Our research on the burden and costs of TRD and MDSR in Portugal had two main objectives. The first was to uncover the disease burden by estimating the component attributed to treatment-resistant depression and major depression with suicide risk based on the following indicators: mortality, morbidity, and Disability-Adjusted Life Years (DALYs) for the year 2017. Our second objective was to estimate the economic cost of diagnosing and treating TRD and MDSR, including direct costs (medical and non-medical costs) and indirect costs (loss of productivity – relevant to the perspective of society) for the year 2017.
The primary source of the epidemiological information used in this study is the “National Epidemiological Study of Mental Health” (NESMH) [3], which took place between 2008 and 2009. This study is the most accurate and robust national data available, assessing primary data collected in a representative sample of the adult population in mainland Portugal. The results of the NESMH were adjusted in our study, and all estimates were calculated using the population and demographic structure of the year 2017. Portugal’s NESMH was part of the World Mental Health Surveys Initiative (WMHSI), promoted by the World Health Organization and Harvard University. WMHSI was coordinated internationally by Prof. Ronald Kessler and is fully described elsewhere [21–26]. The methodology and implementation of NESMH are described in Xavier et al. (2013) [15].
The questionnaire [15] was divided into two parts to reduce the time taken to answer. Part I used a total sample number of 3849 participants to represent the general population and included an initial baseline assessment to diagnose major mental disorders. Part II of the questionnaire was answered by 2060 individuals which included all participants with a mental disorder diagnosis and a random sample of 25% of participants without psychiatric disorder. Two weights were created to accommodate the stratification of the sample. The data in Part I were weighted for the differential probability of selection (between and within households), non-response bias, and discrepancies between the sample and the geographic and sociodemographic distribution of the Portuguese population assessed in the census. The data of Part II was additionally weighted for the differential sampling of participants from Part I to Part II.
Psychiatric disorders were assessed using the World Health Organization World Mental Health Composite International Diagnostic Interview version 3.0 (WMH-CIDI 3.0), a comprehensive and fully structured interview designed to assess mental disorders according to the definitions and criteria of the DSM-IV and ICD-10.
Estimates of the prevalence of Treatment-Resistant Depression (TRD)
Before obtaining epidemiological information related to TRD, the concept of TRD was operationalized within the framework of available epidemiological data, adopting a “proxy” as collected data does not strictly respect any of the possible definitions of TRD. The criteria used to define a TRD case included the presence of a major depressive disorder in the last 12 months assessed objectively with a validated scale, and a negative answer to the question “Have you ever received treatment that you considered to be useful or effective?”, present in the depression module of WMH-CIDI 3.0.
Results show that the global prevalence rate of treatment-resistant depression in the population aged 18 and over in Portugal should be 1.1%. This prevalence corresponds to about 14.9% of the population with major depressive disorder. These values are obtained with a small number of observations (n=43) in a representative sample of the population, using the previously described weights. Combining this prevalence rate, estimated using the referred weights, with the estimates of the Mainland's adult population in 2017 from Portugal’s National Institute of Statistics (INE), a total prevalence for TRD of approximately 90 thousand patients is obtained.
Estimates of the prevalence of Major Depression with Suicide Risk (MDSR)
In the case of MDSR, the operationalization of the concept was achieved by combining major depressive disorder with suicidal ideation. Both situations were reported in the 12 months before the moment of the survey. The results obtained indicate that the prevalence of MDSR is expected to be around 0.8% of the adult population in Mainland Portugal. Multiplying the Mainland adult population by this rate, a total prevalence of MDSR of approximately 64 thousand patients was obtained in 2017.
Estimates of combined prevalence of Treatment-Resistant Depression (TRD) and Major Depression with Suicide Risk (MDSR)
To fully understand the data presented above, it should be highlighted that some patients check both criteria for identifying the different types of depressive pathology and, therefore, appear in both prevalence estimates. This reality reinforces the adequacy and need to study the cost and burden of the two clinical entities together.
The crossing of individual information in the survey with objective and representative data from the Portuguese population shows that about 11.2% of patients with TRD also suffer from MDSR. Alternatively, it appears that about 19.3% of patients with MDSR suffer from TRD. This information implies that the characterization of these patient populations should consider the prevalence of TRD without MDSR, the prevalence of MDSR without TRD, and the joint prevalence of TRD and MDSR. The aggregate results present a combined prevalence of Treatment-Resistant Depression and Major Depression with Suicide Risk of 11283 patients who simultaneously meet the criteria for both types of depression (7.9% of the total study patients).
