This study was approved by the Arizona State University Institutional Review Board (No. 00010309). The study sites were two outpatient integrated care facilities using the same EHR platform. One facility treats patients with SMI conditions (SMI facility), while the other predominantly treats patients with general behavioral health (GBH) conditions (GBH facility). This study used definitions from Grando et al [15] for prescribers and nonprescribers. GBH facility professionals were comprised of 15% prescribers and 85% nonprescribers, while SMI facility professionals were comprised of 10% prescribers and 90% nonprescribers.

Four, 2-hour focus groups were conducted at the study sites: two focus groups at each site, with seven behavioral health professionals each. Such a design allows sufficient time for individuals to share their thoughts while providing a small-group environment for conversation [28-30]. Focus group participants were facility employees, 21 years of age or older, who worked closely with patients with behavioral health conditions. Participants from the GBH facility were selected by executive staff, while participants from the SMI facility were recruited using flyers and were self-selected. A representative sample of prescribers and nonprescribers was sought for both facilities.

Participants were asked to individually complete a survey prior to the focus group. The survey was adapted from Tierney et al [26] and was comprised of nine statements that were rated on a Likert scale with the following responses: “strongly disagree,” “somewhat disagree,” “neutral,” “somewhat agree,” “strongly agree,” and “don’t know/can’t say” (Table 1) [27]. The survey prompts were divided into six specific aims based on measuring concepts and primary impact, patient or professional (Table 1). After each focus group, participants completed the same survey to evaluate changes in opinions of PC-GIS after seeing how actual patients exercise choices. It was hypothesized that discussion of a demonstrative case study would lead to a decrease in behavioral health professionals’ support for PC-GIS. The survey style (Table 1) lends to measuring opinion changes.

The survey analysis was used to determine the presence or absence of directional opinion change following the focus groups. SPSS Statistics for Macintosh (version 27; IBM Corp) was used for descriptive statistics, Wilcoxon signed-rank testing, Cronbach α tests, scree plots, and exploratory factor analysis on pre- and postsurvey results. Exploratory analysis was done with variables, or groups, of interest with appropriate sample sizes of at least 12 for exploratory factor analysis [31-33]. Descriptive statistics were computed using the bootstrapping option (10,000 replicates), and skewness and kurtosis results were used to confirm normal distribution and, thus, verification of data from the survey [34]. Cronbach α tests were run to ensure multidimensionality of the survey; an α value below .80 was considered evidence of a multidimensional scale [35]. Prompt selections of “don’t know/can’t say” were recorded as blank to avoid skewing results. The Likert scale results ranged from 1.00 (“strong disagreement”) to 5.00 (“strong agreement”). For the pre– and post–Likert scale survey results [35,36], the Wilcoxon signed-rank test (power=0.80, α=.05) was applied to each survey prompt to determine significance for changes after the focus group for all participants (N=28) and between participants serving different patient populations (n=14 each).

Intended response concepts of survey prompts were identified based on results from Tierney et al [26] and were labeled as “prompt aim” (Table 1). Exploratory factor analysis showed which aims, or components, were actively measured by the survey and relationships between prompts based on participant response [36,37]. Varimax rotation was used because survey prompts were not designed to correlate [26]. Measured components of the survey were identified by prompt magnitude in the output component matrix. This analysis revealed the component emphasis from pre- to postsurvey that was used to gauge professionals’ perceptions [35]. Scree plots were used to further validate the component results from the factor analysis where viable components had an exponential slope [35].

The focus group was presented in six sections (Figure 1) [27]. In Section 1, baseline perceptions of PC-GIS were elicited using examples and explanations of granular information and sharing [9,17,19,27,38,39]. In Section 2, the Soni et al [4,23], or “card sorting,” study was explained to participants, and a specific case from the study of granular sharing by an actual patient with patient-executed redactions was presented [27]. Participants were then shown the complete patient data set without redaction in Section 3 as if a provider decided to “break the glass,” a term that refers to health care professionals’ retrieval of a patient’s redacted information in an emergency.

During Section 4, the same patient’s data were presented by category (ie, alcohol use and alcoholism, communicable diseases, drug abuse, genetic information, mental health, other addictions, other information, and sexuality and reproductive health) and sensitivity (ie, very sensitive, somewhat sensitive, or not sensitive). Section 5 explored that patient’s choices to share those categories with different health care professionals and institutions.

