Abstract
Aims
Frailty greatly increases the risk of adverse outcome of trauma in older people. Frailty detection tools appear to be unsuitable for use in traumatically injured older patients. We therefore aimed to develop a method for detecting frailty in older people sustaining trauma using routinely collected clinical data.
Methods
We analyzed prospectively collected registry data from 2,108 patients aged ≥ 65 years who were admitted to a single major trauma centre over five years (1 October 2015 to 31 July 2020). We divided the sample equally into two, creating derivation and validation samples. In the derivation sample, we performed univariate analyses followed by multivariate regression, starting with 27 clinical variables in the registry to predict Clinical Frailty Scale (CFS; range 1 to 9) scores. Bland-Altman analyses were performed in the validation cohort to evaluate any biases between the Nottingham Trauma Frailty Index (NTFI) and the CFS.
Results
In the derivation cohort, five of the 27 variables were strongly predictive of the CFS (regression coefficient B = 6.383 (95% confidence interval 5.03 to 7.74), p < 0.001): age, Abbreviated Mental Test score, admission haemoglobin concentration (g/l), pre-admission mobility (needs assistance or not), and mechanism of injury (falls from standing height). In the validation cohort, there was strong agreement between the NTFI and the CFS (mean difference 0.02) with no apparent systematic bias.
Conclusion
We have developed a clinically applicable tool using easily and routinely measured physiological and functional parameters, which clinicians and researchers can use to guide patient care and to stratify the analysis of quality improvement and research projects.
Cite this article: Bone Joint J 2024;106-B(4):412–418.
Take home message
We developed a novel automated tool that can accurately and efficiently assess frailty in trauma settings.
The use of the Nottingham Trauma Frailty Index has the capacity to modify the decision-making process in trauma treatment by offering clinicians a rapid and unbiased assessment of frailty.
This assessment may then be used to tailor therapies to individual patients, ultimately leading to improved outcomes.
Introduction
Older people are at high risk of poor outcomes after major trauma due to both comorbidity and frailty. Frailty is the age-associated loss of resilience to challenge, due to multiple organ and system deficits. Frailty is an independent predictor of many adverse outcomes in trauma in older people, including in-hospital complications, length of stay, unfavourable discharge destination, and death.1
Many frailty detection tools are unsuitable for use in acute settings such as trauma. Tools based upon the Fried frailty phenotype require variables that are unavailable, such as a pre-injury gait speed,2 and others using the Rockwood cumulative deficit3 or hybrid approaches4 require historical information about previous medical history of functioning that also might not be immediately or reliably available in the acute setting.
There are a range of prognostic tools for measuring outcomes in the hip fracture population, most notably work from our unit on the Nottingham Hip Fracture Score (NHFS).5 The NHFS predicts death and length of stay in hip fracture and other fragility fracture patients.6-8 However, measures of frailty are increasingly being used to guide management in populations with a high prevalence of frailty (such as in the National Hip Fracture Database9 and in response to UK policy),10 as they are better at taking account of the increased vulnerability due to multisystem deficits that characterizes the frailty state than other tools. The Clinical Frailty Scale (CFS)11 is one of the most commonly used frailty tools in acute settings, with a score ranging from 1 to 9, and it has been shown to predict a range of outcomes in hip fracture and other trauma patients. The application of CFS requires staff to be uniformly trained in its use because it involves subjective judgement, otherwise its accuracy may be compromised when used by untrained personnel or different staff.12 However, we have previously demonstrated that CFS in measuring frailty can provide useful predictive information for fitter patients, with a plateau in higher frailty hip fracture and potential difficulties in clinicians distinguishing between higher frailty grades in the acute setting.13
One potential way to improve the diagnosis and measurement of frailty in acute trauma in clinical practice is to use routinely collected clinical variables, as this would negate the need to train staff, reduce subjectivity, and would be likely to increase the proportion of patients with a frailty assessment. Researchers too could use such an approach to define patients with frailty in clinical datasets where there is no formal frailty measurement. We therefore set out to develop an easy-to-administer, reliable, and valid objective tool to measure frailty in the older trauma patient, as distinct from a prognostic score such as the NHFS.
