Teen secure messaging is associated with use of sexual and reproductive health services in one health system

Abstract

Background:

Patient portal use may improve access to or use of sexual and reproductive health (SRH) services for adolescents. We examined the association between adolescent secure messaging on a portal and use of SRH services in one health system.

Methods:

We conducted a retrospective cohort study using electronic health records of adolescents aged 13–17 at Kaiser Permanente Washington from 2019 to 2021. Using month of observation as our unit of analysis, we estimated unadjusted and adjusted odds ratios and 95% confidence intervals for associations between secure messages sent and 3 outcomes: (1) sexually transmitted infection (STI) tests ordered in the observed month, and among pregnancy-capable individuals, (2) pregnancy tests and (3) prescription contraceptive methods ordered.

Results:

37,818 unique individuals contributed 667,678 months of individual observation. Among observed months with STI testing, 6.1% sent secure messages compared to 1.1% in months with no STI testing. Observed months with secure messaging had significantly greater odds of STI tests ordered than months without messaging, after adjusting for covariates (adjusted odds ratio (aOR) 3.5, 95% confidence interval (CI) 2.9, 4.3). Among observed months with prescription contraceptive orders, 5.0% sent a portal message compared to 1.4% without prescription contraceptive orders. Observed months with secure messaging among pregnancy-capable individuals had significantly greater odds of pregnancy tests (aOR 2.6, 95% CI 2.2, 3.1) and prescription contraceptive orders (aOR 1.9, 95% CI 1.7, 2.2) than months without messaging.

Conclusion:

The use of secure messaging by adolescents was more common among those with concurrent SRH care needs.

Implications and Contribution:

In this study using electronic health records of more than 30,000 adolescents aged 13–17 in one health system, we identified an association between secure messaging on a patient portal and receipt of sexually transmitted infection testing and reproductive health services. Portal use may improve linkages to healthcare for adolescents.

Background

Most adolescents in the United States (US) become sexually active before age 18.¹ Yet many face unmet sexual and reproductive health (SRH) care needs.^{2, 3} As a result, adolescents and young adults experience disproportionately worse SRH outcomes compared to adults.⁴ For example, young people aged 15 to 24 experience more than 50% of new sexually transmitted infections (STIs) despite comprising only 27% of the US population. Adolescents aged 15 to 19 who become pregnant are also more likely than older individuals to describe their pregnancies as unintended.⁵

Barriers to essential SRH care such as STI testing and prescription contraceptive methods perpetuate the health inequities that adolescents face. Healthcare access varies substantially, depending on one’s age, geographic location, and economic circumstances.³ Even in settings where SRH care is available, adolescents may not have the opportunity to discuss their SRH needs with clinicians due to lack of opportunities to communicate privately with them.⁶ As a result, continued efforts to identify innovative strategies to connect adolescents to SRH care are needed.

While patient access to the electronic health record (EHR) via an online patient portal has become standard for adults for over a decade, only more recently have adolescents begun to have their own independent access to patient portals. Health systems increasingly offer separate patient portal accounts for adolescents and their parents/caregivers, known as proxy users. However, even when portal accounts are available to adolescents, they are often underutilized.⁷

As adolescents have more options to connect with their healthcare teams digitally and healthcare systems have expanded digital care pathways, it is important to understand how this new digital landscape may impact adolescent SRH care access.⁸ Our previous work showed that a small but increasing proportion of eligible adolescents use the patient portal for secure messaging, and those that do are more likely to be older, female sex (vs. male sex), transgender (vs. not transgender), and have private insurance (vs. public insurance).⁷ Prior studies have demonstrated that adolescents are interested in using the patient portal and other digital healthcare tools such as telemedicine for sexual healthcare needs.^9–11

Currently little is known about whether individuals who use an adolescent patient portal are more likely to receive SRH services than those who do not. Thus, the purpose of this study was to examine the association between secure messaging on a patient portal and receiving SRH services by adolescents in one health system. Our hypothesis was that adolescents who send secure messages were more likely to have received SRH services than those who do not, after adjusting for other factors that may determine portal and SRH service use.