Finally, in Section 6, participants were asked to reflect on their understanding, opinions, and recommendations for PC-GIS.

Two qualitative analysis techniques were applied to focus group outputs to provide insight into the survey results. Thematic analysis was used to find and define emergent topics of importance to participants [40]. Audio recordings of the focus groups were transcribed through a third party [41] and screened for accuracy by three researchers working sequentially by visual annotation. The validated transcriptions were analyzed using Braun and Clarke’s [42] thematic analysis guidelines and anthropological methodology through six iterations, resulting in quantifiable codes and themes [35]. The units of analysis were meaningful phrases per participant and themes identified through repetition and frequency in transcripts. MAXQDA software (VERBI GmbH) was used to identify and define emerging themes from transcripts. Themes and codes were defined and refined iteratively by three researchers.

The exploratory factor analysis results from the survey were used to organize emergent themes and subthemes from the focus group thematic analysis in the last iteration. This step permitted complementary analysis of qualitative and quantitative results.

After providing descriptive statistics and exploratory factor analysis results, themes and codes were used to interpret these results and provide insight into opinion shifts regarding PC-GIS. To compare themes between patient populations, quote matrices and complex coding queries (intersection-set) were performed.

A total of 28 participants were recruited (Table 2). Out of these participants, 5 (18%) prescribers (ie, physicians and nurse practitioners) and 23 (82%) nonprescribers were identified [15]. This was a demographically representative sample of professionals at each site. All 28 participants took the presurvey and 27 (96%) took the postsurvey.

Comparison of the pre– and post–focus group survey responses demonstrated two significant changes: (1) change from strong agreement to strong disagreement (mean <2.5, including SE) with patient-focused survey prompts and (2) change from strong agreement to strong disagreement (mean >3.5, including SE) with professional-focused survey prompts (Table 3). Descriptive analysis results provided skewness and kurtosis statistics exhibiting normal distribution of data, a validation of the survey results and usability of participant responses.

Drilling into the specific prompts, as defined in Table 1, patient understanding (prompt 1, P=.001), patient-professional relationship (prompt 5, P=.006), and patient control acceptance (prompt 6, P=.005) demonstrated significant change, with increased concern about patient control. Authorized EHR access (prompt 3, P=.03) and patient rights (prompt 9, P=.02) also showed significant change toward concern, validating the study hypothesis. Of note, patient-professional relationship (prompt 5) is considered a negatively phrased expression [26,27], providing insight into the postsurvey shift in the quality-of-care response.

Presurvey results loaded into four main components: patient control acceptance, professional considerations, patient rights, and patient understanding. In postsurvey loadings, exploratory factor analysis revealed only three components present; patient understanding was now absent (Table 4). After pairing with the descriptive statistics, results suggest that after the focus group, participants became more concerned with patient rights and patient control acceptance and the impact of these aspects on professional matters, such as quality of care and patient-professional relationship.

In the next step of validating the hypothesis, thematic analysis (4025 codes) of focus groups yielded three main themes (Figure 2), complementing the survey components (Table 5). The themes were professional considerations (2234/4025, 55.5%), patient aspects (1046/4025, 26.0%), and PC-GIS technology aspects (745/4025, 18.5%).

The professional considerations theme covers themes that directly impact a provider, including information needed to provide health care services. The patient aspects theme encompasses all topics relating specifically to patient experience and rationale. The PC-GIS technology aspects theme reflects the specific discussion of PC-GIS process and operations (Table 5 and Figure 2).

The survey results show the shift to concern yielding components of patient control acceptance, patient rights, and professional considerations, while thematic analysis shows how professional discussion revolved predominantly around professional considerations, such as necessity of health information (Figure 2). This overall shift toward professional concern around PC-GIS was observed in the focus group discussion and was coded under multiple themes and subthemes; an example quotation is as follows:

Similarly, professionals quickly pointed to the complexity of PC-GIS in their domain and the potential for negative impact on patient care:

The survey results reflect understanding and opinions in a quantitative fashion, while the interviews demonstrate how the concepts are linked.

