Abstract
Keywords
The Mood Disorders Unit (MDU) was established at Prince Henry Hospital in 1985, with clinicians most commonly referring patients with more severe, persistent and treatment-resistant conditions. Consultants generally make a series of treatment recommendations involving alternative or augmenting pharmacological approaches, non-pharmacological strategies (e.g. cognitive-behaviour therapy) or a course of electroconvulsive therapy (ECT). After initial assessment or treatment, patients are referred back to their clinician for ongoing management.
By late 1998, the MDU had assessed more than 1000 patients. Nuances of service components have been reported [1],[2] as has service satisfaction [3],[4], and we now report on cost issues. Economic analyses generally consider both direct and indirect costs. The former include costs of medical consultations, hospitalisation and investigations, while indirect costs include those associated with lost work productivity and any inability to maintain economic roles due to illness and associated disability.
We principally seek to examine how provision of the MDU service impacts on the ongoing direct costs of depression, with pre-service and post-service comparative analyses. As our consultants also assess primary referrals from general practitioners and take responsibility for area patients admitted with a depressive disorder, we also compare formal MDU referrals with this group which approximates more to generalist public hospital practice (our ‘controls’).
We collected data on costs in the 2 years preceding MDU assessment, aiming to avoid limiting pre-MDU cost assessment to an interval where high and protracted morbidity might lead to MDU referral and also artificially inflate costs. To prevent any spurious distortion driven by the MDU assessment/treatment process, we ignore the first 6 months following MDU assessment, and ‘cost’ only the second 6 months (albeit doubled to create an annual cost). Thus, we do not assess the cost implications of a 3-hour outpatient MDU assessment (or of any MDU hospitalisation) on an index episode. Instead, we examine the extent to which MDU assessment and management planning impacts on resulting costs, a more appropriate objective for a tertiary facility.
Method
A ‘cost questionnaire’ was embedded in an MDU intake assessment protocol previously described in this Journal [5], with patients recruited who had had a primary major depressive episode for less than 2 years. Sample members included formal tertiary MDU referrals (tertiary referrals or TRs) as well as patients referred to an MDU consultant's independent practice or routine hospitalised area patients admitted under the consultant's care (control referrals or CRs), with all patients invited to a 12-month follow up. The CRs had no contact with the MDU other than baseline and follow-up assessment.
A pre-assessment form sought self-reported details on treatment for the mood disorder over the 2 years preceding assessment, with each cost parameter (e.g. number of visits and average duration of visits to a practitioner) being listed in a standardised questionnaire. While our baseline assessment sought treatment details for both of those years, many patients had had their index episode occur in the 12 preceding months, and thus we principally analysed data for costs incurred only for that interval. Professional staff contact and hospitalisation details were limited to the depressive condition, be it for its direct impact or consequences such as a suicide attempt.
At follow up, patients completed a similar questionnaire for the preceding 6 months, as well as a self-estimate of the personal impact of the index episode on them. Specifically, a five-item questionnaire sought their judgement of the extent to which their current episode incurred (i) direct financial costs; (ii) indirect financial costs (e.g. loss of income); (iii) social costs (e.g. loss of friends/contacts); (iv) relationship costs (e.g. stress with intimates); and (v) personal costs (e.g. drop in self-esteem and self-confidence). Precoded options were ‘no’, ‘mild’, ‘moderate’, ‘severe’, ‘extreme’ and ‘catastrophic’. Clinical progress across the whole of the review year was assessed according to definitions devised by Frank et al. [6], and which provide operational criteria for remission (partial and full) and recovery, as well as relapse and recurrence.
While the sample was recruited over a 3-year interval, costs were calculated on November 1998 data. Costs for general practitioner and psychiatrist consultations, diagnostic tests and ECT treatment were based on fees as listed then in the Medicare Benefits Schedule book. We sought to derive ‘total societal’ costings. For instance, each cost/patient includes, where relevant, Medicare benefits, private health insurance rebates, hospital ‘overheads’, pharmaceutical benefits, dispensing fee, healthcare card holder benefits and any direct or ‘gap’ cost to the patient. Specific details on certain costings are now noted:
General practitioner and psychiatrist costing for ‘mood’ visits
Costing calculations were derived from the patient's estimate of (i) the average visit length to these practitioners; and (ii) the number of visits per prescribed period.
