Associations Between Patient–Provider Secure Message Content and Patients' Health Care Visits

Abstract

Background:

Between-visit communications can play a vital role in improving intermediate patient outcomes such as access to care and satisfaction. Secure messaging is a growing modality for these communications, but research is limited about the influence of message content on those intermediate outcomes. We examined associations between secure message content and patients' number of health care visits.

Methods:

Our study included 2,111 adult patients with hypertension and/or diabetes and 18,309 patient- and staff-generated messages. We estimated incident rate ratios (IRRs) for associations between taxonomic codes assigned to message content, and the number of office, emergency department, and inpatient visits.

Results:

Patients who initiated message threads in 2017 had higher numbers of outpatient visits (p < 0.001) compared with patients who did not initiate threads. Among patients who initiated threads, we identified an inverse relationship between outpatient visits and preventive care scheduling requests (IRR = 0.92; 95% confidence interval [CI]: 0.86–0.98) and requests for appointments for new conditions (IRR = 0.95; 95% CI: 0.92–0.99). Patients with higher proportions of request denials or more follow-up appointment requests had more emergency department visits compared with patients who received or sent other content (IRR = 1.18; 95% CI: 1.03–1.34 and IRR = 1.14; 95% CI: 1.07–1.23, respectively). We identified a positive association between outpatient visits and the proportion of threads that lacked a clinic response (IRR = 1.02; 95% CI: 1.00–1.03).

Discussion:

We report on the first analyses to examine associations between message content and health care visits.

Conclusions:

Our findings are relevant to understanding how to better use secure messaging to support patients and their care.

Introduction

Patient–provider communication directly impacts intermediate patient outcomes such as increased trust in providers and improved access to care, patient satisfaction, and self-care.^1

–4 In turn these important intermediate outcomes are associated with improved health in numerous studies (e.g., improvements in glycemic levels, hypertension control, and mental health^5

–9). The effectiveness of patient–provider communication is driven by a variety of patient, clinician, and contextual factors,¹⁰ including communication modality (e.g., verbal and electronic).

Between-visit communications may play a vital role in improving patient outcomes since patients typically forget much of the information shared during a health care encounter.^11
–13 One modality for this between-visit communication growing in frequency is secure messaging, which is the exchange of electronic text-based messages (i.e., e-mail) between clinic staff and patients through a secure platform such as a patient portal. Evidence indicates that patients with complex and chronic conditions take advantage of secure messaging. Four in 10 patients with diabetes used secure messaging between clinical visits,¹⁴ and a majority of cancer patients selected secure messaging as their preferred mode of communication with their clinician, over phone and in-person communication.¹⁵ In addition, secure messaging use may improve access to care for patients with higher out-of-pocket expenses: one study found patients with higher out-of-pocket expenses for visits had fourfold greater odds for using secure messaging to communicate with their clinicians.¹⁶

Studies that examined the association between secure message use (i.e., number of messages) and outpatient visits reported inconsistent findings.^17

–22 Several researchers also noted spikes of short duration in health care visits immediately preceding or following secure messaging use.^18,21,22 One-third of patients reported that use of secure messaging decreased their number of office visits and 4 in 10 reported an associated reduction in phone calls to their clinicians.¹⁶ Alpert et al.¹⁵ reported that patients felt that effective communication delivered through secure messaging prevented unnecessary appointments. Although what is communicated between patients and providers is a known predicate to effective communication that may lead to changes in patients' intermediate and health outcomes,¹ no study explored the association of message content—what type of requests patients make and how clinicians respond—and the number of health care visits.

The mixed findings in current research on the association between secure messaging and visit numbers suggest that understanding the relationship between secure messaging and outcomes requires validated content analysis performed in the context of the message thread (the full electronic conversation, including the initiating message and all responses), and outcomes sensitive to information sharing. We attempt to address this gap by exploring whether message content sent by patients and clinic staff is associated with health care visits. To do this, we developed a taxonomy informed by several theories and designed specifically to categorize the content of secure messages.^23,24 Our taxonomy includes taxa (i.e., codes assigned to message content) to capture patient uncertainty in the form of information seeking requests, patient self-care behaviors manifested in different task-oriented requests, and content that might foster patient–provider relationships (e.g., information sharing and social communication). Taxa related to clinical responses include information sharing, information seeking, action responses to address patients' requests (e.g., fulfillment or partial fulfillment), and encouragement.

