Keywords
adolescent medicine - confidentiality - 21st Century Cures Act - information blocking - clinical documentation - communication
Background and Significance
Background and Significance
Since the inception of OpenNotes, the practice of electronically sharing clinical documentation has aimed to provide transparent communication between health care providers and patients.[1]
[2] Following widespread adoption and a push for patients to have easier access to their medical records, the 21st Century Cures Act information blocking final rule has since mandated the immediate and electronic release of more health care data.[3] There is a wide consensus that such sharing is important to increase patient autonomy and improve record keeping around plans of care.[4] While sharing information with patients is prudent, information sharing in the case of adolescents, among other groups, is complicated by difficulties around protecting patient privacy.[5]
[6]
[7] The American Academy of Pediatrics supports the provision of confidential spaces to build trust and provide appropriate care.[8]
[9]
[10]
[11] Protecting these private conversations between adolescents and their health care providers increases the likelihood that an adolescent will communicate with health care providers and access essential health care.[12]
[13]
[14]
[15]
[16]
[17] This confidential information, however, must be documented and granularly segmented such that it can be released electronically to the adolescent while simultaneously being withheld from their assigned proxy, unless consent is given.[18]
[19]
[20] In the case of free-text data, such as clinical notes, confidential information may need to be documented in a separate confidential note type, where its contents can be labeled to prevent inappropriate release.[21] When confidential information is documented in a routine progress note, there is risk of a breach in confidentiality when that note is released through traditional methods of medical record sharing or electronically.[20]
Even with current toolkits and guides from pediatric advocates, there remains a lack of clarity surrounding what exactly is considered confidential. To further complicate the problem, specific privacy laws allowing adolescents to receive confidential medical care vary from state to state, leaving providers needing to practice confidentiality differently depending on their location.[22]
[23] Under California state law, when minors consent for confidential care, it generally also restricts the sharing of this medical information to their guardians unless direct consent is granted by the patient receiving the care.[24] These laws generally cover care for sexual and reproductive health, mental health, and drug and alcohol use, with exceptions for medication use and harm, among others.[25]
Anecdotally, there is a significant amount of confidential information in progress notes, leading many health care systems to opt out of sharing clinical notes of adolescents in fear of incidentally releasing confidential information.[26]
[27] While there has been research into how confidentiality may be breached from internal employees inappropriately accessing the charts of patients, there is a current lack of research defining the risk of leaking confidential information externally from electronic health data.[28] While Parsons et al studied the prevalence of confidential content within a confidential note type, no previous study has applied state-specific confidentiality laws to quantify the prevalence of confidential content in progress notes that could potentially breach adolescent confidentiality.[29]
Objective
The purpose of this study was to quantify the prevalence of confidential information, based on California state confidentiality laws, within progress notes for adolescent patients.
Methods
Defining a Confidential Rubric
To standardize the chart review process, a panel of experts from adolescent medicine, clinical informatics, and hospital privacy developed a consensus interpretation through a modified nominal group process of the California state confidentiality laws in the form of a decision flowchart ([Fig. 1]). This rubric was not intended to be a legal interpretation of confidentiality laws, but a good-faith effort to apply the laws to the reality of clinical care. Key principles were extracted from the “Understanding Confidentiality and Minor Consent in California” toolkit.[25] These principles were then applied to a small sample of notes to identify persistently ambiguous scenarios. These scenarios were discussed among the panel who developed a consensus on a set of key points. Key points included if the reader could infer patient involvement in confidential behavior, if the statement was denying confidential behavior (i.e., patient denies sexual activity or drug use), or if the content was regarding historical events that infer confidential behavior. If any of these points were in the note, the information would be considered confidential. Other topics likely shared in confidence with the provider, such as social conflict, divorce, or other uncomfortable situations, were not considered confidential, as they are not protected types of information in accordance with California law.
Fig. 1 Guiding principles of adolescent confidentiality algorithm.
