Bulletin of the World Health Organization

Schizophrenia treatment in the developing world: an interregional and multinational cost-effectiveness analysis

Dan Chisholm a, Oye Gureje b, Sandra Saldivia c, Marcelo Villalón Calderón d, Rajitha Wickremasinghe e, Nalaka Mendis f, Jose-Luis Ayuso-Mateos g, Shekhar Saxena h

Introduction

The chronic course and debilitating effects of schizophrenia combine to create a disease which imposes very considerable clinical, social and economic consequences on societies throughout the world, resulting in it being a leading contributor to global and regional levels of disability and the overall disease burden.1 The accumulated effectiveness and cost-effectiveness evidence regarding treatment responses to the burden of schizophrenia provides encouraging but hardly remarkable indications. Two Cochrane systematic reviews have clearly shown the superiority but limited acceptability of older antipsychotic drugs such as chlorpromazine or haloperidol over placebo,2,3 while the arrival of newer, “atypical” antipsychotic drugs has made available to patients and clinicians a set of pharmacological treatment options that are as effective and somewhat more tolerable than conventional neuroleptic drugs. There is also now an accumulation of evidence that assesses the cost-effectiveness of these drugs within randomized trials.4–7 There are also positive findings for the cost-effectiveness of family interventions to reduce the impact of family stress8 and of a short psycho-educational programme to improve patient adherence to medication.9 The cost-effectiveness of these interventions in developing regions of the world, however, is far less well established. The aim of the present work was therefore to provide a baseline assessment of cost-effectiveness for different regions of the developing world, using best available data on the epidemiological burden of schizophrenia, treatment costs and coverage, and the effectiveness of different interventions. To test the validity of these estimates, we subsequently carried out several country-specific studies in the WHO African, Americas and South-East Asia regions.10

Methods

Analytical framework

Through its CHOICE work programme (CHOosing Interventions that are Cost-Effective), WHO has developed a standardized form of cost-effectiveness analysis that aims to provide policy-makers with comparable results for interventions related to diseases or risk factors across the entire health sector.11,12 Initial application of WHO-CHOICE focused on the generation of cost-effectiveness databases at the level of epidemiologically-defined WHO subregions of the world.13–17 The existence of such information, however, provides no guarantee that findings and recommendations will actually change health policy or practice at the national level (where policies are determined and resources actually allocated). Current efforts are therefore focused on context-specific analyses at the national level.18 Such a process of contextualisation provides a powerful test of the validity and robustness of regional analyses. Specific efforts to contextualize regional results for schizophrenia and other psychiatric disorders at the national level are reported elsewhere.19–21 Here we present analytical inputs and results for three countries (Chile, Nigeria and Sri Lanka) together with those for their corresponding WHO subregion – Africa, Americas and South-East Asia. Summary results for other WHO subregions are available from the WHO-CHOICE web site (available at: http://www.who.int/choice).

Epidemiology and natural history

Schizophrenia was modelled as a severe, chronic disorder with a high level of disability, excess mortality from natural and unnatural causes, and a low rate of remission over the longer term (Fig. 1). The incidence, prevalence, remission and case-fatality of schizophrenia in WHO subregions has been estimated as part of the Global Burden of Disease study, based on a review of the relevant epidemiological literature (Table 1).22–24 The validity of this epidemiological disease model for schizophrenia has been evaluated at the country level in Spain.25 Since no recent population-wide survey data were available for schizophrenia in any of the three study countries, no revision was made to these regional values. Nevertheless, and in spite of the inherent variability in psychiatric diagnosis and classification, regional prevalence values are consistent with available data, including a recent prevalence study in Chile,26 a meta-analysis that covered low-income countries such as Nigeria27 and a study conducted some 30 years ago in Sri Lanka.28

Fig. 1. Population model
DW, disability weight; HSV, health state valuation (1-DW); x, disability weight for general population; y, disability weight for population with schizophrenia.

Estimating intervention effectiveness

Intervention analysis focused on efficacious and available treatments for schizophrenia, specifically “typical” (older) antipsychotic drugs (e.g. chlorpromazine, haloperidol) and “atypical” (newer) antipsychotic drugs (e.g. risperidone, olanzapine), alone or in combination with psychosocial treatment. Since first onset of schizophrenia is currently not preventable, observed incidence also represents the epidemiological situation that would prevail without intervention. Concerning remission and case-fatality, we find no substantive support for rates being modified by the specific effects of the treatments considered here (as opposed to, for example, spontaneous full remission or the potential contribution of a more supportive social environment), so these rates likewise remain constant for both treated and untreated scenarios. Accordingly, the relative effectiveness of these treatments was evaluated with respect to control of positive and negative symptoms together with associated levels of disability.