Methods

Study design and setting

This retrospective cohort study used EHR data at Kaiser Permanente Washington (KPWA) to examine the association between adolescent secure messaging via the online patient portal and three types of sexual and reproductive health services: orders for STI testing among all teen patients and orders for pregnancy testing and prescription contraceptive methods among teen patients presumed to be pregnancy-capable. KPWA is a not-for-profit healthcare delivery system in Washington State. Over 300,000 members receive care in 32 KPWA-owned facilities staffed by over 1000 KPWA primary care and specialty clinicians. This study was approved by the Kaiser Permanente Inter-regional Institutional Review Board.

KPWA has offered an online patient portal (MyChart) to members since 2003 – initially made available for those over 18 and for parents/caregivers of children aged 12 and younger. In 2017, two features were released for parents of teens aged 13–17 to allow them to secure message with providers and view immunization records. Adolescent members aged 13–17 were not eligible to have their own MyChart accounts until January 2019 after KPWA implemented systems to ensure confidentiality of services protected by state minor consent laws. At that time, KPWA members aged 13–17 and their parents/caregivers became eligible for “teen accounts” and “teen proxy accounts,” respectively – separate accounts with limited, parallel functionality to send and receive messages from select healthcare providers (including their primary care team and some specialists), request refills for medications not deemed confidential by the patient, and view immunization records (excluding human papillomavirus vaccine to limit proxy view of any confidential administrations). In other words, teen accounts did not display potentially confidential information such as STI test results or contraception prescriptions in order to minimize the risk for parents/caregivers to coerce their teens to provide them access to their teen accounts. A plan regarding communication of results and/or follow-up care needs is agreed upon between teen and ordering clinician at the time of testing and could include phone or portal-based outreach depending on the teen’s preference.

Adolescent members must actively register for a MyChart account. This process has evolved iteratively since 2019 to reduce barriers to registration. When first available, adolescents were required to complete a brief registration online, then verify their identity in person at any clinical facility to receive a password or elect to have a temporary password mailed to their primary household address before the account could be activated. During the COVID-19 pandemic, a phone-based activation process was made available as an alternative to in-person or mail. In 2020, teen and proxy MyChart accounts also added the functionality to join video visits directly from MyChart.

Study population

The following eligibility criteria defined the study population: aged 13 to 17 years, has had at least 3 months of continuous enrollment in a KPWA or Medicaid plan (up to 31 days allowed of administrative gaps) and has had at least 1 clinical visit at KPWA in primary or specialty care in the past 36 months OR has been paneled with a KPWA provider during the month of observation. We excluded individuals likely to have severe intellectual disability limiting their ability to communicate independently with their healthcare team. These individuals were identified using a modification of the Pediatric Medical Complexity Algorithm with at least one body system involved being neurologic.¹² The study population was designed to represent all other adolescents receiving care within our health system.

We created an analytic dataset using de-identified data from EHRs observed on the first day of each month (i.e. index day) from January 1, 2019 to December 31, 2021. Monthly observations as the unit of analysis allow for more specificity of age at time of measurement within the dynamic adolescent period and allow values for selected variables to vary with time, such as age, year, insurance status, parent proxy status, and having had a recent preventive health visit. Based on our sample definition, unique adolescents meeting inclusion criteria could potentially contribute a maximum of 36 months of observation if they were continuously enrolled in the 3 months prior to January 1, 2019 and remained enrolled with KPWA until December 31, 2021.

Measures

The goal of this analysis was to assess the association between secure messaging and receipt of SRH services. Secure messaging was defined as one or more secure messages sent from an adolescent portal account during the observed month. We considered three measures of SRH services ordered in the observed month: (1) a composite variable of any test ordered for gonorrhea, chlamydia, or HIV (the 3 most commonly ordered tests for sexually transmitted infections in our health system), (2) pregnancy test, and (3) prescription contraception ordered. We chose to measure orders rather than completed tests or prescription fills to account for the fact that testing or prescription fills may be delayed beyond the observed month or forgone based on patient preference, but an order demonstrates that the service was made available to them.

