Profiles of Web-based Portal Users with Inflammatory Bowel Disease

Abstract

Background

Web-based portals can enhance communication between patients and providers to support IBD self-management and improve care. We aimed to identify portal use patterns of patients with inflammatory bowel disease (IBD) to inform future web portal-based interventions and portal design.

Methods

Patients with IBD receiving care at the University of Michigan between 2012 and 2021 were identified. Meta-data from electronic logs of each patient’s most recent year of portal use were abstracted. Portal engagement was characterized in terms of intensity (ie, frequency of use); comprehensiveness (ie, number of portal functions used); and duration (ie, quarters per year of portal use). We used k-means clustering, a machine-learning technique, to identify groupings of portal users defined in terms of engagement features.

Results

We found 5605 patients with IBD who had accessed their portal account at least once. The average age was 41.2 years (SD 16.7), 3035 (54.2%) were female, and 2214 (39.5%) received immune-targeted therapies. We identified 3 patterns of portal engagement: (1) low intensity users (29.5%); (2) moderate intensity, comprehensive, and sustained users (63.3%); and (3) high intensity, comprehensive, sustained users (7.2%). Patients with more intense, comprehensive, and sustained use of the portal were older, female, with more comorbidities, and were more likely to receive immune-targeted therapies.

Conclusion

Understanding distinct patterns of portal use can inform portal-based interventions and portal design. Patient portals may be particularly helpful in delivering assistance to those with comorbidities and those receiving immune-targeted therapies—many of whom demonstrate more intense, comprehensive, and sustained portal use.

Introduction

Inflammatory bowel disease (IBD) is a chronic disabling condition of the gastrointestinal tract that affects an estimated 3 million Americans.¹ High-quality IBD management requires multidisciplinary care, patient engagement in self-management, and a strong patient-provider relationship.^2,3 The use of digital health technologies to support patient care, including text-based remote monitoring platforms, web-based self-management tools, and web-based patient portals, is increasingly used to improve communication between face-to-face clinical encounters.^4–6

Web-based patient portals are secure websites that provide patients with access to their health information and allow them to schedule upcoming appointments and communicate with their health care team.⁷ Engagement with portals can improve patients’ health knowledge, medication adherence, and use of preventive care services.^8,9 Although many health systems are investing in web-based patient portals, little is known about current portal utilization patterns. This information is important to drive effective portal development and promote engagement while minimizing usage barriers. Given the complex clinical and self-management demands of IBD, understanding portal use among these patients is particularly important. However, current portal use among IBD patients has not been described. In this study, we characterized patterns of web-based patient portal engagement among patients with IBD.

Methods

Design and Participants

We identified all patients with IBD who received care at a University of Michigan (and its associated community health care centers and clinics) between 2012 and 2021.¹⁰ Patients were included if they had accessed the web-based patient portal at least once and had at least 1 year of follow-up data after signing up. The MyUofMHealth Patient Portal provides 24/7 access to appointment scheduling, medical bill payment, messaging with the care team, prescription refills, and viewing parts of the health record. Approval from the institutional review board (HUM00165048) was obtained prior to the initiation of the study.

Variables and Measurements

Metadata from electronic logs of each patient’s portal use were abstracted. Portal engagement during their most recent year of portal enrollment was characterized in terms of frequency of use (intensity), which included the number of instances in which the patient (1) logged into the portal, (2) viewed inpatient notes, (3) viewed test results, (4) viewed radiology results, (5) read or replied to a message, (6) loaded a medication list, or (7) requested a medication refill. We also evaluated the comprehensiveness of portal use, which was defined by the number of different portal functions that were used. Last, we characterized the duration of portal use, which was defined by the number of quarters during the year in which a patient used the portal. Validity of portal activity information was previously confirmed through manual chart review.

Metadata were merged with users’ sociodemographic and clinical characteristics stored in their electronic health record. These data included demographic information (eg, age, gender, race), insurance type (Medicaid vs non-Medicaid), comorbid conditions (summarized using Charlson comorbidity index¹¹), mental health disorders (including anxiety, mood disorder, psychosis, alcohol substance abuse, suicide and self-harm, adjustment disorder, disruptive disorder, or personality disorders), IBD-specific data including IBD type (ulcerative colitis, Crohn’s disease, or indeterminate colitis), and use of immune-targeted therapies (biologics and thiopurines).

