NOTE: As of June 14, 2013, the Project HealthDesign website is now an archive and will not be updated regularly. Please feel free to use the site to learn about our work exploring the power and potential of personal health records. Direct any inquiries to info@projecthealthdesign.org.
  • ShareThis
  • YouTube
HomeResources > Policy > Health Policy Comments and Testimonies > Project HealthDesign comments to NIH on BD2K
 

Project HealthDesign comments to NIH on BD2K

Project HealthDesign
Rethinking the Power and Potential
Of Personal Health Records


March 15, 2013
National Institutes of Health
9000 Rockville Pike
Bethesda, MD 20892


Dear Colleagues,


Thank you for this opportunity to comment on training needs in the BD2K initiative. We strongly support the initiative and agree that new kinds of knowledge are needed to work with large data set analytics. Our experience here at Project HealthDesign bears directly on some of the issues you’re addressing with BD2K. While our data came in small or large sets (sensor data captured every five seconds or observations recorded daily), we have come to recognize that big data means more than just volume.

We would like to see an even further expansion of the BD2K focus to address not only the size of data sets, but the type of data to be analyzed. In order for researchers to be fully prepared to handle future data-intensive enterprises, they need to adopt a broader definition of what constitutes relevant data, particularly when it comes to health care. Clinical profiles of patients rely both on data generated in the clinical area as well as data generated in the everyday lives of patients. We believe that new approaches to behavioral and computational science must account for health data that is generated outside of the clinical or professional setting -- data that is both defined and generated by patients rather than by clinicians. We call this data “Observations of Daily Living” (or ODLs) -- information defined and generated by patients themselves as they go through their daily lives and attend to their own health status. Project HealthDesign has learned that clinical data alone is not enough, and that expanding the health data set to include a patient’s perspective is imperative to a developing workforce charged with analyzing technical innovations, effects of medications, and outcomes of care.

Over the last eight years, our researchers at Project HealthDesign have faced the issues associated with the persistent collection of patient-generated data from a variety of settings through mobile devices and other home-based mechanisms. The sheer volume of health data was enlightening and challenging, but we found even greater challenge and opportunity in processing the rich and varied health data from individual patients, collected in the course of their everyday lives.

Maximizing the value of this rich and varied information must take into account the following idiosyncratic definitions of data elements that sometimes characterize this data:

  • Patient-defined and generated data is unique to each individual. Therefore, possible iterations are infinite and non-translatable among patients. In the case of the Estrellita application, which was designed to collect information from high-risk infants, a “fussy-o-meter” was incorporated that parents could use to track the infant’s temperament across the day. One mother’s data point for a particularly fussy baby may look similar to that of another mother’s calmer one; theidiosyncratic meaning and its implications are markedly different. The inability to map to standard concepts presents an obstacle that is necessary to overcome for high-level analysis.
  • Patient-defined and generated data is often episodic or event-driven—characterized by starts and stops—making it prone to gaps and missing data. The fact that this data is most easily acquired through mobile devices, which are at high-risk for breaches in privacy and security, also leads to situations in which data can be lost. For instance, the application developed by our iN Touch project, which examined low-income teens managing their obesity, featured remote wiping and disabling of devices whose security had been compromised, leading to gaps in the data. Management of persistent data to make it amenable to big data methodology requires the ability to handle missing or inconsistent intervals of data resulting from these events.
  • Patient-defined and generated data does not have to be stored in the EHR, but it must be easily accessible to both the patient and clinician wherever and whenever needed, which necessitates the creation of multiple access points for relevant information. The challenge here is best demonstrated by Project HealthDesign’s BreathEasy application, where patients entered data into a smartphone application which could then be viewed both by clinicians in a web-based dashboard and by patients on a laptop or PC. Analysis of a patient’s health then relied on at least three disparate sources of information: the smartphone application, the web-based dashboard, and the EHR. In order to maintain a comprehensive and accurate understanding of this information, the provenance of the data must be retained at the data element.
  • Patient-defined and generated data accrues greater benefit through accumulation over time and circumstances, enabling judgments about relative health and identification of significant trends. In our dwellSense application, in-home sensors monitored the routine tasks of elders at risk for cognitive decline, such as taking medication, making and receiving phone calls and preparing coffee. In this process, researchers were able to identify patterns where small variations were insignificant, but sudden or large variations potentially heralded a clinical problem. In order to assess these patterns, the entirety of received sensor data must be employed.

As you begin looking at Big Data 2 Knowledge, we advise you to consider not only volume, but high levels of variability and irregularity in the data stream. We thank you again for your efforts in building a workforce that is equipped to analyze health data that will produce the maximum benefit to patients.


Sincerely,
Patricia Flatley Brennan, RN, PhD
eRA Commons Username: pattifbrennan
Director, Project HealthDesign
Moehlman-Bascom Professor of Nursing and Industrial Engineering
info@projecthealthdesign.org

 
Project HealthDesign is a national program of the Robert Wood Johnson Foundation