The aim of this study was threefold: First, to conduct a socioeconomic analysis of Individual Placement and Support (IPS) integrated with specialized substance use disorder treatment. Second, to explore differences in the use and related costs of employment support services from the Norwegian Labour and Welfare Administration (NAV). Third, to assess the validity of a prior simulation study by updating the model with data from the IPS-SUD trial.

The IPS-SUD trial was a pragmatic randomized controlled trial implemented at Oslo University Hospital (OUH), Norway [16]. The participants (n = 187) were adults currently in specialized SUD treatment at OUH. They had a median age of 37 (IQR: 30–45), 22% were women, and the majority had been unemployed for 2 years or more (55%) [14].

The participants were recruited from five outpatient units and two residential units at OUH in the period from March 1st, 2020, to May 31st, 2022. Participants were randomized to receive either IPS or enhanced TAU. The IPS group was offered a maximum of 13 months of integrated employment support alongside their SUD treatment. The enhanced TAU group was offered a self-help guidebook, a three-session workshop in groups of no more than 13 participants, and an individual follow-up session with a group leader, alongside their SUD treatment. Because the IPS-SUD trial did not collect data for a TAU-only group, the current analysis of costs and benefits includes estimates only for the IPS and enhanced TAU groups, thus leaving out data on the TAU-only group, which was included in the previously published pre-trial socioeconomic simulation [17].

The former simulation and the present study use Early Health Technology Assessment (eHTA) to assess the costs and socioeconomic effects of the intervention. eHTA is defined as a health technology assessment conducted to inform decisions about subsequent development, research, and/or investment by explicitly evaluating the potential value of a conceptual or actual health technology [19].

Since the cost of the IPS intervention is delivered by the hospital but its outcomes and savings impact on budgets outside the hospital, we chose to conduct a broad socioeconomic analysis, in contrast to a traditional health-economic analysis. Our economic model estimates effects in terms of direct costs of the intervention for the hospital, as well as its direct and indirect costs and effects for society. All costs and potential savings were reported in Norwegian Kroner (NOK) and converted to Euros using the exchange rate from January 29, 2025 (1 NOK = €0.0845). For modelling purposes, we scaled up each group to represent data for 100 participants over a 10-year time horizon.

We estimated direct costs associated with the intervention from the payer’s perspective (i.e., the healthcare services), which included hospital charges and professional fees. The time horizon for the cost of the intervention was set to one treatment cycle of approximately 1 year; thus, the discount rate was 0%. IPS costs used in our primary analysis were informed by the total cost of having four full-time employment specialists delivering the intervention, in total €684,450. Costs of enhanced TAU were calculated for one person working 20% of full-time hours, based on full-time salary for one employment specialist position (€57,037).

Additionally, we included two sensitivity analyses. In the first we calculated costs of the intervention based on the number of hours of IPS delivery instead of using hospital charges and professional fees. We constructed two different scenarios: one using costs based on IPS intensity data extracted from the trial (10.4 h average IPS consultations per participant) and one using costs based on the assumptions of IPS intensity from the pre-trial simulation (87.5 h average IPS consultations per participant), multiplied by standard hourly rates sourced from SSB (€37). The sensitivity analyses also included a similar way to calculate enhanced TAU cost, which included costs of course coordination (1.5 h of work multiplied by an hourly rate of €30, divided by 13 participants) and course instruction (6 h multiplied by an hourly rate of €38, divided by 13 participants, including an additional hour of individual session per participant). All cost estimations were calculated for 100 participants in each group, giving a total of 1,040 h of IPS in our scenario based on trial data (IPS low intensive) and 8,750 h in the scenario based on pre-trial assumptions (IPS high intensive). The total patient contact for the sensitivity analysis of enhanced TAU (ETAU 2) was 158 h.

