Loneliness, social isolation, and pain following the COVID-19 outbreak: data from a nationwide internet survey in Japan

The aim of cross-sectional study was to investigate the association between loneliness, increased social isolation, and pain following the COVID-19 outbreak. A total of 25,482 participants, aged 15–79 years, were assessed using an internet survey; the University of California, Los Angeles Loneliness Scale (Version 3), Short Form 3-item (UCLA-LS3-SF3) was used to assess loneliness, and a modified item of the UCLA-LS3-SF3 was used to measure the perception of increased social isolation during the pandemic. The outcome measures included the prevalence/incidence of pain (i.e., headache, neck or shoulder pain, upper limb pain, low back pain, and leg pain), pain intensity, and the prevalence of past/present chronic pain. Pain intensity was measured by the pain/discomfort item of the 5-level version of the EuroQol 5 Dimension scale. Odds ratios of pain prevalence/incidence and past/present chronic pain prevalence according to the UCLA-LS3-SF3 scoring groups (tertiles) and the frequency of the perceived increase in social isolation (categories 1–5) were calculated using multinomial logistic regression analysis. The mean pain intensity values among different loneliness and social isolation levels were tested using an analysis of covariance. Increased loneliness and the severity of the perceived social isolation were positively associated with the prevalence/incidence of pain, pain intensity, and the prevalence of past/present chronic pain.

Increased loneliness and social isolation due to the implementation of physical distancing measures and travel restrictions to prevent transmission during the coronavirus disease 2019 (COVID-19) pandemic have adversely affected physical and mental health throughout the world [1][2][3][4] . Loneliness and social isolation are great determinants of health and are associated with quality of life 5 measures and psychological disorders, such as depression 6 , as well as with physical diseases such as cardiovascular diseases 7 and increased blood pressure 8 . Moreover, a systematic review reported that loneliness and social isolation were risk factors of early mortality, with an increased likelihood of death ranging from 26 to 29% 9 .
Loneliness and social isolation are also well-known psychosocial risk factors for the exacerbation of pain 10,11 . A previous epidemiological study using data from the English Longitudinal Study of Ageing (ELSA) cohort reported that loneliness was associated with the higher prevalence of musculoskeletal pain 12 . Other studies demonstrated that loneliness is strongly associated with pain 10,13,14 , and social isolation predicted pain interference 11 . Therefore, a worsening of psychological stress due to the increased loneliness and social isolation experienced during the COVID-19 pandemic may contribute to the onset of pain 15 . Moreover, increased loneliness and social isolation during the COVID-19 pandemic may have a large impact on individuals with chronic pain who have a low tolerance for psychological stress [15][16][17] . However, there is lack of direct evidence of an association between www.nature.com/scientificreports/ pain and increased loneliness and social isolation following the COVID-19 outbreak; this highlights the need for a large population-based study to address an important and urgent health issue. Loneliness is considered to be felt sad and alone, and no consensus definition currently exists. Based on a review of the literature, Peplau & Perlman (1982) offered the following definition: "the aversive state experienced when a discrepancy exists between the interpersonal relationships one wishes to have, and those that one perceives they currently have;" thus, it is an emotionally unpleasant experience 18,19 . In addition, two distinct types of loneliness exist, including emotional loneliness and social loneliness 18 . Emotional loneliness is defined as "results from the lack of a close, intimate attachment to another person" and social loneliness is defined as "results from the lack of a network of social relationships in which the person is part of a group of friends who share common interests and activities" 18 . On the other hand, social isolation is structural and can be described objectively based on the characteristics of the situation 20 ; it is often distinguished from loneliness in an academic context, although loneliness and social isolation are sometimes used collectively, as the concept of social isolation partially overlaps with the concept of social loneliness 18 .
Therefore, the aims of the present study were to investigate the associations between the loneliness and increased social isolation experienced during the COVID-19 pandemic and (1) the prevalence/incidence of pain (i.e., headache, neck or shoulder pain, upper arm pain, low back pain, and leg pain), (2) pain intensity, and (3) the prevalence of past/present chronic pain (pain lasting ≥ 3 months) using data from a large cross-sectional survey of social and health variables related to COVID-19 in Japan.

