Impact of COVID-19 pandemic on incident diagnosis patterns in German refugee centres: quasi-experimental study, 2018-2023

doi:10.21203/rs.3.rs-4122139/v1

Download PDF

Article

Impact of COVID-19 pandemic on incident diagnosis patterns in German refugee centres: quasi-experimental study, 2018-2023

https://doi.org/10.21203/rs.3.rs-4122139/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The COVID-19 pandemic may have affected morbidity patterns of residents in refugee centres, but empirical evidence is scarce. We utilised linked data from a health surveillance network in German refugee centres, employing a quasi-experimental design to examine the effects of the COVID-19 pandemic on newly diagnosed medical conditions. These diagnoses were coded in on-site healthcare facilities for refugee patients. Our analysis encompasses the timeframe from October 2018 to April 2023 and includes individual-level data for 109,175 refugees. This data resulted in 76,289 patient-months across 21 refugee centres, with a total occupancy of 144,012 person-months. We employed segmented regression analyses, adjusting for time trends, socio-demographic factors, centre occupancy, and centre-specific characteristics, to evaluate the COVID-19 pandemic's impact on incident diagnosis patterns among refugees. The COVID-19 pandemic significantly altered diagnosis patterns among refugees in German centres. Notably, incidents of injuries, mental disorders, psychotherapeutic drug prescriptions, and genitourinary diseases rose, while respiratory diseases decreased, later rebounding. An 88% increase in injury-related diagnoses suggests heightened violence experiences during flight or in centres. Mental disorder diagnoses and psychotherapeutic drug prescriptions rose by 73% and 95%, reflecting pandemic-related stressors in refugee centres, highlighting the pandemic's multifaceted impact on refugee health.

Health sciences/Health care/Public health/Epidemiology

Health sciences/Risk factors

The COVID-19 pandemic and related responses have impeded progress towards the Sustainable Development Goals (SDGs) worldwide¹. The pandemic revealed and exacerbated pre-existing inequalities, with migrants and refugees being particularly affected by multiple adverse conditions which put them at higher risk of suffering from COVID-19 or from the consequences of pandemic control measures^2,3. Despite early appeals by the World Health Organization to adequately consider migrants and refugees in pandemic response plans⁴, many countries side-lined migrants and refugees in their global health response^5–7.

Refugees in camps have been at particular risk of acquiring COVID-19 due to poor hygiene and over-crowding^4,8–10. Additionally, many health and immigration authorities responded to the COVID-19 pandemic with strict measures including mass quarantine, which refers to indiscriminate movement restrictions for all camp inhabitants and bans of in-and-out movements. Such measures have not only put refugees at higher risk of COVID-19 infection¹⁰, but also led to disruptions of essential health and social services¹¹, which in turn further increased the pre-existing isolation and marginalisation of refugees. Early guidance of the European Centre for Disease Prevention and Control stated that mass quarantine of refugee centres should be avoided due to the potential negative impact on refugees’ mental health⁷. Nevertheless, such practices were widespread. In Germany, for example, mass quarantine was not only used in the early pandemic phases but continued into 2022¹⁰. Combined with other pandemic control and mitigation measures⁷, life in refugee centres became even more regulated and subject to external control of authorities, severely affecting and interfering with refugees’ autonomy.

In Germany, pandemic measures for refugees involved compulsory PCR testing upon arrival, followed by mandatory quarantine, additional testing, and quarantine upon transfer. Measures included social distancing, mandatory mask-wearing, movement restrictions, and the reduction or suspension of on-site services and activities, such as leisure and shared spaces, to control and mitigate the spread of the pandemic.¹¹

Similar measures, including testing, individual and mass quarantine, and social distancing, were used in COVID-19 and other infectious disease containment efforts in refugee camps beyond the German context as well¹². While such measures are generally intended to protect refugees and prevent disease transmission, refugees may experience them as coercive¹³. This is exacerbated by often insufficient communication and coordination strategies of health and immigration authorities in charge of the centres^11,14. Furthermore, the COVID-19 pandemic led to disruptions of administrative processes of the asylum system, which in turn amplified uncertainty for refugees with precarious residence status or whose decision on their asylum application was pending.

The COVID-19 pandemic and its related responses, combined with disruptions of health and social services, had the potential to function as severe stressor for refugees in institutionalised settings such as camps or refugee centres. However, evidence on the impact of the COVID-19 pandemic on refugee populations is rare¹⁵. The lack of evidence has been attributed to pre-existing weaknesses of health monitoring systems¹⁶, which often fail - even in high-resource countries - to capture health of refugees in a manner that is both comprehensive and comparable across time and space¹⁷. Systematic reviews from early^2,3 and later stages of the pandemic ^15,18, as well as preliminary results¹⁹ of ongoing systematic reviews²⁰, indicate that there is a paucity of quantitative evidence on health impacts of the COVID-19 pandemic in refugee populations.

We used data from a novel multi-centre health surveillance network in German refugee centres²¹ to analyse the impact of the COVID-19 pandemic on incident diagnosis patterns among refugees between 2018 and 2023.

We studied the onset of the pandemic as a quasi-experimental situation, resembling an interrupted time series design, to perform a longitudinal analysis of changes in the incidence proportion of coded medical diagnoses among refugees seeking care in on-site healthcare facilities in refugee centres by using segmented negative binominal regression models with zero-inflation. We thereby considered pre- and peri-pandemic time trends, key socio-demographic characteristics of refugees, number of inhabitants as well as variations stemming from characteristics of refugee centres. We used data from an electronic health records (EHR) software, which is used by on-site healthcare providers across refugee centres in the network for the structured and standardised recording of diagnoses, symptoms, prescriptions and follow-up procedures. We used these individual-level data to determine the incidence of newly diagnosed conditions, and linked this information with aggregated data on the occupancy of refugee centres (i.e. number, age, and sex of inhabitants of refugee centres) provided by immigration authorities to obtain reliable denominators (see methods). The data covers individual-level data of 109,175 refugees (unique individuals), who visited on-site healthcare facilities at least once and of which we derived an analysis set comprising 76,289 patient-months from 21 refugee centres. The centres are located in three German federal states and comprised a total occupancy of 144,012 person-months of refugees during the observation period. We used March 2020 as break point in a segmented regression analysis to investigate the impact of the COVID-19 pandemic on a total of 21 indicators reflecting newly coded diagnoses on non-communicable diseases, physical conditions, mental conditions, and infectious diseases. Indicators were constructed using diagnosis categories based on International Classification of Diseases (ICD-10-GM Version 2021) and drug prescriptions based on the Anatomic Therapeutic Classification (ATC 2023) from the EHR. Further subsets of the available data sources were used to perform sensitivity analyses (details see methods).