Covariates (categorical unless noted) included age in years during the observed month (continuous variable), gender defined using administrative sex and an optional gender identity field (female sex, not transgender or gender diverse [TGD]/gender not recorded; male sex, not TGD/gender not recorded; female sex, TGD; male sex, TGD); racial and ethnic identity (non-mutually exclusive options in administrative data: Asian, Black/African American, Pacific Islander or Native Hawaiian, American Indian or Alaska Native, Hispanic, more than one race selected, white non-Hispanic only, other [not Hispanic], or no race or ethnicity recorded); insurance status (HMO/PPO or Medicaid); parent proxy account (active or inactive/absent); secure messaging prior to observed month (prior use or no prior use); and preventive care visit in past 18 months. Gender data was abstracted from a sexual orientation gender identity (SOGI) form within the EHR, completed by clinical team members and based on self-report from patients during clinical encounters. Race and ethnicity data was included in our analysis to identify differences in SRH service use that may suggest barriers to access and also to adjust for the known disparities in secure messaging by race and ethnicity in our multivariable analyses.⁷ We chose to not collapse race and ethnicity variables into mutually exclusive categories because the data collection process does not allow patients to self-select into mutually exclusive categories, and we did not want to incorrectly assign multiracial individuals.^{13, 14}

Analytic methods

We summarized demographic characteristics of the overall sample and among those assumed to be pregnancy-capable based on administrative sex being female, with subgroups defined by outcome. Generalized estimating equations (GEEs) with a logistic link estimated unadjusted and adjusted odds ratios (aORs) and 95% confidence intervals (CI) for associations between secure messages sent and the 3 types of SRH services: (1) STI tests ordered in the observed month, and among pregnancy-capable individuals, (2) pregnancy tests and (3) prescription contraceptive methods ordered. GEE models were adjusted for baseline covariates described above. Correlation of monthly observations within an individual was modeled via robust sandwich variance estimates (to ensure robustness to any type of correlation structure), obtained by assuming a working independence correlation structure. In order to examine the association with first-time portal use during the observed month, we also conducted GEE analysis accounting for a potential interaction by no prior (“first-time”) or prior portal use.

Since a calendar month is an arbitrary time frame for constricting secure messaging incidents that occurred in close proximity to SRH services ordered during the month of observation, two sensitivity analyses were conducted. In the first analysis, we defined secure messaging as having occurred in the observed month or month prior as the variable of interest. In the second analysis, we defined prior secure messaging as having occurred more than one month prior to the month of observation.

Results

The study sample included 37,818 unique individuals who contributed 667,678 months of individual observation (Figure 1). 18,499 individuals (48.9%) were pregnancy capable, contributing 326,061 months of observation. Table 1 shows descriptive characteristics of all months of observation included in the analysis. 1.2% of all observed months had messages sent from adolescent portal accounts, and among pregnancy-capable individuals, 1.6% of observed months had messages sent. Our sample represented a broad range of racial and ethnic identities. The majority had private insurance (83%), and roughly half (48%) had had a preventive care visit in the past 18 months. Results of bivariate associations between secure messaging and pregnancy tests, contraception, and STI tests ordered during month of observation are displayed in Supplemental Table 5.

Figure 1. Flow diagram to develop analytic dataset.

Abbreviations: PCP=primary care provider

Table 1.