In this study, we aimed to identify routinely collected variables that are predictive of frailty and, based upon these, develop a frailty scale suitable for use in older patients with traumatic injury.
Methods
We analyzed prospectively collected registry data from 2,108 patients admitted to a single major trauma centre over five years. We divided the sample equally into two, creating derivation and validation samples. In the derivation sample, we performed univariate screening followed by multivariate regression analyses using clinical variables to predict our gold standard, the CFS. From this, we created a model to calculate the Nottingham Trauma Frailty Index (NTFI). To assess for systematic measurement biases, we used Bland-Altman analyses in the validation cohort, comparing the NTFI and the CFS.
The Hospital Audit Committee approved this study (approval number 21-040 C). We report our results in accordance with the Transparent Reporting of a multivariable Prediction model for Individual Prognosis or Diagnosis.14
Source of data
The study used an anonymized data extract that included 7,441 trauma patients sequentially admitted to a level 1 trauma centre between 1 October 2015 and 31 July 2020. The data were extracted from the Nottingham Trauma Registry, which has been previously described and combined data from three routinely collected clinical patient groups: fragility fractures, hip fractures, and adult trauma. There were 7,441 adult fractures; 6,002 were fragility fractures, of which 3,801 were hip fractures. We included all datasets available locally to ensure the predictive score was generalizable to the whole available older population. While each dataset is heterogeneous, combining and linking the data yields the whole older population for our centre presenting with a fragility fracture.
Participants
We selected records used for this study from all those in the registry for the study period using the following eligibility criteria: age 65 years and above; presenting with a traumatic fracture requiring hospital admission; and a CFS completed on admission.
Outcome
We aimed to predict frailty as measured by the CFS score,11 recorded on admission in clinical records by clinical staff and subsequently coded by audit staff, all of whom were unaware of this study. This score has nine distinct values, ranging from 1 (robust) to 9 (moribund).
Predictors
We screened all available and feasible variables from the dataset for inclusion in the model. Of the 47 available variables, candidate predictors were 27 variables recorded on admission by clinicians and coded by audit staff, as shown in Table I.
Table I.
Candidate variables grouped by class.
| Demographics | Mental health variables | Chronic diseases | Biological variables | Physical state |
|---|---|---|---|---|
| Age | Abbreviated Mental Test | Diabetes | Glomerular filtration rate | Residential home resident |
| Sex | Alzheimer’s disease | Atrial fibrillation | Platelet count | Living alone |
| Mechanism of injury | Dementia | Hypertension | Lactate level | Nursing home resident |
| Alcohol excess | Osteoporosis | White cell count | Pre-admission mobility | |
| Visual acuity | Malignancy | Haemoglobin level | ||
| Creatinine level | ||||
| Urea level | ||||
| Sodium level | ||||
| Potassium level | ||||
| Calcium level |
Sample size
Different authors advise that multivariate analyses require at least ten to 20 cases per independent variable, and others advise a sample of at least > 500 cases.15 In predictive modelling, however, larger sample sizes result in the development of more robust models, reduce the risk of overfitting, and enhance the precision and accuracy of predictions.16 In our dataset, 2,108/7,441 patients fulfilled our inclusion criteria and had complete data, giving us 1,054 patients in each cohort, which we deemed sufficient for our purposes.
Missing data
Our analyses used complete cases only, because of the many biases introduced by imputation of missing data,17,18 and because we had a sample size that made it unnecessary. We assumed that data in this clinical registry would be missing at random and hence a complete case analysis would not introduce systematic bias.19
Statistical analysis
We cleaned the dataset before we conducted any analyses, by identifying and repairing erroneous data such as outliers, duplicates, and where words rather than figures had been entered.20 We then randomly divided the dataset equally into derivation and validation groups.