Chart Review
This study is a single-center retrospective chart review of outpatient visit notes within Stanford Children's Health (SCH), a predominantly subspecialty academic pediatric network in Northern California. Notes were queried for visits between January 1, 2016 and December 31, 2019, for patients aged 12 to 17 years with a final note author of resident physician, fellow physician, or attending physician. Notes originated by a nurse practitioner or physician assistant and routed to a physician for co-signature were included if finalized by a physician author. This was considered as the prominent workflow for note writing within the organization. A simple random sampling of 1,200 notes (enough to obtain confidence intervals with a margin of error of 3%) was extracted and divided among a team of five physician reviewers (M.B., R.G., J.C., N.R., K.M.), with 20% of notes cross-reviewed by a different reviewer.[30] Annotations were labeled using the Doccano software.[31] Simple random sampling was performed, where every note had an equal probability of being selected. The physician reviewers were trained using the confidential rubric described above. Confidential information was categorized into one of three domains based on California state law including drugs and alcohol, sexual and reproductive health, and mental health. Reviewers were instructed to avoid labeling confidential information clearly in a medication or problem list that was being pulled in from an auto-populated list. Notes could contain more than one type of confidential information. If the reviewer was uncertain of a label, it was labeled as such and re-categorized via adjudication (M.B. and K.M.). If there was a mismatch in the subcategory of the label applied to a note text between two reviewers, both labels were maintained. A conservative approach was adopted and if any reviewer identified confidential content in a note, it was assumed to have confidential content even if another reviewer did not label it.
Data Analysis
Primary analysis to determine the prevalence of confidential terms in clinical progress notes was calculated by the proportion of notes with at least one label of confidential information. A 95% confidence interval was calculated for this prevalence using the asymptotic (Wald) normal approximation. A secondary analysis of characteristics of the patients' gender designated in the electronic health record, self-reported race, and language spoken was performed using a chi-squared analysis. A logistic regression analysis was performed to determine a statistically significant association between patient age and confidential content. A threshold of p <0.05 was used to determine statistical significance. Inter-rater reliability scores were calculated using Fleiss's kappa analysis with a score of 1 representing perfect agreement and 0 representing poor agreement. All statistical analyses were performed using Python. The Stanford University institutional review board determined that this project does not meet the definition of human subject research as defined in federal regulations 45 CFR 46.102 and is exempted from further institutional review board review.
Results
Confidential Terms Analysis
Out of 1,200 adolescent progress notes manually reviewed, 255 notes (21.3%) (95% confidence interval: 19–24%) contained confidential information. The prevalence of each note containing one of the three subcategories at least once, along with inter-rater reliability averages and examples, is outlined in [Table 1]. The most common subcategory was mental health with 145 notes labeled as such (12.1%). Mental health labels included components of the history and physical exam such as affect, suicidal ideation, or depressed mood. The second most common subcategory was drug and alcohol with 98 (8.2%) of notes containing this label. Many of the labels for this category were screening questions obtained during rooming or preappointment questionnaires, asking patients if they used tobacco, alcohol, or any other drugs. Mentions of illicit substance use such as marijuana or alcohol use were notable labels as well. The least prevalent subcategory was sexual and reproductive health, which was found in 78 (6.5%) notes during manual chart review. Labels varied from mentions of birth control to sexually transmitted disease screening and treatment. Mentions of sexual orientation and gender identity also qualified as confidential information. Review of confidential labels found that 38% of the 255 flagged notes contained only statements denying confidential behavior.
Table 1
Percent of notes with confidential information by label type, associated Fleiss' kappa statistic, and examples of confidential information
|
Label type
|
Percent of total notes reviewed
|
Average Fleiss' kappa
|
Examples
|
|
Sexual health
|
6.50%
|
0.67
|
1. “Denies that she is sexually active”
2. “Interested in Nexplanon today”
3. “Chlamydia and gonorrhea was obtained, but infection is unlikely”
|
|
Drugs and alcohol
|
8.17%
|
0.64
|
1. “He reports that he has never smoked. He does not have any smokeless tobacco history on file”
2. “More recently, she smoked marijuana”
3. “Has essentially cut out vaping”
|
|
Mental health
|
12.08%
|
0.21
|
1. “Denies stress, anxiety, or depression”
2. “Appears to have an anxiety disorder with perfectionistic and obsessive features”
3. “Has minimal anxiety, depression, or OCD behaviors at this time”
|
|
Has any confidential label
|
21.25%
|
0.50
|
N/A
|
Abbreviation: OCD, obsessive-compulsive disorder.
The Fleiss' kappa statistic for inter-rater reliability regarding if a note contained any confidential information was 0.50 between the five reviewers. Fleiss' kappa was the lowest for mental health at 0.21, 0.64 for drugs and alcohol, and was the highest at 0.67 for sexual and reproductive health.
Demographic Data
Departments with the highest note counts included Orthopedics and Sports Medicine (19.8%), Dermatology (12.8%), and Endocrinology (10.2%). There was a similar distribution of notes between males (48.1%) and females (51.8%), and amongst all ages, with slightly fewer notes representing 12-year-olds (14.8%). A significant portion of our cohort were white or Caucasian (41.2%) or unknown (41.7%) where the patient declined to state or the data were missing. A majority of our cohort was English speaking (83.9%) compared to Spanish (12.8%) or other languages (3.3%). Demographic data of the 1,200-note cohort are found in [Table 2].