Descriptive results

The main analysis period spanned 56 months (October 2018 – April 2023) across up to 21 refugee centres which were successively included in the surveillance network, yielding a total of 314 centre-months of observations on coded diagnoses. About one third of the observations (36%) stem from central registration- and reception-centres (REG), in which refugees are placed for several weeks to a few months. About two thirds (64%) was derived from peripheral reception centres (REC), to which refugees are transferred until a decision is made on their asylum application, and in which they stay up to 18 months or longer. Of the total occupancy of these centres (144,012 person-months), an average 64% were male and 75% adults (aged 18 years or above) (Table 1).

Table 1

Weighted sociodemographic characteristics of occupancy in refugee centres, 2018–2023, N = 144,012 person-months
	2018	2019	2020	2021	2022	2023	Total
Centre-months	6	14	9	11	12	10	62
Sociodemographic characteristics
% Male
mean ± sd	70 ± 7.6	63 ± 15	63 ± 15	61 ± 8.4	61 ± 12	68 ± 17	64 ± 13
median (Q1, Q3)	70 (63, 73)	66 (55, 76)	66 (49, 71)	64 (50, 68)	66 (53, 68)	72 (66, 81)	66 (58, 72)
min - max	60–81	34–82	40–87	48–71	37–74	36–87	34–87
CI	[62, 78]	[54, 71]	[51, 74]	[56, 67]	[53, 68]	[55, 80]	[60, 67]
W	67 ± 6.7	64 ± 11	64 ± 11	62 ± 8	65 ± 9.2	74 ± 12	64 ± 3.6
% Adults
mean ± sd	79 ± 6.2	78 ± 6.2	76 ± 12	71 ± 8	70 ± 8.9	79 ± 9	75 ± 9
median (Q1, Q3)	80 (76, 84)	77 (73, 82)	80 (68, 82)	72 (70, 78)	72 (69, 75)	81 (73, 88)	76 (70, 81)
min - max	70–86	68–89	58–94	53–81	50–81	65–90	50–94
CI	[73, 86]	[74, 81]	[67, 86]	[66, 77]	[65, 76]	[73, 86]	[73, 78]
W	78 ± 5.5	77 ± 6.1	77 ± 8.6	72 ± 6.7	73 ± 7.6	81 ± 9.1	76 ± 3.9
Average monthly number of inhabitants (occupancy)
mean ± sd	409 ± 374	306 ± 260	350 ± 257	323 ± 257	510 ± 539	529 ± 439	401 ± 367
median (Q1, Q3)	268 (224, 413)	256 (133, 377)	224 (156, 539)	228 (130, 520)	319 (246, 514)	328 (265, 768)	269 (175, 520)
min - max	131–1147	41–1058	132–884	62–903	201–2102	124–1540	41–2102
CI	[16, 801]	[156, 456]	[153, 548]	[151, 496]	[168, 853]	[215, 843]	[308, 494]
W	409 ± 374	306 ± 260	350 ± 257	323 ± 257	510 ± 539	529 ± 439	405 ± 96
Legend: SD: Standard deviation, Q1, Q3: first and third quartile, min: minimum, max; maximum and CI: 95% Confidence interval of the “weighted mean facility observation” values (weighted by ${n}_{occ}$); W 2018_2023: “weighted annual” mean ± standard deviation (weighted by the mean occupancy of a facility in the respective year); W total: mean value ± standard deviation of weighted annual mean values.

The least frequently coded medical diagnosis groups (Fig. 1) were “Diseases of the blood or blood-forming organs and certain disorders involving the immune mechanism” (ICD D50-D90) and “Neoplasms” (ICD C00-D48) with a (weighted) average incidence proportion of 0.3%, respectively. Notifiable infectious diseases (except COVID-19) according to section 36 of the German Infection Protection Act (IfSG) were also found among the least frequently coded indicators with an average (weighted) incidence proportion of 0.41%. The most frequently coded medical diagnoses groups were “Diseases of the respiratory system” (ICD J00-J99), followed by “Certain infectious and parasitic diseases” (ICD A00-B99)” and “Diseases of the digestive system” (ICD K00-K93) with an incidence proportion of 6.0%, 4.8% and 4.7%, respectively. These were followed by coded diagnoses on “Mental and behavioural disorders” (ICD F00-F99) with a (weighted) mean average incidence proportion of 4.0%. The mean (weighted) incidence proportion was 0.77% for prescriptions of psychoactive drugs (Fig. 1).

Means, standard deviations, median, minimum, and maximum as well as 95% confidence intervals (CI) of the weighted cumulative incidences of all 21 indicators stratified by year can be found in the Supplementary table 1.

Impact of the COVID-19 pandemic

The impact of the COVID-19 pandemic on the 21 health indicators can be found in Fig. 2. The mixed-effects zero-inflated negative binomial models were fitted to investigate the average difference in incidence proportions between peri-pandemic and pre-pandemic time periods as incidence rate ratios (IRR), while adjusting for the proportion of males and adults, secular trends (time as discrete variable), as well as potential influences of the characteristics of refugee centres, which we considered as random intercept. The impact was additionally evaluated by the mutually adjusted peri-pandemic time trend. The exact values of the IRR with 95% CI, p-values as well as the results of the fitted models can be found in Supplementary Chap. 1. Sensitivity analyses comprised further adjustment for differences within and between centres in underlying countries of origin over time, but this required using patients (instead of occupancy) as denominators as the information was available in the EHR but not in occupancy data (Supplementary Chap. 2).

The adjusted IRRs of conditions related to “Disabilities” (IRR: 1.88 [1.1–3.21), “Injury, poisoning and certain other consequences of external causes (S00-T98)” (IRR: 1.88 [1.4–2.53]), “Mental and behavioural disorders (F00-F99)” (IRR: 1.73 [1.27–2.35]), “Prescription rates of psychopharmaceutic drugs” (IRR: 1.95 [1.25–3.04]) as well as “Diseases of the genitourinary system (N00-N99)” (IRR: 1.51 [1.16–1.98]) were significantly higher in the peri-pandemic compared to the pre-pandemic time period. The IRRs remained elevated or at least marginally significant with lower point-estimates in sensitivity analyses which further adjusted for countries of origin (Fig. 2). The adjusted incidence of “Diseases of the respiratory system (J00-J99)” significantly declined (IRR: 0.51 [0.38–0.67]) in the peri-pandemic period, and even more under further adjustment for countries of origin (Fig. 2). Despite the use of different denominators for the calculation of incidence proportions, and different adjustment variables, results for these indicators were consistent.

