Secure Messaging and Health Outcomes

Nine hypothesis testing papers evaluated health outcomes and many focused on health conditions pertinent to VA.^{1,2,3,4,5,6,7,8,9} Seven of these studies were randomized controlled trials (RCTs).^{5,6,8,7,1,4,9}

Four papers focused on SM and diabetes or related conditions.^2,3,5,6 In the article by Zhou,² the investigators examined the relationship between SM and performance on standard HEDIS measures for diabetes and hypertension among Southern California Kaiser Permanente patients with these conditions. In regression analysis, patients with diabetes and/or hypertension who participated in SM were statistically significantly more likely to see improvements from baseline in all of the HEDIS measures (diabetes process and intermediate outcome measures and blood pressure control). Two or more emails in a two-month period (as compared to only one) were associated with better performance on some measures (“dose-response effect”). In a separate matched cohort analysis, matching patients who used SM with those who did not, email use was associated with improvement in most of the diabetes measures but not in blood pressure control in patients with hypertension.

Another study from Group Health Cooperative was a cross-sectional analysis evaluating outcomes for patients with diabetes who had completed the process of identity verification required to participate in SM, comparing those who used the service to those who did not.³ They adjusted for a number of patient characteristics including medical comorbidity and depression severity, as well as provider characteristics such as age, sex and specialty. In terms of quality, they found that the rate of A1c<7% was 36% higher in patients with the highest rate of SM use (>= 12 threads/year); though SM was not associated with better blood pressure control and was modestly associated with better lipid control. This analysis could not ascertain whether the utilization of SM was responsible for the improved outcomes as patients who used SM more may simply have been more engaged in their health. Interestingly, primary care visit rates in this study were 32% higher in patients with high use of SM (three to four additional outpatient visits per year).

Two RCTs evaluated web-based educational programs for diabetes that included SM as a component of the system. One RCT evaluated Veteran patients with diabetes who were randomized to usual care versus web-based care management that included receipt of a notebook computer, a blood pressure monitor, a glucometer, free internet service and access to a diabetes care management website.⁵ The intervention patients were able to send and receive secure messages to and from their care manager. They also had access to educational links and web-based diabetes resources. While hemoglobin A1c levels decreased over time in both groups, the intervention group demonstrated a greater decline over time (decrease of 1.6% versus 1.2% in usual care group). Blood pressure declines were also greater in the intervention patients with hypertension (−10 versus −7 mm Hg) while there were no differences in lipid levels. In the second study from the University of Washington,⁶ patients were similarly randomized to usual care versus web-based and in-person care management with access to SM functionality. Hemoglobin A1c levels were 0.7% lower in the intervention group compared to usual care when evaluated nine to 12 months after the start of the study (95% CI 0.2–1.3, p=0.01). There were no differences in blood pressure or lipid control between the groups.

Another study from Group Health Cooperative evaluated the impact of managing depression care using online secure messaging available through a patient website linked to the medical record.⁹ 48% of patients with new prescription for an antidepressant and who were registered for online messaging consented to participate and were randomized to usual care with their prescribing physician versus usual care augmented by three planned contacts via messaging with a mental health nurse care manager, which included structured assessments and nurse communication with the patients’ providers. Intervention patients had lower depression severity at the end of the evaluation period and higher proportions experienced 50% or greater decreases in the depression scale of the Hopkins Symptom Checklist (55% v 41%, OR=1.8, 95% CI 1.0 to 3.1). Utilization rates for visits, telephone encounters and non-study online messages did not differ between the groups.

