Aim

To illustrate the potential utility of a simple filter model in understanding the patient outcome and cost effectiveness implications for depression interventions in primary care.

Steps 1–3: Identify the relevant group data

1. Population: define and quantify the population of interest (e.g., general population, certain age cohorts);
2. Target group: define and quantify the target of the intervention (e.g., depressed individuals);
3. Setting attendance: enter the proportion of the target group attending the setting of interest (e.g., depressed individuals attending primary care);

Steps 4–6: Adding the filters representing different intervention scenarios

When the user enters the relevant proportions for the following steps, the number of persons at each level in the model is calculated and used as output in the subsequent step. For example, the calculated number of persons of the target group attending the setting of interest who are detected for the condition of interest (step 4) are used as the starting figure for reach and adherence (step 5).

1. 4. Detection: the proportion of the target group attending the setting of interest who are detected for the condition of interest. An intervention can be applied here to increase detection using relevant intervention effect (increase in proportion detected) and cost estimates.
2. 5. Reach and adherence: the following are estimated by the user. An intervention can be applied at each of these steps with the relevant intervention effect (increase reach or adherence proportions) and cost estimates required.
   1. Reach: The proportion of individuals with the condition of interest who are offered treatment by their healthcare provider;
   2. Adherence: The proportion of individuals with the condition of interest who would comply with the offered treatment;
3. 6. Effect on outcome: for a dichotomous outcome measure of treatment success, specify the expected proportion of those that are offered the intervention that will achieve a successful outcome following the intervention.

The outputs derived from the above data are outlined in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Model outputs and calculations used to derive outputs.

https://doi.org/10.1371/journal.pone.0246728.t001

Data analytic procedures: Application of the filter model to depression treatment

To demonstrate the potential utility of the decision analytical filter model for improving depression outcomes, data for four scenarios, including ‘usual care’, and three hypothetical conditions, are presented. A successful outcome for the following scenarios was defined as an individual reaching ‘remission’, i.e., below threshold depression levels. The baseline scenario attempted to model standard care in the identification and treatment of depression for primary care patients, using data drawn from the literature (see Table 2). This scenario assumed no interventions were implemented and that usual care was provided. The three hypothetical scenarios represent an improvement above usual care in each of the filters. An estimation of the intervention effect and relevant costs associated with achieving these improvements were entered into the model. While efforts were made to draw these estimates from the literature, the aim of the model, in its current status, is to illustrate hypothetical scenarios. Further, the population and target group can be adjusted for different countries, regions or settings. The Australian population has been used in the model presented for illustrative purposes, however, the derived data is drawn from international references. Filter one, i.e., detection, examined an intervention aimed at increasing detection of depression in primary care. This involved administering a free validated depression measure to primary care patients electronically using a touchscreen computer upon presentation to their appointment. Filter two, i.e., reach, involved increasing the offer of evidence-based depression treatment from providers. This intervention is based on training and support for GPs in appropriate management of depression. Filter three, i.e., adherence consisted of increasing patient adherence to offered treatment. This involved a telephone intervention to monitor the progress of patients undergoing treatment for depression and provide adherence strategies. All filters were assumed to operate independently from one another, for example an increase in the proportion of patients detected for depression did not increase the proportion of patients offered treatment by GPs. Dollar amounts are presented in Australian Dollars (AUD).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Model parameters and output under usual care and three hypothetical intervention scenarios.