Previous studies suggested that effective secure messaging communication prevented some patients from booking unnecessary appointments.¹⁵ We, therefore, hypothesized that patients would have fewer health care visits if the messages included some components of patient-centered communication (i.e., information sharing, self-care management support, and fostering relationships). That is, we hypothesized that patients would have (1) fewer office, emergency department, and inpatient visits if they used secure message to (1a) receive information from clinic staff, (1b) share information with clinic staff, (1c) support their self-care (e.g., prescription-related requests), or (1d) exchange communication with clinic staff to foster relationship building (e.g., Social communication). Conversely, we hypothesized that denying or ignoring patients' secure message requests may lead to increased face-to-face visits as patients seek appropriate responses to their requests. Consistent with these hypotheses, we anticipated positive associations between the (2) number of health care visits and (2a) deferrals or denials sent by clinic staff in response to a patient's request, and (2b) patient-initiated threads that received no response from clinic staff.

Methods

STUDY DESIGN

We conducted a retrospective cohort study using secondary data derived from a large urban academic health system's electronic health record. Our study population included randomly selected adult patients with diabetes or hypertension registered with the patient portal before April 2018. The study period was January 1 through December 31, 2017. We identified patients with diabetes and hypertension if they had at least two outpatient visits or one inpatient visit during 2016 with an E11 or I10 (respectively) International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) code. We excluded patients who did not have at least one outpatient visit during the first 6 months of 2018. Our research received Institutional Review Board approval.

We categorized patients who initiated secure message (SM) threads during the study period (“SM only”), patients who did not initiate message threads (“non-SM users”), and the “full population” that included both. Our full population included 2,111 patients with diabetes and/or hypertension, of whom 1,031 initiated at least one thread in 2017. Thirty-eight percent of patients had only diabetes and 37% only hypertension. Our message sample included 7,346 threads with 10,163 patient-generated messages and 8,146 staff-generated messages. Our clinic staff population totaled 674 and included a range of clinician types, including but not limited to physicians, nursing staff, physician assistants, pharmacists, and administrative staff.

INDEPENDENT VARIABLES

Our independent variables represent patient- and staff-generated message content as identified through our taxonomy. We assigned taxa to all patient- and staff-generated messages that were part of message threads (i.e., an initiating message and all subsequent responses) initiated by patients between January 1 and December 31, 2017. We excluded threads if messages within the thread were sent outside that date range.

Details on the coding process are described elsewhere.²⁴ In brief, a primary coder assigned taxa to all messages and a second coder assigned taxa to a random 10% sample of messages. Discrepant results were discussed, and the primary coder recoded all messages based on any changes in taxa definitions. We assigned at least one code (i.e., taxon) to each message but as many taxa as necessary could be assigned to a message to address all concepts included in each message. We counted a taxon only once per message.

We defined clinic e-mail nonresponse as a message thread that included only patient-generated messages and no messages from clinic staff.

Our study examines each taxon, and clinic nonresponse, as unique independent variables. Because of the strong correlation between the number of threads initiated and taxon occurrence, our independent variables were estimated as a prevalence value for each taxon. The numerator was the number of times the patient sent (patient-generated taxa) or received (staff-generated taxa) a selected taxon and the denominator was the total of all patient- or staff-generated taxa, respectively. We estimated nonresponse prevalence using a denominator of all threads initiated by the patient.

DEPENDENT VARIABLES

Our study examined three dependent variables that captured the number of times a patient was seen for care within the health system during 2017: the number of outpatient visits, the number of emergency department visits, and the number of inpatient visits. We did not include patient visits to other health care institutions. We included all visits occurring between January 1 and December 31, 2017, and we modeled each as discrete count values.

COVARIATES

We included several patient demographic and health status characteristics as covariates based on our understanding of these factors' contributions to health care visits, health status, and taxa use. We included patient sex, race (White, Black, or other), and insurance type (public, private, uninsured, or other), and health condition (diabetes, hypertension, or both conditions) as categorical variables. Our continuous covariates included patient age, number of threads initiated in 2017, and number of diagnoses selected from the most commonly occurring ICD-10-CM diagnosis codes within our population (diabetes, hypertension, lipoprotein metabolism and other lipidemia disorders, overweight and obesity, joint disorders, gastroesophageal reflux disease, back or spine pain, soft tissue disorders such as rheumatism or fibromyalgia, and sleep disorders). We also included the number of outpatient visits as a covariate in analyses where outpatient visits were not the dependent variable. In analyses that only included patients who initiated threads, we added a continuous variable representing the average distance between patients' home addresses and the clinic(s) to which they corresponded, calculated as the distance between zip code centroids.