In the derivation sample, we initially examined the relationship between the CFS and the candidate variables by visually examining the data using bar charts, and box and scatter plots. We excluded variables showing no obvious relationship to the CFS from further analysis. We conducted univariate analyses on remaining variables to screen for eligibility. As the CFS was normally distributed (Supplementary Figure a), continuous variables were analyzed using Pearson’s correlation coefficient (r), and categorical (binary) variables using the independent-samples t-test. Multiple linear regression analysis was then conducted and included candidate variables with a significant (p < 0.05) univariate association with the CFS, and a good correlation (r > 0.2) for continuous variables or a good effect (Cohen’s d > 0.2) for binary variables.
We used the results of the multiple regression analysis to derive an equation to compute the NTFI, and performed verification tests for homoscedasticity and multicollinearity. We examined homoscedasticity (to determine if the variance of the residual, or error term in a regression model is constant) using a scatter plot of the predicted values and residuals. We looked for multi-collinearity by analyzing variable inflation factors (VIFs) in the results of collinearity statistics.
We subsequently tested the regression model derived from the derivation cohort in the validation cohort, by comparing scores generated by the multiple regression model (the NTFI) with the CFS scores using Bland-Altman plots and analyses.21 We initially tested the normality of the difference between the NTFI and the CFS by histogram (Gaussian), then performed an independent-samples t-test to determine whether there was a statistically significant difference between zero and the mean difference between the two instruments. We then created scatter plots where the x-axis represented the mean of the two measures and the y-axis represented the difference between the two paired measures.21 We analyzed the data using SPSS Statistics v. 28 (IBM, USA).
Results
There were 7,441 records in the register over the study period, of which 2,108 (1,054 in both the derivation and validation groups) were selected for this study using the inclusion criteria (Figure 1). In the derivation and validation samples, 926 and 948 complete records were available for regression analysis, respectively.
Fig. 1
Flowchart illustrating the overall procedure for participants.
Univariate analysis
Following visual inspection of distributions, 14/27 variables (alcohol excess, osteoporosis, Alzheimer’s disease, potassium level, platelet count, visual acuity, hypertension, diabetes, living alone, sodium, creatinine, calcium, lactate, and white cell count) had no apparent relationship with CFS and were not analyzed. Univariate analyses were conducted on the remaining 13 variables (age, sex, pre-admission mobility, mechanism of injury, urea (U), abbreviated mental test (AMT), haemoglobin (Hb), glomerular filtration rate (GFR), atrial fibrillation, dementia, residential care home resident, nursing home resident, malignancy) that were eligible (Tables II and IV). All 13 variables demonstrated a significant relationship with the CFS and were put forward to the multiple regression analysis, with the exception of urea (r = 0.1) and malignancy (d = -0.17) due to the low strength of correlation with CFS.
Table II.
Characteristics of patients in the derivation and validation groups.
| Characteristic | Derivation group | Validation group |
|---|---|---|
| Cases, n | 1,054 | 1,054 |
| Mean age, yrs (range) | 83.58 (65 to 105) | 83.60 (65 to 105) |
| Mean CFS (SD) | 4.56 (1.671) | 4.56 (1.686) |
| Median AMT (IQR) | 8 (5 to 10) | 8 (4 to 10) |
| Median haemoglobin, g/l (IQR) | 123 (111 to 134) | 122 (112 to 133) |
| Median urea, mmol/l (IQR) | 7.1 (5.5 to 9.2) | 6.8 (5.6 to 8.9) |
| Pre-admission mobility assistance, n (%) | 665 (63.1) | 673 (60.4) |
| MOI: fall from standing height, n (%) | 988 (93.7) | 990 (93.9) |
| Female sex, n (%) | 746 (70.8) | 743 (70.5) |
| Atrial fibrillation, n (%) | 194 (18.4) | 234 (22.2) |
| Dementia, n (%) | 290 (27.5) | 286 (27.1) |
| Residential home resident, n (%) | 180 (17.1) | 200 (19) |
| Nursing home resident, n (%) | 100 (9.5) | 109 (10.3) |
| Malignancy, n (%) | 207 (19.6) | 196 (18.6) |
| Median glomerular filtration rate, ml/min (IQR) | 71 (32.3) | 71 (54 to 85) |
-
AMT, Abbreviated Mental Test; CFS, clinical frailty scale; IQR, interquartile range; MOI, mechanism of injury; SD, standard deviation.