Severity Level Distribution
The National Epidemiological Study on Mental Health (NESMH) contains information on the severity levels of the disease. The classification of patients by severity levels was based on the criteria adopted in the World Mental Health Survey (WMHS) (See the paper by Xavier et al). For patients with TRD and MDSR, the distribution estimated by levels of severity can be seen in Table 1.
Table 1 - Distribution of estimated TRD and MDSR cases by the level of severity
Severity level
|
TRD
|
MDSR
|
Severe cases
|
27.7%
|
64.4%
|
Moderate cases
|
60.7%
|
31.0%
|
Mild cases
|
11.6%
|
4.6%
|
TRD – Treatment-resistant depression; MDSR – Major Depression with suicide risk
Table 1 shows that the distribution by severity levels indicates a difference between TRD and MDSR, with the latter pathology having significantly higher levels. In the case of the joint prevalence of TRD and MDSR, it is assumed that the distribution by severity levels is that of the MDSR, as it is the most severe condition.
The Matter of Duration
An important component necessary to estimate the burden of TRD and MDSR is the duration of the episodes and, consequently, the fraction of the time affected by the disease in the reference year for the analysis. The distribution of the duration of episodes of major depression appears to be very heterogeneous. Spijker et al (2002) [27] retrospectively used data from the Dutch mental health survey to estimate the duration of an episode of major depression. They found that in 50% of cases the duration was three months or less, in 63% of cases was six months or less, in 76% of cases was 12 months or less, but in 20% of cases the duration was found to be 24 months or more. Spijker et al (2002) [27] reported that the median duration was three months, but the estimated average duration was 8.4 months. No specific references were found for the case of TRD and MDSR. In the present study, the assumption is made that the duration of TRD and MDSR can be approximated by the duration of episodes of major depression. The study by Ferrari et al (2013) [28], which synthesizes information on estimates of the various parameters necessary to calculate the disease burden, indicated that the average duration used in the Global Burden of Disease, resulting from a synthesis of the literature, was 37.7 weeks. The present study has used this value, i.e., it is assumed that depression affects the health of patients at an average of 72.3% of the time in the year in which an episode occurs.
Burden of Disease
The burden of disease was estimated through the disability-adjusted life years (DALYs). The most recent version of the methodology introduced by the World Bank and the World Health Organization (WHO) was adopted in this study [29].
DALYs are a measure, expressed in time, of the amount of health lost due to the disability generated by disease or premature death. The measure includes two indicators: 1) the years lost due to premature death (Years of Life Lost - YLL, in the graphs indicated as DALY due to death), the lost time being operationalized as the difference between age at the time of death and the standard life expectancy for that age; and, 2) the years lost due to disability (Years Lost to Disability - YLD, in the graphs indicated as DALY due to disability), where the time spent suffering a disability is considered [30].
The equation used to estimate the number of DALYs lost by an individual is as follows:
𝐷𝐴𝐿𝑌 (𝑐, 𝑠, 𝑎, 𝑡) = 𝑌𝐿𝐿 (𝑐, 𝑠, 𝑎, 𝑡) + 𝑌𝐿𝐷 (𝑐, 𝑠, 𝑎, 𝑡)
Where c is cause, s is sex, a is age, and t is time.
Disability is measured by a coefficient with values between 0 (without any disability, perfect health) and 1 (total disability or death). Standard life expectancy results from a reference mortality table designed to have universal applicability.
Years of Life Lost due to premature death -YLL
The years of life lost due to premature death (YLL) are calculated by multiplying the number of deaths caused by the disease under analysis and the years of life lost, which are a function of the age at which death occurs. In the case of depression, it is not usually taken as a direct cause of death. However, a significant fraction of suicide deaths can be statistically attributed to depression. In this context, YLL were estimated considering that a fraction of the overall suicide mortality is attributable to depression. Following the approach that uses the concept of attributable fraction [31], the fraction of total mortality attributable to MDSR was determined by the equation:
where RR is the Relative Risk of death in patients with the disease under study and p is the prevalence of the disease. The RR of suicide mortality in major depression considered by Ferrari et al (2013) [28] was 19.9. In the present study, TRD, and especially MDSR, would be expected to have higher RRs than those of major depression in general. In the literature, specific estimates were not found. The adopted methodology was to estimate the attributable fraction of suicides to major depression and, later, through the opinion of experts to obtain the proportion of this attributable fraction (AF) that applies to the two types of depressive pathology under study.
