We targeted patients and/or parents/guardians of patients who were repeat users of the open source indivo PCHR [33], embedded in a hospital patient portal system. We defined repeat PCHR users as those who had logged in to the system on at least five separate occasions, as recorded in system tracking data. Respondents were at least 18 years old and could read English.

The PCHR system with which respondents were familiar contained clinically and administratively entered data about diagnoses, vital statistics, allergies, medications, lab reports, problems and health behaviors populated from the electronic medical record at the host institution. Respondents could review, annotate but not delete from their record and set role-based access permissions for others to see it, including out-of-hospital providers, however sharing with public health authorities was not an option. A secure messaging and an online scheduling system operate with the PCHR and a policy framework governs access and sharing rules in the pediatric domain [34]. An online interactive tutorial and demonstration system provides audio-narrated instruction to orient users to the PHCR and to its functions including reviewing, annotating and communicating with clinicians.

The survey primarily collected closed-ended, fixed-choice data. Some questions had open-ended fields for narrative comment. Domains included perceived value of the PCHR and its capabilities, knowledge and history of sharing information from the PCHR, willingness to share data from the PCHR, conditions and contexts for sharing, demographics and health status. Unwillingness to share specific categories of health information, or reticence, was assessed with the question “Are there any categories of information included in your/your child's PCHR that you would consider sensitive and would not share?” asked separately for sharing with: 1) an out-of-hospital health care provide (“outside provider”); and 2) the state/local public health department. Respondents could indicate which items from a list in the survey were too sensitive to share (multiple selections allowed). The list included: contagious illness, violence, sexually transmitted diseases (STD), tobacco, alcohol, other substances, genetic disorders, mental illness, family information and financial information. Participants were classified as willing to share all categories of information with a target if they either chose “I would share all categories of information” as a response or if they did not check off any information category as too sensitive to share. The context for sharing with an outside provider was to support patient care, and the context for sharing with public health was to support health monitoring and research.

Perceived values of the PCHR and reasons for reticence to share were based upon measures described previously [17,35].

Demographic and self-rated health descriptors were based on standardized measures. Race and ethnicity reports were reduced for analyses to white/non-white categorizations. Respondents were classified as non-white unless they reported solely Caucasian race and also non-Hispanic ethnicity. Patient self- (or parent)-rated health was based on a scale question where values of “good”, “very good” or “excellent” were used to indicate a better health condition and compared to values of “fair” or “poor”. Patient age in years was reported and dichotomized at 13 years and older for categorical analyses. Categorical responses for household income were reported for the values of under $40,000, $40,000-100,000, $101,000-200,000, and over $200,000 and dichotomized at over $100,000 for analyses.

To measure response bias, we compared patient characteristics of responders to the source population of all PCHR patient records that existed at the time of survey (n = 1610 inclusive of responders), obtained from an administrative database. We compared aggregate data about the source population to those for responders for measures of patient: age, race/ethnicity, duration of treatment at the hospital, and in-state residence.

An analysis of system data identified 820 repeat users of the system all of whom were notified by email that a bulletin communication had been posted to their PCHR homepage. To protect patient privacy, this email made no mention of the nature of the bulletin or presence of a survey. Of the notified users, 463 (56.5%) were active in the portal system during the study period, as evident by information about last login, and could reasonably be assumed to have had an opportunity to see the survey notification bulletin. Of these active users, 261 took the survey for a response rate of 56%. Responders did not differ from the source population in measures of patient race, length of care at the hospital, or in-state residence. Patients in the source population were more likely to be over the age of 12 (40% versus 31%, p = .006).

Descriptive statistics characterized response rate and sample demographics. Observations in which respondents selected a “prefer not to answer” option were excluded from analyses except for analyses of income and sharing where missing income data were substantial and responses separately categorized. Narrative responses were reviewed and when appropriate, assigned a coded value by matching responses to a priori fixed response fields, or preserved as written for qualitative annotation. Chi Square tests compared respondents and the source population and assessed associations between willingness to share by topic and demographics. Willingness to share was measured by topic as a dichotomous dependent variable. We used a repeated measures analytic approach to compare sharing probabilities for respondents with respect to different topics and different “targets” defined as outside providers and public health, recognizing that responses to sharing questions are clustered at the individual level. All analyses were performed using SAS 9.2.

Of the 261 respondents, 21 (8%) were patients and the rest were parents or guardians of patients. Demographic and health characteristics of the survey participants are described in Table 1.

