Loneliness in Primary Care Patients: A Prevalence Study

Rebecca A. Mullen, MD, MPH; Sebastian Tong, MD, MPH; Roy T. Sabo, PhD; Winston R. Liaw, MD, MPH; John Marshall; Donald E. Nease Jr, MD; Alex H. Krist, MD, MPH; John J. Frey III, MD

Disclosures

Ann Fam Med. 2019;17(2):108-115. 

In This Article

Methods

This study was a cross-sectional survey of adult patients presenting for outpatient care from April 2017 through January 2018 in 2 practice-based research networks, the State Networks of Colorado Ambulatory Practices and Partners (SNOCAP) and the Virginia Ambulatory Care Outcomes Research Network (ACORN). The study was approved by the Colorado Multiple Institutional Review Board and Virginia Commonwealth University Institutional Review Board.

Setting and Population

SNOCAP is a collaborative of 5 practice-based research networks in Colorado that includes more than 600 primary care clinicians and represents all major demographic populations in Colorado. ACORN, the only practice-based research network in Virginia, consists of over 500 primary care clinicians and includes over 150 rural, suburban, and urban practices. All practices in the participating networks were electronically notified of the study through e-mails and newsletters and were given an opportunity to participate; interested practices e-mailed research coordinators to sign up for the study. Additionally, the research team personally solicited practices likely to be interested collaborators based upon prior experience.

Sixteen practices participated in the study, 7 ACORN-affiliated and 8 SNOCAP-affiliated practices. The participating ACORN practices were located in Richmond, Fairfax, and Front Royal, serving predominently urban underserved, suburban and affluent, and rural populations, respectively. Participating SNOCAP practices throughout metro Denver, Boulder, and across rural Eastern Colorado, similarly serve diverse populations.

Participants were recruited by convenience sampling. Patients aged 18 years and older (the upper limit being people aged 89 years in Colorado due to institutional review board restrictions) were asked to complete a paper survey before their appointment. Patients who were unable to read English were excluded. At ACORN-affiliated practices, student and research coordinators approached all adult patients with an office visit during the study period to complete the survey. Patients were approached on consecutive days until 100 responses per practice were obtained. Due to availability of resources, at SNOCAP-affiliated practices, front desk staff offered the surveys to all eligible patients until either 100 responses were collected or until the end of a consecutive 4 to 7 day collection period, whichever occurred first. Patients were not reimbursed for survey completion and participation was blinded to the clinician. Given the nature of our deidentified study methods, we did not collect overall clinic response rate estimates or monitor if participants completed the survey more than once.

Measures

Survey questions included validated measures of loneliness as well as sociodemographics, health-related quality of life, and health care utilization (Supplemental Appendix, available at http://www.AnnFamMed.org/content/17/2/108/suppl/DC1/). Demographic information included age, sex, zip code, marital status, employment status, and current relationship status (see Supplemental Appendix for survey instrument). Respondent zip code data was used to interpret urban vs rural status based upon the United States Department of Agriculture Rural-Urban Commuting Area Codes.[18]

We used the 3-item University of California, Los Angeles (UCLA) Loneliness Scale to measure loneliness.[19] This shortened screening tool has illustrated reliability and correlation to the full 20-item Revised UCLA Loneliness Scale, and addresses lack of companionship, feeling left out, and feelings of isolation from others.[20] Respondents replied on a 3-point scale from never, sometimes, or often, corresponding to scores of 1–3 for each item, and received a total score of 3–9. A total score of 6 or above was considered lonely.[21]

Participants were asked to report on their health using Healthy Days Measures from the Centers for Disease Control and Prevention measurement of Health-Related Quality of Life.[22] Overall health status was measured by responses to "in general, how would you rate your health" on a 5 point scale (1 = poor, 2 = fair, 3 = good, 4 = very good, 5 = excellent). Respondents provided an estimate of the number of days that his or her physical or mental health was not good during the prior month. The patient questionnaire also included 3 questions asking about emergency department visits, primary care visits, and hospitalizations within the prior 12 months. Question structure was based upon the Centers for Disease Control and Prevention National Health Interview Survey.[23]

Statistical Analysis

Continuous measurements were summarized with means and standard deviations (SD), while categorical measurements were summarized with frequencies and percentages. Differences in patient characteristic distributions between health networks were assessed using χ 2 tests. Associations between the categorical loneliness classification and patient characteristics were evaluated with generalized linear mixed models, with the loneliness classification modeled as a binary outcome against a single fixed effect for each patient demographic, and with Froma practice-level random effect to account for inter-practice variability.

Unadjusted and adjusted odds ratios were estimated between patient characteristics and loneliness classification using generalized linear models. Unadjusted estimates were obtained by separately fitting each characteristic (sex, age, race, employment, relationship, location, health) against the binary loneliness classification, with a practice-based random effect to account for between-practice variability. Adjusted estimates were obtained using 2 approaches: by jointly modeling all characteristics against loneliness classification, and by jointly modeling all significant characteristics against the loneliness classification, though omitting the practice-based random effect due to lack of significance. In both cases, some classifications with sparse data (eg, other, prefer not to answer) were removed and some levels with similar meanings and similar loneliness classifications (eg, widowed, separated, divorced; excellent or very good health) were combined to simplify analysis and interpretation. The GLIMMIX procedure in the SAS statistical software package, version 9.4 (SAS Institute, Inc) was used for these analyses.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....