Cost-effectiveness analysis of combined cognitive and vocational rehabilitation in patients with mild-to-moderate TBI: results from a randomized controlled trial

This study is a cost-effectiveness analysis of a randomized, controlled, parallel-group trial that compared a combined cognitive and vocational intervention to multidisciplinary TAU on employment and other clinical outcomes after mild and moderate TBI. Details of the study design, recruitment, randomisation, blinding and data collection are provided in the study protocol [12]. Briefly, the trial was conducted at a specialized TBI outpatient clinic at Oslo university hospital (OUH), Norway, where patients were randomised to a group-based compensatory cognitive training intervention and individualized supported employment (CCT-SE) or individualized outpatient treatment provided by a multidisciplinary team (TAU) in a 1:1 ratio. Participants were recruited and started treatment 8–12 weeks after sustaining a mild or moderate TBI and received the intervention or control group treatment for a total of 6 months.

The trial was performed in accordance with the Declaration of Helsinki, and registered at ClinicalTrials.gov on 28.03.2017 (#NCT03092713). One-hundred and sixteen individuals aged 18–60 years were recruited to the RCT between July 2017 and April 2019. Two participants dropped out from the CCT-SE group and 1 from the TAU group after randomization. See Additional file 1 for flow chart of participant recruitment. Participants had sustained a mild or moderate TBI assessed by a Glasgow Coma Scale (GCS) score of 10–15, loss of consciousness for < 24 h and post-traumatic amnesia for < 7 days. Criteria from the American Congress of Rehabilitation Medicine were used to establish the presence of mild TBI [13]. All participants were employed in a minimum 50% position at the time of injury, but sick listed 50% or more due to post-concussive symptoms assessed by the Rivermead Post Concussion Symptoms Questionnaire [14] 2–3 months post injury. Table 1 provides participants’ demographic, injury- and work-related characteristics at baseline, in addition to EQ-5D-5L index values. There were no statistically significant between group differences on baseline characteristics, with the exception of EQ-5D-5L index values (p = 0.013). See Additional file 2 for proportion of responses by level of severity for EQ-5D-5L dimensions at baseline per treatment group.

CCT is a manualised intervention targeting post-concussive symptom management and cognitive symptoms [15, 16]. The intervention was provided by a clinical psychologist and a MD in groups of two to five participants for 2 h weekly over 10 weeks. Each session covered the topics through a combination of psychoeducation and compensatory strategy training. The vocational part of the intervention was based on elements from the Individual Placement and Support model of SE [12, 17]. Follow-up was customised to the individual participants’ needs and entailed mapping resources, limitations and work tasks, followed by guidance, advice and work task accommodations. The SE intervention was provided by three employment specialists for a total of 6 months. The CCT and SE interventions were provided in parallel and therapists providing the interventions collaborated closely to ensure implementation of strategies and compensatory techniques at the workplace.

TAU consisted of assessment and treatment provided by a multidisciplinary team (physiatrist, neuropsychologist, physiotherapist, occupational therapist and social worker) at a specialized TBI outpatient clinic at the Dept. Of Physical Medicine and Rehabilitation, OUH. Follow-up consisted of individual contacts and an education group focused on nonspecific education about TBI and discussion of common problems in daily life following TBI. The education group consisted of four weekly meetings of 2 h, each led by a different professional. TAU was provided for a maximum of 6 months.

The costs of the treatment provided in the intervention and control groups were calculated from each therapist’s gross annual salary, while also considering social costs (employers’ costs for mandatory and supplementary pension plans and insurances, factor of 1.4 in Norway). Costs of the CCT intervention and educational group provided in TAU were further adjusted for mean number of participants in each group (3 in CCT and 7 in educational group). The unit costs and frequency of treatment provided in CCT-SE and TAU are presented in Additional file 3.

