The prevalence of frailty and pre-frailty among geriatric hospital inpatients and its association with economic prosperity and healthcare expenditure: A systematic review and meta-analysis of 467,779 geriatric hospital inpatients

Background: Frailty is a common and clinically significant condition among geriatric populations. Although well-evidenced pooled estimates of the prevalence of frailty exist within various settings and populations, presently there are none assessing the overall prevalence of frailty among geriatric hospital inpatients. The purpose of this review was to systematically search and analyse the prevalence of frailty among geriatric hospital inpatients within the literature and examine its associations with national economic indicators. Methods: Systematic searches were conducted on Ovid, Web of Science, Scopus, CINAHL Plus, and the Cochrane Library, encompassing all literature published prior to 22 November 2018, supplemented with manual reference searches. Included studies utilised a validated operational definition of frailty, reported the prevalence of frailty, had a minimum age ≥ 65 years, attempted to assess the whole ward/clinical population, and occurred among hospital inpatients. Two reviewers independently extracted data and assessed study quality. Results: Ninety-six studies with a pooled sample of 467,779 geriatric hospital inpatients were included. The median critical appraisal score was 8/9 (range 7 – 9). The pooled prevalence of frailty, and pre-frailty, among geriatric hospital inpatients was 47.4% (95% CI 43.7 – 51.1%), and 25.8% (95% CI 22.0 – 29.6%), respectively. Significant differences were observed in the prevalence of frailty stratified by age, prevalent morbidity, ward type, clinical population, and operational definition. No significant differences were observed in stratified analyses by sex or continent, or significant associations between the prevalence of frailty and economic indicators. Conclusions: Frailty is highly prevalent among geriatric hospital inpatients. High heterogeneity exists within this setting based on various clinical and demographic characteristics. Pooled estimates reported in this review place the prevalence of frailty among geriatric hospital inpatients between that reported for community-dwelling older adults and older adults in nursing homes, outlining an increase in the relative prevalence of frailty with progression through the healthcare system.


Introduction
Frailty is a multi-dimensional and dynamic condition, theoretically defined as a state of increased vulnerability, resulting from ageassociated declines in reserve and function across multiple physiologic systems such that the ability to cope with every day or acute stressors is compromised (Xue, 2011).Although declines in physiological reserve are associated with senescence in the normal ageing process, frailty is an extreme consequence of this process, where this decline is accelerated and homeostatic responses begin to fail (Ferrucci et al., 2002;Taffett, 2003).Frailty is a common and clinically significant condition amongst geriatric populations, predominantly due to its association with adverse health outcomes, such as hospitalisation, disability, and mortality (Clegg et al., 2013;Fried et al., 2001;Gill et al., 2006;Rodriguez-Mañas, Fried, 2015;Shamliyan et al., 2012;Sourial et al., 2013;Sternberg et al., 2011).
Recently, a scoping review reported a median frailty prevalence of 49% (range 34-69%) in acute care hospital settings (Theou et al., 2018).However, this review had a number of methodological limitations, including the inclusion of the entire sample of any study with a single participant ≥ 65 years, where up to 50% of the sample were not hospital inpatients, and studies that did not report on the method of frailty assessment.Similarly, a recently published systematic review and meta-analysis which examined the prevalence of pre-frailty and frailty together among hospitalised older adults, in studies which also assessed undernutrition risk, found a mean prevalence of 84%, but with limited data from only 11 studies (n = 2,725 patients) eligible for meta-analysis (Ligthart-Melis et al., 2020).
Consequently, there is an evident need for more robust and comprehensive research to thoroughly assess the prevalence of frailty within the overall population of geriatric hospital inpatients.This constitutes an important gap in the literature, which needs to be addressed, with an enhanced understanding regarding the prevalence of a condition within a specific setting, providing a number of potential consequential utilities.These include the enhanced ability to contribute to improvements in the planning and orientation of organisational structures and resources, to meet the needs of populations.This is particularly true regarding the ability to tailor services within settings to the needs of service users.For example, specifically with regard to frailty among geriatric hospital inpatients, the potential implementation of exercise rehabilitation treatments within this setting for this cohort; with physical activity and exercise being proposed as potentially offering the best form of treatment for frail older adults (Theou et al., 2011).Moreover, exercise has been shown to be capable of reducing, and even reversing frailty within older adults, and reversing functional decline associated with hospitalisation among acute geriatric inpatients (Fiatarone et al., 1994;Martínez-Velilla et al., 2018;Tarazona-Santabalbina et al., 2016).
As such, the purpose of this review was to systematically search and analyse the prevalence of frailty among geriatric populations (aged ≥ 65 years) within inpatient hospital settings within the literature.If a meta-analysis proved possible, the aim of this review was also to synthesise pooled estimates of the prevalence of frailty and pre-frailty, as well as the prevalence of frailty stratified by age, sex, operational frailty definition, prevalent morbidities, ward type, clinical population, and geographic location, among geriatric hospital inpatients.Additionally, this review examined the association between the prevalence of frailty among geriatric hospital inpatients, and gross domestic product (GDP) per capita purchasing power parity (PPP) and health care expenditure per capita PPP.Preliminary research into these areas have shown frailty in the community to be correlated with economic indicators (Theou et al., 2013), but note that more research is needed in this regard to better understand this relationship.