Methods
Study design. The Japan COVID-19 and Society Internet Survey (JACSIS) was designed to investigate the social and health situation related to the COVID-19 pandemic using an epidemiological approach. From August 25 to September 30 in 2020, a total of 28,000 respondents, aged 15-79 years, were selected from 224,389 candidates who received an e-mail invitation among the approximately 2.2 million panelists registered with a Japanese internet survey agency (Rakuten Insight, Inc., Tokyo, Japan https:// in.m. aipsu rveys. com). A random sampling method was used to recruit participants using a computer algorithm; the sample was representative of the official Japanese demographic composition as of October 1, 2019 based on categories of age, sex, and living area (i.e., prefecture). All participants provided web-based informed consent before responding to the online self-report questionnaire. More specific information pertaining to the JACSIS, along with a detailed description of the method used to recruit participants is described in the Supplementary Methods. The present study employed a cross-sectional design intended to detect changes in biopsychosocial and socioeconomic factors before and after the COVID-19 outbreak.
Study population. Of 28,000 participants in the JACSIS, we analyzed data from 25,482 participants (12,809 women and 12,673 men) after excluding 2518 participants who provided invalid responses or met other exclusion criteria; these measures to consistently validate the data quality were performed as described in previous studies 21,22 . The participants with invalid responses were detected through the use of a dummy item, which stated, "Please choose the option second from the bottom. " Participants who selected any option other than the one indicated were excluded (n = 1955). Participants using "all" recreational substances and medications (i.e., sleeping pills, anxiolytic agents, legal/illegal opioids, cannabis, cocaine, etc.) were excluded, as were those with "all" chronic diseases (i.e., diabetes, asthma, stroke, ischemic heart disease, cancer, mental disease, etc.); this resulted in the exclusion of 422 and 141 participants, respectively. The characteristics of the 25,482 participants who were included in the final analysis are shown in Supplementary Table 1.

Main outcome measures.
Loneliness during the COVID-19 pandemic. A Japanese version of the University of California, Los Angeles Loneliness Scale (Version 3), Short Form 3-item (UCLA-LS3-SF3) was used to assess loneliness; the scale was originally developed in English, although both the English and Japanese versions have previously been determined to be valid and reliable 23,24 . The items were as follows: (1) "How often do you feel that you lack companionship?"; (2) "How often do you feel left out?"; and (3) "How often do you feel isolated from others?" Participants were asked to rate the frequency in which these feelings were experienced over the past 30 days using a five-point scale, ranging from 1 to 5 [1 (Never), 2 (Rarely), 3 (Sometimes), 4 (Often), 5 (Always)]; this was modified from the original UCLA-LS3-SF3, which used a four-point frequency scale 23,24 . The total UCLA-LS3-SF3 scores in the current study ranged from 3 to 15, with higher scores indicating more severe loneliness 23,24 . We defined those in the first tertile (3 points), the second tertile (4-5 points), and the third tertile (6-15 points) of the UCLA-LS3-SF3 scores as having 'no loneliness' , 'mild loneliness' , and 'moderate-to-severe loneliness' , respectively. The Cronbach's alpha value of the internal consistency for the three items was 0.93. Pain intensity. Pain intensity was assessed as a continuous variable based on the pain/discomfort item of a Japanese adaptation of the 5-level version of the EuroQol 5 Dimension (EQ5D-5L) scale 25 . The item was phrased as follows: "Please choose the option that best describes your health today. " Participants rated their present pain/ discomfort on a scale of 1-5 based on the following descriptions: (1) I have no pain or discomfort, (2) I have slight pain or discomfort, (3) I have moderate pain or discomfort, (4) I have severe pain or discomfort, or (5) I have extreme pain or discomfort.
Past and present chronic pain. Participants were asked if they had experienced chronic pain for three months or more; the responses were based on the following categorical options: "none, " "I have a history of chronic pain but have already recovered", "yes, receiving treatment", or "yes, without treatment. " The data of individuals who had a history of chronic pain but had already recovered (i.e., history of past chronic pain) were collected because those individuals may have had a lower tolerance for psychosocial stressors, such as loneliness and social isolation, similar to the individuals who reported that they were currently experiencing chronic pain. We considered the latter two categories (having chronic pain with or without treatment) as history of present chronic pain, and we classified the participants into the following three categories: those without chronic pain, those with a history of past chronic pain, and those with a history of present chronic pain (currently experiencing chronic pain).