https://doi.org/10.1371/journal.pone.0246728.t002

Comparison with existing literature

Interestingly, the findings of the model in the example above indicated that filter one, which aimed to increase detection through a screening intervention using touchscreen computer tablets, was the most cost-effective filter. While there is some contention regarding screening for depression in primary care [32–34], this method of intervening represents a relatively low cost and low maintenance strategy while greatly increasing the number of successes in the hypothetical scenarios. Further, the considerable investment placed on technological health advancements [35], including mobile health [36, 37], could provide opportunities to explore remote screening options via smartphone messages and apps that link in with primary care medical records. This approach could be conducted automatically and notify healthcare providers if their patients are at-risk of depression, which would further reduce hardware and personnel time associated with in-clinic screening. Electronic screening options therefore present a feasible intervention option to improve administration rates of depression treatment. While interventions relating to reach and adherence are important areas to examine, the greater level of intensity required to undertake these interventions, as well as a smaller proportion of patients accessed, resulted in much higher ICERs for these interventions. As previously stated, the findings presented from the model are intended to be illustrative, however, where possible, the data were drawn from the literature. Therefore, the approach of targeting and comparing these different filters within primary care still warrant consideration.

Strengths and limitations

Use of the model should be considered in light of its limitations. Firstly, the information derived from the model is only as strong as the input data provided. It is therefore limited by any inaccuracies or biases that exist in the studies the data is derived from. Further, some data used in the model were based on research which was more than ten years old. This was due to the model requiring specific information to meet the parameters, such as intervention costs per person, relevant targeted interventions (e.g., provider training) and binary study outcomes (e.g., remission rates). While this limited the data available for use, the parameters of cost and outcome are of high importance to policy and health services decision makers. This limitation is also a strength of the model, as outcomes and costs for a single intervention can also be compared when there is uncertainty in key parameters. Nevertheless, future decision-makers utilising the model to allocate large amounts of resources may consider applying quality checklists to included studies, such as the risk of bias criteria from the Effective Practice of Care (EPOC) collaboration [38] to ensure relevant information can be drawn from high quality research trials.

Additionally the model assumes that intervening at one filter does not alter rates at other filters. For instance, in the worked example, training GPs in depression management may increase the quality of the information provided to patients or improve the likelihood of GP follow-up, which may inadvertently increase patient adherence to treatment. Furthermore, incorporating the complexity of certain conditions in to this simple model is difficult. For example, depression is associated with different levels of severity and treatment pathways differ based on this severity. The presented model only accounted for intervention-related costs and did not include health care costs associated with severity of illness and hence, increased treatment intensity, healthcare provider time, nor the patient perspective in estimating costs. Severity of conditions and health care costs could be considered in future applications of the model. The model is also not able to account for inaccuracy of detection, such as false positives and false negatives and therefore issues such as unnecessary treatment and intervention were not considered. A further strength of the model is its ability to let the user enter a range of values for key input parameters. Hence, the user can vary the value of any of the filter model’s input parameters that might be associated with meaningful levels of uncertainty. This then produces a range of outcome values (or boundaries) that can be considered in decision making.

Implications for research and/or practice

This theoretical filter model, developed using economic cost-effectiveness principles, has the potential to influence discussions on resource allocation involving decision or policy-makers for depression management in primary care. The simplicity and accessibility of the filter model represent the strengths of this approach. It is stored within a Microsoft Excel file and can be used by individuals without specialised skills, including those working within relevant care settings for depression. While the model can be applied at a national level, as was demonstrated in this paper, it could also be used to make decisions at a local level such as within a singular health services. Further, using this tool to examine depression could be expanded beyond the primary care setting to specialised mental health or hospital settings. Such an application of the tool could be used to examine the ICER of different aspects of treatment being offered by these services. The calculator is cost-free which makes it a feasible option for health care facilities to optimise cost-effective treatment outcomes.

Input data required for the tool should be evidence-based regarding target populations, reach and effectiveness. Generally, these data are likely to be available in the scientific literature and population health reports. However, another useful aspect of this model is that it highlights gaps in the current literature that prevent examining certain aspects of selected care pathways. For instance, when considering the depression literature examined, there was a lack of economic analysis reported for intervention studies undertaken. While the effectiveness of interventions in improving depression outcomes is of great importance, the cost-effectiveness is equally valuable for determining the feasibility of implementing research findings in to actual care. Future research examining depression interventions in primary care should carefully consider this aspect of research.