a b s t r a c t

Objective: Patient portal (PP) use has rapidly increased in recent years. However, the PP use status among house- less patients is largely unknown. We aim to determine 1) the PP use status among Emergency Department (ED) patients experiencing houselessness, and 2) whether PP use is linked to the increase in patient clinic visits.

Methods: This is a single-center retrospective observational study. From March 1, 2019, to February 28, 2021, houseless patients who presented at ED were included. Their PP use status, including passive PP use (log-on only PP) and effective PP use (use PP of functions) was compared between houseless and non-houseless patients. The number of clinic visits was also compared between these two groups. Lastly, a multivariate logistic regression was analyzed to determine the association between houseless status and PP use.

Results: We included a total of 236,684 patients, 13% of whom (30,956) were houseless at time of their encounter. Fewer houseless patients had effective PP use in comparison to non-houseless patients (7.3% versus 11.6%, p < 0.001). In addition, a higher number of clinic visits were found among houseless patients who had effective PP use than those without (18 versus 3, p < 0.001). The adjusted odds ratio of houseless status associated with PP use was 0.48 (95% CI 0.46-0.49, p < 0.001).

Conclusions: Houselessness is a potential risk factor preventing patient portal use. In addition, using patient portals could potentially increase clinic visits among the houseless patient population.

(C) 2023

1. Introduction

With the rapid development of the internet, Health information technology (HIT) has been developed and used widely in the field of healthcare across the country in recent years [1,2]. For example, over 96% of hospitals and 86% of physician offices now use electronic health records (EHR) and the majority of EHR systems are tethered to the pa- tient portal (PP) [3,4]. Recent studies showed that EHRs have been widely adopted by healthcare providers (HCP) whereas patient portal

* Corresponding author at: John Peter Smith Health Network, 1500 S. Main St., Fort Worth, TX 76104, USA.

E-mail addresses: [email protected] (C. Holmes), [email protected] (K. Holmes), [email protected] (J. Scarborough), [email protected] (J. Hunt), [email protected] (J.P. d’Etienne), [email protected] (A.F. Ho), [email protected] (N. Alanis), [email protected] (R. Kirby), [email protected] (C.D. Schrader), [email protected] (H. Wang).

has not [3,5]. Lower PP use has been attributed to several reasons [6,7]. Certain Vulnerable patient populations have lower rates of PP use. These vulnerable populations include those who are older, who have less internet access, and who have low socioeconomic status , etc. [6-9]. At present, the status of PP use among patients who experiencing houselessness is largely unknown.

Houselessness is a special condition. Individuals experiencing houselessness tend to have less healthcare support including access and engagement to healthcare, communication with healthcare pro- viders, and shared decision-making on their health conditions [10,11]. On the other hand, houseless individuals often lack positive self- management behaviors which could lead to poor healthcare out- comes [12,13]. In the past, interventions focusing on promoting healthcare outcomes among individuals experiencing houselessness included providing homes, assigning individuals to primary care physi- cians, and establishing dedicated outreach homeless programs, all showed certain levels of improvement, but are quite passive [14-17].

https://doi.org/10.1016/j.ajem.2023.01.029

0735-6757/(C) 2023

These interventions are healthcare provider/third-party driven. The promotion of automaticity among houseless individuals is still lacking. With the recent studies focusing on PP use and healthcare outcomes, results showed that patients who used PP were associated with having better HCP communication [18,19]. Better patient-provider communi- cation is recommended in healthcare and associated with better Disease outcomes [18,19]. More importantly, patients who are using PP demon- strate the enhanced self-management behavior. Positive disease self- management behaviors are also associated with better health outcomes [20,21]. Therefore, it is important to determine the current PP status

among the houseless patient population.

These improved healthcare outcomes include healthcare access (e.g., being able to visit healthcare facilities), healthcare engagement (e.g., having primary care physicians assigned), and healthcare utiliza- tion (e.g., clinical appointments). Specifically, among patients seen in the Emergency Department (ED), with the improvement of PP use, we expect to see an improved transition from ED discharge to clinic visits. If this is proved effective in the houseless population, interventions for increased PP use can be implemented to maximize the HIT benefit. Therefore, in this study, we aim to 1) determine the PP use status among ED patients experiencing houselessness, and 2) determine whether PP use is linked to an increase in subsequent patient clinic visits.