ANALYTIC APPROACH

We estimated unadjusted differences for categorical variables using chi square for categorical variables and t-test for continuous variables. We applied Poisson regression with robust variance estimation to estimate incident rate ratios (IRRs).²⁵ For each dependent–independent variable combination, we conducted two regression analyses: one with the population who initiated threads (SM users) and a second that included the “full population.” For full population analyses, we included all patients who did not initiate a thread in 2017 (non-SM users) and patients who sent or received a message coded with the selected taxon. In the full population analyses by taxon, patients who sent messages not coded with the selected taxon were excluded from these analyses. For example, a full population analysis of the Information seeking/Medical guidance taxon included all patients who did not send messages and all patients who sent a message coded with Information seeking/Medical guidance; in those analyses, we excluded patients whose sampled messages were not coded with the Information seeking/Medical guidance taxon (see Supplementary Data S2 for sample sizes). All analyses were conducted using SAS v9.4.

Results

Table 1 presents unadjusted descriptive characteristics of patients who did and did not initiate threads. Our SM user population had higher percentages of White and female patients, and a higher percentage had private insurance. A higher percentage of our non-SM users had public insurance.

Table 1.

Descriptive Statistics for Patient Characteristics

CHARACTERISTICS	SENT MESSAGES ( n = 1,031)	DID NOT SEND MESSAGES ( n = 1,080)	p
Age in years (mean)	57.91	57.85	0.92
Distance between home and clinic (mean)	31.02	—	—
Conditions (%)
Both	23.18	25.37	0.12
Diabetes only	38.60	38.24	0.60
Hypertension only	38.22	36.39	0.39
Insurance (%)
Other	26.28	26.51	0.52
Private	32.10	19.18	<0.0001
Public	39.96	51.62	<0.0001
Uninsured	1.64	2.69	0.08
Number of visits (mean)
Emergency department	0.34	0.42	0.09
Inpatient	0.39	0.32	0.12
Outpatient	15.89	11.16	<0.0001
Race (%)
Black	40.5	47.22	<0.01
Other	4.85	6.30	0.10
White	54.6	46.30	0.05
Patient sex (%)
Male	35.01	43.52	<0.001
Female	64.98	56.48	0.09

Table 2 presents the mean prevalence of each taxon among patients who initiated threads. The two most prevalent nongrouping level patient-generated taxa were Prescription refills and renewals and Information seeking/Medical guidance requests. Least prevalent were Social communication taxa. On average, patients initiated 2.11 (95% confidence interval [CI]: 1.88–2.36) threads that received no response from clinic staff.

Table 2.

Mean Taxon Prevalence Among Patients Who Initiated a Message Thread in 2017

GENERATED BY	TAXA^a	MEAN TAXON PREVALENCE (95% CI)^a
Patient (n = 10,163)	Information seeking	20.93 (19.51–22.35)
	Logistics	7.31 (6.48–8.13)
	Medical guidance	14.34 (13.09–15.59)
	Information sharing	23.61 (22.17–25.05)
	Clinical update	10.43 (9.38–11.47)
	Response to clinician's message	10.73 (9.91–11.56)
	Self-reporting	2.45 (1.82–3.08)
	Task-oriented requests not reflective of uncertainty	41.38 (39.31–43.46)
	Prescription refills and renewals	14.97 (13.45–16.49)
	Other administrative	5.00 (4.22–5.78)
	Appointment cancellation	4.70 (3.77–5.64)
	Follow-up appointment	4.26 (3.48–5.05)
	Appointment for preventive care or physical exam	1.47 (1.00–1.94)
	Appointment reschedule	10.98 (9.60–12.37)
	Other task-oriented requests	10.67 (9.58–11.76)
	Referral	1.35 (0.93–1.77)
	New or change prescription	4.09 (3.47–4.71)
	Laboratory test or diagnostic procedure	1.74 (1.34–2.13)
	Appointment for new condition/symptom	3.50 (2.77–4.23)
	Social communication	2.63 (2.18–3.09)
	Appreciation or praise	0.67 (0.43–0.91)
	Complaints	0.79 (0.57–1.00)
	Life issues	1.23 (0.90–1.56)
Clinic staff (n = 8,146)	No response to patient-initiated thread^b	28.07 (26.26–29.88)
	Action responses	30.37 (28.63–32.12)
	Acknowledge	3.81 (3.11–4.51)
	Fulfills request	22.58 (20.93–24.23)
	Partially fulfills request	3.98 (3.37–4.59)
	Denies request	1.10 (0.79–1.41)
	Recommendation to schedule an appointment	1.75 (1.40–2.09)
	Information seeking	12.02 (11.04–12.99)
	Information sharing	38.84 (37.24–40.44)
	Medical guidance	13.10 (11.91–14.29)
	Orientation to procedures–treatments, or preventive behaviors	22.45 (21.15–23.76)
	Deferred information sharing	10.46 (9.48–11.44)
	Social communication: Encouragement	0.89 (0.59–1.18)

For patient-generated taxa, the denominator is the total number of patient-generated taxa. The denominator for the clinician-generated taxa is the total number of taxa assigned to clinician-generated messages sent to the patient.