To calculate improvements in disability compared to (untreated) natural history, we adopt a recently developed approach in which treatment effect sizes reported in controlled trials are converted into an equivalent change in disability weight.29,30 Thirty general practitioners were asked to rate a series of vignettes, each adjacent vignette representing one standard deviation increased clinical severity for schizophrenia. The resulting conversion factor of 0.181 can be multiplied by effect sizes reported in the meta-analytic literature31,32 to give the corresponding change in disability weight (Table 2). At the country level, baseline effect sizes for psychosocial treatment were set somewhat higher (20%) than those reported in the international literature31 because of evidence for a more powerful impact of intervention in developing countries from two recent studies of treated schizophrenia in India.33,34

Estimating intervention costs

For the regional analysis, both a hospital-based and community-based outpatient service model were costed. At the national level, country teams focused only on the service model that had policy support for scaling-up (which in all cases was the community-based service model). For the regional analysis, annual expected resource requirements per “average” patient (with ranges denoting differing utilization for acute versus chronic cases) included: daily antipsychotic drug supply (e.g. 300–500 mg chlorpromazine) plus anticholinergic medication (where indicated) and laboratory procedures (1–4 tests); psychosocial intervention (6–12 sessions, where applicable); primary care (6–12 visits); outpatient attendances (20–50% of cases, monthly visits); day care (community model only: 20–50% of cases, 1–2 attendances per week); acute inpatient care (hospital-based model: 25–50% of cases, average length of stay: 21–56 days; community-based model: 10–50% of cases, average length of stay: 14–28 days); longer-term inpatient care (hospital model: 20–50%, length of stay: 90–180 days) or residential care (community model: 10–30%, length of stay: 90–180 days). Country-specific values concerning the frequency and intensity of health care uptake were based on available survey data (Chile) or on local expert opinion and a Delphi consensus panel survey of mental health professionals (Nigeria, Sri Lanka; a description of the Delphi study methodology is presented by C Ferri et al.).35

Resource items were multiplied by their respective unit costs to give an annual cost per treated case, expressed either in international dollars (regional analysis) or local currency units (national analysis). Region- and country-specific unit costs for primary and secondary care services, together with other default resource inputs such as salaries of health professionals, were estimated on the basis of regression analyses carried out on a large, multinational dataset that predicted cost as a function of gross domestic product per capita (plus other explanatory variables).36 Predicted costs at the country level, which were generally found to be robust when compared to available hospital data or pay scales, were replaced by local values wherever reliable data were available. Drug prices for the regional analysis were obtained from the International Drug Price Indicator Guide (available at: http://erc.msh.org/dmpguide) and from hospital pharmacies at the country level.

Results

Effectiveness of treatment

The population-level impact of first-line treatment of schizophrenia with older and newer antipsychotic drug therapies, alone or in combination with psychosocial intervention and case management, is shown in Table 3. At existing treatment coverage rates, which ranged from as low as 20% in Nigeria to as high as 80% in Chile, the current mix of single or combination therapies avert between 20–260 disability-adjusted life years (DALYs) per year per one million total population. By contrast, scaled-up implementation of the most effective treatments (at a target coverage level of 70–90%) is expected to avert between 230–575 DALYs per year per one million population, equivalent to 0.08–0.14 DALYs or 30–50 “disability-free days” per treated case per year. The incremental effect of combining psychosocial intervention with pharmacotherapy is estimated to be substantial (at least 65% more DALYs averted). Country results diverged from regional findings, not only because of different levels of treatment coverage but also due to higher rates of anticipated non-adherence to medication at the national level. In no instance, however, did the rank order of treatment effectiveness change.

Costs

Total costs encapsulating both the use of health-care services as well as associated training and administrative programme expenditures are also shown in Table 3. Costs relate to the annual cost (in millions of currency units) per one million total population, which equates to cost per capita. Only results for the community-based service model are shown here, but it is important to note that our regional analysis estimated that costs of the hospital-based service model exceed those of the community-based service model by approximately 33–50%, reflecting greater use of resource-intensive services such as acute and long-term psychiatric inpatient care.

In the regional analysis, the estimated treatment cost per capita for community-based provision of older antipsychotic drugs was I$ 0.74 (WHO African subregion D), I$ 2.10 (WHO South-East Asia subregion B) and I$ 3.13 (WHO Region of the Americas subregion B), equivalent to I$ 306, I$ 617 and I$ 980 per treated case, respectively; country contextualisation results produced lower figures of I$ 1.52 (Chile), I$ 0.39 (Nigeria) and I$ 0.57 (Sri Lanka). Interventions making use of newer (atypical) antipsychotic drugs (clozapine in Chile, risperidone in Nigeria and Sri Lanka), which in the regional analysis were estimated to be two to four times more costly than older drugs (I$ 3–6 per capita), were found to be lower than regional values in Chile and Sri Lanka (less than I$ 3 per capita), but very much higher than predicted in Nigeria (more than I$ 10 per capita). These results reflect the important influence of national drug procurement mechanisms and resulting supply prices. For example, at the time of the study, 2 mg risperidone could be obtained for 4.5 rupees (US$ 0.06) in Sri Lanka, compared to 255 Naira (US$ 2.50) in Nigeria, a 40-fold difference! Concerning adjuvant psychosocial treatment, additional costs including training were generally very modest on account of the relatively low salary levels prevailing in these regions and countries.