Four additional studies evaluated web-based systems that allowed patient access to medical information and SM but that were not focused solely on SM. ^7,8,1,4 The two most relevant to VA included a study from Group Health and one from the University of Colorado.^7,8 The Group Health study randomized patients with hypertension to usual care versus one of two intervention groups. All participants were given access to Group Health’s patient website and all were encouraged to work with their providers to improve their blood pressure control. One intervention group was given a home BP monitor and training on the use of the patient website while the other intervention group was also given access to a pharmacist care manager who initiated case management via SM and face-to-face visits.⁷ Only the pharmacy case management group demonstrated significant improvements in BP control with 25% more patients with controlled BP compared to usual care (mean decrease −8.9 mm Hg) and 20% more compared to the other intervention group (mean decrease −6.0 mm Hg). The second study randomized patients with congestive heart failure (CHF) to usual care versus access to a web-based medical record system that also included an educational guide and a messaging system.⁸ There were no significant differences between the intervention and usual care groups in the primary outcome of self-efficacy nor in measures of health status. General adherence to medical advice was better in the intervention group (2.3 points higher on a 0–100 scale) but there was no difference in medication adherence; and no difference overall in patient satisfaction. While there were no differences in mortality, hospitalizations or clinic visits, there was a significant increase in emergency room visits in the intervention group. In this study, email messages did not seem to replace telephone calls and particularly early in the intervention, there were more messages (phone and email combined) from intervention patients than phone messages from the control group. The other studies included one focused on Danish and Irish patients with ulcerative colitis who were randomized to a web-based educational and care management program with email and text messaging capabilities versus usual care and found improved adherence to acute care, improved knowledge and quality of life and fewer clinic visits in the intervention group;¹ a second study that randomized couples undergoing in-vitro fertilization (IVF) in the Netherlands to usual care versus access to an internet-based personal health record with email capability and found no differences in patient empowerment (e.g. self-efficacy, knowledge) or secondary outcomes such as patient satisfaction, change in anxiety or change in depression.⁴

Secure Messaging and Patient Satisfaction

Seven hypothesis testing studies evaluated patient satisfaction and SM,^{10,11,8,4,12,13,14} four of which were RCTs or controlled clinical trials.^10,11,8,4 The first RCT involved an academic internal medicine clinic affiliated with the University of Colorado in Denver which was using an electronic medical record (EMR) system.¹⁰ Six hundred and six patients with experience using an internet browser were randomized to have access to a patient portal with SM that allowed requests for appointments, medication refills, and clinical communication with the provider, or usual telephone care. Patient satisfaction was assessed at baseline and at six months and was significantly higher in the intervention group in the areas of communication with the clinic (rated better by 44% of portal group versus 11% of controls), overall rating of clinic services (rated very good or excellent by 59% of portal group versus 48% of controls), and ratings of non-urgent messaging, prescription refills, referral requests and appointment scheduling (generally 20% improvement in portal group compared to controls).

The second RCT involved physicians and patients affiliated with a Penn State faculty-based clinic.¹¹ Four of eight participating physicians were assigned to offer email communication to their patients. Patients of all eight physicians were asked about their home email use and those of the participating physicians who were regular home email users were offered the opportunity to participate in the study and if consented, were given access to a secure email account for a six month period. Patients of the other four physicians continued with regular telephone contact with the clinic. All patients were surveyed at baseline, three months and six months. For the 100 patients analyzed, 67 in the email group and 35 in the control (the study excluded 72 patients who did not respond to all of the questionnaires), patient satisfaction was significantly higher in the email group in the areas of convenience of communicating with their physician and the amount of time spent contacting their physician. One hundred percent of patients in the intervention group and 97% of patients in the control group felt that patients should be able to email their doctors. Providers were not as satisfied as patients though the number of physicians involved in the study (eight) was not sufficient to draw statistical conclusions. There were trends toward increased satisfaction among intervention physicians in the areas of convenience, time spent and volume of messages.

The third randomized trial from the University of Colorado was discussed in the previous section as well and randomized patients with CHF to usual care versus access to a web-based medical record system that also included an educational guide and a messaging system.⁸ As noted, this study did not demonstrate significant differences in patient satisfaction.

The final RCT has also been previously described and randomized couples undergoing in-vitro fertilization (IVF) in the Netherlands to usual care versus access to an internet-based personal health record with email capability and found no differences in patient satisfaction.⁴

Of the three remaining studies evaluating SM and patient satisfaction, two were cross-sectional survey studies, and one was a case-control study of implementation that included a patient survey. In the first, the investigators surveyed patients who use the internet for health information regarding their email practices with physicians.¹² Of the 1881 individuals who completed the survey, 17% reported electronic communication with their physicians or physician staff, and satisfaction was highest in those who reported that their physician always responded (OR 15.9) compared to those who reported responses less than 100% of the time. The two remaining studies evaluated a web-based SM system being implemented at a clinic affiliated with the University of California Davis Health System.^13,14 In the first paper,¹⁴ eight providers, their patients, and their staff were surveyed about the system. Two hundred thirty eight of 645 patients with email addresses (36.9%) responded to the survey with 97% having used the system at least once (49.6% used it only once or twice). Satisfaction was related to receiving a response right away or by the next business day. The second study evaluated patient satisfaction one year later and surveyed 5971 patients from 4320 email addresses with a response rate of 40.3% (1743/4320).¹³ Their findings again associated higher satisfaction with more timely responses.