Denominator is the number of threads initiated by patient.

CI, confidence interval.

Table 3 presents the results from the Poisson regression models that included only patients who initiated threads (full regression results available in Supplementary Data S1). We found a small positive association between clinic nonresponse and outpatient visits. We observed negative associations between outpatient visits and the taxa for Schedule request/Preventive care, Schedule request/New condition or symptom, and the parent-level staff-generated Action responses. We found two taxa positively associated with emergency department visits: Schedule request/Follow-up and clinic staffs' Denial of patient requests. Across most taxa associated with inpatient visits, we observed inverse associations with taxon prevalence. Of the covariates included in the multivariate analyses exploring associations with outpatient visits (see Supplementary Data S1), only those that were health condition related were statistically associated with outpatient visits after controlling for the other covariates in the analyses. For inpatient visit analyses, covariates with statistical significance included patient sex, number of outpatient visits, health condition type, and—for some taxa—payer type. For emergency department visit analyses, covariates with statistical significance included patient race, patient sex (for some taxa), health condition, and number of outpatient visits.

Table 3.

Association between Taxa and Office, Emergency Department, and Inpatient Visits

GENERATED BY	TAXA^a	IRR^b [95% CI] AMONG PATIENTS WHO INITIATED THREADS ( n = 1,008)
GENERATED BY	TAXA^a	OUTPATIENT VISITS	EMERGENCY DEPARTMENT VISITS	INPATIENT VISITS
Patient	Information sharing/Response to clinician's message	–	–	0.87 (0.75–1.00)
	Scheduling request/Follow-up appointment	–	1.14 (1.07–1.23)	–
	Scheduling request/Preventive care	0.92 (0.86–0.98)	–	–
	Scheduling request/New condition or symptom	0.95 (0.92–0.99)	–	0.63 (0.43–0.93)
	Task-oriented/Referral request	–	–	0.64 (0.44–0.95)
	Task-oriented/New or change prescription request	–	–	0.74 (0.60–0.92)
Clinic staff	No response to patient-initiated thread^c	1.02 (1.00–1.03)	–	–
	Action responses^d	0.98 (0.97–1.00)	–	–
	Action response/Denies request	–	1.18 (1.03–1.34)	–
	Information seeking	–	–	0.84 (0.73–0.96)

Independent variable numerator is the number of times the patient sent (patient-generated) or received (clinic staff-generated messages coded with the selected taxon and the denominator is the total of all taxa coded to messages the patient sent or received, respectively.

Represents an incident rate ratio associated with a 10-percentage point increased prevalence of the selected taxon. Each cell represents a separate regression model where the independent variable is the row header and the dependent variable is the column header. Taxa not presented in the table were not statistically significant at p < 0.05.

Independent variable is the percentage of threads that did not receive a clinical response.

Parent-level taxon that is aggregate of Acknowledge, Fulfill, and Partially fulfill.

–, not statistically significant at p < 0.05.

Supplementary Data S2 include the results from the regression analyses that used the full population. We observed positive associations between outpatient visits and taxon prevalence for most patient-generated and clinic staff-generated taxa. Similar to analyses using the population who initiated threads, the full population regression models identified negative associations between taxon prevalence and emergency departments except for Scheduling request/Follow-up (IRR = 1.11; 95% CI: 1.03–1.20). We observed negative associations between inpatient visits and the other task-oriented requests grouping taxon and several of its child taxa.

Discussion

We report on the first analyses to examine associations between message content and health care visits. Our analyses found that patients who initiated message threads had more outpatient visits and fewer emergency department visits. We found reduced inpatient and outpatient visits associated with content indicators of patient self-management (i.e., Scheduling request/Preventive care, Scheduling request/New condition, Referral request, New or change prescription request). We confirmed that communication that lacked components of patient-centered communication was associated with an increase in outpatient visits (i.e., clinic nonresponse) and emergency department visits (i.e., clinic staff denials of patient requests). Counter to our hypotheses, we observed a positive association between emergency department visits and Schedule request/Follow-up and positive associations between outpatient visits and taxa in full population analyses.

Temporal information would be helpful to better understand the association between message content and visits. For example, it is a common practice for patients to receive guidance to follow-up with their primary care providers after discharge from an emergency department visit. The positive association we report in this study between Schedule request/Follow-up and emergency department visits may, therefore, be evidence of patients following that guidance. Without the temporal context for when these requests were made relative to emergency department visits, we cannot be certain if this explains the association.