Several indicators showed significant increases (IRR and 95% CI > 1.0) in the peri-pandemic period when using occupancy numbers as denominators, but the effects disappeared in the sensitivity analyses under adjustment for countries of origin and usage of patient numbers as denominator for calculating incidences (Fig. 2). Among these were “Endocrine, nutritional and metabolic diseases (E00-E90)”, “Diseases of the musculoskeletal system and connective tissue (M00-M99)”, “Neoplasms (C00-D48)”, and “Diseases of the ear and mastoid process (H60-H95)”. The results for these indicators were less consistent and amenable to variation after adjustment for countries of origin.

The incidence of “Certain infectious and parasitic diseases (A00-B99)” was non-significant when using occupancy numbers as denominator (IRR: 0.87 [0.7–1.09]), but showed a significant decline during the pandemic compared to pre-pandemic time periods when using patients as denominator and adjusting for countries of origin (IRR: 0.63 [0.41–0.95]). A similar pattern was observed for notifiable infectious diseases according to German national law (IfSG), for which the incidence was unaffected by the pandemic when using occupancy as denominator (IRR: 1.25 [0.83–1.88]) but turned significant with IRR < 1.0 in the sensitivity analysis with patients as denominator and further adjustment for countries of origin (Fig. 2).

No impact of the COVID-19 pandemic was found on all other indicators, which consistently showed non-significant results both in main and sensitivity analyses (Fig. 2) Note, that the model for the variable “Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism” (short label: Blood) did not converge in sensitivity analysis 1. Therefore, the results of this variable are given for the main analysis only).

Results of further sensitivity analyses using other data subsets with different inclusion or selection criteria confirmed the patterns described above and did not show substantially different model estimates (Supplementary Chap. 2.).

The peri-pandemic time trends were non-significant or only marginally significant for 17 of the 21 indicators in the main analysis (Fig. 2 and Supplementary Chap. 1.1.), indicating no major time trends beyond the mutually adjusted average impact of the pandemic. Only two indicators (Respiratory conditions, and Hypertension) showed significant peri-pandemic time trends over and above the average impact and the effects of co-variables: For each additional month in the peri-pandemic time period, the incidence of “Hypertensive disorders (I10-I15)” increased by 3%. “Diabetes Mellitus (E10-E14)” and conditions related to “Pregnancy, childbirth and the puerperium (O00-O99)” were marginally significant. Adjusting for country of origin (in sensitivity analyses with patients as denominator) resulted in non-significant time trends. While the average impact of the pandemic led to a decline in “Diseases of the respiratory system (J00-J99)”, the peri-pandemic time trends increased by 4% per each additional month in the peri-pandemic period (Fig. 2 and Supplementary Chap. 1.1.). In the sensitivity analysis, patterns remained unaffected when adjusting for country of origin

Counterfactual analysis of selected indicators

To further consolidate the findings of our analysis, we performed a counterfactual analysis by predicting the expected incidence proportions of analysed health indicators from the models, while considering the respective age- and sex distribution of refugees in the occupancy. The counterfactual measures reflect the incidence proportions that would have been expected if the COVID-19 pandemic had not happened. We visualised the expected versus the observed incidence as well as estimated counterfactual incidence of selected outcomes for which the above-mentioned estimates (for both the main analysis and sensitivity analysis) showed consistent results with either significantly elevated (Figs. 3–6) or reduced adjusted IRR as impact of the COVID-19 pandemic (Fig. 7).

The COVID-19 pandemic had a considerable impact on incident diagnosis patterns among refugees residing in refugee centres in Germany. Between October 2018 and April 2023, the incidence of “Injury, poisoning and certain other consequences of external causes”, “Mental and behavioural disorders”, “Prescriptions of psychotherapeutic drugs”, and “Diseases of the genitourinary system” significantly and consistently increased, holding sociodemographic and centre-related factors constant. While “Diseases of the respiratory system” decreased in average, related diagnoses showed a rising trend in the later peri-pandemic time period after the initial decline.

The 88% rise in diagnoses related to injury and consequences of external causes may indicate an increased experience of violence during flight or during refugees’ stay in the centres. Systematic reviews of media reports found that violent incidences in refugee centres during COVID-19 outbreaks were common. These were related to social tensions within the centres or to enforcement of the strict pandemic rules which partially required, or were related to, police operations¹⁰. Previous research also showed that refugees living in reception centres are at a higher risk of experiencing mental health disorders due to crowded living conditions, reduced autonomy, lack of privacy and other adverse experiences often associated with life in such centres^22–26. In addition, most reception centres, do not provide psychosocial care services, making it difficult for patients to access appropriate care^27,28. As a result, the estimated incidence of mental health disorders identified in our analysis (4% of patients) is likely to underestimate the true burden of mental health disorders. Our study also revealed a 73% increase in diagnoses related to mental disorders and a 95% increase in prescriptions of psychotherapeutic drugs following the onset of the pandemic. This suggests that the pandemic and related containment and mitigation measures have added to existing stressors, exacerbating the already challenging living conditions in refugee centres, and negatively impacting mental health.