Secure Messaging and Adherence

There were two papers of clinical trials that evaluated SM and adherence, both of which were randomized.^15,8 One specifically evaluated SM reminders and one evaluated an online medical record with SM as one component. In a study from Northwest Permanente in Portland, OR,¹⁵ investigators with an established EMR randomized 2100 patients with SM accounts who were due for colorectal cancer screening to usual care or to receive a letter reminder or a secure email reminder to pick up a fecal occult blood test at the lab. The cohort was subsequently reduced to 1409 patients because of logistical issues with performing the intervention leaving 494 in the usual care group, 458 in the letter group and 457 in the SM group. The two intervention groups were significantly more likely to undergo screening than the usual care group, and at similar rates (7.8% usual care, 23.6% letter, 22.7% email). They concluded that email was as effective a prompt as a letter. The final an online medical record for CHF from the University of Colorado which has been previously described, assessed adherence to medications and general adherence to medical regimens and found that the latter was significantly improved in the intervention group users versus controls.⁸

Secure Messaging and Efficiency/Utilization

There were 18 studies that evaluated SM and some aspect of efficiency or healthcare utilization.^{1,16,17,3,18,19,20,21,22,14,23,6,24,25,26,8,13,7} Six studies were randomized controlled trials,^{1,20,6,26,8,7} ten were observational analyses,^{16,17,3,19,21,22,14,23,24,13} and two were survey analyses.^18,25 Four studies evaluated provider productivity and provider attitudes,^13,22,14,26 seven studies evaluated characteristics of users of SM systems,^{18,16,25,24,21,17,19} and eight studies evaluated the impact of SM on utilization of healthcare services.^{24,23,3,8,6,7,20,1}

There were four papers that evaluated SM and physician productivity.^13,22,14,26 Three papers described results of the same intervention with slightly different types of analyses of physician productivity.^13,22,14 These papers evaluated a secure web-messaging system being implemented at a clinic affiliated with the University of California Davis Health System that was described in the patient satisfaction section. The first was a pre-post study that found that intervention providers had higher overall averages of visits/day for the 5- month study period compared to the same period in the previous year (increasing from 24.8 to 29.5 visits/day, p=0.02).¹³ The second study compared the intervention providers with providers at a control clinic site before and after the intervention.²² Again, intervention physicians averaged 11% more visits/day than those in the control group (25.5 vs. 22.9) and averaged 10% more RVUs/day (relative value units) than the controls, though RVUs per visit did not differ. This was viewed as a positive finding for this fee-for-service environment. The third study retrospectively compared SM and telephone volumes for the intervention clinic and a control clinic.¹⁴ Case call volume averaged 18.2% less than control (21.6 vs. 26.4, p=0.002) and case message (phone plus web) volume averaged 13.7% less than control (phone only) (22.8 vs. 26.4, p=0.025). Patient enrollment rose rapidly during this time, presumably accounting for the decrease in contacts/1000 panel patients. The fourth paper on physician productivity evaluated the introduction of a web-based communication tool for patients in a University of Michigan academic primary care practice and found that physicians in the intervention group (whose patients were given access to the web-messaging tool) did not experience a decrease in the volume of personal email and telephone messages, which remained similar to that of the control physicians despite the addition of a modest number of web-based messages.²⁶

Two studies from Group Health Cooperative evaluated the characteristics of patients who used SM.^18,16 The first compared these patients to those who had registered for access to the patient web site but had not utilized SM. SM users were more likely to be middle aged (50–65 years old), female, and with a higher overall level of morbidity.¹⁸ They were less likely to be insured by Medicaid. As would be expected, patients treated by providers with higher levels of SM use and with a higher proportion of younger patients on their panels were also more likely to use SM. The second study focused on patients with diabetes in Group Health and evaluated predictors of initial use of SM when it became available in 2003, and then subsequent use.¹⁶ Initial use was associated with younger age, male sex, living in a higher socioeconomic status neighborhood and greater provider level of SM use. Older patients were more likely to subsequently sign up for SM if they switched to a high-use physician. Increasing morbidity also was associated with subsequent SM use. A third study similarly evaluated characteristics of patients who used the SM application associated with “PAMFOnline,” a personal health record system operated by the Palo Alto Medical Foundation.²⁵ Users had opted to pay an annual fee of $60 for unlimited messaging ability in addition to the general functionalities of the PHR. Compared to the general PAMF patient population, PAMFOnline messaging users were older, had more active problems, took more medications and used PAMF to conduct more office visits. In a fourth study from Kaiser Permanente Northwest that compared SM users to non-users, users were older and more likely to have diabetes.²⁴

