximize response rates.15
Variables and Analysis
Three use variables were constructed at the physicianlevel. These included (1) weekly patient Webmessages per 100 scheduled visits (number of patientWeb messages per week divided by the mean numberof scheduled visits per week during the study period Ã—100), (2) weekly patient e-mails per 100 scheduled visits(number of physician-reported patient e-mails perweek divided by the mean number of scheduled visitsper week during the study period Ã— 100), and (3) weeklytelephone calls per 100 scheduled visits.
Physician survey variables included a "Web benefits"scale that indicated attitudes toward using the Web withpatients, with higher scores indicating more favorableattitudes (4 items, α = .88), and a general communicationscale indicating attitudes toward communicationwith patients and staff (4 items, α = .82). Finally, weexamined preferences for different modes of communicationfor different clinical issues.
Variables from the patient survey included the following:(1) A Web benefits scale that indicated attitudestoward using Web communication with healthcareproviders, with higher scores indicating more favorableattitudes (3 items, α = .90). (2) Preferences for differentmodes of communication using questions similar tothose on the physician survey. (3) A general communicationscale indicating attitudes toward communicationwith and access to physicians and staff outside of visits,with higher scores indicating more favorable attitudes(4 items, α = .85).
Poisson regression analysis was used to examine differencesin trends in counts of the different use measuresacross the 5 periods between the physician studygroups, with physician as the unit of analysis. For example,in one model, telephone volume was the dependentvariable, and independent variables included group(intervention vs control), periods (1 through 5), groupand period dummy variables, physician status (residentvs faculty physician), and dummy variables indicatingstudy clinic site. Differences in physician and patientattitude scale scores were examined using ordinaryleast squares regression analysis. The dependent variableswere the various scales specified as intervals.Independent variables for the physician scale modelsincluded study group, physician type (resident vs facultyphysician), and clinic. Independent variables for thepatient scales models included study group, age, sex,and clinic. The study was approved by the University ofMichigan Institutional Review Board.
Baseline Characteristics of the Physician Sample
Intervention and control physician groups did notdiffer with regard to the distribution of faculty physicians(27.4% in the intervention group vs 32.3% inthe control group, = .27), female sex (40.1% vs 41.2%, = .59), or mean number of scheduled visits per week.
One thousand thirty-eight patients registered onthe Web site and sent 2238 messages (41.8% ofpatients sent no messages, 19.4% sent one, 16.3% sent2-3, and 22.5% sent ≥ 4 messages). The growth ofpatient registration and messaging was uniform duringthe study period. About half of the messages wererelated to health-related follow-up, inquiries, or requestsfor test results. Approximately 20% wereappointment related, 15% were referral requests, and12% were prescription requests or renewals.
Table 1 gives Web, e-mail, and telephonevolume for 5 one-week data collectionperiods during the studyamong the intervention and controlgroup physicians. The trend in Webmessage use (the mean number ofweekly Web messages per 100 scheduledvisits) was modest, approachingthat of e-mail use for the interventionphysician group. E-mail and telephonevolume trends (the mean number ofweekly messages reported in the priorweek per 100 scheduled visits) weresimilar between study groups (Waldtest, 8.1; = .09 for e-mail; and Waldtest, 9.1; = .18 for telephone; controllingfor physician type and clinic).
The response rate to the survey was 71.2% (37 facultyphysicians and 57 residents). Table 2 gives perceivedbenefits of Web communication with patients.Intervention physicians perceived greater benefits ofWeb communication than control physicians across allitems. For example, intervention physicians weremore likely than control physicians to believe thatWeb communication would be "a good way for patientsto contact me" (56.3% vs 41.3%, = .04). The meanscore for the Web benefits scale was 4.0 in the interventiongroup vs 1.1 in the control group ( = .008, controllingfor physician type and clinic).
Figure 1 shows physician preferences for differentmodes of communication by health issue. Physicianstended to prefer face-to-face visits as health issuesincreased in complexity or sensitivity. For example, preferencefor visits was low for simpler issues such as a cholesteroltest results and normalPap smear or prostate-specificantigen test results (11.7% and12.7%, respectively) comparedwith more complex issuessuch as back pain (50.2%) ormore sensitive issues such asprolonged sadness or anxiety(84.7%). By contrast, onlinemessaging decreased markedlyrelative to other modes asissues became more complexand sensitive. Preference fortelephone communication washighest for issues that weremoderately complex (such asback pain and abnormal Papsmear or prostate-specific antigen test results), but waslower for simple issues (such as cholesterol test results)or the most complex and sensitive issues (such as breastor testicular pain and prolonged sadness or anxiety).Physicians in the intervention group endorsed onlinemessaging with patients at higher rates than controlphysicians only for the less complex or sensitive issues,including normal test results and symptoms such as sorethroat or back pain (data not shown).