Information regarding the use of healthcare services, informal care and sickness absence was collected at 3, 6 and 12 months after study inclusion. The information was self-reported by the participants, including only trauma-related treatment and follow-up, using a questionnaire specifically designed for that purpose (developed by EAA). The questionnaire included number of visits to a general practitioner, physical therapist, chiropractor, contract specialists (dentist, neurologist, ophthalmologist, orthoptist/optician, otorhinolaryngologist and psychologist) and out-of-pocket services (naprapath and osteopath). Information about informal care was collected by asking participants if they had received informal care by friends or family since last follow-up and, if so, number of hours receiving assistance per week. We also recorded information about gross annual salary and productivity loss, which included work status in terms of sickness absence (percentage and duration) at each follow-up. We did not assess the use of medications, distance and transportation related to the healthcare utilisation and, hence, could not calculate medication and transportation costs. The cost of productivity loss was based on the participant’s gross annual salary, and was calculated by subtracting weekly wage for full time work (i.e., pre-injury level of work) from weekly wage for actual number of hours worked per week at each follow-up (i.e., taking into account sickness absence). Costs of informal care was calculated using the opportunity cost method (i.e. time spend providing informal care was valued as paid working time [18], multiplying the mean hourly wage in Norway [19] by number of hours of informal care per week. Health service utilisation and related costs, including formal primary and secondary care reimbursed by the government, were categorized according to the CCT-SE and TAU groups. Total care costs also included informal care and out-of-pocket services (i.e. services not reimbursed by the government). Total societal costs comprised total care costs in addition to costs related to productivity loss. Cost categories and unit costs are presented in Table 2. The majority of unit costs were based on the reimbursement schemes. Costs were estimated on a present-value basis of Euro (€) in 2019 (€1 = Norwegian Kroner [NOK] 10).

The primary cost-effectiveness analysis was based on the difference between CCT-SE and TAU in number of days until return to pre-injury work level (i.e., cost per day earlier back to work). As the secondary outcome, cost-effectiveness was evaluated in terms of quality-adjusted life-years, QALYs, where 1 QALY is equal to 1 year in full health. This was measured using utility scores derived from the 5-level EQ-5D version (EQ-5D-5L) based on five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression [20]. EQ-5D-5L was administered at baseline and follow-ups (3, 6 and 12 months), and valued using the Danish tariff as a Norwegian tariff is currently not available [21].

Descriptive statistics were used to present participants’ characteristics, costs of the CCT-SE intervention and TAU, health- and informal care and productivity loss, with mean and standard deviation (SD) or median and inter-quartile range (IQR) for continuous variables, and proportions and percentages or range for categorical variables.

In the cost-effectiveness analysis, the differences in outcomes of CCT-SE and TAU were compared to difference in costs.

The primary cost-effectiveness analysis was defined by the incremental cost-effectiveness ratio (ICER), given by

where the nominator in Eq. (1) is the incremental cost, while the denominator in Eq. (1) can be interpreted as number of days earlier back to pre-injury work level in the CCT-SE group. The costs included in the primary analysis were total healthcare costs (i.e., costs of the CCT-SE intervention or TAU and other healthcare services).

The secondary cost-effectiveness analysis was based on QALYs as the health outcome. There were 15 participants with missing EQ-5D-5L values. The two missing values in the TAU group at baseline were replaced by the average value for the TAU group. When a participants had a missing observation between two observation points on the time line, we assumed the relationship between the adjacent recorded observations to be linear. For instance, for missing values at 3 months we assumed a linear trend from the observations at baseline to 3 months was carried onwards to 6 months. Further, for participants with reported EQ-5D-5L at baseline and 3 months, but not at 6 and 12 months, we assumed that the trend from baseline to 3 months were carried onwards to 6 months, which resulted in two additional observations being replaced. For missing values at 12 months, we assumed the same EQ-5D-5L value as observation at 6 months, which resulted in 11 observations being replaced.

We estimated the QALYs for each individual by area under the curve (AUC), where we used the EQ-5D-5L values at baseline, 3, 6 and 12 months follow-ups and assumed linear changes in HRQoL over time [22]. From Table 1 we observed that the EQ-5D-5L baseline values were significantly different between CCT-SE and TAU (p = 0.013). To adjust for the difference, we applied two methods. Firstly, based on the method presented by Manca et al. [22], we used regression analysis to adjust estimated QALYs for EQ-5D-5L value at baseline and treatment group (see Additional file 4). Secondly, we parallel shifted the observation for the intervention group upwards and equal to the difference at baseline, equal to 0.065. Analysis based on the raw data are presented in Additional file 5. The costs included both total healthcare and societal costs. Lastly, we have included an analysis comparing the improvement in HRQoL values from baseline to 12 months. In this alternative, the incremental effect will be interpreted as the difference in incremental improvement on the HRQoL value over the 12 months follow-up period.