Methods
This systematic review and meta-analysis was designed and conducted in accordance with PRISMA standards (Moher et al., 2009;Page et al., 2021).A comprehensive review protocol was developed and adhered to for all steps of this review, and has been published as a protocol paper elsewhere (Doody, Aunger et al., 2019).

Data sources and searches
Searches were conducted on the platforms of Ovid (incorporating the databases of Journals @Ovid full text, EMBASE, CAB abstracts, Ovid MEDLINE® In process and other non-indexed citations, Ovid MED-LINE®, and PsycINFO), and Web of Science (incorporating the databases of Science Citation Index Expanded (SCI-Expanded), Conference Proceedings Citation Index -Science (CRI-S), and Emerging Sources Citation Index (ESCI)), and the databases of CINAHL Plus, Scopus, and the Cochrane Library databases (the Cochrane Database of Systematic Reviews (CDSR), the Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Methodology Register (CMR), the Database of Abstracts of Reviews of Effect (DARE), Health Technology Assessment database (HTA) and the NHS Economic Evaluation Database (EED)), encompassing all available literature published prior to 22/11/2018 (Appendix 1), and supplemented with manual reference searches of all included articles.

Study selection
Inclusion criteria required studies to have: a minimum age of ≥ 65 years; utilised a clearly defined and validated operational definition for the classification of frailty (i.e., one which takes into consideration the multi-dimensional nature of the condition, and has been specifically validated for the assessment of frailty; either through comparison with existing validated frailty tools, or its predictive value regarding negative health outcomes associated with frailty); either assessed (or attempted to assess) the whole ward, department, unit, hospital, or specific clinical population, or employed some form of randomised selection of participants; occurred within a hospital setting, in, or including, hospital inpatients (operationally defined as any patient admitted to hospital who remains overnight, or were initially expected to remain overnight), and; report the prevalence of frailty or provide sufficient data to allow its calculation.If a study examined a mixed cohort, only data pertaining to hospital inpatients were included in this review.Exclusion criteria were all studies whose full text was not available in the English language, and studies where the sample were not hospital inpatients (i.e., outpatients, day patients, or community-dwelling individuals).
Prior to the commencement of title and abstract screening by three independent reviewers (PD, EA, and JA), duplicates were removed using EndNote (VX 8.2).The succeeding reduced list of studies was further manually screened for the removal of any remaining duplicates.All reviewers were provided with an instructional screening form, and a.ris file containing all studies captured within the platform and database searches.This screening form outlined the eligibility criteria and P. Doody et al. instructions on setting up the file for screening within a reference manager (Appendix 2).
The title and abstract of all studies were independently screened by the three reviewers, with each reviewer placing potentially eligible studies into a separate folder.On completion, potentially eligible studies from all three reviewers were placed into a 'master folder' and the results collated.Duplicates were removed, leaving the final combined list of studies for the full text screening phase.All reviewers independently screened the full text of remaining studies utilising the screening form and maintained separate files for included and excluded studies (including reasons), as well as for studies for which they believed there was need to contact the authors for clarification or additional information.
On completion, a full text screening master file was formulated by the lead reviewer displaying each reviewer's full text screening decision for each study (Appendix 3).All three reviewers subsequently met to discuss the decisions of each study and endeavoured to come to an agreement on studies for which there was not initial unanimous consensus.During this process, a full list of included (Appendix 3) and excluded studies (with reasons) (Appendix 4), and studies for which reviewers agreed to contact authors for additional information or clarification (Appendix 5) was formed by the lead reviewer.Subsequently, the lead reviewer contacted the relevant study authors and, on receipt of clarification or additional information, forwarded this information to the two other reviewers for independent assessment.All reviewers subsequently met to further discuss and come to resolution on the eligibility of all such studies (Appendix 5).
Manual screening was also employed by reviewers and included the reference lists of all included studies, as well as excluded but potentially relevant studies or systematic reviews captured within the screening.As part of the grey literature search of this review, in process publications were also searched and conference abstracts followed up with authors to ascertain if full text's relating to these data were available.Studies of the same cohort were included only once, specifically, the study which provided the most information about the cohort relevant to this review.In the event two or more studies reported an identical quantity of data relevant to the review, the study which was published first was given precedence for inclusion.