Statistical Analysis

We performed analyses using SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA) and R (version 4.1.2). Our goal was to identify distinct groups (ie, clusters) of patients based on their pattern of portal use activities and assess differences in patient characteristics between clusters. In this study, the word cluster is used interchangeably with group.

Prior to clustering, each of the 9 portal engagement measures were transformed to standardize their distribution around a mean of 0 and a standard deviation (SD) of 1.^12,13 We then applied the k-means clustering algorithm to the standardized metadata; k-means clustering is an unsupervised machine-learning technique used to find groupings by measuring dissimilarity (distance) between observations and assigning them to separate clusters (k). We used Euclidean distance, one of the classical methods for measuring distance, to perform this analysis. The objective is to identify clusters, which contain observations that are as similar as possible, and observations from different clusters that are as dissimilar as possible.

After applying the k-means clustering technique, we used the elbow method to determine the optimal number of clusters (k).¹⁴ In the elbow method, we plot varying number of clusters (k) against their within-cluster sum of squares. The within-cluster sum of squares is highest at a k = 1. As the number of clusters (k) increases, the within-cluster sum of squares starts to decrease, creating a plot that looks like an elbow. The optimal k value is the k value that corresponds to the point when the within-cluster sum of squares changes more rapidly and becomes parallel to the x-axis. To confirm the optimal number of clusters, we also visualized the clusters. We compared the patient-specific characteristics of the clusters or groupings resulting from the k-means clustering technique using separate bivariate analyses (ie, Student’s t test and χ² test for continuous and categorical data) and a combined multinomial logistic regression. The clustering solution was confirmed using hierarchical agglomerative clustering, which resulted in similar groupings. Unlike k-means clustering, hierarchical clustering methods do not require a prespecified number of clusters to be generated. Using this technique, each observation begins as an individual cluster and merges with other observations until unique clusters or groupings are formed. The resulting clusters are visualized as a dendrogram, a tree-based representation of merging clusters. The study protocol was reviewed and approved by the institutional review board (#HUM00165048).

Results

Overall, we identified 6972 patients with IBD who received their IBD care within a University of Michigan between 2012 and 2021 (Figure 1). A majority, 99.7% (n = 6952), were signed up for the web-based patient portal, and 5826 patients (83.8%) accessed the portal. A minority, 221 patients, had insufficient follow-up that was less than 1 year between sign-up date and their last clinic interaction; these patients were excluded.

The analytic cohort was composed of 5605 patients (80.4%), with a mean age of 41.2 years (SD, 16.7) and a mean Charlson comorbidity index of 1.0 (SD, 2.0; Table 1). Overall, 3035 patients (54.2%) were female, 4752 (84.8%) were White, and 1894 (33.8%) carried a comorbid diagnosis of a mental health disorder. Additionally, 2680 (56.1%) had Crohn’s disease, 2663 (47.5%) had ulcerative colitis, and 262 (4.7%) had indeterminate colitis; 2214 (39.5%) received immune-targeted therapies (ie, thiopurine or biologic medications; Table 1). Patients’ most recent year of portal engagement varied. Continued portal use was common in later years, with 411 (7.3%) patients discontinuing portal use after 2019, 652 (11.6%) patients discontinuing portal use after 2020, and 3919 (69.9%) patients continuing to engage with the portal in 2021 at the end of the study period (Table 1).

In their most recent year of portal engagement, patients most often used the patient portal for reading and replying to messages from their clinical team (mean count 143.9, SD 22.5) and reviewing test results (mean count 52.9, SD 132.0; Table 2). A Pearson’s correlation matrix between metadata variables (ie, portal functions) showed no strong correlations (>0.7) between any of the variables (Figure 2, Supplemental Table 1). Clustering analyses identified 3 patterns of portal engagement (Supplemental Figure 1). A small number of patients (n = 63) had usage patterns that fell outside of these 3 dominant groups and are not described. The 3 portal engagement patterns were as follows: (1) low intensity users (29.5% of users, n = 1,655); (2) moderate intensity, comprehensive, and sustained users (63.3% of users, n = 3548); and (3) high intensity, comprehensive, sustained users (7.2% of users, n = 402; Table 1). When examining these portal engagement patterns by year, moderate and high intensity use was higher in more recent years, and low intensity use was higher in earlier years, though these differences preceded the coronavirus 2019 (COVID-19) pandemic in 2020 (Table 1).

