Digital interventions to reduce social isolation and loneliness in older adults: An evidence and gap map

The review examined any digital intervention aimed at reducing social isolation or loneliness in adults aged 60 and older. Digital interventions included:

**Video calling** (e.g., Skype, FaceTime, Zoom calls with family or peers)

**Social media platforms** (e.g., Facebook, Twitter, online forums)

**Online group activities** (e.g., virtual book clubs, exercise classes, discussion groups)

**Skill-building programs** (e.g., internet training, tablet loan programs, digital literacy courses)

**Combined interventions** (e.g., tablet provision plus training plus ongoing video calls)

**Comparators** varied across studies:

No intervention (waitlist control)

Usual activities (no change to routine)

In-person versions of the same activity

Non-digital alternatives (e.g., telephone calls, printed materials)

**Outcome measures** were:

**Loneliness** (subjective feeling of being alone or disconnected)

**Social isolation** (objective measure of contact frequency, network size, or participation)

**Social support** (perceived availability of help or companionship)

**Quality of life** (general well-being)

**Depression** (mood symptoms)

**Anxiety** (worry or nervousness)

**Cognitive function** (memory, attention, processing speed)

**Physical function** (mobility, daily activities)

**Adverse events** (any harm from the intervention)

The review included 71 studies with a total of approximately 10,000 participants. Specific demographics varied widely:

**Age range:** 60 to 98 years old (mean age across studies ranged from 65 to 85)

**Gender:** Predominantly female (60–80% across most studies)

**Setting:** Community-dwelling older adults (living at home, not in nursing homes or hospitals), plus some studies in assisted living facilities and retirement communities

**Health status:** Mixed — some studies included healthy older adults, others included those with chronic conditions (e.g., heart disease, diabetes, arthritis), cognitive impairment (mild to moderate), or mobility limitations

**Technology experience:** Ranged from complete beginners (never used a computer) to regular internet users

**Geographic distribution:** Studies from 20 countries, mostly high-income (USA, UK, Canada, Australia, Netherlands, Japan, South Korea), with a few from middle-income countries (China, Brazil, Thailand)

**Sample sizes per study:** Ranged from 10 to 1,200 participants (median around 60–80)

Studies used a variety of validated and non-validated instruments. The most common were:

**Loneliness:**

**UCLA Loneliness Scale (Version 3):** 20 items, scored 20–80, higher = more lonely. Used in ~40% of studies.

**De Jong Gierveld Loneliness Scale:** 11 items, scored 0–11, higher = more lonely. Used in ~20% of studies.

**Single-item loneliness question:** "How often do you feel lonely?" (never, sometimes, often). Used in ~15% of studies — less reliable.

**Social isolation:**

**Lubben Social Network Scale (LSNS-6):** 6 items, scored 0–30, lower = more isolated. Used in ~15% of studies.

**Social Network Index (SNI):** Counts number of regular contacts across 12 relationship types.

**Self-reported contact frequency:** "How many times per week do you see/talk to friends/family?"

**Other outcomes:**

**Quality of life:** SF-36, WHOQOL-BREF, EQ-5D

**Depression:** Geriatric Depression Scale (GDS-15 or GDS-30), PHQ-9

**Anxiety:** GAD-7, State-Trait Anxiety Inventory

**Cognitive function:** Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA)

**Physical function:** Barthel Index, Timed Up and Go test

**Measurement timing:** Most studies measured outcomes at baseline and immediately post-intervention (4–12 weeks). Only ~20% of studies included follow-up beyond 3 months. Very few measured beyond 6 months.

### Study design

This is a **systematic review with an evidence and gap map** — a structured synthesis of existing research that identifies what has been studied and where evidence is missing. The authors searched 12 electronic databases (including MEDLINE, PsycINFO, CINAHL, Cochrane CENTRAL, and Web of Science) for studies published between January 2000 and December 2022.