For the secondary cost-effectiveness analysis where QALYs were the main health outcome, we estimated the net monetary benefit (NMB), which is defined by

where λ refers to the threshold value for a QALY gained. If the NMB is lower or equal to zero, the intervention is considered cost-effective. According to Norwegian guidelines the threshold value depends on the severity of the condition defined by the absolute shortfall (number of lost years in good health), and varies from about € 27 500 to € 1 000 00 [23]. In the estimation of NMB, we applied € 27 500 as the threshold value.

To estimate cost-effectiveness with the change in HRQoL from baseline to 12 months, the results were presented by ICER, defined by Eq. (1) with differences in change in HRQoL.

To account for uncertainty in the outcomes (number of days back to work, QALYs, change in HRQoL and costs) between individuals, we conducted a sensitivity analysis by the bootstrap methods with replacement including 1000 iterations. The results of the 1000 iterations of mean ICERs are presented in a cost-effectiveness plane, and for the secondary cost-effectiveness analysis the cost-effectiveness acceptability curves (CEACs) were presented, reporting the likelihood of the CCT-SE intervention to be cost-effective (number of simulations falling below the threshold) according to different threshold values.

Three different cost-effectiveness analyses were conducted. The primary analysis was based on the number of days until return to pre-injury work levels. From Table 5, we see that from a healthcare perspective, the ICER was € 56 per day earlier back to work (897/16) for the CCT-SE group compared to TAU. The mean production loss per day for the CCT-SE group was € 275, thus outweighing the cost per day earlier back to work.

The second and third analyses were based on the secondary outcome, QALYs. When estimating the incremental QALYs by adjusting for EQ-5D-5L values at baseline, the incremental QALYs was 0.009 (see the regression output reported in Additional file 4 ). Given a threshold value of € 27,500, the NMB from the healthcare perspective was € -561 (0.009*27,500–979), indicating that CCT-SE was not a cost-effective alternative to TAU at the given threshold. From the societal perspective, the NMB was € 1,566 (0.009*27,500-(-1,319)), where the incremental costs are negative indicating CCT-SE to be a cost saving alternative compared to TAU.

In the next alternative, where we parallel adjusted the EQ-5D-5L values equal to the difference at baseline, the NMB was € -132 (0.031*27,500–979) and € 2,336 (0.037*27,500-(-1,319)) from a healthcare and societal perspective, respectively. Lastly, when the health outcome was measured as a change in HRQoL from baseline to 12 months, the ICER was € 19,260 per QALY gained (979/0.042), and € – 109,900 per QALY gained (-1,319/0.012) for health care and societal costs, respectively. In Fig. 1, we have presented the uncertainty analysis for the primary outcome (number of days earlier return to work). All scatters represent estimated mean incremental health care costs and number of days earlier back to work. Scatters above the horizontal axes indicate that CCT-SE is more costly than TAU (99.9% of the scatters), while scatters to the right of the y-axes indicate a positive effect of CCT-SE on number of days earlier back to work (76% of the scatters). In Fig. 2, scatterplots and CEACs are presented for three health outcomes. The upper two figures present the results from the analysis where the QALYs gained was adjusted for the difference in baseline EQ-5D-5L value. All scatters are above the horizontal axes, indicating that CCT-SE is more costly than TAU when only health care costs are included. From the CEAC we see that the probability of CCT-SE being cost-effective is higher than TAU for threshold values above € 98,000.

The two figures in the middle present the results based on a parallel shift of the EQ-5D-5L values equal to the difference at baseline. From the scatterplot we see that 92% of the iterations have a positive incremental QALY. The CEAC illustrates that at a threshold of € 32,000, the likelihood for CCT-SE and TAU to be cost-effective is equal to 50%. For threshold values above € 32,000, the likelihood for CCT-SE to be cost-effective is higher than TAU reaching 80% at values above € 69,000. In the last row we have reported results based on difference in the improvement of HRQoL scores from baseline to 12 months between CCT-SE and TAU. The costs are the same as for the two other alternatives, while 94% of the scatters indicate a positive effect of CCT-SE on the change in HRQoL over 12 months. For threshold values above € 22,000, CCT-SE is more likely to be cost-effective compared to TAU. In Fig. 3, the above analyses were re-run, but included societal costs. We see that the proportion of iterations with a negative incremental value (indicating CCT-SE to be cost-saving) was equal to about 63%.

