Impact of long COVID on health-related quality-of-life among Japanese adults: findings of CARE Japan study Authors: Sabera Sultana, Yusuke Asai, Haruhiko Ishioka, Shinichiro Ikeda, Anna Ohtera, Nobuaki Matsunaga, Norio Ohmagari, Shinya Tsuzuki Published: 2026 ABSTRACT AbstractObjectivesThis study aimed to comprehensively assess the implications of long COVID on the health-related quality-of-life (HRQoL) among Japanese adults and to identify its associated factors.MethodsThis study used prospective cohort data from the CARE Japan Study between January 2022 and January 2023. The outcome of this study was HRQoL, which was measured using SF-12 questionnaire. Self-reported long COVID was the primary independent variable. We fitted adjusted beta regression models to calculate beta regression coefficients with 95% confidence intervals (CI) and average marginal effects (AME) to explore the determinants of HRQoL. We also performed latent class analysis (LCA) to identify unobserved patterns of long COVID symptoms.ResultsFinal sample size was 1,285. Compared to the participants with no long COVID, the HRQoL among long COVID patients (β: −0.25; 95% CI: −0.36 to −0.14; AME: −0.036) was significantly lower. The effect of long COVID on HRQoL was the most pronounced among the respondents with pre-existing lung diseases (β: −0.72; 95% CI: −1.29 to −0.16; AME: −0.114). LCA identified three subgroups of long COVID patients - class 1, 2, and 3. Compared to the participants with no long COVID, participants belonged to class 1 (β: −0.47; 95% CI: −0.57 to −0.36; AME: −0.065), class 2 (β: −0.48; 95% CI: −0.60 to −0.35; AME: −0.066), and class 3 (β: −0.93; 95% CI: −1.06 to −0.79; AME: −0.148) had poorer HRQoL.ConclusionsLong COVID patients had reduced HRQoL. Female gender, young-age, thin BMI or pre-existing psychological disorders were associated with lower HRQoL.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11136-026-04284-y.Keywords: Long COVID symptoms, Health-related quality-of-life, Post-COVID conditions, Post-COVID-19 syndrome, Post-acute sequelae of COVID-19, SARS-CoV-2 infection, Japanese adult population, Psychological disorder, Latent class analysis Introduction Many of the survivors of COVID − 19 suffer from a complex, multisystemic persistent or relapsing symptoms for weeks, months, or even years after the initial infection of COVID-19. When the symptoms are persistent or relapsing even after 12 weeks of initial acute infection of COVID-19, it is termed as long COVID [1]. One meta-analysis with 144 contributing studies reported that, the global pooled long COVID prevalence was 36% (95% confidence interval: 33%–40%) as of 2025 [2]. Common long COVID symptoms include extreme tiredness, anxiety, memory problems, dizziness, taste or smell issues, insomnia, dyspnoea, cough, headache, palpitation, and digestive problems [2–4]. Increasing evidence across the globe is suggesting that long COVID is adversely affecting physical, emotional, and social aspects of life of the patients [5–7]. OpenPROMPT, a UK cohort study in 2024 that measured the impact of long COVID on health-related quality-of-life (HRQoL) in 6,070 participants found a consistent impact on HRQoL, with a higher odd of reporting loss in quality-of-life (OR: 4.7, 95% CI: 3.72–5.93) among people with long COVID compared with those who did not have long COVID [6]. Several studies were conducted to measure the effect of long COVID in the quality of life among Japanese adults. Tsuzuki and colleagues conducted a cross-sectional study in 526 patients in 2021 on long COVID and HRQoL [8] and Honda and colleagues conducted another cross-sectional study in 2022 [9]. Both of these studies found lower HRQoL among patients with long COVID but they used older definition of long COVID (symptoms persistent for 4 weeks or more). Yagi and colleagues conducted a longitudinal study on long COVID and HRQoL in 986 patients with COVID-19 with 1 year of follow-up data but the data were collected between January 2020 and February 2021 when Wuhan type of COVID-19 was predominant in Japan [10]. Maeda and colleagues ran another longitudinal study on long COVID and HRQoL starting from April 2023 when Omicron variant of COVID-19 was predominant in Japan but the sample size was only 379 and the follow-up period was only 6 months [11]. Therefore, the existing studies had small sample size with short duration of follow-up period. Further, among long COVID patients which factors are affecting which aspects of life and the effects of long COVID symptoms in clusters on HRQoL has not been studied comprehensively. Owing to this knowledge gap, our study aimed to comprehensively assess the impact of COVID-19 infection on the HRQoL among Japanese adults after three months of acute phase of infection using monthly