In our second sensitivity analysis we accounted for sample error from the RCT using probabilistic sensitivity analysis (PSA) [20]. We ran 2000 non-parametric bootstrap simulations with replacement for our input variables employment rate, working capacity and income for both the IPS and ETAU groups. The distribution of the simulated employment data can be found in Supplementary Material (Supplementary Table S2). All data related to costs and disability pension were identical to that of the main analysis and were set as fixed, due to them being based on external data only available as point estimates. Our socioeconomic analysis was then run on our simulated data distributions, which resulted in mean socioeconomic estimates including uncertainty in the form of 95% confidence intervals (95% CI).

Results from our PSA are shown in Supplementary Material (Supplementary Tables S1–S3).

As the intervention was financed through public budgets, we included a tax financing cost of 20 cents per euro (20%) in the model. This reflects the marginal cost of collecting an extra euro in taxes, as this may lead to inefficiencies. The estimated measure of 20% is based on the standard assumption in Norwegian public finance that a value creation on the private side of 20 cents is needed in order to finance one euro in a public project [21]. For a complete overview of resources and costs per participant, see Table 1; Supplementary Table S1.

We included indirect costs of production loss from disease and disability, which are estimated in terms of the intervention’s effect on disability pension and absence from the labor market. We assumed that patients who permanently exited the labor market after the intervention were expected to receive disability pension. The 2019 average annual disability pension of €21,379 per recipient [22], was used to show yearly effects of SUD patients excluded from the labor market. For patients who obtained employment, disability pension was reduced according to their working capacity. As disability pension is considered a social transfer payment, we only included the associated tax financing cost of the measure (€0.20 per €1.00) in our model.

From the vocational data in the IPS-SUD trial [14] we used the employment rate for the two groups as the direct effect in our model. This input data was restricted to participants who achieved at least 3 months of paid employment, a threshold we considered a reasonable indicator of long-term employment and appropriate for the 10-year time horizon of our model. These employment rates were 29% among those who received IPS and 26% for those receiving enhanced TAU.

The average salary among those who achieved at least 3 months of paid employment was €22,841 in the IPS group and €10,762 in the enhanced TAU group. The salary included employer’s contribution and social costs, used for value creation per year for a person who entered the labor market after the intervention. We assumed an income tax of 25% and an employer’s contribution of 14.1% in our calculations of increased tax revenue [23].

These future benefits were discounted at a rate of 4%, which is in line with recommendations from the Norwegian Ministry of Finance on public investment projects [24].

The second objective for the present study was to evaluate the impact of IPS on the participants' use of employment support services administered by the Norwegian Labour and Welfare Administration (NAV). In this separate analysis, we extracted the self-report data for those participants in our trial who responded at 6 months follow-up (n = 123) and/or 12 months follow-up (n = 122). Some participants received several services. Different employment support services were categorized as: employment preparation training, employment training, IPS, work capacity assessment, and “other.” Twice as many in the enhanced TAU group reported using an employment support service from NAV, compared to the IPS group (18/58 vs. 9/64). The difference in proportion of employment support service use in the enhanced TAU group and the IPS group was multiplied by the cost associated with each service and summarized as a cost-difference in Table 3. The cost associated with the different employment support services was based on a report prepared for the Ministry of Labour and Social Inclusion for the first three services, and the fourth service, work capacity assessment, was based on expert opinion [25]. We did not manage to price a fifth category, “Other”.

A summary of socioeconomic effects per participant at 10 years is shown in Table 2, with the potential socioeconomic gain of implementing IPS for 100 SUD patients over 10 years illustrated in Figure 1. Given the assumptions in the model, scaled for 100 participants in each group, the socioeconomic value of IPS was estimated to exceed enhanced TAU during the first year (year 0–1) after end of the intervention (€992,224 for IPS, and €778,121 for enhanced TAU), equal to a difference of €214,103. The 5-year estimates were €4,219,207 for IPS and €2,235,751 for enhanced TAU. The 10-year estimates had increased to a gain of €7,602,247 for IPS and €3,763,872 for enhanced TAU, with a difference of €3,838,375.