### Inclusion criteria

**Population:** Adults aged 60+ (or mean age ≥60)

**Intervention:** Any digital technology designed to reduce social isolation or loneliness

**Comparator:** Any (no intervention, usual care, alternative intervention, or pre-post comparison)

**Outcomes:** Social isolation, loneliness, social support, quality of life, depression, anxiety, cognitive function, physical function, or adverse events

**Study design:** Randomized controlled trials (RCTs), quasi-experimental studies, pre-post studies, cohort studies, and qualitative studies

### What they found in terms of study quality

Of the 71 included studies:

**30 were RCTs** (randomized controlled trials) — the strongest design for proving causation

**22 were quasi-experimental** (non-randomized comparison groups)

**19 were pre-post studies** (no comparison group — weakest design)

Among the RCTs, the authors assessed risk of bias using the Cochrane Risk of Bias tool:

**Only 5 of 30 RCTs** had low risk of bias across all domains

**18 of 30 RCTs** had high or unclear risk of bias due to:

- Lack of blinding (participants and staff knew who got the intervention)

- Small sample sizes (many had <50 participants total)

- High dropout rates (20–40% in some studies)

- Self-reported outcomes (participants knew they were in the intervention group)

- No pre-registration of study protocols

### What this design can and cannot prove

**What it can prove:**

The review can identify **patterns across studies** — e.g., which types of digital interventions have been tested, in which populations, and with what general direction of results

It can highlight **gaps in the evidence** — e.g., no studies on certain populations (rural, low-income, non-English speaking) or certain outcomes (long-term effects, adverse events)

It can provide a **comprehensive map** of the existing literature

**What it cannot prove:**

It **cannot** give a precise, reliable estimate of effect size because the included studies are too heterogeneous (different interventions, populations, measures, and durations)

It **cannot** determine causality — many studies lacked proper control groups or randomization

It **cannot** rule out placebo effects — most interventions were unblinded, so participants who received the intervention may have reported feeling less lonely simply because they expected to

It **cannot** tell you which specific intervention works best for which person — the evidence is too sparse and inconsistent

### Major methodological weaknesses flagged by the authors

1. **Short follow-up periods:** Most studies lasted 4–12 weeks. We don't know if effects persist beyond 3 months.

2. **Small sample sizes:** Many studies had <50 participants, making it hard to detect small-to-moderate effects.

3. **High risk of bias:** Lack of blinding, no randomization, high dropout rates.

4. **Heterogeneous outcome measures:** Studies used different scales for loneliness, making direct comparison difficult.

5. **Publication bias:** Studies with null or negative results are less likely to be published, so the review may overestimate positive effects.

6. **Limited diversity:** Most participants were white, female, urban-dwelling, and relatively well-educated. Findings may not generalize to men, rural populations, or those with limited digital literacy.

### Primary outcomes (loneliness and social isolation)

**Loneliness (30 studies measured this):**

**12 of 30 studies** reported a statistically significant reduction in loneliness in the digital intervention group compared to control

**Effect sizes ranged from small to moderate** (Cohen's d = 0.20 to 0.55) — meaning the intervention reduced loneliness by about 0.2 to 0.6 standard deviations

**No meta-analysis was possible** due to heterogeneity, so the authors could not calculate a pooled effect size

**The largest effects** were seen in studies that combined multiple components (e.g., tablet + training + video calls + online group activities)

**The smallest effects** were seen in studies that only provided access to social media or online forums without any training or support

**Social isolation (22 studies measured this):**

**8 of 22 studies** reported a statistically significant reduction in social isolation

**Effect sizes ranged from small to moderate** (Cohen's d = 0.15 to 0.45)

**Studies that measured objective social contact** (e.g., number of calls, visits, or interactions) showed smaller effects than those that measured perceived social support

### Secondary outcomes

**Social support (18 studies):**

**7 of 18 studies** reported a statistically significant increase in perceived social support

Effect sizes: Cohen's d = 0.20 to 0.50

**Quality of life (15 studies):**

**5 of 15 studies** reported a statistically significant improvement

Effect sizes: Cohen's d = 0.15 to 0.40

**Depression (12 studies):**

**4 of 12 studies** reported a statistically significant reduction in depressive symptoms

Effect sizes: Cohen's d = 0.20 to 0.45

**Anxiety (8 studies):**

**2 of 8 studies** reported a statistically significant reduction

Effect sizes: Cohen's d = 0.15 to 0.30

**Cognitive function (6 studies):**

**2 of 6 studies** reported a statistically significant improvement

Effect sizes: Cohen's d = 0.20 to 0.35

**Physical function (4 studies):**

**1 of 4 studies** reported a statistically significant improvement

Effect size: Cohen's d = 0.25

**Adverse events:**

**Only 3 of 71 studies** reported on adverse events

No serious adverse events were reported, but this is likely due to lack of monitoring rather than true absence of harm