On average, almost 3 in 10 of our patients' threads did not receive a message response. Many organizations utilize a triage system when responding to patients' messages, although the complexity of such triage systems have been reported by some clinicians as a barrier to use.²⁶ Such complexity could lead to messages being overlooked which could have a negative impact on patient–clinician relationships. This may have significant impact on the trust between patient and clinic staff and reduce patients' ability to manage their care if information-seeking about health conditions or task-oriented requests related to self-management are not addressed. In one study, patients noted that the quality and content of clinicians' SM responses could alter their relationships with those clinicians; for example, frustration increased when questions were left unanswered.¹⁵ Administrators may want to examine ways to simplify or streamline triage workflows to ensure nonresponse is avoided whenever feasible.

We identified a positive association between outpatient visits and clinic nonresponse. The prevalence of nonresponses increased with the number of outpatient visits. We did not observe similar associations with emergency department and inpatient visits; however, the average number of emergency department and inpatient visits our population experienced in 2017 was low, making association detection challenging. It is important to consider that if a patient requested information or that a task be completed with no SM response from the clinic, then the patient likely needed to find another avenue to obtain the answer or complete the task. In keeping with this, Lanham et al.²⁷ found that half of their messages lacking an electronic message response were responded to through another modality (e.g., phone and in-person visit).

Consistent with our hypothesis, patients who received denial responses from their clinical team visited the emergency department more frequently. At 1% prevalence among patients who initiated messages, clinic denials were not a common occurrence. Although ours is an interesting result, it is important to remember that our findings make no indication of causality nor do we have context for the denial. Our analyses were based on individual taxa assigned to messages and do not consider the full conversation within the message thread. A denial accompanied by clinic staff seeking additional information may have a different impact on patient outcomes relative to a denial without information or context. Similarly, denials to certain types of message content may have more impact on outcomes than others. Future research should analyze the impact of the full call-and-response context of the thread to try to tease out these nuances.

LIMITATIONS

In addition to the limitations noted earlier, there are two others we feel are important to call out. First, this is an analysis of message content found within a single large urban medical center's patient portal. The patient population and their engagement with the patient portal, communication practices between patients and clinic staff, and patient outcomes may not be generalizable to other populations. The practice of triaging messages and the characteristics of patients using the portal are similar to other published work, so we expect differences to be minimal.

Second, the variety of taxa sent by a patient increases as the number of threads initiated by the patient increases. To account for this, we represented our independent variables as prevalence values. The use of prevalence allowed us to account for the number of taxa sent or received by the patient, whereas not losing the taxon volume. It does, however, make interpretation of the results somewhat challenging.

Future Research

Our comparisons using the full population found statistically significant associations between outpatient visits and many taxa. It is possible that these associations reflect differences in the two patient populations that we could not control for in our analyses. For example, we may be observing patient activation and engagement represented through secure messaging use rather than associations between specific taxa and our dependent variables. One of the final stages of patient activation is taking action to improve or maintain health,²⁸ and it could be argued that patients who make requests of their clinical team between outpatient visits are taking that action. In fact, one study found a positive association between patients' between-visits communications and activation rates.²⁹ The results presented herein may have been impacted by unmeasured moderating or mediating variables, one of which could have been patient activation. Our findings are consistent with this hypothesis in that many of the differences we detected by taxon in the full population analyses did not persist in models that only included patients who initiated threads. Future research should examine patient activation as a mediator in these relations.

It is important to consider that as with other forms of communication, secure messaging use varies by patient characteristics, including age, education, race, ethnicity, primary language, health literacy, home location, income, and patient sex.^{30

–43} These findings may indicate that secure messaging may not be an appropriate form of communication for all patients; in fact, the Institute of Medicine reports that patient–clinician communication should be provided through the patient's preferred modality.⁴⁴ In addition, research has demonstrated that SM content varies by both patient and staff characteristics.⁴⁵ Our study's focus, therefore, was primarily to determine if message content was associated with patients' health care utilization among patients who opted to use that communication modality. By also comparing SM users' to nonusers' health care utilization in our secondary analyses, we provide early data on the differences in utilization between users and nonusers.

As noted earlier, however, these findings may be confounded by unmeasured mediating variables such as patient activation and patients' communication preferences. To mitigate the impact on known mediators of communication, we controlled for many patient characteristics in our multivariable analyses. Future studies should continue to integrate as many of these patient characteristics as is feasible and continue to compare the impacts of secure messaging use and nonuse on intermediate and health outcomes. Because research has demonstrated impacts to health outcomes based on secure messaging use and content,^{17,36
–38,46

–49} it will be important to understand whether these differences have the potential to further exacerbate existing health disparities and if expanding use to populations not using secure messaging could promote better health equity.