In the sensitivity analysis based on hourly delivery (low vs. high intensive), the socioeconomic value estimates from both IPS scenarios were expected to surpass enhanced TAU (ETAU 2) during the intervention year (year −1 to 0) following the intervention start (Enhanced TAU = €415,101; IPS low intensive = €878,384; IPS high intensive = €536,058).

In the PSA, the re-estimated point estimates for the three employment variables were quite similar to those in the main analysis (Supplementary Table S2). The resulting socioeconomic analysis based on these numbers show similar trends as the main analysis. The socioeconomic value estimate of IPS surpasses ETAU during first year after intervention with a mean difference of 324,960 (Supplementary Figure S1). However, the PSA showed large uncertainties in this point estimate (95% CI: −462,357 to 1,124,683). In case of best scenario (upper bound of the 95% CI) for both groups, net gain of IPS occurs within the intervention year, while in a worst case scenario (lower bound of the 95% CI) for both groups, a net gain of IPS occurred within the second year after intervention end (year 1–2). However, in case of worst case performance of IPS compared to best case ETAU, IPS does not surpass ETAU (as illustrated by overlapping CIs).

The yearly socioeconomic gain, including mean differences between each group and a graph depicting them, can be found in Supplementary Material (Supplementary Table S3; Supplementary Figure S1).

Table 3 shows the proportion of employment support services used in the IPS group and the enhanced TAU group, the cost per participant for each service, and the cost difference between the two groups. In total, the IPS participants had less service use in four out of five employment support service categories, totaling a group cost saving of €157,593 relative to the participants receiving enhanced TAU.

In this updated eHTA, the estimated socioeconomic gain of IPS was found to exceed that of enhanced TAU during the first year after the intervention, and this change was observed 1 year later than in the pre-trial eHTA. Our sensitivity analysis of costs estimated IPS to exceed enhanced TAU already during the intervention year, which is similar to the pre-trial eHTA. Our sensitivity analysis using PSA gave the same conclusion as the main analysis, but showed that the mean difference estimate has large uncertainty (Supplementary Table S3).

Our results should be interpreted with some caveats. First, the extrapolation of study results over 10 years represents an early assessment of potential benefits; thus, the results are merely formative and should be reassessed when more long-term data is available. Second, we chose to extrapolate the employment input only from participants with employment exceeding 3 months, resulting in a small dataset on salary and FTEs in both groups. Thus, there is uncertainty in the amount and time horizon of the cost and potential savings estimates. Third, though our analysis of costs and benefits includes information from health and social welfare services, it lacks information from the criminal justice sector, and it does not include potential new treatment contacts or rehospitalizations from the health sector. It may be assumed that employment is associated with reduced crime and reduced risk of rehospitalization, meaning that if these factors were included, the estimates would potentially show an even larger gain from IPS. Fourth, the generalizability of our results may be most relevant to countries with comparable welfare systems. Lastly, our analysis is based on the effect of IPS during strict periods of social distancing due to the COVID-19 pandemic, which may have disproportionately impacted the IPS group and thereby reduced the socioeconomic impact of IPS and also the generalizability of our results.

Our updated eHTA model estimates that IPS, despite being modeled with a conservative high cost, achieves a greater net socioeconomic gain during the first year after the end of the intervention, compared to enhanced TAU. The conservative scenario of our earlier eHTA, which was based on registry data and expert opinions pre-RCT, provided valid estimates comparable with this updated eHTA based on post-RCT data. Compared with the enhanced TAU group, the IPS group showed cost savings due to less use of services from NAV. Our socioeconomic analysis based on costs and benefits from budgets across different sectors, can provide valuable information on the societal effect of IPS, supplementing estimates resulting from traditional health-economic analysis. From the perspective of decision-makers, it will be informative to consider the overall socioeconomic effect when assessing the organization and implementation of IPS, especially since IPS directly impacts beyond hospital budgets. Future research on the long-term duration of employment for this patient group would be an important next step to provide better data on estimating socioeconomic gain.