Table III.
Pearson correlation results for eligible continuous variables. Dependent variable: Clinical Frailty Scale.
| Independent variables | Pearson correlation (95% CI) | Relationship* | p-value |
|---|---|---|---|
| Age, yrs | 0.35 (0.30 to 0.40) | Moderate positive | < 0.001 |
| Urea, mmol/L | 0.11 (0.04 to 0.17) | Weak positive | < 0.001 |
| Haemoglobin, g/l | -0.25 (-0.31 to -0.19) | Moderate negative | < 0.001 |
| GFR, ml/min | -0.21 (-0.27 to -0.15) | Moderate negative | < 0.001 |
| AMT (0 to 10) | -0.59 (-0.62 to -0.54) | Strong negative | < 0.001 |
-
*
Weak: 0.01 to ≤ 0.20; moderate: > 0.20 to 0.50; strong: > 0.50.
-
AMT, Abbreviated Mental Test; GFR, glomerular filtration rate.
Table IV.
Independent-samples t-test results for eligible categorical variables. Dependent variable: Clinical Frailty Scale.
| Variable | Mean (SD) | Independent-samples t-test | Cohen’s d | p-value | |
|---|---|---|---|---|---|
| Yes | No | ||||
| Female sex | 4.69 (1.65) | 4.25 (1.69) | 3.91 | 0.27 | < 0.001 |
| Atrial fibrillation | 4.89 (1.53) | 4.49 (1.70) | -3.02 | -0.24 | 0.003 |
| Dementia | 5.90 (1.51) | 4.05 (1.51) | -18.49 | -1.28 | < 0.001 |
| Malignancy | 4.79 (1.53) | 4.50 (1.70) | -2.23 | -0.17 | 0.026 |
| Residential home resident | 6.14 (1.20) | 4.24 (1.57) | -15.39 | -1.26 | < 0.001 |
| Nursing home resident | 6.15 (1.23) | 4.39 (1.62) | -10.50 | -1.10 | < 0.001 |
| Falls from standing height | 4.63 (1.64) | 3.54 (1.82) | -4.69 | -0.66 | < 0.001 |
| Pre-admission mobility assistance | 4.94 (1.49) | 3.51 (1.70) | -13.12 | -0.92 | < 0.001 |
-
SD, standard deviation.
Multiple regression analysis
Five of the 11 variables included in the multiple regression analysis (age, mechanism of injury, haemoglobin, AMT, and pre-admission mobility) remained significantly associated with CFS scores (Table V), but six (sex, atrial fibrillation, dementia, care home resident, nursing home resident, and GFR) were non-significant (Figure 2).
Table V.
Multiple linear regression to demonstrate relationship between Clinical Frailty Scale and significant patient characteristics.
| Model | Unstandardized coefficient, B (SE; 95% CI) | Standardized coefficient, β | t | Collinearity statistics | p-value | |
|---|---|---|---|---|---|---|
| Tolerance | VIF | |||||
| Constant | 6.383 (0.69; 5.03 to 7.74) | 9.23 | < 0.001 | |||
| AMT | -0.234 (0.01; -0.26 to -0.21) | -0.50 | -20.42 | 0.919 | 1.088 | < 0.001 |
| Hb | -0.011 (0.00; -0.02 to -0.01) | -0.12 | -4.92 | 0.953 | 1.050 | < 0.001 |
| Age | 0.021 (0.01; 0.01 to 0.03) | 0.10 | 4.06 | 0.885 | 1.130 | < 0.001 |
| MOI | 0.445 (0.20; 0.05 to 0.85) | 0.05 | 2.19 | 0.991 | 1.009 | 0.029 |
| Pre-admission mobility | -1.100 (0.09; -1.27 to -0.93) | -0.31 | -12.64 | 0.927 | 1.079 | < 0.001 |
-
Dependent variable: Clinical Frailty Scale.
-
AMT, Abbreviated Mental Test; CI, confidence interval; Hb, haemoglobin; MOI, mechanism of injury; SE, standard error; VIF, variable inflation factor.
