Factors Associated With Loss of Usual Source of Care Among Older Adults

Stephanie K. Nothelle, MD; Cynthia Boyd, MD, MPH; Orla Sheehan, MD, PhD; Jennifer L. Wolff, PhD


Ann Fam Med. 2018;16(6):538-545. 

In This Article


USC status was determined from round 1 (baseline) survey responses to the question, "Is there a doctor that you think of as your regular doctor, that is, a doctor you usually go to when you are sick and need advice about your health?" Loss of a USC was defined as the first transition from reporting having a USC in round 1 to reporting no USC in a subsequent round.

We examined older adults' demographic factors, health status, function, and financial and social supports. Demographic factors included respondent type (proxy vs self), sex, age, race/ethnicity, marital status, type of residence, educational attainment, current census division, and whether the participant reported moving to a new residence since the last round. Health status indicators included self-rated health and hospitalization in the last 12 months. We categorized depressive symptoms using previously established cut points for responses to the PHQ-2.[28] Presence of chronic conditions was determined using self-report of a physician's diagnosis of heart attack, heart disease, high blood pressure, arthritis, osteoporosis, diabetes, lung disease, stroke, dementia or Alzheimer's disease, or cancer. Functional status was determined by a report of help needed with self-care activities in the last 12 months. Report of any fall in the last 12 months was also recorded. Income quartiles were constructed using the first round assignment of income.[29] A composite measure of supplemental insurance coverage was created based on report of either supplemental (Medi-gap) or drug coverage. Medicaid coverage was noted separately. Presence of a transportation barrier is a composite measure of participant's response of yes to any of the following: "In the past month, did a transportation problem ever keep you from (visiting friends and family, attending religious services, participating in clubs, classes, or other organized activities, or going out for enjoyment)." As item nonresponse was limited, ranging from 0% to 5%, with most variables at <1%, we assigned missing values the modal value.

Statistical Analysis

Participants who reported having a USC in round 1 (n = 7,226) were followed to identify whether they subsequently reported no USC (n = 469) (Figure 1). As we were interested in loss of USC status, the unit of analysis was the transition from one round to the next for each person-period. In this format, each participant could contribute up to 5 observations. Participants entered the study at round 1 and were followed until they experienced 1 of the following events: loss of USC, death, loss to follow-up, or study period completion (round 6), whichever occurred first.

Figure 1.

Participant flow diagram: loss of usual source of care.
FQ = Facility questionnaire only, no participant interview; USC = Usual source of care

We first examined the National Health and Aging Trends Study sample by baseline report of USC using descriptive statistics, comparing group differences using X2 analyses. We then used multivariable pooled logistic regression to examine which demographic, clinical, and social factors are associated with a loss of USC. Pooled logistic regression is considered equivalent to Cox proportional hazards models when the outcome is uncommon[30] and is intended for data collected at discrete time intervals.[31]

We also conducted a sensitivity analyses in which we removed individuals who reported living in residential care or a nursing home to assess the sensitivity of results to differential loss to follow-up.

All analyses were conducted in Stata version 14 (StataCorp LLC) using the svy commands and weights to account for the complex sampling strategy and to produce population estimates.