Measuring loneliness: guidance for use of the national indicators on surveys

This guidance is for people who may be considering including measures of loneliness on their surveys. It will be useful to those in central or local government considering how best to measure loneliness in keeping with the Government’s Loneliness Strategy announced in October 2018. It will also be helpful to those working in academia or in the private sector who would like to measure loneliness as part of their work. Using the approach recommended here will help to align your survey with this wider initiative and, in doing so, make a more positive contribution to the evidence base on loneliness.

Further guidance specifically intended for third sector organisations interested in measuring their impacts on loneliness is being developed by the What Works Centre for Wellbeing.

In keeping with the Loneliness Strategy, loneliness is defined here as: “a subjective, unwelcome feeling of lack or loss of companionship. It happens when we have a mismatch between the quantity and quality of social relationships that we have, and those that we want.” This is based on a definition first suggested by Perlman and Peplau in 1981¹.

It is important not to confuse loneliness and social isolation. They are different concepts requiring different approaches to measurement. The focus here is on the subjective emotional experience of loneliness, not the objective experience of how often we are alone. They may overlap, but loneliness and social isolation are different and this guidance deals only with the measurement of loneliness.

It is important to measure loneliness because, although it has been found to be associated with a range of poor health outcomes, the evidence on loneliness is currently quite patchy. We have much more robust and extensive data on loneliness in older people, but much less for other age groups including children and young people. We also need to understand more about what factors are most associated with loneliness, what the effects of loneliness are for different people, and how we can prevent or alleviate it. If more people measure loneliness in the same way, we will build a much better evidence base more quickly. That’s why the Prime Minister asked Office for National Statistics to develop national indicators of loneliness for people of all ages, suitable for use on major studies.

Notes for: What is loneliness and why should you measure it?

1. For more information, see Perlman D and Peplau LA (1981) 'Toward a Social Psychology of Loneliness', which can be found in Gilmour R and Duck S (editors) (1981), Personal relationships. 3, Personal relationships in disorder, London: Academic Press, pages 31 to 56.

Loneliness can occur at any point in life and is an experience likely to affect most of us at some point. It becomes a more serious issue, associated with poorer health outcomes, when it is a frequent experience. The Loneliness Strategy summarised some of the most important known effects of loneliness in the following way:

“Feeling lonely frequently is linked to early deaths. Its health impact is thought to be on a par with other public health priorities like obesity or smoking. Research shows that loneliness is associated with a greater risk of inactivity, smoking and risk-taking behaviour; increased risk of coronary heart disease and stroke; an increased risk of depression, low self-esteem, reported sleep problems and increased stress response; and with cognitive decline and an increased risk of Alzheimer’s. What’s more, feeling lonely can make a person more likely to perceive, expect and remember others’ behaviour to be unfriendly. This can increase social anxiety and cause them to withdraw further, creating a vicious cycle.

“…Feeling lonely frequently has a direct impact on individuals and can also have wider effects for society. For example, lonely people are more likely to be readmitted to hospital or have a longer stay. There is also evidence that lonely people are more likely to visit a GP or A&E and more likely to enter local authority funded residential care. At work, higher loneliness among employees is associated with poorer performance on tasks and in a team, while social interaction at work has been linked to increased productivity. A study by the Co-op and New Economics Foundation attempted to calculate the cost of this, estimating that loneliness could be costing private sector employers up to £2.5 billion a year due to absence and productivity losses.”

Many different approaches have been used to measure loneliness. These include both loneliness multi-item scales and single-item measures. Some measures ask about loneliness directly while others ask about emotions associated with loneliness from which loneliness is then inferred. There are advantages and disadvantages associated with each.

Office for National Statistics (ONS) undertook a programme of scoping work and consultation with experts on existing approaches to loneliness measurement and, from this, two measures were selected as meeting many criteria we required in loneliness indicators. We then did further testing of these measures to understand how they work for people of different ages and backgrounds and how well they performed on different types of surveys.

Based on this, we suggest that the “gold standard” is to use both direct and indirect measures of loneliness where possible. This is the approach currently taken by the English Longitudinal Study of Ageing and the Understanding Society study. This enables us to measure responses on a scale that has been assessed as valid and reliable, as well as allowing the respondent to say for themselves whether they feel lonely, providing further insight into the subjective feeling of loneliness for different people.

