ABSTRACT
Objective Screening has an essential role in preventive medicine. Ideally, screening tools detect patients early enough to manage the disease and reduce symptoms. We aimed to determine the cost-effectiveness of routine screening schedules.
Methods We used a discrete-time nonstationary Markov model to simulate the progression of depression. We adopted annual transition probabilities, which were dependent on patient histories, such as the number of previous episodes, treatment status, and time spent without treatment state based on the available data. We used Monte Carlo techniques to simulate the stochastic model for 20 years or during the lifetime of individuals. Baseline and screening scenario models with screening frequencies of annual, 2-year, and 5-year were compared based on incremental cost-effectiveness ratios (ICER).
Results In the general population, all screening strategies were cost-effective compared to the baseline. However, male and female populations differed based on cost over quality-adjusted life years (QALY). Females had lower ICERs, and annual screening had the highest ICER for females, with 11,134 $/QALY gained. In contrast, males had around three times higher ICER, with annual screening costs of 34,065$/QALY gained.
Conclusions Considering the high lifetime prevalence and recurrence rates of depression, detection and prevention efforts can be one critical cornerstone to support required care. Our analysis combined the expected benefits and costs of screening and assessed the effectiveness of screening scenarios. We conclude that routine screening is cost-effective for all age groups of females and young, middle-aged males. Male population results are sensitive to the higher costs of screening.
INTRODUCTION
Depression is one of the most common mental health conditions and a leading cause of disability that results in substantial impairment. (1) It co-occurs highly with other physical illnesses, which causes health and economic burdens to individuals and society. It is estimated that depression cost $210 billion in 2010, including direct medical and indirect costs (e.g., loss of productivity). (2)
On average, 22.9% of females and 15.1% of males experience at least one episode of major depression in their lifetime, and 43.3% of patients are not receiving any treatment. (3) Considering high prevalence and low detection rates, improved management of disease reduces health spending and improves patients’ quality and quantity of life.
Both major depression and minor depression present with either depressed mood or loss of interest or pleasure in usual activities along with other symptoms, and the signs are present for two weeks or longer. Minor depression, with a few mild symptoms, can be resolved without treatment, but it may evolve into major depression. Major depression with moderate or severe symptoms significantly impacts life quality and may require treatment from a mental health specialist. However, most depression cases are treated in general medical settings in practice. (3) Depression is a recurrent disease, and at least full or partial remission is required to distinguish episodes. (4)
Between 1998 and 2018, suicide rates increased by 35% in the overall population. (5) A 50% increase in suicide rates was observed among women from 2000 to 2016. (6) Due to the high cost and the negative impacts on patients’ overall health, improved detection efforts attract policymakers and healthcare agencies.
Routine depression screening can be used to improve the recognition of depressive episodes. The U.S. Preventive Services Task Force (USPSTF) suggests screening of patients with a system that accurately diagnoses, treats, and follows up with the patients. Still the net benefit of screening, optimal screening interval, and timing is unknown. (7) Studies in literature questioned the USPSTF recommendations, (8) and most of them are limited to show the impact of screening. (9)
In some health systems, screening is commonly used; in other settings, implementation is limited. (10) Therefore, in current clinical practice, screening measures are variable. According to the National Ambulatory Medical Care Survey (NAMCS) in the United States, only 2.29 percent of primary care visit patients were screened for depression in 2010. (11) Whereas, in postpartum populations, screening is implemented on a large scale (79%). (12) Given this uncertainty, exploring the benefits of screening can provide insight into decisions about routine depression screening.
In this study, we aimed to determine the cost-effectiveness of routine screening schedules. A natural history model was introduced where probabilities were assigned to each health state based on patient characteristics such as age, gender, and history of depression. We used Monte Carlo simulation to evaluate disease progression over the years. We assessed the various routine screening frequencies based on cost and benefits.
METHODS
Model Structure
We used a discrete-time nonstationary Markov chain model. The model consists of ten health states: healthy, major depression without treatment, major depression with treatment, partial remission, full remission from major depression, minor depression without treatment, minor depression with treatment, full remission from minor depression, suicide, and death from other causes (Figure 1). States were similar to Valenstein et al.’s (9) model. But distinct from their work, we used more detailed remission states, e.g., partial and full remission from different types of depression.
Markov Health State Model
We classified the existence of 2 to 4 depressive symptoms as minor depression. (13) We considered patients in partial remission of current major depression when they had some symptoms of major depression; however, the full diagnostic criteria of major depression were not satisfied. (4) Healthy state included individuals who were never depressed (minor or major).