follow-up data of up to one year and to identify psychosocial factors associated with the HRQoL after COVID-19 infections. Methods Data source We used prospective cohort data from the CARE Japan Study (https://japan.helpstopcovid19.com/), which has been conducted by IQVIA Japan in a national center hospital in Tokyo since January 2021. In this study, we used a non-probability convenience sampling method. Patients who visited the hospital with flu-like symptoms, including fever, cough, sore throat, chill, body aches, runny nose etc. were provided an explanation about the purpose of this study and were invited to participate. Upon collecting patients’ consent via Internet, baseline data were collected. Patients were then e-mailed questionnaires every month for follow-up. The CARE Japan Study collects data on participants’ demographics at baseline. Information on diagnostic test results for COVID-19, vaccination status, symptoms related to COVID-19, and SF-12 questionnaire was collected at baseline and at each 4–6 weeks during the one year of follow-up period. In this study we used IQVIA data collected between January 2022 and January 2023. Study design and participant selection criteria This was a longitudinal study conducted among Japanese male and female adults. After initial registration in the CARE Japan Study, participants were excluded if they did not have a history of COVID-19 infection, had less than 90 days of follow-up after the initial infection, or had missing covariate data (< 1%). If any participant had a second COVID-19 infection at any time point within the 1-year follow-up period, the participant was excluded from the study to avoid biases in outcome. No imputation was performed, as the proportion of missing covariate data was small. Figure 1 shows the participant selection criteria in this study. Participant selection flow-chart Dependent variable The outcome of this study was HRQoL, which was a patient reported outcome measure. 12-item Short Form (SF-12) questionnaire measured patients’ perceived and reported outcomes reflecting several domains of functioning, including physical functioning (limitations in physical activities due to health problems), role limitations (impact of physical functioning limitations & emotional problems on the activities of daily life), bodily pain, vitality (energy levels and fatigue and perception of general health), social functioning (impact of physical or emotional problems on social activities), and mental health (psychological well-being, including feelings of anxiety and depression). The score of each domain was weighted following the methods described by Braizer (2004) and combined to produce the HRQoL score [12]. We calculated HRQoL score at each follow-up point over one year of follow-up duration, and then, we took the median value of HRQoL scores. Scores on the SF-12 ranged from 0 to 1, with higher scores indicating better HRQoL. It is a valid and reliable measure with internal consistency estimates exceeding 0.70 [13]. Independent variables Self-reported long COVID was the primary independent variable of this study. We considered 25 pre-established long COVID symptoms with a follow-up duration of up to 1 year after the initial onset of COVID-19 infection. long COVID was defined as per the CDC definition: the presence of at least one infection-associated chronic condition. at the entry period of the cohort (more than 90 days after acute phase of COVID-19 infection), or at any of the follow-up responses [14, 15]. As covariates we included age (younger: 39 years or less, middle-aged: ≥ 40 and ≤ 64 years, and older: ≥ 65 years), sex (male or female, assigned at birth), body mass index (BMI), vaccination history, and pre-existing medical conditions. Unlike the WHO guidelines, Japan adopts a lower BMI cutoff point. According to the guidelines of the Japan Society for the Study of Obesity, we categorized BMI as underweight (≤ 18.5 kg/m2), normal weight (BMI > 18.5 kg/m2 to <25 kg/m2), and overweight/obese (≥ 25 kg/m2) [16]. Participants were considered vaccinated if they received a COVID-19 vaccine at least 14 days before the confirmation date of COVID-19 infection. Vaccination included two categories: none or primary doses (1st and 2nd), and booster doses (primary doses plus 3rd or 4th). For each participant, we also considered the pre-existing medical conditions including psychological disorders (depression or anxiety or insomnia), metabolic disorders (diabetes or hypertension), lung diseases, and seasonal allergies. Statistical analysis We used descriptive statistics including, frequency and percentage to show the study sample characteristics, median to report the distribution of HRQoL score, and inter-quartile range (IQR) to report the distribution of participants’ follow-up. We