Overall sample characteristics

	Observed months among all individuals		Observed months among pregnancy-capable individuals
	N	%	N	%
1 or more messages sent from teen portal
No	659,883	98.8	320,980	98.4
Yes	7,795	1.2	5,081	1.6
Calendar year
2019	216,187	32.4	105,875	32.5
2020	225,521	33.8	110,083	33.8
2021	225,970	33.8	110,103	33.8
Age (years)
13	133,407	20.0	65,036	19.9
14	134,610	20.2	65,705	20.2
15	134,990	20.2	66,212	20.3
16	137,658	20.6	67,178	20.6
17	127,013	19.0	61,930	19.0
Gender
Male sex, not TGD	339,340	50.8	-	-
Female sex, not TGD	317,852	47.6	317,852	97.5
Male sex, TGD	2,277	0.3	-	-
Female sex, TGD	8,209	1.2	8,209	2.5
Race & Ethnicity*
Asian	105,385	15.8	52,865	16.2
Black/African American	71,105	10.6	36,113	11.1
Pacific Islander and Native Hawaiian	17,856	2.7	8,790	2.7
American Indian and Alaska Native	12,999	1.9	6,759	2.1
Hispanic	58,771	8.8	29,234	9.0
White non-Hispanic only	343,707	51.5	167,419	51.3
More than one race**	65,753	9.8	33,654	10.3
Other (not Hispanic)	16,481	2.5	7,703	2.4
No race or ethnicity recorded	66,615	10.0	30,189	9.3
Insurance status
Private (HMO/PPO)	556,220	83.3	272,469	83.6
Medicaid	111,458	16.7	53,592	16.4
Preventive care visit in past 18 months
No	344,812	51.6	167,378	51.3
Yes	322,866	48.4	158,683	48.7
Prior portal use for secure messaging
No	641,417	96.1	309,949	95.1
Yes	26,261	3.9	16,112	4.9
Parent proxy account active
No proxy account, or not active	429,124	64.3	208,921	64.1
Yes	238,554	35.7	117,140	35.9
Total	667,678	100	326,061	100

Abbreviations: TGD = transgender or gender diverse, HMO/PPO: Health Maintenance Organization/Preferred Provider Organization

^*Race/ethnicity categories are not mutually exclusive.

^**More than one race/ethnicity category was reported. These observations are also counted in individual race/ethnicity categories, so percentages will add up to more than 100

The unadjusted and adjusted regression analyses are presented in Table 2 for all individuals and Tables 3 and 4 for pregnancy-capable individuals. In our adjusted analysis of all observed months, months with secure messages sent had significantly greater odds of STI tests ordered in the same month than observed in months without secure messaging (aOR 3.5, 95% CI 2.9, 4.3). Analyses including prior secure messaging as an interaction term showed significantly greater odds of STI tests ordered among observations with no prior secure messaging, i.e. “first-time” use, (aOR 4.9, 95% CI 3.9–6.2) than observations with prior secure messaging (aOR 2.5, 95% CI 1.9–3.3).

Table 2.

Unadjusted and adjusted odds of secure messaging among observed months with STI tests orders compared to those without, stratified by history of portal use for secure messaging

	No STI test (GC/CT or HIV) orders	1 or more STI test (GC/CT or HIV) orders	Unadjusted analysis			Adjusted analysis
	N (%)	N (%)	OR	95% CI	p-value**	OR	95% CI	p-value**
1 or more messages sent from patient portal
No	657,070 (98.9)	2,813 (93.9)	1.00	ref		1.00	ref
Yes	7,611 (1.1)	184 (6.1)	5.65	4.80, 6.64		3.54	2.92, 4.30
Messages sent during observed month among first-time portal users					0.000			0.000
No	635,707 (99.5)	2,651 (97.0)	1.00	ref		1.00	ref
Yes	2,978 (0.5)	81 (3.0)	6.52	5.21, 8.17		4.88	3.86, 6.17
Messages sent during observed month among prior portal users
No	21,363 (82.0)	162 (61.1)	1.00	ref		1.00	ref
Yes	4,633 (17.8)	103 (38.9)	2.93	2.25, 3.82		2.53	1.95, 3.29

^**p-value for interaction term

OR=odds ratio; CI=Confidence interval

Table 3.

Unadjusted and adjusted^* odds of secure messaging among observed months^* with pregnancy test orders compared to those without, stratified by history of portal use for secure messaging

	No pregnancy tests orders	1 or more pregnancy test orders	Unadjusted analysis			Adjusted analysis
	N (%)	N (%)	OR	95% CI	p-value**	OR	95% CI		p-value**
1 or more messages sent from patient portal
No	316,725 (98.5)	4,255 (94.4)	1.00	ref		1.00	ref
Yes	4,828 (1.5)	253 (5.6)	3.90	3.33, 4.57		2.62	2.22, 3.10
Messages sent during observed month among first-time portal users					0.000				0.000
No	304,125 (99.4)	3,963 (97.7)	1.00	ref		1.00	ref
Yes	1,769 (0.6)	92 (2.3)	3.99	3.23, 4.94		3.46	2.78, 4.30
Messages sent during observed month among prior portal users
No	12,600 (80.5)	292 (64.5)	1.00	ref		1.00	ref
Yes	3,059 (19.5)	161 (35.5)	2.27	1.83, 2.82		2.17	1.73, 2.72

^*Among pregnancy-capable adolescents

^**p-value for interaction term

OR=odds ratio; CI=Confidence interval

Table 4.