Our analysis also showed a 51% increase in “Diseases of the genitourinary system” (N00-N99) following the onset of the COVID-19 pandemic. This finding contrasts with available studies, which have found a reduction in genitourinary care utilisation and related diagnoses²⁹. These were connected to general lockdowns but also to reductions in the scope, availability, and accessibility of genitourinary services often deprioritised in response to COVID-19 ³⁰. In contrast, with very few exceptions, care provision in the on-site health facilities of the refugee centres included in this study has continued throughout the pandemic at roughly constant levels. The reduction in genitourinary service provision in surrounding hospitals may have increased on-site care-seeking among residents in refugee centres, potentially contributing to the increase in genitourinary conditions found in our data. However, as access to routine care outside of the centres is always low due to logistical, administrative, and language barriers this effect is likely to be minor and other disease areas do not follow similar patterns. It is hence more likely that the increase in genitourinary conditions is due to early symptoms or complications^31,32 of infection with the Sars-CoV-2 virus. While our data do not include individual-level information on COVID-19 infections (due to parallel reporting systems), we are aware of several outbreaks in some of the studied refugee centres, particularly during the early stages of the pandemic. Several literature reviews have shown that the virus can lead to wide range of urogenital complications, including acute kidney injury and lower urinary as well as genital infections.^33–35

Our study shows that respiratory conditions considerably declined during the pandemic, mostly likely as a consequence of the strict pandemic control measures implemented in the refugee centres. This finding is in line with international studies in refugee populations³⁶ and national studies in the resident population in Germany³⁷. The rising time trend, on the other hand, may be an indication of the subsequent and sequential relaxation of measures over time which gradually individual behaviour change and lower adherence to and enforcement of social distancing or wearing of masks.

Overall, the results derived from the routine health surveillance network provide valuable insights into the temporal patterns of 21 health related indicators among refugees in one of the largest refugee-receiving countries in Europe. The identified pattern provides evidence of the detrimental impact of the COVID-19 pandemic on health, reflected by a rise in potential consequences of violence (and other external causes) and mental health problems. The only indicator that significantly and consistently declined during the pandemic related to respiratory conditions. We found no sufficiently robust evidence for effects on other indicators.

The strength of our analysis lay in the quasi-experimental situation, in which data was collected before and during the COVID-19 pandemic in a comparable and consistent way by means of unified EHR. We covered a time period of more than 4 years, and adjusted for potential influences on the outcomes related to individual and centre-related aspects. However, we had no information on provider-related variables, which means that we could not adjust for potential confounding caused by coding behaviour of health professionals. Our data was also limited by a lack of information on countries of origin when using occupancy data. We hence controlled for differences in the composition of the refugee population of the 21 centres with respect to countries of origin over time by using patient numbers as denominator. Despite the use of different data subsets, several of the estimates proved to be robust (Fig. 2). Another limitation of our data is the lack of information on length of stay in the centres, which in essence means that we cannot attribute a coded diagnosis to the situation in the centres directly. A condition may be diagnosed in the centre, but be obtained or attributable to experiences made during the migration process before emigration to Germany or before reaching the centre. However, movement and travel restrictions were very strict in the beginning of the pandemic, and restrictions were only gradually relaxed and lifted, so that we the risk of “conflating effects” due to high immigration is unlikely. This is confirmed by national asylum statistics, which shows that asylum applications comparatively low in 2020, gradually rising to pre-pandemic levels and above until 2023 ³⁸. Another limitation is that we had no data on confirmed COVID-19 cases from all centres. Despite the fact that the EHR contained a module to record COVID-19 testing and test results, only one centre used the module after the onset of the pandemic, while the other centres used parallel systems mandated by respective authorities or public health services to record or notify respective cases. The micro-level fragmentation of health information systems, and the scattered nature of data between immigration and health authorities ¹⁷, hence prevented us from considering and analysing the role of COVID-19 outbreaks in changing diagnoses patterns. The divide between health data governance and immigration data governance^17,39 was reflected in our study in terms of different data sources and mandates for health data versus occupancy data in refugee centres. This divide could be partially overcome by linking health data with aggregate data on age, sex, and total numbers of inhabitants of refugee centres which we prospectively obtained from authorities. However, direct or indirect linkage at individual level would have been beneficial and would have allowed for more robust analyses. This calls for enhanced linkage methods and approaches to overcome the fragmentation of data in the context of migration and health¹⁷.

Further limitations relate to the potential underestimation of incidences of rare events and/or incident diagnoses in small refugees centres due to our method of anonymisation, which required to set observations less than 3 to zero. While we sought to address this underestimation by applying zero-inflation models (i.e. treating zeros as ‘true’ values instead of missing data), underestimation cannot be ruled out. However, we accounted for this limitation by weighting our findings based on the size of the occupancy (or patient) population, so that centre-months with larger sample sizes or observations had a higher weight in the estimates. Finally, there were reliability issues with the occupancy data, due to manual entry by authorities and discrepancies between totals and age- and sex strata. However, our sensitivity analyses (Supplementary Chap. 2.) using different subsets of data show that this did not affect our estimates. The reliability problems in occupancy data underline the relevance of obtaining and capturing reliable denominator data for health studies in the context of forced migration ^40,41.

Overall, our study provides robust evidence for changing patterns of morbidity among refugees in refugee centres during the pandemic. Patterns changed to the considerable disadvantage of refugees, with higher incident disease burdens related to injury, violence, and mental health conditions. An exception was observed only for respiratory conditions, which declined most likely due to the implemented pandemic measures. Future pandemic preparedness and response strategies must better consider the specific conditions in refugee centres and mitigate the negative consequences of health emergencies. Further research is required to better understand the rise in genitourinary conditions.

Quasi-experimental study design

Our study resembles an interrupted time series (ITS)design⁴² with months per centre as observation units obtained from a retrospective (open) cohort of health surveillance data from the PriCarenet surveillance network. The impact of COVID-19 on incident diagnosis patterns among refugees was evaluated using a segmented regression approach.

Setting and data sources

The analysis was conducted within the framework of PriCarenet, a health surveillance network²¹. PriCarenet is overseen by the University Hospital Heidelberg and comprises healthcare providers operating healthcare facilities on-site as of October 2018. Since December 2023, these facilities are distributed across 24 state-level registration and reception centres, along with one district-level accommodation centre for refugees in Germany. These 25 centres are situated in the German states of Baden-Wuerttemberg, Bavaria, and Hamburg. These states collectively host approximately 30% of the asylum-seeking population in Germany, as determined by administrative quotas ⁴³. Within PriCarenet, healthcare providers are equipped with a customized Electronic Health Record (EHR) system known as Refugee Care Manager (RefCare). RefCare not only includes standard medical record-keeping features but also incorporates a built-in health surveillance module^21,44 (Table 2). The surveillance module comprises an automated analysis of locally stored medical routine data using predefined indicators. The indicators are constructed using diagnosis categories based on International Classification of Diseases (ICD-10-GM Version 2021) and drug prescriptions based on the Anatomic Therapeutic Classification (ATC 2023) as defined and outlined in Table 3, and operationalized through a standardised analysis script ^21,44. To protect data anonymity, any observations with counts less than 3 are adjusted to 0. More detailed information about the surveillance infrastructure in PriCarenet, and the local analysis of indicators can be found in previous reports ^21,44.