did Wilcoxon rank-sum test to assess whether the HRQoL score differed significantly by different sociodemographic groups. We fitted adjusted beta regression models with a logit link function to examine the determinants of HRQoL, reporting coefficients with 95% confidence intervals (CIs). Individual long COVID symptoms were not included simultaneously as predictors to avoid conceptual overlap. The models were adjusted for sex, age, BMI, vaccination status, psychological disorders, metabolic disorders, lung diseases, and seasonal allergies. Covariates were selected a priori based on previous studies and data availability [9–11] and no formal data-driven variable selection procedures were applied. As beta regression coefficients are estimated on the logit scale (representing changes in the log-odds of the outcome proportion), we additionally calculated average marginal effects (AMEs) to improve interpretability. AMEs represent the average change in HRQoL (on the original proportion scale) associated with a one-unit change in each predictor, holding other variables constant. Model assumptions were assessed by confirming that the outcome variable was bounded within the (0,1) interval, and evaluating overall model fit. We also conducted subgroup analysis for age, sex and pre-existing medical conditions using adjusted beta regression models to explore the effect of long COVID on HRQoL score. Given the number of predictors and subgroup analyses conducted, there is a possibility of type I error due to multiple comparisons. As these analyses were primarily exploratory, the findings should be interpreted with caution. We performed latent class analysis (LCA) to identify underlying subgroups based on observed symptom patterns. Models with increasing numbers of classes were fitted sequentially. The optimal number of classes was determined using a combination of statistical criteria, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample-size adjusted BIC, with lower values indicating better model fit. Entropy was used to assess classification accuracy, with higher values indicating better separation between classes. Model selection also considered the interpretability of classes and avoidance of very small class sizes. Average posterior probabilities were examined to ensure adequate classification certainty. To reduce potential multicollinearity among symptom variables, pairwise correlations were assessed and, among variables with correlation > 0.75, only one representative symptom was retained. As a result, 17 long COVID symptoms were included in the LCA. Supplementary Figure S1 shows the correlogram of the long COVID symptoms included in LCA. All statistical analyses were performed using R (v4.4.1; The R Foundation for Statistical Computing, Vienna, Austria). LCA was performed using the poLCA() function from the poLCA R package v1.6.0.1 (Linzer DA, Lewis JB 2011). AME was estimated using margins() function from margins R package v0.3.28 (Ben Bolker 2017). A p-value of < 0.05 was considered statistically significant. Results 1,285 respondents were included in the final analysis; among them 349 respondents completed only one follow-up and 936 respondents completed more than one follow-up during 4th to 12th months of follow-up period after initial COVID-19 infection. In the study population, the mean age was 52 years and median HRQoL score was 0.83. Young participants and participants with no long COVID were less likely to complete more than one follow-up (Supplement Table S1). Table 1 summarizes the sociodemographic characteristics and background medical conditions of the study sample, along with the median HRQoL score of each group. 45% of the respondents were female and booster dose vaccine coverage was 85%. 70% of the participants reported long COVID over one year period of follow-up. The median HRQoL score of long COVID patients was 0.79 whereas it was 0.92 among the participants without long COVID. 16%, 3.7%, and 44% of the respondents had prior psychological disorder, lung disease and seasonal allergies respectively. Respondents with prior medical conditions showed significantly lower HRQoL score compared to the respondents without prior medical conditions. Table 1Basic characteristics of the participantsCharacteristicN = 1,285 n (%)HRQoL scoreMedian, (95% CI)P value Gender < 0.001Male704 (55.0)0.86 (0.72, 0.92)Female581 (45.0)0.79 (0.66, 0.86) Age < 0.001Young123 (9.6)0.78 (0.66, 0.86)Middle-age1,041 (81.0)0.83 (0.72, 0.92)Old-age121 (9.4)0.89 (0.75, 0.92) BMI < 0.001Normal892 (69.0)0.85 (0.72, 0.92)Obese304 (24.0)0.86 (0.69, 0.92)Thin89 (6.9)0.72 (0.66, 0.86) Vaccine doses 0.092>21,098 (85.0)0.85 (0.72, 0.92)≤2187 (15.0)0.80 (0.68, 0.92) Long COVID < 0.001No388 (30.0)0.92 (0.80, 0.92)Yes897 (70.0)0.79 (0.66, 0.89)Number