Other mental health professional costs
Consultation with hospital administrators and professional bodies (e.g. Australian Psychological Society) suggested the following costs: hospital-based clinical psychologist and social worker (base salary + 25%% overhead), $36/hour; hospital-based psychologist, $26/hour; private psychologist, $150/hour; and trained nurse (community-based), $25/hour. Such costs were calculated for those patients seeing the practitioners as out-patients or in community-based programs. If a patient had been hospitalised, staff costs were absorbed into the generic hospital cost.
Cost of pathology services
Patients were asked the number of blood test occasions for their mood state. We derived an ‘occasion’ cost of $78, made up of three representative general and specialised components (i.e. a full blood count, liver function tests and thyroid function tests).
Cost of diagnostic imaging services
Cost per patient consisted of the scheduled fee plus 25%% for hospital overheads. For a computerised tomography (CT) brain scan, the cost was then $252 and for a magnetic resonance image (MRI) scan, $594.
Cost of electroconvulsive therapy
Cost per treatment for ECT was calculated as including a fee of $53 for the consultant psychiatrist in attendance and of $72 for the anaesthetist, plus 25%% for hospital overheads. The total ECT cost (per defined period) was calculated from the total individual treatment cost ($157) multiplied by the number of treatments received.
Cost of antidepressant medication
Since selective serotonin re-uptake inhibitors (SSRIs) are now the commonest antidepressant type prescribed in Australia, this class was used as the costing basis, with a prescription cost of $51 (including government pharmaceutical benefit, healthcare card benefit and cost to patient). The cost per patient was calculated from prescription costs for the duration of the patient's taking the medication prior to baseline assessment. Regrettably, we did not collect appropriate data for the subsequent year.
Hospital costs
Hospital cost information was obtained from discussion with the New South Wales Department of Health, and from public and private hospital administrators. While estimates varied, costs did not appear to differ between private psychiatric units, public psychiatric units and general ward or casualty units. The cost of staying in any type of hospital was therefore derived from a range of estimates as $400/day.
Cost of social welfare
Social welfare costs were derived for those who were unemployed, on a sickness benefit or a disability pension due to their ‘mood’ state, and obtained from the University of New South Wales Social Policy Research Centre. Since we did not have data on what type of benefit patients were receiving, we calculated social welfare costs based on an estimate of unemployment or sickness benefit ($321/fortnight) and disability pension ($355/fortnight). Our estimate of $170/week is for a single person and does not take into account rent assistance or means testing for other income.
Results
Patients
The total baseline sample consisted of 270 patients with depression, of whom 204 were TRs and 66 CRs. For the whole sample, 172 (64%%) were female, with the average age being 43.3 (SD = 14.9) years, while 101 (37%%) were assessed as inpatients. Some 29%% were single, while 27%% were separated/divorced or widowed. Twenty-five per cent were in full-time work, 13%% were in part-time work, 14%% were involved in home duties or were students, 9%% had retired, 12%% were unemployed and 27%% were on benefits. Twenty-three (9%%) were diagnosed clinically as having a psychotic depression (PD), 85 (31%%) endogenous depression (ED), 94 (35%%) neurotic depression (ND) and 68 (25%%) reactive depression (RD). Complete costing data were available for all 270 for the 12 months prior to baseline assessment.
Depression details
For 80 (30%%), the current depressive episode was their first. Over their lifetimes, sample members had had a mean number of 14 depressive episodes and a mean duration of 102 weeks of significant mood disturbance and, for the 44%% of the sample who had taken time off work, a mean leave time of 41 weeks. Some 65%% had been hospitalised.
In the 12 months prior to MDU assessment, 24%% had been hospitalised in a public psychiatric hospital (mean duration 6.3 weeks), 14%% in a private psychiatric hospital (mean duration 8.1 weeks) and 14%% in a general hospital ward or casualty department (mean duration 6.5 days). Over that period, and for their mood state, 50%% had had to take time off work (for a mean duration of 20 weeks), 20%% had required social services, 70%% had visited a psychiatrist (with a mean of 17 visits) and 68%% had visited a general practitioner, 20%% had received a brain CT and 6%% an MRI, while 45%% had had diagnostic blood testing for their mood state.