Potential confounders. Potential confounders were mainly adapted from previous population-based cohort studies that have examined the associations between psychosocial factors and pain experience [26][27][28] . These included demographic factors, socioeconomic factors, lifestyle factors, and a history of diseases related to the explanatory and outcome variables.
Socioeconomic factors. We collected data on the level of education completed (less than high school, high school, vocational school, junior or technical college, university, graduate school, or other), marital status (married or common-law, single, divorced, or widowed), whether the individual was living alone (yes or no), employment status (company executive, owner of a family-operated business, employee of a family-operated business, management-level employee, full-time employee, contract employee, part-time employee/on-the-side worker, student, retired, full-time homemaker, or unemployed), and equivalized income (categorized into quintiles).
Equivalized income was calculated by dividing the median value of the multiple-choice annual household income before-tax by the square root of the number of people living in the household. We defined poverty as an annual equivalized income of less than 1.22 million JPY, which was the poverty line in 2018 as defined by the Organisation for Economic Co-operation and Development 25 .
Life style. We collected data on smoking status (non-smoker, ex-smoker, or current smoker), alcohol consumption (never, ex-drinker, social drinker, < 23 g per day, 23-45 g per day, or ≥ 46 g per day), changes in physical activity before and after the COVID-19 outbreak (decreased, no change, or increased), and sleep duration (< 4 h, 4-5 h, 6-7 h, 8-9 h, ≥ 10 h, or hard to respond/unsure). We defined any individual with alcohol consumption ≥ 46 g per day as a heavy drinker.
History of mental diseases. We collected data on the history of depression (categorized as none, a history of depression but already recovered, or comorbid depression) and the history of mental illnesses other than depression (categorized as none, a history of mental illness other than depression but already recovered, or comorbid mental illness other than depression). Statistical analysis. First, we determined the P for trend of the means and proportions of potential confounders according to the severity of loneliness based on the UCLA-LS3-SF3 scoring groups and the frequency in which participants felt increased social isolation during the COVID-19 pandemic using a general linear model. Second, we examined the association of loneliness and increased social isolation and the prevalence/incidence of pain experienced during the COVID-19 pandemic. The odds ratios (ORs) of the prevalence and incidence of pain (i.e., headache, neck or shoulder pain, upper limb pain, low back pain, or leg pain) according to the UCLA-LS3-SF-3 scoring groups were calculated using a multinomial logistic regression model, with adjustment for potential confounders.
Third, the overall adjusted mean values of pain intensity for all participants and for individuals with any pain symptoms according to the UCLA-LS3-SF3 scoring groups and the frequency in which feelings of increased social isolation were experienced during the COVID-19 pandemic were tested using an analysis of covariance, with Dunnett's test. Individuals with any pain symptoms were defined as those with headache, neck/shoulder pain, upper limb pain, low back pain, leg pain, or the presence of chronic pain. P values were calculated to compare those reporting a lack of loneliness or those who did not feel an increase in social isolation during the COVID-19 pandemic at all to those of all the other categories. www.nature.com/scientificreports/ Finally, we examined the association between the loneliness and increased social isolation felt during the COVID-19 pandemic and the prevalence of past and present chronic pain. The ORs of the prevalence of past and present chronic pain according to the UCLA-LS3-SF3 scoring groups and the frequency of feelings of increased social isolation were calculated using a multinomial logistic regression model, adjusted for potential confounders.
A multinomial logistic regression analysis is often used to estimate multiple categorical outcomes (i.e., the prevalence/incidence of pain and the prevalence of past/present chronic pain). In addition to the ORs, the P for trends according to the UCLA-LS3-SF3 scoring groups and the frequency of feelings of increased social isolation during the COVID-19 pandemic were calculated using a general linear model. Model 1 was adjusted for age and sex, whereas Model 2 was adjusted for age, sex, level of education, marital status, living arrangement (living alone or not), employment status, equivalized income, smoking status, alcohol consumption, physical activity, sleep duration, history of depression, and history of mental illnesses other than depression.
