Abstract
Introduction
Frailty is a multidimensional and dynamic clinical syndrome characterized by increased vulnerability to stressors.1,2 It is associated with increased mortality, hospitalization, falls, and admission to long-term care.1,3 Patients with frailty also experience impaired quality of life and heightened loneliness.4,5 Frailty is closely associated with increasing age as it is characterized by an accumulation of cellular and molecular deficits. 6 Multimorbidity, defined as the coexistence of two or more chronic conditions, 7 is also increasingly prevalent as populations age. The prevalence of multimorbidity varies depending on the definition used, population considered, and methods used for diagnosing 8 ; a study from the Scottish primary health care found that 23% of the general population live with multimorbidity, in Denmark, the prevalence is 22%. 9 However, in patients above 85 years, it is more than 70%.10,11 Whilst the prevalence of multimorbidity is lower among individuals under 65 years compared to older adults, the absolute number of younger adults experiencing multimorbidity exceeds that of older persons. 10 Individuals with multimorbidity often face complex health challenges involving both physical and mental health issues. 12 In individuals with multimorbidity, more than one third had both a physical and a mental health diagnosis, with the likelihood of mental illness increasing as the number of physical conditions rises. 13 These interconnected health issues contribute to the reduced quality of life,14,15 diminished physical functioning, prolonged hospital stays, and increased healthcare utilization.9,11,16
Several tools for measuring frailty exist, such as the Clinical Frailty Scale, 17 Fried’s Frailty Phenotype, 3 and The Frailty Index. 18 However, most of these focus on the physical domain, and even fewer include the patient’s self-perceived physical, psychological or social health status. 19 Most frailty scales furthermore require assessment by a specialized health care professional, making such evaluations time- and resource-consuming.20,21
The Tilburg Frailty Indicator (TFI) is a self-reported tool developed to assess frailty across physical, psychological, and social domains, allowing for a comprehensive evaluation of the individual’s health vulnerabilities. 22 Studies have confirmed the TFI’s reliability and validity in predicting poor outcomes, such as mortality and hospital readmissions, particularly in elderly community-dwelling populations,23,24 but also among acutely admitted older adults. 25 A Danish study has demonstrated the practical application of the TFI in clinical settings, showing that a TFI score of 5 or above was significantly associated with increased readmission and mortality risks in acutely admitted patients. 26 The authors also investigated which elements of the TFI predicted readmission and death, finding social frailty to predict both outcomes, whereas psychological frailty only predicted readmissions. 25
Previous studies evaluating the TFI have focused primarily on older, community-dwelling adults. However, because multimorbidity and frailty are closely correlated, and psychological and social problems are highly prevalent among patients with multimorbidity regardless of age, evaluating the predictive value of the TFI in this population may be clinically relevant. Moreover, the TFI provides a multidimensional 15-point scale, yet most previous reports have relied on a dichotomised cut-off, potentially oversimplifying a complex and dynamic syndrome. 27 Finally, little is known about how the TFI performs across the broader adult age span, despite the fact that frailty is not limited to older adults. Addressing these gaps may clarify whether the TFI can identify high-risk patients with multimorbidity across different ages and care settings, and whether a point-by-point approach yields additional information beyond dichotomisation.
This study aimed to extend prior work in three ways; First, to evaluate the predictive value of the TFI in adults with multimorbidity recruited across diverse clinical settings in relation to mortality, readmission, and number of healthcare contacts. Second, to explore potential age-related differences in the association between frailty and adverse outcomes. Third, to examine the TFI both as a dichotomous and a continuous (point-by-point) measure to characterize graded risk patterns that may be obscured by dichotomization. Finally, we explored whether specific questions in the TFI are more prevalent as drivers of the outcomes than others.
Methods
Study design and setting
The study was designed as a prospective observational study complemented with retrospective data collected from Danish National Registers. Participants were recruited from inpatient and outpatient facilities within the Regional Hospital Central Jutland (Department of Internal Medicine or the Department of Cardiology, Viborg and Skive hospitals). Additionally, at two general practitioners, and via the primary nursing care of Skive Municipality.