### Subgroup analyses (where reported)

**Age:** No clear difference between young-old (60–75) and old-old (75+)

**Gender:** Studies with higher proportions of women showed slightly larger effects, but this may reflect women's greater willingness to engage with digital interventions

**Baseline loneliness:** Studies that recruited lonely participants showed larger effects than those that recruited general populations

**Intervention type:** Multi-component interventions (training + equipment + ongoing support) showed larger effects than single-component interventions (e.g., just providing a tablet or just offering a Facebook group)

**Duration:** Interventions lasting 8–12 weeks showed larger effects than those lasting <4 weeks or >12 weeks (though the latter may reflect dropout)

Because the review could not perform a meta-analysis, precise effect sizes are not available. However, based on the range of reported effects:

**Loneliness reduction:** On the UCLA Loneliness Scale (20–80), a Cohen's d of 0.35 corresponds to roughly a 3–4 point reduction. This is equivalent to moving from "sometimes lonely" to "rarely lonely" on several items. For context, a 3–4 point reduction is about half the effect of a structured in-person social group intervention.

**Social isolation reduction:** On the Lubben Social Network Scale (0–30), a Cohen's d of 0.30 corresponds to roughly a 2–3 point increase. This means gaining about 1–2 additional social contacts or feeling more connected to existing contacts.

**Depression reduction:** On the Geriatric Depression Scale (0–15), a Cohen's d of 0.35 corresponds to roughly a 1–2 point reduction. This is a small effect — about one-third the effect of antidepressant medication in clinical trials.

**Quality of life improvement:** On the SF-36 (0–100), a Cohen's d of 0.30 corresponds to roughly a 5–8 point improvement in the social functioning subscale. This is noticeable but not transformative.

**In plain English:** If you're an older adult feeling moderately lonely, a well-designed digital intervention (training + equipment + regular video calls with peers) might reduce your loneliness by about 10–20% — enough to feel a meaningful difference, but not enough to eliminate loneliness entirely. The effect is roughly comparable to joining a weekly in-person social group, but with lower logistical barriers.

### What the authors acknowledge

1. **High heterogeneity** across studies prevented meta-analysis and precise effect size estimation

2. **Short follow-up** (most studies <12 weeks) means long-term effects are unknown

3. **Poor reporting** of adverse events and dropout reasons

4. **Limited diversity** in study populations (mostly white, female, urban, well-educated)

5. **Publication bias** likely overestimates positive effects

6. **No cost-effectiveness data** — we don't know if these interventions are worth the investment

### What a critical reader would add

7. **Lack of blinding is a major problem.** Most participants knew they were receiving a digital intervention, so improvements in self-reported loneliness could be due to placebo effects, social desirability bias (wanting to please the researchers), or simply the attention of being in a study (Hawthorne effect).

8. **No active control groups.** Most studies compared digital interventions to "no intervention" or "usual activities." This means we can't tell if the digital component itself is helpful, or if any structured social activity (even a phone call or a weekly visit) would produce the same effect.

9. **Technology access confounds results.** Participants who agreed to join a digital intervention study were likely already interested in technology and had some baseline digital literacy. Results may not generalize to those who are tech-averse or lack internet access.

10. **Small sample sizes** mean many studies were underpowered to detect small but meaningful effects. The fact that only 40% of studies found significant results could mean the true effect is small, or that most studies simply couldn't detect it.

11. **No standardized intervention.** Each study used a different combination of hardware, software, training, and support. It's impossible to know which specific components are essential.

12. **The review is from 2023** and includes studies up to 2022. Given the rapid pace of digital technology change, some interventions (e.g., specific apps or platforms) may already be outdated.

For someone running their own n=1 experiment (testing a digital intervention to reduce your own loneliness or social isolation):

### What to test

**Best bet:** A multi-component intervention combining:

- **Equipment:** A tablet or computer with a camera and internet connection (if you don't already have one)

- **Training:** 2–4 sessions of basic digital literacy (how to use video calling, email, social media, or online groups)

- **Regular video calls:** At least 2–3 scheduled video calls per week with family, friends, or a peer group (e.g., a virtual book club, exercise class, or discussion group)

- **Ongoing support:** Access to tech support or a "buddy" who can help troubleshoot

**Alternative to test:** A structured online group program (e.g., a 6-week virtual workshop on a topic you enjoy — gardening, history, music, etc.) with weekly group video calls and a private online forum for between-session chat

**Dose:** Aim for at