Abstract
Background
COVID-19 infection and hospitalization have been associated in literature with increased post-acute healthcare expenditure, but social inequalities in this association remain largely unstudied. Therefore, this study aimed to investigate post-acute healthcare expenditure among patients hospitalized with and without COVID-19 and to identify socioeconomic and sociodemographic factors associated with increased post-acute health costs.
Methods
A matched cohort of 10,380 patients with COVID-19 and 26,270 patients without COVID-19 hospitalized in Belgium between 14 September and 31 December 2020 was created. A health systems approach was adopted and the direct individual post-acute healthcare expenditure in the calendar year 2021 in the context of the Belgian compulsory health insurance was available from the Intermutualistic Agency. The association of socioeconomic and sociodemographic variables with post-acute healthcare costs was investigated using multivariable generalized linear models with a negative binomial distribution and a log-link function.
Results
COVID-19 hospitalization was associated with 18.5% (95% confidence interval: 16.9% to 20.1%) lower healthcare costs in the calendar year following hospitalization compared to non-COVID hospitalization, because patients without COVID-19 during the pandemic might have been admitted to the hospital for more severe health conditions. Lower education level, lower income, living in a collective household, and older age were associated with higher post-acute healthcare costs among COVID-19 patients, whereas COVID-19 patients having a migration background experienced lower healthcare costs.
Conclusions
This study identified different socioeconomic and sociodemographic characteristics associated with increased post-acute healthcare expenditure among COVID-19 patients, emphasizing the vulnerability of hospitalized COVID-19 patients with lower social status.
Keywords: COVID-19, Hospitalization, Healthcare expenditure, Health costs, Social inequalities, Socioeconomic status
Background
The clinical presentation of coronavirus disease 2019 (COVID-19) following infection by the SARS-CoV-2 virus varies widely. It ranges from asymptomatic cases, over mild cases experiencing a fever and a cough, to severe cases with respiratory failure requiring hospital admission or resulting in death [1]. Furthermore, about 45% of COVID-19 patients do not return to their baseline health state within three months following infection and experience post-acute sequelae of COVID-19 (PASC) [2, 3]. This condition is characterized by persistent or new symptoms (such as fatigue, shortness of breath, or cognitive dysfunction) following SARS-CoV-2 infection, usually three months from onset, lasting for at least two months, without alternative diagnosis explaining the symptoms [4, 5].
In Belgium, considerable socioeconomic (SE) and sociodemographic (SD) inequalities have been observed in COVID-19 preventive behavior [6, 7], SARS-CoV-2 infection [8], COVID-19 hospitalization [9], PASC [3], and COVID-19-specific mortality [10]. Additionally, socially disadvantaged groups experience higher rates of various comorbidities, including hypertension, diabetes, and asthma [11]. A syndemic exists between COVID-19 and several comorbidities, as they cluster in socially vulnerable populations and negatively interact with each other, resulting in worse health outcomes [12–16]. Therefore, the COVID-19 pandemic is characterized as a “syndemic pandemic” which exacerbates existing social health inequalities.
Social inequalities also persist in healthcare expenditure and utilization in the general Belgian population. Lower household education level and income are associated with higher average annual health costs [17]. Nevertheless, these groups also report the highest rates of unmet medical needs and delay in seeking healthcare [18, 19]. Additionally, disparities in healthcare utilization by education level and income group remain, even after adjustment for health needs; individuals with lower education level or lower income more often visit the general practitioner, but are less likely to access preventive care or specialist consultations [20].
International research shows that individual healthcare expenditure increased not only during acute COVID-19 infection, but also during the first months following infection, both when compared to their individual health costs before the infection [21–24] or to a control group without COVID-19 infection [25–33]. Additionally, some studies have explored differences by age, sex, comorbidities, vaccination state, and severity of COVID-19 infection, suggesting that older age, female sex, the presence of comorbidities, absence of vaccination, and higher severity of infection further increase post-acute healthcare costs [21–23, 27, 30]. However, the association between SE and SD characteristics and post-acute healthcare costs following COVID-19 remains largely unstudied.
In this study, we focus on COVID-19 hospitalized patients in Belgium as captured in the national COVID-19 surveillance system (Clinical Hospital Surveillance, CHS). Our primary objectives were (i) to assess how COVID-19 hospitalization affects post-acute healthcare expenditure during the first year following hospitalization, relative to non-COVID-19 hospitalization, and (ii) to identify SE and SD factors associated with increased post-acute healthcare expenditure among COVID-19 hospitalized patients, and how these patterns may differ from those observed in non-COVID-19 patients. To investigate these objectives, we examined the direct individual healthcare costs in the calendar year 2021 among patients hospitalized with and without COVID-19 during the last trimester of 2020 in Belgium. The present research explores, for the first time, post-acute healthcare expenditures following COVID-19 hospitalization in Belgium, providing insights into how socioeconomic factors, pre-existing disease burden, and pandemic-related pressures shaped care, while contextualizing these patterns alongside non-COVID-19 hospitalizations.
Methods
Data sources
This research fits within the HELICON project, investigating the indirect and long-term effects of COVID-19 and social inequalities therein in Belgium [34]. A cohort of patients hospitalized with and without COVID-19 during the last trimester of 2020 in Belgium was selected from the HELICON data linkage [35]. This data linkage combined data from three Belgian national administrative databases at the individual level using a pseudonymized version of the Belgian social security identification number (SSIN). The first employed data source was the Clinical Hospital Surveillance (CHS), which collected individual-level data on hospital stays of patients hospitalized with a lab-confirmed COVID-19 infection across all 103 Belgian hospitals. Hospital staff completed online forms at admission and discharge of the COVID-19 patients [36, 37]. Secondly, Statistics Belgium (Statbel) provided information on SE (education level, income, and employment status) and SD (migration background, household type, age, sex, and region) factors for all identified hospitalized COVID-19 patients and a 10% random sample of the Belgian population from which patients hospitalized without COVID-19 were selected. Statbel derived data from multiple registers; SD indicators were obtained from the national registry, whereas the income was obtained from the tax registry, and education level and employment status were extracted from different regional, national, and international databases using algorithms designed by Statbel. All SE and SD variables reflected the pre-pandemic situation, i.e. the situation on the 1st of January 2020 or before. Furthermore, Statbel provided mortality and emigration data up to the 1st of January 2022 to be able to exclude deceased or emigrated patients. An overview of the Statbel variables, their sources, and reference dates are available in Table 1. Lastly, the Intermutualistic Agency (IMA) summarized individual healthcare utilization and costs in the context of the compulsory Belgian health insurance, which covers more than 99% of the Belgian population. Lower coverage rates are observed in Brussels, among males, young adults, and Belgians living abroad. Moreover, some highly vulnerable populations, such as undocumented migrants or asylum seekers, are not covered but were also not included in the definition of the Belgian population used for this calculation [20]. IMA created two datasets; the patient dataset summarized healthcare costs and utilization per patient over one calendar year, whereas the hospital dataset recorded all different hospital stays (COVID and non-COVID) per patient. For both datasets, a 2020 and a 2021 version was obtained. The HELICON data linkage falls under Article 9§ 2(j) of the General Data Protection Regulation (GDPR) and therefore, no informed consent was obtained from the patients [38, 39]. Formal approval for the linkage was granted by the Information Security Committee on Social Affaires and Health [40].
Table 1.
Overview of variables, sources and reference dates obtained from Statbel
| Variable | Source | Reference date | |
|---|---|---|---|
| Socioeconomic (SE) | Education level | Multiple regional and international databases | 01/01/2017 |
| Income | Tax registry (IPCAL) | Fiscal year 2019 | |
| Employment status | Multiple national and international databases | Week preceding 01/01/2020 | |
| Sociodemographic (SD) | Migration background | National registry | 01/01/2020 |
| Household type | National registry | 01/01/2020 | |
| Age | National registry | 01/01/2020 | |
| Sex | National registry | 01/01/2020 | |
| Region | National registry | 01/01/2020 | |
| Mortality | National registry | 01/01/2022 | |
Study population
In total, 61,074 patients were admitted to and discharged from the hospital between 14 September and 31 December 2020 in Belgium and had at least one overnight stay in the hospital. 11,488 patients were excluded because their age or sex was not available, they were younger than 25 years old, they died in 2020, or they deregistered from the national registry or emigrated in 2020 or 2021. Additionally, 139 patients were excluded because they had a negative total healthcare cost in 2021 in the administrative IMA database. Negative cost amounts can result from corrections of earlier wrongly invoiced costs. Although IMA applies algorithms to resolve these issues, some residual inconsistencies remain [41].
To contextualize healthcare spending among COVID-19 inpatients, we relied on the CHS linkage to distinguish patients hospitalized with COVID-19 (exposed) from those hospitalized without COVID-19 (non-exposed). Because the CHS provided coverage of about 80% of all COVID-19 hospitalized patients in Belgium for the considered study period, some non-differential misclassification of truly exposed patients as non-exposed patients was expected, on average resulting in bias towards the null [36, 42]. 23,177 patients had a link to the CHS after which patients without admission or discharge form available, discharged the same day as admission, and not tested for COVID-19 because of associated symptoms (e.g. systematic screening, to exclude asymptomatic cases) were excluded (n = 12797). In the end, the exposed study population consisted of 10,380 hospitalized COVID-19 patients, corresponding to roughly 45% of the COVID-19 hospitalizations initially identified in the dataset before exclusion criteria were applied. On the other hand, 26,270 hospitalized patients without link to the CHS were considered as hospitalized without COVID-19 (Fig. 1).