Participants
Adult patients (age ≥18 years) with two or more chronic conditions, with at least one being treated at a hospital outpatient clinic at the Department of Internal Medicine or the Department of Cardiology, Viborg and Skive hospitals met inclusion criteria. We used a report by Barnett et al., which included the 40 most prevalent medical conditions in Scotland’s general population 13 to define chronic conditions. Nine conditions (pain condition, hearing loss, alcohol problems, other psychoactive substance misuse, diverticular disease of intestine, glaucoma, blindness and low vision, chronic sinusitis, and learning disability) not routinely handled in outpatient settings were excluded from the inclusion diagnoses. The instructions used by the clinical personnel responsible for including patients are displayed in Supplemental Table S1. Patients with dementia, cognitive deficits, or life expectancy of less than three months were considered ineligible.
The participants were recruited by a member of their health-care team from either the municipality nursing care, general practitioner, or the hospital clinics. During the study inclusion period, all patients in contact with these institutions were screened for eligibility.
Recruitment started November 1st 2019, and was planned to proceed until May 1st 2020, however, it was terminated on March 12, 2020 due to the COVID-19 lockdown in Denmark. Recruitment was reinitiated on September 1st 2021 until January 31st 2022. Eligible patients were included in the study on the day they provided oral and written informed consent for participation.
Data collection and sources
Patient data
Demographic data, level of education, current labour market affiliation, current use of municipal health care, social serviceswere obtained from each patient at inclusion. Information regarding the participants’ chronic conditions was self-reported, but the study nurse had access to the medical records and was allowed to include information from those sources. Also, body weight and height were measured. All inclusion data was captured and stored in Research Electronic Data Capture (Redcap).28,29
Registry data
The Danish Civil Registration System assigns residents a unique number, enablingaccurate linkage between Danish national registries. 30 For each patients, data for initiation and termination of hospital contacts (physical, virtual, or home visit) was obtained from the Danish National Patient Registry (DNPR). The number of consultations at the general practitioner (in clinic, telephone, e-mail, or home visit) or specialist medical practitioners was obtained from the Danish National Health Service Registry, 31 see Supplemental Table S2 for an overview of the included codes.
Exposure
The exposure was patient-reported frailty assessed by the TFI. At the day of inclusion, the patient completed the TFI questionnaire. The questionnaire has been translated and validated in Danish. 32 It consists of two parts, where the first (part A) comprise of basic determinants of frailty, such as age, sex, marital status, and socio-economic variables. The second part (part B) entails 15 questions, subdivided into eight questions targeting the physical domain, four questions directed at the psychological domain, and three questions regarding the social domain. 22 The TFI score is calculated from part B, with each question scored as 0 (not present) or 1 (present). The scale ranges from 0 to 15, where 15 indicates the highest level of frailty. We used the original dichotomous version with a cut-off of 5 points. 22 Thus, a TFI score of 5 or higher indicated frailty; however, to further examine the properties of the TFI, we also used it as a continuous score of 1 to 15.
Outcomes
1. Data on 180-days all-cause mortality was obtained from the civil registration registry, which records the exact date of death. 2. Readmission was defined as the first unplanned admission (hospital contact lasting >12 hours based on DNPR data) within 30 days after discharge date from the initial admission. For participants enrolled during a hospital admission, the index discharge was that admission; for others, the index discharge was discharge from the first hospital admission occurring after enrolment within the 180-day follow-up. Participants contributed only one 30-day risk window and only the first readmission counted as an event. 3. Healthcare contacts included both primary sector and hospital contacts, excluding prescheduled hemodialysis treatment. Hospital contacts included admissions (and readmissions), virtual and home visits, and outpatient visits. Contacts lasting ≤12 hours without another contact ≤4 hours before or after, and then totalling>12 hours, were considered outpatient visits (this definition was necessary due to the structure of the database). Embedded hospital contacts during admissions were included, such as cardiology outpatient echocardiography during an admission at another department. Primary sector contacts were defined as contacts to the general practitioner (clinic visits, telephone, e-mail, or home visits) or private practice specialist clinic consultations. A sub-analysis differentiated between the primary sector and hospital contacts.
Timing of measurements
Baseline measurements, including the TFI, were obtained at the time of enrolment. For participants included during a hospital admission, the follow-up period started on the date of discharge; for outpatients, general practice, or municipal recruits, follow-up started on the enrolment date. Participants were censored at death or at the end of the 180-day follow-up period, whichever came first.
Statistical analysis
Descriptive statistics are presented using counts with percentages and means with standard deviation (SD). Summary statistics were performed for the entire group and for frail or non-frail participants separately.