Patients with more intense, comprehensive, and sustained use of the portal were more likely to be older and female, with more comorbidities. The mean age of high intensity, comprehensive, sustained users was 49.0 years (SD, 16.4) compared with 40.7 years (SD, 16.3) among moderate intensity, comprehensive, and sustained users and 40.4 years (SD, 17.1) among low intensity users (P < 0.001). Women comprised 62.9% (n = 253.0) of high intensity, comprehensive, sustained users compared with 54.3% (n = 1927.0) of moderate intensity, comprehensive, and sustained users and 51.9% (n = 859.0) of low intensity users (P < .001). The mean Charlson comorbidity index of high intensity, comprehensive, sustained users was 2.5 (SD, 3.3) compared with 1.0 (SD, 1.9) and 0.7 (SD, 1.6), respectively, for moderate intensive, comprehensive, and sustained users and low intensity users (P < .001). Finally, high intensity, comprehensive, sustained users were also more likely to have a comorbid mental health disorder than moderate intensity and low intensity users with comorbid mental health diagnoses among 55.2% (n = 222.0), 34.2% (n = 1214.0), and 27.7% (n = 458.0), respectively. High intensity, comprehensive, and sustained users (46.3%, n = 186.0) and moderate intensity, comprehensive, and sustained users (48.7%, n = 1726.0) were also more likely to receive biologic medications than low intensity users (39.3%, n = 650.0; Table 1). These findings were confirmed on multinomial logistic regression, where high intensity portal use was independently associated with older age (odds ratio [OR], 1.02; 95% confidence interval [CI], 1.01-1.03; P < .001), female gender (OR, 1.68; 95% CI, 1.33-2.14; P < .001), a higher Charlson comorbidity index (OR, 1.29; 95% CI, 1.23-1.36; P < .001), ulcerative colitis (OR, 1.36; 95% CI, 1.07-1.74; P = .013), comorbid mental health diagnoses (OR, 2.39; 95% CI, 1.88-3.04; P < .001), and biologic use (OR, 1.74; 95% CI, 1.35-2.24; P < .001) compared with low intensity use (Table 3). Similarly, in a multinomial logistic regression, moderate intensity portal use was independently associated with female gender (OR, 1.14; 95% CI, 1.02-1.29; P = .028), a higher Charlson comorbidity index (OR, 1.12; 95% CI, 1.07, 1.16, P < .001), ulcerative colitis (OR, 1.15; 95% CI, 1.01-1.30; P = .032), comorbid mental health diagnoses (OR, 1.24; 95% CI, 1.08-1.41; P = .002), Medicaid coverage (OR, 0.77; 95% CI, 0.66-0.90; P = .001), biologic use (OR, 1.47; 95% CI, 1.29-1.68; P < .001), and thiopurine use (OR, 1.22; 95% CI, 1.07-1.39; P = .003; Table 3). In sensitivity analysis using the hierarchical clustering technique, we found similar results in terms of number of clusters and cluster characteristics (Table 4, Supplemental Figure 2 and 3).

Discussion

In this large sample, we found that patients with IBD have varying patterns of web-based patient portal engagement that can be characterized into 3 distinct groupings: (1) low intensity users; (2) moderate intensity, comprehensive, and sustained users; and (3) high intensity, comprehensive, and sustained users. Understanding these distinct groups of portal users can inform web portal-based interventions that leverage existing patterns of portal use or strive to optimize portal usage to facilitate high-quality IBD care.

Nearly two-thirds of patients were categorized as moderate intensity, comprehensive, and sustained users of the web-based patient portal. Compared with low intensity users, these moderate intensity users are older and have more comorbidities and mental health diagnoses; they are more likely to have ulcerative colitis than Crohn’s disease; they are less likely to receive Medicaid coverage and are more likely to take immune-targeted therapies. These findings suggest that web-portal interventions that require moderate, comprehensive, and sustained use would be feasible for many patients with IBD to engage in. However, nearly one-third of patients with IBD are categorized as low intensity portal users. It is important to take these low intensity portal users into consideration, as they would be more likely to have low fidelity in engaging with portal-based interventions.

