Impact of hospital safety-net status on in-hospital and short-outcomes after hematopoietic stem cell transplantation.

Introduction

Hematopoietic stem cell transplantation (HSCT) is a potentially curative treatment for malignant and non-malignant hematological diseases 1 – 3 . In recent years, the expansion of indications for HSCT has led to a significant increase in transplant volume 4 – 6 . Although advancements in HSCT methods and supportive care have markedly improved patient outcomes, the in-hospital mortality rate for allogeneic HSCT (allo-HSCT) ranges from 5 to 11% 7 , 8 , while that for autologous HSCT (auto-HSCT) ranges from 1.5% to 3.8% 9 , 10 . HSCT is a complex and resource-intensive therapeutic procedure. Patients who undergo HSCT often bear an enormous financial burden, and complications such as bone marrow suppression, infection and organ dysfunction during bone marrow transplantation further increase patient costs 11 – 16 .

Safety-net hospitals (SNHs) are healthcare facilities in the United States that serve patients enrolled in Medicaid or Medicare, as well as those without health insurance. By providing medically underserved populations with care, SNHs work to reduce inequities in health care. They make a significant contribution to public health and ensure equitable access to healthcare for all socioeconomic groups. However, increasing evidence suggests that solid organ transplantation at SNHs is associated with poor clinical outcomes.

But no relevant studies on HSCT have verified this relationship. We used the National Readmission Database (NRD) to evaluate the in-hospital and short-term outcomes of HSCT at SNHs.

Methods

Study design and population

The NRD, as part of the Healthcare Cost and Utilization Project, includes data representing 95% of U.S. hospitalizations. In healthcare facilities, such as hospitals and clinics, physicians use International Classification of Diseases (ICD) codes to document patient diagnoses and procedures during hospital stays. The NRD includes patient diagnoses, demographic data, procedure information, hospital characteristics, and outcomes (e.g., mortality, complications). The NRD is the most extensive database for researching and evaluating readmission trends and supporting national readmission analyses. We searched the NRD to identify hospitalized patients who underwent HSCT from 2011–2022 using the relevant ICD-9 and ICD-10 codes, which were used in our previous research 17 .

Using the source of the hematopoietic cells as the grouping factor, HSCTs were divided into allogeneic HSCT (allo-HSCT) and autologous HSCT (auto-HSCT) 18 . All the ICD-9 and ICD-10 codes used in this research are listed in the Supplementary Table 1 . Following the methodology of the Agency for Healthcare Research and Quality, the proportion of patients who were uninsured (self-pay) or covered by Medicaid was calculated for each hospital annually using the NRD 19 . Hospitals were categorized into quartiles based on this proportion, which was derived from all hospitalizations captured within the NRD during the study period 20 , 21 .

Ethics statement

As this study analyzes data from the public database NRD, it is an administrative and de-identified database that does not contain any personal information. Considering the need to comply with HUCP related information during the use of the NRD database, this study was approved by the Human Ethics Committee of the Affiliated Hospital of Qingdao University (QYFYWZLL29760). Our study was conducted in accordance with the Declaration of Helsinki, and none of the transplanted organs or tissues came from prisoners.

Endpoints

The primary aim of this study was to compare in-hospital mortality rates between HSCT patients during hospitalization in SNHs and those in non-SNHs. The secondary outcomes of interest included in-hospital complications, including infectious complications (including sepsis and bacterial pneumonia), acute kidney injury, cardiogenic shock and stroke. Cardiogenic shock was identified using predefined ICD codes and confirmed only when accompanied by cardiogenic etiologies. For all the HSCT patients, we also analyzed the short-term outcomes as well as the rates of all-cause readmission within 30-, 90-, and 180-days.

Statistical analysis

The R software (version 4.0.2) was utilized for all the statistical analyses. Descriptive statistics are expressed as means ± standard deviations, percentages, medians, and interquartile ranges (IQRs). Normally distributed count data are reported as the mean plus or minus the standard deviation and were compared via a t test; conversely, nonnormally distributed data are expressed as medians (25th quartile, 75th quartile) and were compared via the Wilcoxon signed-rank test. Categorical data are expressed as percentages, and group comparisons were performed using the Chi-square test or Fisher’s exact test. Propensity score matching (PSM) was performed, with patients being matched at a 1:1 ratio to reduce bias. The matching factors in this article are placed in Supplementary Table 2 . Multivariate logistic regression was employed to account for the impact of baseline characteristics and comorbidities on outcomes.

