How we measure loneliness
The scales we administer, how Mary asks the questions, and how scores combine into the composite index used across care team views.
What we measure
Two validated scales, used together, not interchangeably. Each was designed for a different question.
UCLA-3
3 items, score 3 to 9, asked every call
Question it answers: how lonely does this person feel right now, on a scale that catches change between calls?
The three items:
- How often do you feel that you lack companionship?
- How often do you feel left out?
- How often do you feel isolated from others?
Scoring: each item 1 (hardly ever), 2 (some of the time), 3 (often). Total 3 to 9. Threshold of clinical concern: 6 or higher.
Strength: short, sensitive to change, validated for repeated measurement.
Limit: unidimensional. It tells you "how much" but not "what kind".
DJG-6 (De Jong Gierveld)
6 items, 2 sub-scales of 3, asked monthly
Question it answers: is this person missing an intimate attachment (emotional loneliness), a broader network (social loneliness), or both? Different interventions follow.
Emotional sub-scale (3 items):
- I experience a general sense of emptiness
- I miss having people around
- I often feel rejected
Social sub-scale (3 items, reverse-scored):
- There are many people I can trust completely
- There are enough people I feel close to
- There are plenty of people I can rely on if I have problems
Scoring: dichotomised 0 or 1 per item. Emotional 0 to 3, social 0 to 3. Threshold of concern on either sub-scale: 2 or higher.
Strength: tells you what kind of loneliness, which determines what to do about it.
Why both, not one?
UCLA-3 catches change between calls. DJG-6 distinguishes emotional loneliness (missing one specific person, e.g. a spouse who died) from social loneliness (missing a network, e.g. moved away from community). The clinical response is different in each case. Restoring one close connection helps emotional loneliness; activity and group settings help social loneliness. Without the split, every lonely person gets the same intervention plan and roughly half of them do not benefit.
This is also why population reports show two axes. A facility may have residents who score similarly on UCLA-3 but split cleanly into emotional and social, and they need different programming.
How Mary administers the items
Mary administers the scale items in conversation, verbatim, in her own warm voice. She does not summarise items, does not paraphrase, and does not infer answers from general conversation. A score is only recorded when the item was actually asked and the person actually answered.
After the call, the transcript is reviewed by an extraction step that requires a verbatim quote from the resident for each scored response. If a quote is missing, the item is marked as not administered rather than guessed.
Cadence is automated:
- UCLA-3: administered every call where it makes sense in conversation. Daily-to-weekly cadence in practice.
- DJG-6: administered once a month per person. The system tracks the next-due date and prompts Mary on the call closest to it.
Mary never prescribes after a loneliness response. She does not tell the person to "join a club" or "call your family". The scores travel to the care team, who decide on the response with the person.
The composite index (v1)
Each scale stands on its own and has its own report. The composite combines them into a single 0 to 100 number used for ranking and triage, never for diagnosis.
How it is calculated
| Component | Weight | Why |
|---|---|---|
| Acute (latest UCLA-3) | 30% | How the person feels right now, on the most recent administration. |
| Chronicity (last 12 admins) | 25% | Persistence matters more than any single bad day. Share of recent administrations above the clinical threshold. |
| Trajectory (4-week slope) | 20% | Direction of change. Worsening trajectories warrant a different response from stable ones. |
| DJG-6 emotional dimension | 15% | Absence of an intimate attachment, e.g. a spouse who has died, an estranged child. Different intervention pathway. |
| DJG-6 social dimension | 10% | Absence of a broader network, e.g. moved away from community. Different intervention pathway. |
A prosody-flatness adjustment can move the composite up or down by no more than 10 points, applied only when voice expressiveness sits in the top or bottom decile across the call. This is a cross-check on self-report, not a separate diagnosis.
Weights are configurable per organisation. The default weights shown here are what ship out of the box. Any change is recorded in the version log below.
Severity bands
Dimension classifier
When DJG-6 is recent, the composite labels a primary dimension:
- Emotional: emotional sub-scale at or above threshold, social below. Pathway: restore one close connection.
- Social: social sub-scale at or above threshold, emotional below. Pathway: increase network breadth, activity programmes.
- Mixed: both sub-scales at or above threshold. Both pathways together.
- Unclear: DJG-6 not yet administered or out of date. The next monthly DJG-6 will disambiguate.
What family see vs what care staff see
Family-side view
Direction only. Family see whether their loved one's connection feels steady, stretched, or improving, with conversation prompts. No raw scores, no item-level answers, no thresholds.
This is intentional. Loneliness scores are a clinical signal. Family receive enough to know when to lean in; the underlying numbers stay with the care team.
Care team view
Full numbers. Each scale's trajectory, threshold crosses, dimension classifier, composite with sub-scores, recommended intervention pathway.
Sortable cohort ranking. Per-call detail showing exactly which items were asked and how the person answered, in their own words.
References
Every choice on this page is grounded in published work. Follow the links to the source.
Version log
The composite formula is stamped on every read with a version. When weights or components change, the change is recorded here and the prior version stays available for audit.