- Physical therapists assess inpatient mobility and functional status and make discharge destination recommendations.
- The deMorton Mobility Index score and independence in toilet transfers best predicted which communitydwelling patients were not likely to go directly home.
- The AlphaFIM score and independence in walking best predicted which community-dwelling patients were likely to go directly home.
- Clinicians will need to decide whether to correctly identify patients likely to go directly home, at the risk of misclassifying people unlikely to be discharged directly home, or vice versa.
Health care utilization by older people is disproportionately high around the world. In the United States, people aged 85 years and older account for 2% of the total population but account for 9% of hospital discharges and 11% of hospital days.1 In Australia, people aged 65 years and older account for 41% of hospital admissions and almost half of all patient days (48%), despite making up only 15% of Australia's population.2 Older patients often present to the hospital with both medical and social issues that may impact on discharge destination.3 As such, discharge planning is an integral component of patient care.
Discharge planning involves early identification of a patient's likely care needs prior to hospital discharge and aims to reduce unplanned readmissions and minimize costs.4 Patients from an acute hospital may be discharged directly home. Alternatively, patients may be discharged to a subacute facility for ongoing multidisciplinary therapy or discharged to a residential facility for more permanent care.5,6 Discharge planning requires consideration of clinical, functional, cognitive, and social domains and relies on input from all members of the multidisciplinary team, including physicians, nurses, physical therapists, and other health professionals. As experts in mobility and function, physical therapists are well placed to assess mobility and functional status and make recommendations about ongoing requirements.
Discharge planning can be inefficient and discharge can be delayed for nonmedical reasons on 30% to 61% of occasions.7–9 Reasons for delays include inadequate decision making, complexity of patients, unavailability of ongoing care supports, and inadequate patient assessment.7–9 Thus, thorough patient assessment by the multidisciplinary team is critical. Such assessment may utilize tools to accurately measure patient's health and function and facilitate goal setting for discharge planning.
Previous research has found that a number of patient factors and assessment tools have been used to predict discharge across various discharge destinations.10–14 These factors and tools span a wide variety of domains, including performance in activities of daily living, cognition, mobility, premorbid living situation, and medical factors such as comorbidity and admission diagnosis. For example, tools such as the Modified Barthel, the Northwick Park Dependency Scale, and the Abbreviated Mental Test have been shown to discriminate between discharge destinations in general medicine, while tools such as the Orpington Score have been used in the population with stroke.10,15 However, the majority of the literature in this field is dated and based on data collected more than 10 years ago.11,16 Given the aging population and significant changes to the health care system during this time, an updated investigation is warranted. In addition, prior work has been conducted predominately in the United States and European countries, with a paucity of research from the Australian health care setting.
Therefore, the aim of this study was to investigate the association between different patient factors and patients' discharge destination (home or “not home”) from acute general medical wards.
Study Design and Setting
This was a prospective, single-site observational study conducted on the general medical units of The Royal Melbourne Hospital, a large metropolitan tertiary hospital in Australia from July 2016 to August 2017. Ethical approval was obtained from The Royal Melbourne Hospital and The University of Melbourne. The study was registered prospectively on the Australia and New Zealand Clinical Trials Registry (ACTRN12616000870459).
Patients were eligible if they were referred to physical therapy and admitted to a general medical unit. Admissions to these units are emergency and nonelective. Patients were excluded if they were for palliation or previously recruited. Patients were also excluded if they were transferred from another unit or if they were initially seen on a weekend or public holiday, as staff trained in assessment procedures were not available on weekends. Secondary to staffing resources, a maximum of 5 patients could be recruited daily. Where more than 5 patients were eligible, names were randomly selected from an opaque hat. This randomization strategy was chosen as it could be completed within the constraints of a brief daily handover meeting. Verbal consent was obtained from all recruited patients or their next of kin.
