Perceptions and Preferences About Granular Data Sharing and Privacy of Behavioral Health Patients

Abstract

Little is known about data sharing preferences for care and research of behavioral health patients. Eighty-six behavioral health patients (n =37 Latinos; n = 32 with serious mental illness) completed questionnaires, in either English or Spanish, with items assessing their views on privacy and sensitivity of health record information. Most patients (82.5%) considered mental health information as sensitive. In general, there was a direct correspondence between perceived sensitivity of information and willingness to share with all or some providers. A main motivation for sharing data with providers was improving the patient’s own care (77.8%). Most participants (96.5%) indicated they would be extremely to somewhat willing to share their data for research with their care facilities and universities. Follow-up patient interviews are being conducted to further elucidate these findings.

Introduction

Behavioral health problems include substance use disorders, serious psychological distress, suicide, and mental disorders [1]. In the U.S., behavioral health conditions affect over 44 million (18.3%) adults, including 10.4 million (4.2%) adults who have a serious mental illness (SMI) such as major depression [2]. On average, SMI patients have 3.5 times higher emergency room visits, 4 times higher primary care visits, and 5 times higher specialist visits than the general population [3]. It is estimated that 17.8% of the U.S. population is Latino or Hispanic [4]. Of those, 15.6% (over 8.9 million) had a diagnosable mental illness in the past year. The rate of illicit drug use for Hispanic individuals ages 12 and up was 8.9%, while the national average was 10.2% [5].

The growth of health information exchanges (HIEs) and healthcare technologies have stimulated interest in integrated care and data sharing. Behavioral health patients often visit multiple health care providers when receiving behavioral and physical health treatments, and care coordination among those providers could be advantageous [3]. While data sharing among care providers may improve care, it may also impact patient privacy [6,7]. Privacy concerns may lead patients to avoid discussing their problems with providers, delay care and withhold information from providers [8-10]. Additionally, patients with behavioral health conditions frequently experience social stigma, employment and insurance discrimination, legal concerns, and worry regarding disclosing information to others [11,12].

The Office for the National Coordinator for Health Information Technology (ONC) recommends giving patients more granular level control over how and with whom their health record information is shared [13]. Granular control could include giving patients more authority over with which providers, under which circumstances, and for which purposes their data are shared. For example, ability to make decisions over whether or not to share alcohol abuse-related information with a physical health provider. Studies indicate that patients desire more control over their health data for care and research [11,14,15]. A recent study of 394 patients indicated that they were reluctant to share clinical data with for-profit research organizations, and that 32% of them expressed a desire for choices regarding the data sharing category (for example, mental health, substance abuse history, etc.) and data recipient when sharing medical records for care [14]. Understanding the level of granularity that individuals desire is still an open question.

Little is known about data sharing preferences for care and research of behavioral health patients. Grando et al. surveyed 50 English speaking behavioral health patients with no SMI concerning their data sharing preferences [11]. The study showed that behavioral patients may wish granularity over who can access their personal health data for care and research. Understanding patients’ data sharing preferences and perceptions may improve education and consent processes that influence their decisions to release or withhold health information.

The purpose of this study was to survey English and Spanish speaking behavioral health patients, including those with SMIs, on their perceptions regarding data sensitivity, willingness to share health data for care and research and related motivations.

Methods

Study Sites

Study site 1 is a community clinic in Arizona providing general mental health (GMH) treatment and social services to adults of all ages. Site 2, also in Arizona, offers case management services to adult patients with SMI.

Survey

We found validated English surveys, such as [16], developed to assess patients’ data sharing choices. But there was a lack of bilingual (English and Spanish) data sharing questionnaires validated with behavioral health patients.

Our survey was based on [11]. Demographic information was categorized based on U.S. Census Bureau classifications, except diagnoses, which were adapted per National Institute of Mental Health categorization [17,18]. The sensitive categories used in our instrument to ask questions related to sensitive data were based on those used by the National Committee on Vital and Health Statistics [19]. The resulting survey was translated to Spanish and back-translated to English by native Spanish speakers.

Survey Reliability Testing

Reliabilities of questionnaire items were examined using a test-retest approach with 31 Spanish and English-speaking adult behavioral health patients from study sites 1 and 2. Participants completed the questionnaire, in either English or Spanish, on two occasions, 14-21 days apart. Questionnaire items were revised based on the outcomes of the reliability analyses. The revised questionnaire was used in the current study.

Study Participants

Participants for the study described here were recruited from the same study sites where the reliability study was conducted. Potential participants were identified by study site staff members during routine clinical visits and referred to the recruiters. After the recruiter met with the prospective participant at the facility and explained the study to him/her (in either English or Spanish), the recruiter assessed the participant’s decision-making capability (using the UBACC test)[20]. We excluded participants with low consent comprehension (i.e., with UBACC scores < 15). Adult patients (21 years old or older) diagnosed with GMH or SMI who agreed to complete the questionnaire in English or Spanish and were deemed capable of giving informed consent were considered eligible to participate.