Also, there is variation in how people understand the term “loneliness” and some people might be reluctant to admit to loneliness, and this might be particularly true of certain groups such as older men. A multi-item measure that does not mention loneliness directly can be helpful to address these issues. The recommended measures are well-tested, and have a positive track record for performing well in general population surveys using different approaches to data collection.

Specifically, we recommend four questions to capture different aspects of loneliness. The first three questions are from the University of California, Los Angeles (UCLA) three-item loneliness scale. The wording of the UCLA questions and response options are taken from the English Longitudinal Study of Ageing. The last is a direct question about how often the respondent feels lonely, currently used on the Community Life Survey. For those aged 16 years and over, the loneliness measures should be as in Table 1.

If survey space is a major constraint, and the “gold standard” is not possible, we would recommend using the direct question on loneliness on its own: “How often do you feel lonely?”. This should be used in conjunction with the five-item response scale detailed in Table 1. To maintain consistency with other established surveys, we recommend using the question and response scale as outlined in Table 1. This will provide comparability with other surveys and enable benchmarking for the minimum amount of survey time and space.

An adapted version of the measures is recommended for use with children and young people aged 10 to 15 years. The wording for the children’s measure was changed to a more “plain English” version, reflecting concerns that the words “companionship” and “isolation” are difficult for children to read and may be interpreted in a range of different ways.

We revised the questions and tested them qualitatively (to understand children’s ease of use and interpretations) on a survey conducted among children by The Children’s Society. The findings showed that the revised questions were interpreted appropriately and that children generally preferred the simplified wording. For those aged between 10 and 15 years, the loneliness questions should be as in Table 2.

As with the questions for adults, we would recommend use of all four questions as the “gold standard”, but where this is not possible due to space constraints, we advise use of the single direct question on loneliness.

An introduction to the questions may not always be necessary and current survey practice varies as to whether questions on loneliness are introduced to respondents or not. An important consideration is how well the questions flow on from preceding questions in the survey and whether it is necessary to introduce the questions to signal a change of topic. The decision may also depend on whether all four questions are used or just one.

If an introduction is used, we recommend that the word “loneliness” is not included because this may affect responses to the first three questions, which are intended to measure loneliness without mentioning the word.

The following introduction was used while testing the questions on the Opinions and Lifestyle Survey (OPN), which involves a telephone interview:

“The next questions are about relationships with others. For each one, please say how often you feel that way.”

The response categories were read out with each question. Feedback from the testing suggests that this general introduction worked well and the same approach could be used in other interview formats and among children and young people, as well as adults.

Aim of the loneliness measures

The questions were designed for use on surveys and are not intended to be used as a tool for diagnosing loneliness in individuals.

Ethical considerations

The potential sensitivity of the questions should be fully considered as part of survey ethics review procedures. In survey question testing, we did not find that people were upset by the questions and as noted, the recommended questions have been in use with general population samples for several years. Similar questions have also been used in mental health service evaluations with no adverse outcomes reported (for more information, see the AdCare evaluation).

In surveys where the sample is more focused on those at greater risk of frequent loneliness, such as people with physical or mental health challenges, carers or unemployed people, we suggest that the questions can be used, but particular thought should be given in advance to any aftercare or support that could be offered to people who identify themselves as lonely.

One of the aims of the Loneliness Strategy is to remove the stigma associated with loneliness and enable people to talk more freely about it. The questions could be a useful way to do that, with the right support in place to continue the conversation after the survey with people who say they are lonely. At the end of the survey, we would suggest providing information about other sources of support that may help to alleviate loneliness. This could include encouraging people to speak to their GP or to contact other locally available services such as befriending or community connector services. Additionally, signposting people to relevant national services such as Childline or The Silver Line may also be helpful.

As part of the survey, in keeping with good practice, it should be made clear to respondents at the outset that all responses are voluntary, that they do not have to answer any questions with which they are uncomfortable and that their responses will be kept confidential and anonymous and used only for the purposes specified at the outset.

Briefing interviewers

If the questions are used on surveys involving interviewers, it is important that interviewers are fully briefed about the reasons for asking the questions, the fact that the questions are in use on national surveys and that they have not been found to affect respondents adversely or cause distress. Interviewers’ own possible apprehensions about raising the subject of loneliness may need to be addressed in advance of the survey, along with approaches to dealing with any problems they encounter and the types of support that can be offered to respondents if required.

Proxy responses

Due to the subjective and sensitive nature of the questions, we suggest that they should not be answered by anyone other than the respondent. For this reason, we would advise that proxy responses not be accepted.