Table 2

Software features of the electronic health records “Refugee Care Manager” (RefCare):
Management Features	Patient Medical Records
Patient Management Task and daily lists External document storage User management External doctor management Facility and clinic data Local export of patient lists for follow-up	Record patient contact (patient history, clinical findings, diagnosis, therapy, etc.) Display and filter contact history Printable medication plan and immunisation status Generate doctors’ letters COVID-19 documentation Print function (e.g. for prescriptions) Patient interface for multilingual communication
Health Surveillance	Medical Records Transfer
On-site data analysis “at the click of a button” Review local results for planning purposes Export anonymised results for meta-analysis and reporting	Encrypted transfer of patient records between participating institutions Transfer of patient records to/from other facilities on request or in anticipation of patient transfer

Table 3

Indicator definitions based on diagnoses (ICD-10 Codes) and prescriptions (ATC-Codes) recorded in the electronic health record
Indicator labels	Indicator definition	Operationalisation (ICD-10 or ATC-Codes)
Indicators based on recorded diagnoses		ICD-10-Codes
Disability	Disabilities	H54, R47, H90-H91, H80-H82, Q71-Q73, M20-M21, Z89, G82, F06-F07, I68, P91, F7, F1
Skin	Diseases of the skin and subcutaneous tissue	L00-L99
Cons.ext.causes	Injury, poisoning and certain other consequences of external causes	S00-T98
Digestive syst.	Diseases of the digestive system	K00-K99
Blood	Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism	D50-D90
Inf.diseases	Certain infectious and parasitic diseases	A00-B99
Inf.notify	Notifiable infectious diseases	B30.0, B30.1, A05.1, A23.0, A23.1, A23.3, A23.8, A23.9, A04.5, A92.0, A00, A81.0, A97, A36, A98.4, A04.4, B67, A04.3, A75.0, A84.1, A95, A07.1, A41.3, A49.2, G00.0, J09, J14, J20.1, P23.6, A98.5, B15, B16, B17.1, B18.2, B19, B16.0, B16.1, B17.0, B17.2, B17.8, B20-B24, D59.3, M31.1, J09, J10, J11, A37, A07.2, A96.2, A68.0, A48.1, A48.2, A30, A27, A32, P37.2, B50-B54, A98.3, B05, A39, A41.0, A49.0, G00.3, P36.2, A22, B26.8, B26.9, A08.1, A70, A01.1, A01.2, A01.3, A01.4, A20, A80, A78, A08.0, P35.0, B06.8, B06.9, A0, A03, A50, A53, A82, Z20.3, P37.1, B75, A15 - A19, P37.0, O98.0, A21, A01.0, A92.0, A92.4, A96, A98.0, A98.1, A99, B02, P35.8, A04.6
Circulatory syst.	Diseases of the circulatory system	I00 – I99
Hypertension	Hypertension	I10-I15
Metabolic	Endocrine, nutritional and metabolic diseases	E00-E90
Diabetes	Diabetes mellitus	E10-E14
Musculoskelet. syst.	Diseases of the musculoskeletal system and connective tissue	M00-M99
Neoplasm	Neoplasms	C00-D48
Nervous syst.	Diseases of the nervous system	G00-G99
Ear.mastoid	Diseases of the ear and mastoid process	H60-H99
Eye.adnexa	Diseases of the eye and adnexa	H00-H59
Pregn.condition	Pregnancy, childbirth and the puerperium	O00-O99
Psych.condition	Mental and behavioral disorders	F00-F99
Genitourinary syst.	Diseases of the genitourinary system	N00-N99
Respiratory syst.	Diseases of the respiratory system	J00-J99
Indicators based on recorded prescriptions		ATC-Codes
Psych. prescrip.	Psychoactive drug prescriptions	N05, N06A, N06B, N06C, N07BB
Legend: ICD-10: International Classification of Disease. ATC: Anatomical Therapeutic Chemical Classification.

The data used in this paper covers the time span from October 2018 to April 2023. The facilities included in this study joined the surveillance network at different dates (Supplementary Chap. 3.). Some centres have since departed from the network due to closures or changes in healthcare providers, but still contributed their anonymous health surveillance data for the purpose of this study. Provided data consequently varies per centre (Supplementary Chap. 3.).

RefCare is used by health professionals, who are the data holders of the individual-level patient data in on-site health care facilities. The respective authorities in the three federal states are responsible for immigration data, and are data holders of the occupancy data, i.e. the sociodemographic information of the refugee centres’ inhabitants.

The flowchart in Fig. 8 provides an overview of the data selection process, the nature of used data sources and the four derived data subsets (Subset 1–4).

Electronic Health Records (RefCare) data set (subset 1)

Using the 25 refugee centres and months as units of analysis, subset 1 contains 833 observations (i.e. 833 “centre-months”) of recorded medical data with an average of ${mean(n}_{pat})=259$ (standard deviation $sd\left({n}_{pat}\right)=287$) patient-months. The sample comprised 215.864 patient-months (= $\sum _{i=1}^{833}{n}_{pat}^{i}$, where ${n}_{pat}^{i}$ is the number of refugee patients of “centre-month” $i$) of a total of 109.175 refugee patients between October 2018 and April 2023 (Fig. 8). For these 833 centre-months, we used reported monitoring data on the number of male, female, adult (≥18 years of age) and underage (<18 years of age) patients; as well as data on the incident coding of diagnoses for 21 indicators (based on ICD-10 Codes) by centre and month. Furthermore, in sensitivity analysis 1 we used data on the country of origin of the patients from the EHR to run models which account for compositional differences in the refugee population within and between refugee centres over time (Supplementary Chap. 2.1.). Estimates for the COVID-19 impact from this sensitivity analysis are reported in Fig. 2.

Occupancy data and aggregate-level socio-demographics

Furthermore, we gathered information on occupancy of each refugee centre within the PriCarenet surveillance network through a monthly online survey conducted with the responsible authorities of these centres. This prospective census survey was initiated in October 2018 and encompasses count data concerning the number of residents on the 15th day of each respective month, categorized by age (adults: ≥18 years and children: ≤18 years), and biological sex (male/female) (Fig. 8). To determine the total occupancy of each centre for every month, we combined the reported counts of male and female adults separately for the adult population and likewise for the children. These cumulative counts of children and adults were then summed to calculate the total occupancy for each centre and month. Furthermore, the overall (unstratified) number of the occupancy was collected.