Fig. 2
Variables selection processes.
NTFI specification
The significant regression equation (F (5,920) = 176.74; p < 0.001; β = 6.383 (95% CI 5.028 to 7.740); p < 0.001, R2 = 0.49) was used to generate a predicted frailty score (the NTFI) using the following equation:
NTFI = 6.383 + 0.021 (age) − 1.100 (pre-admission mobility) + 0.445 (MOI) − 0.234 (AMT) – 0.011 (haemoglobin)
where pre-admission mobility is coded as 1 = requires mobility aid(s) and 2 = freely mobile without aids; MOI is coded as 1 = other MOI and 2 = falls from standing height; AMT is the ten-point Abbreviated Mental Test score; and haemoglobin is measured in g/l. The NTFI can be treated in a similar way to the CFS; while the values are continuous, the same bounds range from 1 (robust) to 9 (moribund).
Verification
The standardized residuals from the multiple regression were calculated and displayed using Q-Q plots, which indicated homoscedasticity and the normal distribution of standardized results: the scatter plot assumed (approximately) rectangular shape, with scores clustered in the centre (around 0) and dispersed in a rectangular pattern (Supplementary Figure b). Multicollinearity was not present, as indicated by the collinearity statistics columns of Table V.
Bland-Altman analysis
The NTFI was validated in the validation cohort. The histogram graph showed that the difference between the NTFI and CFS was normally distributed (Gaussian), as illustrated in Supplementary Figure c). There was no significant difference between the measurements and zero (t(947) = 0.51; p = 0.612, independent-samples t-test). Figure 3 shows a Bland-Altman plot, including the 95% CIs, illustrating the high level of agreement between the NTFI and the CFS (mean difference 0.02 (SD 1.14)).
Fig. 3
Bland-Altman plot. CFS, Clinical Frailty Scale; NTFI, Nottingham Trauma Fracture Index.
Discussion
We have developed a novel score to predict frailty in major trauma patients using five variables from routinely collected data (age, mechanism of injury, haemoglobin level, AMT score, and pre-admission mobility) and shown that it can accurately replicate the CFS.
A strength of this study is that we used data collected by clinicians rather than research tools collected by staff, which means that our results are likely to be applicable in routine clinical practice. The large sample size increases reliability, and we did not need to use imputation, which reduced the risk of bias. Testing the generated model in a large different dataset from the one we used to derive it reduced the risk of over-fit. However, we could only examine routinely collected variables in this particular dataset: it is possible that there are other routinely collected data items that could improve upon our model. Another limitation is that the derivation and validation samples were collected in the same centre: it is possible that the model could perform less well in a dataset from a different hospital. This study has not tested the predictive value of the NTFI against frailty outcomes following trauma.
We believe we are the first to develop a frailty index for patients with trauma using routinely collected clinical data. Age, cognitive function, and mobility are commonly used in frailty assessment,2,11,22-29 but the haemoglobin level and mechanism of injury are novel indicators of frailty. Mechanism of injury (falling from a standing height) is likely to reflect frailty in this trauma population because robust people usually require stronger forces to produce injury: bony injuries caused by a fall from a standing height are often referred to as fragility fractures and to indicate osteoporosis, which often accompanies frailty.30,31 It is clearly acting here as a trauma-specific frailty variable, confirming a frailty role in the widely known association between ground-level falls and poor outcomes in older trauma patients.32-34 Haemoglobin level is also likely to be trauma-specific because it reflects not only generic frailty deficits, such as the degree of pre-morbid anaemia and malnutrition, but also the blood loss from injury. Haemoglobin has previously been found to be predictive of outcome in trauma, and is included in the NHFS.5 Because of these two trauma-specific items, the NTFI is a trauma-specific score and would not be expected to predict the presence of frailty in non-trauma populations.