Sample outcome at 12 months
Of the 182 (67%%) who accepted the 12-month follow up, the outcome of the index episode was as follows: 57%% had met criteria [6] for a ‘recovery’ (an euthymic state lasting more than 2 months) on average some 21 weeks after initial assessment, and 6%% met criteria for a ‘full remission’ (an euthymic state lasting less than 2 months). Of those recovering or remitting, some 23%% had a ‘relapse’ to a full major depressive syndrome (on average 23 weeks after baseline) and 10%% had a ‘recurrence’ or a new major depressive episode (on average 34 weeks past baseline).
Differences between those attending and not attending follow-up
Such comparisons of baseline variables identified only a few differences. Thus, those attending follow up had had fewer lifetime depressive episodes (11 vs 19, p < 0.05), scored as less likely to exhibit disordered personality function (on a range of measures, including fewer DSM-defined Cluster B personality traits) and were less likely to have ever used marijuana (30%% vs 49%%, p < 0.01), but did not differ significantly in frequency of other illicit drugs or cigarette use. The groups did not differ by sociodemographic variables, age of first depressive episode, severity of baseline episode or diagnostic subtype, TR or CR group representation (66%% vs 73%%) and, most importantly here, generated strikingly similar total costs in the 12 months prior to baseline assessment (t = 0.3).
Differences between ‘tertiary’ and ‘control’ referrals
The TRs and CRs did not differ significantly by age (43.9 vs 41.7 years), sex (63%% vs 67%% female), marital status (χ2 = 1.1), age at first depressive episode (30.9 vs 31.4 years), number of lifetime depressive episodes (14.6 vs 12.2), baseline depression severity on two measures (one self-report and one clinician-rated), or their chance of having taken any illicit drugs over their lifetime (40%% vs 35%%) or having smoked cigarettes (34%% vs 35%%). The groups did differ in employment status (χ2 = 13.8, df 4, p < 0.01), with the TRs more likely to be in receipt of social services. The diagnostic profile in the TRs and CRs, respectively, was 7.4%% versus 12.1%% PDs; 35.8%% versus 18.2%% EDs, 38.2%% versus 24.2%% NDs and 18.6%% versus 45.5%% RDs, with the RD representation identifying the greatest diagnostic difference across the samples. The TRs and CRs had similar rates of bipolar patients (11%% vs 8%%).
Theoretically, we would expect the TRs to have more persistent and treatment resistant conditions, and thus greater disability, engendering higher costs. Analyses established that they were less likely to be having their first depressive episode (i.e. 28%% vs 40%%, χ2 = 4.2, p < 0.05) and had a longer current episode (35 vs 27 weeks, t = 2.0, p < 0.05), but tended to be less likely to be inpatients (i.e. 36%% vs 42%%, not significant).
Comparison of mean costs for depression incurred (i.e. to patient, Commonwealth, insurance company, hospital) by consultant referrals (CR) and tertiary referral (TR) patients in the 12 months prior to baseline assessment
The unit cost per individual (UCPI) data enrich the picture. The TRs had significantly higher general practitioner costs, but a non-significant trend to higher psychiatrists' costs. Distinctions were most evident for hospitalisation costs (mean costs of $22 000 vs $7000 for CR for all hospital groupings), and with a higher percentage using each of the assessed hospital facility types. The UCPI differences were not distinct for diagnostic tests or treatment costs, or even social service costs, so that the greater overall mean UCPIs incurred by the TRs ($14000 vs $3000) were principally generated by hospitalisation costs.
Group total costs and user mean cost for depression incurred (i.e. to patient, Commonwealth, insurance company, hospital) over two annualised intervals for tertiary referrals (TRs)
Group total costs and user mean cost for depression incurred (i.e. to patient, Commonwealth, insurance company, hospital) over two annualised intervals for consultant referrals (CRs)
Comparison of utilisation data across the subsamples indicated that, in the 6 months prior to follow up, the two groups differed significantly on only one study variable, in that the TRs were significantly more likely (21%% vs 8%%, χ2 = 3.8, p < 0.05) to be in receipt of social services. The non-significant differences allowed the conclusion that the two subgroups had approximated following baseline assessment. The TRs and CRs did not differ in the rate of hospitalisation (14%% vs 10%%), length of any such hospitalisation (5.9 vs 7.2 weeks), casualty or general ward admission (4%% each, and 2.2 vs 4.0 days), taking time off work (24%% vs 27%%) or in its total duration (21 vs 13 weeks), duration of receipt of social services (24 vs 32 weeks), whether there was out-patient psychiatrist follow up (78%% vs 73%%) or number of visits (8.5 vs 5.1), whether the patient visited a general practitioner for their depression (34%% vs 27%%) or number of visits (6.2 vs 3.5) or whether the patient was currently being in contact with a psychiatrist (67%% vs 58%%).