The three remaining studies in this area included one from a single surgical practice that found that email users were significantly younger than non-users but found no sex differences in use;²¹ a study from Europe that identified young (<25), higher educated and urban patients as most likely to consider the availability of email access when choosing a new doctor;¹⁷ and a study of a Florida physician survey which found that physicians in large group practices (50 or more physicians) were more likely, and Asian-American physicians were less likely to communicate with patients via email.¹⁹

Data regarding healthcare utilization and SM differed among the various studies. There were two studies evaluating implementation of SM at two different Kaiser Permanente regions.^24,23 The first involved patients from KP Northwest and evaluated utilization in a pre-post fashion for a retrospective cohort of members registered for KP HealthConnect Online (the SM application) for longer than 13 months and compared utilization for a sample of this cohort to a group of non-registered patients matched for age, sex, chronic conditions and primary care physician.²⁴ In the pre-post cohort analysis, annual primary care office visit rates decreased significantly from 2.47 to 2.24 visits per member per year. Similarly, SM user visit rates decreased more than the visit rate for non-users (10.3% vs. 3.7%, (p<0.003). The second study evaluated the impact of KP HealthConnect implementation in Kaiser Permanente Hawaii and compared healthcare utilization post-implementation (2007) to that of 2004.²³ This study found significant decreases in age/sex-adjusted total office visits per member (decrease of 26.2%) and significant increases in total scheduled telephone visits per member. Total contacts including secure messages and office and telephone visits increased significantly from 5.18 contacts per member per year to 5.61. A third study from Group Health Cooperative that has been previously described compared patients with diabetes who had completed the process of identity verification required to participate in SM to those who did not.³ Controlling for a number of variables including overall medical comorbidity and diabetes severity, primary care visit rates were statistically significantly higher in the SM users (32% higher), as were specialty visits and emergency visits. These results indicate that data is conflicting regarding whether SM results in increased or decreased visit rates and there may be unmeasured factors in these studies related to the system, the patients or the healthcare environment which contribute to the findings.

Three studies of online medical records with SM capability also looked at utilization.^8,6,7 One study involving patients with CHF that has been previously described that randomized patients to access to the system found that intervention patients had significant increases in overall emergency department visits but no differences in mortality, hospitalizations or heart failure clinic visits.⁸ Another study in patients with diabetes randomized to access to an online care management application that has also been previously described found no differences in primary care or specialty care visits or inpatient days though there was additional time spent by the care managers in the intervention group.⁶ The third involved patients with hypertension randomized to usual care, access to the web-based application and access to the web-based application with pharmacy case management, and has also been previously described.⁷ This study found no differences in primary care visits, inpatient or urgent care/emergency use at one year. Telephone encounters were higher in the pharmacy arm than the other two though these patients had a modest but statistically significant decrease in specialty office visits.

Finally, two European studies found that patients randomized to SM with their general practitioner had fewer office visits than those without access, though there were no differences in the number of telephone consultations between groups,²⁰ and patients with ulcerative colitis who had access to a web-based care management program with messaging had fewer routine visits than control patients.¹

Summary for Key Question #1

The hypothesis testing studies that include secure messaging support the following conclusions:

• There is moderate strength evidence that secure messaging (especially as part of a web-based management system) can improve glucose outcomes in patients with diabetes. (GRADE = moderate)
• There is low strength evidence that adding a web-based pharmacist to secure messaging improves blood pressure outcomes in patients with hypertension. (GRADE = low)
• There is low strength evidence that secure messaging with a web-based management system improves ulcerative colitis symptoms. (GRADE = low)
• There is moderate strength evidence that secure messaging increases patient satisfaction. (GRADE = moderate)
• There is low strength evidence that secure messaging improves adherence to colorectal cancer screenings or heart failure management. (GRADE = low)
• There is low strength evidence that secure messaging influences either positively or negatively efficiency or utilization. (GRADE = low)