The NTFI can be used to identify frailty in trauma patients using routinely collected clinical data, without the need to train staff in the use of a specific frailty tool (Supplementary Material 4). This could be useful in clinical services where there are large numbers of frequently changing staff, in whom training in the reliable use of frailty tools is difficult. This is an important addition to the range of assessment tools available, as assessment of frailty has been identified as an important component of managing the older trauma population35 and part of policy.36 Our new frailty index (NTFI) overlaps with the Nottingham Hip Fracture Score used for the prediction of outcomes such as survival, because they have two common items (haemoglobin level and age), which presumably are related both to frailty and survival. But the NTFI includes measures of musculoskeletal performance and independence in ADLs (pre-admission mobility), cognitive function (AMT), and the fact that an injury was sustained after low-velocity trauma (fall from standing height): these factors are more closely related to the frailty concept and hence may best guide management that depends upon the frailty state of the patient and better predict frailty outcomes. The integration of NHFS and NTFI in older trauma patients with hip fractures could assist clinicians in predicting potential outcomes and addressing frailty. Future research may investigate the interplay between these two tools.
NTFI scores could be used not only to guide patient management, but also to stratify patients in quality improvement projects. The NTFI could also be used in existing clinical and research datasets where frailty has not been measured, to explore the impact of frailty. An important implication is that NTFI would quickly enable clinicians to stratify frailty risk, thereby reducing intervention and waiting times, and facilitating the triage of severely injured frail patients. Further work could aim to automate the calculation of the NTFI score from electronic patient records so as to provide a frailty assessment in real time. Further research is also required to examine its performance in other datasets, including whether it is predictive of frailty outcomes such as the complication rate, length of stay, need for long-term care, and survival.
References
1. Alqarni AG , Gladman JRF , Obasi AA , Ollivere B . Does frailty status predict outcome in major trauma in older people? A systematic review and meta-analysis . Age Ageing . 2023 ; 52 ( 5 ): afad073 . Crossref PubMed Google Scholar
2. Fried LP , Tangen CM , Walston J , et al. Frailty in older adults: evidence for a phenotype . J Gerontol A Biol Sci Med Sci . 2001 ; 56 ( 3 ): M146 – 56 . Crossref PubMed Google Scholar
3. Rockwood K . Frailty and its definition: a worthy challenge . J Am Geriatr Soc . 2005 ; 53 ( 6 ): 1069 – 1070 . Crossref PubMed Google Scholar
4. Joseph B , Pandit V , Zangbar B , et al. Validating trauma-specific frailty index for geriatric trauma patients: a prospective analysis . J Am Coll Surg . 2014 ; 219 ( 1 ): 10 – 17 . Crossref PubMed Google Scholar
5. Maxwell MJ , Moran CG , Moppett IK . Development and validation of a preoperative scoring system to predict 30 day mortality in patients undergoing hip fracture surgery . Br J Anaesth . 2008 ; 101 ( 4 ): 511 – 517 . Crossref PubMed Google Scholar
6. Rushton PRP , Reed MR , Pratt RK . Independent validation of the Nottingham Hip Fracture Score and identification of regional variation in patient risk within England . Bone Joint J . 2015 ; 97-B ( 1 ): 100 – 103 . Crossref PubMed Google Scholar
7. Lisk R , Yeong K , Fluck D , Fry CH , Han TS . The ability of the Nottingham Hip Fracture Score to predict mobility, length of stay and mortality in hospital, and discharge destination in patients admitted with a hip fracture . Calcif Tissue Int . 2020 ; 107 ( 4 ): 319 – 326 . Crossref PubMed Google Scholar