Physician satisfaction with communication withpatients and staff was similar among intervention andcontrol physicians. More than three quarters of all physicianswere satisfied with their communication and relationshipwith patients, as well as communication withstaff. However, fewer physicians in both groups were satisfiedwith patient communication outside of clinic visits(41.7% in the intervention group vs 47.9% in the controlgroup, = .35). General communication scale scores didnot differ between the 2 groups (mean score, 1.6 in theintervention group vs 1.9 in the control group; = .58).
The overall patient survey response rate was 62.5%(531/850). The mean age was 49.8 years, two thirdswere female, and 16.5% were nonwhite. There were nosignificant differences between respondents and nonrespondents.Only 16.2% had a high school education orless, 30.2% had some college, and nearly one half had acollege degree or higher. (Percentages do not total 100%due to missing responses.) Two thirds reported using emailat least daily. About 80% of patients reported regularWeb use, but only half reported using the Web toshop for items or to bank online.
Attitudes toward electronic communication weresomewhat more favorable for intervention vs controlgroup patients, but differences were small. For example,61.9% of intervention patients vs 54.2% of controlpatients agreed or strongly agreed that they would likeusing the Web to communicate with providers ( = .15).Differences in Web benefits scale scores were small (2.5in the intervention group vs 2.1 in the control group, = .39). Figure 2 shows differences in patient preferencesfor various modes of communication about selectedhealth topics. Some percentages within health topicssum to more than 100% because respondents couldendorse more than 1 mode. The pattern of patient preferencesfor visits was similar to that of the physicians.Patients preferred visits for more complex or sensitiveissues and preferred online communication for lesscomplex or sensitive issues. However, compared withtheir physicians, the proportion of patients whoendorsed online communication was greater across allhealth topics. There were no significant between-studygroup differences in thesemeasures.
The role of clinic staff inelectronic communication isone area in which patientsand physicians do not agree:43.5% of patients disagreedthat "it is OK for staff to handlemy e-mail messages" vs13.7% of physicians ( < .01).The results for Web messagingwere similar: 33.5% ofpatients disagreed that "it isOK for staff to handle my Webmessages" vs 11.7% of physicians( < .01). These figuresdid not differ significantlybetween study groups. Only44.4% of study group patientsand 47.4% of control grouppatients agreed that they havesufficient access to physicians or clinic staff outside ofvisits. General communication scale scores were similarbetween groups (1.5 in the intervention group vs 1.2 inthe control group, = .34).
We performed a randomized controlled trial of aWeb-based patient-provider communication tool toevaluate its effect on other modes of communicationand patient and provider satisfaction. We found thatthe Web-based tool increased online communicationvolume, but only modestly. At its peak, Web communicationvolume approached that of personal physician email,but it was far lower than telephone volume. Resultssuggested that only about 10% of patients who were regularusers of the Internet registered on our online systemduring the study period. Personal patient e-mailvolume and telephone volume were similar for the studygroups. Therefore, the Web messaging did not appearto offset e-mail and telephone volume.
The lower-than-expected amount of Web messagingreflected several factors relevant to a start-up phase ofonline communication in clinical practice. First, wefound that many patients did not have sufficient Web-basedexperience to navigate the intervention Web site.National surveys show that many adults are not regularWeb users, and there are large age, education, and ethnicdisparities in use patterns.16,17 The report by Houston etal11 in this issue of the showed that, even amongadults who sought healthcare information on theInternet, fewer than one fifth reported e-mailing theirphysician, and there were significanteducation and ethnicitygradients in use patterns.Second, the Web interface wasaimed at clinic staff, as wecould not connect physiciansdirectly. Some patients mayhave resisted using the Webtool, because many patientsindicated that they did notfavor staff as intermediaries inonline communication withtheir physicians. Finally,patient and staff feedback suggestedthat the features of theWeb interface we developedhad some limitations that mayhave reduced its appeal overregular e-mail and telephone.