Participation of authorities in this survey is voluntary. We collected occupancy data from 22 centres, resulting in a comprehensive dataset covering 417 centre-months spanning from October 2018 through June 2023 (Fig. 8). The average occupancy stands at ${mean(n}_{occ})=411$ individuals per centre per month, with a standard deviation of $sd\left({n}_{occ}\right)=435$.

Description of derived datasets and variables

We matched the EHR data with the monthly occupancy data for each centre, wherever possible (Fig. 8). In 64 cases, the occupancy count was lower than the number of patients (i.e., ${\text{n}}_{\text{o}\text{c}\text{c}}<{\text{n}}_{\text{p}\text{a}\text{t}}$). This occurrence is reasonable in situations where refugee centres experience a rapid turnover of individuals, such as a high influx of new arrivals and frequent transfers. In such instances, individuals may seek on-site healthcare services but stay within the centres for only a brief period, leading to a temporary misalignment between occupancy figures and the number of patients receiving healthcare services. These observations were excluded for the main analysis which resulted in a total of 314 centre-months between October 2018 and April 2023 of 21 centres (with ${mean(n}_{pat})=243$, $sd\left({n}_{pat}\right)=240$, ${mean(n}_{occ})=459$ and $sd\left({n}_{occ}\right)=462$; subset 2).

In 75 cases, the sum of the reported strata counts (female/male x adult/children) did not equal the reported total occupancy. Therefore, we repeated the main analysis on subset 3 (sensitivity analysis 2), where the occupancy totals equal the totals in occupancy age-and sex-strata AND ${\text{n}}_{\text{o}\text{c}\text{c}}\ge {\text{n}}_{\text{p}\text{a}\text{t}}$. (Supplementary Chap. 2.2.). Furthermore, in sensitivity analysis 3, we repeated the main analysis again (which was performed on subset 2), but instead used subset 4 of the linked data which contained no restrictions, i.e. all observations of the linked dataset (Supplementary Chap. 2.3.).

Furthermore, we calculated the following variables (for each subset respectively):

time: discrete variable indicating time from the start up to the end of the observation period October 2018 to April 2023 with time ID = {1, …, 56}
covid: coded 0 for pre-covid time points and 1 for post-covid time points (0: < March 2020, 1: $\ge$ March 2020). This variable captures the impact of the COVID-19 pandemic in peri-pandemic time periods, with pre-pandemic time periods used as reference.
postslope: coded 0 up to the last point before COVID-19 and coded sequentially from 1 thereafter (0: < March 2020, 1: March 2020, 2: April 2020, …, 37: April 2023). This variable captures the peri-pandemic time trend.

It should be noted that there are two levels in the data: months and centres. As a result, there are multiple observations of these levels per year. Therefore, in order to report the mean incidence (Table 1), we determined weighted mean values averaging over the months for each facility per year, so that there is only one observation per year of a facility. The weighting was based on ${n}_{occ}$, i.e. months with high occupancy were assigned a higher weighting when calculating the weighted mean incidence for each facility (“weighted mean facility observation”). Table 1 shows the mean value with standard deviation (mean+-sd), median with 25th and 75th quartiles (Q1, Q3), minimum and maximum (min - max) and 95% confidence interval (CI) weighted mean values of facility observations. Furthermore, the annual weighted mean value and weighted standard deviation are given, whereby the weighting was accordingly to the mean occupancy of a facility within one year. That is, if the mean occupancy of a facility in one year is higher, the observation weighs more (“weighted annual”: mean and standard deviation of the “mean facility observation” values within one year weighted by the mean occupancy of the respective facility; compare row W, 2018–2023). Additionally, the mean value and the standard deviation of the weighted annual mean values were calculated (row W, last column).

Description of the regression model

In order to assess the impact of the COVID-19 pandemic on the incident health indicators we fitted a negative binominal model with zero-inflation model on the matched data for each indicator. The model allows the conditional mean to depend on the percentage of adult and male occupancy, overall number of occupancy (${\text{n}}_{\text{o}\text{c}\text{c}}$) as well as randomly on centres, while β₀ captures the baseline level of the outcome at time 0 (beginning of the observation period), β_time estimates the structural trend or growth rate, independently from COVID-19, β_covid estimates the immediate impact of COVID-19 or the change in the outcome of interest after COVID-19 and β_postslope reflects the change in the trend or growth rate in the outcome after COVID-19. Furthermore, the model assumes structural zeros (Supplementary Chap. 1.). The model can be represented by the following set of equations:

$${\mu }=\text{E}\left(\text{c}\text{o}\text{u}\text{n}\text{t}|u,\text{N}\text{S}\text{Z}\right)=\text{exp}\left({{\beta }}_{0}+{{\beta }}_{\text{a}\text{d}\text{u}\text{l}\text{t}}+{{\beta }}_{\text{m}\text{a}\text{l}\text{e}}+{{\beta }}_{{\text{n}}_{\text{o}\text{c}\text{c}}}+ {{\beta }}_{\text{t}\text{i}\text{m}\text{e}}+{{\beta }}_{\text{c}\text{o}\text{v}\text{i}\text{d}}+{{\beta }}_{\text{p}\text{o}\text{s}\text{t}\text{s}\text{l}\text{o}\text{p}\text{e}}+u\right),$$

$$u\sim\mathcal{N}(0, {\sigma }_{u}^{2})$$

$${{\sigma }}^{2}=\text{V}\text{a}\text{r}\left(\text{c}\text{o}\text{u}\text{n}\text{t}|u, \text{N}\text{S}\text{Z}\right)= {\mu }\left(1+\frac{{\mu }}{{\theta }}\right),$$

$$\text{l}\text{o}\text{g}\text{i}\text{t}\left(\text{p}\right) = {{\beta }}_{0}^{\left(\text{z}\text{i}\right)}$$

where u is a centre specific random effect, $\text{N}\text{S}\text{Z}$ is the event “non-structural zero”, $\text{p}=1-\text{P}\text{r}\left(\text{N}\text{S}\text{Z}\right)$ is the zero-inflation probability and ${\beta }$’s are the regression coefficients with subscript denoting the covariate and with 0 denoting the intercept ⁴⁵. The chosen parameterization of the negative binomial uses a logarithmic link and denotes the variance increasing quadratically with the mean as ${{\sigma }}^{2}= {\mu }(1 + {\mu }/{\theta })$, with ${\theta }>0$ ⁴⁶ (Supplementary Chap. 1). The analysis was performed with R-programming language using the glmmTMB-package ⁴⁵.