Among survey respondents, interest in the PCHR was nearly universal with 97.1% indicating that they would like to maintain their record as they/their child matures. Knowledge about sharing through the PCHR was limited; 42.3% were aware that they could allow other individuals to view their record. Only 22.6% of the total sample knew that they could choose to share only portions of their record. Only 13.8% (n = 33) of all respondents indicated that they had previously shared their PCHR. Among these, the most common target for sharing was with an outside provider, reported by 66%. Those who reported having shared their record were no more likely to report knowledge of PCHR sharing capabilities than those with no history of sharing (16 reported knowledge of sharing; OR 1.3, 95% CI 0.6, 2.8, p = .4). A history of sharing the PCHR was not associated with willingness to share PCHR information about different topics with public health or providers.

Patient perceptions about the value of the PCHR focused mainly on personal control over and access to health records (Figure 1). Large percentages of respondents also reported valuing the PCHR as a tool for interacting with or sharing records with other providers or institutions. When asked to indicate which single function was most important of the many they valued, 64.7% chose unlimited access to their record, followed by centralizing health records (9.1%). Value selections were not associated with child age, health status, length of treatment at the hospital, history of sharing, or willingness to share information with public health or an outside provider. No clear pattern was evident in associations among specific value areas and other factors, including number of children in a household, race or income. Interest in keeping the record was not uniformly associated with perceived value areas.

More respondents were willing to share all categories of information with the state/local public health authority (63.3%) than with an outside provider (54.1%) (OR 1.5, 95% CI 1.1, 1.9; p = .005). Few respondents reported they would not share any category of information with an outside provider or public health authority, respectively 5.2% and 7.9%.

Reticence to share PCHR information varied by health topic and was lowest with respect to sharing contagious illness information; only 8.7% and 9.2% of respondents were unwilling to share this information with a provider outside their home institution, or with the state/local public health authority, respectively (Figure 2). Reticence was highest with respect to sharing financial information; 37.6% and 32.3% of respondents were unwilling to share this information with an outside provider and a public health authority, respectively. Subjects were more likely to be reticent to share information on sexually transmitted disease with an outside provider than with public health (OR 1.6, 95% CI 1.03, 2.5, p = .037). No other differences in sharing across targets by topic were found.

When asked to characterize willingness to share PCHR information with an outside provider, reticence was higher for all exemplary health topics compared to contagious illness (Figure 3). For example, respondents were 6.3 times more likely to report reticence to share information about money than contagious illness with an outside provider (p < .0001). For all topics assessed, willingness to share information with an outside provider was not associated with patient’s age, health status, number of children in household, or race. Reporting an income of $100,000 or lower, and reporting an income at all (versus not providing information about one’s income) were both positively associated with willingness to share health information for multiple health topics (Figure 4).

The most common objections to sharing data with an outside provider were relevance to patient care and the potential for discrimination by insurance companies (Figure 5). Narrative responses indicated that a decision about sharing categories of information would be made on a case-by-case basis, depending on the reason for consulting an outside provider, the patient/family relationship with the provider, and consideration of why information was needed or requested. A few parents noted that they would not share any information that might prejudice an outside provider’s judgment in diagnosing or treating their child.

Compared to reticence about sharing contagious illness information with a public health authority, respondents were significantly more likely to report reticence to share information about all topics except violence, sexually transmitted diseases, and tobacco (Figure 6). For example, respondents were 4.9 times more likely to report reticence to share information about money compared to contagious illness with a public health authority (p < .0001). Willingness to share health data by category was not associated with patient’s age, race or health status, the number of children in the household, or income. Those who declined to provide their family income were less likely to share information from every category but mental illness, contagious illness and violence (Figure 7). There was no association between reported income level and willingness to share with public health in any information category.

Lack of trust in both the anonymity of data and in a government agency handling the data were the primary concerns reported for sharing with public health (Figure 8). In narrative comments, participants questioned the relevance of providing health data to a public health authority. Several respondents also commented that even with anonymized data, inclusion of personal information in a small sample could jeopardize privacy.

This is a single site study—findings may not generalize to other populations. We assessed attitudes across several but not all health areas, for two main but not all possible targets. A majority of eligible respondents participated. Demographic data about patients were available for the source population of PCHR users in aggregate. Comparison of aggregate sample characteristics for respondents and the source population identified no differences in patient race, length of treatment at the hospital or in-state geographic residence but found that patients in the source population were older. Available data do not afford direct comparison of responders and non-responders. Survey research has inherent limitations around capturing nuance and this may be important to understanding sharing of information from a still novel technology. Analyses of narrative data and reports about reasons for reticence are included to deepen our understanding of sharing. Alternative approaches were not used but might include surveying respondents about sharing under specific patient scenarios or focus group research.