Our intervention positivelyinfluenced physicians' attitudestoward online communication. Interventionphysicians reported perceiving greater benefits of Webuse with patients, especially for simpler issues andrequests such as normal laboratory results. This suggeststhat Web-based communication systems can bedeployed in a way that improves physician and staff attitudestoward this mode of communication. Yet, physicians'attitudes toward general communication withpatients and staff did not appear to be affected by ourintervention. In particular, many physicians reportedbeing dissatisfied with the quality of communicationwith patients outside of visits—the type of communicationthat may be most subject to improvement via electronicinterventions. This may reflect the fact that thetelephone remains the primary mode of communicationbetween visits, especially for patients requiring themost between-visit contact. The intervention did notaffect patient attitudes toward communication in general,and patients reported dissatisfaction with between-visitcommunication with their providers.
Our results suggest that patients and physicians differin their preferences regarding modes of communication.Physicians appeared to "titrate" their preferences formode based on the complexity and sensitivity of theissues. Physicians favored visits for the most complex orsensitive health issues, while favoring online communicationfor simple requests. Physician preferences fortelephone communication appeared to bridge theseother 2 modes of communication, as physicians' telephonecommunication preferences were highest forhealth issues that were neither too simple nor too complexor sensitive. Compared with physicians, patientsgenerally favored online communication over visits evenfor more complex or sensitive health issues. This likelyreflects the increasing desire to avoid the hassles andcosts of visits and telephone calls using technology thathas now become a common mode for many people. Italso may reflect what Houston et al11 found—that electroniccommunication emboldens patients to ask questionsthey might not feel comfortable asking in person oron the telephone.
There were also important differences betweenphysicians and patients with regard to preferences forthe organization of online communication, as physiciansfavored a role for intermediary "triage staff," whilepatients did not. Patients' expectations and desires for a"direct connect" to their physicians appeared to begreater for e-mail than for the Internet. This may reflectpatients' general experiences with these modes of communication,as e-mail is used to communicate directlywith 1 or more individuals, while Web communicationis used to surf for information or perform services,which are not generally aimed at a specific individual orgroup of individuals.18
Staff involved with responding to messages on theWeb system (mostly nurses) generally reported favorableattitudes toward online communication (data notshown) and were comfortable responding to online messages.They agreed that the Web messaging system wasa good way to communicate with patients, but thoughtthat it added to the daily workload. In general, theywere satisfied with getting online with patients andvoiced frustration with other modes of communicationsuch as the telephone.
Several limitations of this study deserve comment.First, university-affiliated primary care clinics wereused; therefore, results may not be generalizable toother settings. Second, limitations of the technology didnot allow us to connect physicians to the Web tool.Patient demand for online communication and the contentof the interactions may be different if physiciansand staff are connected. Third, modest patient use ofthe Web interface limited power to detect interventioneffects on e-mail and telephone volume and visit distribution.Post hoc power calculations showed that thestudy had 80% power to detect a difference of 12 weeklytelephone calls per 100 scheduled visits (52 vs 64)and an absolute difference of 5% in no-show rates (7% vs12%). Fourth, we did not measure factors that mayaffect resource use such as work flow associated withdifferent modes of communication. Finally, "contamination"among patients of control physicians may haveoccurred through promotion of the intervention withinclinics used by patients of intervention and controlphysicians. However, control patients could not accessthe intervention Web site, and e-mail use among controlphysicians remained low throughout the study period,suggesting contamination was not an issue.
Implications for Further Research
Many patients and providers are dissatisfied withnon—visit-related communication with physicians andtheir staff. Online communication has the potential toaddress this unmet need and improve the efficiency andeffectiveness of care delivery. Indeed, clinical practicesare beginning to make serious investments in onlinecommunication. Many providers believe that suchinvestments will pay off quickly by reducing telephonecalls and inappropriate visits and by increasing the efficiencyof work flow. However, several factors may limitcost savings such as unmet need for communication,insufficient integration, persistent mismatch betweencommunication task and mode, and patient selectionfactors. One important lesson from our research, whichmirrors the experience of others, is the potential forslow uptake of Web messaging during initial phases ofdeployment. However, advances in technology andincreased patient facility with Web-based tools shouldincrease early uptake among patients in new initiatives.
A second lesson from our research is that there is asubstantial "digital divide" between patients and theirphysicians with regard to appropriate content of messagingand the role of staff in the communication flow.This underscores the need for patient and provider promotionstrategies.19 Results from early demonstrationprojects suggest that patients will "get the message" andadhere to clinic guidelines about the "dos and don'ts" ofonline communication.20
Future research should focus on the effect of onlinecommunication on quality of care. Although randomizationmay not be possible in all settings, observationalstudy designs with appropriate control groups mayyield valuable information about such outcomes asadherence to preventive care, test result follow-up,medication compliance, and improved self-managementof chronic conditions. The prospects for online communicationbetween patients and their providers are brightas improvements in technology fuel strong underlyingdemand. Appropriate evaluation of online communicationwill provide the essential information to helpadvance the use of these new tools in clinical practiceand ultimately improve patient care.