Table 2
Confidential prevalence categorized by patient sex, age, race, and language spoken
|
Manually reviewed notes (percent of total notes)
|
Identified as confidential (percent of total notes)
|
|
Total
|
1,200 (100%)
|
255 (21.3%)
|
|
Patient sex
|
|
Male
|
577 (48.1%)
|
91 (35.7%)
|
|
Female
|
621 (51.8%)
|
163 (63.9%)
|
|
Unknown
|
2 (0.2%)
|
1 (0.4%)
|
|
Patient age (y)
|
|
12
|
178 (14.8%)
|
26 (10.2%)
|
|
13
|
209 (17.4%)
|
34 (13.3%)
|
|
14
|
215 (17.9%)
|
34 (13.3%)
|
|
15
|
202 (16.8%)
|
47 (18.4%)
|
|
16
|
193 (16.1%)
|
63 (24.7%)
|
|
17
|
203 (16.9%)
|
51 (20.0%)
|
|
Patient race
|
|
Asian
|
174 (14.5%)
|
34 (13.3%)
|
|
Black or African American
|
21 (1.8%)
|
6 (2.4%)
|
|
Native Hawaiian or Other Pacific Islander
|
11 (0.9%)
|
1 (0.4%)
|
|
White or Caucasian
|
494 (41.2%)
|
115 (45.1%)
|
|
Other/unknown/declines
|
500 (41.7%)
|
99 (38.8%)
|
|
Patient language
|
|
English
|
1,007 (83.9%)
|
229 (89.8%)
|
|
Spanish
|
153 (12.8%)
|
23 (9.0%)
|
|
Other
|
40 (3.3%)
|
3 (1.2%)
|
Secondary Analysis
Confidential information was more likely to be found in notes for female patients as compared to males (p < 0.05) as well as for English-speaking patients as compared to Spanish-speaking or “other” (p < 0.05). When comparing patient-reported race, we found no differences in the prevalence of notes with confidential information amongst all reported races and those who preferred not to state their race (p = 0.42). Using a logistic regression model, visualized in [Fig. 2], we found that as patients got older, there was a higher prevalence of their notes containing confidential information (p < 0.05). Of departments with at least 10 notes reviewed, the highest prevalence of confidential notes were Gynecology (67%), Endocrinology (51%), and Eating Disorder Clinic (44%).
Fig. 2 The probability of confidential information in notes by age with logistic regression analysis.
Discussion
This is the first study to report detailed counts of confidential information in adolescent clinical progress notes. With an overall prevalence of 21.3%, this study demonstrates that progress notes pose a significant risk to breach adolescent confidentiality if released electronically in their entirety. For adolescents, conversations held in confidence with their health care providers may be leaked to parents, leading to a further decrease in trust and adolescents seeking less health care for sensitive topics. Providers should continue to adapt protected methods of documenting confidential information while still providing the standard of care for this age group. This study suggests that there are certain populations that are at higher risk of obtaining and leaking confidential information including females and older adolescents, supporting results previously reported by Lewis Gilbert et al.[32] This could be explained by women requiring medical care for hormonal contraceptive use and other sexual and reproductive care, or potential provider bias in documenting confidential information for females more often. It has also been found that females more often seek outpatient care for mental health services.[33] Notes for non-English speaking patients may be less likely to contain confidential information given barriers to communication.[34] Directing provider education at high-risk departments is a prudent first step in a resource-limited environment to protect confidential information.
Even with a consensus from our team of adolescent medicine physicians and a review of California adolescent toolkits, there remains variation on what a specific provider may identify as confidential, as evidenced by low inter-rater reliability scores across most domains. The lowest score for mental health may be due to variation in interpretation of the California state law and its complexity. Some discrepancies in labeling included information that reported parental concern for mental illness and individual reviewers labeling portions of a physical exam like affect, while others did not. While there are regulations describing general guidance for consent and confidentiality, as of this study, there is little detail on the granular details of confidentiality. While an exhaustive list of confidential topics and scenarios would be useful to guide providers, state variation and consensus makes this a difficult task.[23] If a group of highly trained reviewers are unable to agree on what is considered confidential, it is unlikely that the entire physician workforce can address this in a standardized way. The rubric used in this study represents an interpretation of California confidentiality laws and recognizes that there are other valid decision points when attempting to define what is confidential. For example, while our conservative approach considered statements denying confidential behavior as a positive label, others may not consider this to be legally confidential. This rubric represents a starting point to quantify the prevalence of confidential content and when applied broadly, organizations should come to an agreement of their own definition of confidentiality and remain consistent in how they document.