Considering both healthcare and societal costs, costs associated with healthcare services were relatively low and accounted for 8% of the total societal costs. This could be related to the injury severity levels of the participants included in this study; patients with mild and moderate TBI are less likely in need of complex and expensive healthcare services compared to those with severe TBI [24]. However, at the population level, costs following mild TBI far exceed those of more severe injuries as a result of the high incidence of mild TBI [25, 26]. Moreover, indirect costs related to loss of productivity are generally considered to exceed the direct costs associated with healthcare utilization after TBI [27].

Healthcare costs were higher in the CCT-SE group which may in part be due to the small number of patients in some of the CCT groups leading to use of more healthcare resources per patient and consequently higher healthcare costs. The intervention was provided as part of a research project operating under resource and time restrictions, while providing the intervention in a clinical setting could allow for larger group sizes as the intervention developer notes that CCT may be delivered in groups of up to 7 participants [15]. The baseline EQ-5D-5L index values were significantly lower in the CCT-SE group compared to TAU despite randomisation. Lower initial HRQoL could indicate a greater need for healthcare services, thus might have contributed to the higher consumption and costs of healthcare services observed in the CCT-SE group. However, these findings may also indicate that the CCT-SE intervention did not cover the participants specific needs for healthcare services such as treatment of somatic problems (headache, dizziness, nausea, vision problems) leading them to seek more additional services.

Costs related to productivity loss constituted the largest costs in both groups, accounting for 84% of total costs. These costs were lower in the intervention group supporting the main findings of the RCT, showing a significant difference in work participation between CCT-SE and TAU at 3 months follow-up. Additionally, participants in the CCT-SE group returned to preinjury work levels earlier which might indicate a beneficial effect of the combined cognitive and vocational intervention [10].

A substantial proportion of support required to live at home after TBI is provided by informal caregivers, family members and friends, but these costs are seldom reported in clinical trials. In this study, informal care costs accounted for 8% of the total costs, and were lower in the CCT-SE group compared to TAU. The CCT intervention provides psychoeducation about TBI and strategies to manage cognitive and other symptoms which may have had a positive influence on independence in functioning. An alternative explanation may be found in differences in the demographic characteristics of the participants. Compared to CCT-SE, a higher proportion of participants in the TAU group (22% vs 17%) reported that they were single and thus may have needed more support from their environment in managing the consequences of their TBI.

This is the first study to evaluate the cost-effectiveness of a combined cognitive and vocational rehabilitation intervention in patients with mild and moderate TBI. Overall, the findings show that the incremental health care costs are positive, while the incremental societal costs are negative, which implies that CCT-SE intervention is a cost-saving alternative to TAU from the societal perspective. The incremental effect depends on the analysis. In the primary analysis, the CCT-SE intervention resulted in earlier return to work. In the secondary analysis, when measuring the health outcome in QALYs, the size of the incremental effect depended on whether and how the imbalance in HRQoL at inclusion were adjusted. Both groups improved their HRQoL, and improvements over 12 months were in line with previously reported estimates of minimally clinical important difference in EQ-5D index values ranging from 0.03 to 0.52 [28]. The change in HRQoL was, however, greater in the CCT-SE group, suggesting a positive effect of intervention.

In 2016, the suggested threshold value for a QALY gained in Norway was € 27 500 – € 100 000 per QALY for the lowest to severest grade patients [23]. In this study, when we parallel adjusted the EQ-5D-5L values equal to the difference at baseline, the cost-effectiveness of CCT-SE and TAU was equal at the threshold of € 32 000. However, for threshold values above € 32 000, the likelihood for CCT-SE to be cost-effective was higher than TAU. Further, when considering the between-group incremental difference in HRQoL from baseline to 12 months, CCT-SE is more likely to be cost-effective compared to TAU from threshold values above € 22 000. Mild TBI is a highly prevalent condition with a high impact on productivity loss and societal costs and could in this context be considered as a more severe condition. In this study, costs associated with productivity loss accounted for 84% of the total costs, highlighting the impact of mild TBI on societal costs.