According to NESMH, the prevalence of major depression in the year prior to the survey interview was 6.8% [3]. The AF that results from the above equation is therefore 56.24%.
The next step was to estimate the proportion of these suicides attributable to the prevalence of TRD and MDSR. Our research assumed that, as in these data, suicide did occur, so ex post it is a case of MDSR. In practice, considering the diagnostic criteria for major depressive disorder in its different presentations, the possibility of some suicides occurring in patients who would not be diagnosed with MDSR, even having a diagnosis of major depression, should be considered. Thus, it is conservatively assumed that 90% of suicides attributable to major depression in general can be more specifically attributed to MDSR or TRD. Assuming 90%, the final attributable fraction is 50.6%. In summary, this is the percentage of the disease burden and costs generated by suicide deaths that will be attributed to DGRS and DRT in the present study.
Years of Life Lost due to Disability -YLD
DALY indicator, as a metric of disease burden, estimates, in addition to mortality, the disease burden generated by morbidity, considering that the time lived with a disease contributes to the years of life lost as that such a disease is disabling. The equation used to estimate the YLD number is as follows:
𝑌𝐿𝐷 (𝑐, 𝑠, 𝑎, 𝑡) = 𝑃 (𝑐, 𝑠, 𝑎, 𝑡) × 𝐷𝑊 (𝑐, 𝑠, 𝑎)
Where P is Prevalence of cause (c), by age (a) and sex (s), in year (t); and DW is Disability Weight specific to the cause (c), age (a) and sex (s).
The disease burden was estimated from the indicators: prevalence, mortality, disease duration.
The estimation of YLDs requires the use of disease-specific weighting or disability coefficients and is calibrated according to the different levels of disease severity. The most current version of the weights was published by Salomon et al (2015) [32] and the weights by severity level for the case of depression are shown in Table 1.
The use of these weights in the case of TRD and MDSR depends on the distribution of patients by severity levels. This information, from NESMH is shown in Table 1, and can be reviewed in columns (3) and (4) of Table 2. The average values of the weights are very high compared to other pathologies, probably because an expressive proportion of patients with major depression are not actually being treated, which in the case of MDSR is reinforced by the fact that even in the target patients of treatment they do not evaluate it as being effective.
Table 2 – Disability weights considered in the burden of disease estimates
Severity level
(1)
|
Disability weight
(2)
|
Proportion in the prevalence of TRD
(3)
|
Proportion in the prevalence of MDSR
(4)
|
Mild
|
0.1451
|
11.6%
|
4.6%
|
Moderate
|
0.396
|
60.7%
|
31%
|
Severe
|
0.658
|
27.7%
|
64.4%
|
Average weight
|
-
|
0.439
|
0.553
|
Sources: Salomon et al (2015) [32] regarding (2), and experts’ qualified opinions and author’s calculations to (3) and (4).
Costs of Illness
Direct Costs
The direct costs of TRD and MDSR resemble the monetary appreciation of the resources consumed in treating these diseases. The study of direct costs is based on the previously presented estimates of disease prevalence and the information on the pattern of use of resources contained in the “National Epidemiological Study of Mental Health” (NESMH) [3] and in the opinion of their experts. Microdata available in the 2017 Hospital Morbidity Database regarding inpatient and outpatient episodes registered using the International Classification of Diseases, version (ICD-10 CM) and billing in Homogeneous Diagnostic Groups (HDG), were considered. Microdata was used to estimate the number of relevant hospitalizations and outpatient episodes, as well as the respective costs. The study also used aggregated data on the consumption of drugs associated with the treatment of depression, from IQVIA and hmR, and expert opinions were used to estimate costs in areas where databases or other quantifiable sources of information are not known officially or academically recognized. Finally, in this study, the unit costs of hospitalizations, hospital consultations, and complementary means of diagnosis and therapy were obtained from the prices defined in Portuguese Law (Ordinance No. 207/2017, of 11 July).