For outcome 1 (all-cause mortality), absolute estimates were obtained using the Kaplan-Meier estimator, which was also used for the illustration of the survival curves. The Log Rank test was used to test for significance. For outcome 2 (readmission), the absolute estimates were obtained using the Aalen-Johanson estimator, treating death as a competing risk, as this is recommended for the study of non-fatal outcomes in populations with a high mortality.33,34 Gray’s test was used to test for significance.
For outcomes 1 and 2, we used a Cox regression to estimate hazard ratios (HRs) in both a crude and an analysis adjusted for age and sex. All estimates are reported with 95% confidence intervals (CIs). In the main analysis, estimates were reported for the entire population, comparing frail participants to non-frail. Readmission (outcome 2) was modeled primarily using cause-specific Cox regression. This model estimates the association between frailty and readmission among participants who remain alive at the time of event.35,36 As a sensitivity analysis, we also fitted Fine–Gray subdistribution hazards models to assess whether results were consistent when estimating effects on the cumulative incidence of readmission in the presence of the competing risk of death.
Furthermore, in sub-analysis, age was dichotomized at 70 years. A recent review emphasizes that frailty is not restricted to a single chronological definition, and the choice of cut-off depends on the study aim and context. 2 The 70-year cut-off was chosen pragmatically to ensure sufficient events in both strata.
As a measure of the Cox models’ predictive precision, we used the C-index as a measure for model discrimination. This value corresponds to the area under the receiver operating characteristics (ROC) curve, where 1 represents 100% correct classification of all participants with the outcome, and 0.5 represents chance.
For outcome 3, we used a Poisson regression with robust estimation of variance adjusted for age and sex. This method considers the total number of contacts as a count in relation to the time the participants are “at risk” of having the outcome. The output is an incidence rate ratio (IRR) with 95% confidence intervals comparing frail with non-frail individuals..
Study size
The study size was pragmatic and determined by the recruitment period and available resources. Recruitment was curtailed by the COVID-19 pandemic, which limited the total number of participants. We therefore did not perform an a priori sample size calculation. As observed power calculations are determined by the observed p-values, they do not add information beyond effect estimates and confidence intervals.37,38
An additional summary of the prevalence of positive answers for each of the TFI part B questions was computed both overall and according to each outcome. This was performed to observe for variations in data. No formal testing was performed, we considered a 10%-point difference in prevalence for each question between the total cohort and participants with the specific outcomes to be a relevant difference. For this summary, outcome 3 was arbitrarily categorized into groups of participants having 8-14 contacts during follow-up and those having 15 or more contacts.
The study is reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE). 39 Statistical analyses were performed on the dedicated research remote servers of the Danish Health Data Authorities using STATA (version 18, StataCorp) and R (version 4.3.3, R Foundation for Statistical Computing).
Results
The study included 471 participants. 446 were included at the hospital (353 from the outpatient clinics, 93 participants during admission), 15 participants from general practice and 10 from Skive municipality. The cohort had a mean age of 67.4 years (SD 12.5) and an average of 6.0 (2.3) chronic conditions. Male participants comprised 58% of the cohort. The TFI questionnaire completion rate was 100%.
Patient characteristics at inclusion.
aincludes unemployment benefits, Danish student grants, and social assistance.
Abbreviations: TFI: Tilburg Frailty indicator, SD: Standard deviation.
Mortality
At 180 days, 41 participants had died, corresponding to an all-cause mortality rate of 8.7%. Frail participants had a notably higher risk of mortality compared to non-frail participants (Table 3). Adjusted Cox regression analysis revealed that frailty was associated with a HR of 4.6 (95% CI 2.2, 9.7) for death. Additionally, when examining frailty as a continuous variable, each one-point increase in the TFI score resulted in a 22% increase in the hazard for mortality (HR 1.2 (95% CI 1.1, 1.3). As shown in Figure 1, the Kaplan-Meier curve demonstrated a clear separation in survival between frail and non-frail participants (p <0.001), with frail participants having substantially lower survival rates over the 180-day period. 180-day all-cause mortality among frail and non-frail patients.
Stratified analysis
In participants aged ≤70 years, frail individuals had an adjusted HR of 12.2 (95% CI 1.5, 98.0) compared to non-frail, while frail participants >70 years had a HR of 3.7 (95% CI 1.7, 8.4) compared to non-frail (Table 3). Estimates in this stratified analysis were imprecise and CIs were wide, due to a low number of events among participants ≤70 years (9 events/248 at risk).
Specific TFI questions
Tilburg Frailty Indicator part B questions overall and according to outcome.
n represents the number of patients with a positive answer (=the answer that assigns 1 point to the score).