Prior to the multivariable regression, we assessed multicollinearity by calculating the variance inflation factor and subsequently excluded variables with high collinearity. The results are expressed in ORs. Baseline patient characteristics and hospital-related variables were included in the multivariable regression analysis. All covariates incorporated into the multivariable model are listed in Supplementary Table 3 . All analyses were two-tailed, with statistical significance set at P < 0.05.

Results

Baseline characteristics

We identified 130,506 patients who underwent HSCT from the NRD, including 100,393 patients who underwent auto-HSCT and 30,113 patients who underwent allo-HSCT. The screening and exclusion process is presented in Fig. 1 . At SNHs, 24,717 patients (24.6%) underwent auto-HSCT, and 6,207 patients (20.6%) underwent allo-HSCT. A greater proportion of patients who were treated in SNHs experienced complications such as anemia, mild liver disease, diabetes, and renal disease, regardless of whether they underwent autologous or allogeneic HSCT. The detailed baseline characteristics of the HSCT patients are shown in Supplementary Table 4 .

Considering the influence of demographic information and comorbidities on the results, we performed a PSM analysis to reduce the probability of an intergroup error. After matching, there were 24,715 pairs of patients who underwent auto-HSCT and 6,207 pairs who underwent allo-HSCT. The baseline characteristics were not different between the SNHs group and the non-SNHs group. The baseline characteristics of the hospitalized HSCT patients are shown in Table 1 .

In-hospital outcomes

Before matching, the in-hospital mortality rate was 2.4% for the auto-HSCT patients and 5.3% for the allo-HSCT patients. The mortality rate was higher among the patients who underwent auto-HSCT (2.7% vs. 2.3%; P < 0.001) and allo-HSCT (6.5% vs. 5.1%; P < 0.001) at SNHs than at non-SNHs (Supplementary Tables 5 and 6 ). After matching, the risk remained high among patients who underwent auto-HSCT (2.7% vs. 2.4%; P = 0.026) and allo-HSCT (6.5% vs. 5.3%; P = 0.006) at SNHs. Multivariate logistic regression analysis revealed that the mortality rate of patients who underwent auto-HSCT at SNHs was 1.14 times higher than that of patients who underwent HSCT at non-SNHs (OR = 1.14; 95% CI: 1.02–1.27; P = 0.024), and the patients who underwent allo-HSCT at SNHs had a 1.24 times higher mortality rate than those who underwent HSCT at non-SNHs (OR = 1.24; 95% CI: 1.07–1.44; P < 0.001).

In the SNHs group, the incidence of infectious complications during hospitalization was significantly higher among auto-HSCT patients (OR = 1.16; 95% CI: 1.07–1.25; P < 0.001) and allo-HSCT patients (OR = 1.12; 95% CI: 1.03–1.22; P < 0.001). Additionally, these patients were at increased risk of requiring mechanical ventilation, regardless of undergoing auto-HSCT (OR = 1.28; 95% CI: 1.14–1.43; P < 0.001) or allo-HSCT (OR = 1.35; 95% CI: 1.16–1.58; P < 0.001). Significant differences in the incidence of in-hospital stroke, cardiogenic shock or intracranial hemorrhage were not found between SNHs and non-SNHs. The in-hospital complications are described in detail in Tables 2 and 3 . In a sensitivity analysis adjusting for primary payer status (Supplementary Table 7 ), the associations between SNH status and in-hospital outcomes remained broadly consistent.

Follow-up results

The 30-day readmission rates were higher in the allo-HSCT patients and the auto-HSCT patients treated at SNHs. Among the auto-HSCT patients, the SNHs group had a readmission rate of 16.3%, while the non-SNHs group had a rate of 15.2%. Even after adjusting for confounding factors, this difference remained significant (aOR: 1.05; 95% CI: 1.02–1.09; P < 0.001). Similarly, the readmission rates for allo-HSCT patients were 31.6% in the SNHs group and 28.0% in the non-SNHs group. This difference also remained significant after adjustment (aOR: 1.13; 95% CI: 1.05–1.23; P = 0.002). At both 90 and 180 days, all-cause readmission rates continued to be higher in the SNHs group. The detailed results of the multiple logistic regression analyses are provided in Fig. 2 . The distribution of 30-day readmission causes after HSCT, categorized into eight clinically meaningful groups based on ICD code prefixes, is summarized in Supplementary Fig. 1 .

