This appendix documents the methods, assumptions and evidence sources used in the LonelyLessDCE decision aid tool. It explains how DCE results are translated into predicted uptake and willingness to pay, how cost components and health outcomes are parameterised for the Australian context and how benefit metrics are derived.
The main manuscript describes the development of the DCE, the survey sample and model results for older adults in Australia. This technical appendix should be read alongside the main manuscript and the online appendix.
The DCE elicited preferences for key features of loneliness support programmes for older adults. Attributes included the type of support programme, delivery mode, frequency of sessions, duration of sessions, distance/accessibility and out of pocket cost per session. Attribute levels were selected through a review of the literature on loneliness interventions and consultation with experts, in line with recommended good practice for health related DCEs.
Respondents aged 65 years and older completed a series of paired choice tasks in which they chose between two hypothetical programmes and an opt out option. The experimental design used a blocked D-efficient design, with respondents randomly allocated to blocks.
The primary DCE analysis estimated an error component logit model with a random alternative specific constant (ASC) for the programme options to capture panel correlation and unobserved preference heterogeneity, while treating non cost attributes as fixed. The utility for individual n choosing alternative j in choice situation t is written as
Unjt = Vnjt + εnjt, where the observed component of utility for the programme option is Vprogram = ASCprogram + βX − βcCost and for the opt out alternative Voptout = ASCoptout. Here X is a vector of coded attribute levels and Cost is the out of pocket cost per session. The cost coefficient βc is treated as continuous.
For any given configuration of attribute levels, the tool constructs the deterministic utility index for the programme alternative by inserting the coded attributes and cost into the model coefficients. Predicted uptake is then obtained using the standard logit formula for the choice between joining the programme and opting out. The probability of joining the programme is
P(programme) = exp(Vprogram) / [exp(Vprogram) + exp(Voptout)].
This probability is interpreted as the expected uptake proportion among older adults, holding constant the distribution of preferences implicit in the DCE sample. In the decision aid, a base cohort size of 250 eligible older adults is assumed. The expected number of participants is calculated as this cohort size multiplied by the predicted uptake probability. Users can rescale the results to other cohort sizes through linear extrapolation.
Willingness to pay (WTP) for each attribute level is obtained from the DCE model by dividing the attribute coefficient by the absolute value of the cost coefficient. For attribute k with coefficient βk and cost coefficient βc, the implied marginal WTP is WTPk = βk / |βc|. This yields a monetary value in Australian dollars per session that represents the additional fee older adults would be willing to pay for a change from the reference level of that attribute, holding other attributes constant.
The WTP outcomes tab in the tool presents point estimates of WTP per participant per session for each attribute level. These values are obtained from the error component logit model used for the tool and are rounded for display. Methods for deriving WTP from DCEs and interpreting the resulting trade offs are consistent with established guidance on stated preference methods in health and patient centred outcomes research.
For any specific programme configuration, the tool calculates the per participant WTP per session by summing the attribute level WTP values corresponding to the selected levels. For example, if the configuration involves a group based community engagement programme delivered in person at a local venue, meeting weekly for 2 hours at a given cost per sesion, the WTP contribution for each of these design features is summed.
The per participant WTP for the complete programme is then obtained by multiplying the per session WTP by the number of sessions. The current implementation assumes 12 sessions by default, which aligns with the group based educational friendship enrichment intervention used in Australian modelling of loneliness interventions for older women. Total WTP based benefits for the scenario are calculated by multiplying the programme level WTP by the expected number of participants predicted by the uptake model.
The WTP based benefit cost ratio is defined as the ratio of total WTP benefits to total economic costs. Values greater than one indicate that, on average, the stated value that older adults place on the configured programme features exceeds the estimated economic cost of delivering the programme to the target group.
Cost components and unit values in the decision aid are based on Australian costing work for loneliness interventions, particularly the modelling of the Friendship Enrichment Programme and volunteer led internet and computer training interventions for older women. These interventions were evaluated using a Markov model that reported programme costs, healthcare savings, productivity gains and return on investment over three and five year horizons. The National Mental Health Commission also developed a technical brief and workbook that set out detailed assumptions for intervention delivery and resource use. These sources provide the conceptual template for the LonelyLessDCE cost structure, which is adapted to a cohort of 250 older adults.