Similar to other research, we hypothesized a general positive or negative association between taxa and visit count. The goal in improving patient outcomes, however, should not be reducing visits but rather reducing inappropriate or unnecessary care. It may be that an average of 15 visits per patient is what our population needed to achieve good outcomes. Future research must consider the value of measuring utilization overall, which is an important component to overall health care cost and delivery, with the association of utilization on outcomes. When considering the communication pathway and impacts to patients' health outcomes, researchers may wish to integrate alternative metrics such as whether patients received appropriate guidance-based care (e.g., preventive care, laboratory tests, and screenings).

We included in our analyses a covariate that approximated the average travel distance between the patient's home and the clinic. This continuous variable was not statistically significant in any of our multivariate analyses after controlling for other variables. Other published research found that patients who lived closest to the clinic were less likely to seek certain types of information and share clinical updates compared with those who lived further.⁴⁵ It is unclear, however, whether those differences reflect a communication choice based on travelling distance or are a proxy for other characteristics not included in the analyses. Future research should explore the influences of SM use among patients with different travelling distances and include in those analyses key potential confounders of such use (e.g., rural vs. urban location, accessibility of internet and transportation, poverty, and education).

Finally, there is minimal research available on the cost implications of secure messaging use. Estimates of the time needed by clinic staff to respond to patient messages ranges between 3.5 and 15 min^47,50,51; this time is typically not reimbursed by health insurance plans. Conversely, by estimating productivity in terms of Relative Value Units, physicians who used secure messaging averaged 11% more visits and $95 a day more than their counterparts who did not use secure messaging.⁵² Future research should explore not only the patient intermediate and health outcomes associated with secure messaging use, but the financial impact on patients and health care organizations as well.

Conclusion

Through the application of a theory-informed taxonomy, this study explores one step in the Street et al.¹ pathway that demonstrates the link between communication functions applicable to secure messaging and patients' health outcomes. Our dependent variables are proxies for Street's intermediate outcomes of health care access and self-management. A recent study found that patients felt that effective communication delivered through secure messaging prevented unnecessary appointments.¹⁵ This study represents the first step to better understanding the types of content that might be leveraged to improve that communication and achieve the goals of advancing appropriate health care utilization. Next steps for advancing this research include analyses that integrate temporality into the analyses, exploring the full context of the conversation relative to outcomes rather than a single taxon, and identifying different utilization metrics that align with SM use. It is clear, however, that our first-of-its-kind research demonstrates linkages between utilization and the content sent by patients and clinicians. Clinic staff and health care administrators should be aware of these associations when establishing message triage workflows and responding to SMs.

Footnotes

Disclaimer

The primary author's affiliation with The MITRE Corporation is provided for identification purposes only and is not intended to convey or imply MITRE's concurrence with, or support for, the positions, opinions, or viewpoints expressed by the author.

Disclosure Statement

The authors have no competing or financial interests to declare.

Funding Information

No funding was supplied to support this research.

Supplementary Material

Supplementary Data S1

Supplementary Data S2

References

Street

Jr. , Makoul

, Arora

, Epstein

RM.

How does communication heal? Pathways linking clinician–patient communication to health outcomes. Patient Educ Couns, 2009; 74:295–301.

Epstein

, Street

. The values and value of patient-centered care. Ann Fam Med, 2011; 9:100–103.

Epstein

, Street

Jr . Patient-centered communication in cancer care: Promoting healing and reducing suffering. National Cancer Institute, NIH Publication No. 07-6225. Bethesda, MD, 2007. Available at http://outcomes.cancer.gov/areas/pcc/communication (last accessed August 20, 2020 ).

Epstein

, Duberstein

, Fenton

, et al. Effect of a patient-centered communication intervention on oncologist-patient communication, quality of life, and health care utilization in advanced cancer: The VOICE randomized clinical trial. JAMA Oncol, 2017; 3:92–100.

Lafata

, Morris

, Dobie

, Heisler

, Werner

, Dumenci

. Patient-reported use of collaborative goal setting and glycemic control among patients with diabetes. Patient Educ Couns, 2013; 92:94–99.

Schoenthaler

, Kalet

, Nicholson

, Lipkin

Jr . Does improving patient-practitioner communication improve clinical outcomes in patients with cardiovascular diseases? A systematic review of the evidence. Patient Educ Couns, 2014; 96:3–12.