We used Monte Carlo techniques to simulate the stochastic model for 20 years or during the lifetime of individuals. We ran each scenario for 1000 replications. At the entry, patients were assigned their initial states based on prevalence, and they transitioned between states at the end of each year. The initial population of 100,000 individuals was introduced to the system based on U.S. age distribution (14) and a constant number of people preserved in the system. In each time unit, the number of 18-year-old individuals entering the system was equal to the number of people leaving the system from states of suicide or death from other causes.
Model Parameters
We derived the model parameters from published data. Time-sensitive transition probabilities were used, where the mortality probabilities were increased by age. We used annual transitions that depend on patient history, such as the number of previous episodes, treatment status, time spent without treatment, and demographics (age, gender) if applicable. Table 1 shows the parameters that were used in the baseline model and their sources.
Major and minor depression parameters
When there was no consensus in the literature for a parameter value, we chose conservative estimates. For example, for the incidence of major depression, we used the lower bound of rates reported in the ECA study. (15) Consequently, our results were not biased towards screening.
We obtained all the varying values of sensitivity analysis parameters from the literature (Supplementary Table S1). We used the medium and high values of incidence and prevalence (3, 15) and matched two of these with using recall bias rates. (16)
Prevalence and Incidence
The ECA study’s incidence rates yield a 50% lifetime prevalence of major depression, (17) which is higher than reported in the literature. There are arguments around the potential discordance between incidence and lifetime prevalence rates of major depression. (18)
To be consistent with the reported incidence rates from national surveys, we used a lower bound of incidence. (15) We integrated the matching lifetime prevalence using a recall bias of 31.9% for females and 16.3% for males, and a 12-month prevalence recall bias of 25.5% for females and 9% for males, which were obtained from the literature. (16, 19)
Detection and treatment
We had a detection rate of 45% for major depression and 30% for minor depression in the baseline model. (9) 43 to 60% (3) of patients who initiated the treatment from major depression depend on the age of onset, whereas 20% (9) of cases with minor depression received treatment. The multiplication of these two probabilities ((detection rate)*(initiation of treatment)) represent the transitions from without treatment to with treatment state.
Treatment increased the full remission rate by 35% (0.37 without treatment vs. 0.50 with treatment) for major depression and 29% (0.55 without treatment vs. 0.71 with treatment) for minor depression. (Table 1)
Early treatment efforts had an impact on the outcome of antidepressant treatment. (20) Duration of untreated illness (DUI) longer than 6 months decreased the effect of antidepressant treatment. Patients who spent more than 12 months without treatment status have reduced remission likelihood by one third. (21)
Transition to remission state depended on the healthcare provider and the type of treatment. We assumed that 31% of major depression and 25% of minor depression cases were treated by a specialist. (9) Patients were treated with medications in primary care (9) Psychotherapy, medication, or a combination of these two was used during the treatment of a specialist.
Relapse and recurrence
Major depression is a recurrent mental health condition; around 67% of patients have at least one recurrence every 10 years. (22) Patients with residual symptoms were 3 times more likely to relapse than the fully recovered patients (0.76 from partial remission vs. 0.25 from full remission). (23) The number of previous episodes and treatment status affected the probability of recurrences. We obtained base recurrence rates for patients who had a history of cases pertaining to depression from (22) and patients who had one previous case and determined that they indicated an increased risk of recurrence by 16% compared to the base. Patients with 2 or more previous episodes had an additional 32% risk of recurrence. (22)
Patients who recovered from major remission without appropriate medication or other treatments were more likely (relative risk of 1.52 to 2.69 (24-27) for 52 weeks) to relapse. We increased the recurrence rate by 30% for patients transitioning to a remission state from major depression without treatment. (28)
Suicide and Death
We assumed that depression-related suicides occur only among major depressive patients regardless of treatment status. We derived the transition probabilities for suicide from the modifications noted in (9) for age and gender-specific rates. The validity of suicide rates was ensured by comparing the suicide prevalence ratios from 2001 to 2018. (29, 30) We assumed that at least 50% of suicides related with depression and 50-80% of older adults who die by suicide have been shown to have major depression. (29)
Age-specific death from other causes was utilized, (31) and we assumed that the major depressive patients had increased their risk of death based on other reasons by a ratio of 1.58. (32)
Screening Tools
For the baseline model, the detection rate of depression in usual care settings increased by 50% (9) for major depression (baseline 45% vs. with screening 68%) and 23% (9) for minor depression (baseline 30% vs. 37% with screening). (Table 1)
Costs and Utilities
We conducted our analysis from the societal perspective, which took into account direct and indirect costs. (Table 2) We assumed that patients who received treatment from primary care settings had 4 visits plus a follow-up visit in a year. (9) Patients treated by mental health specialists had around 11 visits. (9) 31% of patients received treatment from mental health specialists (26% self-refer and 5% referred by primary care physician) in a year. (9) Additionally, 1% of patients have inpatient visits in a 12-month period with an average of 11.6 days of stay. On the other hand, the cost of minor depression was estimated as two-thirds of major depression. (33) We calculated the average treatment cost of partial remission (34) as $1095, based on the number of remaining symptoms. We assumed there is no direct or indirect cost to capture during full remission states, as for the healthy state.