Table 1 summarises the main cost components in the tool, together with the unit cost, quantity assumptions and principal sources or justification. All costs are expressed in 2024 Australian dollars. Where necessary, costs were updated from published values using Australian consumer price index data and average weekly earnings to approximate wage growth for relevant staff categories.
| Cost component | Description | Unit cost (AUD) | Unit | Source or justification |
|---|---|---|---|---|
| Local advertising | Paid advertisement in local newspaper or community newsletter to promote the programme | 1,500 | Per campaign | Adapted from National Mental Health Commission modelling of educational friendship programmes, updated to 2024 prices |
| Leaflet printing | Design and printing of programme information leaflets | 0.50 | Per leaflet | Unit cost for colour printing and basic design, consistent with patient information material costings in Australian preventive health programmes |
| Postage | Mailing of leaflets and invitations to eligible older adults | 1.20 | Per postal item | Australia Post standard letter rate in 2024 plus handling allowance |
| Administrative staff time | Time spent on recruitment, scheduling, record keeping and general coordination | 45 | Per hour | Approximate hourly wage for administrative officers derived from Australian Bureau of Statistics average weekly earnings |
| Facilitator or trainer time | Delivery of group sessions by trained facilitator or allied health professional | 80 | Per hour | Aligned with hourly costs for community based mental health or social work professionals in Australian cost effectiveness studies |
| Session materials | Printed materials, workbooks and refreshments per participant per session | 10 | Per participant per session | Based on National Mental Health Commission assumptions for group education programmes, inflated to 2024 values |
| Venue hire | Hire of community venue for group sessions | 90 | Per session | Typical hourly hire for community centres and meeting rooms in Australian capital cities |
| Participant travel cost | Travel expenses incurred by participants to attend sessions | 8 | Per return trip | Approximate fuel and public transport costs over a short local trip, consistent with Australian travel cost assumptions in health economic evaluations |
| Participant time cost | Opportunity cost of time spent in sessions and travel for older adults | 25 | Per hour | Set at roughly half of average full time hourly earnings from Australian Bureau of Statistics average weekly earnings, reflecting lower opportunity cost in retirement |
The decision aid separates fixed costs that do not vary with uptake from variable costs that depend on the number of participants and sessions. Fixed costs include local advertising, programme setup, development of materials and a portion of administrative time. Variable costs include venue hire, facilitator time, session materials and participant travel. When the user indicates that the opportunity cost of participant time should be included, time spent in sessions and travel is monetised using the hourly rate in Table 1 and added to variable costs.
For a chosen configuration, total economic cost is calculated as the sum of fixed costs plus the number of sessions multiplied by the per session variable cost per participant and the predicted number of participants. When users change the number of sessions, delivery mode or co payment, these changes propagate automatically through the costing engine. The costs and benefits tab of the tool displays total economic cost, cost per participant and cost per loneliness free day alongside the benefits described in later sections.
The tool allows the user to apply optional state or territory specific multipliers to reflect differences in cost of living and service delivery costs across Australia. These multipliers are applied to both programme costs and participant fees. The default values are illustrative and were calibrated using publicly available data on relative wage levels and consumer prices across jurisdictions. Users with access to more precise local cost information can adjust these multipliers to better reflect their context and can then save configurations for later use.
The link between loneliness interventions and health outcomes in the decision aid is based on modelled cost effectiveness work evaluating the impact of group based loneliness interventions on depression and associated health states among older adults in Australia. In this work, reducing loneliness through preventive programmes led to gains in quality adjusted life years by lowering the incidence of depression and other adverse health outcomes over time. The analysis reported incremental cost effectiveness ratios for the Friendship Enrichment Programme and a volunteer led internet and computer training intervention, as well as estimates of cost savings and return on investment over five years.
To translate these findings into a flexible decision aid, the tool allows users to select low, moderate and high QALY gain scenarios per participant. These values are calibrated so that the implied cost per QALY gained and cost per loneliness free year fall within the range reported in the published studies. The default monetary value per QALY is set at AUD 50,000, which is consistent with values commonly used in Australian cost effectiveness analyses and health technology assessment guidelines.
Total QALY based benefits for a configuration are calculated as the product of the QALY gain per participant, the predicted number of participants and the value per QALY. The QALY based benefit cost ratio is obtained by dividing total QALY benefits by total economic costs. Both the benefits and the ratio are reported in the costs and benefits tab and are used to construct example scenarios discussed in the main manuscript and online appendix.