Fig. 1.
Flowchart selection of study population of patients hospitalized with and without COVID-19. Patients hospitalized in Belgium between 14 September and 31 December 2020. IMA: Intermutualistic Agency, SE: socioeconomic, SD: sociodemographic, CHS: Clinical Hospital Surveillance
Included variables
This study examined the total direct individual healthcare expenditure (expressed in euros, €) in the calendar year 2021 among patients hospitalized with and without COVID-19 during the last trimester of 2020 as outcome. Cost data were provided by IMA, which compiled healthcare costs available from reimbursement claims of general practitioner, specialist or hospital visits, and prescription medicines purchased in public pharmacies in the context of the Belgian compulsory health insurance. This health insurance encompasses a wide range of healthcare services for which patients receive full or partial reimbursements. Nevertheless, approximately 18% of healthcare expenditure in 2021 in Belgium consisted of out-of-pocket payments paid by the patient [20]. A system of increased reimbursement is in place for people with low income or chronically ill patients. Additionally, a maximum billing system caps high health costs for all patients. A health systems perspective [43] was applied in the study, with the total direct costs calculated as the sum of costs paid by the patient (co-payments and supplements) and paid by the insurance (variable reimbursed costs, reimbursed co-payments when the maximum bill is reached, and an estimate of the fixed costs per hospital stay). These fixed hospital costs were directly paid from the health insurance to the hospitals and were not available from the IMA dataset. Hence, they were estimated by multiplying the total length of hospital stay in 2021 with the average annual 100% per diem cost, an approach previously described [17, 44, 45].
Different SE and SD characteristics reflecting the pre-pandemic situation were included in the analysis. SE variables included education level, income, and employment status. Education level was based on the International Standard Classification of Education (ISCED 2011) and coded as “Low” (lower secondary education or less, ISCED 0 to 2), “Middle” (upper secondary education, ISCED 3 to 4), “High” (higher education, ISCED 5 to 8), or “Missing”. Income data was available as deciles of the yearly net taxable income per household and was considered as “Low” (deciles 1 to 3), “Middle” (deciles 4 to 6), “High” (deciles 7 to 10), or “Missing”. For patients living in collective households, the deciles of personal income were used. Lastly, employment status was defined as “Employed”, “Unemployed”, “Retired or receiving capital income”, and “Other or missing”. The latter category consisted of a heterogeneous group, including students, stay-at-home parents, and people on long-term sick leave. Missingness in SE variables was not randomly distributed (Table 2), and the probability of being missing likely depended on SE category itself (e.g. income level more often missing for people engaging in undocumented work with low wages). Therefore, the missing-indicator method was used and a separate missing category was introduced for all SE variables [46].
Table 2.
Missingness in socioeconomic variables among patients hospitalized with and without COVID-19
| All SE variables available, n (%) | At least one SE variable missing, n (%) | P-value | |
|---|---|---|---|
| Education level | P < 0.001* | ||
| High | 7390 (95.31%) | 364 (4.69%) | |
| Middle | 8322 (88.63%) | 1068 (11.37%) | |
| Low | 13,645 (89.02%) | 1683 (10.98%) | |
| Missing | 0 (0%) | 4178 (100%) | |
| Income | P < 0.001* | ||
| High | 12,099 (88.56%) | 1563 (11.44%) | |
| Middle | 9629 (81.84%) | 2137 (18.16%) | |
| Low | 7629 (72.81%) | 2849 (27.19%) | |
| Missing | 0 (0%) | 744 (100%) | |
| Employment status | P < 0.001* | ||
| Employed | 13,061 (89.4%) | 1549 (10.6%) | |
| Unemployed | 973 (70.76%) | 402 (29.24%) | |
| Retired or capital income | 15,323 (91.12%) | 1493 (8.88%) | |
| Other or missing | 0 (0%) | 3849 (100%) | |
| Migration background | P < 0.001* | ||
| Belgian natives | 23,569 (86.97%) | 3531 (13.03%) | |
| Second-generation | 2619 (78.86%) | 702 (21.14%) | |
| First-generation EU | 1525 (58.99%) | 1060 (41.01%) | |
| First-generation non-EU | 1644 (45.12%) | 2000 (54.88%) | |
| Household type | P < 0.001* | ||
| Couple without children | 11,465 (87.03%) | 1708 (12.97%) | |
| Couple with children | 7608 (74.71%) | 2576 (25.29%) | |
| Single person household | 7076 (80.28%) | 1738 (19.72%) | |
| Single parent family | 1875 (69.68%) | 816 (30.32%) | |
| Collective household | 819 (75.9%) | 260 (24.1%) | |
| Other type | 514 (72.5%) | 195 (27.5%) | |
| Sex | P < 0.001* | ||
| Male | 14,390 (82.67%) | 3017 (17.33%) | |
| Female | 14,967 (77.78%) | 4276 (22.22%) | |
| Age | P < 0.001* | ||
| 25–44 years | 5572 (70.41%) | 2342 (29.59%) | |
| 45–64 years | 8458 (72.01%) | 3288 (27.99%) | |
| 65–84 years | 12,659 (90.6%) | 1314 (9.4%) | |
| 85 + years | 2668 (88.43%) | 349 (11.57%) | |
| Region | P < 0.001* | ||
| Flemish Region | 17,216 (82.8%) | 3576 (17.2%) | |
| Walloon Region | 10,247 (80.09%) | 2547 (19.91%) | |
| Brussels-Capital Region | 1894 (61.81%) | 1170 (38.19%) | |
| Underlying health conditions | P < 0.001* | ||
| 0 | 7282 (77.05%) | 2169 (22.95%) | |
| 1 | 10,977 (82.09%) | 2395 (17.91%) | |
| 2 | 7578 (80.91%) | 1788 (19.09%) | |
| 3 or more | 3520 (78.91%) | 941 (21.09%) |
Patients hospitalized in Belgium between 14 September and 31 December 2020. P-value of chi-squared test investigating missingness in at least one socioeconomic (SE) variable at random by SE variables, sociodemographic variables, or the number of underlying health conditions. Row-wise percentages are shown. * p < 0.05. SE: socioeconomic, EU: European Union
The SD characteristics included migration background, household type, age, sex, and region. Migration background was based on country of birth and the parents’ nationality. The variable distinguished between “Belgian natives”, “Second-generation migrants”, “First-generation EU-27 (European Union) migrants”, and “First-generation non-EU migrants”. Next, household type contained “Couples without children”, “Couples with children”, “Single person household”, “Single parent family”, “Collective households” (including for instance nursing homes, prisons, or religious communities), and “Other types of private households” (such as cohabiting siblings or friends). Region of residence consisted of the “Walloon Region” (southern part of Belgium, mainly French-speaking with a German-speaking minority, more sparsely populated), “Flemish Region” (northern part, mostly Dutch-speaking, more densely populated), and “Brussels-Capital Region” (capital city in the center of the country, both Dutch and French-speaking, very densely populated). No missingness in SD variables was observed in the study population.
Lastly, the number of underlying health conditions was considered to adjust for health needs. This was derived from the IMA 2020 patient data and based on so-called “pseudo-pathologies” defined by a panel of experts from the National Institute for Health and Disability Insurance (NIHDI) using algorithms based on prescribed medication dispensed in public pharmacies. These algorithms rely on a minimum consumption of 90 Defined Daily Doses (DDD) during the calendar year 2020 for medication of pre-defined Anatomical Therapeutical Chemical (ATC) classes in combination with a minimum age of the patient [47, 48]. Additionally, cancer status was based on nomenclature codes for reimbursement of multidisciplinary oncology consultation, radiotherapy, or chemotherapy. The pseudo-pathologies were combined into underlying health conditions, for which cardiovascular, pulmonary, metabolic, neurological, hepatic, renal, and immunological conditions, transplantation, and cancer were considered (Table 3). These were further summarized as “0”, “1”, “2”, or “3 or more” underlying health conditions present.
Table 3.