Data extraction and quality assessment
Data extraction of eligible studies was performed by two reviewers (PD and BS) independently.In the event of any discrepancies between the two reviewers, an attempt was made to reach a consensus by discussion.A contingency plan was in place, regarding obtainment of the opinion of a third reviewer, in the event that a full consensus could not be reached between the two reviewers after an exhaustive discussion, with the majority consensus taken.However, ultimately, this contingency plan was not utilised, as both reviewers came to agreement after discussion in all cases.
The following data, where available, were extracted from all eligible studies.If any data were not immediately available, the authors of these studies were contacted in an attempt to retrieve all applicable data: Study details: authors, year of publication, study title, journal of publication, and aim.Study methods: setting, ward/department/unit/ hospital type, clinical population, study design, recruitment duration, subject characteristics (age of participants (mean and standard deviation, range)), sex (proportion of male/female participants), country/ continent, sample size, diagnosis/prevalent morbidity (if applicable), any other relevant characteristics, criteria utilised for the operational definition of frailty.Results: Number of frail participants, number of prefrail participants, number of robust/non-frail participants, prevalence of frailty, prevalence of pre-frailty, prevalence of robustness/non-frailty, number of male participants, number of frail male participants, number of pre-frail male participants, number of non-frail/robust male participants, prevalence of frailty in male participants, prevalence of pre-frailty in male participants, prevalence of non-frailty/robustness in male participants, number of female participants, number of frail female participants, number of pre-frail female participants, number of nonfrail/robust female participants, prevalence of frailty in female participants, prevalence of pre-frailty in female participants, prevalence of non-frailty/robustness in female participants, and finally authors ' and reviewers' comments (Appendix 6).
External to the studies, data were additionally extracted with regard to the 5-year average GDP per capita PPP (current international $) of the country in which each study was conducted, incorporating the 5 years directly preceding the commencement of recruitment to the study (International Monetary Fund, 2019).External data were also extracted with regard to the 5-year average healthcare expenditure per capita PPP (current international $) of the country in which each study takes place, incorporating the 5-years directly preceding the commencement of recruitment to the study (World Health Organisation, 2019).Each calendar year of the study was also included provided recruitment continued through to > 6 months in the preceding year (Appendix 6).
The quality of eligible studies was independently assessed by two reviewers (PD and EA) using the Joanna Briggs Institute critical appraisal tool for studies reporting prevalence data (Munn et al., 2015) (Appendix 7).In the event of any discrepancies between the two reviewers, an attempt was made to reach a consensus by discussion.Similar to the process for data extraction, a contingency plan was in place to obtain the opinion of a third reviewer, in the event a consensus could not be reached, with the proceeding majority consensus taken as final.However, ultimately this contingency plan was not utilised, as the two reviewers came to successful resolution in all cases.

