Who are Portal Users vs. Early E-Visit Adopters? A Preliminary Analysis

Abstract

Patient portals and eVisits are gaining momentum due to increasing consumer demand for improved access to clinical information and services, availability of new technologies to deploy them and development of reimbursement initiatives by major payers. Despite increasing interest in online health consultation by consumers, adoption has been slow and little is known about the users of such services. In this study, we analyze the key features that distinguish early adopters of eVisits from portal consumers, in aggregate and in four distinct ambulatory practices, using data from a major healthcare provider in Western Pennsylvania. Preliminary results indicate that out of 10,532 portal users, the 336 patients who submitted 446 eVisits between April 1, 2009 and May 31, 2010 are younger on average, predominantly female, not retired, but in poorer health condition. They access the portal more frequently, indicating that they are potentially more involved in managing their health. Using fixed-effects logistic regression models to compare across practices, we note that practice indicator is a significant predictor of eVisit usage, perhaps due to the varying strategies used to build awareness and encourage adoption. Despite the small difference in out-of-pocket payment for eVisits covered by insurance vs. otherwise, insurance coverage for eVisits significantly contributes to increased usage. In ongoing work, additional characteristics of patients and practices that have access to the patient portal will be used to better delineate patients’ choice of eVisit vs. the traditional office visit.

Introduction

Patient health portals have become a critical component of a healthcare organization’s service delivery strategy, empowering patients to access their clinical information and interact with their healthcare team [1, 2]. Through patient portals, users have the ability to self-service and research their own health information and health issues. By providing them with access, they can review and validate portions of their medical record. Interactions with the office also become more user-friendly and efficient. Requests for prescription refills, appointments, medical advice such as appropriate medication use, and other related information can be received electronically, automatically routed to the correct resource, and managed in a timely fashion that integrates into workflow with minimal disruption to the patient or staff [3]. Perhaps one of the most valuable features of patient portals is the provisioning of services to treat patients for non-urgent health conditions via eVisit deployment [4, 5, 6]. Furthermore, this approach can evolve into a service that assists patients in managing chronic health conditions. By providing the tools to enter data such as blood glucose levels, weight, and blood pressure, and resources needed to monitor and control their health conditions over time, patients have an improved ability to actively participate in their healthcare and achieve more favorable health outcomes [2, 3].

A recent survey of nearly 5,300 patients by Forrester Research reported that health reform will necessitate online consultations between physicians and patients as more consumers seek access to doctors [7]. Patient eVisits are thus gaining momentum due to increasing consumer demand for improved access to clinical services, availability of new technologies to deploy such services and development of reimbursement initiatives by major payers [5, 6, 7]. The eVisit service provides patients with an online consultation through a series of structured, secure message exchanges with a physician using portal technology, providing an alternative for onsite office visits and non-reimbursed phone-based care, with the potential to increase the volume of patient access to providers [4, 5, 6]. Online consultation is a very important service to patients with internet access [8], has been shown to prevent an office visit for about 40% of the patients who signed up for the service [5], and the availability of this service decreased spending on the clinic visit [9]. Despite increasing interest in online health consultations by consumers, adoption has been slow and little is known about the users of such services other than that they are more educated and have higher household income compared to nonusers [7, 10]. A national interview survey of health IT use among 18–64 year old consumers confirmed that despite wide interest in accessing health information, women are overwhelmingly more likely than men to look up online health information, request prescription refills, seek medical advice and other interactions with their health provider [7]. As portals and eVisits become more widely available to patients and more increasingly sought, it is important to understand what are the key characteristics of actual consumers of online health consultations, in contrast to survey respondents, in order to design appropriate services and deliver the most satisfactory experiences for providers and patients alike.

In this study, we address this question by analyzing the deployment of structured eVists within a patient portal environment at four primary care practices associated with a major health system in Western Pennsylvania. We examine the key demographics of consumers accessing the portal and eVisit service, respectively, from April 1, 2009 to May 31, 2010. Using data from the four practices that include eVisit records, patient demographics, portal logon information, and diagnoses and medications, we estimate the likelihood of eVisit usage based on individual patient demographic and some clinical characteristics and their past utilization of online healthcare information on the portal. We further delineate the effects of varying practice environments and insurance availability using fixed-effects regression models.