List of underlying health conditions, pseudo-pathologies, and anatomical therapeutic chemical (ATC) or nomenclature codes
| Underlying health condition | Pseudo-pathologies | Pseudo-pathology codes | ATC/nomenclature codes |
|---|---|---|---|
| Cardiovascular | Cardiovascular disorders (general) | PSEUDOPATH_0101 | C01, C02, C03, C07, C08, C09 |
| Thrombosis (antithrombotics) | PSEUDOPATH_01A01 | B01A | |
| Pulmonary | Chronic obstructive pulmonary disease | PSEUDOPATH_0301 | R03BB, R03DA04, R03A, R03BA |
| Asthma | PSEUDOPATH_0401 | R03DC01, R03DC03, R03DX05, R03A, R03BA | |
| Cystic fibrosis | PSEUDOPATH_0501 | R05CB13, R07AX02, A09AA02 | |
| Metabolic | Diabetes mellitus | PSEUDOPATH_0601 | A10A, A10B |
| Neurological | Psychosis in people aged 70 and under | PSEUDOPATH_1301 | N05AA, N05AB, N05AC, N05AD, N05AE, N05AF, N05AG, N05AH, N05AN, N05AX, N07XX06 |
| Psychosis in people aged 70 and over | PSEUDOPATH_1401 | N05AA, N05AB, N05AC, N05AD, N05AE, N05AF, N05AG, N05AH, N05AN, N05AX, N07XX06 | |
| Parkinson’s disease | PSEUDOPATH_1501 | N04AB, N04AC, N04B | |
| Epilepsy and neuropathic pain | PSEUDOPATH_1601 | N03 | |
| Multiple sclerosis | PSEUDOPATH_1901 | L03AB07, L03AB08, L03AX13, L05AA23, L05AA27, L05AA31, L05AA34, N07XX09 | |
| Alzheimer’s disease | PSEUDOPATH_2101 | N06DX01, N06DA | |
| Hepatic | Chronic hepatitis B and C | PSEUDOPATH_1801 | L03AB04, L03AB05, L03AB09, L03AB10, L03AB11, J05AF08, J05AF10, J05AE11, J05AE12, J05AE14, J05AX15, J05AX65, J05AB04, J05AF05 |
| Renal | Renal failure | PSEUDOPATH_2201 | A11CC03, A11CC04, A11CC06, V03AE02, V03AE03, A12AA12, V03AE04, V03AE01, H05BX01 |
| Immunological | Rheumatoid arthritis, Crohn’s disease, ulcerative colitis, psoriatic arthritis | PSEUDOPATH_1201 | L04AA11, L04AA12, L04AB01, L04AB02, L04AA13, A07EC01, A07EC02, L04AA24, L04AB04, L04AB05, L04AB06, L04AC07 |
| Acquired immunodeficiency syndrome | PSEUDOPATH_1701 | J05AF05, J05AE, J05AF, J05AX | |
| Transplantation | Organ transplantation | PSEUDOPATH_2001 | L04AA02, L04AA06, L04AA10, L04AA18, L04AC02, L04AD02, L04AA01, L04AD01 |
| Cancer | Cancer (based on reimbursement for multidisciplinary oncology consultation) | CANCER_MOC_YN | 350,232, 350,254, 350,265, 350,276, 350,280, 350,291, 350,302, 350,372, 350,383, 350,394, 350,405, 350,416, 350,420, 350,453, 350,464, 350,475, 350,486 |
| Cancer (based on reimbursement for chemotherapy or radiotherapy) | CANCER_CHEMORT_YN | L01, 444,113, 444,124, 444,135, 444,146, 444,150, 444,161, 444,172, 444,183, 444,216, 444,220, 444,253, 444,264, 444,290, 444,301, 444,312, 444,323 |
Used algorithms rely on a minimum consumption of 90 Defined Daily Dose (DDD) during the calendar year 2021 for medication of pre-defined Anatomical Therapeutical Chemical (ATC) classes in combination with a minimum age of the patient. Additionally, the presence of cancer was defined based on nomenclature codes for reimbursement of multidisciplinary oncology consultation, radiotherapy or chemotherapy. More information on these algorithms is available through the Intermutualistic Agency (IMA). The pseudo-pathologies were combined into underlying health conditions, which were further summarized as “0”, “1”, “2”, or “3 or more” underlying health conditions present
Statistical analyses
A chi-squared test was performed to investigate missingness in at least one SE variable at random by SE variables, SD variables, or the number of underlying health conditions. The missing-indicator method was applied and a missing category was introduced for the SE variables [46]. To identify baseline differences between patients hospitalized with COVID-19 (exposed) or without COVID-19 (non-exposed), a logistic regression model was fitted with COVID-19 exposure status as binary outcome and SE variables, SD variables, and number of underlying health conditions as predictors. A likelihood ratio test (LRT) investigated significant baseline differences in composition of the two groups for each considered variable adjusting for all other variables. Since various significant baseline differences between the exposed and non-exposed groups were observed, a matching strategy using stabilized inverse probability weighting (IPW) was applied to mitigate confounding bias. Weights were derived from the predictions of the previously fitted logistic regression model using the following formula:
 |
where w represents the weight, X the COVID-19 exposure status, SE the socioeconomic variables, SD the sociodemographic variables, H the number of underlying health conditions, and i the individual patient [49]. A chi-squared test investigated differences in prevalence of individual underlying health conditions between patients hospitalized with and without COVID-19. Post-acute healthcare costs were described by calculating the mean, median, quartiles, minimum and maximum and using a histogram truncated at the overall 95th percentile, separately for patients hospitalized with and without COVID-19. For the main analysis, generalized linear models (GLMs) with a log-link function and a negative binomial distribution were fitted using the healthcare costs as outcome variable and applying the weights from IPW. To investigate the first objective, we fitted a model with COVID-19 status as exposure, both unadjusted and adjusted for SE characteristics, SD variables, and the number of underlying health conditions. The second objective of the study looked at differences in SE and SD factors associated with increased healthcare expenditure between patients hospitalized with or without COVID-19. In the statistical model, we introduced interaction terms between the COVID-19 status on the one hand, and SE and SD variables, and the number of underlying health conditions on the other hand. Statistical tests were performed at the 5% nominal significance level. All analyses were performed using R (version 4.4.0).
Sensitivity analysis
To account for survival bias, we additionally excluded patients who died in the calendar year of 2021 from the sensitivity analysis. The resulting study population consisted of 9666 patients hospitalized with and 24,459 patients hospitalized without COVID-19. Summary and descriptive statistics were calculated among all hospitalized (both COVID and non-COVID) patients by mortality state. The same statistical methodology as for the main analysis was employed; a logistic regression model was fitted, an LRT test was performed, stabilized IPW were calculated, and GLMs with a negative binomial distribution and a log-link function were fitted.
Results
Description of the study population
The study included 10,380 patients with COVID-19 and 26,270 patients without COVID-19 admitted to and discharged from a Belgian hospital between 14 September and 31 December 2020. Regarding the SE variables, most patients had a low education level (42%), compared to a middle (26%), high (21%), or missing (11%) education level (Table 4). The high-, middle-, and low-income groups comprised of about one-third of the patients. 40% of patients were employed, whereas 46% were retired or received capital income, 4% were unemployed, and 11% were coded as other or missing. By migration background, 74% of patients were native Belgian, whereas 9% were second-generation, 7% were first-generation EU, and 10% were first-generation non-EU migrants. Most patients were part of a couple without children (36%), couple with children (28%), or single person household (24%), whereas patients living in single parent families (7%), collective households (3%) and other types of private households (2%) were a minority. The study population comprised of slightly more females (53%) than males (47%). In total, 22% of the patients were between 25 and 44 years old, 32% were aged between 45 and 64, 38% were between 65 and 84, and 8% were 85 years and over. Most patients resided in the Flemish (57%) and Walloon Region (35%), with a smaller number living in the Brussels-Capital Region (8%). Almost three out of four patients suffered from underlying health conditions (36% having one and 38% having multiple health conditions). Patients hospitalized with COVID-19 more often had cardiovascular, pulmonary, renal, metabolic, neurological and immunological conditions, whereas hospitalized non-COVID-19 patients were more frequently diagnosed with cancer (Table 5). Males, older patients, patients residing in Wallonia, patients living in a collective household, first-generation non-EU migrants, patients with lower education level, retired patients, and patients with underlying health conditions were overrepresented among patients hospitalized with COVID-19. Hence, stabilized inverse probability weights were applied in subsequent statistical analyses.
Table 4.