Readmission
Cox regression analyses: associations between Tilburg Frailty Indicator, all-cause mortality and 30-days readmissions.
HR: Hazard ratios (95% CI), TFI: Tilburg Frailty Indicator.
aAdjusted for age and sex.
30-day readmission among frail and non-frail patients.
Stratified analyses
Frail individuals aged ≤70 years had a significantly higher risk of readmission, with an adjusted HR of 4.7 (95% CI 2.1, 10.4), compared to non-frail, whereas frail participants >70 years had an adjusted HR of 1.7 (95% CI 1.0, 3.1) compared to non-frail (Table 3). Again, CIs were wide and reflecting imprecise estimates due to low power.
Specific TFI questions
Weight loss (38% vs. 20% overall), difficulty walking (63% vs. 47% overall), physical tiredness (64% vs. 54% overall), feeling down (51% vs. 38% overall), and feeling nervous or anxious (43% vs. 30% overall) were all more than 10%-point more frequent among participants who were readmitted within the 180-day follow-up period.
Healthcare contacts
Poisson regression analyses: associations between Tilburg Frailty Indicator and health care contacts, overall and subdivided on primary sector or hospital contacts.
IRR: Incidence rate ratios (95% CI), PY: Person years, TFI: Tilburg Frailty Indicator.
aAdjusted for age and sex.
Stratified analyses
In participants aged ≤70 years, frailty was associated with significantly more healthcare contacts, with an adjusted IRR of 1.8 (95% CI 1.5, 2.1) compared to non-frail individuals. A one-point increase in TFI score was associated with an adjusted IRR of 1.1 (1.1, 1.3) (Table 4). Among participants aged >70 years, frailty was not statistically significantly associated with a higher risk of health care contacts, although the point estimates pointed in the same direction as for younger participants (Table 4).
When analyzing healthcare contacts separately according to sector (Table 4), the observed pattern of the estimates was similar to the overall analysis for both primary sector and hospital contacts, with frail participants having a higher risk of health care contacts: IRR 1.5 (95% CI 1.3, 1.8) for primary sector contacts and IRR 1.4 (95% CI 1.2, 1.6) for hospital contacts. Similarly, the risk was more pronounced among participants ≤70 years, whereas the associations were not statistically significant among individuals >70 years.
Specific TFI questions
None of the individual TFI questions showed a 10%-point change in prevalence among participants with 8-14 or ≥15 healthcare contacts compared to the overall cohort (Table 2). Generally, we observed a pattern that participants with 8-14 contacts reported problems less common than the overall cohort, while participants with ≥15 healthcare contacts reported problems more frequently, however all differences were below 10%-point. The largest difference in prevalence among participants with ≥15 healthcare contacts was observed for difficulty walking (54% vs. 47% overall) and feeling down (45% and 38% overall), both showing a 7%-point increase.
Discussion
In this observational study of 471 adult participants with multimorbidity, we found that frailty, as measured by TFI, significantly increased the hazard of all-cause mortality, 30-days readmission risk, and the risk of having contact with the health care system within 180 days following the frailty assessment. Furthermore, each one-point increase in TFI increased the hazard of mortality and readmission, while the risk of being in contact with the health care system also increased. The number of healthcare contacts increased equally in both sectors, indicating that frail participants had an increased use of both the primary sector and hospital services.
Stratified analyses of participants below and above 70 years of age showed an even higher risk of all outcomes in the youngest group, although some outcomes were imprecise due to a lower statistical power.
Among frail participants who died or were readmitted, weight loss, difficulty walking, physical tiredness, feeling down, and feeling nervous or anxious were the challenges that were most prevalently reported compared to the entire cohort.
