Scientific Papers

Multimorbidity healthcare expenditure in Belgium: a 4-year analysis (COMORB study) | Health Research Policy and Systems

Description of Image

Summary of findings

This study provides insights into numerous multimorbidity profiles in Belgium. In addition to examining the cost by morbidity counts as well as demographic and socioeconomic subgroups, the study explored the costs of a large number of dyads (171) and triads (969) associated with the 19 most prevalent chronic conditions. People with multimorbidity constituted nearly half of the studied population and their total healthcare cost constituted around three quarters of the healthcare cost of the studied population. The cost of multimorbidity increased with age and morbidity count, and individuals with lower socioeconomic status were more prone to higher healthcare costs than those of higher socioeconomic status. The most common dyad, arthropathies + dorsopathies, with a prevalence rate of 14%, accounted for 11% of the total national health expenditure. The most frequent triad, arthropathies + dorsopathies + hypertension, with prevalence rate 5%, contributed 5%. Prevalent morbidity combinations, rather than high-cost ones, made a greater contribution to total national health expenditure. The average annual direct cost per person with dyad was €3515 (95% CI 3093–3937), while the average annual direct cost per person with triad was €4592 (95% CI 3920–5264). Dyads and triads associated with cancer, diabetes, chronic fatigue, and genitourinary problems had the highest costs. In most cases, the cost associated with an individual with multimorbidity was lower or not substantially different from the combined cost of the same conditions observed in separate patients.

General trend of multimorbidity and cost

In 2018, multimorbidity affected approximately 48% of the population aged 15+ in Belgium. This rate surpassed the reported global multimorbidity prevalence of 37.2% (95% CI 34.9–39.4) in the community setting, as well as the regional rate for Europe of 39.2% (95% CI 33.2–45.2) [36]. However, it is important to approach these comparative statements with caution. The presence of heterogeneity in study methodologies, sample selections, data collection approaches, definitions of multimorbidity and the scope of included chronic conditions introduce challenges in directly comparing prevalence rates of multimorbidity.

The five most common chronic conditions in dyads and triads were arthropathies, dorsopathies, hypertension, high cholesterol, and allergy. Several results are consistent with those from the United States, England and France [37,38,39]. In the United States, arthritis, high cholesterol, and hypertension were also the most common in multimorbidity, alongside diabetes [37, 38]. Hypertension was reported as the most common condition in morbidity combinations in England, alongside diabetes, chronic kidney disease, and asthma [17]. However, chronic kidney disease had low prevalence in Belgium. Diabetes and asthma (as part of the respiratory disease group) were also common, but secondary to those mentioned.

This study supports previous evidence indicating that the cost of multimorbidity increases with age and morbidity count [40,41,42,43,44,45]. However, after adjusting for morbidity counts, adding age to the analysis did not significantly improve the ability to explain the variation in costs [46]. Nevertheless, individuals aged 70+ who were affected by multimorbidity played a significant role in terms of healthcare costs. Despite constituting a smaller portion of the total population, this group accounted for a substantial share of healthcare expenditures. This underscores the importance of addressing the unique healthcare needs and challenges faced by the elderly population, particularly those dealing with multiple health conditions simultaneously.

Further, our study confirms that multimorbid individuals of lower socioeconomic status had higher healthcare cost compared with those of higher status. Similar research in the United Kingdom also showed that healthcare cost increased with greater levels of deprivation; conversely, individuals from higher socioeconomic backgrounds were more likely to experience better health, leading to lower care needs [47,48,49].

Cost of multimorbidity in Belgium in comparison with other countries

In comparison with the few high-income countries for which data are available, the average cost of multimorbidity per person in Belgium appears to be lower, with dyads and triads costing an average of €3515 and €4592 per person per year, respectively. In England, the average annual cost per person with dyad was €5013 (£3717) and with triad was €7116 (£5276), but this only included secondary care costs and excluded primary care and pharmaceutical expenses – the actual cost may be much higher [17]. In the United States, the median costs of dyads and triads were €6751 ($6208) and €8892 ($8177), respectively [37, 50,51,52]. The most expensive dyad in our study was cancer + chronic fatigue, with an estimated cost of €8345; which was lower than the estimated €11,381 for cancer + neurological diseases in New Zealand [53]. The most prevalent dyad in our study, arthropathies + dorsopathies, cost €3044 – lower than the cost of treating osteoarthritis + back pain in Sweden (€5358) [54]. One of the most prevalent triads in our study was dorsopathies + high cholesterol + hypertension (€1824), the cost of which was significantly lower than the estimated cost of treating low back pain + hypertension + hyperlipidemia (€22,906) in the United States [55]. Comparing our findings with those of other studies is challenging because few other studies have a comparable scope. Further, it should be noted that comparing costs between countries is difficult due to variations in the disease burden, methodology, data collection, sample representativeness and differences in healthcare systems [2]. Despite this, on average, the cost of multimorbidity in Belgium was found to be notably lower than that of other countries with similar economic contexts.