1. Methods
   1. Study design and setting

This is a single-center retrospective observational study. The study hospital is a public-funded, urban tertiary referral center, a level-1 trauma center, a chest pain center, and a comprehensive stroke center. The hospital network includes over 50 primary care and subspecialty clinics and one dedicated houseless clinic across Tarrant County in the Fort Worth, Texas area covering an estimated two million people. The hospital ED has approximately 120,000 patient visits per year. This study has been approved by the regional institutional review board (IRB# 1600198-2) with waived informed consent.

• 1. Study participant

We collected patient socioDemographic and clinical data via EHR and PP from March 1, 2018, to February 28, 2021. This study enrolled all patients who presented to the study hospital ED from March 1, 2019, to February 28, 2021. For each patient, we collected information on the patient’s historical use of PP in the past 12 months from the pa- tient index ED visit (e.g., if a certain patient presented at ED on 1/1/ 2020, we reviewed this patients’ PP use history from 1/1/2019 to 12/ 31/2019). Considering PP use can be changed dynamically if the same patient had multiple ED visits during the study period, each ED visit was counted as a separate visit. Since all data was collected via EHR with minimal incomplete (i.e., missing, incorrect, or error) information from very few key variables, patients with incomplete key variable information were excluded from this study.

• 1. Data retrieval and validation

Since all study data were retrieved from the EHR, we performed data validation. Two dedicated people in the field of Medical Informatics and information technology retrieved and validated study data and were initially blinded to this study (i.e., before the main results were open for all the individuals who participated in this project). In addition, we also randomly selected 20 patient datasets from the entire dataset to manually check and validate the accuracy of data retrieval.

• 1. Variables

Key variable: patients experiencing houselessness. We included all houseless patients presenting to the study hospital ED during the study period. Houseless patients were defined as those who met the US Department of Housing and Urban Development (HUD) definition of houselessness at the time of entry into EHR system [16]. Such patients were initially identified via our EHR system and then further paired with the Tarrant County Houseless Management Information System (HMIS) database archived in Fort Worth, TX, USA. In addition to syste- matically identifying houseless patients, the street medicine team (a dedicated care team for houseless patients) at the study hospital had the ability to further identify houseless patients who may have been missed due to data matching issues or did not exist in the HMIS system. For situations where data did not match between systems, houseless- ness status was verified by the street medicine team, and patients were entered into EHR manually.

Other variables: we collected Patient demographic characteristics:

1. age; 2) sex (male and female); 3) race/ethnicity (divided into 4 groups based on the definition of basic racial and ethnic categories from Federal Statistics and Program Administrative Reporting:
   1. Non-Hispanic White, b) Non-Hispanic Black, c) Hispanic/Latino, and

d) others). The other race/ethnicity category included: American Indian, Alaska Native, Asian or Pacific Islanders. Patients’ social characteristics included 1) insurance status in two categories: patients’ having no insurance coverage (self-pay) and patients’ having any insurance. Such insurance includes government-assisted insurance (e.g., Medicaid, Medi- care, etc.), commercial (e.g., Cigna, BCBS, United Health, etc.), hospital- issued insurance (e.g., COOKs, study hospital sponsored insurance), and special insurance (e.g., work-related compensation, VA insurance, etc.),

1. patient Preferred language in two categories: English versus non- English. Since there are over 80 different languages, we included all patients who preferred non-English into one group. Patient healthcare in- formation: 1) presence of chronic disease conditions: we categorized patients into three groups: a) patients without any chronic condition;

b) patients with only one chronic condition; and c) patients with multimorbidity. 2) dedicated Primary care physician : a) yes, and

b) no. In addition, we also included one coronavirus disease 2019 (COVID-19) variable: Whether patients presented to ED prior to the COVID-19 pandemic (yes versus no). The COVID-19 pandemic refers to March 11, 2020, World Health Organization (WHO) announcement of the US COVID-19 pandemic. We categorized into two groups:

• 1. patients who visited ED prior to the COVID-19 pandemic and
  2. patients who visited ED during the COVID-19 pandemic.
  3. Outcome measures

Study primary outcome measures are 1) Passive PP use (i.e., only logged on to the PHR). This was defined as patients who only logged on their PP (i.e., Mychart). Mychart is a PP used in the study hospital sys- tem and is tethered with EHR (Epic). If patients logged on to their Mychart without using any functions, this will be defined as passive PP use. For example, a certain patient only used Mychart to review their laboratory or image results, check their billing information, or review their clinical appointments without doing anything. 2) Ef- fective PP use (i.e., using PP functions). This was defined as patients using at least one of the three basic PP functions: a) using PP to com- municate with HCPs, b) using PP to arrange new clinical appoint- ments or revise existing appointments, and c) using PP to refill patient medications. Upon patient presentation at the ED, PP use statuses in the past 12 months were collected. Our secondary out- come is to measure the number of clinic visits in the past 12 months and compare patient clinic visits between patients who use PP versus ones not.