Discussion

This study demonstrates differences in outcomes between patients undergoing HSCT at SNHs and those undergoing HSCT at non-SNHs. The main findings of our research are as follows: (1) In the United States, a subset of patients (approximately 24%) undergo HSCT at SNHs. (2) The in-hospital mortality rate was greater for patients undergoing HSCT at SNHs than those at non-SNHs. (3) The incidences of infectious complications and acute kidney injury were higher in the SNHs group compared to the non-SNHs group in HSCT patients. (4) The readmission rate for patients treated in SNHs was higher than that for patients treated in non-SNHs. Overall, although these differences were statistically significant, the absolute differences in outcomes were modest, suggesting limited clinical significance.

To our knowledge, this study is the first contemporary analysis to assess the effect of SNHs on HSCT outcomes. SNHs primarily serve low-income patients, the uninsured, or those covered by Medicaid. SNHs are an important component of the healthcare system, providing care for approximately 25% of patients. The prognosis of HSCT is influenced by various factors, such as the disease type and stage, the patient’s general health, and the donor’s source and compatibility 22 – 25 .

Patients treated at SNHs often experience delays in seeking medical care due to financial constraints, which may lead to diagnosis at a more advanced stage of hematologic malignancy 26 . Specifically, many patients undergoing HSCT, regardless of hospital type, have multiple comorbidities such as diabetes, hypertension, and cardiovascular disease 22 , 30 . These conditions can complicate post-HSCT management and influence outcomes. In our study, patients at SNHs exhibited a higher prevalence of comorbidities, likely due to the patient population served rather than any intrinsic property of SNHs.

Even after PSM, in-hospital mortality remained higher at SNHs, indicating that unmeasured hospital-level factors may be associated with the observed differences. Although payer composition may contribute to the observed disparities, sensitivity analyses yielded broadly consistent results. These findings suggest that the differences are not solely driven by payer status and may reflect both patient- and hospital-level factors. Previous studies have indicated that institutional experience and transplant volume may influence HSCT outcomes 27 . For example, higher institutional case volumes are associated with lower short-term and long-term mortality following HSCT. In a large European cohort study, center patient numbers and program duration were significantly associated with improved survival rates after allogeneic and autologous transplants 28 . This discrepancy has been attributed to the experience accumulated through long-term transplantation activities at the center and the availability of multidisciplinary support. It should be noted that the relationship between case volume and transplant outcomes is not entirely consistent across different studies. The database used in this study did not include information on transplant center case volume or specific resource allocation, making it impossible to directly evaluate these factors. However, these hospital-level differences may be associated with the persistently higher in-hospital mortality observed at SNHs after PSM.

The type of hospital facility is known to have an influence on early mortality rates and overall survival outcomes in patients with hematologic malignancies 29 . SNHs primarily serve socioeconomically and medically underserved populations and often operate under persistent financial pressures. Previous studies have indicated that funding constraints, particularly inadequate disproportionate share hospital (DSH) payments and lower reimbursement rates, may limit hospitals’ resource allocation and diminish their capacity to provide high-complexity medical services 30 – 32 . Although these background factors are important for understanding system-level differences, they were not measured in our dataset and therefore cannot be directly linked to the results observed in our cohort. Infections remain a key determinant of in-hospital mortality after HSCT. Consistent with previous reports, patients treated at SNHs are at increased risk of infectious complications 33 – 35 . However, the mechanisms underlying these differences cannot be determined from our data.

This is the first study to focus on the short-term readmission rates in HSCT patients at SNHs. Patients who underwent HSCT in SNHs had significantly higher readmission rates than those who underwent HSCT in non-SNHs, which may be associated with unmeasured socioeconomic factors. Patients typically treated at SNHs often experience significant socioeconomic challenges. Psychosocial issues, financial difficulties, and a lack of insurance coverage can also increase psychological stress among low-income populations in need of transplantation. Financial barriers to treatment can induce anxiety and depression in patients, which in turn negatively impacts treatment adherence and overall engagement with care 17 . Additionally, limited access to preventive care and recommended therapies may further affect transplant outcomes. Our findings may have implications for interpreting hospital performance metrics, such as those reported by the Center for International Blood and Marrow Transplant Research (CIBMTR). However, due to the lack of linkage between the NRD and registry-based data, direct comparisons are not possible, and further studies integrating these data sources are warranted.