Descriptive statistics of patients hospitalized with and without COVID-19
| All, n (%) | COVID-19, n (%) | No COVID-19, n (%) | P-value | |
|---|---|---|---|---|
| All | 36,650 (100%) | 10,380 (100%) | 26,270 (100%) | |
| Education level | P < 0.001* | |||
| High | 7754 (21.16%) | 1693 (16.31%) | 6061 (23.07%) | |
| Middle | 9390 (25.62%) | 2298 (22.14%) | 7092 (27%) | |
| Low | 15,328 (41.82%) | 5026 (48.42%) | 10,302 (39.22%) | |
| Missing | 4178 (11.4%) | 1363 (13.13%) | 2815 (10.72%) | |
| Income | P = 0.103 | |||
| High | 13,662 (37.28%) | 3479 (33.52%) | 10,183 (38.76%) | |
| Middle | 11,766 (32.1%) | 3424 (32.99%) | 8342 (31.75%) | |
| Low | 10,478 (28.59%) | 3245 (31.26%) | 7233 (27.53%) | |
| Missing | 744 (2.03%) | 232 (2.24%) | 512 (1.95%) | |
| Employment status | P < 0.001* | |||
| Employed | 14,610 (39.86%) | 3506 (33.78%) | 11,104 (42.27%) | |
| Unemployed | 1375 (3.75%) | 321 (3.09%) | 1054 (4.01%) | |
| Retired or capital income | 16,816 (45.88%) | 5616 (54.1%) | 11,200 (42.63%) | |
| Other or missing | 3849 (10.5%) | 937 (9.03%) | 2912 (11.08%) | |
| Migration background | P < 0.001* | |||
| Belgian natives | 27,100 (73.94%) | 6964 (67.09%) | 20,136 (76.65%) | |
| Second-generation | 3321 (9.06%) | 912 (8.79%) | 2409 (9.17%) | |
| First-generation EU | 2585 (7.05%) | 898 (8.65%) | 1687 (6.42%) | |
| First-generation non-EU | 3644 (9.94%) | 1606 (15.47%) | 2038 (7.76%) | |
| Household type | P < 0.001* | |||
| Couple without children | 13,173 (35.94%) | 3710 (35.74%) | 9463 (36.02%) | |
| Couple with children | 10,184 (27.79%) | 2861 (27.56%) | 7323 (27.88%) | |
| Single person household | 8814 (24.05%) | 2453 (23.63%) | 6361 (24.21%) | |
| Single parent family | 2691 (7.34%) | 653 (6.29%) | 2038 (7.76%) | |
| Collective household | 1079 (2.94%) | 495 (4.77%) | 584 (2.22%) | |
| Other type | 709 (1.93%) | 208 (2%) | 501 (1.91%) | |
| Sex | P < 0.001* | |||
| Male | 17,407 (47.5%) | 5786 (55.74%) | 11,621 (44.24%) | |
| Female | 19,243 (52.5%) | 4594 (44.26%) | 14,649 (55.76%) | |
| Age | P < 0.001* | |||
| 25–44 years | 7914 (21.59%) | 1018 (9.81%) | 6896 (26.25%) | |
| 45–64 years | 11,746 (32.05%) | 3531 (34.02%) | 8215 (31.27%) | |
| 65–84 years | 13,973 (38.13%) | 4709 (45.37%) | 9264 (35.26%) | |
| 85 + years | 3017 (8.23%) | 1122 (10.81%) | 1895 (7.21%) | |
| Region | P < 0.001* | |||
| Flemish Region | 20,792 (56.73%) | 4741 (45.67%) | 16,051 (61.1%) | |
| Walloon Region | 12,794 (34.91%) | 4599 (44.31%) | 8195 (31.2%) | |
| Brussels-Capital Region | 3064 (8.36%) | 1040 (10.02%) | 2024 (7.7%) | |
| Underlying health conditions | P < 0.001* | |||
| 0 | 9451 (25.79%) | 1943 (18.72%) | 7508 (28.58%) | |
| 1 | 13,372 (36.49%) | 3722 (35.86%) | 9650 (36.73%) | |
| 2 | 9366 (25.56%) | 3171 (30.55%) | 6195 (23.58%) | |
| 3 or more | 4461 (12.17%) | 1544 (14.87%) | 2917 (11.1%) |
Patients hospitalized in Belgium between 14 September and 31 December 2020. P-value of likelihood-ratio test (LRT) test investigating baseline differences in patients hospitalized with or without COVID-19 for each socioeconomic, sociodemographic variable, or underlying health conditions, adjusted for all other variables in table. Column-wise percentages are shown. * p < 0.05. EU: European Union
Table 5.
Underlying health conditions in patients hospitalized with and without COVID-19
| All, n (%) | COVID-19, n (%) | No COVID-19, n (%) | P-value | |
|---|---|---|---|---|
| All | 36,650 (100%) | 10,380 (100%) | 26,270 (100%) | |
| Underlying health conditions | P < 0.001* | |||
| 0 | 9451 (25.79%) | 1943 (18.72%) | 7508 (28.58%) | |
| 1 | 13,372 (36.49%) | 3722 (35.86%) | 9650 (36.73%) | |
| 2 | 9366 (25.56%) | 3171 (30.55%) | 6195 (23.58%) | |
| 3 or more | 4461 (12.17%) | 1544 (14.87%) | 2917 (11.1%) | |
| Cardiovascular | P < 0.001* | |||
| Present | 23,343 (63.69%) | 7,571 (72.94%) | 15,772 (60.04%) | |
| Not present | 13,307 (36.31%) | 2,809 (27.06%) | 10,498 (39.96%) | |
| Pulmonary | P < 0.001* | |||
| Present | 5,863 (16%) | 2,127 (20.49%) | 3,736 (14.22%) | |
| Not present | 30,787 (84%) | 8,253 (79.51%) | 22,534 (85.78%) | |
| Renal | P < 0.001* | |||
| Present | 987 (2.69%) | 344 (3.31%) | 643 (2.45%) | |
| Not present | 35,663 (97.31%) | 10,036 (96.69%) | 25,627 (97.55%) | |
| Hepatic | P < 0.817 | |||
| Present | 97 (0.26%) | 29 (0.28%) | 68 (0.26%) | |
| Not present | 36,553 (99.74%) | 10,351 (99.72%) | 26,202 (99.74%) | |
| Metabolic | P < 0.001* | |||
| Present | 6,995 (19.09%) | 2,615 (25.19%) | 4,380 (16.67%) | |
| Not present | 29,655 (80.91%) | 7,765 (74.81%) | 21,890 (83.33%) | |
| Neurological | P = 0.005* | |||
| Present | 4,122 (11.25%) | 1,245 (11.99%) | 2,877 (10.95%) | |
| Not present | 32,528 (88.75%) | 9,135 (88.01%) | 23,393 (89.05%) | |
| Immunological | P < 0.001* | |||
| Present | 1,351 (3.69%) | 444 (4.28%) | 907 (3.45%) | |
| Not present | 35,299 (96.31%) | 9,936 (95.72%) | 25,363 (96.55%) | |
| Transplantation | P = 0.282 | |||
| Present | 255 (0.7%) | 64 (0.62%) | 191 (0.73%) | |
| Not present | 36,395 (99.3%) | 10,316 (99.38%) | 26,079 (99.27%) | |
| Cancer | P < 0.001* | |||
| Present | 3,566 (9.73%) | 606 (5.84%) | 2,960 (11.27%) | |
| Not present | 33,084 (90.27%) | 9,774 (94.16%) | 23,310 (88.73%) |
Patients hospitalized in Belgium between 14 September and 31 December 2020. P-value chi-squared test investigating differences in presence of underlying health conditions among patients hospitalized with and without COVID-19. * p < 0.05
Objective 1: Post-acute healthcare expenditure by COVID-19 exposure status
The individual healthcare expenditure in 2021 averaged to €17,219 with a median of €8495, as the distribution of healthcare costs was highly skewed to the right. On average, observed post-acute costs were lower in patients hospitalized with compared to without COVID-19 (Table 6; Fig. 2). Adjusted for SE and SD variables, and the number of underlying health conditions, COVID-19 hospitalization was associated with 18.5% (95% confidence interval (CI): 16.9% to 20.1%) lower healthcare costs in the calendar year following hospital admission compared to non-COVID hospitalization (Table 7).
Table 6.
Summary statistics of post-acute healthcare expenditure in 2021 among hospitalized patients with and without COVID-19
| n | Min | Q1 | Median | Mean | Q3 | Max | |
|---|---|---|---|---|---|---|---|
| All | 36,650 | €196 | €4511 | €8495 | €17,219 | €20,043 | €567,333 |
| COVID-19 | 10,380 | €982 | €3996 | €7945 | €17,209 | €21,058 | €283,022 |
| No COVID-19 | 26,270 | €196 | €4749 | €8679 | €17,233 | €19,694 | €567,333 |
Patients hospitalized in Belgium between 14 September and 31 December 2020. Min: minimum, Q1: first quartile, Q3: third quartile, Max: maximum
Fig. 2.
Histogram of post-acute healthcare costs in 2021 among patients hospitalized with (n = 10380) and without (n = 26270) COVID-19. Patients hospitalized in Belgium between 14 September and 31 December 2020. Post-acute healthcare costs were truncated at the overall 95th percentile
Table 7.
