Pet Ownership, Living Alone, and Cognitive Decline Among Adults 50 Years and Older

Key Points Questions Is pet ownership associated with cognitive decline in older adults, and how does pet ownership mitigate the association between living alone and the rate of cognitive decline? Findings In this cohort study of 7945 participants 50 years and older, pet ownership was associated with slower rates of decline in verbal memory and verbal fluency among individuals living alone, but not among those living with others. Pet ownership offset the association between living alone and declining rates of verbal memory and verbal fluency. Meaning These findings suggest that pet ownership may be associated with slower cognitive decline among older adults living alone.


Assessment of Cognitive Function
Verbal memory was assessed by testing the immediate and delayed recall of 10 unrelated words.The verbal memory score was the sum of words correctly recalled in the immediate and delayed word recall tests, ranging from 0 to 20.Verbal fluency was evaluated using a verbal fluency task in which participants were required to list as many animal names as possible within 1 min.The number of listed animal names was counted as the verbal fluency score, without upper limits.For both tests, a higher score suggests better cognitive performance.In addition, the cognitive assessment date of each participant was recorded.

Assessments of Potential Covariates
Race/ethnicity was self-reported and was categorized as White or other (including Asian, Asian British, Black, Black British, mixed ethnic group, and other). 1 Participants selfreported the highest educational qualification obtained.Education level was classified into high level (university degree or equivalent), middle level (A-level/higher education below degree), and low level (no qualifications/O-level or equivalent). 2Wealth was based on a comprehensive assessment of economic resources (e.g., financial, housing, and physical wealth) excluding pension wealth, and was divided into quintiles (1 = poorest; 5 = richest). 3cial isolation was evaluated using the previously reported Social Isolation Index.4,5 This index included marital status (1 point for unmarried/not cohabiting), contacts with children (1 point for contacting in person or by telephone/written/e-mail less than monthly), contacts with other family members (1 point for contacting in person or by telephone/written/e-mail less than monthly), contacts with friends (1 point for contacting in person or by telephone/written/e-mail less than monthly), and social participation (1 point for not being a member of any organizations such as religious groups, gyms/sports clubs, and committees).
The social isolation score ranges from 0 to 5, with a higher score representing more severe social isolation.According to the frequency of drinking in the past year, we classified drinking status into less than weekly, 1−4 days a week, and 5−7 days a week. 3Physical activity was assessed by asking participants how often they participated in vigorous, moderate, and light physical activities, and was categorized into vigorous, moderate, and light physical activities. 6Depressive symptoms were assessed using the 8-item version of the Center for Epidemiologic Studies Depression Scale.The total score ranges from 0 to 8, and a higher score indicates more severe depressive symptoms.A cut-off value of ≥ 4 was used to identify older adults with depressive symptoms. 7

Inverse Probability Weighting
Referring to previous studies, 8,9 weights were derived from a logistic regression analysis between a set of measures that were independently predictive of missing data (i.e., age, sex, race/ethnicity, education level, employment status, wealth, smoking status, drinking status, physical activity, and the social isolation score, self-rated general health, depressive symptoms, hypertension, diabetes, and cardiovascular disease).We predict the probability that each individual is a complete case (i.e., non-missing) using logistic regression.The inverse of the above probabilities is the weight.Minimal missing data on indicators used to derive weights were singly imputed as the modal value of continuous variables or mean value of categorical variables (all indicators had < 6% of values missing).The Hosmer-Lemeshow test was used assess the fit of the missingness model, with results showing no indication of poor fit (P-value < 0.05).Weights ranged from 1.01 to 7.03.The inverse probability weighting analysis is performed by adding weights in the main analysis.Living alone (no) and pet ownership (no) × time -0.002 (-0.009 to 0.005) .61 Living alone (yes) and pet ownership (yes) × time -0.007 (-0.019 to 0.005) .27 Living alone (yes) and pet ownership (no) × time -0.029 (-0.038 to -0.020) <.001