Our findings revealed differences in in-hospital and short-term outcomes between patients treated at SNHs and those treated at non-SNHs. These findings suggest differences in patient outcomes across different types of hospitals, but do not establish a causal relationship. However, SNHs already provide essential medical care for vulnerable populations, and additional research is needed to elucidate the underlying reasons for the differences in patient outcomes between these types of hospitals. SNHs require increased financial, infrastructural, and workforce support to enhance perioperative care, optimize patient management, and ultimately reduce the incidence of complications associated with HSCT.

Study limitations

The NRD is a valuable tool for researching trends in hospital readmission, as it offers data for a large sample of matched-patients hospitalized across the United States to support medical quality assessment and policy-making. However, this study has several limitations. First, our definition of SNHs may not fully capture all aspects of a hospital’s safety-net status, as it is based on the payer composition of hospitalized patients captured in the NRD rather than the overall population served by each hospital. In addition, the NRD does not allow us to disentangle whether the observed associations are driven by patient-level socioeconomic factors or hospital-level characteristics; thus, the findings should be interpreted as reflecting a combination of both. Furthermore, our findings may be influenced by patient selection or Berkson’s bias. Such biases may be systematic differences in the characteristics of inpatients in different hospitals, and this is usually an important source of changes in hospital care outcomes. Second, the NRD includes only inpatient data, excluding data collected during outpatient visits and data from patients who present to the emergency department but do not require hospitalization, which may lead to an underestimation of the overall burden on the healthcare system. Furthermore, the database does not provide detailed clinical information, such as blood test results, disease stratification, staging, bone marrow matching, or medication data, limiting its ability to assess disease severity. It also lacks information about general infrastructure differences, such as standard procedures in different types of hospitals or staff-to-patient ratios, which may affect outcomes. Finally, the NRD lacks key variables such as socioeconomic factors, lifestyle behaviors, long-term follow-up information, and patient adherence, restricting a comprehensive analysis of the factors contributing to readmissions. The NRD provides encounter-level clinical and demographic information but does not include details on hospital staffing, clinical workflow protocols, resource allocation, or transplant-specific operational infrastructure.

Conclusions

This comprehensive study compares the outcomes of patients undergoing HSCT at SNHs and non-SNHs. Patients at SNHs had higher rates of in-hospital mortality, infectious complications, and acute kidney injury than those at non-SNHs, along with significantly higher short-term readmission rates. These differences, while statistically significant, were modest in magnitude. This study highlights the potential the challenges faced by SNHs and calls for increased financial, infrastructural, and workforce support to improve perioperative care, optimize patient management, and enhance HSCT outcomes.

Supplementary Information

Supplementary Information.

Author contributions

Q.Z. Data curation, formal analysis, resources, software, writing—original draft. H.W. Data curation, formal analysis, resources, software, writing—original draft. Q.C. Data curation, methodology, resources, software, writing—original draft. HJ.D. Data curation, resources, software. QY.Y. investigation, supervision, writing—review and editing. N.L. Data curation, resources, software, writing—review and editing. W.L. methodology, writing—review and editing. N.Z. Data curation, methodology, software, writing—review and editing. K.Z investigation, project administration, writing—review and editing. M.F. Conceptualization, investigation, methodology, project administration, supervision, writing—review and editing. C.Z. Conceptualization, investigation, methodology, project administration, supervision, writing—review and editing.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

Our study was conducted in accordance with the Declaration of Helsinki, and none of the transplanted organs or tissues came from prisoners. As all data from the NRD are de-identified, informed consent from individual patients was not required.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Qi Zhang and Hui Wang These authors contributed equally to this work and should be considered co-first authors.

Contributor Information

Min Fang, Email: fangdamin2023@126.com.

Chi Zhou, Email: 13206397160@163.com.