Costs of Hospitalizations and ambulatory hospital visits
Estimates of hospital activity related to hospitalization episodes associated with TRD and MDSR are reported. The estimates are based on an analysis of the Hospital Morbidity Database for 2017. In this database, the use of ICD 10 hinders a finer analysis with the separation of TRD and MDSR. Thus, for greater accuracy, the hospitalization episodes associated with TRD and MDSR are presented together.
Registered and coded episodes are included in this database, including coded and registered hospital outpatient episodes. The identification of relevant episodes associated with TRD and MDSR was based on the International Classification of Diseases. The selection of relevant cases was made by the clinical team and experts. The episodes on which the subsequent analysis is based have all been classified with the ICD 10 CM. The use of the ICD 10 - CM classification was evaluated to guarantee the compatibility between the selected episodes and the diseases under study. The starting point was given by the ICD 10 - CM encodings. Major depressive pathology, single episode, and ICD 10 - CM. 0-9 Major, recurrent, depressive pathology. It was also necessary to add some episodes that were considered to be relevant and that were not part of the preliminary analysis. Thus, a set was added to the selected episodes in which the main diagnosis was “suicidal ideation” (ICD 10-CM R45851) provided that secondary diagnoses (from d2 to d50) were included or a sub-item of diagnoses F32 (Pathology major depression, single episode), or a sub-item of diagnoses F33 (major, recurrent depressive pathology). The use of episodes was further refined by considering additional information on the GDHs of the selected episodes.
Costs of pharmacological therapy
Estimates of drug costs in the treatment of TRD and MDSR are based on the intersection information from the EENSM regarding the consumption of medicines, information on the drug market from IQVIA and hmR, and finally, information on the drug market required selection and quantification criteria designed by experts.
The following classes of drugs were studied, following the terminology of the Anatomical Therapeutical Chemicals classification (ATC): N6A Antidepressants and Mood Stabilizers, N5A Antipsychotics, N5C Tranquilizers / Anxiolytics, and N5B Hypnotics / Sedatives.
Detailed data were obtained at the level of pills or equivalent since it was not possible to have access to market statistics specifying the quantities based on Defined Daily Doses (DDD). The available data made it necessary to make assumptions about the average consumption for each class of drugs. Average consumption patterns of one, two, or three tablets per day were chosen, depending on the class of medication.
Costs with complementary means of diagnosis and therapeutics
According to the experts, in the context of the usual follow-up, patients with TRD and MDSR tend to do routine tests twice a year. In addition to these consumptions, about 5% of patients undergo thyroid tests. Finally, about 5% of patients are submitted to imaging tests whose main objective is to overlook the possibility of other somatic pathologies. In two-thirds of the cases, the exam is a Computed Axial Tomography (CT) and, in the remaining cases, a Magnetic Resonance Imaging (MRI).
Based on data from NESMH, it is possible to obtain the proportion of patients who had any contact with the health system during the year prior to the study, both in mental health andprimary care. According to that information, 26.3% of the patients with TRD had some contact with mental health services, and 39% had some contact with primary health care services. Assuming that the contact probabilities in the two areas are independent, the probability of having at least one contact for patients with TRD is given by 1- (1-0.263) x (1-0.39) = 0.55. Consequently, this result will calibrate the estimates that follow, as it is assumed that only patients in contact with the health system generate consumption of complementary means of diagnosis and therapeutics. Specifically, for the case of patients with TRD (not including joint prevalence with MDSR), the following analysis assumes that the pattern of resource use of complementary means of diagnosis and therapeutics applies to 55% of patients.
In regard to patients with MDSR, estimates based on NESMH indicate that 43.2% will have some contact with mental health services and 57.7% with primary health care services. As in the previous case, independence of the probabilities of contact is assumed, resulting in an estimate of the percentage of patients with some type of contact with the health system of 76%. Weighting the two percentages by the proportion of patients with TRD only (55.5%) and patients with MDSR (44.5%), an average percentage of patients with contact of 64.4% is obtained. Applying this percentage to the prevalence in 143163 patients results in 92147 patients who generate a consumption of complementary means of diagnosis and therapeutics. It should be noted that the estimates presented treat the cases in which the patients have TRD and MDSR together as equivalent to those of the patients with the most severe situation, a methodology already adopted in other parts of this study.