The next section summarizes recent research on online healthcare delivery. The details of the eVisit process, technology and the specifics of the study site and data are addressed in the eVisit Service section. Preliminary results are provided in the Descriptive Analysis and Regression Result section. The final section outlines potential extensions incorporating additional characteristics about the practices, their participating vs. non-participating physicians, patient insurance details and healthcare utilization profiles, among others.

Background

Secure, online messaging between patients and providers is an increasingly desired communication channel for healthcare delivery for reasons of convenience, documentable communication, less chance of miscommunication or lack of connectivity compared to telephone, access to consolidated information, and less interruption to the other services since physicians can review the inquiry after the current task is completed and patients can wait for the reply while doing other activities [11]. However, challenges such as privacy issues, cost, unstructured documentation, unsettled reimbursement structure, physicians’ concerns about being overwhelmed by emails from patients, and lack of clear guidelines for successful implementation have resulted in a reluctance to adopt online messaging via emails [11, 12]. Furthermore, integration with the Electronic Health Record (EHR) is necessary to enhance the online communication process which poses greater challenges for smaller clinics. It is also difficult to place a charge on each email message [12]. Therefore, the adoption has been slow and only large academic medical centers have initiated pilot studies in recent years [2, 4, 5, 6].

Instead of email messaging, this study examines the use of a structured, secure, online consultation system developed within the health system’s web portal service and integrated with the EHR to better understand the profile of users of such services. This structured eVisit service has the potential to not only improve access to care similar to that of secure email messaging, but it can also enhance awareness of the specific health conditions treated via the service and improve self-health management [13]. This new channel structures questions about treated health conditions using evidence-based templates, and subsequent questions related to each condition follow logically so that patients can fill in the details without using free text form; free text is allowed if and when they want to provide critical additional information. Thus, while answering the questions, patients can become familiar with the relevant symptoms associated with their condition and the key questions to ask about it. Understanding the characteristics of the early adopters of eVisit service can potentially help organizations to better design its delivery to meet the healthcare and information needs of patients seeking online care.

Methods

Study Setting:

The study site is a major academic medical center in Western Pennsylvania which has developed an online health portal that allows patients to take a more active role in managing their own health by providing secure and convenient electronic access to their eRecord health information. The portal is integrated with the ambulatory electronic health record which allows the health care team to interact with patients through their current applications and workflow. The technology utilizes the underlying technical infrastructure and solutions offered by Epic Corporation (EpicCare EMR: http://www.epicsystems.com/software-clinical.php and MyChart Patient Portal: http://www.epicsystems.com/software-ehealth.php). The portal has been in use for more than five years, has over 19,000 current patient users, and is continuously growing. There are several options that are provided to portal users that offer a variety of services to patients. The clinical component of the portal provides patients with the ability to view lab results and diagnostic studies, solicit medical advice from their healthcare team (such as questions regarding a specific medication), request prescription refills, receive health maintenance reminders, and request or schedule medical appointments. The business component of the portal offers automated scheduling, registration and billing services that are standardized across the enterprise and integrated with the EHR. This integration offers patients self-service solutions for appointment scheduling, pre-registration to update select information like address and payer information, and correspondence with the business office.

In August, 2008, the medical center released structured eVisits as an additional function within the clinical component. The eVisits service provides patients with a questionnaire template-driven, online consultation through a series of secure message exchanges with a physician. The primary objective of this new online service was to provide an alternative for onsite office visits and reimbursable phone-based care. At the time of this study, 8 conditions were covered by eVisits: cough, red-eye, vaginitis, diarrhea, sore throat, urinary tract infection, and back pain as well as a generic category of “other.” A standardized template creates structured documentation of the consultation, is easy to use and integrated with practice workflow. It also captures information that is stored in the EHR. On April 1, 2009, the eVisit service was deployed in four primary care practices and reimbursed by some health plans. Patients incurred a $15 to $20 copayment when the service was covered by their insurance plan and a $30 charge otherwise, as clearly indicated on the portal site. Physicians received reimbursement for each eVisit irrespective of whether the service was covered by an insurer. The physicians and staff at the offices encouraged patients to sign up for portal access and use eVisits for treatment of the specified, episodic illnesses. Unlike other eVisit studies in the literature [5], there is no family joint account and the service is limited to adult patients at the current time. Each patient has to sign up for their own portal account. Thus, use of the service was purely voluntary by patients and providers.