In our study, the dyads associated with the highest costs predominantly consisted of cancer, diabetes, chronic fatigue, and genitourinary problems. These results are, in part, consistent with previous studies that have identified cancer and diabetes, as standalone conditions, associated with high costs, and dyads/triads that included these conditions were also costly to treat [2, 53, 56,57,58]. However, the reasons why chronic fatigue and genitourinary problems frequently appeared in the top most expensive dyads/triads are less obvious. Further investigation into literature found that chronic fatigue is a complex chronic illness that causes widespread pain, cognitive impairment, can incapacitate individuals for a long period of time, and with a poor prognosis [59]. Diagnosis relies on assessing patient-reported symptoms and extensive testing to exclude other illnesses or factors, as there is no specific laboratory-based diagnostic test [59]. Extensive testing and the impact on quality of life leading to possible homecare service use, to some extent, explain the high cost of chronic fatigue. Genitourinary problems encompass a broad range of conditions, including urinary incontinence, kidney stones, chronic cystitis and prostate problems. Studies have indicated that patients with genitourinary problems not only incurred high healthcare cost for testing and treatment, but also for behavioural therapy, devices and routine care items [60, 61].

In the low-cost group, common chronic conditions within the combinations were allergy, stomach ulcer, and osteoporosis. Allergy is often excluded from multimorbidity studies, making it difficult to compare the cost of combinations involving allergy across countries. As the health outcomes used in this study was self-reported by patients, it was uncertain whether the person received a clinical diagnosis and whether healthcare was sought. Furthermore, for many types of allergies, avoidance of the suspected allergen is the only treatment, and healthcare is typically only sought in the event of a reaction [62, 63]. This may explain the relatively low healthcare cost associated with allergy. Regarding stomach ulcers, complications are uncommon and most cases are treated with pharmacotherapeutics [64, 65]; thus it is reasonable that healthcare costs are relatively low. Regarding osteoporosis, a study conducted in Belgium in 2004 reported having osteoporosis cost €535 per person per year and the author recognised that this figure is low, possibly because osteoporosis remains under-treated in Belgium [66]. Hence, factors such as underreporting, limited healthcare utilisation and cost-effective management contribute partially to the lower costs observed in certain dyads and triads.

The importance of interaction effects in multimorbidity costing

To ensure precise estimation in multimorbidity costing studies, it is essential to consider interaction effects to avoid the potential risks of overestimating or underestimating cost. Few studies have explored the interaction of conditions and its effect on cost [15, 39, 67]. In our study, the cost of most dyads and triads did not differ significantly from the summed cost of the same conditions existing in different individuals. For those that differed significantly, more sub-additive effects were observed in dyads and super-additive effects in triads. For dyads, one out of two pairs with super-additive interaction effects was discordant and one out of four with sub-additive effects was concordant. For triads, four out of five triads with super-additive interaction effects were discordant and one out of three with sub-additive effects was concordant. Concordant conditions share similar pathophysiologic risks or disease management plans, while discordant conditions are those with unrelated/indirectly related pathophysiologic risks and disease management plans [68, 69]. A third type – dominant conditions – are severe conditions that may limit life expectancy or require extensive medical treatment [68, 69]. Based on this premise, the majority of our results are reasonable as they align with this classification. For example, having eye disease + hypertension (discordant) or dorsopathies + genitourinary problems + high cholesterol (discordant) increased spending, while having diabetes + hypertension (concordant) or arthropathies + dorsopathies + osteoporosis (concordant) reduced spending. Although this does not enable us to explain all of our findings, it serves as a foundation for investigating the healthcare-seeking behaviour of individuals with dyads or triads that displayed super- or sub-additive healthcare expenditure. More frequent utilisation, complex disease trajectories or complications, polypharmacy and inadequate coordination between services are potential explanations why some dyads/triads resulted in super-additive healthcare spending [7, 70, 71].

However, our study showed that most dyads and triads resulted in lower or comparable healthcare costs to the summed cost of the same conditions in different individuals. Although this could be perceived as positive news from an economic perspective, it could also indicate that patients with multimorbidity are “backgrounding” one condition for another and may not be receiving sufficient care for all their conditions. This phenomenon aligns with the Shifting Perspectives Model of Chronic Illness that suggests people living with multimorbidity may place illness in the foreground or the background of their “world”, depending on the context [72].