Association between COVID-19 exposure status and post-acute healthcare expenditure
| Model | Group | Exp(B) [95% CI] | P-value |
|---|---|---|---|
| Unadjusted |
No COVID-19 (ref.) COVID-19 |
1 0.860 [0.841, 0.880] |
< 0.001* |
| Adjusted1 |
No COVID-19 (ref.) COVID-19 |
1 0.815 [0.799, 0.831] |
< 0.001* |
Exponentiated regression coefficients (Exp(B)) with 95% confidence intervals (CI) for the association COVID-19 exposure status and post-acute healthcare expenditure in 2021 among patients hospitalized in Belgium between 14 September and 31 December 2020 with (n = 10380) and without (n = 26270) COVID-19. 1Adjusted for education level, income, employment status, migration background, household type, sex, age, region, and number of underlying health conditions. * p < 0.05. CI: confidence interval
Objective 2: SE and SD variables associated with post-acute healthcare expenditure
Important socioeconomic inequalities in post-acute expenditure were found by education level, employment status, and income, both for COVID and non-COVID patients, although with some differences between the groups. Regarding the education level, COVID-19 hospitalized patients with a middle and low education level experienced 9.5% (95% CI: 4.3% to 15.0%) and 5.5% (95% CI: 0.4% to 10.8%) higher total healthcare costs compared to patients with a high education level, respectively. However, a different trend was observed in patients hospitalized without COVID-19, where total costs were 4.1% (95% CI: 1.1% to 7.0%) lower, for people with middle education level compared to high education level. People with a low education level also experience lower costs than people with high education level, although not significant. Higher costs were also observed for people with lower income. Compared to patients with a high income, COVID-19 patients with a middle or low income experienced 10.7% (95% CI: 5.9% to 15.7%) and 16.7% (95% CI: 10.4% to 23.4%) higher costs, respectively, whereas patients with a middle or low income without COVID-19 experienced 3.1% (95% CI: 0.3% to 6.0%) and 4.2% (95% CI: 0.6% to 7.9%) higher costs, respectively. Lastly, unemployed COVID-19 patients experienced 14.5% (95% CI: 6.2% to 22.0%) lower total costs compared to employed patients, whereas unemployed patients without COVID-19 showed 6.1% (95% CI: 0.2% to 12.4%) higher costs compared to employed patients. On the other hand, both for COVID and non-COVID, significantly higher costs were observed for people on retirement or receiving capital income compared to employed people.
Compared to native Belgians, first-generation EU COVID-19 patients had 14.3% (95% CI: 8.2% to 20.0%) lower healthcare costs, whereas no difference was observed in patients without COVID-19. Additionally, first-generation non-EU migrants had lower healthcare costs both in patients with and without COVID-19. On the other hand, no difference was observed between native Belgians and second-generation migrants. Regarding the household type, COVID-19 patients living as couple with children experienced 5.8% (95% CI: 0.7% to 11.2%) higher costs compared to couple without children, whereas non-COVID patients living as couple with children experienced 4.4% (95% CI: 1.5% to 7.2%) lower costs. On the other hand, both for COVID and non-COVID patients, single person household and single parent family experienced higher costs. Compared to patients living as a couple without children, COVID-19 patients living in a collective household had 117.7% (95% CI: 96.5% to 141.2%) higher costs and non-COVID patients living in a collective household had 108.2% (95% CI: 95.2% to 122.2%) higher costs. Female patients hospitalized without COVID-19 experienced 4.2% (95% CI: 2.2% to 6.2%) lower total costs than males, whereas the sex difference was not significant among COVID-19 hospitalized patients. A gradient in total costs was observed with advancing age; both in patients hospitalized with and without COVID-19, higher costs were experienced by older patients. In comparison to the Flemish Region, health costs of COVID-19 hospitalized patients were 3.9% (95%CI: 0.2% to 7.7%) higher in the Walloon Region and 16.2% (95% CI: 8.9% to 23.9%) higher in the Brussels-Capital Region. On the other hand, health costs in patients without COVID-19 were 5.1% (95% CI: 3.0% to 8.2%) lower in the Walloon Region and 7.6% (95% CI: 3.3% to 12.0%) higher in the Brussels-Capital Region compared to the Flemish Region (Fig. 3).
Fig. 3.
Association socioeconomic and sociodemographic variables with post-acute healthcare expenditure by COVID-19 exposure state. Exponentiated regression coefficients (Exp(B)) with 95% confidence intervals (CI) for the association between SE and SD characteristics and post-acute healthcare costs in 2021, adjusted for the number of underlying health conditions, by COVID-19 exposure status among patients hospitalized in Belgium between 14 September and 31 December 2020 with (n = 10380) and without (n = 26270) COVID-19. P-values for interaction test statistical significant differences in the association between hospitalized patients with and without COVID-19. * p < 0.05. CI: confidence interval, EU: European Union
Sensitivity analysis
To account for mortality during follow-up, we excluded patients deceased in 2021 from the sensitivity analysis. On average, patients deceased in 2021 had much higher post-acute healthcare expenditure than patients who survived (Table 8). Moreover, various baseline covariate differences were observed between patients who died in 2021 and those who survived (Table 9). After exclusion of patients deceased in 2021, patients hospitalized with COVID-19 experienced 19.1% (95% CI: 17.5% to 20.7%) lower healthcare costs compared to patients without COVID-19 (Table 10). Regarding the association of post-acute healthcare costs with SD and SE variables, some small differences from the main model were observed (Fig. 4). First, among non-COVID patients, the difference between middle and high income as well as the difference between employed and unemployed was no longer significant. Among COVID-19 patients, the difference between couple with children and couple without children was no longer significant. On the other hand, we observed lower costs among second-generation COVID-19 patients compared to native Belgians when removing deceased patients. Lastly, both among patients hospitalized with and without COVID-19, no sex difference was observed anymore.
Table 8.
Summary statistics of post-acute healthcare expenditure in 2021 among hospitalized patients by 2021 mortality state
| n | Min | Q1 | Median | Mean | Q3 | Max | |
|---|---|---|---|---|---|---|---|
| Not deceased in 2021 | 34,125 | €196 | €4317 | €7806 | €15,272 | €17,060 | €567,333 |
| Deceased in 2021 | 2525 | €1115 | €22,234 | €35,199 | €43,528 | €55,363 | €283,022 |
Patients hospitalized in Belgium between 14 September and 31 December 2020. Min: minimum, Q1: first quartile, Q3: third quartile, Max: maximum
Table 9.
Descriptive statistics of patients hospitalized with and without COVID-19 by 2021 mortality state
| All, n (%) | Not deceased in 2021, n (%) | Deceased in 2021, n (%) |
P-value | |
|---|---|---|---|---|
| All | 36,650 (100%) | 34,125 (100%) | 2525 (100%) | |
| COVID-19 exposure status | P < 0.001* | |||
| COVID-19 | 10,380 (28.32%) | 9666 (28.33%) | 714 (28.28%) | |
| No COVID-19 | 26,270 (71.68%) | 24,459 (71.67%) | 1811 (71.72%) | |
| Education level | P = 0.003* | |||
| High | 7754 (21.16%) | 7428 (21.77%) | 326 (12.91%) | |
| Middle | 9390 (25.62%) | 8934 (26.18%) | 456 (18.06%) | |
| Low | 15,328 (41.82%) | 13,866 (40.63%) | 1462 (57.9%) | |
| Missing | 4178 (11.4%) | 3897 (11.42%) | 281 (11.13%) | |
| Income | P = 0.143 | |||
| High | 13,662 (37.28%) | 13,150 (38.53%) | 512 (20.28%) | |
| Middle | 11,766 (32.1%) | 10,851 (31.8%) | 915 (36.24%) | |
| Low | 10,478 (28.59%) | 9419 (27.6%) | 1059 (41.94%) | |
| Missing | 744 (2.03%) | 705 (2.07%) | 39 (1.54%) | |
| Employment status | P < 0.001* | |||
| Employed | 14,610 (39.86%) | 14,287 (41.87%) | 323 (12.79%) | |
| Unemployed | 1375 (3.75%) | 1353 (3.96%) | 22 (0.87%) | |
| Retired or capital income | 16,816 (45.88%) | 14,794 (43.35%) | 2022 (80.08%) | |
| Other or missing | 3849 (10.5%) | 3691 (10.82%) | 158 (6.26%) | |
| Migration background | P < 0.001* | |||
| Belgian natives | 27,100 (73.94%) | 24,986 (73.22%) | 2114 (83.72%) | |
| Second-generation | 3321 (9.06%) | 3189 (9.35%) | 132 (5.23%) | |
| First-generation EU | 2585 (7.05%) | 2410 (7.06%) | 175 (6.93%) | |
| First-generation non-EU | 3644 (9.94%) | 3540 (10.37%) | 104 (4.12%) | |
| Household type | P < 0.001* | |||
| Couple without children | 13,173 (35.94%) | 12,211 (35.78%) | 962 (38.1%) | |
| Couple with children | 10,184 (27.79%) | 9943 (29.14%) | 241 (9.54%) | |
| Single person household | 8814 (24.05%) | 7955 (23.31%) | 859 (34.02%) | |
| Single parent family | 2691 (7.34%) | 2540 (7.44%) | 151 (5.98%) | |
| Collective household | 1079 (2.94%) | 829 (2.43%) | 250 (9.9%) | |
| Other type | 709 (1.93%) | 647 (1.9%) | 62 (2.46%) | |
| Sex | P < 0.001* | |||
| Male | 17,407 (47.5%) | 15,986 (46.85%) | 1421 (56.28%) | |
| Female | 19,243 (52.5%) | 18,139 (53.15%) | 1104 (43.72%) | |
| Age | P < 0.001* | |||
| 25–44 years | 7914 (21.59%) | 7866 (23.05%) | 48 (1.9%) | |
| 45–64 years | 11,746 (32.05%) | 11,357 (33.28%) | 389 (15.41%) | |
| 65–84 years | 13,973 (38.13%) | 12,596 (36.91%) | 1377 (54.53%) | |
| 85 + years | 3017 (8.23%) | 2306 (6.76%) | 711 (28.16%) | |
| Region | P < 0.001* | |||
| Flemish Region | 20,792 (56.73%) | 19,396 (56.84%) | 1396 (55.29%) | |
| Walloon Region | 12,794 (34.91%) | 11,847 (34.72%) | 947 (37.5%) | |
| Brussels-Capital Region | 3064 (8.36%) | 2882 (8.45%) | 182 (7.21%) | |
| Underlying health conditions | P < 0.001* | |||
| 0 | 9451 (25.79%) | 9342 (27.38%) | 109 (4.32%) | |
| 1 | 13,372 (36.49%) | 12,670 (37.13%) | 702 (27.8%) | |
| 2 | 9366 (25.56%) | 8402 (24.62%) | 964 (38.18%) | |
| 3 or more | 4461 (12.17%) | 3711 (10.87%) | 750 (29.7%) |
Patients hospitalized in Belgium between 14 September and 31 December 2020. P-value of likelihood-ratio test (LRT) test investigating baseline differences in hospitalized patients deceased or not in 2021 for COVID-19 exposure state, each socioeconomic, sociodemographic variable, or underlying health conditions, adjusted for all other variables in table. Column-wise percentages are shown. * p < 0.05. EU: European Union
Table 10.