The eVisit Service Process:

The eVisit process is initiated by a patient who logs into the patient portal to submit an eVisit for a non-urgent health condition. The patient is introduced to information regarding eVisits, including overview, warnings, and frequently asked questions, linked to “Submit an eVisit” option as well as a video demo of an eVisit for patient education. The patient must accept the terms and conditions of eVisit comprising emergency disclaimers and privacy policy before accessing the main forms to list symptoms associated with any of the 7 conditions covered by the eVisit service (Figure 1). An ‘Other’ category is available to allow specification of conditions that are not included in the 7 well defined areas (Figure 2). The patient may select a pharmacy for any prescriptions needed for the visit or add their own, and review health issues, medications, and allergies.

Figure 1.

Screen Shot of eVisit Request Form

Figure 2.

eVisit Questionnaire for ‘Other’ Category

In the case of the 7 specific conditions, a questionnaire with branching specific to the chosen condition and related symptoms is completed and free text added for symptoms not on the list. Once the eVisit is submitted, the message goes to a support staff pool, a successful message submission acknowledgement is received by the patient as well as a notification of subsequent steps. These include forwarding the eVisit to a physician who is on call to provide a timely response during regular business hours. The physician reviews the eVisit, makes appropriate diagnosis, and replies to the patient about how to proceed. Once the patient receives the information and it is deemed to have addressed the health condition, a satisfaction survey is completed. If the patient has additional concerns, a few request-response exchanges take place before the physician closes the encounter and notifies the support staff. This completes the eVisit. If the physician decides that the condition cannot be treated online, the patient is scheduled for an office visit and the eVisit submission is cancelled without charge to the patient.

Data:

Data for analyzing usage and trends came from several sources. These included portal activation data for eligible members of the medical center as well as the clinics participating in the pilot study, de-identified patient demographics and eVisit transaction data, and payer information. Transaction data related to eVisits submitted between April 1, 2009 and May 31, 2010 included a unique message identifier, associated eVisit, patient and physician identifiers, date, time and subject of the message (one of the 8 conditions), and whether the message was from/to the patient. Patient demographics included age, gender, ethnicity, marital and employment status. Total patient population of the study clinics and the subset who had activated portal access were also available. Overall, 446 eVisits were submitted by 336 unique patients, with 72 patients submitting requests more than once. 1,523 total messages were exchanged between patients and physicians in this time period. Practices 1 – 4 had 51, 12, 8 and 3 physicians, respectively, with participation rate in the eVisit service at 12% (6 physicians), 100%, 50%, 33%, respectively. Since the service is provided by voluntarily participating physicians, we can expect more eVisits from a practice with higher physician participation rate. Therefore, it is not surprising that the number of eVisits is the highest in practice 2.

Models:

In this study, we distinguish two categories of patients – those who utilized only the portal service (‘PP only’ patients) which allows them to make appointments, update demographic information, submit billing/insurance information, and find related health information; and those who additionally submitted at least one eVisit (‘eVisit’ patients). Since both groups are online service users, there is no discrepancy regarding web accessibility and their use of healthcare information which minimizes possible selection bias. Descriptive analysis using simple t-tests delineated some characteristics that significantly differentiated eVisit patients from PP only patients. We then estimate the effect of demographic characteristics, clinical characteristics such as the total number of diagnoses and medications, and frequency of past portal usage on the odds of eVisit submission by using an empirical model with practice fixed effects. Some of these variables have also been suggested by previous studies which have found that females are dominant in online health consultations [5, 6] and likely to have more chronic diseases such as diabetes [6], indicating that online health services users may have more health concerns. Studies have found conflicting results regarding age; portal users were found to be older, on average, than regular members in [6], but of working age (mean = 38 years) in [5]. Although [6] does not distinguish between secure messaging users and portal users, it is implied that secure messaging service is used by the portal users, thus it is reasonable to assume that patients using emails as a communication channel have the above characteristics.