Missing values of potential confounders were used as dummy variables. P values < 0.05 (two-tailed tests) were considered statistically significant. All statistical analyses were performed using Statistical Analysis Software (SAS), Version 9.4 (SAS Institute Inc., Cary, NC, USA). Table 1. Mean values and proportions for participant characteristics (n = 25,482). COVID-19 coronavirus disease 2019, SD standard deviation. Loneliness was measured using theJapanese version of the University of California, Los Angeles Loneliness Scale (Version 3), Short Form 3-item tool. Poverty was defined as an equivalized income less than 1.22 million Japanese yen. Individuals with alcohol consumption ≥ 46 g per day were defined as heavy drinkers. P for trend values were calculated using a generalized linear model. The internet survey agency respected the Act on the Protection of Personal Information in Japan. All participants provided web-based informed consent before responding to the online questionnaire. This study was exempted from the obligation to obtain informed consent from the parents/guardians of minors under the age of 18 in Japan. The Ethical Guidelines for Medical and Health Research Involving Human Subjects enforced by Japan's Ministry of Education, Culture, Sports, Science and Technology and Japan's Ministry of Health, Labour and Welfare addressed, "when the research individuals have completed junior high school or another relevant schooling, or is 16 years or older, and is considered to have enough judgment concerning the research to be implemented on themselves, and the following matters are prescribed in the research protocol, and the chief executive of the research implementing entity approves to carry out the research after relevant ethical review committee deliberation, informed consent shall be obtained not from representative but from the said research subject. (1) The research to be implemented does not involve any invasiveness; and (2) Information concerning the implementation of the research, including the purpose of the research and how specimens or information will be handled, is made public, and opportunities to refuse that the research is commenced or continued on the research subject are ensured for persons who exercise parental authority over the said research subject and guardians of the minor 29 . " All participants completed junior high school, the present study did not involve any invasiveness, and the approval of the Research Ethics Committee of the Osaka International Cancer Institute for the study protocol was obtained as aforementioned. A credit point known as "Epoints", which could be used for internet shopping and cash conversion, was provided to the participants as an incentive.

Results
Descriptive statistics. Of   www.nature.com/scientificreports/ 33.3%, and 17.8%, respectively. The incidences of headache, neck or shoulder pain, upper limb pain, low back pain, and leg pain after the start of the COVD-19 outbreak were 2.0%, 2.4%, 1.8%, 2.1%, 1.5%, respectively. In terms of pain intensity, a total of 7% reported experiencing moderate-to-extreme pain. A total of 6.3% of participants had a history of chronic pain but had already recovered (i.e., a history of past chronic pain), 3.9% had a history of present chronic pain but were receiving treatment, and 6.5% had a history of present chronic pain that was not being treated.
The mean values and proportions related to the characteristics of the participants according to their UCLA-LS3-SF3 scoring group and the frequency of feelings of increased social isolation are indicated in Table 1. Compared with participants who did not report experiencing loneliness (the first quartile of the UCLA-LS3-SF3 score), those within the second to fourth quartiles based on the UCLA-LS3-SF3 scores were more likely to be younger and female, to have a lower educational level, to be divorced, to live alone, to be experiencing poverty, to be current smokers, to have decreased their level of physical activity during the COVID-19 pandemic, to sleep less than six hours per night, to have comorbid depression, and to have comorbid mental illnesses other than depression; they were less likely to be obese and heavy drinkers. The prevalence of unemployment did not vary among the UCLA-LS3-SF3 scoring groups.