Health care organizations recognize that while the responsibility lies on individual providers to understand and follow state law when documenting confidential information, they must create efficient and simple workflows to make this process easier for busy clinical providers. These workflows must also relay confidential information during chart review for other providers to intuitively find when necessary. Interventions may include developing new note types or providing just in time education with disappearing note text to help clinicians decipher this nuanced topic.[35] At SCH, providers can document in a specific adolescent confidential note type with the future intent to implement electronic release to the adolescents' electronic health portal while being withheld from a proxy. This note type was released in 2020 at SCH, 1 year after the study period. Since this study analyzed notes prior to the implementation of the 21st Century Cures Act, it can be inferred that the organizational release of historical notes would pose a large risk to confidentiality. Organizations that have not yet implemented a confidential note type may have a quantifiably larger amount of confidential information in progress notes in their historical note release.
Limitations to this study include an inability to garner the entire context of information if deemed confidential. For example, if a parent was the historian for the confidential information, it logically would not be considered legally confidential to them. However, if not documented, a reviewer is unable to know if that is the case. Instances where the adolescent consented to the sharing of this information but it was not explicitly documented is another gap in our ability to identify confidential information. SCH has a high rate of subspecialty care, where these providers may be less likely to have confidential conversations and the prevalence of confidential information may be different in a heavily primary care setting where it is routine to obtain a Home, Education, Activity, Drugs, Sexuality, Suicidality (HEADSS) assessment. There remains ambiguity in the definition of confidential information after defining a confidential rubric and training reviewers, including defining confidential physical exam information, negation, and psychotropic medication use. This may have biased the counts of confidential information depending on individual reviewer interpretations. Reproducibility of this study may be limited by physician agreement on what is labeled confidential as evidenced by moderate Fleiss' kappa scores. Providers had access to a general confidential note type during the period of this study which may have appropriately lowered the amount of potentially confidential information in progress notes. Generalizability of this study is limited by California state law on adolescent confidentiality.
Nationally, there must be a continued voice for adolescent patients who are unlikely to have the means to protect themselves against breaches in confidentiality. Electronic health record vendors should partner with professional organizations such as the American Academy of Pediatrics and the Society for Adolescent Health and Medicine to create innovative solutions to documenting and segmenting confidential information. Professional organizations can work toward creating a consensus standard to define what is considered confidential information, such as behavioral negation or variation based on state law. While health care organizations continue to release health care data electronically while protecting patient privacy, the Office of the National Coordinator should release guides and considerations for these nuanced scenarios, acknowledging that there are technical limitations to granularly segmenting confidential information. While the spirit of the OpenNotes initiative and mandates of releasing confidential information is a step in the right direction, it places adolescents in a vulnerable situation where they may not have a voice to protect themselves.[20] The technical infeasibility exception to the 21st Century Cures Act information blocking rules may currently protect adolescent confidentiality while workflows for documenting confidential information are established.[36]
Conclusion
Our study found that confidential information is present in 21.3% of our sampled adolescent outpatient progress notes. If these historical progress notes were released electronically to proxies, there would be a significant concern for breaching adolescent confidentiality. There is a need for a scalable solution that can continue to expand access to patient health care data while protecting the privacy of the adolescent and preventing potential breaches of confidentiality.
Clinical Relevance Statement
Clinical Relevance Statement
In this retrospective chart review that included a random sampling of 1,200 adolescent outpatient progress notes, there was a 21.3% prevalence of confidential information. The electronic release of adolescent progress notes to proxies has a significant potential to break adolescent confidentiality. This work identifies gaps in clinical documentation workflows that require innovative approaches to information sharing following the 21st Century Cures Act.
Multiple-Choice Questions
Multiple-Choice Questions
-
What federal regulation requires the immediate and electronic release of electronic patient health information?
-
The Health Insurance Portability and Accountability Act.
-
The Health Information Technology for Economic and Clinical Health Act.
-
The 21st Century Cures Act.
-
The Affordable Care Act.
Correct Answer: The correct answer is option c, the 21st Century Cures Act. The final rules on information blocking require the immediate and electronic release of all electronic health information by October 2022.
-
In California, electronic release of confidential information regarding an adolescent's sexual and reproductive care should be released to which account type, unless explicit consent is given?
Correct Answer: The correct answer is option b, the Adolescent's Account. The information blocking rules defer to state law, where in California this includes sexual and reproductive health care. Unless explicit consent to share this information with a guardian is given by the adolescent, the information is covered by another information blocking exception, or is not considered confidential, it should only be released to the adolescent patient themselves.