Our t-test results in the next section indicate that employment status as well as insurance coverage are significant, and thus we include these as explanatory variables alongside gender, age, and health condition (total number of diagnoses and medications). The unit of analysis is each patient record and dependent variable is whether a patient has experienced an eVisit during the study period which begins on April 1, 2009. The total portal usage per patient is obtained for the period from September 1, 2008 to March 31, 2009. Since eVisit itself is a part of the patient portal service, utilizing eVisit inevitably adds to the frequency of portal usage. Therefore, in order to minimize bias, we use the frequency of portal access immediately preceding our study period.

Four models are developed in the empirical analysis. The basic model, Model 1, includes only patient demographics as explanatory variables across all the practices, in aggregate. Next, we incorporate available clinical data into Model 1 by including diagnoses and medication information as well as past portal access as predictors to estimate Model 2 since the differences in these features between eVisit and PP only users are significant in the descriptive summary. The descriptive statistics in Table 2 indicates that there are also distinct differences between the four practices in demographics, portal access and user characteristics, thus we estimate Model 3 after adding practice fixed effects to Model 2, as well as insurance information. Our data does not provide detailed information on insurance coverage for eVisits, but approximate information was available, so the results presented here are an upper bound on the actual effect. Model 4 then replaces practice fixed effects with physicians’ eVisit participation rate in each practice to see whether the differences across practices can be explained by the physician participation factor. The model specifications are as follows:

• Model1

• eVisit = α + βDemographics + ɛ

• Model2

• Model1 + γPortal Access + δ₁Diagnoses + δ₂Medications

• Model3

• Model2 + ωInsurance + θPractice fixed effects

• Model4

• Model2 + ωInsurance + θPhysician Particiation

Table 2:

Summary of demographics by practice

	eVisit patients				PP only patients
	Practice1	Practice2	Practice3	Practice4	Practice1	Practice2	Practice3	Practice4
# of patients logged	61	254	15	6	4,978	3,630	888	700
Average age	45.17	46.88	55.07	48	50.81	52.04	56.27	52.31
Female percentage	83.33%	77.87%	60%	83.33%	65.15%	63.98%	69.03%	50.43%
Fulltime employed	73.33%	64.03%	60%	66.67%	65.33%	58.44%	60.02%	61.86%
Average diagnoses	1.46	1.82	1.17	1.6	0.24	0.21	0.19	0.24
Average PP access	44.72	13.21	6.5	8.67	17.31	6.29	6.2	4.37
Insurance holders	98.4%	95.6%	100%	100%	67%	81.1%	83.4%	87.4%

There are several categories in patient’s employment status, marital status and ethnicity, but we have chosen to compare ’full time, retired and others’ as to employment, ’married or else’ as to marital status, and ’white or others’ as to ethnicity because other categories either do not have significant variation between eVisit users and PP only users or have very small number of observations. We use multivariate logistic regression as our preliminary analysis method.

Results

Descriptive Analysis:

Out of 13,279 patients who signed up for the portal service, 10,532 patients (79.3 percent) accessed the portal at least once, but only 3.2 percent of those portal users experienced eVisit service. 40 percent of eVisit patients had sinus/cold symptoms as a major complaint, thus the frequency of eVisit use is higher during the winter time (Figure 3) whereas the frequency of PP access does not show any seasonal variation (Figure 4). 63 percent of eVisits were submitted during clinic business hours (8am to 4pm), and the median response time (first response to patient’s original submission) to an eVisit submitted during business hours was 45 minutes.

Figure 3:

Monthly eVisit frequency

Figure 4:

Monthly PP frequency

Figure 5 presents the differences in age distribution across the two categories of users. Interestingly, PP users are distributed across all age groups, making it close to normal distribution, whereas patients who submitted eVisits are concentrated in the younger age groups. Also, there is a noticeable difference in gender composition between eVisit users and PP only users; eVisit group has significantly higher proportion of female users (Figure 6).