• 1. Data analysis

We analyzed the continuous variable (i.e., age, number of clinical visits) using both Student t-tests for mean (standard deviation) com- parison and Wilcoxon Rank Sum test for median (interquartile range) comparisons between patients experiencing houselessness (houseless group) and ones without (non-houseless group). Meanwhile, we use the Pearson Chi-Square test for categorical data (e.g., sex, race/ethnicity, Mychart use, etc.) comparisons between these two groups. We used multivariate logistic regressions to examine the association between PP use and patients’ houseless status with the adjustment of age, sex, race/ethnicity, insurance status, preferred language, PCP status, chronic conditions, and COVID-pandemic. Such associations were determined by adjusted odds ratio (AOR) with 95% confidence intervals (CI). All analyses were performed using STATA 16.0 software (College Station,

Table 1

General Characteristics of Study Patient Information.

Patients experiencing houselessness

	(n = 30,956)	(n = 205,692)
Age
Mean (SD) year	48 (13)	43 (17)
Median (IQR)	50 (39,58)	42 (29, 56)
Sex
Male (n, %)	19,791 (64)	100,994 (49)
Female (n, %)	11,165 (36)	104,698 (51)
Race/ethnicity — (n, %) Non-Hispanic White	14,046 (45)	64,988 (32)
Non-Hispanic Black	13,249 (43)	65,319 (32)
Hispanic/Latino	3216 (10)	64,666 (31)
Others *	445 (1.4)	10,719 (5.2)

Insurance Status — (n, %)

Patients without houselessness

TX) with a p-value <0.05 considered statistically significant.

1. Results

English	30,596 (99)	172,752 (84)
Non-English	360 (1.2)	32,940 (16)
Due to the study hospital ED’s unique features (urban, public-funded Primary care physician — (n, %)

Have any insurance

No insurance (self-pay) Unknown

Preferred language — (n, %)

24,741 (80)

6215 (20)

31 (0.1)

129,895 (63)

75,797 (37)

371 (0.2)

tertiary referral center), we have 30,956 patients experiencing house-	Yes	21,007 (68)	106,663 (52)
lessness presented at ED seeking for healthcare, accounting for approx-	No	9949 (32)	99,029 (48)
imately 13% of total ED visits. The detailed study flow diagram is shown Chronic disease conditions —
in Fig. 1.	(n, %) No chronic condition	2760 (8.9)	60,578 (29)
Patients who are experiencing houselessness (referred to as house-	One chronic condition	3887 (13)	44,018 (21)
less patients) tended to be elderly, male, English speaking, and	Multimorbidity	24,309 (79)	101,096 (49)
Non-Hispanic White as compared with patients without houselessness	ED Visits — (n, %)

(referred to as non-houseless patients, Table 1). Houseless patients in this study also had higher insurance coverage and higher rates of PCP assignment. In addition, more houseless patients had multimorbidity in comparison to non-houseless patients. A similar trend occurred with decreased ED visits after the COVID-19 pandemic (Table 1).

Similar passive PP use rates are found in both the houseless and non- houseless patients. However, both groups show the low involvement of passive PP use (<10%). When effective PP use was analyzed, less effec- tive PP use rate was found among houseless patients in comparison to the non-houseless patients (7.3% vs. 11.6%, p < 0.001). Such less effective PP use was found across all three basic PP functions. Meanwhile, the study also shows that the use of one PP function (i.e., refill medication) was extremely low in both groups (Table 2). Overall PP use is 19.9% among non-houseless patients and 15.4% among houseless patients (p < 0.001, Table 2).

When the number of clinic visits was measured and compared between patients who used PP versus ones who did not use PP, we

Abbreviations: n, number; COVID-19, Coronavirus disease 2019. *other race/ethnicity include American Indian, Alaska native, Asian or Pacific Islanders.