References

1. D’Souza, A. et al. Current use of and trends in hematopoietic cell transplantation in the United States. Biol. Blood Marrow Transplant. 26 , e177–e182. 10.1016/j.bbmt.2020.04.013 (2020).

2. Xu, L. P. et al. Hematopoietic stem cell transplantation activity in China 2019: A report from the Chinese Blood and Marrow Transplantation Registry Group. Bone Marrow Transplant. 56 , 2940–2947. 10.1038/s41409-021-01431-6 (2021).

3. Xu, L. et al. The consensus on indications, conditioning regimen, and donor selection of allogeneic hematopoietic cell transplantation for hematological diseases in China-recommendations from the Chinese Society of Hematology. J. Hematol. Oncol. 11 , 33. 10.1186/s13045-018-0564-x (2018).

4. Niederwieser, D. et al. Hematopoietic stem cell transplantation activity worldwide in 2012 and a SWOT analysis of the Worldwide Network for Blood and Marrow Transplantation Group including the global survey. Bone Marrow Transplant. 51 , 778–785. 10.1038/bmt.2016.18 (2016).

5. Xu, L. P. et al. A review of hematopoietic cell transplantation in China: Data and trends during 2008-2016. Bone Marrow Transplant. 52 , 1512–1518. 10.1038/bmt.2017.59 (2017).

6. Gratwohl, A. et al. One million haemopoietic stem-cell transplants: a retrospective observational study. Lancet Haematol 2 , e91-100. 10.1016/s2352-3026(15)00028-9 (2015).

7. Majhail, N. S. et al. Transplant center characteristics and survival after allogeneic hematopoietic cell transplantation in adults. Bone Marrow Transplant. 55 , 906–917. 10.1038/s41409-019-0748-1 (2020).

8. Godara, A. et al. Length of stay and hospital costs for patients undergoing allogeneic stem-cell transplantation. JCO Oncol. Pract. 17 , e355–e368. 10.1200/op.20.00170 (2021).

9. Sanchez, L. et al. In-hospital mortality and post-transplantation complications in elderly multiple myeloma patients undergoing autologous hematopoietic stem cell transplantation: A population-based study. Biol. Blood Marrow Transplant. 23 , 1203–1207. 10.1016/j.bbmt.2017.03.012 (2017).

10. Marini, C. et al. Real-life data on safety and efficacy of autologous stem cell transplantation in elderly patients with multiple myeloma. Ann. Hematol. 98 , 369–379. 10.1007/s00277-018-3528-x (2019).

11. Maziarz, R. T. et al. Health care costs among patients with hematologic malignancies receiving allogeneic transplants: A US payer perspective. Blood Adv. 8 , 1200–1208. 10.1182/bloodadvances.2023011033 (2024).

12. Maziarz, R. T. et al. Estimating the Lifetime Medical Cost Burden of an Allogeneic Hematopoietic Cell Transplantation Patient. Transplant Cell Ther 29 (637), e631-637.e639. 10.1016/j.jtct.2023.06.013 (2023).

13. Perales, M. A. et al. Real-world economic burden associated with transplantation-related complications. Biol. Blood Marrow Transplant. 23 , 1788–1794. 10.1016/j.bbmt.2017.06.017 (2017).

14. Khera, N., Zeliadt, S. B. & Lee, S. J. Economics of hematopoietic cell transplantation. Blood 120 , 1545–1551. 10.1182/blood-2012-05-426783 (2012).

15. Mariotto, A. B., Yabroff, K. R., Shao, Y., Feuer, E. J. & Brown, M. L. Projections of the cost of cancer care in the United States: 2010-2020. J. Natl. Cancer Inst. 103 , 117–128. 10.1093/jnci/djq495 (2011).

16. Weisdorf, D., Ruiz-Arguelles, G. J., Srivastava, A., Gómez-Almaguer, D. & Szer, J. Economic challenges in hematopoietic cell transplantation: How will new and established programs face the growing costs?. Biol. Blood Marrow Transplant. 23 , 1815–1816. 10.1016/j.bbmt.2017.07.026 (2017).

17. Zhang, Q. et al. Psychological risk factors on hematopoietic stem cell transplantation patients: Insights from National Readmission Database analysis. Psychooncology 33 , e70032. 10.1002/pon.70032 (2024).