It is difficult to compare results from this study with previous studies due to substantial variation regarding methodology, patients’ clinical characteristics and intervention content. The majority of studies assessing the cost-effectiveness of rehabilitation after TBI are carried out on severe TBI [29‐31], with results supporting the cost-effectiveness of early transfer to a dedicated neurocritical care unit or specialized rehabilitation. Our intervention can also be considered as an early, specialized intervention launched 8–12 weeks after mild-to-moderate TBI. Thus, the study findings give further support to the notion that targeted early intervention programs for patients with TBI might be cost-effective, in this case due to the costs associated with informal care and productivity loss.

This study systematically evaluated and compared the costs related to an intervention program and standard care both during and after the intervention phase for working age patients with mild-to-moderate TBI. The cost evaluation did not only include costs related to the study treatments and healthcare services, but also productivity loss and informal care from family caregivers related to daily life consequences of the injury, reflecting meaningful impacts of TBI on families and society. The societal costs were dominating both with regard to the total costs as well as the incremental costs, which resulted in CCT-SE being a cost-saving alternative. Guidelines for economic evaluations in several countries recommend analysis to be conducted from a health care perspective [32]. This study has illustrated that for some patient groups and types of interventions, a health care perspective does not fully represent the consequences an intervention could imply.

There was a significant difference in baseline EQ-5D index values despite following randomisation procedures. The lower baseline index value in the CCT-SE group may have allowed for greater improvement in HRQoL over time compared to TAU and should be considered when interpreting results from this study. There were 28 missing observation points for EQ-5D-5L amounting to approximately 6% of the total number of observations. By manual imputation, we reduced the number of missing information to 11 EQ-5D-5L values (2.4%), which resulted in 56 and 55 observations of QALYs in the CCT-SE and TAU group, respectively. Alternatively we could have applied multiple imputation, but as the number of observations was relatively small, we do not believe that it would change the results. There were also missing information in the estimation of production loss and informal care, resulting in a reduction in number of observations to 46 and 44 for CCT-SE and TAU, respectively. Replacing production loss and informal care, was not possible based on the information in the study. In situations where missing values are not at random, the results in the cost-effectiveness analysis could be biased. The number of missing values in the analysis applying the health care perspective is relatively small, hence it is likely that the findings are representative. With regard to analysis from the societal perspective, future research should focus on ensuring data collection that reduce the number of missing information to validate the results from our study. The study was conducted within the specialist healthcare service in Norway, where the study intervention was compared with TAU which is a standard follow-up practice in the study hospital, but may not be representative of practice in other hospitals in Norway or internationally. Thus, the results may not be generalizable to other health care systems with different care practices or economic conditions.

The cost of informal care was valued in terms of mean hourly wage multiplied by hours spend on providing care (i.e. productivity loss of the caregiver). However, we did not register whether the informal care was provided during work hours or leisure time, thus time spent providing care did not necessarily equal lost productivity for the informal caregivers. Moreover, hours of informal care per week was reported by the participants which may have led to an over- or under-estimation. The operationalization and valuation of informal care is a subject of discussion [18]. While some argue that valuing informal care as lost wage may lead to an over-estimation of costs, others argue that leisure time should be valued equal or above market wage rate. We applied the human capital approach [33] in the analyses, assuming that a person on sick leave cannot be replaced and hence all absence from work is included as a cost. Alternatively, we could have applied the friction cost approach, where only the costs related to replacing a person on sick leave is accounted for in the cost [33]. This approach would most likely have resulted in lower societal costs. Obtaining sick-leave and healthcare claims for the participants would provide a more valid measure of healthcare use and cost; however, the self-reported utilization is necessary when including costs for non-covered services such as informal care and out-of pocket healthcare services. Moreover, self-reported productivity loss and healthcare utilization could be considered a limitation if respondents report false values or have difficulty remembering. However, the study participants had sustained mild-to-moderate TBIs, and their knowledge of personal work-related data was not suspected to be notably affected. In addition, neuropsychological assessment at baseline showed that severe memory or attention problems were not present to such a degree that one would expect invalid reports [34].