8. Archunan MW , Subhash S , Attwood J , et al. Nottingham Hip Fracture Score: does it predict mortality in distal femoral fracture patients? Cureus . 2021 ; 13 ( 10 ): e19139 . Crossref PubMed Google Scholar
9. No authors listed . National Hip Fracture Database (NHFD) annual report 2017 . Royal College of Physicians . 2017 . https://nhfd.co.uk/files/2017ReportFiles/NHFD-AnnualReport2017.pdf ( date last accessed 19 January 2024 ). Google Scholar
10. No authors listed . NHS . 2019 . https://www.longtermplan.nhs.uk/wp-content/uploads/2019/08/nhs-long-term-plan-version-1.2.pdf ( date last accessed 19 January 2024 ). Google Scholar
11. Rockwood K , Song X , MacKnight C , et al. A global clinical measure of fitness and frailty in elderly people . CMAJ . 2005 ; 173 ( 5 ): 489 – 495 . Crossref PubMed Google Scholar
12. Hörlin E , Munir Ehrlington S , Henricson J , John RT , Wilhelms D . Inter-rater reliability of the Clinical Frailty Scale by staff members in a Swedish emergency department setting . Acad Emerg Med . 2022 ; 29 ( 12 ): 1431 – 1437 . Crossref PubMed Google Scholar
13. Ikram A , Norrish AR , Marson BA , Craxford S , Gladman JRF , Ollivere BJ . Can the Clinical Frailty Scale on admission predict 30-day survival, postoperative complications, and institutionalization in patients with fragility hip fracture?: a cohort study of 1,255 patients . Bone Joint J . 2022 ; 104-B ( 8 ): 980 – 986 . Crossref PubMed Google Scholar
14. Collins GS , Reitsma JB , Altman DG , Moons KGM . Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD Statement . BMC Med . 2015 ; 13 ( 1 ): 1 . Crossref PubMed Google Scholar
15. Harrell FE . Regression Modeling Strategies . In : Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis . New York, NY : Springer , 2001 . Crossref Google Scholar
16. Riley RD , Ensor J , Snell KIE , et al. Calculating the sample size required for developing a clinical prediction model . BMJ . 2020 ; 368 : m441 . Crossref PubMed Google Scholar
17. Gorelick MH . Bias arising from missing data in predictive models . J Clin Epidemiol . 2006 ; 59 ( 10 ): 1115 – 1123 . Crossref PubMed Google Scholar
18. Bertsimas D , Pawlowski C , Zhuo YD. . From predictive methods to missing data imputation: an optimization approach . J Mach Learn Res . 2017 ; 18 ( 1 ): 7133 - 71 . Google Scholar
19. Pepinsky TB . A note on listwise deletion versus multiple imputation . Political Analysis . 2018 ; 26 ( 4 ): 480 – 488 . Crossref Google Scholar
20. Koszalinski R , Tansakul V , Khojandi A , Li X . Missing data, data cleansing, and treatment from a primary study: implications for predictive models . Comput Inform Nurs . 2018 ; 36 ( 8 ): 367 – 371 . Crossref PubMed Google Scholar
21. Giavarina D . Understanding Bland Altman analysis . Biochem Med (Zagreb) . 2015 ; 25 ( 2 ): 141 – 151 . Crossref PubMed Google Scholar
22. Castell M-V , Sánchez M , Julián R , Queipo R , Martín S , Otero Á . Frailty prevalence and slow walking speed in persons age 65 and older: implications for primary care . BMC Fam Pract . 2013 ; 14 ( 1 ): 1 – 9 . Crossref PubMed Google Scholar
23. Thillainadesan J , Scott IA , Le Couteur DG . Frailty, a multisystem ageing syndrome . Age Ageing . 2020 ; 49 ( 5 ): 758 – 763 . Crossref PubMed Google Scholar
24. Maxwell CA , Mion LC , Mukherjee K , et al. Preinjury physical frailty and cognitive impairment among geriatric trauma patients determine postinjury functional recovery and survival . J Trauma Acute Care Surg . 2016 ; 80 ( 2 ): 195 – 203 . Crossref PubMed Google Scholar
25. Young J , Inouye SK . Delirium in older people . BMJ . 2007 ; 334 ( 7598 ): 842 – 846 . Crossref PubMed Google Scholar
26. Salluh JIF , Wang H , Schneider EB , et al. Outcome of delirium in critically ill patients: systematic review and meta-analysis . BMJ . 2015 ; 350 : h2538 . Crossref PubMed Google Scholar
27. Levkoff SE , Evans DA , Liptzin B , et al. Delirium. The occurrence and persistence of symptoms among elderly hospitalized patients . Arch Intern Med . 1992 ; 152 ( 2 ): 334 – 340 . Crossref PubMed Google Scholar