Counterfactual analysis

We performed a counterfactual analysis by predicting the expected values of the 21 health indicators given that the pandemic had not happened (variable covid set at “0”) while considering the socio-demographic characteristics of the underlying refugee population in respective centres and time periods. We plotted the estimated counterfactual, observed, and estimated outcome values given by incidence rates in percent (i.e. the number of cases divided by occupancy and multiplied by 100) of selected indicators together in box plots over the observation period (compare Figs. 3–7).

Acknowledgements

We would like to thank all participating reception centres for asylum seekers for supporting the establishment of the PriCare Surveillance Network (PriCareNet-Consortium).

Author contributions

KB conceived the study. KB and SE designed the statistical methodology. SE, SR, KB, RJ collected and curated the data. SE prepared the data, conducted the analysis and created figures and tables. KB and SE wrote the first and final draft of the manuscript. SR and RJ reviewed and edited revisions for important intellectual content. KB, RJ, SR verified data analysis and validated the findings. All authors (KB, SE, SR, RJ) have access to the data in the study and had final responsibility for the decision to submit for publication.

Competing interests

Financial competing interests: The authors acknowledge research support and institutional funding received by the German Federal Ministry of Health in line with a resolution passed by the German Bundestag (Grant no: 2516FSB415, Grant holder: KB) in the period 2016-2020. Funds were received for salaries and equipment to develop, validate and implement the surveillance methodology, technology, and infrastructure. We acknowledge further funding received by the State Ministry of Justice and Migration (Baden-Württemberg) and Regional Authorities in Bavaria as well as care provider organisations (Klinikum Würzburg, St. Joseph Klinik, MKT) for operational running costs of the use and implementation of the electronic medical records software RefCare in the scope of the surveillance network within a non-for-profit licensing model (Grant holder: KB). The funders had no role in design, analysis, or interpretation of data or in the decision to publish.

Personal financial interests: KB and RJ are registered at University Hospital Heidelberg as co-inventors of the electronic medical records software RefCare in line with the Employee Invention Act (ArbnErfG). The invention is related to the underlying software and concept for surveillance, without receiving any individual financial benefits from licenses or use and implementation of the software.

Non-financial competing interests: None declared.

The authors declare no competing interests.

Data availability

The datasets generated and/or analysed during the current study are not publicly available due to the data-use and -access (DUAC) regulations of the PriCareNet Consortium. The generated and analysed datasets are available for scientific purposes from the PriCareNet Consortium upon reasonable request by contacting the spokesperson (Kayvan Bozorgmehr, [email protected]).

