Electronic health records (EHRs) have multiple benefits for clinicians and patients, but there have been unintended consequences, including clinician burnout.
In Endocrinology and Metabolism Clinics of North America, Martin Kathrins, MD, a urologist in the Division of Urology at Brigham and Women’s Hospital, and Anna Goldman, MD, an endocrinologist at the Brigham and at Atrius Health, discuss how EHRs can optimize the treatment of hypogonadal men.
Clinical Decision Support Systems (CDSSs)
Clinicians should familiarize themselves with CDSSs for male hypogonadism. CDSSs have led to improved clinical outcomes by encouraging evidence-based test selection, reinforcing best practices and clinical guidelines, and cross-referencing drug allergy and medication interactions.
Biomedical Research
EHRs make it possible to detect previously unknown disease correlations and discover new patterns for classification and prediction of patient outcomes. A list in the article highlights several contributions to the understanding of male hypogonadism that have already come from research using EHRs.
There are limitations on the validity of EHR data, though. It can be affected by:
- Ascertainment bias—Clinicians may code for more common diagnoses, such as hypogonadism, when it is difficult to find codes for uncommon, related conditions or when evaluating a patient with suspected but not confirmed hypogonadism
- Report bias—Available data on prescribing behaviors for the diagnosis of hypogonadism within a given institution may not reflect real-world practice
EHR-based research done with data from health services warrants additional scrutiny compared with research based on government databases and national insurance claims databases, which may not suffer from those biases.
Artificial Intelligence–driven Algorithms
Researchers are working toward applying natural language processing (NLP) to CDSSs in multiple ways:
- Statistical modeling—Probabilistic or machine-learning algorithms for text analysis might be used to monitor prostate cancer survivors for biochemical recurrence by incorporating prostate-specific antigen monitoring over time
- Prospective databases—NLP could be used to extract numerous data points from narrative prostate pathology reports and create a database of information such as the Gleason score and the number of biopsy cores positive for cancer
- Temporal information extraction will require further refinement of NLP technology but is expected to lead to advances such as the ability to capture a diagnosis of erythrocytosis before or after initiation of testosterone supplementation therapy or model the long-term course of patients with chronic disease
It might be possible to link EHR data with genetic data to provide genotype-based decision support.
The authors also discuss the benefits of EHR databases for billing optimization and support of telehealth. The disadvantages they review include information clutter (substantial repetition of information within the same note), depersonalization of the patient–physician interaction, the burden on responding to numerous patient messages, pop-up alert fatigue, and the downsides of using scribes.