18. Saad, A. et al. Hematopoietic cell transplantation, version 2.2020, NCCN Clinical Practice Guidelines in Oncology. J. Natl. Compr. Canc. Netw. 18 , 599–634. 10.6004/jnccn.2020.0021 (2020).

19. Sutton, J. P., Washington, R. E., Fingar, K. R. & Elixhauser, A. in Healthcare Cost and Utilization Project (HCUP) Statistical Briefs (Agency for Healthcare Research and Quality (US), 2006).

20. Branche, C. et al. Utilization of minimally invasive colectomy at safety-net hospitals in the United States. Surgery 176 , 172–179. 10.1016/j.surg.2024.03.036 (2024).

21. Ando, T. et al. In-hospital outcomes of ST-segment elevation myocardial infarction complicated with cardiogenic shock at safety-net hospitals in the United States (from the Nationwide Inpatient Sample). Am. J. Cardiol. 124 , 485–490. 10.1016/j.amjcard.2019.05.037 (2019).

22. Copelan, E. A. Hematopoietic stem-cell transplantation. N. Engl. J. Med. 354 , 1813–1826. 10.1056/NEJMra052638 (2006).

23. Vittayawacharin, P., Kongtim, P. & Ciurea, S. O. Allogeneic stem cell transplantation for patients with myelodysplastic syndromes. Am. J. Hematol. 98 , 322–337. 10.1002/ajh.26763 (2023).

24. Gagelmann, N. et al. Optimized EBMT transplant-specific risk score in myelodysplastic syndromes after allogeneic stem-cell transplantation. Haematologica 104 , 929–936. 10.3324/haematol.2018.200808 (2019).

25. Zeiser, R. Introduction to a how i treat series on major complications after allogeneic stem cell transplantation. Blood 135 , 1609. 10.1182/blood.2020005811 (2020).

26. Lee, T. C. et al. Liver transplantation at safety net hospitals: Potentially vulnerable patients with noninferior outcomes. Surgery 166 , 1135–1141. 10.1016/j.surg.2019.06.020 (2019).

27. Lim, L. et al. The association between institutional case volume of hematopoietic stem cell transplantation and mortality. Transplant. Proc. 55 , 1715–1725. 10.1016/j.transproceed.2023.05.019 (2023).

28. Gratwohl, A. et al. Economics and outcome after hematopoietic stem cell transplantation: A retrospective cohort study. EBioMedicine 2 , 2101–2109. 10.1016/j.ebiom.2015.11.021 (2015).

29. Bhatt, V. R. et al. Early mortality and overall survival of acute myeloid leukemia based on facility type. Am. J. Hematol. 92 , 764–771. 10.1002/ajh.24767 (2017).

30. Cole, E. S., Walker, D., Mora, A. & Diana, M. L. Identifying hospitals that may be at most financial risk from Medicaid disproportionate-share hospital payment cuts. Health Aff (Millwood) 33 , 2025–2033. 10.1377/hlthaff.2014.0109 (2014).

31. Neuhausen, K. et al. Disproportionate-share hospital payment reductions may threaten the financial stability of safety-net hospitals. Health Aff (Millwood) 33 , 988–996. 10.1377/hlthaff.2013.1222 (2014).

32. Hoehn, R. S. et al. Effect of hospital safety-net burden on cost and outcomes after surgery. JAMA Surg. 151 , 120–128. 10.1001/jamasurg.2015.3209 (2016).

33. Law, A. C. et al. In-hospital vs 30-day sepsis mortality at US safety-net and non-safety-net hospitals. JAMA Netw. Open 7 , e2412873. 10.1001/jamanetworkopen.2024.12873 (2024).

34. Barbash, I. J. & Kahn, J. M. Sepsis quality in safety-net hospitals: An analysis of Medicare’s SEP-1 performance measure. J. Crit. Care 54 , 88–93. 10.1016/j.jcrc.2019.08.009 (2019).

35. Ngo, L., Quintero, L. & Parreco, J. P. Nationwide propensity-matched comparison of surgical site infections in safety net hospitals. JAMA Surg. 159 , 588–590. 10.1001/jamasurg.2023.7863 (2024).

Associated Data

Supplementary Materials

Supplementary Information.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, upon reasonable request.