Unadjusted and adjusted^* odds of secure messaging among observed months^* with contraception prescription/procedure orders compared to those without, stratified by history of portal use for secure messaging

	No contraception prescriptions/procedure orders	1 or more contraception prescriptions/procedure orders	Unadjusted analysis			Adjusted analysis
	N (%)	N (%)	OR	95% CI	p-value**	OR	95% CI	p-value**
1 or more messages sent from patient portal
No	310,972 (98.6)	10,008 (95.0)	1.00	ref		1.00	ref
Yes	4,549 (1.4)	532 (5.0)	3.63	3.24, 4.08		1.94	1.74, 2.17
Messages sent during observed month among first-time portal users					0.000			0.000
No	298,983 (99.4)	9,105 (98.0)	1.00	ref		1.00	ref
Yes	1,674 (0.6)	187 (2.0)	3.67	3.15, 4.27		2.85	2.42, 3.35
Messages sent during observed month among prior portal users
No	11,989 (80.7)	903 (72.4)	1.00	ref		1.00	ref
Yes	2,875 (19.3)	345 (27.6)	1.59	1.38, 1.84		1.62	1.40, 1.87

^*Among pregnancy-capable adolescents

^**p-value for interaction term

OR=odds ratio; CI=Confidence interval

In our adjusted analyses of observed months among pregnancy-capable individuals, months with secure messages sent had significantly greater odds of both pregnancy test and prescription contraception orders than observed months without secure messages sent. (pregnancy test orders aOR 2.6, 95% CI 2.2, 3.1; prescription contraception orders aOR 1.9 95% CI 1.7, 2.2). Analyses including prior secure messaging as an interaction term showed significantly greater odds of pregnancy test and prescription contraception orders among observations with no prior secure messaging, i.e. “first-time” use, (pregnancy test orders aOR 3.5, 95% CI 2.8, 4.3; prescription test orders aOR 2.8, 95% CI 2.4, 3.3) than observations with prior secure messaging (pregnancy test orders aOR 2.2, 95% CI 1.7, 2.7; prescription contraception orders aOR 1.6, 95% CI 1.4, 1.9).

Findings for sensitivity analyses examining the association between secure messaging in the observed or prior month and SRH services can be found in Supplementary Tables 6–9.

Discussion

This study demonstrated that observed months in which teens aged 13–17 sent secure messages in our health system had greater odds of SRH services orders than observed months without messages. Our findings showed that overall, a small proportion of observed months involved secure messages sent from adolescent portal accounts. Our analyses examining the interaction of prior secure messaging suggest that needing or having recently received SRH services may prompt adolescents to send messages for the first time using a patient portal.

These findings imply that access to secure messaging on patient portals may support linkages to SRH care for adolescents in integrated health systems. Our study addresses an important gap in examining the process-outcome link pertaining to secure messaging and adolescent health services outcomes. Previously, Huang and colleagues examined outcomes of medical note sharing among adolescents with chronic disease.¹⁵ Youth in their survey study who used a portal reported high satisfaction and adequate comprehension with documentation at visits. We did not identify prior studies among adolescents that explored the association between portal use and receipt of SRH services.

Our study findings align with others that show low but increasing rates of adolescent engagement on the patient portal.¹⁶ Thompson et al reviewed over 2 years of teen and parent portal enrollment in a Florida-based academic pediatric health system, finding that only 3% of the enrolled adolescent population used the portal.¹⁷ Calman and colleagues described in an implementation study that 11% of eligible patients aged 10–17 had activated a portal, compared to 31% of the general population in a New York-based academic health system.¹⁸ In an Australian study, Torrens et al identified that adolescents were least likely to register for a portal account when compared to any other age group.¹⁹ These studies focused on enrollment, which is a necessary but not sufficient element of adolescent portal use to meet their SRH care needs. While a number of studies have highlighted benefit and satisfaction from teens with secure messaging use,^{18, 20–22} these studies taken together with our findings suggest a need for support and awareness building for teens to both overcome enrollment barriers and also become comfortable using the portal to engage with their healthcare team, particularly for confidential care needs as they transition to adulthood.²³ Notably, a survey study by Wright and colleagues of over 300 emerging adults aged 18–29 showed that three-quarters of respondents had used at least one portal function, underlining the importance of this skill building during adolescence as a key aspect of transition readiness.^{24, 25}