Quantitative synthesis (meta-analysis)
Where a sufficient quantity of identified studies were comparable, meta-analysis, pooling the aggregated data from each study, was performed using Review Manager (RevMan) version 5.3 (The Nordic Cochrane Centre -The Cochrane Collaboration, 2014).Clinical heterogeneity was assessed by two reviewers based on their judgement of the available data, and any disagreements discussed thoroughly with the aim of reaching unanimous consensus, which occurred in all cases.Statistical heterogeneity was assessed through the utilisation of a Cochran Q test and considered present at p < 0.05 (Higgins and Thompson, 2002).An I 2 test was performed to assess the magnitude of this heterogeneity, with I 2 values of 25%, 50%, and 75% being considered low, moderate, and high, respectively (Higgins et al., 2003).Where the Cochran Q statistic test detected statistically significant heterogeneity, combined with the researcher's assessments concluding that variation in effect size between studies could not be fully explained by the sampling error within each study, i.e., that the true effect-size was not identical for all studies, a randomised-effects model was utilised (Borenstein et al., 2010).
Stratified analysis was also conducted according to age (65-74 years, 75-84 years, and 85 + years), sex, operational frailty definition, ward type, prevalent morbidity, clinical population, and geographic location (country and continent) where possible.These variables were specifically chosen for stratified analysis due to an enhanced knowledge of these areas being of practical utility to researchers and clinicians; stemming from empirical evidence persistently showing variation in these factors to impact on the prevalence of frailty (Andela et al., 2010;Purser et al., 2006;Santos-Eggimann et al., 2009).As such, stratified analysis facilitated provision of a more in-depth and thorough insight into the prevalence of frailty among geriatric patients within this setting.
Clinical heterogeneity for stratified analysis was assessed by two reviewers based on their judgement of the available data.Any initial disagreements were discussed thoroughly, with a unanimous consensus reached in all cases.Statistical heterogeneity for stratified analysis was assessed as above through the utilisation of Cochran Q tests, with I 2 tests P. Doody et al. performed to assess the magnitude of this heterogeneity (Higgins and Thompson, 2002;Higgins et al., 2003).All pooled estimates of the prevalence of frailty were reported with 95% confidence intervals.
Correlation analysis was also employed to examine the relationship between the prevalence of frailty among geriatric hospital inpatients and economic prosperity (GDP per capita PPP) (current international $), and healthcare expenditure (per capita PPP) (current international $).In addition, multi-linear regression analysis was employed to examine the predictive value between economic prosperity and healthcare expenditure and the prevalence of frailty among geriatric inpatients, using IBM Statistical Package for Social Sciences (SPSS) version 27 (IBM Corp, 2020).

Qualitative synthesis
A brief systematic narrative analysis of all outcomes was also performed, with findings presented in both textual and tabular formats.

Role of the funding source
This review was supported by the European Commission Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement (675003).The funding source had no role in the design, conduct, or reporting of the review, or the decision to publish the manuscript.The authors have no competing interests to disclose.

Results
Systematic searches yielded a combined total of 4,757 results, of which 1,549 were removed as duplicates.Four additional articles were identified within the reference list of included studies during manual screening.The remaining 3,208 articles were screened by title and abstract by the three independent reviewers and the results collated, leaving 655 studies for full text screening.344 of these articles were initially excluded due to ineligibility: minimum age < 65 years (n = 122); utilisation of a non-validated operational definition for the classification of frailty (n = 91); sample were not hospital inpatients at the time of frailty assessment (n = 89); did not assess (or attempt to assess) the entire ward/clinical population or employ some form of randomised selection of participants (n = 5); other reasons (predominantly duplicate cohorts) (n = 37); multiple (combination of the above reasons) (n = 117).
A further 235 studies screened by full text were deemed to not be initially possible to definitively include or exclude based on available data.As such it was agreed by the three reviewers to contact the study authors for additional information or clarification regarding eligibility.The corresponding author of all 235 studies was contacted via email by the lead reviewer to obtain the relevant additional data, or clarification, to facilitate inclusion / exclusion.A response was received from 99 of the 235 corresponding authors.Of the 136 studies without an initial response from the corresponding author, a second author (typically first or senior author) of all 136 studies were contacted by the lead reviewer, a minimum of 14-days after the initial inquiry to corresponding authors.A response was received for 37 of these 136 studies, giving a combined response rate of 57.9% (n = 136) for the 235 studies.Ultimately this process resulted in an additional 20 studies being deemed eligible for inclusion in the review, resulting in 96 eligible studies in total (Fig. 1).However, this process did add considerably to the timeline for this review beyond the initial search period.(All inquiries to study authors, and responses received are detailed in Appendix 5).
Of these 96 eligible studies, only four initially reported the full range of data sought for stratified analysis.The corresponding author of the remaining 92 studies were contacted in an attempt to obtain these data.If a response was not received within 14-days, a second author was contacted.This process resulted in successful obtainment of additional data for 58 of the 92 studies with initially incomplete data for all elements of stratified analysis (All inquiries to study authors, and responses received are detailed in Appendix 5).
A detailed list of all 96 included studies, reporting selected relevant study characteristics is displayed in Table 1:

Methodological quality assessment
The median score of the Joanna Briggs Institute critical appraisal tool for studies reporting prevalence data for the 96 included studies was 8 out of 9 (range 7-9) (Appendix 7).
Additionally, of the 96 included studies, one study each was conducted among patients identified as primarily possessing dermal (Madni et al., 2018); oral (Martín et al., 2018); and renal (Morton et al., 2018) related morbidities.These studies were not included in the above pooled prevalence analysis stratified by prevalent morbidity due to the lack of multiple comparable data points to facilitate stratified pooled analyses in the above regard.
Additionally, one study each utilised one of the ten additional validated operational definition of frailty.However, these studies were not included in the above pooled prevalence analysis stratified by operational definition due to the lack of multiple comparable data points to facilitate stratified pooled analyses in the above regard.

Association between the prevalence of frailty and economic indicators
A detailed list of all 96 included studies, reporting selected relevant study characteristics regarding the prevalence of frailty and economic indicators is displayed in Table 2:

Gross domestic product per capita purchasing power parity
As data were not normally distributed, a Spearman's rank correlation coefficient was employed to examine the association between the prevalence of frailty among geriatric hospital inpatients and GDP per capita PPP.No significant correlations were observed between the prevalence of frailty among geriatric hospital inpatients and GDP per capita PPP (r = − 0.081, p = 0.452), the prevalence of pre-frailty among geriatric hospital inpatients and GDP per capita PPP (r = 0.107, p = 0.423), or a combination of prevalence of frailty and pre-frailty, and GDP per capita PPP (r = 0.24, p = 0.857).

Health care expenditure per capita purchasing power parity
Similar to the GDP per capita PPP analysis, these data were not normally distributed, and as such a Spearman's rank correlation coefficient was employed to examine the association between the prevalence of frailty among geriatric hospital inpatients and healthcare expenditure per capita PPP.No significant correlations were observed between the prevalence of frailty among geriatric hospital inpatients and healthcare expenditure per capita PPP (r = 0-0.197,p = 0.071), the prevalence of pre-frailty among geriatric hospital inpatients and healthcare expenditure per capita PPP (r = 0.220, p = 0.097), or a combination of prevalence of frailty and pre-frailty, and healthcare expenditure per capita PPP (r = 0-0.146,p = 0.275).