Special considerations for using the questions on surveys of children and young people

Based on consultation with experts from our Technical Advisory Group, we suggest that the adapted questions for use with children and young people are suitable from the age of 10 years. It is from approximately this age that children are generally able to understand and respond in a meaningful way to these types of subjective survey questions and our own testing focused on children and young people from the ages of 10 to 15 years. After this point, young people are usually included in general population surveys of adults.

Some experts on the Technical Advisory Group raised the possibility that less able children may struggle to read or understand the word “occasionally” in the response options for the direct question on loneliness. The Children’s Society conducted additional cognitive testing to explore this. This involved three mini-groups with children aged 10, 12 and 13 years, all of whom were part of youth or activity groups run by The Children’s Society. Their participation was voluntary and each young person was given a high street store voucher to thank them for taking part.

Across all three groups, young people were first asked to answer the questions independently on their own. They were able to answer the questions quickly and easily. After answering the questions, they were asked about their interpretations of the questions and response options. This suggested some confusion about the differences between the words “sometimes” and “occasionally” and which word indicated greater frequency. However, when presented as incremental response options in a scale running from “never” lonely to “often or always” lonely, they were able to select the response option closer to “often or always” as indicating more frequent loneliness.

These findings suggest that children from the age of 10 years are able to use the response scale as intended. To make the scale as clear as possible to younger respondents, we suggest that it is helpful to present the response scale horizontally with “often or always” lonely at one end and “never” lonely at the other end. This may make it easier to choose the response option that best suits them based on how close it is to “often or always” or “never” lonely.

We also recommend that children should be invited to participate in surveys directly as well as asking for parental or guardian consent. As with adults, information about the study, its purposes, who is conducting it and what will happen to their information including issues of anonymity and confidentiality should be provided before the survey, with special attention given to the use of plain English. It should also be emphasised that they do not have to answer any questions with which they are not comfortable.

Being inclusive

If your survey focuses on people with learning disabilities or includes a more accessible version, we suggest that the questions tested for use with children and young people may be a good option. This is because they use plain English that is intended to be easier to read and understand.

How and where people will answer the questions

Our testing work among children and young people indicated that privacy is an important issue in answering questions about loneliness. How and where people will answer the questions” requires a wording change: “Whether the questions are asked in a survey at school or at home was less relevant to most young people than the desire to answer them in privacy, so their answers could not be seen or overheard by others. Although many young people also told us it may be important to seek support from others when you are feeling lonely, they did not necessarily want others to know about their feelings unless and until they are ready to share them.

Although we have not done similar testing with older people, it is likely that similar issues may apply to other age groups as well. Existing surveys using the loneliness questions vary in terms of practices for addressing this issue. For example, the English Longitudinal Study of Ageing (ELSA) uses a self-completion approach to these questions, enabling respondents to complete these more sensitive questions separately from the rest of the survey, which is interviewer-led.

From a survey perspective, this suggests that it may be important to consider the use of self-completion elements within the survey and to look for ways to ensure that people’s privacy will be respected in completing the survey. If the survey is administered in an institutional setting (for example, school, college, university, work, care homes), it is important to ensure that privacy and confidentiality are respected and that respondents do not feel under any pressure to share their responses with others. If the respondents’ privacy cannot be ensured, it would be beneficial to have some indication of whether the respondent was alone when the questions were answered as this may impact data quality.

Mode of survey administration

As we have not tested for mode effects, nor have we found other studies which have, we cannot confidently say whether the mode of survey administration (such as, face-to-face interview, telephone survey, online survey, self-completion questionnaire) affects responses. The cognitive testing suggested that privacy may best be ensured via self-completion modes and that a perceived lack of privacy may affect data quality. The issue of maintaining respondent confidentiality and privacy should therefore be given careful consideration in whichever mode is used.

Question order

We recommended that the three questions of the University of California, Los Angeles (UCLA) scale should precede the direct question about loneliness. The UCLA scale is designed to measure loneliness without mentioning the word, so placing the direct measure before the UCLA questions can be problematic in that respect. Additionally, our survey testing shows that it produces different results than asking the three UCLA questions first, followed by the direct question. We also suggest that the UCLA questions should be kept together and in the order in which they are presented in Section 4 and Section 5 to ensure comparability of the findings with other surveys.