In qualitative studies, low portal access among youth is attributed to lack of awareness among adolescents, concerns about confidentiality, ^{21, 26} and lack of engagement with young people when designing tools and workflows.^{23, 27} Over the past decade, the adolescent health professional community has underscored the critical importance of protecting adolescents’ confidential health information in the context of online patient portals and particularly parent proxy access.²⁸ Among clinicians and health informatics leaders, there have been calls for institutional resources and commitment to build awareness and engagement, and to add adolescent health-centered features and standard practices within EHR products in order to deliver on the promise of confidentiality for adolescent patients who need it.^{29, 30} Yet providers who care for adolescents continue to face challenges in engaging with them on the patient portal, and more research and programming are needed.³¹

Strengths and limitations

We leveraged a large sample of adolescents receiving community-based care in an integrated care and coverage model, using several years of utilization data. To our knowledge, no other quantitative studies have examined provision of specific services in association with portal use in adolescents. Our sample included a diverse population of adolescents in terms of gender, racial and ethnic identities, although the proportion of individuals with private insurance (83%) was higher than national estimates of all children (66%).³² We acknowledge that our use of administrative sex on EHR data may incorrectly categorize transfeminine patients who have changed the sex listed on their legal documents but may not be capable of pregnancy or use prescription contraception. Furthermore, given our approach, we could not confirm that messages from adolescent portal accounts were written by the adolescents themselves. Of note, a 2021 study of more than 3000 eligible adolescent accounts across 3 institutions found that 64 to 75% of adolescent portal accounts with outbound messages were accessed by parents/caregivers. KPWA’s verification process may prevent parent/caregiver users from accessing adolescent accounts, but the fidelity of this in our study setting is unknown. Our study design did not allow for causal inference. In fact, secure messaging could occur after an SRH service encounter during the month of observation (i.e. to inquire about results or navigate follow-up needs). We did not examine the content of secure messages to confirm whether message exchanges were related to communicating results and/or recommending treatment. For this reason, we were unable to determine whether concerns about confidentiality and/or urgency of follow-up care influenced the use of a portal versus other means of communication (e.g. phone) to address SRH care needs. However, through examining the interaction of prior portal use we demonstrated that first-time secure message users were more likely to have an SRH service ordered during the month of observation than prior secure message users. This suggests that secure messaging might be a key factor in facilitating SRH services access, but more research is needed to confirm this hypothesis. We did not measure actual SRH testing or pharmacy fills but rather orders for these services, so are also unable to examine the association between secure messaging and receipt of services. Additional analyses would be needed to differentiate whether secure messaging is more strongly associated with SRH care than other types of care. Finally, SRH services received outside of our health system or not requiring a prescription (e.g. emergency contraception, condoms) were not captured. Adolescents in need of confidential SRH care may intentionally seek care from a community (non-KPWA) provider, so our sample may be biased to adolescents who choose to involve their parent/caregiver in their SRH care.

Next steps

Substantial gaps remain in our understanding of the association between secure messaging and other digital healthcare tools and health outcomes. Future research should include qualitative studies specific to portal use and SRH care among youth and the development and study of interventions to promote adolescent use of patient portals for confidential and non-confidential healthcare needs. Given that parent proxy support becomes unavailable at age 18, all adolescents need support to ensure they have skills to independently access portal-based care and maintain health and well-being as young adults.

Abbreviations:

95% CI

95% confidence interval

KPWA

Kaiser Permanente Washington

HMO

Health Maintenance Organization

PPO

preferred provider organization

TGD

transgender or gender diverse

GEE

generalized estimating equations

RR

relative risk

aRR

adjusted relative risk