Arora

, Weaver

, Clayman

, Oakley-Girvan

, Potosky

. Physicians' decision-making style and psychosocial outcomes among cancer survivors. Patient Educ Couns, 2009; 77:404–412.

Stewart

, Brown

, Donner

, et al. The impact of patient-centered care on outcomes. J Fam Prac, 2000; 49:796–804.

Naik

, Kallen

, Walder

, Street

Jr . Improving hypertension control in diabetes mellitus: The effects of collaborative and proactive health communication. Circulation, 2008; 117:1361–1368.

10.

Epstein

, Franks

, Fiscella

, et al. Measuring patient-centered communication in patient–physician consultations: Theoretical and practical issues. Soc Sci Med, 2005; 61:1516–1528.

11.

da Assunção

, Neely

, Lochab

, Mizumi-Richards

, Barnett

, Pandit

. Patient recall of surgical information after day case knee arthroscopy. Knee Surg Sports Traumatol Arthrosc, 2013; 21:1510–1515.

12.

Tarn

, Flocke

. New prescriptions: How well do patients remember important information?. Fam Med, 2011; 43:254–259.

13.

Kessels

RP.

Patients' memory for medical information. J R Soc Med, 2003; 96:219–222.

14.

Lyles

, Grothaus

, Reid

, Sarkar

, Ralston

. Communication about diabetes risk factors during between-visit encounters. Am J Manag Care, 2012; 18:807–815.

15.

Alpert

, Markham

, Bjarnadottir

, Bylund

. Twenty-first century bedside manner: Exploring patient-centered communication in secure messaging with cancer patients. J Cancer Educ, 2021; 36:16–24.

16.

Reed

, Graetz

, Gordon

, Fung

. Patient-initiated e-mails to providers: Associations with out-of-pocket visit costs, and impact on care-seeking and health. Am J Manag Care, 2015; 21:e632–e639.

17.

Harris

, Haneuse

, Martin

, Ralston

. Diabetes quality of care and outpatient utilization associated with electronic patient-provider messaging: A cross-sectional analysis. Diabetes Care, 2009; 32:1182–1187.

18.

North

, Crane

, Chaudhry

, et al. Impact of patient portal secure messages and electronic visits on adult primary care office visits. Telemed J E Health, 2013; 20:192–198.

19.

Liss

, Reid

, Grembowski

, Rutter

, Ross

, Fishman

. Changes in office visit use associated with electronic messaging and telephone encounters among patients with diabetes in the PCMH. Ann Fam Med, 2014; 12:338–343.

20.

Bergmo

, Kummervold

, Gammon

, Dahl

. Electronic patient–provider communication: Will it offset office visits and telephone consultations in primary care?. Int J Med Inform, 2005; 74:705–710.

21.

Zhou

, Garrido

, Chin

, Wiesenthal

, Liang

. Patient access to an electronic health record with secure messaging: Impact on primary care utilization. Am J Manag Care, 2007; 13:418–424.

22.

Meng

, Palen

, Tsai

, McLeod

, Garrido

, Qian

. Association between secure patient–clinician email and clinical services utilisation in a US integrated health system: A retrospective cohort study. BMJ Open, 2015; 5:e009557.

23.

Heisey-Grove

, DeShazo

. Look who's talking: Application of a theory-based taxonomy to patient–clinician e-mail messages. Telemed J E Health, 2020; 26:1345–1352.

24.

Heisey-Grove

, Rathert

, McClelland

, Jackson

, DeShazo

. Classification of patient- and clinician-generated secure messages using a theory-based taxonomy. Health Sci Rep, 2021; 4e295.

25.

Hilbe

JM.

Modeling Count Data. Cambridge University Press, CA, 2014.

26.

Wooldridge

, Carayon

, Hoonakker

, Musa

, Bain

. Technology-mediated communication between patients and primary care clinicians and staff. Proc Hum Factors Ergonomics Soc Ann Meet, 2016; 60:556–560.

27.

Lanham

, Leykum

, Pugh

. Examining the complexity of patient-outpatient care team secure message communication: Qualitative analysis. J Med Internet Res, 2018; 20:e218.

28.

Hibbard

, Stockard

, Mahoney

, Tusler

. Development of the patient activation measure (PAM): Conceptualizing and measuring activation in patients and consumers. Health Serv Res, 2004; 39:1005–1026.

29.

Alexander

, Hearld

, Mittler

, Harvey

. Patient-physician role relationships and patient activation among individuals with chronic illness. Health Serv Res, 2012; 47:1201–1223.

30.

Cronin

, Davis

, Shenson

, Chen

, Rosenbloom

, Jackson

. Growth of secure messaging through a patient portal as a form of outpatient interaction across clinical specialties. App Clin Inform, 2015; 6:288–304.