It is well known that a high correlation between multimorbidity and frailty exists - a recent systematic review reported that 72% of patients with frailty also had multimorbidity, 40 and furthermore, worsening of frailty – frailty transitions – are also connected to multimorbidity. 41 However, the prevalence of frailty among (primarily) community-dwelling adults with multimorbidity was only 16-17% in these studies. 40 Both used the Fried’s phenotype criteria for frailty 3 which consists of five phenotypical traits (weight loss, exhaustion, reduced grip strength, slow walking speed, and reduced physical activity). These measures are closely related to five of the 15 TFI questions covering the physical domain. Interestingly, the Frail Scale identifies even fever patients as frail, with prevalence below 10% among the same population. 42 However, as noted above, the TFI also addresses the psychological and social aspects of frailty, which are common issues among patients with multimorbidity.12,13 A recent review identifies multidimensional frailty in more than 40% of community dwelling older adults. 43 This correlates with our study, where TFI identifies almost 50% of participants as frail. This discrepancy is likely due to a combination of the TFI categorizing a higher proportion of individuals as frail than the frailty phenotype model, as observed by others, 44 and possibly a selection of more frail patients with multimorbidity in this study. Consistent with previous studies,25,26 our results indicate that frailty serves as a reliable predictor of all-cause mortality, readmission, and number of healthcare contacts. This predictive capacity is evident in both elderly and multimorbid populations when followed for a period of 180 days. This broader applicability highlights the importance of frailty assessment as a valuable clinical tool, particularly when evaluating patients with complex health conditions. Our study shows that a one-point increase in TFI is associated with increased risk of these outcomes, emphasizing that frailty exists on a continuum, and higher level of frailty is associated with a higher risk of adverse outcomes.
This study establishes an association between frailty and an increased number of healthcare contacts in both the primary and secondary care sectors. While the TFI has previously been linked to markers of healthcare utilization, 45 this study is, to our knowledge, the first to demonstrate a significant association with increased planned and unplanned contacts across both sectors. These findings highlight the necessity for cross-sectoral interventions that integrate primary and secondary care, balancing the need for individualized, patient-centered care with sustainable resource allocation.
Integrated care models for patients with complex healthcare needs, emphasize the need for cross-sectoral collaboration and flow of information between sectors, while focusing on individualized interventions. 46 However, limited evidence support the impact of these models on health-related quality of life, mental health, or mortality, and cost-effectiveness varies.46–48Research is needed to define optimal care structures that align with the complex needs of frail patients, while ensuring healthcare system. 49
Although results in our stratified analyses were exploratory and underpowered, they indicate that the size of the association between frailty and adverse outcomes may be even higher in younger adults than among older adults. Furthermore, the C-index reported in Table 3 suggests that the TFI has a higher predictive precision among the younger patient group. The age-associated reduction of predictive performance has also been reported by others, 50 and may indicate that the risk of adverse outcomes becomes increasingly multifactorial with rising age and therefore more complex to predict. Younger patients with frailty likely have more advanced or unstable disease burden and less age-related physiological decline. This impact their risk of adverse outcomes. Additionally, the municipal health care system may be less attuneed to younger patients, hindering preventive measures more difficult. If frailty represents the same syndrome across age groups, 51 it remains unclear which interventions are most relevant for younger patients with multimorbidity and frailty, compared to older patients. 52
A study by Gobbens and Andreasen
25
investigated the association between TFI components and mortality and readmission, reporting that physical and social frailty were associated with both outcomes, whereas psychological frailty related only to readmission. Although our participants were younger with higher levels of comorbidity, data from Table 2 partially indicate a similar pattern. Elements from the physical domain, including
TFI has been compared to other frailty instruments and has been recommended in contexts where health care professionals have limited time for interview and examination. 23 While no consensus regarding the precise elements that constitute the frailty syndrome, with more than 50 different frailty evaluation systems reported, 53 the most frequently used systems focus on physiological frailty. 19 It is noteworthy, although not unexpected, that our study, and Gobbens and Andreasen, 25 found both psychological and social aspects to be significant and associated with adverse outcomes. Thus, considering targeted interventions toward these important aspects is advisable. However, a more detailed description of the causes and reasons for the adverse outcomes is needed to design such interventions. Healthcare systems worldwide are already struggling with limited resources 54 and studies have shown that the demographic changes of the future will put even further strain on healthcare. 55 This emphasizes the growing need for tools that can effectively identify adults in risk of unfavorable health outcomes, and tailored interventions to address the complex needs of these patients. 47 TFI is a self-reported tool that does not require an advanced and time-consuming geriatric assessment performed by healthcare professionals. Self-reported tools could potentially streamline patient assessment, allowing for efficient identification of high-risk individuals. However, like all self-reported tools, caution must be exercised when evaluating patients with cognitive impairment, as reliability may diminish when cognitive impairment becomes pronounced. 56 Cognitive decline is, however, strongly associated with both physical and overall frailty,17,23 and should always raise considerations for a more extensive evaluation of level of function the patients’ need for support.