Li, C., et al. Responses to the COVID-19 pandemic have impeded progress towards the Sustainable Development Goals. Communications Earth & Environment 4, 252 (2023).
Hintermeier, M., et al. SARS-CoV-2 among migrants and forcibly displaced populations: a rapid systematic review. Journal of migration and health, 100056 (2021).
Hayward, S.E., et al. Clinical outcomes and risk factors for COVID-19 among migrant populations in high-income countries: A systematic review. Journal of Migration and Health 3, 100041 (2021).
Kluge, H.H.P., Jakab, Z., Bartovic, J., D'Anna, V. & Severoni, S. Refugee and migrant health in the COVID-19 response. The Lancet 395, 1237-1239.
Bozorgmehr, K. Power of and power over COVID-19 response guidelines. The Lancet 396, e67 (2020).
Knipper, M., Bozorgmehr, K., Offe, J. & Orcutt, M. Germany's new global health strategy: leaving migrants behind. The Lancet 397, 20-21 (2021).
ECDC. Guidance on infection prevention and control of COVID-19 in migrant and refugee reception and detention centres in the EU/EEA and the UK. (European Centre for Disease Prevention and Control Stockholm, 2020).
Kondilis, E., et al. The impact of the COVID-19 pandemic on refugees and asylum seekers in Greece: A retrospective analysis of national surveillance data from 2020. EClinicalMedicine 37, 100958 (2021).
Kondilis, E., Puchner, K., Veizis, A., Papatheodorou, C. & Benos, A. Covid-19 and refugees, asylum seekers, and migrants in Greece. bmj 369(2020).
Jahn, R., Hintermeier, M. & Bozorgmehr, K. SARS-CoV-2 attack rate in reception and accommodation centres for asylum seekers during the first wave: systematic review of outbreak media reports in Germany. Journal of Migration and Health, 100084 (2022).
Biddle, L., et al. COVID-19 in collective accommodation centres for refugees: assessment of pandemic control measures and priorities from the perspective of authorities. Bundesgesundheitsblatt, Gesundheitsforschung, Gesundheitsschutz, 1 (2021).
Führer, A., Özer Erdogdu, I., Kompa, P., Yilmaz-Aslan, Y. & Brzoska, P. COVID-19 pandemic in shelters for asylum seekers: a scoping review of preventive measures. BMJ Open 12, e058076 (2022).
Rast, E., Perplies, C., Biddle, L. & Bozorgmehr, K. Between Care and Coercion: Asylum Seekers’ Experiences With COVID-19 Containment and Mitigation Measures in German Reception Centres. International Journal of Public Health 68(2023).
Gold, A.W., et al. [Early crisis response to the COVID-19 pandemic in collective accommodation facilities for refugees : Analysis of cross-actor working arrangements from the perspective of the reception authorities in Germany]. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz (2023).
Hintermeier, M., et al. COVID-19 among migrants, refugees, and internally displaced persons: systematic review, meta-analysis and qualitative synthesis of the global empirical literature. medRxiv, 2023.2008.2003.23293586 (2023).
Bozorgmehr, K., Biddle, L., Rohleder, S., Puthoopparambil, S. & Jahn, R. What is the evidence on availability and integration of refugee and migrant health data in health information systems in the WHO European Region? in Health Evidence Network (HEN) Synthesis Report 66 (WHO Regional Office for Europe, Copenhagen, 2019).
Bozorgmehr, K., et al. Integration of migrant and refugee data in health information systems in Europe: advancing evidence, policy and practice. The Lancet Regional Health – Europe 34(2023).
Hintermeier, M., et al. Refugees, migrants, internally displaced people and COVID-19: protocol for an updated systematic review. (2021).
Bozorgmehr, K. & Profili, C. 8.Q. Workshop: Covid-19 and pandemic measures: reducing or reproducing social and health inequities among migrants? European Journal of Public Health 33(2023).
Bozorgmehr, K., et al. Unintended consequences of public health and social measures implemented in the setting of reception and collective accommodation centres for refugees to prevent SARS‐CoV‐2 infection and its consequences. (York).
Jahn, R., et al. Health monitoring of refugees in reception centres for asylum seekers: Decentralized surveillance network for the analysis of routine medical data. Journal of Health Monitoring 6(2021).
Ziersch, A. & Due, C. A mixed methods systematic review of studies examining the relationship between housing and health for people from refugee and asylum seeking backgrounds. Social Science & Medicine 213, 199-219 (2018).
Rast, E., Hintermeier, M., Bozorgmehr, K. & Biddle, L. Housing and health: A multidimensional, qualitative analysis of the experiences of asylum seekers and refugees living in German reception centres. SSM - Qualitative Research in Health 5, 100407 (2024).
Mohsenpour, A., et al. Type of Refugee Accommodation and Health of Residents: A Cross-Sectional, Population-Based Cluster Analysis in South-West Germany. Int J Public Health 68, 1605786 (2023).
Bozorgmehr, K. Assessing Refugee Accommodation: From Broken Windows Index to Heterotopic Spaces. in Refugee Camps in Europe and Australia: An Interdisciplinary Critique (eds. Razum, O., Dawson, A., Eckenwiler, L. & Wild, V.) 65-78 (Springer International Publishing, Cham, 2022).
Mohsenpour, A., Biddle, L. & Bozorgmehr, K. Exploring contextual effects of post-migration housing environment on mental health of asylum seekers and refugees: A cross-sectional, population-based, multi-level analysis in a German federal state. PLOS Global Public Health 3, e0001755 (2023).
Wahedi, K., et al. [Healthcare provision for asylum seekers in reception centres: Qualitative survey from the perspective of healthcare providers]. Bundesgesundheitsblatt-Gesundheitsforschung-Gesundheitsschutz 63, 1460-1469 (2020).
Karato, Y. Flucht & Gewalt: Psychosozialer Versorgungsbericht Deutschland 2023. 57 (2023).
Rajwa, P., et al. Changing patterns of urologic emergency visits and admissions during the COVID-19 pandemic: a retrospective, multicenter, nationwide study. Arch Med Sci 17, 1262-1276 (2021).
Morlacco, A., Motterle, G. & Zattoni, F. The multifaceted long-term effects of the COVID-19 pandemic on urology. Nature Reviews Urology 17, 365-367 (2020).
Bani-Hani, M., et al. Evaluation of lower urinary tract symptoms among male COVID-19 patients during the second wave: An observational study. Urol Ann 14, 372-376 (2022).
Tayebi, S., et al. Lower Urinary Tract Symptoms (LUTS) and COVID-19. Continence Reports 9, 100044 (2024).
Ebner, B., Volz, Y., Mumm, J.-N., Stief, C.G. & Magistro, G. The COVID-19 pandemic — what have urologists learned? Nature Reviews Urology 19, 344-356 (2022).
Chan, V.W.-S., et al. A systematic review on COVID-19: urological manifestations, viral RNA detection and special considerations in urological conditions. World Journal of Urology 39, 3127-3138 (2021).
Inouye, D., Baker, Z., Peña, A. & Vasquez, E. Epididymitis, orchitis, and epididymo-orchitis associated with SARS-CoV-2 infection in pediatric patients: A systematic review. Frontiers in Urology 2(2023).
Altare, C., et al. COVID-19 epidemiology and changes in health service utilization in Azraq and Zaatari refugee camps in Jordan: A retrospective cohort study. PLoS medicine 19, e1003993 (2022).
Ullrich, A., et al. Impact of the COVID-19 pandemic and associated non-pharmaceutical interventions on other notifiable infectious diseases in Germany: An analysis of national surveillance data during week 1–2016 – week 32–2020. The Lancet Regional Health – Europe 6(2021).
BAMF. Aktuelle Zahlen (11/2023). (Nuernberg, 2023).
Wickramage, K. & Annunziata, G. Advancing health in migration governance, and migration in health governance. The Lancet 392, 2528-2530 (2018).
Correa-Velez, I. & Gifford, M. When the right to be counted doesn't count: the politics and challenges of researching the health of asylum seekers. Critical Public Health 17, 273-281 (2007).
Bozorgmehr, K., et al. Integration of migrant and refugee data in health information systems in Europe: advancing evidence, policy and practice. The Lancet Regional Health – Europe 34(2023).
Reeves, B.C., Wells, G.A. & Waddington, H. Quasi-experimental study designs series-paper 5: a checklist for classifying studies evaluating the effects on health interventions-a taxonomy without labels. J Clin Epidemiol 89, 30-42 (2017).
GWK. Königsteiner Schlüssel. . (Gemeinsame Wissenschaftskonferenz (GWK), Bonn, 2017).
Nöst, S., et al. Health and primary care surveillance among asylum seekers in reception centres in Germany: concept, development, and implementation. Bundesgesundheitsblatt, Gesundheitsforschung, Gesundheitsschutz 62, 881-892 (2019).
Brooks, M.E., et al. glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. The R journal 9, 378-400 (2017).
Hardin, J.W., Hardin, J.W., Hilbe, J.M. & Hilbe, J. Generalized linear models and extensions, (Stata press, 2007).

There is NO Competing Interest.

supplementaryinformationtableS1final.xlsx
Supplementary table 1
supplementaryinformationtableS2final.xlsx
Supplementary table 2
supplementaryinformationtableS3final.xlsx
Supplementary table 3
supplementaryinformation.pdf
Supplementary information

Download PDF

Version 1

posted

You are reading this latest preprint version

Impact of COVID-19 pandemic on incident diagnosis patterns in German refugee centres: quasi-experimental study, 2018-2023

Status:

Version 1

Abstract

Figures

Introduction

Results

Descriptive results

Impact of the COVID-19 pandemic

Counterfactual analysis of selected indicators

Discussion

Methods

Declarations

Acknowledgements

Author contributions

Competing interests

Data availability

References

Additional Declarations

Supplementary Files

Status:

Version 1