31.

Tang

, Black

, Young

. Proposed criteria for reimbursing eVisits: Content analysis of secure patient messages in a personal health record system. AMIA Annu Symp Proc 2006:764–768.

32.

White

, Moyer

, Stern

, Katz

. A content analysis of e-mail communication between patients and their providers: Patients get the message. J Am Med Inform Assoc, 2004; 11:260–267.

33.

North

, Crane

, Stroebel

, Cha

, Edell

, Tulledge-Scheitel

. Patient-generated secure messages and eVisits on a patient portal: Are patients at risk?. J Am Med Inform Assoc, 2013; 20:1143–1149.

34.

Haun

, Patel

, Lind

, Antinori

. Large-scale survey findings inform patients' experiences in using secure messaging to engage in patient-provider communication and self-care management: A quantitative assessment. J Med Internet Res, 2015; 17:e282.

35.

Graetz

, Gordon

, Fung

, Hamity

, Reed

. The digital divide and patient portals: Internet access explained differences in patient portal use for secure messaging by age, race, and income. Med Care, 2016; 54:772–779.

36.

Shimada

, Allison

, Rosen

, Feng

, Houston

. Sustained use of patient portal features and improvements in diabetes physiological measures. J Med Internet Res, 2016; 18:e179.

37.

Chung

, Panattoni

, Chi

, Palaniappan

. Can secure patient-provider messaging improve diabetes care?. Diabetes Care, 2017; 40:1342–1348.

38.

Price-Haywood

, Luo

, Monlezun

. Dose effect of patient–care team communication via secure portal messaging on glucose and blood pressure control. J Am Med Inform Assoc, 2018; 25:702–708.

39.

Tarver

, Menser

, Hesse

, et al. Growth dynamics of patient-provider internet communication: Trend analysis using the Health Information National Trends Survey (2003 to 2013). J Med Internet Res, 2018; 20:e109.

40.

Wade-Vuturo

, Mayberry

, Osborn

. Secure messaging and diabetes management: Experiences and perspectives of patient portal users. J Am Med Inform Assoc, 2013; 20:519–525.

41.

Smith

, O'Conor

, Aitken

, Curtis

, Wolf

, Goel

. Disparities in registration and use of an online patient portal among older adults: Findings from the LitCog cohort. J Am Med Inform Assoc, 2015; 22:888–895.

42.

Schickedanz

, Huang

, Lopez

, et al. Access, interest, and attitudes toward electronic communication for health care among patients in the medical safety net. J Gen Intern Med, 2013; 28:914–920.

43.

Masterman

, Cronin

, Davis

, Shenson

, Jackson

. Adoption of secure messaging in a patient portal across pediatric specialties. AMIA Annu Symp Proc, 2016; 2016:1930–1939.

44.

Institute of Medicine (US) Committee on Quality of Health Care in America. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: National Academies Press, 2001.

45.

Heisey-Grove

, Rathert

, McClelland

, Jackson

, DeShazo

. Patient and clinician characteristics associated with secure message content: Retrospective cohort study. J Med Internet Res, 2021; 23:e26650.

46.

Heisey-Grove

, Laura

, Rathert

, Tartaglia

, Jackson

, DeShazo

. Associations between patient health outcomes and secure message content exchanged between patients and clinicians: Retrospective cohort study. J Med Internet Res, 2020; 22:e19477.

47.

Zhou

, Kanter

, Wang

, Garrido

. Improved quality at Kaiser Permanente through e-mail between physicians and patients. Health Aff (Milwood), 2010; 29:1370–1375.

48.

Ralston

, Cook

, Anderson

, et al. Home blood pressure monitoring, secure electronic messaging and medication intensification for improving hypertension control: A mediation analysis. App Clin Inform, 2014; 5:232–248.

49.

Harris

, Koepsell

, Haneuse

, Martin

, Ralston

. Glycemic control associated with secure patient-provider messaging within a shared electronic medical record. Diabetes Care, 2013; 36:2726–2733.

50.

Anand

, Feldman

, Geller

, Bisbee

, Bauchner

. A content analysis of e-mail communication between primary care providers and parents. Pediatrics, 2005; 115:1283–1288.

51.

Lin

C-T

, Wittevrongel

, Moore

, Beaty

, Ross

. An internet-based patient-provider communication system: Randomized controlled trial. J Med Internet Res, 2005; 7:e47.

52.

Liederman

, Lee

, Baquero

, Seites

. Patient-physician web messaging. The impact on message volume and satisfaction. J Gen Intern Med, 2005; 20:52–57.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.10 MB

0.06 MB

0.00 MB