Thematic analysis also conveys the complexity of participant perceptions, with subthemes interweaving patient and professional considerations. Regarding 879 codes, the reactions subtheme included general feedback about PC-GIS (n=255, 29.0%) as well as specific concerns (n=334, 38.0%), predominantly relating to patient safety and health. Patient safety and health encompassed issues ranging from missing medications and the potential for drug-drug interactions to the need for improved physical and mental health integration. A minority (n=149, 17.0%) of participants felt that data sharing in health care as an environment should never be granular: “[This is] not a place to be granular.” Other issues surfacing in the health professional concern area included the mismatch of patients’ interpretation of information versus health professionals’ interpretation.

While the patient aspects theme included two subthemes corresponding to exploratory factor analysis components, the major subtheme of patient perspective considered drew greatly on health professionals’ thoughts on patients’ reasoning to share or not share health information. Indeed, 65.0% of 722 codes (n=469) within the patient perspective considered subtheme were specifically related to patient reasoning to not share, such as patient fears or fear of discrimination (n=113, 24.0%). Many instances of patient reasoning (n=201, 42.9%) dealt with patients’ understanding and comprehension.

To provide further context for the quantitative results, participant opinions from within the reactions subtheme (Section 2 codes: 81/116, 69.8%; Section 4 codes: 35/116, 30.2%) were divided into three groups: concerned, supportive, and mixed opinions. For Section 2 questions, regarding the redacted case study, 46% (37/81) expressed mixed opinions, 36% (29/81) were concerned, and 19% (15/81) were supportive. Within Section 4 questions, regarding patient categorization and sensitivity, 43% (15/35) expressed concerned opinions, 37% (13/35) expressed mixed opinions, and 20% (7/35) were supportive. Of note, PC-GIS unease revolved around patient safety:

To that end, participants suggested prioritizing a risk-benefit analysis tool for patients as an adjunct to professional-patient PC-GIS interactions:

A common theme among all focus group participants was that role-specific, essential health information access should be considered when granting PC-GIS authority:

The problem is that “essential” information depends on the circumstances of the patient:

Exploration of potential differences between health professionals using descriptive statistics and drilling down on qualitative data led to identification of two distinct patient populations: GBH and SMI. Descriptive statistics and Wilcoxon signed-rank tests were performed separately for each group (Table 6). Differences were observed in the presurvey, where participants treating an SMI population showed agreement with all positively phrased prompts. Meanwhile, those treating the GBH population showed agreement only with prompts 3 and 7: authorized EHR access (positive phrasing) and quality of care (negative phrasing). Differences were compounded in the postsurvey, where participants treating an SMI population showed agreement only with prompt 5: patient-professional relationship (neutral phrasing perceived as negative) [26,27]. Those treating a GBH population agreed with the negatively perceived prompts (ie, patient-professional relationship and quality of care) and disagreed with a positively worded prompt (ie, patient control acceptance; Table 6). Participants treating an SMI population initially perceived PC-GIS positively, then made a significant shift to neutral or mixed opinion with concern over the patient-professional relationship prompt (P=.007). The participants treating a GBH population showed concern over PC-GIS with a shift to concern over the following prompts: patient control acceptance (prompt 2, P=.007; prompt 4, P=.009), quality of care (P=.01), and patient understanding (P=.03).

Quote matrices were applied to specify differences between GBH and SMI professionals on survey aims, with some subtopics within themes considered when applicable. GBH professionals showed greater frequency of discussing negative reactions (32/44, 73% of codes), positive reactions (14/22, 64%), professional risks and rights (44/67, 66%), and outside institutions (52/69, 75%). SMI professionals more frequently discussed the following subtopics: do not need to know (21/23, 91%), patient-professional relationship (124/180, 68.9%), trust (22/32, 69%), patient aspects (669/1046, 64.0%), and costs (36/46, 78%). These results reflect large differences in frequencies of thematic analysis coded topics. However, while a topic may be suggested, the discourse may not contain an opinion. Therefore, a second layer of qualitative analysis to identify subthemes was performed.

To identify differences in participant perceptions on themes, subthemes, and topics within subthemes, complex coding queries were used to categorize negative, positive, or mixed perception codes. GBH professionals perceived costs and trust negatively, overall (Table 7). SMI professionals referred to professional risks and rights topics with a negative slant, while the other topics were presented in a mixed or positive fashion. The complex coding queries highlight the complexity of the topic and suggest an impact of patient population on subthemes and topics.