Verbal fluency
Living alone (no) and pet ownership (yes) × time 0

eFigure 5 .
Estimated Cognition z Scores (SD Units) During Waves 5 to 9 by Pet Ownership in Wave 5, Stratified by Living Alone in Wave 5 Covariates were set to the following values: age = 65 years, female, White race, high education level, retired, the third quintile of wealth, the social isolation score = 1, current non-smokers, drinking less than once a week, moderate physical activity, good self-rated general health, and not having depressive symptoms, hypertension, diabetes, and cardiovascular disease.

eFigure 6 .
Estimated Cognition z Scores (SD Units) During Waves 5 to 9 by the Combination of Pet Ownership and Living Alone in Wave 5Covariates were set to the following values: age = 65 years, female, white race, high education level, retired, the third quintile of wealth, the social isolation score = 1, current non-smokers, drinking less than once a week, moderate physical activity, good self-rated general health, and not having depressive symptoms, hypertension, diabetes, and cardiovascular disease.

eTable 1. Baseline
Characteristics of Participants Included in the Analysis and Participants Lost to Follow-

Up Characteristic Participants included Participants lost to follow-up P-value
Rates of Loss to Follow-Up During Waves 5 to 9 a Modifying Role of Living Alone in Wave 5 in the Association of Pet Ownership in Wave 5 With The model included pet ownership, living alone, time, pet ownership × time, living alone × time, pet ownership × living alone, pet ownership × living alone × time, and covariates (i.e., age, sex, race/ethnicity, education level, employment status, wealth, smoking status, drinking status, physical activity, the social isolation score, self-rated general health, depressive symptoms, hypertension, diabetes, and cardiovascular disease in wave 5).The model included living alone, time, living alone × time, and covariates (i.e., age, sex, race/ethnicity, education level, employment status, wealth, smoking status, drinking status, physical activity, the social isolation score, self-rated general health, depressive symptoms, hypertension, diabetes, cardiovascular disease, and pet ownership in wave 5).The Inverse Probability Weighting Analysis: Modifying Role of Living Alone in Wave 5 in the Association of Pet Ownership in Wave 5 With Cognitive Decline During Waves 5 to 9 a The model included pet ownership, living alone, time, pet ownership × time, living alone × time, pet ownership × living alone, pet ownership × living alone × time, and covariates (i.e., age, sex, race/ethnicity, education level, employment status, wealth, smoking status, drinking status, physical activity, the social isolation score, self-rated general health, depressive symptoms, hypertension, diabetes, and cardiovascular disease in wave 5).The Inverse Probability Weighting Analysis: Association of Pet Ownership in Wave 5 With Cognitive Decline During Waves 5 to 9, Stratified by Living Alone in Wave 5 The Inverse Probability Weighting Analysis: Joint Associations of Pet Ownership and Living Alone in Wave 5 With Cognitive Decline During Waves 5 to 9 a a Wilcoxon rank-sum tests were used to compare the distribution of follow-up time.bKruskal-Wallisranktests were performed to compare the distribution of follow-up time.eTable 3. a Pearson Chi-squared tests were performed to compare the distribution of rates of loss to follow-up.eTable4.a a a Abbreviation: CI, confidence interval.eTable8.
Association of Time-Varying Living Alone With Cognitive Decline During Waves 5 to 9 a The model included living alone, time, living alone × time, and covariates (i.e., age, sex, race/ethnicity, education level, employment status, wealth, smoking status, drinking status, physical activity, the social isolation score, self-rated general health, depressive symptoms, hypertension, diabetes, cardiovascular disease, and pet ownership in wave 5).
a The model included the combination of living alone and pet ownership, time, the combination of living alone and pet ownership × time, and covariates (i.e., age, sex, race/ethnicity, education level, employment status, wealth, smoking status, drinking status, physical activity, the social isolation score, self-rated general health, depressive symptoms, hypertension, diabetes, and cardiovascular disease in wave 5).Abbreviation: CI, confidence interval.eTable 10. a Abbreviation: CI, confidence interval.