Prior to COVID-19 pandemic	18,325 (59)	114,066 (55)
After COVID-19 pandemic	12,631 (41)	91,626 (45)

found that the number of clinic visits was higher among patients who used PP regardless of their houseless status (Table 3). In addition, the number of clinic visits was even higher among patients who use PP effectively (Table 3).

To determine the association between houseless status and overall PP use, both univariate logistic and multivariate logistic regression anal- yses were performed. We find that risks affecting overall PP use are multifactorial. Female gender, having insurance coverage, English lan- guage preference, presence of chronic conditions, and COVID-19 all in- crease patients’ use of PP at various levels. Hispanic or Non-Hispanic Black race/ethnicity and being houseless decreased patients’ use of PP

2,074 patients were excluded due to missing /incorrect

/error key variables including sex(18), race/ethnicity

(1,385), insurance (901), and language (1,014)

Effective PP use: 2,267 Patients

Passive PP use: 2,494 Patients

Effective PP use: 23,867 Patients

Passive PP use: 17,090 Patients

30,956 Patients had

experienced homelessness

205,692 Patients without experiencing homelessness

236,648 Patients enrolled in a final analysis

A total of 238,722 Patients

(from March 1,2019 to February 28, 2021)

Fig. 1. Study flow diagram.

Table 2

A Comparison of Patient Portal Use Between Houseless and Non-Houseless Patients.

Houseless Patients (n = 30,956)

Non-Houseless Patients (n = 205,692)

P value

Overall PP use — (n, %) 4761 (15.4) 40,957 (19.9) <0.001

Passive PP use — (n, %) 2494 (8.1) 17,090 (8.3) 0.134

Effective PP use — (n, %)

Use PP to Communicate with HCPs

Use PP to arrange/revise clinical appointments Use PP to refill medications

Abbreviations: n, number; HCP, Healthcare provider; PP, Patient Portal.

2267 (7.3)

2259 (7.3)

917 (3.0)

33 (0.1)

23,867 (11.6)

23,771 (11.6)

9215 (4.5)

369 (0.2)

<0.001

<0.001

<0.001

0.004

Table 3

A Comparison of Patient Clinic Visits Between Patients with and without Patient Portal Use.

rapidly [1,22]. However, healthcare disparity still occurred among some vulnerable populations. It is reported that the use of patient por-

Houseless Patients Number of clinic visits Median (IQR)

Non-Houseless patients Number of clinic visits Median (IQR)

P value

tals (PP) is still lacking among some minorities [6,18]. However, the sta- tus of PP use among the houseless patient population is rarely reported. In this study, we found similar findings of current PP use status among

Overall PP use 26 (13, 45) 14 (6, 28) <0.001

Overall Non-PP use 17 (6, 38) 2 (0, 9) <0.001

Effective PP use	30 (16, 48)	18 (9, 34)	<0.001
Non-Effective PP use	17 (6, 38)	3 (0,10)	<0.001

Abbreviations: n, number; SD, Standard deviation; IQR, Interquartile range; PP, patient portal.

(AOR 0.48, 95% CI 0.46-0.49, p < 0.001) even after adjusting for all other independent factors (Table 4).

1. Discussion

In recent years, with the implementation of HITECH and ACA (Af- fordable Care Act), health information technology (HIT) has grown

Table 4

Model to predict patient portal (MyChart) usage.

UOR		95% CI	AOR	95% CI
Patient houseless condition Patients with no houseless Reference Reference Reference Reference condition
Patients experiencing	0.73	0.71-0.76	0.48	0.46-0.49
houselessness
Age	1.01	1.01-1.01	0.99	0.99-0.99
Sex
Male	Reference	Reference	Reference	Reference
Female	2.07	2.03-2.11	1.81	1.77-1.85
Race/ethnicity
Non-Hispanic White	Reference	Reference	Reference	Reference
Non-Hispanic Black	0.71	0.70-0.73	0.59	0.58-0.61
Hispanic/ Latino	0.72	0.70-0.74	0.82	0.80-0.85
Others	1.16	1.11-1.22	1.17	1.11-1.23
Insurance status

Non-insurance (self-pay) Reference Reference Reference Reference Any insurance 3.71 3.61-3.81 2.17 2.11-2.24