of symptoms* (median, IQR)2 (0,4)–< 0.001 Psychological disorder < 0.001Yes212 (16.0)0.72 (0.65, 0.86)No1,073 (84.0)0.86 (0.72, 0.92) Metabolic disorder 0.140Yes252 (20.0)0.86 (0.72, 0.92)No1,033 (80.0)0.82 (0.72, 0.92) Lung disease 0.002Yes47 (3.7)0.72 (0.64, 0.86)No1,238 (96.3)0.84 (0.72, 0.92) Seasonal allergies < 0.001Yes571 (44.0)0.80 (0.66, 0.92)No714 (56.0)0.86 (0.72, 0.92)n (%); Median; * Number of symptoms were calculated among respondents with long COVID (n= 897);HRQoL, health-related quality of life; BMI, body mass index; IQR, interquartile range; CI, confidence interval Basic characteristics of the participants n (%); Median; * Number of symptoms were calculated among respondents with long COVID (n= 897); HRQoL, health-related quality of life; BMI, body mass index; IQR, interquartile range; CI, confidence interval We checked the HRQoL score among participants with and without long COVID at each follow-up point as shown in Fig. 2(A). We found that the differences were clinically significant (Minimal Clinically Important Difference > 0.05) at each of the follow-up points [12]. As shown in Fig. 2(B), there were significant differences in perceived health quality between the participants with and without long COVID in each domain of HRQoL and the differences were mostly occurring in social function, role limitation and mental health domains. Among the participants without long COVID, 25%-35% had suboptimal scores in social function, role limitation and mental health domains whereas the proportion was 50%-65% among the long COVID patients. On vitality domain more than 85% of the respondents had suboptimal score regardless of their long COVID status. A Health-related quality of life score (HRQoL) at each follow-up point; B Proportion of respondents with health loss on each domain of HRQoL In Fig. 3, beta regression analysis showed that, compared to men, the HRQoL score among women (β: −0.25; 95% CI: −0.33 to −0.16; AME: −0.037) was lower. Further, HRQoL score was lower among young-age group (β: −0.20; 95% CI: −0.39 to −0.01; AME: −0.03) compared to old-age group and among participants with thin BMI (β: −0.16; 95% CI: −0.31 to −0.01; AME: -0.024) compared to the participants with normal BMI. Compared to the participants with no long COVID, the HRQoL among long COVID patients (β: −0.25; 95% CI: −0.36 to −0.14; AME: −0.036) was significantly lower and increase in long COVID symptoms by one decreased the HRQoL score by 0.09 (β: −0.09; 95% CI: −0.10 to −0.07) point. Psychological disorders (β: −0.26; 95% CI: −0.36 to −0.15; AME: −0.041) showed significant inverse association with HRQoL whereas metabolic disorder, lung disease, and seasonal allergies did not show any significant relation with HRQoL. Determinants of health-related quality of life (HRQoL) and its subdomains. Notes: 1 The models were adjusted for gender, age, BMI, vaccine doses, long COVID, number of symptoms, psychological disorders, metabolic disorders, lung diseases, seasonal allergies. 2 Reference groups for each predictor – Gender (male), age (old age), BMI (normal), vaccine doses (≤ 2), long COVID (no), psychological disorder (no), metabolic disorder (no), lung disease (no), seasonal allergies (no). We also explored the effect of long COVID on HRQoL score among different subgroups and the findings are shown in Supplementary Figure S2. Among all the three age groups and in both the genders long COVID patients showed poorer HRQoL score compared to the participants with no long COVID. The inverse association between long COVID and HRQoL score was the worst among the respondents with pre-existing lung diseases (β: −0.72; 95% CI: −1.29 to -0.16; AME: −0.114) and considerably low among the respondents with pre-existing seasonal allergies (β: −0.27; 95% CI: −0.37 to −0.04; AME: −0.032). Number of long COVID symptoms showed strong positive associations with all the subdomains of HRQoL. Using LCA, we identified three subgroups of long COVID patients - latent class 1, 2, and 3. Figure 4 shows the probability distribution of the long COVID symptoms among the latent classes. In latent class 1, all the long COVID symptoms had minimal probability and none of the symptoms was predominant (probability > 0.5). 55.2% of the respondents belonged to this class. In latent class 2, headache, sore throat, and nasal congestion were the predominant (probability > 0.5) long COVID symptoms and 25.7% of the respondents belonged to this class. 