Patient Access to Medical Records through Tethered Systems

Of the articles evaluating patient access to their own medical record through a system tethered to a provider-based electronic record, seven were RCTs,^{8,6,27,4,7,28,29} four were observational hypothesis testing studies,^16,30,31,32 and nine were descriptive quantitative studies.^{33,34,35,36,37,38,39,40,41} Only one study was a cost-benefit analysis study.⁴²

The systems described in these studies included a number of functionalities as part of their intervention. These are listed in Table 1. As can be seen, almost all systems included secure messaging, access to test results and progress notes/visit summaries, and education/disease management, meaning the effects cannot be disentangled and these four functionalities represent a “core” set for patient portals.

Table 1

Tethered Personal Health Record Systems and Functionalities.

Summary for Key Question #2 – Tethered Systems

The evidence reviewed regarding patient access to their own medical records, when this is tethered to an electronic health record, supports the following conclusions:

• There is low strength evidence that patients’ access to their own medical record improves health outcomes for chronic diseases such as diabetes, hypertension, or heart failure (GRADE = low).
• There is low strength of evidence that patient access to their own records, combined with the ability to submit an electronic care plan can improve medication decisions at a subsequent visit for patients with diabetes (GRADE = low).
• There is insufficient evidence to reach conclusions regarding the potential effect of patients’ access to their own medical record on efficiency (GRADE = insufficient).
• There is low strength evidence that patients’ access to their own medical record influences utilization (GRADE = low).
• There is insufficient evidence to reach conclusions regarding the potential effect on patients’ access to their own medical record on patient attitudes.
• There is moderate strength evidence that use of patient portals is lower among certain racial or ethnic groups, in particular African Americans (GRADE = moderate).
• There is moderate strength of evidence that use of patient portals is lower in persons with lower literacy or education level (GRADE = moderate).

Patient Access to Medical Records: Non-Tethered Systems

There were five papers focusing on patient access to their own medical records using systems that were not necessarily tethered to a provider-based EHR.^{53,54,55,56,57} All were survey studies and four were from Europe or Canada. The one US study from the University of Washington evaluated the use of a web-based personal health record by disadvantaged people living at a low-income housing facility.⁵⁶ The study utilized nursing students to help residents utilize the system and found that utilization was much higher on the days that the students were available or on the day after. 12 of 24 registered users completed a survey. Only two were able to complete the health record entry on their own, though 92% were satisfied with the system and most felt the system helped them manage their health information. A Swedish study evaluated patient attitudes towards a national pharmacy register using a web survey.⁵³ Of 7860 unique site visitors, 22% completed the survey. Respondents generally rated the system highly and felt that it would help with medication safety and healthcare, particularly that given by the pharmacist. A study from Switzerland evaluated the attitudes of patients given “CardioCard,” a small CD-ROM containing cardiac medical information including study reports and reports of clinical examinations.⁵⁵ Seventy three percent of 536 patients who received a CardioCard responded to the survey. Seventy three percent found it very useful or useful and 78% were willing to pay for it. Most (73%) did not access the card themselves and 88% did not have their physicians access the card. Only 5% were worried about data security.

The final two studies surveyed patients on their attitudes regarding having access to online health information.^57,54 One from England invited 1050 patients to participate in a survey with a 60% response rate (629 responses).⁵⁷ Analyses were conducted for patients aged 21 and older. Most felt they should have access to their health records (83%) and 79% thought it was a good idea to make them electronic. 35% preferred print-outs of records and 50% had concerns about the security of their records held on computers. A study from Canada surveyed Chief Executive Officers of Canadian public and acute care hospitals to understand organizational readiness for patient accessible records.⁵⁴ Eighty-three of 213 CEOs responded (39% response rate). Fifty-four percent of hospitals had some sort of EHR. 87% of respondents identified hospital financial resources as the most important barrier to providing patient access to the EHR.

Data for the non-tethered systems confirms the generally positive attitudes from other survey studies of patients regarding their interest in having access to their medical records. The one study in a low income and elderly population identifies potential barriers related to computer and internet skills.