Compared to participants who did not report feeling an increase in social isolation, those who felt more socially isolated during the COVID-19 pandemic were more likely to be younger and obese, to have a lower educational level, to live alone, to be unemployed and living in poverty, to have decreased their level of physical activity during the COVID-19 pandemic, to sleep less than six hours per night, to have comorbid depression, and to have comorbid mental illnesses other than depression; they were also less likely to be heavy drinkers. The prevalences of those who were divorced and those who currently smoked did not vary based on the perceived increase in social isolation.  www.nature.com/scientificreports/ Prevalence/ incidence of pain. Table 2, Figs. 1, 2, 3, and 4 indicate that both loneliness and the perception of increased social isolation during the pandemic were positively associated with the prevalence and incidence of all types of pain (i.e., headache, neck or shoulder pain, upper limb pain, low back pain, and leg pain). Table 3 and Fig. 5 indicate the differences in reported pain intensity according to the UCLA-LS3-SF3 scoring groups and the frequency with which participants reported an increase in feelings of social isolation. Compared to participants who did not experience loneliness or increased social isolation, those  pandemic. Odds ratios of each pain symptom in model 2 according to the frequency of feelings of social isolation compared to never felt feelings of social isolation were indicated. X-axis indicated odds ratio. Bars indicated 95% confidence intervals. Abbreviation: OR odds ratio. www.nature.com/scientificreports/ who did reported more severe pain intensity. In Model 2, for individuals who reported any pain symptoms, the adjusted mean pain intensity values among those reporting a lack of loneliness, mild loneliness, and moderateto-severe loneliness were 1.5, 1.6, and 1.8, respectively; among those reporting increased feelings of social isolation during the COVID-19 pandemic, the adjusted mean pain intensity values for the five frequencies were 1.6, 1.7, 1.8, 1.9, and 2.0.

Pain intensity.
History of past and present chronic pain. Table 4 and Fig. 6 indicate that loneliness and a feeling of increased social isolation were both positively associated with the history of past and present chronic pain.

Discussion
The predominant location where participants reported experiencing pain during the COVID-19 pandemic was the neck or shoulders, followed by low back pain. Of the total number of participants, 6.3% reported a history of past chronic pain, and 10.4% were presently experiencing chronic pain. Both loneliness and a perception of increased social isolation during the COVID-19 pandemic were positively associated with the prevalence/ incidence of all types of pain, pain intensity, and the prevalence of past/present chronic pain in a dose-response manner.
These results were consistent with previous studies which investigated the negative effects of loneliness and social isolation with pain 10,11,13,14 . However, the association between social isolation and a higher risk of pain was partly contrary to a previous study that reported that people who experienced chronic musculoskeletal pain had a higher risk of loneliness but had a lower risk of being socially isolated 12 . They measured structural social isolation (e.g., marital/cohabiting status, social contact, and social participation) using a questionnaire 12 , which was different from the social isolation we used, which was similar to the awareness of social loneliness. This difference in the assessment of social isolation may induce different results. While they assumed that people with musculoskeletal pain might increase in contact with family and friends or social networks in support of taking them to healthcare appointments, people with pain symptoms in the present study may find it difficult to contact or make social network-related healthcare appointments during the COVID-19 pandemic.
To the best of our knowledge, the present study is the first to provide direct evidence of the impact that loneliness and the perception of increased social isolation during the COVID-19 pandemic has had on individuals with pain and a prior history of chronic pain. We assumed several underlying mechanisms that could explain these findings. Table 3. The association of loneliness and a feeling of increased social isolation with pain intensity. Model 1: Adjusted for age and sex. Model 2: Adjusted for age, sex, educational level, marital status, living alone, employment status, equivalized income, smoking status, alcohol consumption, physical activity, sleep duration, history of depression, and history of mental illnesses other than depression. COVID-19 coronavirus disease 2019, OR odds ratio, CI confidence interval. Loneliness was measured using the Japanese version of the University of California, Los Angeles Loneliness Scale (Version 3), Short Form 3-item assessment tool. Pain symptoms were headache, neck/shoulder pain, upper limb pain, low back pain, leg pain, or the presence of chronic pain. P values were calculated to compare participants who without loneliness / did not feel increased social isolation during the COVID-19 pandemic at all to those in the other categories using the analysis of covariance with Dunnett's test. ***p < 0.001. n = 25,482. www.nature.com/scientificreports/ In general, loneliness and social isolation are factors associated with psychological stress and distress that can induce depressive symptoms 30,31 , and a previous meta-analysis revealed that loneliness had a moderate impact on depression 6 . Depressive symptoms have been shown to be comorbid with pain and increased pain intensity 32,33 ; in turn, loneliness and social isolation would be associated with increased pain responses or perceptions mediated by depressive symptoms, even in the absence of an actual COVID-19 infection.
