Economic evaluation of a Decision Support Tool to guide intensity of mental health care in general practice: the Link-me pragmatic randomised controlled trial

In Australia, common mental health problems are initially diagnosed and treated by general practitioners (GPs) [1]; however, under- and over-treatment is common. Australian mental health reforms have promoted stepped care to improve the efficiency and quality of mental health care. Evidence suggests that stepped care can be effective for the treatment of depressive and anxiety disorders [2, 3]; few studies have explored the value for money credentials of stepped care approaches for the treatment of anxiety and depression in primary care settings [3].

The Link-me approach to stepped care comprises a patient-completed digital Decision Support Tool (DST) to predict the severity of depressive and anxiety symptoms in three months’ time, support individuals to identify their mental health treatment priorities, and deliver prognosis-based treatment recommendations.

In a pragmatic, stratified randomised controlled trial (RCT), the Link-me intervention resulted in greater reductions in psychological distress at six months over usual care[4]. This paper reports on the economic evaluation incorporated in the RCT [5].

The Link-me RCT was conducted within 23 general practices in three Australian states (New South Wales, Victoria, Queensland) across metropolitan, outer metropolitan and regional locations in collaboration with three Primary Health Networks (PHNs). PHNs are not-for-profit companies commissioning health services in 31 geographically defined areas throughout Australia [6]. The study protocol and analysis plan for clinical and economic outcomes are described in full elsewhere [4, 5].

Briefly, adults attending a participating general practice for any reason were invited to complete an eligibility screening tool and the DST (Supporting Information, section S1) on a tablet device in the waiting room. Eligible patients (ages 18–75 years; proficient in English; providing a phone number and email address; having a Medicare card; reporting current anxiety or depression symptoms or use of medication for mental health) were classified into three prognostic groups. Participants categorised into the minimal/mild and severe groups were individually randomised to the intervention or control group. Participants classified as minimal/mild and allocated to the intervention group were offered a selection of low intensity treatment options available through local in-person services, online, by telephone or mobile app. The recommended options were viewed on the tablet immediately after completing the DST and a copy was emailed to participants including links to the relevant services. The severe prognostic group allocated to the intervention were offered care navigation, a model of clinical care coordination informed by the principles of collaborative care and motivational interviewing. Participants could receive up to eight contacts with a trained care navigator to develop and implement a structured care plan to address priorities they identified in the DST. Participants could also access additional care package funding, through the participating PHNs, to support service access when cost was a barrier to care. Participants randomised to the control group for both minimal/mild and severe prognostic groups received advice to discuss any mental health concerns with their GP. The moderate group were excluded from the study based on the availability of mental health service offerings appropriate to this level of need [7].

This is the first economic evaluation of a digital platform designed to predict the severity of symptoms and coordinate care for people with depression or anxiety in primary care. Our results suggest that Link-me was associated with additional health sector and societal costs but resulted in achieving additional improvement in mental health symptoms. Incremental costs were highest for the severe prognostic group which also achieved the largest incremental improvement in mental health symptom scores.

Given the time-limited nature of the trial, reported costs may be considered the upper estimate of the true costs associated with delivering care navigation (i.e., training and development) and would likely reduce over time once the intervention became established. The Link‐me intervention was most likely not running at full economic efficiency because the rollout of the actual process of care navigation was being fine‐tuned with the cost of likely learning effects being incurred. If the learning effects were finalised and Link‐me was running at steady state, the throughput of participants would be increased, leading to reductions in average cost per person and contact.

Link-me is a novel approach; therefore, published evaluations of directly comparable interventions are lacking. An economic evaluation of the precursor digital platform to Link-me used to manage the treatment of people with depression found it to be dominant (more effective and less costly) over usual care from both health sector and societal perspectives [18]. It utilised the Assessment of Quality of Life-8D (AQoL-8D) questionnaire [19] to measure utility values and calculate QALYs in contrast to the current evaluation using the EQ-5D-5L. Despite improvements in mental health symptom scores measured with the K10 in the current evaluation, no significant differences were found between groups on the EQ-5D-5L utility scores or QALYs. Economic analyses of collaborative care interventions for treating depression in primary care found mixed results for cost-utility analyses [20, 21]. Studies using the Short Form quality of life measure detected significant differences in QALYs, while those using the EQ-5D found negative or non-significant incremental QALYs [20]. This may reflect lower sensitivity of the EQ-5D to changes in quality of life resulting from improvement in mental health symptoms, despite previous studies indicating that most multi-attribute utility instruments can discriminate between differing levels of symptom severity for people with depression [22]. Link-me provided statistically significant improvement in K10 scores over existing care at six-month follow up, however the mean difference did not reach the prespecified minimal clinically relevant difference of 2.4 points (equivalent to a standardised mean difference of -0.3). However, half of the participants allocated to the Link-me intervention did not accept the offer to engage in care navigation. When a complier average causal effect analysis was undertaken to assess the magnitude of change in K10 scores associated with engagement in the different steps of care navigation, the intervention effect size increased and led to clinically meaningful differences [4]. This suggests that there is an opportunity to optimise the Link-me intervention to improve engagement potentially leading to improved clinical outcomes.

Average health sector and societal costs across all participants for both the Link-me and control groups were higher than reported in previous economic evaluations of stepped and collaborative care interventions for people with depression [3, 20, 21]. The addition of anxiety symptoms to the prognostic model of the DST may have contributed to this result. Previous research found that health sector and productivity costs for comorbid depression and anxiety were significantly higher than either diagnosis alone [23]. Another explanation is that there was increased time missed from work since study participants in both randomised groups were seeking additional treatment for their mental health concerns.

The Link-me intervention was hypothesised to be cost-effective through triaging participants in the minimal/mild prognostic group to low intensity services associated with lower costs and offsetting the increased costs of care in the severe prognostic group. However, participants in the minimal/mild prognostic group who were offered low intensity services had a significantly higher average mean health sector cost at 12‐month follow‐up than their usual care counterparts, in part due to their greater use of psychology services. This may be due to the pragmatic nature of this real-world trial since we could not limit access to more intensive forms of care. Participants in the Link-me group utilised psychological services at a similar rate to the comparison group among the minimal/ mild prognostic group (25% intervention; 19% comparison). The feedback and opportunity for reflection provided by the Link-me DST may have prompted increased help-seeking through psychologists, in preference to the suggested low-intensity options, partially accounting for the higher costs observed.

The Link-me approach to stepped mental health care would not be considered cost-effective utilising the cost/QALY outcome metric commonly used by health technology assessment agencies. Link-me could be viewed as trending toward cost-effectiveness by providing improvement in mental health symptoms, assessed by the K10, for additional cost. Given the Productivity Commission’s recommendation for a national digital mental health platform to support a person-centred approach to mental health care in Australia [27], Link-me should be appraised for this important function.