Association between COVID-19 exposure status and post-acute healthcare expenditure, excluding deceased
| Model | Group | Exp(B) [95% CI] | P-value |
|---|---|---|---|
| Unadjusted |
No COVID-19 (ref.) COVID-19 |
1 0.855 [0.835, 0.874] |
< 0.001* |
| Adjusted1 |
No COVID-19 (ref.) COVID-19 |
1 0.809 [0.793, 0.825] |
< 0.001* |
Exponentiated regression coefficients (Exp(B)) with 95% confidence intervals (CI) for the association COVID-19 exposure status and post-acute healthcare expenditure in 2021 among patients hospitalized in Belgium between 14 September and 31 December 2020 with (n = 9666) and without (n = 24459) COVID-19 after excluding patients deceased in 2021. 1Adjusted for education level, income, employment status, migration background, household type, sex, age, region, and number of underlying health conditions. * p < 0.05. CI: confidence interval
Fig. 4.
Association socioeconomic and sociodemographic variables with post-acute healthcare expenditure by COVID-19 exposure state, excluding deceased. Exponentiated regression coefficients (Exp(B)) with 95% confidence intervals (CI) for the association between SE and SD characteristics and post-acute healthcare costs in 2021, adjusted for the number of underlying health conditions, by COVID-19 exposure status among patients hospitalized in Belgium between 14 September and 31 December 2020 with (n = 9666) and without (n = 24459) COVID-19 after excluding patients deceased in 2021. P-values for interaction test statistical significant differences in the association between hospitalized patients with and without COVID-19. * p < 0.05. CI: confidence interval, EU: European Union
Discussion
This study investigated post-acute healthcare expenditure in the first calendar year following hospitalization among patients hospitalized with COVID-19 relative to patients hospitalized without COVID-19 in Belgium. Our study provides one of the few empirical assessments of post-acute healthcare expenditures following COVID-19 hospitalization in Belgium and situates these findings within the context of non-COVID-19 hospitalizations. Contrary to international literature findings [32, 50], we observed lower post-acute health costs for patients hospitalized with compared to without COVID-19, possibly due to the nature of hospital admission of the non-exposed group. Included patients were admitted to the hospital between 14 September and 31 December 2020, during the second wave of the COVID-19 epidemic in Belgium. Because COVID-19 policy measures in place at that moment reduced or suspended non-essential care [51], patients hospitalized without COVID-19 likely were admitted for more severe acute health conditions, with potentially resulted in higher post-acute healthcare expenditure. Although we do not have information on the severity of the comorbidity, our sample of hospitalized COVID-19 patients more frequently had cardiovascular, pulmonary, renal, metabolic, neurological and immunological underlying conditions, whereas hospitalized non-COVID-19 patients were more often diagnosed with cancer. It is possible that these cardiovascular, pulmonary, renal, metabolic, neurological and immunological conditions may not have triggered as intensive or costly post-discharge interventions. In contrast, cancer hospitalizations often involve acute interventions that are inherently costlier during and after discharge, although considerable differences have been reported by cancer type in Belgium [52]. We showed in the sensitivity analysis that survival bias could not explain the observed results; when excluding patients deceased in 2021, lower post-acute healthcare costs among patients with compared to without COVID-19 were still observed.
As a second objective, we investigated the association between SE and SD factors and post-acute healthcare expenditure among patients hospitalized with COVID-19. We observed higher costs for patients with a lower education level, lower income, without migration background, living in a collective household, and at older age. Our results are in line with previous observational research investigating risk factors for increased 6-month post-acute cost of care following COVID-19 infection in Italy. In that study, higher post-acute costs were observed among females, older patients, and patients with more comorbidities, as defined by the Charlson comorbidity index. Additionally, absence of COVID-19 vaccination, having a symptomatic infection, and requiring hospital admission for COVID-19 were identified as risk factors for increased post-acute costs in Italy [27]. Some other studies also observed higher post-acute healthcare expenditure among older COVID-19 patients [21, 29, 30] or among patients with a COVID-19 requiring hospital admission [22, 23, 30].
Higher health expenditure following COVID-19 hospitalization might be attributable to increased medical needs due to the development of PASC. Also within COVID-19 patients, those developing PASC have been shown to experience higher post-acute medical costs [24]. In Belgium, 47% of COVID-19 patients report PASC three months following infection, with higher rates in females, people with a lower education level, and people with more comorbidities [3]. These findings align with our results, as we observe higher post-acute health costs for these groups in hospitalized COVID-19 patients. However, advancing age was not identified as a risk factor for the development of PASC [3]. Moreover, the PASC development status was not known in our study and could therefore not be adjusted for.
Lower education level was associated with increased post-acute healthcare expenditure among hospitalized COVID-19 patients compared to a high education level. However, among patients hospitalized without COVID-19, no difference was observed in post-acute health costs between the different education levels. Different mechanisms can explain the association between education level and post-acute health expenditure in hospitalized COVID-19 patients. First, the delay between onset of COVID symptoms and diagnosis might play a role. In a cohort of healthcare workers in the US, longer time between COVID-19 symptoms and diagnosis was observed among people with a lower education level [53]. As a later diagnosis results in worse acute COVID-19 outcomes [54], it might also result in poorer post-acute outcomes and higher healthcare costs. Additionally, not being able to telework could to some extent explain the higher post-acute healthcare costs observed in hospitalized COVID-19 patients with a low education level. People with a higher education level work more often from home than people with a lower education level in Belgium [55]. In general, teleworking tends to reduce stress and improve the health, well-being, and quality of life of employees, although the effects differ by support given by the employer and the personal situation at home [56]. Moreover, the vast majority of the study population was not employed.
To our knowledge, this is the first study to provide individual-level, population-based evidence on post-acute healthcare expenditures following COVID-19 hospitalization in Belgium. By linking the CHS with administrative reimbursement data, we captured a substantial proportion of hospitalized COVID-19 patients (about 45% of the original population in the national COVID-19 surveillance) and a contemporaneous non-COVID-19 comparison group, allowing contextualization of expenditure patterns during crisis time. These features make our findings robust and informative for understanding healthcare utilization and planning in the Belgian context.
Several limitations of the current study need to be acknowledged. A first major limitation of our study is possible selection bias as some exclusion criteria were specific to COVID-19 patients (e.g. the exclusion of patients without available admission and discharge forms or patients not tested for COVID-19 symptoms). However, we partly mitigated this limitation by applying using IPW in the analysis. Misclassification bias is also expected because of the partial coverage (80%) of all hospitalized COVID-19 patients in Belgium by the CHS, resulting in estimates that are on average biased towards the null [42]. Since our study only included hospitalized patients with COVID-19, statistical inference should be limited to this population to avoid collider bias [57]. Hence, our results cannot be extrapolated to all COVID-19 patients. Secondly, by considering healthcare expenditure in the context of the Belgian compulsory health insurance using data available from IMA, not all healthcare-related costs were taken into account. For instance, non-reimbursed over-the-counter drugs were not included. Furthermore, we only focused on direct post-acute healthcare costs, since data on indirect costs, such as costs for transportation to the hospital or due to absenteeism from work because of illness, were not available. Next, coverage of the Belgian compulsory health insurance is not complete, with lower insurance rates in males and young adults, and no coverage for undocumented migrants and asylum seekers [20]. Therefore, some highly vulnerable populations could not be included in the current analysis. To adjust for health needs, we introduced the number of underlying health conditions based on the consumption of certain categories of medication as a confounder in the model. However, we should be careful in interpreting these regression coefficients as both the post-acute healthcare costs and the number of underlying health conditions are based on reimbursement requests to the Belgian compulsory health insurance. Additionally, it has been shown that not all (combinations of) underlying health conditions will affect healthcare expenditure to the same extend [17]. Moreover, we were not able to adjust for the COVID-19 vaccination status, which is also characterized by social inequalities [7, 10]. The vaccination campaign for the general population was only launched in January 2021 [58] and thus could not have influenced the disease course of COVID-19 patients during hospitalization. Nevertheless, patients could have acquired a SARS-CoV-2 infection in the calendar year 2021 requiring medical care and, as a result, could have experienced associated healthcare expenditure. Lastly, as the different SE and SD variables are not independent, the obtained regression coefficients are not causal effect estimates and should be interpreted with caution.