Discussion and conclusions
In this systematic review and meta-analysis, 96 studies were identified with an overall pooled sample of 467,779 geriatric hospital inpatients aged ≥ 65 years, which utilised a validated operational     definition of frailty, attempted to assess the whole ward/clinical population, occurred in a hospital setting, in or including hospital inpatients, and reported, or provided sufficient information to allow the calculation of, the prevalence of frailty.Included studies were conducted in 21 countries, across five continents.The overall pooled estimate of frailty was 47.4%; although this varied significantly based on prevalent morbidities, age, ward type, clinical population, and the operational definition utilised for the classification of frailty.To the authors' knowledge, this is the largest and most comprehensive systematic review and metaanalysis of the prevalence of frailty among older adults conducted in any setting, and the first well-evidenced systematic review and metaanalysis among geriatric hospital inpatients.The overall pooled prevalence estimate of frailty of 47.4%, places the prevalence of frailty among geriatric hospital inpatients between that reported for community-dwelling older adults at 10.7% (Collard et al., 2012), and older adults in nursing homes at 52.3% (Kojima, 2015); outlining an increase in the relative prevalence of frailty with progression through the healthcare system.The overall pooled prevalence of pre-frailty of 25.8% is lower than that reported for both community-dwelling older adults at 41.6% (Collard et al., 2012), and nursing home residents at 40.2% (Kojima, 2015); while the combined prevalence estimates of both frailty and pre-frailty increase from 52.3% among community-dwelling older adults, to 73.2% among geriatric hospital inpatients, and to 92.5% among nursing home residents.This underlines that differences in the relative prevalence of frailty status between community, and hospital inpatient settings, are the result of an increase in the relative prevalence of frailty, and similar reductions in the relative prevalence of both pre-frailty and robustness.However, differences in the relative prevalence of frailty status between hospital inpatient and nursing home settings, these data show, are primarily the result of a relative increase in the prevalence of pre-frailty, and reductions in the prevalence of robustness.
The overall pooled frailty, and pre-frailty, prevalence estimates of 47.4% (95% CI 43.7-51.1%),and 25.8% (95% CI 22.0-29.6%)respectively, are relatively consistent with, though more precise than, estimates reported within a recent systematic review and meta-analysis which examined the prevalence of frailty and pre-frailty among hospitalised older adults in 11 studies which also assessed undernutrition risk, at 47% (95% CI 37-57%) and 36% (95% CI 29-44) respectively (Ligthart-Melis et al., 2020).Similarly, the pooled prevalence estimates of frailty on acute wards of 51.1% (95% CI-35.9-66.2%),as well as among all acute hospital inpatients, of 47.3% (95% CI 42.8-51.8%),are relatively consistent with findings of a recent scoping review, which reported a median frailty prevalence of 49% (range 34-69%) in acute care hospital settings (Theou et al., 2018).Further, no significant differences in the prevalence of frailty were observed in stratified analyses by sex.This is in contrast to systematic reviews and meta-analysis of the prevalence of frailty among community-dwelling older adults (Collard et al., 2012;Siriwardhana et al., 2018;He et al., 2019).However, consistent with the findings of systematic reviews and meta-analysis among other clinical populations of older adults such as nursing home residents (Kojima et al., 2015).These findings contribute to the literature illustrating sex differences in the prevalence of frailty among community dwelling older adults, may dissipate among clinical geriatric populations.
No significant associations were observed between the prevalence of frailty among geriatric hospital inpatients and GDP per capita PPP, and healthcare expenditure per capita PPP.This contrasts with previous research among community-dwelling older adults within 14 European countries, and Israel, conducted utilising data from the Survey of Health, Ageing, and Retirement in Europe (SHARE).This cross-sectional analysis examined the association between GDP per capita PPP, and health expenditure as a percentage of GDP, and the prevalence of frailty among community-dwelling older adults assessed by the frailty index.Fifteen observations of the weighted national prevalence of frailty for community-dwelling older adults in each country were correlated with both national economic indicators, and reported strong correlation between GDP per capita PPP (r = − 0.71, p < 0.01), and healthcare expenditure as a percentage of GDP (r = − 0.63, p < 0.05), and the prevalence of frailty among community-dwelling older adults (Theou et al., 2013).
It is possible that these associations, while present in the community, are not present in inpatient hospital settings.Given the inherent nature of hospital inpatient settings, i.e., institutions for chronically or acutely unwell patients, this association may be more sensitive among the general population of community-dwelling older adults; however, more large-scale and comprehensive studies are required in a variety of settings.Given the lack of statistically significant differences in the pooled prevalence of frailty stratified by continent within this present review alone, this may not be surprising, however, significant differences in the prevalence of frailty were observed between countries.In this regard, an additional limitation of these analyses is that included studies were predominantly from economically-developed countries, as there is presently limited evidence regarding the prevalence of frailty in lowincome countries; an issue which has been observed previously in a meta-analysis of the prevalence of frailty among community-dwelling older adults in middle-, and low-income countries (Siriwardhana et al., 2018).To the authors' knowledge, this present review is the first study of any design to examine the association between the prevalence of frailty among geriatric hospital inpatients and national economic indicators.It has been postulated that increases in economic prosperity may limit the prevalence and burden of frailty within national health systems (Theou et al., 2013).However, these findings bring this postulation into question among geriatric hospital inpatients, and as such reliance of non-direct intervention such as economic development, to improve the prevalence and burden of frailty on health systems alone, appears, at least partially, to be misplaced.As such the findings of this review further suggests the need for more direct interventions to address the burden of frailty among this population.Future research examining the prevalence of frailty among geriatric hospital inpatients in low-income countries may facilitate further elucidation of this relationship, as these data become available for less economically developed regions of the world.Although, it may be that this relationship does not exist in the same capacity as it appears to among community-dwelling older adults, to the authors' knowledge the study by Theou et al. (2013) is the only study to previously examine this relationship.As such, additional studies, in a variety of settings, may aid in elucidating this relationship further.This systematic review and meta-analysis had many strengths, including extensive systematic searches of 17 databases; manual screening of the reference lists of all included articles (and relevant studies or systematic reviews captured within platform and database searches); the screening of grey literature, including in process publications, and conference abstracts, which were followed up with study authors to ascertain if a full text relating to these data were available; employment of three independent reviewers during the screening phase of the review, ensuring high internal reliability and consistency of included articles; the utilisation of meticulously defined eligibility criteria; the employment of two independent data extractors and quality assessors; an extensive data procurement strategy, including contacting 517 authors to obtain additional information relevant to inclusion within different aspects of the review; robust analysis of the prevalence of frailty stratified by clinically useful variables; and a comprehensive record of all information pertaining to the review process available as supplementary materials.