Cost-effectiveness

Relative to the natural history situation of no intervention for schizophrenia, combination interventions that are based on a community-based service model and which use older rather than newer antipsychotic drugs were found to be most cost-effective (Table 3). The cost per DALY averted (or healthy life year gained) for this intervention across the three WHO subregions ranged from I$ 2350 in WHO African subregion D to I$ 7158 in WHO Region of the Americas subregion B; at the national level, cost-effectiveness ratios ranged between I$ 1670 (Nigeria) and I$ 3400 (Chile). At the regional level, the cost-effectiveness of newer antipsychotic drugs implemented within a community-based service model without adjuvant psychosocial treatment was estimated to range between I$ 13 000 and I$ 20 000. Following contextualisation, the range widened substantially, from below I$ 9000 in Chile and Sri Lanka to more than I$ 80 000 in Nigeria. In all regions and countries, this represented the least efficient of the treatment strategies that were evaluated.

Uncertainty analysis

Deterministic uncertainty analysis showed that discounting the future stream of health gains by 3% per year (as done in the baseline analysis) has a relatively negligible impact on results, whereas removing age-weights – which give higher value to the middle years of life – significantly reduced the total estimated health gain (by 25%). In a second step, probabilistic uncertainty analysis showed the influence that stochastic variability in key drivers of cost and effect had on cost-effectiveness baseline results. For effectiveness, the conversion factor for translating effect sizes into disability weight change was the variable imbued with greatest uncertainty. For costs, drug prices together with average duration and the unit cost of a hospital inpatient stay were allowed to vary. Allowing for the correlation that exists between total costs and effects (range: 0.78–0.98), scatter plots with 1000 simulations were generated with the software programme MCLeague.37 Fig. 2 shows the results for Chile. When viewed in logarithmic terms (since total cost and effect variables were found to have a bivariate lognormal distribution), the uncertainty “clouds” for each intervention do not overlap, indicating that results are robust to plausible variability in these key input parameters. Greater uncertainty surrounds interventions that use newer, atypical antipsychotic drugs.

Fig. 2. Cloud graph showing stochastic uncertainty around costs and effects of schizophrenia interventions in Chile

Discussion

In light of the public health consequences of schizophrenia, this study set out to examine the cost-effectiveness of interventions capable of reducing its clinical and societal burden in the developing world. The purpose of such an exercise is to locate the relative position of schizophrenia interventions within a wider cost-effectiveness and priority-setting framework in the health sector. This is relevant because treatment of schizophrenia is all too often regarded as overly expensive or unaffordable.

Policy implications

There are several policy implications that arise from this attempt to generate economic evidence for the treatment of schizophrenia. First, it is important to note that the current situation is not the most efficient way to proceed. A similar conclusion has been drawn in the context of the Australian mental health system.29 Use of a “do-nothing” baseline comparator, as was implemented here, reveals that there are alternative strategies that represent a more cost-effective use of resources than the current mix (which is weighed down by the prevailing reliance on relatively high-cost institutional care). Second, all alternatives carry an additional cost to the current situation because they assume – on equity grounds – a higher level of coverage within the population in need. But that cost need not be excessive; for example, an extra 11 Naira or I$ 0.27 per capita would need to be invested each year to move from a treatment coverage of 20% to 70% in Nigeria.

Third, the least costly option is to restrict treatment to older antipsychotic drugs provided within a community-based service model. However, it makes better clinical and economical sense to go further and provide adjuvant psychosocial treatment to patients and families. For a modest extra cost there is a substantial additional impact on the disability or functioning levels of persons living with schizophrenia. In the three countries represented here, the cost per healthy year of life for this intervention is in the range of I$ 1700–3400, a value that falls below the international yardstick of average income per capita in all but the lowest-income countries.38

Fourth, and by contrast, switching first-line treatment to newer antipsychotic drugs has a very modest expected incremental impact on health outcomes but, depending on the price reached for these drugs, has a potentially ruinous effect on the financial feasibility of scaled-up provision of treatment (as would be the case in Nigeria, for example). This conclusion is in line with recent empirical research in Australia, the United Kingdom and the United States of America.7,8 The Assessing Cost-Effectiveness in Mental Health (ACE-MH) project in Australia39,40 also found an unfavourable level of cost-effectiveness for atypical antipsychotic drugs (an incremental cost of A$ 48 000–92 000 per DALY compared to conventional neuroleptics). However, when generic forms of these medications become more widely available, the picture could change dramatically with regard to cost-effectiveness, so our results should not be taken to imply that low-income countries should permanently exclude these newer medications from their public health systems.