Study Design

After initial screening and consenting of eligible participant, the recruiter offered the participant the option of completing the questionnaire either in English or Spanish, and either electronically or on paper. The recruiter was present to help the participant with any questions or technical difficulties. Participants were compensated for their time.

Data Analysis

We used univariate statistics (e.g., frequencies, means, standard deviations, percentages) and plots to summarize the data. Parametric inferential statistical methods were used to analyze perceptions of data sensitivity and willingness to share data among English and Spanish-speaking, Latino and non-Latino participants from GMH and SMI populations.

Results

Demographics

Of the 88 participants recruited, 2 were excluded because of inability to understand and follow the study protocol, as measured by the UBACC test. Table 1 shows the demographics of participants included in the sample. The majority (n = 54; 62.8%) of patients had a GMH condition, while the rest were patients with SMI diagnoses. Most participants (n = 71; 82.5%) opted to have the questionnaire administered in English; the remainder opted for Spanish.

Table 1:

Demographic of participants

Participant characteristics (n=86)	Freq. (%)
Age (Years)
21-30	19 (22.1)
31-40	24 (27.9)
41-50	16 (18.6)
51-60	15 (17.4)
61-70	9 (10.5)
>70	2 (2.3)
Unknown	1 (1.2)
Gender
Male	26 (30.2)
Female	59 (68.6)
Other	1 (1.2)
Race/Ethnicity
White Alone, Not Hispanic or Latino	34 (39.5)
Black or African American	11 (12.8)
Hispanic or Latino	37 (43.0)
Native American or Alaskan Native	3 (3.5)
Other, Unknown	1 (1.2)
Income
≤$10000	50 (58.2)
$10001-$20000	23 (26.7)
$20001-$30000	10 (11.6)
>$30001	3 (3.5)
Education
No Schooling	1 (1.2)
Middle school (grades 6-8)	9 (10.5)
Some high school (no diploma)	14 (16.3)
High school graduate	19 (22.1)
Some college (1-4 years, no degree)	24 (27.9)
Associate degree	14 (16.2)
Bachelor’s degree	5 (5.8)
Diagnoses *
Anxiety or panic disorder	65
Bipolar Disorder	34
Depression	65
Impulse Control Problems	10
Identity or memory problems	22
Eating disorder	5
Obsessive compulsive disorder	9
Personality disorder	13
Schizophrenia or other psychosis	14
Drug or alcohol addiction	18
Post-traumatic stress disorder or
adjustment disorder	36
Chronic pain or somatic disorder	24
Other	1

^*As participants may have more than one diagnosis, the percentages are not reported.

Data Sharing for Care

We asked questions to understand participants’ desire for granular data sharing control based on type of information, information recipient and purpose of data usage. We asked participants how likely they were to share their behavioral health data with different behavioral and non-behavioral providers (Figure 1). Participants were most willing to always or sometimes share their health information with the behavioral providers at the study sites, followed by emergency providers, other non-behavioral providers at the study sites (e.g., primary and specialty care providers, pharmacists), behavioral providers outside the sites, and lastly with other non-behavioral providers outside the study sites. No significant differences in responses were seen in comparisons of English vs. Spanish speakers (χ²=1.27, p=0.866), Latino vs. non-Latino participants (χ²=0.78, p=0.941), or GMH vs. SMI patients (χ²=0.12, p=0.998).

Figure 1:

Behavioral health data sharing preferences, based on the type of medical provider

In assessing participants’ perceptions about how sensitive different types of health information are, we provided them with eight health information categories: mental health, psychotherapy notes, sexual and reproductive health, domestic violence and abuse information, information on sexually transmitted diseases, drug or substance abuse, alcohol abuse, and genetic data. Most participants considered mental health information the most sensitive, followed by psychotherapy notes (Figure 2). For several categories, the most common single response was ‘It does not apply to me’. Unfortunately, we did not collect with the survey information that could be used to check if the participant did not have certain types of sensitive medical records.

Figure 2:

Health categories classification as sensitive information

Participants were then asked about the likelihood of sharing sensitive health information with providers outside the study sites (Figure 3). We computed the mean percentage of patients who wanted choices regarding sharing their data with different types of providers. On average, when self-reporting having sensitive information in their medical records (the option ‘It does not apply for me’ was not selected), many participants (64.15%) wanted to restrict those records from some or all health care providers.

Figure 3:

Willingness to share sensitive health data with providers outside study sites

In general, we observed a direct correspondence between perceived sensitivity of information and willingness to share. The main exception was genetic data. While participants considered genetic data the eighth most sensitive type of information, they ranked it as the third most sharable. However, with the exception of sexually transmitted diseases (p<.05, Fisher’s exact test), none of the associations between willingness to share a particular category of information and perceived sensitivity of the information).

We asked participants about providers’ access to health information when prescribing a new medication. Most participants (78.0%) responded that the providers should have access to all their health data, 12.0% thought that providers should see only the data to which a patient provides the access, and 10.0% indicated that the providers should see all the health data only when the new medication may have any harmful interactions or effects.