Self-rated ‘costs’
Self-estimate cost questionnaire data and correlation of scores on each parameter with financial costs
Impact of diagnostic subtype
The influence of diagnostic subtyping was evident across a number of cost parameters for the whole sample. For example, it was a significant discriminator (F = 5.7, p < 0.001) of psychiatrist visit costs in the year before baseline assessment, with the average cost for the NDs being $2009, as against $868–923 for the other three (PD, ED and RD) groups, and with a similar pattern in the preceding year. In the year prior to baseline assessment, hospitalisation costs were highest for the PDs at $15 000, as against $9000 for the EDs, $5500 for the NDs and $4500 for the RDs. Electroconvulsive therapy costs were highest (F = 4.0, p < 0.01) in the PDs ($496), followed by the EDs ($265), NDs ($115) and RDs ($107). By contrast, antidepressant costs were highest in the EDs ($385), as against $337 in the NDs, $230 in the PDs and $197 in the RDs (F = 7.4, p < 0.001). The cost gradient over the 12-month interval was quite striking, with the UCPI being $17 326 for the PDs, $13 245 for the EDs, $10 950 for the NDs and $8238 for the RDs.
For each year prior to baseline assessment, the 28 bipolar patients consistently generated higher blood test and antidepressant costs than the remaining unipolar patients, and tended to generate higher hospitalisation costs ($2300 vs $1978 for the earlier pre-baseline period, and $11 100 vs $6861 for the second pre-baseline period). By contrast, after baseline assessment, their total costs were approximately half that of the unipolar patients, largely due to lower hospitalisation costs (i.e. $875 vs $2429).
Impact of other study variables
We examined a number of additional study variables as impacting on costs, across all three formalised intervals. Receipt of social welfare was a significant predictor of higher total costs across each interval. Patients with a history of self-injury generated higher costs over the earliest assessment period ($8906 vs $2665, p < 0.01) and post-baseline assessment ($7852 vs $3265, p < 0.05). Those with a history of illicit drug use or heavy alcohol use, or who had smoked cigarettes, generated slightly higher pre-MDU assessment costs (i.e. $5583 vs $4238) as well as post-MDU costs (i.e. $6465 vs $3549), but such differences were not significant. A longer lifetime duration of depression consistently predicted higher costs across each interval (r = 0.45, 0.16 and 0.32, respectively). Those rated as having personality dysfunction generated higher costs over the earliest assessment period (r = 0.24, p < 0.001) and across the post-baseline interval (r = 0.27, p < 0.001).
Predictors of costs in 2 years prior to baseline assessment
Such univariate analyses suggested a number of variables for several multivariate analyses. A multiple regression examined predictors of total costs over the 2 years prior to the baseline assessment. The refined set of significant variables (overall F = 12.6, p < 0.001) included patients having a longer lifetime duration of depression (beta weight of 0.22), having higher levels of disordered personality function (B = 0.22), being on social welfare in the 12 months prior to baseline assessment (B = 0.18), having a bipolar disorder (B = 0.14), receiving a clinical diagnosis of ED (B = 0.14), and not having a diagnosis of RD (B = −0.20). Repeating that analysis after excluding those patients whose initial depressive episode occurred only in the preceding previous year generated two significant predictors in the whole sample (F = 30.4, p < 0.001): being on social welfare and not receiving an RD diagnosis. The same analyses in subsamples identified differing predictors. Thus, for the CRs, predictors of higher total costs were: higher personality dysfunction scores, using marijuana and not receiving an RD diagnosis (F = 9.6, p < 0.001), and, for the TRs, predictors were being on social welfare, a longer lifetime duration of depression and receiving an ED diagnosis (F = 10.4, p < 0.01).