19.2% of the respondents belonged to latent class 3, where all the respondents had fatigue (probability = 1.0). Cough, headache, sore throat, nasal congestion, and aches and pain were the other predominant (probability > 0.5) long COVID symptoms in latent class 3. Median HRQoL scores were 0.80 (95% CI: 0.77 to 0.80), 0.79 (95% CI: 0.78 to 0.79), and 0.66 (95% CI: 0.65 to 0.66) respectively in latent class 1, 2, and 3. Adjusted beta regression analysis showed that, compared to the participants with no long COVID, participants belonged to latent class 1 (β: −0.47; 95% CI: −0.57 to -0.36; AME: −0.065), latent class 2 (β: −0.48; 95% CI: −0.60 to −0.35; AME: −0.066) and latent class 3 (β: −0.93; 95% CI: −1.06 to −0.79; AME: −0.148) had significantly lower HRQoL. Distribution of long COVID symptoms among latent classes of long COVID patients Discussions The twofold aim of this study was to assess the impact of long COVID on the HRQoL among Japanese adults after three months of acute phase of initial COVID-19 infection using monthly follow-up data of up to one year and to identify the psychosocial factors associated with the HRQoL. We found that, long COVID had a consistent negative impact on HRQoL across the whole 1 year of follow-up. Multivariable regression showed that, being women or young-age, having long COVID or thin BMI or pre-existing psychological disorders were associated with lower HRQoL. HRQoL was decreasing when the number of long COVID symptoms were increasing. Overall, long COVID patients after 4 months to 1-year of initial COVID-19 infection had significantly worse HRQoL than those without long COVID and HRQoL did not improve significantly over 1 year of follow-up in the cohort. Similar results have been reported in several other studies that, long COVID can persist for over 1 year and is associated with reduced HRQoL scores compared with those without long COVID [17–19]. We found that female respondents were more likely to report lower HRQoL. Previous studies also supported this finding [2, 17, 20]. Our results showed a positive association between age and HRQoL which might be explained by the positive association between age and mental health domain of HRQoL (Fig. 3). Similar to our findings, several other studies also reported that, young adults were more vulnerable to a decreasing mental health in the pandemic compared to older individuals [7, 21, 22]. Another previous study in Japan reported that, the adjusted odds ratios for the association between COVID-19 infection and HRQoL in the age categories of less than 30 years, 30s, 40s, 50s, and 60 years or higher were 0.54 (0.15–1.92), 1.70 (1.03–2.81), 1.14 (0.82–1.57), 1.05 (0.77–1.42), and 0.87 (0.46–1.64), respectively which also supports our findings [23]. Respondents who had prior psychological disorders were more likely to report poor scores in role limitation, social functioning and mental health domain. Whereas, Respondents who had prior lung disease were more likely to report poor scores in role limitations, social functioning and vitality. Therefore, long-term, multidisciplinary interventions with specific focus on mental health among women, young and middle age respondents, and respondents with psychological disorders or lung diseases are needed. The prevalence of long COVID symptoms in our study was 70% which was comparatively higher than most of the previous studies [2, 6]; because we reported the period prevalence of long COVID over 1 year of follow-up period instead of point prevalence. Owing to the relapsing nature of long COVID, period prevalence can estimate the whole burden of long COVID in a population over a period of time. However, the broad definition of long COVID (≥ 1 symptom persisting for at least 3 months) used in this study may have contributed to higher prevalence estimates by capturing symptoms not specific to long COVID, particularly in self-reported data. In addition, a proportion of participants contributed only a single follow-up, and follow-up completion appeared to differ by long COVID status. Participants with persistent or more severe symptoms may have been more likely to remain in the study, whereas those with symptom resolution may have been less likely to continue responding. This differential retention could have resulted in an overestimation of the cumulative prevalence. Therefore, the findings should be interpreted in the context of this retained online cohort. LCA showed that long COVID symptoms were the most concentrated among the respondents who belonged to latent class 3 and the least concentrated among the respondents who belonged to latent class 1. We found that, one in every 5 of long COVID patients belonged to latent class 3 and were more likely to develop at least 6 symptoms (Cough, headache, sore throat, nasal congestion, aches and pain, and fatigue) with median HRQoL as low as 0.66. Even the long COVID patients who belonged to latent class 1 had significantly lower HRQoL compared to the respondents with no long COVID. Similarly, one previous study in Japan reported that, the patients who had only one long COVID symptom showed a significantly lower