We also considered the routine analyses that patients with TRD and MDSR would do twice a year, on average. A second set of analyses, related to the thyroid test, is carried out annually by about 5% of the patients. In addition, about 5% of patients undergo a CT scan (2/3 of the cases) or an MRI (remaining 1/3) to screen for other pathologies. In 50% of these cases, it is necessary to use contrast, increasing costs.
Costs of emergency department visits
According to the billing rules of the Portuguese NHS, episodes of urgency followed by hospitalization are integrated into hospitalization prices. It is then assumed that the Homogenous Diagnostic Groups (HDG) hospitalization prices are estimates of the overall costs of hospitalization, including the costs of the previous emergencies that generated these hospitalizations. For this reason, the costs of emergency department visits will estimate only the costs of emergency episodes without hospitalization.
Indirect Costs
Indirect costs result from the loss of productivity of patients and are defined in the present study as the value of production losses attributed to treatment-resistant depression and major depression with suicide risk. These may include absenteeism as short-term disability, premature exit from the labor market as long-term disability, and productivity lost by premature death.
The sources of information to identify these costs include academic literature, Portuguese databases, observational studies, and surveys conducted in Portugal. Other variables, such as the average wages by sex and age, will be estimated based on data from the 2017 Personnel Tables of the Portuguese Ministry of Labor, Solidarity, and Social Security.
Labor costs, given by gross wages and employers' social security contributions, are the best measure of the productivity of potential workers, following the Human Capital theory. The average salary, by gender and age group, is added by the employer's contribution to Social Security (23.75%). The resulting value is multiplied by 14 to obtain an estimate of annual productivity.
Long-Term Indirect Costs: Effects on Employment
The performed analysis takes into account employment until the age of 65 to consider a better approximation to the effective age of leaving the labor market. This effective age reflects that not all workers retire at the official retirement age, given the existence of multiple exceptions: the receipt of disability pensions, early retirements after long-term unemployment, and other situations of an early exit from the labor market.
The employment rates of the population with TRD were approximated by the employment rates of the population with Major Depression, and the employment rates in the population with MDSR were approximated by the employment rates in the group of people with suicidal ideation in the last 12 months. Following the principle of considering people with both types of depression have the most serious disease, joint cases of TRD and MDSR were included in the estimates for TRD, as it exhibits a greater impact on employment rates.
To monetize the lost production due to the lower employment levels, the Human Capital approach was employed, and lost production was approximated by the wage costs that workers would receive.
Short Term Indirect Costs: Absenteeism and Presenteeism
To estimate the daily productivity lost due to absenteeism, the average annual salary was divided by 230, corresponding to the number of working days per year, given that absenteeism, by definition, only occurs on these days.
The next step in estimating the indirect costs of absenteeism and presenteeism is to estimate the employment of patients with TRD and MDSR, which is achieved by combining estimates of disease prevalence and employment rates by gender, age group, and disease used in the previous section. For the reasons previously indicated, patients with TRD and patients with TRD and MDSR are linked together.
A viable way to identify the incremental effect of the diseases under analysis on absenteeism was to consider the difference between the days of absenteeism in the population with the diseases under study and the days of absenteeism in the general population.
There is no single convention on estimating the cost of presenteeism (see Mattke et al, 2007 [33]). The methodology adopted to estimate the cost of presenteeism assumed that one day of presenteeism has a weight of 0.25 days of absenteeism. Thus, an estimate is obtained for the total effect of each disease measured by additional equivalent days of absenteeism per year.
Indirect Costs of Premature Suicide Mortality
The indirect costs generated by suicide match the current value of all future production that would have been carried out by the deceased if he/she had survived. The updated rate used in this analysis is 4%, as outlined in the current guidelines for conducting health technology assessment studies in Portugal [34]. The convention of estimating future values of employment rates and wages by age and gender according to the statistics for 2017 was followed. The population's probabilities of survival are also used to calculate the expected value of future productivity. It is assumed that patients deceased due to suicide would have the survival probabilities given by the 2016-2018 Mortality Tables for men and women in general.