Predictors of cost charges after assessment
Equally importantly, we examined (using logistic regression analyses) for predictors of increased or decreased costs after baseline consultation; although, to prevent confounding, costs in the preceding 2-year period were not included. In the whole sample, the only predictor of higher costs was a diagnosis of ED (F = 4.5, p < 0.05). In the CRs, increased costs were predicted by a longer lifetime duration of depression, a higher personality dysfunction score, a diagnosis of ED and lower trait anxiety (F = 7.6, p < 0.001), while in the TRs, higher scores were predicted by being male and not having a diagnosis of RD.
Discussion
Apart from unpublished studies examining comparative cost-benefits of antidepressant drugs, there has been no previous attempt to cost depression and its treatment in Australasia. The Global Burden of Disease Report [7] has, however, had a major impact in creating awareness about the disability, and associated personal and financial costs, caused by depression.
While the present report has a number of obvious limitations (e.g. it examines select samples of depressed patients in relatively small numbers, and primary data relies on patient self-report details), it also has a number of strengths, and a key one is that we went to considerable trouble to detail a methodology for assessing the direct costs of depression, and particularly its management, in the local setting. Regrettably, we did not collect data on antidepressant costs for the post-baseline period, but their contribution to overall cost is likely to be minor, as they contributed less than 3%% in the TRs and 10%% in the CRs for the year preceding baseline assessment. While some unit costs will vary across services and time, we believe that we have identified relevant parameters, and have reported a wide range of analytic strategies, so that other services (and health planners) can replicate or modify such strategies to cost their service, although there is always the risk of quite erroneous conclusions being drawn from any data set. Some will be noted in this discussion.
We also proceeded beyond the common focus on financial costs by examining personal costs for patients. When asked to rate the impact of depression on their lives, the patients rated indirect financial costs as having more impact than direct ones, rated social and relationship costs even higher, but clearly put the ‘personal’ costs of depression highest. Thus, to have a depressive disorder is potentially depressogenic in and of itself. It would be useful if this finding were pursued in qualitative studies to determine to what degree such a high personal cost is driven by the illness itself, associated stigma and any other factors.
Across the whole sample we identified a range of individual factors that had substantial economic cost implications. The list includes obvious components (such as receipt of social welfare, hospitalisation and therapist costs) but also diagnostic type, ‘track record’ (e.g. previous hospitalisation, lifetime duration of depression), personality dysfunction, selfinjurious behaviours, illicit drug-taking and cigarette smoking. Such factors, as well as ones not measured (e.g. compliance with treatment and medication) all have substantial potential to impact on costs.
The last set of factors is not surprising to the clinician but may be misinterpreted in many cost studies. For example, several studies (e.g. [8]) have quantified greater disability in those with sub-syndromal depression than in those with major depression. Such a counter-intuitive finding may reflect a focus on measuring Axis I symptoms in such analyses. Those with sub-syndromal (and the so-called ‘minor’ depressions) may be more likely to have personalities and life styles that drive depressive symptoms and disability, generating economic costs which might be unfairly costed to ‘depression’ in such studies. Here, there was evidence that a dysfunctional personality style (as measured directly and indirectly from variables such as self-injurious behaviours) does contribute to higher costs in those with depression, but the nature of this contribution needs to be conceptualised more clearly in cost and disability studies.
Diagnostic type produced particularly interesting findings about costs. A diagnosis of RD was significant in many analyses, indicating that those with a precipitant-weighted or adjustment disorder generated comparatively low costs. By contrast, those receiving a clinical diagnosis of ND (and where it might be assumed that anxiety and/or personality style was contributing to a depressive syndrome) generated the highest out-patient psychiatrist costs prior to assessment. In addition, as they had the lowest rates of recovery, they tended to generate the highest hospitalisation and social welfare costs longitudinally. By contrast, those with the more ‘biological’ depressive subtypes (i.e. PD and ED) generated the highest hospitalisation, investigatory and ECT costs prior to assessment, but comparatively lower costs when followed longitudinally. This might indicate that, despite the greater severity, associated disability and greater need for investigation and hospitalisation for those with these illnesses, active treatment for those with psychotic and melancholic depression is distinctly cost beneficial. The same pattern appeared to hold for bipolar patients (who generated higher costs than unipolar patients prior to assessment but subsequently lower ones), and again could indicate the differentially superior benefits of active treatment.