The mixed methodology analysis focused on the differences between SMI and GBH professionals, where SMI professionals displayed lower levels of concern regarding the process of PC-GIS, more frequently citing the following topics: do not need to know (21/23, 91% of codes), patient-professional relationship (124/180, 68.9%), and trust (22/32, 69%).

Results show that behavioral health professionals had fewer positive views on PC-GIS after the focus group, with a significant opinion shift toward concern on the following topics: patient understanding (P=.001), authorized EHR access (P=.03), patient-professional relationship (P=.006), patient control acceptance (P=.005), and patient rights (P=.02). Qualitative methodology supported these results, as themes and subthemes, such as professional considerations (2234/4025, 55.5% of codes) and necessity of health information (260/766, 33.9%), identified aspects of PC-GIS concerns; indeed, participant opinions after viewing the case study without redactions (Section 4) showed increased levels of concern (7% overall change).

Mixed methodology results after the focus group showed concerns that PC-GIS could negatively impact behavioral health professionals’ ability to deliver optimal care. This perception shift was evident in qualitative results from discussions dominated by patient health and safety topics (combined, 60.3% [70/116] of codes of concerned reaction). Our results show that behavioral health professionals remained highly concerned about patient granular control for a variety of reasons [18]. Health professionals in our study highlighted potential negative effects of granular sharing, including impact on the professional-patient relationship and lack of access to necessary health information, reflected in the professional considerations theme. A minority of professionals (149/879, 17.0%) considered health care as simply “not a place to be granular,” while most acknowledged acceptance of the trend toward increasing PC-GIS and offered concrete recommendations for proactive processes that could help ensure patient safety while preserving record sharing choice.

Much of the literature focuses on patient perspectives of PC-GIS, demonstrating that patients respond positively to having granular control over data sharing [5,17,18,24], while casting doubt on a patient’s understanding of information relevancy or professionals’ “need to know” [5,15]. Our study reflects similar concerns by behavioral health professionals, where subthemes included necessity of health information (260/766, 33.9% of codes) and patient concept comprehension (13/115, 11.3%). Our thematic analysis demonstrated that when professionals are shown results of a patient’s granular sharing choices, they view the choices from the patient perspective, while expressing apprehension that necessary role-specific information may not be appropriately shared [4]. Employing shared decision-making using specialty-tailored methods may help alleviate such concerns [5,15,27].

A trusting bond is important in health care delivery and is a continued underlying basis in quality of care and patient outcomes in health literature [2,43-46]. Our qualitative results refer to trust when subthemes such as talking with patient, patient-professional relationship, and patient education directly deal with strengthening the relationship and understanding between health professionals and patients. A recent study by Esmaeilzadeh [47] showed that patients’ trust in providers influences their trust in information sharing technology. Our qualitative results exemplify some processes that health professionals may use to strengthen trust: “Tell them [patients] the reason why we’re asking, the importance of it, and to help them understand why we need the information.” Indeed, this type of approach to strengthening the professional-patient relationship was a common recommendation in our study, as well as in existing literature, to alleviate professionals’ worry over patients not sharing appropriate information [15,27,44,46,47].

Proceeding further into the topic of education, professionals involved in PC-GIS must have the knowledge, background, and tools to assist patients in making safe sharing choices. This is exemplified in the case where a professional supports a patient’s choice to withhold behavioral health diagnoses and medications from a patient’s dentist. In reality, oral health and behavioral health have many important intersections, including substance use disorder and eating disorders [48]. Therefore, organizations and institutions must ensure that their PG-GIS process and professionals are prepared to provide sound advice to ensure patient safety. While health care institutions need to consider PC-GIS use in integrated and coordinated care, attention should be paid to critical policies, such as Title 42 CFR (Code of Federal Regulations) Part 2 and the CARES (Coronavirus Aid, Relief, and Economic Security) Act [8,38,49], and safe implementation of health care technologies relevant to health information exchange and patient EHRs [11,22]. Professionals must be actively engaged in the creation, implementation, and monitoring of data sharing policies that integrate relevant statutes as well as advances in technology and biomedicine. Our results show that when provided with an in-depth explanation of tools affecting health care delivery, health care professionals can begin the necessary dialogue regarding concerns and recommendations for improvement [27]. Identification of concerns for all stakeholders is necessary for successful implementation of health care technologies, such as PC-GIS.