On the other hand, high intensity users comprise only 7% of patients. Compared with moderate intensity users, these high intensity users are older and have a higher proportion of multimorbidity and mental health diagnoses. However, the use of immune-targeted therapies is similar among moderate and high intensity users, suggesting similar IBD severity. Although high intensity users represent a minority of patients, they account for a relatively high proportion of the web-based portal burden for clinical staff. For example, high intensity users read or replied to about 39-fold more messages than low intensity users and approximately 4-fold more messages than moderate intensity users. As a second example, high intensity users requested medication refills about 7-fold more often than low intensity users and approximately 2-fold more often than moderate intensity users. These patterns of high intensity web-based portal use are similar to well-established patterns of health care utilization, with a minority of patients accounting for a disproportionate amount of emergency department visits and hospitalizations. The increased burden of patient-generated health data from high intensity web-based portal use can contribute to clinician burn out.¹⁵ Future studies should focus on understanding the broader health care utilization patterns of these high intensity web-based portal users. Established interventions for the highest health care utilizers, such as care coordination programs or whole person health models (eg, IBD-specialized medical home), could be investigated as a potential option for addressing the needs of this subgroup of patients to reduce clinician burden and improve the quality of IBD care.

Patients with IBD most often used the portal for reading and reviewing messages or for reviewing test results. Both activities are a form of patient-provider communication. This speaks to one key role of web-based patient portals: facilitating communication as a means of improving care and health outcomes.^16,17 The role of communication in improving the quality of IBD care should be considered as we increase our mechanistic understanding of the factors that lead to high-quality IBD care.^18,19

Web-based portals have the potential to be leveraged to improve IBD care and health outcomes.²⁰ For example, a web portal-based diabetes program that incorporated activities such as web messaging and uploading of blood glucose data led to improved glucose control among patients with diabetes.²¹ In IBD, one study by Reich et al demonstrated that IBD patients who received disease-specific information through a web-based portal had a trend towards in improved quality of life.²² Further, systematic review of digital health interventions has demonstrated an association with lower rate of health care utilization and health care costs compared with standard of care.²³ Therefore, it is important understand web-based portal use among low-intensity users in efforts to improve IBD outcomes. Low intensity users had fewer comorbidities than moderate intensity users. They were less likely to be taking immunomodulators or biologics, and they were also less likely to have a comorbid mental health diagnosis. Perhaps this suggests that patients who require more complex care, such as patients who receive immune-targeted therapy or those with multiple medical problems, may also be more likely to use the web-based portal for self-care. It is possible that they may also be more fluent in its use. Ultimately, the heterogeneity in web-based portal use behaviors needs to be considered in the development of web portal-based interventions, as interventions tailored to varying portal use behaviors may be more effective than a “one size fits all” approach.

It is interesting to note that we had very high rates of web-based portal users in our institutional overall, which was likely facilitated by prior institutional efforts to increase web portal use.²⁴ Most of our patients, 99.7% (n = 6952) had signed up for the portal, and 83.8% accessed their portal. These findings are consistent with prior studies that have demonstrated that IBD patients have high acceptance of telemanagement.²⁵ This further supports efforts to improve IBD care through web-based portal interventions.

The limitations of this study include its retrospective nature and lack of a control group. The population of this study is based on a data from a tertiary referral center. It is therefore unknown if the findings of this study can be extrapolated to other clinical settings. In addition, we characterized patients based on metadata from electronic logs of each patient’s web-based portal use. There are more subtle features of portal use behaviors that cannot be further characterized using these data, and future qualitative studies may therefore help build on our study’s findings. Finally, adoption of portable technology devices has increased over time. Therefore, patients enrolled later in the study may have higher utilization of the web-based portal.

In conclusion, patients with IBD have varying patient portal engagement behavior that can be characterized into distinct patterns. Understanding these distinct patterns of portal use can inform web portal-based interventions that leverage existing patterns of portal use or strive to optimize portal usage to facilitate patient-provider communication. An example of this might be web-based portal interventions that utilize tailored messaging to support patients in monitoring and managing their symptoms, communication with the health care team when early symptoms of flare are detected, and coping with stressors that may trigger their symptoms. Patient portals may be particularly helpful in delivering assistance to those IBD patients with comorbidities and receiving immune-targeted therapies—many of whom demonstrate more intense, comprehensive, and sustained patient portal use.

Author Contributions

S.C.M., M.N., K.N., J.Z., and A.K.W. contributed to the conceptualization of the study, methodology, drafting of the manuscript, and critical revisions. S.C.M. and A.K.W. also contributed to supervision. M.N. also contributed to data curation. P.D.R.H., J.P., K.R., and J.B. contributed to methodology and critical revisions. All authors approved the final version of the manuscript.

Funding

S.C.M. received funding from the National Institute of Health through the Michigan Institute for Clinical and Health Research (KL2TR002241). A.K.W. received funding from a VA Health Services Research and Development Merit Award. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs or the National Institute of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of Interest

The authors have no other conflicts of interest to disclose.

References