Figure 5:

Age distribution of eVisit &PP users

Figure 6:

Gender composition of eVisit & PP users

Summarizing other demographic characteristics of the two user groups, Table 1 shows that the majority of patients only used the portal service; eVisit patients are younger on average, primarily female, with a smaller retired proportion. These differences in demographic characteristics, patient conditions, and volume of PP access between eVisit users and PP only users in aggregate can also be observed amongst the four different primary care practices (Table 2). The average age tends to be higher and the female proportion substantially lower in practice 4 within the eVisit group and in practice 3 within PP only group. Employment status also varies across the practices as well as average frequency of PP access. Additionally, one of the four practices has a higher proportion of eVisit users relative to its patient pool (see Practice 2 in Table 3), which may be partly due to each practice having a different strategy for building awareness and promoting the eVisit service. Therefore, practice fixed effect regression model is used (Model 3) in order to account for these observed differences among the four practices, and the results are presented in the next section.

Table 1:

Demographics of eVisit vs. portal users

	eVisit patients	PP only patients
Number of patients	336	10,196
Average age*	46.88	51.73
Min (Max) age	21 (78)	19 (97)
Female percentage*	78%	64%
Married percentage	66.4%	67.1%
Fulltime employed	65.5%	62.2%
Student – Full time	2.98%	3.84%
Retired*	6.0%	13.6%
Ethnicity – white	76.5%	76.4%
Average diagnoses*	1.72	0.24
Average Medications*	1.23	0.11
Average PP access*	18.5	12.5
Insurance coverage*+	96.4%	74.1%

^*The difference between two groups is significant at α = 0.001

⁺In the absence of actual coverage information, assumptions are made to approximate whether a particular policy covers eVisit.

Table 3:

Patient proportion and number of physicians in practices

Practice	eVisit users	eVIsit volume	PP only users	Patient proportion	Number of Physicians
Practice1	18.15%	69	48.82%	43%	51
Practice2	75.6%	353	35.61%	36%	12
Practice3	4.46%	17	8.71%	13%	8
Practice4	1.79%	7	6.86%	8%	3
Total	100%	446	100%	100%	74

Regression Results:

Table 4 summarizes the effect of the variables on the likelihood of eVisit use. Although the magnitude of the effect is not large, the frequency of past portal access is a significant indicator of eVisit use throughout all models; for each additional portal access, the odds of being an eVisit user increases by approximately 0.5%. Female patients are more likely to use the service than male patients as we expected from the descriptive statistics (Table 1), and the age and retired employment status works negatively on the odds of becoming an eVisit user. If a patient’s age increases by 1, the odds decreases by over 2 percent in all scenarios. Retired status is the significant factor that differentiates eVisit users from PP only users from the t-test result. However, the effect is not significant overall across the regression models. As expected from the descriptive statistics, ethnicity of being white is also not significant. The work practices of a clinic plays an important role in encouraging patients to use eVisit, hence practice indicator is a significant explanatory variable when added to the model (p<0.001 in Model 3). As expected, the more complex a patient’s health condition as indicated by the number of diagnoses and medications, the more likely that the patient uses the eVisit service. This supports prior research in which use of e-health services was shown to be highest among patients with greater medical need [14]. As seen in Models 3 and 4, one additional diagnosis increases the odds of using eVisit by 18 percent, and one more medication increases the odds by 29 percent. The significance of portal usage, gender, age, diagnoses and medications remain the same. With the removal of the practice fixed effect and addition of the physician participation rate in Model 4, the estimated coefficients across all variables stay similar in Model 3 and Model 4, which implies that physician’s participation rate in the eVisit service explains the differences in practices sufficiently. One additional percentage increase in physician’s participation rate will increase the odds of using eVisit by 1.02 percent.

Table 4:

Effect on the likelihood of eVisit use

	Model 1	Model 2	Model 3	Model 4
gender	0.617*** (0.136)	0.6127*** (0.1403)	0.6033*** (0.0927)	0.6057*** (0.1418)
Age	−0.021*** (0.005)	−0.0246*** (0.0051)	−0.0235*** (0.0012)	−0.0239*** (0.0052)
Fulltime	0.003 (0.127)	0.1388 (0.1337)	0.2351*** (0.0326)	0.2345+ (0.1364)
retired	−0.371 (0.277)	−0.2567 (0.2877)	−0.2425* (0.1177)	−0.2565 (0.2876)
married	0.247+ (0.247)	0.3380* (0.1324)	0.0766 (0.1092)	0.0711 (0.1354)
white	0.034 (0.132)	0.1310 (0.1389)	0.0342 (0.0501)	0.0309 (0.1408)
Portal		0.0028** (0.0009)	0.0051*** (0.0011)	0.0052*** (0.0010)
diagnoses		0.1264** (0.0383)	0.1682* (0.0824)	0.1652*** (0.0401)
medication		0.3158*** (0.0538)	0.2507*** (0.0282)	0.2524*** (0.0580)
e_insurance			0.8785* (0.3471)	0.8714** (0.3308)
participation				0.0227*** (0.0017)