While some research exists that considers health professionals’ opinions on granular sharing and consent, few studies have explored the behavioral health realm, where care generates and uses highly sensitive patient information [26,50,51]. Our study has added to knowledge on behavioral health professionals’ perceptions [27], while exposing divergence between behavioral health professionals who treat differing patient populations. Results comparing behavioral health professionals’ predominant patient populations indicate that those working with an SMI population displayed lower levels of concern and focused on patient aspects (669/1046, 64.0% of codes), as compared to participants working with the GBH population (patient control acceptance, P=.004; patient understanding, P=.02).

Our results provide a perspective on the relevance of studies from physical health settings application to behavioral and integrated health groups [13]. Based on a pilot study where some patients were given PC-GIS capabilities, Tierney et al [26] showed that 63% of the providers strongly agreed that granular information restriction will likely reduce the quality of care delivered. Our study found that after the focus group intervention, behavioral health professionals agreed overall with Tierney et al’s findings, moving from strong agreement to strong disagreement regarding patient-focused survey prompts. Such outcomes show that behavioral health professionals have similar levels of concern about PC-GIS as compared to other health professionals. However, their concerns may differ in scope and application, as behavioral health patients may be more vulnerable to addictions, discrimination, and influence from outside institutional pressures [52-54].

The divergence observed became more visible when looking at the two facilities. Survey analysis showed that participants from the SMI facility not only viewed PC-GIS more positively than those from the GBH facility, but they also displayed differences in what survey aims (ie, patient-professional relationship and patient rights) we found to have significant change from pre- to postsurvey. Professionals from the SMI facility displayed a significant (P=.007) shift from neutral to mixed opinions regarding the patient-professional relationship, an aim falling within the patient aspects category. Results from the quote matrices reinforced these outcomes, as the conversation by SMI facility professionals focused more heavily on patient aspects (669/1046, 64.0% of codes). On the contrary, GBH facility participants showed consistent agreement (Table 6) on professional-focused prompts and disagreement with patient-focused prompts. Health professionals working with an SMI population may view PC-GIS with less concern because they typically interact with their patients more frequently over longer periods of time [55,56]. One approach to alleviating PC-GIS concerns may be in bolstering patient-professional relationships and communication.

Integrated care that emphasizes transparency and patient-centeredness are health system goals [9,10]. This study highlights key aspects of PC-GIS that must be considered for its broader and deeper integration in the care environment. Improving process transparency benefits patients and professionals by exposing gaps in care, differences in patient safety and outcomes, and drivers of costs [2,9,15,18,57]. Understanding diverse professional perspectives is critical in developing granular consent systems that balance patient-professional information needs. Results of this study will be used in the development of a PC-GIS tool, My Data Choices, to allow patients with behavioral health conditions to choose which medical records (eg, mental health information) to share with whom (eg, behavioral health providers) and for what purpose (eg, health care).

While our study had a small number of participants, it was a demographically representative sample within each integrated care study site, capturing the essence of the integrated health care team and their role-inspired concerns and needs. The sample size may affect the study replicability; further research with a larger sample size and in a variety of behavioral health and integrated care settings is needed. Another limitation in our study is the pragmatic difference in recruitment methods to accommodate facility leadership preference. Future work may consider semistructured interviews to more clearly identify differences in health professionals’ needs. Finally, our study presented a single exemplar case derived from Soni et al [4,23]. Future studies could leverage additional patient data sharing scenarios.

This study enhanced what is known about PC-GIS by systematically exploring the rationale behind behavioral health professionals’ perceptions, using results from a study of PC-GIS by real patients using their own data. Outcomes show that as health care professionals learn about PC-GIS implementation, they develop greater levels of concern. However, professionals balanced their concern with material recommendations for PC-GIS process improvement that ranged from patient and staff education to strengthening patient trust. Participants agreed that an informed and transparent system for health information sharing is needed to foster the mutual trust required to implement robust PC-GIS.