While one focus was on our tertiary MDU referral patients, we included a comparison group. However, as it was a small group and constituted both by outpatients referred to our consultants and by a significant percentage of hospitalised area patients, it cannot be regarded as representative. Clearly, a comparison group treated by a different service or set of psychiatrists might generate quite different costs.
For these CRs, as for our TRs, we can conclude that: (i) therapist, investigatory and antidepressant medication costs are relatively inexpensive in comparison with hospitalisation, ECT and social welfare costs; and (ii) unit costs are not of great importance in and of themselves, as total group costs are more influenced by the prevalence of the individual components and their duration. Thus, a course of ECT was expensive but received only by one patient prior to and following baseline assessment. By contrast, rare cost sources for brief periods (e.g. hospitalisation) or extending over time (e.g. social welfare), or common cost sources (e.g. visiting a psychiatrist, receiving an antidepressant medication) contribute strongly to group costs.
Against expectation, and in comparison with the TRs, annualised costs increased rather than decreased in the CRs. This is likely to reflect two factors. First, 42%% were having their first depressive episode and might be expected to then have low annual costs prior to baseline assessment. Second, and perhaps of greater salience, treatment generates costs. The increased costs here did not relate to health practitioner costs (with a distinct drop in the percentage being so treated by a general practitioner, and a modest increase in the percentage being treated by a psychiatrist), but by distinct increases in hospitalisation, ECT and social welfare costs.
Against this backdrop we now focus on our key interest: the extent to which the MDU assessment process may have impacted on costs. For any episodic condition, it can be relatively easy to demonstrate (albeit spuriously) that a service is cost-effective. Migraine offers such an example, with patients often having episode epochs which encourage seeking assistance (and thus generating costs). Subsequent improvement to a quiescent interval may reflect the impact of any intervention or, alternatively, may reflect no more than a ‘regression to the mean’ phenomenon or a spontaneous improvement in illness pattern. The latter will reduce ‘costs’, and the intervention service might then claim that the treatment was cost-effective. To reduce that possibility and because we sought to focus on any reduction in ongoing (and disability-weighted) costs, we collected data for the 2 preceding years and ignored the 6-month interval following baseline assessment.
Overall group costs in our MDU-assessed TRs were reduced (despite increased costs for ECT and social welfare) by nearly 40 per cent, reflecting some reduction in general practitioner and psychiatrist costs, but most distinctly by a reduction in hospitalisation costs. Despite increased hospitalisation costs for those admitted, the percentage hospitalised dropped from 45%% to 15%%, resulting in a saving of more than $700 000. The extent to which changes in use of practitioners, facilities and services reflects disorder resolution occurring ‘naturally’ or as a direct consequence of the MDU management recommendations and their implementation by referring therapists cannot be answered, revealing a further limitation to such costing studies. To show true cost benefit, the MDU service would need to demonstrate that its ‘intervention’ generated lower subsequent costs than no intervention.
Importantly, however, MDU assessment was not associated with cost increases. We suspect, and have some data to support, that it may have most distinctly improved the outcome trajectory of those with the more ‘biological’ depressive disorders (i.e. bipolar disorder, psychotic and melancholic depression), presumably achieved by review and modification of pharmacological treatments, recommendations for ECT for some, and attention to second-order factors through pointers to treatments such as cognitive-behaviour therapy and strategies such as anxiety management.
Effective services can increase, maintain or decrease costs, as has been reported following the introduction of case management models. Evaluation clearly should be weighted to the objectives of any clinical service. However, to the extent that it is preferable to have a tertiary referral service assessment associated with a lowering of the cost trajectory, the MDU appears to meet that criterion.
Footnotes
Acknowledgements
The authors wish to acknowledge Dusan Hadzi-Pavlovic, Marie-Paule Austin, Heather Brotchie, Yvonne Foy, Ian Hickie and Christine Taylor for study assistance, an NH&MRC Program Grant (993208) and an Infrastructure Grant from the New South Wales Department of Health for funding support.