Implications for future research, health system and policy

This study assessed morbidity combinations that are most expensive and/or prevalent, which can help to identify where cost savings can be achieved through care reorganisation and prevention. For those with super-additive health expenditure, further research can be conducted into the level of care integration and health seeking behaviour. For concordant diseases, it may be more efficient to seek efficiency gains; for instance, by appointing the concordant disease management to the same medical doctor, resulting in time/cost saving and more effective communication [73]. For common concordant conditions, the first line could be the appropriate level of daily care, with a transmural component of annual visits to a medical specialist [8]. Such models are already in place for single disease care pathways [74]. For those with sub-additive spending, further research can be conducted to understand their health service utilisation patterns. The results can serve as a case study for best practices or identify whether the patient is having unmet needs.

To support decision-makers and researchers to predict and monitor the costs of morbidity combinations, a multimorbidity costing tool can be developed, embedding the models from this study to provide a user-friendly platform that can automatically generate the costs of over a thousand morbidity clusters with fine-tuneable parameters, based on the user’s interest. The potential applications of this tool are extensive, ranging from policy-makers and practitioners to insurance companies, patients and families. Its potential impact in informing health policy and decision-making processes cannot be overstated.

Strengths and limitations

This study represents the first of its kind in Belgium, at a time when population-level studies on the cost of multimorbidity remain scarce in Europe and globally [2]. By including a large number of morbidity combinations, accounting for both dyads and triads, our study provides a comprehensive assessment of the cost of multimorbidity. While the list of 19 chronic conditions included in the study is not exhaustive, it encompasses the conditions most prevalent in the population, satisfying the criteria of including at least 12 chronic conditions suggested by Fortin et al. for an accurate measurement of multimorbidity [75].

Our use of linked longitudinal data from an exhaustive claim database, a reliable source of information for studying the cost of chronic conditions, increases the reliability of our findings [15, 46, 76]. Notably, linked data is still not widely utilised in research on the cost of multimorbidity, highlighting the novelty of our approach [2]. Furthermore, the use of health insurance claim databases has been endorsed for conducting cost-of-illness studies, further adding to the robustness of our study [77, 78].

Our sample size is relatively large and representative of the population, providing a strong basis for generalisation of findings. We took into account interaction effects, providing more accurate estimations and avoiding the risk of over-/underestimating costs.

There are several limitations and challenges that should be considered when interpreting these findings. Conducting multimorbidity research is inherently challenging due to the vast number of possible morbidity combinations and scenarios. This is further complicated by computational limitations and the need to strike a balance between model fit and parsimony.

The health outcomes were derived from a cross-sectional design survey, which may be limited by patients’ subjective reporting of their health status, unclear diagnoses, overlapping symptoms and other factors that can affect the accuracy of the reported data. For practical reasons, we also assumed that the patients had had the same chronic conditions across all four years, and excluded certain chronic conditions with low prevalence (stroke, cirrhosis of the liver, kidney diseases, Parkinson’s disease, hip fracture, and gallstones) to increase the robustness of our findings. However, this may have led to an overestimation of costs. Regarding covariates, data on proximity to death was insufficient for inclusion, despite its recognised significance as an explanatory factor for healthcare costs, even more so than age [79].

Additionally, diverse methodologies exist for the selection, inclusion and classification of chronic conditions. The list of conditions incorporated in our study was formulated after numerous deliberations within the team, acknowledging that alternative approaches may also exist. For instance, some conditions could be classified as symptoms or risk factors, rather than standalone chronic conditions. While cancer represents a condition characterised by clear and comprehensive clinical manifestations, high cholesterol and hypertension, on the other hand, are risk factors predisposing individuals to the development of future diseases. Chronic fatigue remains a topic of debate within medical circles due to its association with numerous other conditions, presenting varied levels of reporting and diagnostic confirmation. The conditions included in the study exhibit a diverse spectrum in how they manifest clinically, their impact on health outcomes and the diverse physiological pathways underlying each condition. This is inevitable given the complex nature of chronic conditions and multimorbidity, and the lack of a consensus on definitions and terminologies [3, 75, 80]. Moreover, some conditions were collapsed to form broader morbidity groups (Fig. 1) and there may be potential overlaps across groups. The limited number of included conditions could potentially have resulted in an underestimated prevalence of multimorbidity. Nonetheless, our study aimed to capture all “available” conditions from the database, particularly those that could impact a person’s health-related quality of life and incur healthcare expenditure over a long period. Indirect costs, estimated using a different data source, were not presented here but in a separate study.

Finally, although beyond the scope of our study, we find it important to acknowledge the potential impact of COVID-19 on our results. Upon examining our data and consulting official figures from the Ministry of Social Security, we observed only a marginal increase in total healthcare expenditure in 2020 compared with 2019. While COVID-19 may have influenced healthcare expenditure in 2020, our analysis spanned multiple years, suggesting that any effect is likely minimal and largely irrelevant for the purpose of this analysis and the insights it provides. Other minor limitations were the reporting of cost per average year and the assumption that different multimorbidity scenarios were equally affected by any COVID-19 impact over the relatively short time window.

Description of Image

Source link