Preferred language

Non-English Reference Reference Reference Reference

English 1.52 1.48-1.58 1.70 1.63-1.77

Primary care physician

the houseless patient population. Houseless patients are less likely to

use PP, especially on effective PP use. Though other risks (e.g., other so- ciodemographic features, primary care physician status, patient chronic diseases, etc.) can still affect patients’ using PP. Being houseless is one of the independent risks preventing patients from using PP in general. Therefore, interventions advocating the use of PP should be emphasized among houseless patient populations. Meanwhile, we also found pa- tients who used PP were associated with higher clinic visits regardless of their houseless status. This indicates the potential link between PP use and clinic outcomes. Our study adds extra evidence for HIT use in another vulnerable population and recognizes HIT disparity among the houseless patient population which has not been reported before in the literature. Our findings can serve as a foundation for future inter- vention studies to further improve healthcare quality among the house- less population.

In this study, we divided PP use into two major categories: passive and effective PP use. We believe effective PP use may be more closely re- lated to patient clinical outcomes, especially among ones with chronic conditions. Houseless patients tend to have high rates of multimorbidity [15,23]. Houseless patients are also reported to have a high incidence of substance abuse or psychiatric disorders [24,25]. With such higher psychosocial risks, houseless patients often have suboptimal clinical outcomes (e.g., higher mortality, higher hospitalizations, higher disease complications, etc.) [10,15,23]. On the other hand, disease self- management behavior plays an important role in chronic disease control [20,21]. Houseless patients often lack positive self-management behavior [12,13]. However, effective PP use can act as one of the positive self-management behaviors, which allows patients to be actively involved in managing their personal healthcare and potentially improve their chronic disease care. This study’s second outcome shows the increased clinic visits among PP users can indirectly reflect such an association. However, no available data in this study can prove this causative effect between PP use and increased clinic visit, and future study is needed to further investigate the direct association.

With the implementation of the Affordable Care Act, some states (e.g., California) expanded their Medicaid program to houseless patients

Not having primary care physician

Reference Reference Reference Reference

[26]. In addition, some houseless programs also provided primary care physician (PCP) coverage [14]. Our study showed similar findings.

Having primary care physician 5.64 5.49-5.79 3.99 3.87-4.10

Chronic disease condition

The study hospital has abundant experience in assisting healthcare

among houseless patients in the past, interventions that help house-

Patients with no chronic condition

Patients with one chronic condition

Reference Reference Reference Reference

1.97 1.89-2.04 1.80 1.73-1.88

less patients to get engaged in healthcare include 1) providing hospital-sponsored healthcare insurance coverage; 2) assigning PCPs to houseless patients, and 3) providing a dedicated houseless

Patients with multimorbidity 4.21 4.08-4.35 3.27 3.15-3.39 COVID-19 pandemic

(March 11,2020)

Prior to the COVID-19 pandemic Reference Reference Reference Reference After the COVID-19 pandemic 1.36 1.33-1.39 1.42 1.39-1.45

Abbreviations: UOR, Unadjusted odds ratio; CI, Confidence interval; AOR, Adjusted odds ratio; COVID-19, Coronavirus disease 2019.

clinic. We believe these interventions resulted in increased insur- ance coverage and higher rates of PCP assignment among the house- less patient population in this study. Other key sociodemographic characteristics are similar when compared to previous studies. The trend of houseless patients tends to be elder, male predominant, with more multimorbidities [27,28]. In addition, we found that

fewer ED visits occurred after the COVID-19 pandemic, a similar trend was reported in other studies [29,30].

In our study, our secondary outcomes measured the number of clinic visits. We found that the number of clinic visits was much higher among houseless than non-houseless patients. This may partially be due to the interventions that the study hospital performed. The study hospital has a dedicated houseless clinic that mainly sees all houseless patients [17]. This specific clinic is a walk-in clinic with no appointment needed though it is preferred to make appointments beforehand [17]. In addi- tion, the hospital has a unique street medicine outreach program that allows physicians to go out to shelters or outside where houseless pa- tients commonly gather to register them online and provide healthcare at the site. These events also count for patient clinic visits which result in relatively higher clinic visits in comparison to non-houseless patients.