⁺p<0.1,

^*p<0.05,

^**p<0.01,

^***p<0.001

( ): standard deviation

Note that contrary to an earlier study on willingness to pay (WTP) for online medical services which used surveys to reveal that WTP does not vary across age groups [8], our study indicates that older patients are less likely to use paid online consultation service. This illustrates the discrepancy between surveys and reality in which patients face copayment for electronic visit. From a financial perspective, we observed that the top 10 health insurance organizations subscribed to by the two groups, eVisit users and PP users, differed significantly. Since only few payers reimbursed for eVisits at the time of this data collection, this implies that a patient’s health insurance may be an important factor that influences eVisit use. Accordingly, our results show that holding an insurance that is assumed to cover eVisits increases the odds of eVisit use significantly.

Discussion and Conclusions

Despite the differences in perceived online skills between men and women, our findings suggest that women are more likely to use the eVisit service [4, 10, 15]. The same observation was made in a previous pilot study at Mayo clinic, and the primary reason attributed to this occurrence is the role of a female as the primary family caretaker [5]. However, unlike previous literature, our study does not include pediatrics, and the patients in our data are not people who submit eVisit inquiries but the patients who are to be treated, thus the role of the family caretaker has limited contribution. In fact, it is known that while perceived skills and actual skills are different, men and women are not different in their actual online behavior [7, 9]. Our result also follows that in the email messaging study [6] regarding females being more likely to utilize healthcare information, and shows that even in actual service usage, females show more interest in online care delivery via eVisits.

When practice fixed effect is added in Model 3, employment status becomes a significant factor. This reinforces the fact that the patient populations differ across practices, and this explanatory variable does affect whether a portal user becomes an eVisit user. The higher use of the portal service by the retired population may be due to the reimbursement policies of the health insurer in which the retired population is enrolled, more personal time, and health concerns that do not necessarily support eVisit transactions. The portal service provides health information free of charge, thus using portal only service may provide sufficient benefits to them.

A limitation of the study is the potential lack of generalizability of the results due to small number of eVisit users compared to PP only users. Only 3.2% of portal users experienced eVisit during our study period which provides some preliminary insights for future extensions. Other limitations include assumptions about insurance information and eVisit coverage, and the inclusion of medications and diagnoses that was derived from incomplete medical history. Only limited data on diagnoses and medications was available at the time of this study. Ongoing work proposes to examine the impact of detailed clinical history on eVisit use. From the data, we can address not only the probability of eVisit use, but also the response time of physicians from the threads between patients and physicians that demonstrates how eVisit enables active communication between the two parties. Additional analysis is needed to further study how likely are patients to repeat eVisits, conditioning on their prior user experience [16]. Unlike other online services and e-commerce transactions where trust and reputation systems built on other users’ experiences play an important role [17], eVisit users are mostly aware of what to expect because they know about the clinic and their primary care physician. Thus patients’ repeat eVisits should depend mostly on their own individual experiences rather than others’ ratings.

With availability of physician response time and other important attributes such as diagnoses and prescribed medications, future work will extend our study to focus on predicting patient behavior in choosing eVisit in place of the traditional office visit using more refined econometric models. In particular, categorizing diagnoses into chronic and non-chronic conditions to distinguish healthcare needs, adding healthcare utilization profiles, and detailed insurance coverage information may provide new insights into eVisit and portal utilization. At the same time, patients’ concerns about privacy issues and providers’ concerns about health outcomes and work flow impacts [4, 17] may be discouraging factors in promoting eVisit service. As the eVisit service gets deployed in a large number of additional practices, future work will examine these issues using more recent and extensive data from these care delivery settings.