Conclusion
This study investigated post-acute healthcare costs among patients hospitalized with and without COVID-19 during the last trimester of 2020 in Belgium and social inequalities in their healthcare expenditure. Patients hospitalized with COVID-19 experienced lower post-acute healthcare costs compared to non-COVID patients, as hospitalized patients without COVID-19 during the pandemic might have been admitted for more severe health conditions. Lower education level, lower income, living in a collective household compared to other household types, and older age were associated with higher post-acute healthcare costs among COVID-19 patients, whereas COVID-19 patients having a migration background experienced lower healthcare costs compared to Belgian natives.
This study identified different socioeconomic and sociodemographic characteristics associated with increased post-acute healthcare expenditure, emphasizing the vulnerability of hospitalized COVID-19 patients with lower social status. Future research is needed to investigate the healthcare expenditure of COVID-19 hospitalized patients over longer periods of time and how social inequalities in healthcare expenditure evolve over time. Moreover, the drivers and underlying mechanisms of these social inequalities remain to be studied.
Acknowledgements
Not applicable.
Abbreviations
- COVID-19
coronavirus disease 2019
- PASC
post-acute sequelae of COVID-19
- SE
socioeconomic
- SD
sociodemographic
- SSIN
social security identification number
- CHS
Clinical Hospital Surveillance
- Statbel
Statistics Belgium, IMA: Intermutualistic Agency
- ISCED
International Standard Classification of Education
- EU
European Union
- NIHDI
National Institute for Health and Disability Insurance
- DDD
Defined Daily Doses
- ATC
Anatomical Therapeutical Chemical
- LRT
likelihood ratio test
- IPW
inverse probability weighting
- GLM
generalized linear model
- CI
confidence interval
Author contributions
BD, RDP, DDS, SG, KV, NV, and LVdB conceptualized the study and obtained funding. EB, LC, and LVdB designed the methodology. EB performed statistical analyses. EB wrote the original draft of the paper. All authors read and approved the final manuscript.
Funding
This research was funded by the Belgian Science Policy Office (BELSPO) within the BRAIN-be 2.0 framework supporting pillar 3 Federal societal challenges (grant number B2/202/P3/HELICON). The funders had no role in conceptualization, design, data collection and analysis, decision to publish, or preparation of the manuscript.
Data availability
Due to the General Data Protection Regulation (GDPR) legislations in Belgium, these data are not publicly available. Data requests must be addressed to the Information Security Committee (ISC).
Declarations
Ethics approval and consent to participate
The study was approved by the Belgian Information Security Committee (ISC) Social Security and Health (IVC/KSZG/22/034). The ISC confirmed that this study falls under Article 9§ 2(j) of the General Data Protection Regulation (GDPR) and, in compliance with these GDPR legal grounds of data processing, no informed consent was obtained from the patients. Furthermore, the study was conducted in accordance with the Declaration of Helsinki and ethical approval was obtained from the Ghent University Hospital ethics committee (B.U.N. 1432020000371).
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Lu S, Huang X, Liu R, Lan Y, Lei Y, Zeng F, et al. Comparison of COVID-19 induced respiratory failure and typical ARDS: similarities and differences. Front Med (Lausanne). 2022;9:829771. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Fernández-de-las-Peñas C, Palacios-Ceña D, Gómez-Mayordomo V, Florencio LL, Cuadrado ML, Plaza-Manzano G, et al. Prevalence of post-COVID-19 symptoms in hospitalized and non-hospitalized COVID-19 survivors: A systematic review and meta-analysis. Eur J Intern Med. 2021;92:55–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Smith P, Charafeddine R, Drieskens S, De Pauw R, De Ridder K, Demarest S et al. COVIMPACT studie - COVID-19 infectie en de lichamelijke, geestelijke en sociale gevolgen op lange termijn: resultaten van 3 en 6 maanden opvolging na infectie [Internet]. 2022 [cited 2025 Jan 3]. Available from: https://www.sciensano.be/en/biblio/covimpact-studie-covid-19-infectie-en-de-lichamelijke-geestelijke-en-sociale-gevolgen-op-lange.
- 4.Soriano JB, Murthy S, Marshall JC, Relan P, Diaz JV, WHO Clinical Case Definition Working Group on Post-COVID-19 Condition. A clinical case definition of post-COVID-19 condition by a Delphi consensus. Lancet Infect Dis. 2022;22(4):e102–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.WHO, Post. COVID-19 condition (Long COVID) [Internet]. 2022 [cited 2025 Mar 6]. Available from: https://www.who.int/europe/news-room/fact-sheets/item/post-covid-19-condition.
- 6.Cavillot L, van Loenhout JAF, Devleesschauwer B, Wyndham-Thomas C, Van Oyen H, Ghattas J, et al. Sociodemographic and socioeconomic disparities in COVID-19 vaccine uptake in belgium: a nationwide record linkage study. J Epidemiol Community Health. 2023;78(3):176–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hubin P, Van den Borre L, Braeye T, Cavillot L, Billuart M, Stouten V, et al. Area and individual level analyses of demographic and socio-economic disparities in COVID-19 vaccination uptake in Belgium. Vaccine: X. 2024;18:100496. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Meurisse M, Lajot A, Devleesschauwer B, Van Cauteren D, Van Oyen H, Van den Borre L, et al. The association between area deprivation and COVID-19 incidence: a municipality-level spatio-temporal study in Belgium, 2020–2021. Archives Public Health. 2022;80(1):109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Bruyneel A, Dauvergne JE, Dauby N, Goffard JC, Rea A, Racape J. Social health gradient and risk factors among patients hospitalized for COVID-19 and pre-pandemic respiratory infections. A linked National individual case-control study in Belgium. Front Public Health. 2024;12:1426898. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cavillot L, Van den Borre L, Vanthomme K, Scohy A, Deboosere P, Devleesschauwer B, et al. Unravelling demographic and socioeconomic patterns of COVID-19 death and other causes of death: results of an individual-level analysis of exhaustive cause of death data in Belgium, 2020. Archives Public Health. 2024;82(1):209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Van der Heyden J, Charafeddine R, Gezondheidsenquête. 2018: Chronische ziekten en aandoeningen [Internet]. Sciensano; 2019 [cited 2025 Feb 10]. Available from: https://www.sciensano.be/en/biblio/gezondheidsenquete-2018-chronische-ziekten-en-aandoeningen-0.
- 12.Bambra C, Riordan R, Ford J, Matthews F. The COVID-19 pandemic and health inequalities. J Epidemiol Community Health. 2020;74(11):964–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Chung GKK, Chan SM, Chan YH, Yip TCF, Ma HM, Wong GLH, et al. Differential impacts of Multimorbidity on COVID-19 severity across the socioeconomic ladder in Hong kong: A syndemic perspective. Int J Environ Res Public Health. 2021;18(15):8168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Horton R, Offline. COVID-19 is not a pandemic. Lancet. 2020;396(10255):874. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Saqib K, Goel V, Dubin JA, VanderDoes J, Butt ZA. Exploring the syndemic impact of COVID-19 and mental health on health services utilisation among adult Ontario population. Public Health. 2024;236:70–7. [DOI] [PubMed] [Google Scholar]
- 16.Singer M, Bulled N, Ostrach B, Mendenhall E. Syndemics and the biosocial conception of health. Lancet. 2017;389(10072):941–50. [DOI] [PubMed] [Google Scholar]
- 17.Tran PB, Nikolaidis GF, Abatih E, Bos P, Berete F, Gorasso V, et al. Multimorbidity healthcare expenditure in belgium: a 4-year analysis (COMORB study). Health Res Policy Syst. 2024;22(1):35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Cès S, Baeten R. Inequalities in access to healthcare in Belgium | OSE [Internet]. 2020 [cited 2024 Oct 21]. Available from: https://www.ose.be/publication/inequalities-access-healthcare-belgium.
- 19.Drieskens S, Berete F, Van der Heyden J, Demarest S, Charafeddine R, Gisle L et al. sciensano.be. [cited 2025 May 16]. Gezondheidsenquête. 2018: Gebruik van gezondheidsdiensten. Samenvatting van de resultaten. Available from: https://www.sciensano.be/en/biblio/gezondheidsenquete-2018-gebruik-van-gezondheidsdiensten-samenvatting-van-de-resultaten.
- 20.Gerkens S, Lefèvre M, Brouckaerts N, Levy M, Maertens de Noordhout C, Obyn C et al. Performance of the Belgian health system: Report 2024. Health Services Research (HSR) Brussels Belgian Health Care Knowledge Centre (KCE). 2024;(KCE Reports 367 C).