Cost and utilities for each health state.
The indirect cost was 20% lower in untreated patients ($1600 for treated vs. $3360 for untreated). The overall cost of major depression with treatment was estimated at $3087 vs. $3360 for major depression without treatment, which was similar to the estimates from Cuijpers. (33)
Utility estimates for all health states were obtained from the literature. (Table 2) We assumed that depression with and without treatment states had the same QALYs. We ignored the discount rate of utilities.
Incremental Cost-effectiveness Ratio (ICER)
For all screening scenarios, costs and effects were compared to a situation without screening. Strategies were evaluated based on the incremental cost-effectiveness ratio (ICER), which was calculated as the incremental costs per incremental QALY gained (Equation (1)). The scenario up to the willingness-to-pay threshold of $50 000/QALY gained was classified as optimal. (35)
RESULTS
We considered the following scenarios i) baseline (no screening), ii) annual screening, iii) screening in every 2 years, and iv) screening in every 5 years. We evaluated the case of screening for the overall population, starting at the age of 18 and over 20 years or the patient’s lifetime.
Screening Impact on Outcome Measures
The mean duration in treatment was estimated from the simulation model as 1.26 years for major depression and 1.32 months for minor depression among females. The suicide rates in the baseline model were 5.11 for females and 18.08 for males per 100,000 people in a year. Annual screening could prevent 24.3 cases of suicide per 10,000,000 people in a year (Table 3). Annual screening also shortened the initiation of the treatment time by 5.3 (vs. 2.8 for 2-year screening and 1.1 for 5-year screening) months for females and 3.2 (vs. 1.7 for 2-year screening and 0.6 for 5-year screening) months for males. Annual screening increased the depression-free months up to 5.5 months per year. All the screening scenarios extended time spent in remission (11 days to 1.9 months for females and 7 to 29 days for males based on screening schedule).
Cost-effectiveness measures of screening scenarios (average of 1000 replications)
Table 3 shows that, in the general population, all screening frequencies were cost-effective compared with the baseline. However, there exists a difference between the male and female population based on cost/QALY gained. Females had lower ICERs, and the annual screening had the highest ratio, with 11,134 $/QALY gained. On the other hand, males had around 3 times higher ICER, with annual screening costs of 34,072 $ per QALY gained.
Uncertainty among Monte-Carlo Simulations
Figure S2 (in supplementary) showed every single average incremental cost and QALY pairs for 1000 replications. For the female population, simulation results showed that all ICERs were less than the willingness-to-pay threshold of $50,000. Additionally, less than 10 percent of simulation runs for males were greater or equal to the willingness-to-pay threshold in every scenario.
Sensitivity Analysis
We performed a one-way sensitivity analysis on critical parameters. (9) In our baseline analysis, we observed that screening was cost-effective for both genders with low incidence and prevalence (conservative case). Consequently, the screening cost under medium and high incidence rates was cost-saving for females (2 and 5-year screening). For females, lower ICERs were observed compared to those for males.
Additional analysis on the other parameters such as the remission rate from major depression treatment, QALY of major depression, cost of screening, treatment, and indirect costs showed that female ICERs were not sensitive to any changes in these parameters. However, the ICERs of the male population was higher than $50,000 by using the upper limit of the screening. (Figure 2)
One-way sensitivity analysis results
Age-Specific Analysis
Three different age groups, young (18-34), middle (35-64), and old (65+) were analyzed by evaluating differences in cost-utilities among these subgroups; see Figure S3 (in Supplementary).
All scenarios were cost-effective for each age group of the female population. For the oldest population, about 10% of the simulations have ICERs between $50,000 and $75,000. (please see Supplementary Figure S1)
Young and middle age groups of the male population had ICER values up to $32,090/QALY gained. On average, screening with any frequency was not cost-effective for the oldest male population.
DISCUSSION
Annual screening of all adults 18 years old or older for depression would cost $11,134 and $34,065 per QALY gained for females and males, respectively. In comparison, screening every 5-years resulted in $8627 per QALY earned for females and $26,892 per QALY gained for males. We found that the benefit of early detection and treatment results in the improvement of quality of life. For both genders, annual screening produced greater costs. The ICER decreased with increasing screening frequency for males, whereas it rose from 2-year to 5-year screening in the female population.