Similarly, we asked patients about emergency providers’ access to data in a life-threatening situation. Most of the participants (70.0%), reported that providers should have access to all their data, 19.0% endorsed giving emergency providers access only to data shared by the patients, and 11.0% indicated that providers should have access to all health data only when the emergency may be life threatening.

Participants endorsed sharing their data when it can benefit their own care and treatment (77.8%) or if/when their providers asked them to share their data (61.1%). Large majorities of participants trusted the providers at the study sites overall (87.8%) and trusted them to share only the health data that they consented to share (93.3%).

Large majorities also reported that they would be upset if their providers shared their health data without asking them (83.3%) and that they might react by leaving such providers (65.6%). Only 30% of the participants reported worrying about providers knowing that they receive mental health treatment.

Data sharing for Research

Participants were generally willing to share health information with researchers when their own care (91.1%) or care for others (78.9%) could be improved. About half (51.1%) of participants, indicated they would always share their data for research, while 35.6% indicated that they would share their data for research if they were paid for it. Finally, we asked participants how likely they would be to share their health information with researchers (Figure 4). Participants indicated they would be extremely to somewhat willing to share their data for research purposes with their care facilities (96.5%). Participants appeared less willing to share their health information with drug development companies and government agencies. For each participant, we looked at whether their willingness level varied between different types of organizations. For example; varying willingness to share data with care facility compared to drug companies. Most (78.9%) participants desired control over how they want to share data with different research organizations.

Figure 4:

Willingness to share data for research with different types of organizations.

There were no significant differences in data sharing preferences between English vs. Spanish speakers (χ²=0.29, p=0.990), Latino vs. non-Latino participants (χ²=0.59, p=0.964), or GMH vs. SMI patients (χ²=0.25, p=0.993).

Discussion

Consistent with previous studies on behavioral health patients, participants wanted control over how to share sensitive health data with health providers [11]. In general, there was a direct correspondence between perceived sensitivity of information and willingness to share with all or some providers. When we contrast our results to studies from patients without behavioral conditions, it has been reported that patients with and without sensitive information prefer to restrict the sharing of sensitive versus less sensitive EHR information [14,15]. As reported in the literature, most of the participants appeared to be motivated to share health data unconditionally to avoid medical emergencies or drug-drug interactions [15,21,22]. As in previous studies, our participants trusted their providers at the study sites and trust in providers was an important motivation for sharing health information [11,16,23,24]. Additionally, improvement in a patient’s own care and treatment was an important motivating factor for sharing health data with providers.

As in previous studies [11,14], patients wanted control over how to share health data with researchers. Consistent with literature, willingness to share data decreases when the recipient is a for-profit research organization and important motivations to share health information for research were benefiting own care or improving care for others [11,25].

The ‘It does not apply to me’ response was frequently used when asked to assess the sensitivity of health data and willingness to share sensitive data with providers (34.4% for all types of data, and 29.7% for mental health, psychotherapy notes, drug or substance abuse and alcohol abuse). Participants’ lack of understanding of the meaning of certain sensitive data categories, inability to form opinions regarding sensitive categories, or stigma related to disclosing this information could be potential explanations for this response. For some categories, like sexually transmitted disease or substance abuse, it is highly probable that the question did not apply to the participants. For other categories, such as genetic data, the recruiters received frequent requests from participants for clarifications. These results highlight the need for better on-demand education material to address patients’ varying data sharing preferences and levels of health literacy.

A limitation of our study is that study participants were sampled from only two outpatient clinics in similar geographic areas with similar social demographics. Additional studies should be conducted on a larger sample of the population to capture more diverse views.

The outcomes of this survey and previous formative studies will guide follow-up card sorting interviews [11,25]. Thirty-six survey participants have given access to their health medical records available through the HIE. In the upcoming interviews, data privacy questions will be asked while study participants have access to a subset of their medical records.

Outcomes from that study will influence the design of an e-consent tool based on the Consent2Share software developed by the Substance Abuse and Mental Health Services Administration (SAMHSA) [26]. The My Data Choices tool will support patient-driven data access based on data sharing interoperability standards. Patient data will be shared in compliance with federal and state confidentiality laws, including protection of confidentiality of substance use disorder (42 CFR Part 2) [27]. On-demand multimedia patient education material will be embedded in the tool to illustrate risks and benefits of cross-organizational data sharing. We aim to pilot test the tool with 270 behavioral health patients in a prospective study.

Conclusions

A better understanding of behavioral health patients’ attitudes towards data sharing is needed. The outcomes of this survey indicate consistency between the perspectives on data sharing and privacy of behavioral health patients and other previously surveyed populations of patients with or without sensitive medical records.

In future work, we plan to apply lessons learned from the completed survey to conduct follow-up interviews with a subset of the surveyed patients. The knowledge gained from the interviews will be used for the development of an e-consent tool that will support patient-driven data sharing control and on-demand educational resources to better inform data sharing choices.