From the Division of General Medicine, Department of Internal Medicine (SJK, NN),Department of Health Management and Policy, School of Public Health (SJK), and GlobalREACH (Research, Education, and Collaboration in Health) (CAM), University of Michigan;and the Veterans Affairs Ann Arbor Healthcare System (SJK); Ann Arbor, Mich.
This study was supported by a research grant from The Intel Corporation, Inc, SantaClara, Calif, to the University of Michigan.
Address correspondence to: Steven J. Katz, MD, MPH, Department of HealthManagement and Policy, School of Public Health, University of Michigan, 300 N Ingalls,Suite 7E12, Box 0429, Ann Arbor, MI 48109-0429. E-mail: email@example.com.
Int JMed Inf
1. Ball MJ, Lillis J. E-health: transforming the physician/patient relationship. . 2001;61: 1-10.
Health Aff (Millwood)
2. Starr P. Health care reform and the new economy. .2000;19(6):23-32.
Health Aff (Millwood)
3. Kassirer JP. Patients, physicians and the Internet. .2000;19(6):115-123.
J Fam Pract
4. Grover FJ, Wu H, Blanford C, Holcomb S, Tidler D. Computer-using patientswant Internet services from family physicians. . 2002;51:570-572.
J Gen Intern Med
5. Katz S, Moyer C. The emerging role of online communication between patientsand their providers. . 2004;19:978-983.
J Med Inf
6. Sittig DF, King S, Hazlehurst BL. A survey of patient-provider e-mail communication:what do patients think?. 2001;61:71-80.
Am J Manag Care
7. Moyer CA, Stern DT, Katz SJ, Fendrick AM. "We got mail": electronic communicationbetween physicians and patients. . 1999;5:1513-1522.
8. Kleiner KD, Akers R, Burke BL, Werner EJ. Parent and physician attitudesregarding electronic communication in pediatric practices. . 2002;109:740-744.
Am J Manag Care
9. Moyer CA, Stern DT, Dobias KS, Cox DT, Katz SJ. Bridging the electronicdivide: patient and provider perspectives on e-mail communication in primarycare. . 2002;8:427-433.
J Gen Intern Med
10. Katz SJ, Moyer CA, Cox DT, Stern DT. Effect of a triage-based e-mail systemon clinic resource use and patient and physician satisfaction in primary care: a randomizedcontrolled trial. . 2003;18:736-744.
Am J Manag Care
11. Houston T, Sands D, Jenskes M, Ford D. Experiences of patients who wereearly adopters of electronic communication with their physician: satisfaction, benefits,and concerns. . 2004;10:601-608.
Health Aff (Millwood)
12. Kleinke JD. Vaporware.com: the failed promise of the health care Internet.. 2000;19(6):57-71.
13. Baur C. Limiting factors on the transformative powers of e-mail in patient-physicianrelationships: a critical analysis. . 2000;12:239-259.
Health Aff (Millwood)
14. Goldsmith J. The Internet and managed care: a new wave of innovation.. 2000;19(6):42-56.
J Contin Educ Nurs
15. Anema MG, Brown BE. Increasing survey responses using the total designmethod. . 1995;26:109-114.
16. Cole J. Surveying the digital future: year three: UCLA Center for CommunicationPolicy: The UCLA Internet Report 2003. Available at: http://www.digitalcenter.org. Accessed August 31, 2003.
17. Madden M. The changing picture of who's online and what they do. America'sOnline Pursuits, Pew Internet & American Life Project 2003. Available at:http://www.pewinternet.org/pdfs/PIP_Online_Pursuits_Final.PDF. Accessed March20, 2003.
Am J Manag Care
18. Bundorf M, Singer S, Wagner T, Baker L. Consumers' use of the Internet forhealth insurance. . 2004;10:609-616.
J Am Med InformAssoc
19. Prady SL, Norris D, Lester JE, Hoch DB. Expanding the guidelines for electroniccommunication with patients: application to a specific tool. . 2001;8:344-348.
J Am MedInform Assoc
20. White CB, Moyer CA, Stern DT, Katz SJ. A content analysis of e-mail communicationbetween patients and their providers: patients get the message. . 2004;11:260-267.