HRQoL compared to the patients who had no long COVID symptom and patients who had 5 or more long COVID symptoms HRQol degraded severely among them [19]. Further, the number of long COVID symptoms showed strong inverse relation to each of the domains of HRQoL. These findings inform primary health care physicians, nurses, and physical therapists that, even for minimal long COVID symptoms patient’s experience need to be validated and informs policy makers regarding the need of comprehensive primary health care and integrated long-term monitoring system for patients with multiple long COVID symptoms. This study has several strengths. First, we used monthly follow-up data of up to one year 1 year after the first three months of acute phase of COVID-19 infection to measure long COVID and associated HRQoL which is rare in the context of Japan and also globally. Further, our study used data collected between January 2022 and January 2023, when Omicron variant of COVID-19 was dominant in Japan. So, the findings of this study particularly acknowledge the effect of Omicron variant of COVID-19 on the HRQoL among Japanese adults. This study has several limitations as well. In the first place, due to online recruitment, the study population may be subject to selection bias. In addition, loss to follow-up may be informative, and differential attrition could have introduced bias. Therefore, the findings should be interpreted cautiously and are most applicable to the retained online cohort rather than the general population. Further, pre-post comparison was not possible due to the unavailability of HRQoL data prior to COVID-19 infection. Additionally, due to unavailability of data, we couldn’t include the duration and intensity of long COVID symptoms and socioeconomic status of the respondents, which are important contributors to HRQoL [24, 25]. Because of the observational nature of the study no causation was possible that, long COVID was solely responsible for decreasing HRQoL score. However, findings of this study will give policy makers a more comprehensive understanding of the full spectrum of long COVID and will facilitate better resource allocation for long term monitoring, multidisciplinary treatment and rehabilitation of long COVID patients. Conclusions HRQoL was significantly worse among participants with long COVID compared to the participants without long COVID between 4 months to 1-year after initial acute phase of COVID-19 infection. Female gender, young-age, thin BMI or pre-existing psychological disorders were associated with lower HRQoL. The number of long COVID symptoms was a strong determinant of HRQoL and the presence of minimal long COVID symptoms was associated with lower HRQoL. Cough, headache, sore throat, nasal congestion, aches and pain, and fatigue were the most prevalent long COVID symptoms among the long COVID patients who had the lowest HRQoL. For COVID-19 cases with lung disease as a comorbid condition, the risk of reduced quality of life remains high even after recovery, necessitating follow-up care that includes the post-recovery period. Findings of this research will enhance the understanding of the implications of long COVD on individual’s lives and will underscore the need of coordinated approach for long term treatment, care, and prevention of long COVID through policies, interventions, and integrated health care system. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file 1. Acknowledgements We thank Analisa Avila, MPH, ELS, of Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript. Abbreviations Coronavirus disease Health related quality of life Body mass index Latent class analysis Average marginal effect Confidence interval World Health Organization Author contributions SS, SS and YA: conceived the study, design of the work, analysis, interpretation of the data, writing or review of the article. HI, NM, NO: conceived the study, interpretation of the data, writing or review of the article. SI, AO: acquisition of data, writing or review of the article. Funding Open Access funding provided by Japan Institute for Health Security. This study was supported by Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research (KAKENHI 23K27865). Data availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Conflict of interest I.S. and A.O. are staffs of IQVIA solutions Japan. Other authors have no competing interests. Ethical approval and consent to participate Ethical approval of this study was taken from the Certified Review Board of Japan Institute for Health Security (approval number: NCGM-S-004318-02). Patients’ consent was collected via online. Consent for publication Not applicable. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary file 1. Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.