Methodological considerations
This study was a prospective observational study using routinely collected healthcare data from Danish nationwide registries. The use of these data sources for research is widely used and highly prevalent, and the validity of the registries is considered high. 30 Access to both hospital treatment, general practitioner, and municipality health care is universally free of charge in Denmark. Participants were recruited from several sectors (hospital, general practitioner, and municipality nursing care). As the inclusion criteria required that participants were treated in at least two hospital outpatient clinics for at least two chronic conditions, our study cohort is not population-based but represents a group of multimorbid patients with more severe/complex diagnoses than the general population of multimorbid patients. The findings are therefore most generalizable to adults with multimorbidity in contact with hospital services. Although eligibility was open to all adults (≥18 years), relatively few younger participants were included, and the mean age was close to 70 years. Frailty and multimorbidity are strongly age-related, and our age-stratified analyses should therefore be regarded as exploratory. Consequently, conclusions regarding age-related differences must be interpreted with caution. Due to the COVID-19 pandemic, the study inclusion was paused after 230 participants had been included, and consequently, the 180-day follow-up period for these participants was partially during the first COVID lockdown, where the risk of COVID infection affected healthcare utilization by limiting physical consultations. The remaining participants were included after all COVID restrictions had ceased. The sample size was pragmatic, limited by the recruitment window and prematurely curtailed by the COVID-19 pandemic. Based on the effect estimates and confidence intervals, the main analysis appears reasonably powered, whereas the subgroup analyses, particularly among younger adults, seem underpowered and should be interpreted with caution.
Cause-specific Cox regression was chosen as the primary approach, reflecting our etiologic aim. However, complementary Fine–Gray models produced similar results, underscoring the robustness of our findings irrespective of competing risk methodology.
Patients with dementia were excluded, limiting the external validity of our results. However, use of the TFI is not feasible in patients with severe dementia, considering the self-reported nature of this instrument. Such patients should always be considered frail, and the level of frailty often corresponds to the severity of the cognitive impairment and the degree of physical frailty.17,22 Only two general practitioners out of 15 registered clinics in the municipality participated in the recruitment. However, general practitioners in Denmark typically have several thousand patients in their practice, and as generalists, they cover a broad spectrum of the general population. Therefore, we find the risk of selection bias due to this circumstance to be minimal.
The completeness of the TFI at baseline was 100% due to meticulous attention to detail shown by the responsible personnel during inclusion of participants.
Conclusion
Frailty measured by TFI was associated with adverse outcomes in an adult multimorbid population. The risk of mortality, 30-day readmission, and healthcare utilization was increased with every one-point increase in TFI, and exploratory age-stratified analyses suggested possible differences in associations by age, but estimates for younger adults should be interpreted with caution. Our findings support the potential value of TFI as a brief multidimensional tool to identify high-risk multimorbid patients across care settings.
Supplemental material
Supplemental Material - Association of the Tilburg Frailty Indicator with mortality, readmission, and healthcare contacts in adult patients with multimorbidity
Supplemental Material for Association of the Tilburg Frailty Indicator with mortality, readmission, and healthcare contacts in adult patients with multimorbidity by Thomas J. Hjelholt, Thomas Veedfald, Anne Frølich, Charlotte Appel, Henrik Holm Thomsen, Anders Prior, Anne Dorthe Bjerrum, Marianne Balsby
Footnotes
Acknowledgments
We would like to thank all patients who participated in the study. We would also like to thank our collaborators at Healthcare Center Skive, Skive Municipality, Viborg Regional Hospital, and the participating general practitioners. A special thanks to study nurse Jette Knudsgård Hørup.
ORCID iDs
Ethical considerations
According to Danish law, approval from ethic committee is not required for the study. The study was approved by the Danish Data Protection Agency (case number: 1-16-02-225-19) and was carried out in accordance with the Helsinki Declaration and Good Clinical Practice guidelines.
Consent to participate
Informed consent was obtained from all participants included in the study.
Author contributions
LKN, MB, and ADB designed and conducted the study. TV, TJH, ASJ, AF, CWA, HHT, and AP provided scientific input for the study design, analysis, and interpretation. ASJ and TJH performed the statistical analyses. All authors participated in the development of the manuscript and approved the final version before submission.
Funding
This study was funded by the Regional Council in Central Denmark Region, Clinical Academic Group for Multimorbidity, Regional Hospital Central Jutland and the Health Research Foundation of Central Denmark Region.
Declaration of conflicting interests
All authors declare no conflicts of interest.
Data Availability Statement
To protect the privacy of patients, individual level data are not allowed to be publicly disclosed. The statistical code can be made available upon reasonable request.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