• 1. Limitations

Our study has its limitations. First, this is a single-center retrospec- tive observational study, and potential incomplete data (e.g., incorrect, missing, error) data may subsequently affect the study analysis. Second, our system updates patient houseless statuses periodically, but not in real-time. Patient houseless status can change dynamically, so this data lag can cause relatively incorrect houseless information which could affect our study findings. Third, we did not exclude patients who may not have access to PP. In our healthcare system, PP was pro- vided after patients’ first healthcare visit. However, most of our house- less patients had already visited the study healthcare system with their PP access prior to the current enrollment period (i.e., from March 1, 2019, to February 28, 2021, there were 96.3% of houseless patients had at least one clinic visit 12 months prior to the index ED visit, data not shown). With limited data, we were not able to determine whether the remaining 3.7% of patients had either no PP access or had their PP ac- cess prior to March 1, 2018. Due to the extremely small portion of the sample size, we analyzed our data with the assumption that all patients had their PP access before their index ED visits. Fourth, there are many other PP user functions (e.g., uploading personal monitored data to PP, downloading disease-related knowledge, etc.) that we did not analyze in this study. We only investigated three essential PP functions which may not be enough to determine the current PP status. However, with very little use of these three essential PP functions, we assumed that very few houseless patients may otherwise use more advanced PP func- tions. We are investigating other PP functions in our ongoing study and will report our findings in the future. Fifth, there are other specific risks that might affect patients’ using PP such as whether patients have access to internet, whether patients have a smartphone, education level, and income level, and such risks are not analyzed in this study. Lastly, this study only included ED houseless patients, but houseless patients that presented at other Healthcare facility settings were not involved. There- fore, a large-scale prospective multi-center study is warranted to fur- ther investigate the PP use among the houseless patient population.

1. Conclusions

Houselessness is a potential risk factor preventing patients from using patient portals. However, using patient portals could potentially increase clinic visits among the houseless patient population.

Ethical approval statement

This study was approved by the UNTHSC regional IRB.

CRediT authorship contribution statement

Chad Holmes: Writing - review & editing, Writing - original draft, Supervision, Investigation, Data curation, Conceptualization. Katherine Holmes: Writing - review & editing, Validation, Investigation,

Conceptualization. Jon Scarborough: Writing - review & editing, Vali- dation, Methodology, Investigation, Data curation. Joel Hunt: Writing - review & editing, Validation, Resources, Methodology, Investigation, Conceptualization. James P. D’Etienne: Writing - review & editing, Supervision, Conceptualization. Amy F. Ho: Writing - review & editing, Validation, Project administration, Investigation, Formal analysis, Data curation. Naomi Alanis: Writing - review & editing, Validation, Resources, Methodology, Investigation. Ryan Kirby: Writing - review & editing, Validation, Supervision, Resources, Investigation. Chet D. Schrader: Writing - review & editing, Validation, Supervision, Project administration. Hao Wang: Writing - review & editing, Writing - original draft, Validation, Supervision, Project administration, Methodol- ogy, Investigation, Formal analysis, Conceptualization.

Declaration of Competing Interest

All authors stated that there was no conflict of interest.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2023.01.029.