As a biophysiological mechanism, social isolation may change the threshold of pain sensitivity 34 . In animal models, social isolation in rodents has been shown to result in changed pain thresholds 34,35 ; for example, in group-housed mice with immobilization stress, increased pain thresholds have been reported following morphine administration, although this effect was not observed in socially isolated mice 34 . Thus, social isolation due to the COVID-19 pandemic may change the perception of pain in humans, as increased pain intensity could be mediated by a change in the perception of pain.
A previous cohort study investigated the impact of a catastrophic event on pain responses in humans; the study examined those with pain before and approximately six months after the September 11th terrorist attacks in New York and New Jersey, showing that residents reported an increased prevalence of musculoskeletal pain, but not pain from fibromyalgia, after the terrorist attacks 36 . Although the impact of a terrorism event on health may differ from that of a pandemic, and although we did not investigate fibromyalgia, the results of the present study are consistent with the reported changes in musculoskeletal pain.
The findings of the present study indicate that it is important for healthcare providers and policymakers to pay attention to the increased loneliness and social isolation experienced during the COVID-19 pandemic because they may be associated with an increased incidence and worsening of pain symptoms, and the changes may be exacerbated in individuals who have a history of present chronic pain.
A strength of the present study was that it included a large, population-based sample of approximately 25,000 individuals, providing sufficient statistical power for analysis. The present study had several limitations, however. First, we could not determine whether individual participants had been diagnosed with COVID-19 or not. An actual COVID-19 infection could be a potential cofounder in the present study, as individuals who were COVID-19-positive would have been required to adhere to strict social isolation procedures over a couple of weeks. Second, we examined participants' perceptions of increased social isolation and did not objectively examine the impact of this social isolation on pain (e.g., based on geographical conditions and events). Objective details about social isolation could provide additional useful information. Third, the web-based survey may have biased the distribution and the present study may not be truly representative of the general population, although the demographic profile of the participants was consistent with the official Japanese demographic composition www.nature.com/scientificreports/ based on the Japanese Vital Statistics for age, sex, and living area as of October 1, 2019 37 . The current prevalence of chronic pain (pain duration ≥ 3 months; 10.4%) was one-quarter of that described in a previous report (39.3%) that was based on a postal survey of Japanese adults, aged 20 years or older, living in a single city 26 . The prevalence of present chronic pain reported in previous web-based surveys was approximately 20%, although these studies were limited by the fact that they only assessed those with moderate-to-severe pain intensity (numerical rating scale score of ≥ 5 out of 10) 27,28 . This discrepancy may be a result of the fact that the present study included more healthy individuals compared with previous surveys. In addition, the participants who were recruited in previous surveys may have had a more particular interest in pain compared with the general population; therefore, the current study population may have been more representative than those of previous reports. Finally, there may have been some recall bias in this study, as it used a cross-sectional design; thus, temporal aspects cannot be discussed. For example, individuals experiencing pain who felt increased social isolation following the COVID-19 www.nature.com/scientificreports/ outbreak may be more likely to recall the onset of their pain as being related to the pandemic. Future studies will need to clarify this using a prospective design.
In conclusion, in accordance with their severity, loneliness and the perception of increased social isolation during the COVID-19 pandemic were positively associated with the prevalence/incidence of pain, pain intensity, and the prevalence of past and present chronic pain. These findings suggest that loneliness and social isolation should be considered, especially as they relate to pain experienced during the COVID-19 pandemic period.

Data availability
The data used in this study are not available in a public repository because they contain personally identifiable or potentially sensitive patient information. Based on the regulations for ethical guidelines in Japan, the Research Ethics Committee of the Osaka International Cancer Institute has imposed restrictions on the dissemination of the data collected in this study. All data enquiries should be addressed to the person responsible for data management, Dr. Takahiro Tabuchi at the following e-mail address: tabuchitak@gmail.com.