Limits of economic modelling

The pursuit of a process of contextualization at country level represents an important test and validation of the model and results developed by WHO-CHOICE for schizophrenia, but there remain several concerns and limitations to modelling exercises of this kind. Most importantly, models can only be as good as the data that underlie them, and in this respect many of the epidemiological and efficacy input values used in this analysis still rely on international or regional estimates. There remains an ongoing need to enhance the empirical basis upon which estimation of the use, cost, outcome and impact of services for people with schizophrenia can be made at the national level. A further concern relates to the use of summary measures such as disability-adjusted life years as an outcome measure for schizophrenia treatment, since it is not as sensitive to clinical change as condition-specific measures, does not deal adequately with the issue of comorbidity (particularly substance abuse) and does not reflect the impact of treatment on carers or families. For the purpose of establishing the relative efficiency of schizophrenia care in relation to other investment possibilities in the health sector, however, the DALY is a suitable choice because it provides a direct link to disease burden estimates and enables results to be compared to the many studies of other health interventions using this metric in the particular context of developing countries.13–17

No attempt was made in the present analysis to measure costs or effects incurred outside the health system (such as time spent seeking or providing care by the patient and informal carer) or the non-health benefits of schizophrenia treatment including workforce and household productivity gains. Where such measurements have been attempted,34 a sharp decline in family burden and informal caregiving has been observed, which, in the context of a broader evaluative framework, would enhance the attractiveness of assessed interventions. Concerning employment outcomes, several targeted intervention schemes have been tested and found to have some success on labour force participation, but the potential economic benefits of such initiatives are not expected to be substantial.41

Other decision-making criteria

Although it is new and informative, evidence for the comparative cost-effectiveness of schizophrenia interventions in different regions or countries of the world provides only one input into the decision-making process. The ACE-MH project, for example, employed a set of “second stage” criteria – strength of evidence, equity, feasibility and acceptability – to qualify efficiency findings and promote the use of a broader set of criteria in setting priorities.39,40 Similar (unpublished) exercises were undertaken in Nigeria and Sri Lanka which indicated that equity considerations – which give more weight to interventions that promote wider access and target more severe conditions and younger age groups – were collectively given at least as much weight as the efficiency criterion of cost-effectiveness. While treatment of schizophrenia does not particularly benefit children, onset does typically occur in adolescence or the earlier years of adulthood; treatment addresses and alleviates a very severe and often long-term condition of health; and it can also alleviate the financial burden and associated impoverishment of families if it is made available to underserved populations.34

In conclusion, economic arguments against the widespread or scaled-up provision of treatment for persons with schizophrenia are fuelled by current inefficiencies (mainly related to disproportionate budgetary allocations to mental hospitals) and exacerbated by misinformation and stigma. These analyses suggest that efficient interventions are in fact reasonably cost-effective (against the international yardstick of average per capita income) and deserve particular consideration in terms of other key priority-setting criteria. ■

Acknowledgements

The authors thank the following contributors: Amala de Silva (Sri Lanka), Lola Kola (Nigeria), Benjamin Vicente (Chile), Jose-Miguel Caldas de Almeida (Pan-American Health Organization), Marge Reinap and Taavi Lai (Estonia), Pedro Gutierrez-Recacha (Spain) and Maria-Elena Medina-Mora (Mexico). We also thank Jeremy Lauer for technical support relating to the stochastic uncertainty analysis.

Funding: Jose-Luis Ayuso-Mateos is supported by the Instituto de Salud Carlos III (REM-TAP Network) of the Spanish Ministry of Health.

Competing interests: None declared.

References

Affiliations

  • Department of Health Systems Financing, World Health Organization, Geneva, Switzerland.
  • Department of Psychiatry, University College Hospital, Ibadan, Nigeria.
  • Department of Psychiatry and Mental Health, Universidad de Concepción, Concepción, Chile.
  • Faculty of Medicine, University of Chile, Santiago, Chile.
  • Department of Public Health, Faculty of Medicine, University of Kelaniya, Sri Lanka.
  • Department of Psychological Medicine, University of Colombo, Sri Lanka.
  • Department of Psychiatry, Hospital Universitario de la Princesa, Universidad Autonoma de Madrid, Spain.
  • Department of Mental Health and Substance Abuse, World Health Organization, Geneva, Switzerland.
Share
Email Twitter Facebook Google Delicious LinkedIn
More...