Loneliness is associated with greater use of primary care and hospital services, higher demand for mental health care and increased workplace absenteeism and presenteeism among working age older adults. Australian analyses of the economic costs of loneliness have shown that the total cost across the adult population is substantial with older adults accounting for a sizable share of healthcare and productivity costs. These findings are consistent with international reviews that highlight loneliness and social isolation as important determinants of health care use and economic outcomes.
The LonelyLessDCE tool parameterises cost savings and productivity gains using per participant values derived from the National Mental Health Commission modelling of friendship enrichment and internet training interventions for older women. In that work, national rollout of the programmes was estimated to generate healthcare savings and productivity gains that exceeded the intervention costs, with a return of around 2-3 dollars in savings for each dollar invested over five years for some scenarios. By dividing the reported aggregate healthcare savings and productivity gains by the number of programme participants in that modelling, the decision aid derives per participant savings over five years and scales these to the predicted number of participants in each configured scenario.
Total cost savings and productivity gains for a given configuration are therefore calculated as the per participant savings multiplied by the expected number of participants. These totals are displayed in the costs and benefits tab and are used to calculate a cost saving based benefit cost ratio, defined as the ratio of five year savings to the total economic cost of delivering the configured programme.
In addition to monetary outcomes, the National Mental Health Commission modelling reported loneliness free days as a primary outcome of the friendship enrichment programme. Over three and five year horizons, the interventions were estimated to generate several million additional days in which participants were not lonely, relative to a comparator of no intervention. These loneliness free days can be interpreted as a patient relevant benefit that is closely aligned with the goals of the intervention.
The decision aid converts the published loneliness free day totals into a per participant value over five years by dividing by the number of participants in the modelled cohort. This per participant figure is then multiplied by the predicted number of participants for the configured programme, assuming proportional scaling. The result is an estimate of total loneliness free days attributable to the programme over five years. This metric is reported alongside QALY gains and monetary outcomes to give users a direct sense of improvements in loneliness over time.
For each scenario, the decision aid reports several benefit cost ratios and associated net benefit measures. First, a QALY based benefit cost ratio is defined as monetised QALY benefits divided by total economic costs, with the corresponding net benefit equal to QALY benefits minus costs. Second, a cost savings based benefit cost ratio is calculated as the ratio of total health system and productivity savings over five years to total economic costs, with net benefit given by total savings minus costs. Third, a WTP based benefit cost ratio is obtained by dividing total WTP benefits by total economic costs, with net benefit equal to WTP benefits minus costs. These three summary indicators offer complementary perspectives on value: health outcome gains, fiscal and productivity savings and user centred valuation.
In the Commission modelling that underpins the calibration of the tool, the ratio of total cost savings and productivity gains to programme costs corresponds to a conventional return on investment measure. In that analysis, the base case return on investment over five years was reported to be greater than two for some intervention scenarios, meaning that each dollar spent generated more than two dollars in savings. The LonelyLessDCE tool adopts the same interpretation when reporting cost savings based benefit cost ratios and uses parameter values that preserve this broad magnitude for illustrative scenarios.
Users should bear in mind that the underlying estimates are based on specific interventions, populations and time horizons. When applying results to new settings, it is important to consider differences in health system structures. The main manuscript highlights this limitation and cautions against treating the benefit cost ratios as definitive decision rules. The tool is intended to support scenario analysis.
The decision aid relies on several simplifying assumptions that are important to acknowledge. The DCE model represents average preferences across older adults in the sample and does not explicitly model preference heterogeneity by subgroups such as sex, cultural background, income or region, although the random ASC captures some unobserved heterogeneity. Cost parameters and savings estimates are drawn from specific interventions and time periods and may not fully reflect current wage rates, service configurations or policy priorities in all jurisdictions. QALY gains and loneliness free days are represented by scenario values that are calibrated to the published modelling rather than by intervention specific estimates for every possible programme design.
As a result, the tool is best used as a structured scenario generator that supports comparative analysis, stakeholder engagement and clear communication of trade offs. It should not be used as a stand alone decision rule. Formal funding decisions should be supported by context specific economic evaluations that use up to date local data and explore uncertainty more fully, including sensitivity analyses around key parameters.
Additional background literature on loneliness interventions, healthcare utilisation and economic evaluation is found in the National Mental Health Commission technical materials and in the journal articles listed above.