References

1. Hsiao CJ, King J, Hing E, Simon AE. The role of health information technology in care coordination in the United States. Med Care. 2015;53:184-90.
2. Zhang X, Hailu B, Tabor DC, et al. Role of health information technology in address- ing health disparities: patient, clinician, and system perspectives. Med Care. 2019;57 Suppl 6 Suppl 2:S115-20.
3. Patel V, Johnson C. Individual’s use of online medical records and technology for health needs. , 40ONC Data Breif: Office of the National Coordinator for Health Information Technology: Washington DC; 2018; 2018.
4. Roehrs A, da Costa CA, Righi RD, de Oliveira KS. Personal health records: a systematic literature review. J Med Internet Res. 2017;19:e13.
5. Irizarry T, DeVito DA, Curran CR. Patient portals and Patient engagement: a state of the science review. J Med Internet Res. 2015;17:e148.
6. Turner K, Clary A, Hong YR, Alishahi TA, Shea CM. Patient portal barriers and group differences: cross-sectional national survey study. J Med Internet Res. 2020;22: e18870.
7. El-Toukhy S, Mendez A, Collins S, Perez-Stable EJ. Barriers to patient portal access and use: evidence from the health information national trends survey. J Am Board Fam Med. 2020;33:953-68.
8. Roy M, Purington N, Liu M, Blayney DW, Kurian AW, Schapira L. Limited English pro- ficiency and disparities in health care engagement among patients with breast cancer. JCO Oncol Pract. 2021:OP2001093.
9. Guzman-Clark J, Farmer MM, Wakefield BJ, et al. Why patients stop using their home telehealth technologies over time: predictors of discontinuation in veterans with heart failure. Nurs Outlook. 2021;69:159-66.
10. Lombardi K, Pines JM, Mazer-Amirshahi M, Pourmand A. Findings of a national dataset analysis on the visits of Homeless patients to US emergency departments during 2005-2015. Public Health. 2019;178:82-9.
11. Wang H, Johnson C, Robinson RD, et al. Roles of disease severity and post-discharge outpatient visits as predictors of Hospital readmissions. BMC Health Serv Res. 2016; 16:564.
12. Constance J, Lusher JM. Diabetes management interventions for homeless adults: a systematic review. Int J Public Health. 2020;65:1773-83.
13. Webel AR, Cuca Y, Okonsky JG, Asher AK, Kaihura A, Salata RA. The impact of social context on self-management in women living with HIV. Soc Sci Med. 2013;87: 147-54.
14. Hwang SW, Burns T. Health interventions for people who are homeless. Lancet. 2014;384:1541-7.
15. Aldridge RW, Story A, Hwang SW, et al. Morbidity and mortality in homeless indi- viduals, prisoners, sex workers, and individuals with Substance use disorders in high-income countries: a systematic review and meta-analysis. Lancet. 2018;391: 241-50.
16. Wang H, Nejtek VA, Zieger D, et al. The role of Charity care and primary care physi- cian assignment on ED use in homeless patients. Am J Emerg Med. 2015;33: 1006-11.
17. Holmes CT, Holmes KA, MacDonald A, et al. Dedicated homeless clinics reduce inap- propriate emergency department utilization. J Am Coll Emerg Physicians Open. 2020;1:829-36.
18. Turner K, Hong YR, Yadav S, Huo J, Mainous AG. Patient portal utilization: before and after stage 2 electronic health record meaningful use. J Am Med Inform Assoc. 2019; 26:960-7.
19. Diffin J, Byrne B, Kerr H, et al. The usefulness and acceptability of a personal health record to children and Young people living with a complex health condition: a realist review of the literature. Child Care Health Dev. 2019;45:313-32.
20. Bodenheimer T, Lorig K, Holman H, Grumbach K. Patient self-management of chronic disease in primary care. JAMA. 2002;288:2469-75.
21. Grady PA, Gough LL. Self-management: a comprehensive approach to management of chronic conditions. Am J Public Health. 2014;104:e25-31.
22. Buntin MB, Jain SH, Blumenthal D. Health information technology: laying the in- frastructure for national health reform. Health Aff (Millwood). 2010;29: 1214-9.
23. Morrison DS. Homelessness as an independent risk factor for mortality: results from a retrospective cohort study. Int J Epidemiol. 2009;38:877-83.
24. Moulin A, Evans EJ, Xing G, Melnikow J. Substance use, homelessness, mental illness and Medicaid coverage: a set-up for high emergency department utilization. West J Emerg Med. 2018;19:902-6.
25. Lee KH, Jun JS, Kim YJ, et al. Mental health, substance abuse, and suicide among homeless adults. J Evid Inf Soc Work. 2017;14:229-42.
26. Winetrobe H, Rice E, Rhoades H, Milburn N. Health insurance coverage and healthcare utilization among homeless young adults in Venice. CA J Public Health (Oxf). 2016;38:147-55.
27. Bissonnette A, Hijjazi KH, Elder homelessness.. A community perspective. Nurs Clin North Am. 1994;29:409-16.
28. Bowen M, Marwick S, Marshall T, et al. Multimorbidity and emergency department visits by a homeless population: a database study in specialist General practice. Br J Gen Pract. 2019;69:e515-25.
29. Kang S, Ahn TK, Seo YH, Suh YJ, Paik JH. Comparison of emergency department utilization trends between the COVID-19 pandemic and control period. Medicine (Baltimore). 2021;100:e26847.
30. Yakobi M, Cheng MD. Retrospective analysis of the effects of the COVID-19 pan- demic on the emergency department walk-out rate in an acute care community hospital during the USA national lockdown. Cureus. 2021;13:e16660.