Question placement

The placement of the measures within the survey should be carefully considered. We recommend that surveys do not end with the loneliness questions but finish on another, more neutral subject to avoid leaving respondents in a potentially negative frame of mind.

In the Opinions and Lifestyle Survey (OPN) survey testing, the loneliness questions followed questions on personal well-being. Interviewer feedback suggests this worked well and created a natural flow of topics. The personal well-being questions are also suitable for use on surveys of children and young people, with the questions on life satisfaction, feeling that activities in life are worthwhile, and happiness regularly included on The Children’s Society survey.

Length of time required to ask the questions

In our testing, we found that all four questions and an introduction could be asked in two minutes or less.

OPN survey testing, involving a computer-assisted telephone (CATI) interview, found that it took approximately two minutes to complete the questions, including the introductory preamble. Responses gathered during the OPN interviewer debriefing also specified that the questions did not take long to complete, citing this as one of their main benefits.

Although the questions were tested on The Good Childhood Index Survey, no information was available on the length of time it took for children to complete the online survey. As a rough indication, during cognitive testing, children and young people were read the questions and gave their answers in approximately two minutes as well.

The first three questions for adults and children are those used for the University of California, Los Angeles (UCLA) three-item scale (or an adaptation of this scale for children). Using the questions as intended by the developers of the scale involves assigning a score (as detailed in this section) to each response and creating a total score by summing the individual scores.

For example:

• "Hardly ever or never” equals 1

• “Some of the time” equals 2

• “Often” equals 3

The lowest possible combined score on the loneliness scale is 3 (indicating less frequent loneliness) and the highest is 9 (indicating more frequent loneliness). There is no standard accepted score for which a person would definitely be considered lonely. Instead, we suggest that it may be more helpful to use the average score across the sample to monitor and report changes over time. Additionally, it may also be useful to compare the average scores of different groups of people within the sample.

It is also helpful to look at responses to the individual questions separately, for example, by reporting percentages of people giving each response to each question.

The individual responses to the UCLA questions may also be helpful in interpreting responses to the final question, “How often do you feel lonely?”. This question is not part of the UCLA scale and responses to this question should not be included in the UCLA combined score. Doing so would make your results incomparable with other surveys using the UCLA scale.

The final question also has a different response scale to that used in the first three questions. This is to maintain comparability with an important survey that has been regularly collecting information on loneliness in England for several years, the Community Life Survey. We suggest that this can be used either as a stand-alone measure of loneliness, or in combination with the three questions from the UCLA scale.

There is no special approach to reporting findings from the direct question on loneliness, but it may be helpful to refer to the approach used by the Community Life Survey. They have produced a factsheet on loneliness, which provides a user-friendly approach to reporting the findings.

In reporting the prevalence of loneliness in your study, we suggest using the responses from the direct question, “How often do you feel lonely?”. Although previous studies have shown that some groups, such as men, may under-report loneliness when responding to direct questions, this is currently the best measure we have of understanding people’s own self-perceived experience of loneliness. If the results of the two approaches to measurement suggest differences between groups of people who may be more or less likely to report loneliness (or feelings associated with it), it would be very helpful to indicate this in your reporting of the findings and to share it more widely.

In its first loneliness strategy for England, the UK government has committed to using our recommended loneliness measures as its standard way of measuring loneliness.

All four questions will be adopted in these surveys:

• Community Life Survey

• National Travel Survey

These surveys will include the direct loneliness question only:

• Active Lives Adult Survey

• English Housing Survey

• Taking Part Survey

• Tri-service families continuous attitude survey

Given the important link between health and loneliness, the government is working with various agencies including Public Health England, NHS England and NHS Digital to include the loneliness measures in key surveys, such as the Health Survey for England.

Government will continue to look for opportunities to roll out the recommended loneliness indicators in other relevant surveys, research and evaluations of programmes to reduce loneliness.

Please use these measures and share your findings widely to help us build the evidence base for loneliness. The recommended questions on loneliness have been accepted as interim harmonised principles for use across the Government Statistical Service and Office for National Statistics (ONS) would welcome your feedback on how well they work and any improvements you would suggest before submitting them for final approval as harmonised principles in 2020.

The What Works Centre for Wellbeing would also welcome your feedback on the findings from your research to include in their ongoing review of what works to prevent or alleviate loneliness.

Please get in touch with us by contacting:

• QualityofLife@ons.gov.uk

• Harmonisation@Statistics.gov.uk

• info@whatworkswellbeing.org