In the literature, the combined results from trials showed that the depression screening increased the treatment rates between 2 to 50%. (36) In our model, patients spent an average of 3.65 years to initiate treatment for first-onset major depression. The World Health Organization’s (WHO’s) World Mental Health Survey showed that 35% of patients had 4 years of untreated time before initiating the treatment. (37) We observed that screening decreased the detection time up to 5 months and increased the remission time up to 1.9 months.
The age-specific analysis showed that screening of old populations had higher ICERs than the young and middle age groups. In the old male population, ICERs of all screening scenarios were below the willingness-to-pay threshold of $50,000. The main reason for gender differences in the cost-effectiveness of screening was the varying incidence and prevalence of major depression. In literature, men had lower incidence rates than women, (15) although there is concern that these rates were not reflecting the truth. (38) There were a couple of factors that may cause the underestimation of the rates for males. Firstly, the measurement bias was observed in data because of the inadequate attendance to surveys and lower primary care visits. (38) Secondly, men were less likely than women to have symptoms of depression that fit standard measurement tools. They experienced more externalizing symptoms, such as aggression, violence, and substance abuse. Studies showed that including these alternative symptoms with traditional symptoms as diagnosis criteria had increased the male prevalence rates to have equal proportions with females. (39) We evaluated the alternative incidence, prevalence, and screening scenarios for males with the same frequency as females. Our results are summarized in the Supplementary file.
We evaluated the external consistency of our model. By running the open network model for a longer time, where new individuals were introduced into the system at the age of 18 with the mortality rate, we observed that our initial distribution was equal to steady-state distribution. This was the case when transient behavior was not observed in the system. To obtain a consistent system, we adjusted the lifetime prevalence of major depression using reported recall bias rates. (16)
This paper evaluated the general U.S. population during non-crises period (e.g., economic problems, or pandemics). Recent studies showed that during COVID-19, incidence and prevalence were reported three times higher for moderate and severe depression. (40) We further included the prevalence rate changes between 2020 and 2021 in our analysis. Our results indicated that the depression screening is cost-saving for females at 2 and 5-year screening frequencies and around $11,000 to $15,000 cost per QALY gained for males based on the screening interval between 2020 to 2040 (please see Supplementary file).
LIMITATIONS
The cost-effectiveness of screening may be enhanced by targeting groups with a higher incidence of depression based on ethnicity, comorbidities, or poverty level. However, we analyzed screening scenarios that are valid for the general adult population. Because of incidence, disease progression, and suicide rate difference between females and males, we only considered the gender-specific model. We also extended our analysis for different age groups.
Our model did not consider the differences among the screening tools; we used average test sensitivity for 9 standard screening instruments. We assumed that the screening frequency was not changing at any time during the screening scenario. In our calculations, false-positive cases were not taking into account.
We did not consider the severity of major depression; however, we included the details about elevated risk based on age, gender, history of depression (major or minor), number of previous episodes, remaining symptoms for onset depression, treatment status, type of treatment and time spent without treatment states. We did not exclusively include medication drop-out rates (transition from treatment to without treatment) in our analysis. However, the treatment rates obtained from the literature implicitly include the response of adherent and non-adherent patients who are not fully complied the clinical treatment guidelines.
Like Valenstein’s model, (9) we assumed that suicide might happen when the patient had active major depression. Patients did not have medical and nonmedical costs when they were in full remission; however, staying there for a long time decrease the risk of relapses. We did not consider the medication costs that arose when patients were in remission. In the annual transition model, patients spent at least one year in treatment when they transition to treatment states, which is more than suggested acute therapy (6-12 weeks) and continuation therapy (4-8 months). We assumed that patients in partial remission with residual symptoms had receive maintenance treatment.
CONCLUSIONS
Depression is a common health condition that affects an individual’s everyday life. Considering the high lifetime prevalence and recurrence rates of depression, detection and prevention efforts play an essential role. Furthermore, the increasing trend in suicide rates becomes an emerging public health problem. We conclude that routine screening is cost-effective for all age groups of females and young and middle-aged males. Male population results are sensitive to the higher costs of screening, which indicates that if the screening cost is 44% higher than the average costs, the screening of the male population is not cost-effective. Our analysis combines the expected benefits and costs of screening and assesses the effectiveness of screening scenarios. Screening can be one of the cornerstones to support required care. Our baseline model could be utilized to evaluate the potential consequences of medication strategies or alternative intervention scenario.
Data Availability
All data used in the study is publicly available as described in the paper.
Footnotes
Conflicts of Interest: None reported.
Financial Support: This research was supported in part by the Center for Health and Humanitarian Systems, the William W. George Endowment at Georgia Tech, the following Georgia Tech benefactors: Andrea Laliberte, Claudia L. and J. Paul Raines, and Richard E. “Rick” and Charlene Zalesky, and the Allison Distinguished Professorship at NC State University.
REFERENCES