This review also had a number of important limitations that should be considered when interpreting these findings.Firstly, only studies with a full text available in the English language were eligible for inclusion, as this was the only shared language between the three independent reviewers.As such included studies may be relatively overrepresentative of Western nations (Europe, Australasia, and the Americas), and there is a possibility that this review does not include otherwise eligible studies whose full texts are not available in the English language.However, in this regard, any potentially eligible studies, with an English translated abstract, and full text in other languages, were followed up with study authors in an attempt to obtain an English full text to facilitate thorough screening.Secondly, high heterogeneity was reported across many analyses, and persisted across many univariate stratification analyses.Thirdly, a strength, but also a limitation of this review, was with regard to the specific eligibility criteria employed within this present review, requiring prospectively eligible studies to either assess (or attempt to assess) the whole ward, department, unit, hospital, or specific clinical population, or employ some form of randomised selection of participants.Any exclusion criteria employed within individual studies, in order to meet this criterion, had to meet one of two stipulations: (1) the criterion was essential to defining the clinical population; (2) the criterion is related to insurmountable impracticalities which precluded inclusion of certain individuals.Provided all of a study's exclusion criteria adequately met either of these two stipulations during screening, they were deemed to have sufficiently satisfied the above eligible criterion for the review of having either assessed, or attempted to assess, the entire ward/department/unit/clinical population or employed some form of randomised selection of participants.While such comprehensive stipulations prevented inclusion of any studies with active bias in the recruitment process, those that could be not be recruited in some studies due to impracticalities of inclusion, may also in many cases, be more likely to be frail e.g., those receiving end of life care in a study utilising an objective operational definition for the classification of frailty.Fourtly, an important limitation regarding the economic analyses, is that these data while collected in a systematic manner, incorporating all relevant data which exist in this regard, are unlikely to be precisely nationally representative as they have not (1).assessed the entire population of geriatric hospital inpatients within each country; or (2).been weighted against for example a hypothetical nationally representative databases of geriatric hospital inpatients with regard relevant variables in each country.Future research should further attempt to determine and examine precisely nationally representative data.However, availability of nationally representative data employing appropriate weighting for geriatric hospital inpatients by relevant variables may be difficult, and likely pose substantial feasibility issues regarding accurate facilitation, particularly across nations, without considerable resource investment.Finally, while contributing substantially to the obtainment of further data for these analyses, contacting several hundred authors for these additional data added to the timeline for this review beyond the initial search period.
Through providing a highly detailed analysis of the prevalence of frailty among older people within this setting, the aim of this present review was to provide a resource, which can aid in the facilitation of improvements in the planning, and orientation of organisational structures and resources, to meet the needs of this population, and ultimately enhance the care of older adults with frailty in inpatient hospital settings.Future research, particularly in developing countries, may help to further elucidate any potential relationship regarding national economic indicators and the prevalence of frailty among geriatric hospital inpatients.As frailty is a relatively new concept, particularly as an operationally defined one, with most studies cited within this review published in the past 20 years, it is the intention of the authors to update this review periodically, to examine the potential change in frailty over time, particularly as it relates to national policy directives, and economic indicators as data become available for less developed regions of the world.
More generally the authors have several recommendations with regard to improving reporting in future frailty research among hospitalised older adults, as well as within other settings.These recommendations arise from the following issues which are persistent in the frailty literature, and were continually observed during the screening process for this review (Appendix 3-6): (1) studies often reported participants as frail without a frailty assessment; (2) studies often claimed to utilise validated operational definitions for the classification of frailty, however, adapt these definitions, or classification criteria, which resulted in the definitions becoming not only non-standardised, but also non-validated; (3) the use of the nomenclature for different operational definitions of frailty varied widely, even among studies utilising the same operational definition; (4) often, useful data regarding P. Doody et al. the prevalence of frailty (such as pre-frailty, a sex breakdown of frailty, or occasionally the overall prevalence of frailty itself) were not reported.
Reporting in this regard may be improved by a brief standardised checklist for studies reporting frailty data.The authors suggest the following items for inclusion: (1) accurate citation of the validation study for the specific operational definition utilised for the classification of frailty; (2) accurate use of the nomenclature of the operational definition of frailty utilised in accordance with the initial validation study to maintain reliability and validity, or prominent subsequent study establishing the nomenclature; (3) reporting of the number of frail, pre-frail (if applicable), and robust participants; (4) a sex breakdown of the number of frail, pre-frail, and robust participants.
Given the association of frailty at the individual level with increased healthcare costs, combined with projected population demographics, future research should focus on interventions to reduce the prevalence of frailty among geriatric hospital inpatients.Particularly as hospitalisation is associated with a further decline in functional capacity, interventions to mitigate this decline, and reduce the rate of subsequent rehospitalisation of older adults with frailty are important issues to be addressed.This is particularly the case as future demographic trends predict the overall number of frail older adults to increase dramatically in developed countries in the coming decades as the population ages (Hoogendijk et al., 2019).This will be further exacerbated by declining fertility rates in economically developed countries, which are projected to cause an increase in dependency ratios across the developed world (Murray et al., 2018;United Nations Department of Economic and Social Affairs Population Division, 2019;Vollset et al., 2020).It is in this context that frailty, particularly in older age, has been described as "without question, one of the most serious public health challenges we will face in this coming century" (Dent et al., 2019).
In summary, this systematic review and meta-analysis found that approximately half of all hospital inpatients aged ≥ 65 years are frail, and approximately another 25% are pre-frail.These patients may benefit from interventions targeted at improving frailty status and preventing the functional decline associated with hospitalisation in this population, which can lead to further functional deterioration, recurrent readmission, and adverse health outcomes among these patients.

