Issue Brief - September 2018

Predictive Models for Social Determinants of Health in KP Members and Communities

Kaiser Permanente (KP) Primary Authors:


The World Health Organization defines social determinants of health as “conditions in which people are born, grow, work, live and age, and the wider set of forces shaping the conditions of daily life.” ( Social determinants are the upstream causes that put people at “risk of risks” (Glass, 2006), and that may manifest in specific basic resource needs such as unsafe housing or inadequate access to food or transportation. These non-medical needs, in turn, are critical determinants of health and the effectiveness of health care (Phelan, 2005; Braverman, 2014).

The executive leadership of Kaiser Permanente (KP) has committed to “identify and address individuals’ most pressing basic human needs as a standard part of quality healthcare and to achieve health equity,” in partnership with the communities KP serves. As part of this initiative, KP Community Health established the Social Needs Network for Evaluation and Translation (SONNET), an interdisciplinary consortium of researchers and evaluators with expertise in assessment of social needs, pragmatic interventions, and implementation science. Consistent with the KP Research Strategy, SONNET provides consultation to clinical and operational programs, promotes communication of new developments in KP, and leads projects to advance the social needs agenda of the organization. Out of the broad array of social determinants, SONNET focuses primarily on five basic resource needs (housing, food, transportation, energy/ utilities, and medical costs) that underpin health behaviors and participation in health care and affect health outcomes.

In this Issue Brief, a working group of SONNET investigators and colleagues from other KP departments discusses the role of predictive analytics in identifying basic resource needs and assessing their effect on health outcomes. We define predictive analytics as the development of statistical models “combining a number of characteristics (e.g. related to the patient, the disease, or treatment) to predict a diagnostic or prognostic outcome” (Amarasingham, 2014), Throughout, we highlight approaches to conduct predictive analytics at both the individual (member) level and the community level.

Full version of SONNET's Predictive Models Issue Brief:

SONNET's Predictive Models Issue Brief is only accessible to KP or medical group employees at this time. The information is in an unofficial and informal state, generated for internal quality improvement purposes.

KP Staff View

Registration process for new users:

  1. First time KP users must click the link to Register as a new user.
  2. Register by providing your KP email and creating a password (password must be at least 8 characters long, and contain all of the following: one uppercase, one lowercase, one digit, and one special character).
  3. Once you click Register, an email (subject: SONNET - Confirm your email) will be sent to the address provided.
  4. Confirm your email by clicking the link within the email - a new browser window will open with the following message: Thank you for confirming your email. Please click here to log in.
  5. You will find the Scoping Review under the Documents tab.

If you have any questions regarding the approval process, please contact our Operations Core team member, Caroline Joyce - Email:

KP Staff View

You will be directed to a secure login page outside SONNET's website.

Share this post