- 21.Koumpias AM, Schwartzman D, Fleming O. Long-haul COVID: healthcare utilization and medical expenditures 6 months post-diagnosis. BMC Health Serv Res. 2022;22(1):1010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Scott A, Ansari W, Chambers R, Reimbaeva M, Mikolajczyk T, Benigno M, et al. Substantial health and economic burden of COVID-19 during the year after acute illness among US adults not at high risk of severe COVID-19. BMC Med. 2024;22(1):47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Scott A, Ansari W, Khan F, Chambers R, Benigno M, Di Fusco M, et al. Substantial health and economic burden of COVID-19 during the year after acute illness among US adults at high risk of severe COVID-19. BMC Med. 2024;22(1):46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Tene L, Bergroth T, Eisenberg A, David SSB, Chodick G. Risk factors, health outcomes, healthcare services utilization, and direct medical costs of patients with long COVID. Int J Infect Dis. 2023;128:3–10. [DOI] [PubMed] [Google Scholar]
- 25.Chambers LC, Park A, Cole M, Lovgren L, Zandstra T, Beaudoin FL, et al. Long-term health care costs following COVID-19: implications for pandemic preparedness. Am J Manag Care. 2023;29(11):566–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.DeMartino JK, Swallow E, Goldschmidt D, Yang K, Viola M, Radtke T, et al. Direct health care costs associated with COVID-19 in the united States. JMCP. 2022;28(9):936–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Formoso G, Marino M, Formisano D, Grilli R. Patterns of utilisation of specialist care after SARS-Cov-2 infection: a retrospective cohort study. BMJ Open. 2023;13(3):e063493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.McNaughton CD, Austin PC, Li Z, Sivaswamy A, Fang J, Abdel-Qadir H, et al. Higher Post-Acute health care costs following SARS-CoV-2 infection among adults in Ontario, Canada. J Multidiscip Healthc. 2024;17:5749–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Pike J, Kompaniyets L, Lindley MC, Saydah S, Miller G. Direct medical costs associated with Post-COVID-19 conditions among privately insured children and adults. Prev Chronic Dis. 2023;20:E06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Wolff Sagy Y, Feldhamer I, Brammli-Greenberg S, Lavie G. Estimating the economic burden of long-Covid: the additive cost of healthcare utilisation among COVID-19 recoverees in Israel. BMJ Glob Health. 2023;8(7):e012588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Rizal MF, Hafidz F, Kusila GR, Aisyiah W, Revelino D, Dhanalvin E, et al. Hospital services utilisation and cost before and after COVID-19 hospital treatment: evidence from Indonesia. PLoS ONE. 2024;19(7):e0305835. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Chua KP, Conti RM, Becker NV. Out-of-pocket spending for health care after COVID-19 hospitalization. Am J Manag Care. 2022;28(8):398–402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Łukomska E, Kloc K, Kowalska M, Matjaszek A, Joshi K, Scholz S. Healthcare resource utilization (HCRU) and direct medical costs associated with long COVID or Post-COVID-19 conditions: findings from a Literature Review. 2025. [DOI] [PMC free article] [PubMed]
- 34.HELICON [Internet]. 2020 [cited 2025 Feb 25]. HELICON. Available from: https://www.brain-helicon.be/index.php?option=com_content&view=article&id=2:home&catid=8.
- 35.De Pauw R, Van den Borre L, Baeyens Y, Cavillot L, Gadeyne S, Ghattas J, et al. Social inequalities and long-term health impact of COVID-19 in belgium: protocol of the HELICON population data linkage. BMJ Open. 2023;13(5):e069355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.De Mot L, Robalo Q, Dockx Y, Vandromme M, De Pauw R, Serrien B et al. COVID-19 Clinical Hospital Surveillance Report. Sciensano. 2023.
- 37.Van Goethem N, Vilain A, Wyndham-Thomas C, Deblonde J, Bossuyt N, Lernout T, et al. Rapid establishment of a National surveillance of COVID-19 hospitalizations in Belgium. Archives Public Health. 2020;78(1):121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Regulation – 2016/679 - EN -. gdpr - EUR-Lex [Internet]. [cited 2025 Jun 10]. Available from: https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng.
- 39.Guidelines 03/2020 on. the processing of data concerning health for the purpose of scientific research in the context of the COVID-19 outbreak | European Data Protection Board [Internet]. [cited 2025 Jun 10]. Available from: https://www.edpb.europa.eu/our-work-tools/our-documents/guidelines/guidelines-032020-processing-data-concerning-health-purpose_en.
- 40.Informatieveiligheidscomité Kamer Sociale Zekerheid en Gezondheid. Beraadslaging Nr. 22/014 van 11 Januari 2022 met betrekking tot de dededeling van persoonsgegevens die de gezondheid betreffen van de Healthdata COVID-19 Clinic Databank, Statbel en het Intermutualistisch Agentschap Aan Sciensano En L’université Libre De Bruxelles met het oog op het bestuderen van COVID-19 hospitalisaties in het kader van het HELICON-project [Internet]. 2022 [cited 2025 Nov 18]. Report No.: 1 IVC/KSZG/22/034. Available from: https://www.ehealth.fgov.be/ehealthplatform/file/cc73d96153bbd5448a56f19d925d05b1379c7f21/c05cb231fbb5c274bfc383d13c2750144e2fc899/22-014-n034-helicon-project.pdf.
- 41.Metadata. var SS00060-ZIV-bedrag [Internet]. [cited 2025 May 16]. Available from: https://metadata.ima-aim.be/nl/app/vars/SS00060_Gz.
- 42.Jurek AM, Greenland S, Maldonado G, Church TR. Proper interpretation of non-differential misclassification effects: expectations vs observations. Int J Epidemiol. 2005;34(3):680–7. [DOI] [PubMed] [Google Scholar]
- 43.Sittimart M, Rattanavipapong W, Mirelman AJ, Hung TM, Dabak S, Downey LE, et al. An overview of the perspectives used in health economic evaluations. Cost Eff Resour Alloc. 2024;22:41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Gorasso V, Van der Heyden J, De Pauw R, Pelgrims I, De Clercq EM, De Ridder K, et al. The health and economic burden of musculoskeletal disorders in Belgium from 2013 to 2018. Popul Health Metr. 2023;21:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.De kostprijs van. een ziekenhuisopname | RIZIV [Internet]. [cited 2025 May 16]. Available from: https://www.riziv.fgov.be/nl/thema-s/verzorging-kosten-en-terugbetaling/wat-het-ziekenfonds-terugbetaalt/verzorging-in-ziekenhuizen.
- 46.Choi J, Dekkers OM, le Cessie S. A comparison of different methods to handle missing data in the context of propensity score analysis. Eur J Epidemiol. 2019;34(1):23–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Berete F, Demarest S, Charafeddine R, Bruyère O, Van der Heyden J. Comparing health insurance data and health interview survey data for ascertaining chronic disease prevalence in Belgium. Arch Public Health. 2020;78(1):120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.IMA. EPS R13 – FLAGS Release 20190201 NL. 2019; Available from: https://ima-aim.be/IMG/pdf/eps_r13_-_flags_release_20190201_nl_-_vs2.pdf.
- 49.Chesnaye NC, Stel VS, Tripepi G, Dekker FW, Fu EL, Zoccali C, et al. An introduction to inverse probability of treatment weighting in observational research. Clin Kidney J. 2021;15(1):14–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Lukomska E, Kloc K, Kowalska M, Matjaszek A, Joshi K, Scholz S et al. Healthcare resource utilization (HCRU) and direct medical costs associated with long COVID or Post-COVID-19 conditions: findings from a. Literature Rev. 2025;13(1). [DOI] [PMC free article] [PubMed]
- 51.Directorate-General Healthcare of the Federal Public Service Health. Food Chain Safety and Environment. Key data in health care: COVID-19. 2022.
- 52.Gorasso V, Vandevijvere S, Van der Heyden J, Pelgrims I, Hilderink H, Nusselder W, et al. The incremental healthcare cost associated with cancer in belgium: A registry-based data analysis. Cancer Med. 2024;13(3):e6659. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Baymon DE, Vakkalanka JP, Krishnadasan A, Mohr NM, Talan DA, Hagen MB, et al. Race, Ethnicity, and delayed time to COVID-19 testing among US health care workers. JAMA Netw Open. 2024;7(4):e245697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Chen YJ, Jian WH, Liang ZY, Guan WJ, Liang WH, Chen RC, et al. Earlier diagnosis improves COVID-19 prognosis: a nationwide retrospective cohort analysis. Ann Transl Med. 2021;9(11):941. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Statbel. Working from home [Internet]. 2024 [cited 2025 Mar 11]. Available from: https://statbel.fgov.be/en/themes/work-training/labour-market/working-home#news.
- 56.Oakman J, Kinsman N, Stuckey R, Graham M, Weale V. A rapid review of mental and physical health effects of working at home: how do we optimise health? BMC Public Health. 2020;20(1):1825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Learoyd A, Nicholas J, Hart N, Douiri A. Application of information from external data to correct for collider bias in a Covid-19 hospitalised cohort. BMC Med Res Methodol. 2024;24(1):149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Epistat [Internet]. 2025. Overall roll-out of COVID-19 vaccination. Available from: http://epidata.sciensano.be/epistat/dashboard/#covid_vaccination.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Due to the General Data Protection Regulation (GDPR) legislations in Belgium, these data are not publicly available. Data requests must be addressed to the Information Security Committee (ISC).