28. Cipriani G , Lucetti C , Carlesi C , Danti S , Nuti A . Depression and dementia. A review . Eur Geriatr Med . 2015 ; 6 ( 5 ): 479 – 486 . Crossref Google Scholar
29. Hodkinson HM . Evaluation of a mental test score for assessment of mental impairment in the elderly . Age Ageing . 1972 ; 1 ( 4 ): 233 – 238 . Crossref PubMed Google Scholar
30. Greco EA , Pietschmann P , Migliaccio S . Osteoporosis and sarcopenia increase frailty syndrome in the elderly . Front Endocrinol (Lausanne) . 2019 ; 10 : 255 . Crossref PubMed Google Scholar
31. Li G , Thabane L , Papaioannou A , Ioannidis G , Levine MAH , Adachi JD . An overview of osteoporosis and frailty in the elderly . BMC Musculoskelet Disord . 2017 ; 18 ( 1 ): 46 . Crossref PubMed Google Scholar
32. Kehoe A , Smith JE , Edwards A , Yates D , Lecky F . The changing face of major trauma in the UK . Emerg Med J . 2015 ; 32 ( 12 ): 911 – 915 . Crossref PubMed Google Scholar
33. Spaniolas K , Cheng JD , Gestring ML , Sangosanya A , Stassen NA , Bankey PE . Ground level falls are associated with significant mortality in elderly patients . J Trauma Acute Care Surg . 2010 ; 69 ( 4 ): 821 – 825 . Crossref PubMed Google Scholar
34. Ayoung-Chee P , McIntyre L , Ebel BE , Mack CD , McCormick W , Maier RV . Long-term outcomes of ground-level falls in the elderly . J Trauma Acute Care Surg . 2014 ; 76 ( 2 ): 498 – 503 . Crossref PubMed Google Scholar
35. No authors listed . Guideline for the care of people living with frailty undergoing elective and emergency surgery that encompasses the whole perioperative pathway . British Geriatrics Society . 2021 . https://www.bgs.org.uk/cpocfrailty ( date last accessed 5 January 2024 ). Google Scholar
36. No authors listed . Dementia, disability and frailty in later life – midlife approaches to delay or prevent onset . National Institute for Health and Clinical Excellence . 2015 . https://www.nice.org.uk/guidance/ng16/resources/dementia-disability-and-frailty-in-later-life-midlife-approaches-to-delay-or-prevent-onset-pdf-1837274790085 ( date last accessed 5 January 2024 ). Google Scholar
Author contributions
A. G. Alqarni: Data curation, Formal analysis, Methodology, Validation, Visualization, Writing – original draft.
J. Nightingale: Writing – review & editing.
A. Norrish: Writing – review & editing.
J. R. F. Gladman: Project administration, Supervision, Writing – review & editing.
B. Ollivere: Methodology, Project administration, Resources, Software, Supervision, Writing – review & editing.
Funding statement
The authors disclose receipt of the following financial or material support for the research, authorship, and/or publication of this article: the study is funded by University of Nottingham and King Saud University. The role of the funding was to cover the expenses of the study meetings.
ICMJE COI statement
This study was conducted as part of A. G. Alqarni’s PhD programme at the University of Nottingham, under the supervision of B. Ollivere and J. R. F. Gladman, with funding from the King Saud University in Saudi Arabia.
Data sharing
The datasets generated and analyzed in the current study are not publicly available due to data protection regulations. Access to data is limited to the researchers who have obtained permission for data processing. Further inquiries can be made to the corresponding author.
Ethical review statement
Institutional audit permission was obtained for this study. Approval number 21-040C.
Open access statement
This article is distributed under the terms of the Creative Commons Attributions (CC BY 4.0) licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium or format, provided the original author and source are credited.
Supplementary material
Supplementary graphs for Nottingham Trauma Frailty Index (NTFI) construction and validation, and the NTFI calculator.
Social media
Follow A. G. Alqarni on X @ag_alqarni
Follow B. Ollivere on X @benollivere
This article was primary edited by M. Hossain.