Registration
PROSPERO registration number 79202.

Funding source
European Commission Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement (675003).

Fig. 1 .
Fig. 1.PRISMA flow diagram of systematic review and meta-analysis process.

P
.Doody et al.

Fig. 2 .
Fig. 2.Forest plot of the prevalence of frailty in the 96 studies identified through the systematic review process, including a total of 467,779 geriatric hospital inpatients.

Table 1
Selected characteristics of the 96 included studies.
2015; Sanchis et al., 2015; Sikder P. Doody et al. (continued on next page) P. Doody et al.Table 1 (continued ) (continued on next page) P. Doody et al. (continued on next page) P. Doody et al.Table 1 (continued ) (continued on next page) P. Doody et al. (continued on next page) P. Doody et al. (continued on next page) P. Doody et al.Table 1 (continued ) (continued on next page) P. Doody et al.Table 1 (continued )

Table 1 (
continued ) * =Data not initially reported, or possible to derive from available data.Obtained, or derived, from correspondence with study authors.

Table 2
Selected study characteristics relating to economic analysis of included studies.

Table 2
(continued ) = 5 years prior to commencement of data collection for the study.Each calendar year of the study was also be included provided recruitment continues through to > 6 months in the preceding year.**= Data not initially reported, or possible to derive from available data.Obtained, or derived, from correspondence with study authors.