Variables and Measurement
Data were collected within 72 hours of initial physical therapy assessment. Interpreters were used when the patient's preferred language was not English. A range of assessment tools were selected to assess function, cognition, mobility, premorbid living situation, and medical domains. Tools were selected on the basis of their psychometric properties and suitability for use in an acute clinical environment with a culturally diverse population. Data included premorbid function assessed using the Blaylock Risk Assessment Screening Score.17 This 10-item tool measures age, living situation, function, mobility, cognition, behavior, sensory deficits, active medical problems, number of prescribed medications, and previous admissions to the emergency department.17 It is scored from 0 to 40, with scores of 20 or more suggesting that patients are high risk for not returning directly home.17 Comorbidity was measured using the Charlson Comorbidity Index. This is a valid tool in hospitalized general medical patients that has been used to predict mortality on the basis of comorbidity.18
Mobility and balance were assessed using the de Morton Mobility Index (DEMMI), which takes around 15 minutes to complete.19 This 15-item Rasch-analyzed tool provides a score out of 100. This assessment tool was selected as it is a reliable and valid measure of mobility in the general medical population and mobility is an important consideration when discharge planning.13,19 Previous research suggests that patients needing inpatient rehabilitation or residential care at discharge have significantly lower DEMMI scores at acute hospital discharge than those discharged home.13,20 It has a strong correlation with the Barthel Index (Spearman ρ = 0.76; 95% confidence interval (CI), 0.65-0.84) and a minimally clinically important difference of 10 points.13
Function was assessed using the Alpha Functional Independence Measure (AlphaFIM). The AlphaFIM is an abbreviated version of the Functional Independence Measure (FIM), specifically for use in an acute hospital.21 It evaluates 6 tasks (4 motor and 2 cognitive) on a 7-point scale with higher scores indicating increased independence.21 It takes approximately 5 minutes to complete and has good correlation with its subacute counterpart (the FIM), thereby allowing easy comparison of function between the acute and subacute settings.21 It has been validated in acute medical population and has excellent interrater reliability (interclass correlation coefficient, 0.92).21
Cognition was assessed using the Rowland Universal Dementia Assessment Scale (RUDAS).22 This 6-item tool assesses multiple cognitive domains including memory, praxis, language judgment, drawing, and body orientation to give a score out of 30,22 and takes around 10 minutes to complete.23 Scores less than 23 have been shown to detect dementia with both high sensitivity and specificity (89% and 98%, respectively).22,24 It has also been shown to have high interrater (interclass correlation coefficient, 0.99) and test-retest (interclass correlation coefficient, 0.98) reliability.22 The RUDAS is at least as accurate as the Mini-Mental State Examination in diagnosing dementia (Pearson correlation coefficient: 0.83, P < .05)22,23 and takes a similar time to perform. An advantage of the RUDAS is that it is not affected by education level and can also be administered in other languages using an interpreter.24
Demographic data collected included discharge destination from the acute setting, length of stay, and 28-day readmissions. Discharge destinations were dichotomized into “home” and “not home,” which included discharge to subacute care, residential care, discharge to other destinations (such as a psychiatric facility), and death. Subacute care was defined as a temporary facility that provides multidisciplinary intervention after an acute illness (eg, rehabilitation), whereas residential care was defined as a more permanent facility that provides varying assistance to manage activities of daily living (eg, nursing homes).5,6
Data collection occurred for 12 months to allow for seasonal variation. It was estimated that recruitment over 12 months would enable 480 patients to be included in this pragmatic sample. A minimum sample size of 120 was required to fit a logistic regression model with 4 covariates, assuming that one-third of the sample was in the “not home” destination. This pragmatic sample easily exceeds this minimum.
Data were analyzed using IBM SPSS Statistics (Version 23). Baseline descriptive data are presented as frequency and percentage (for categorical data) and as median and interquartile range for numerical data (as most variables were not normally distributed). Independent t tests were used to determine the difference in means from the assessment tools between those patients discharged “home” and “not home.” Alpha was set at .05.
Preliminary logistic regression was completed for each variable, including the items of a specific tool, as well as the entire score of the tool itself. Tools were assessed as continuous data. Items with more than 2 categories were collapsed to dichotomous variables (eg, “independent with or without an aid” vs “not independent,” where patients required assistance or supervision from another person), as independent performance of tasks is important for independent living. The data set was randomly partitioned into training (50%, n = 208), validation (25%, n = 105), and test (25%, n = 104) data sets. Forward stepwise selection was used to develop candidate models, which were fit to the training data set and assessed on the basis of their performance on the validation data set. Assessing the model performance on a different subset of the data prevents bias caused by overfitting, that is, models that appear to perform well because they explain chance variation in the training subset. Variables were added to the candidate model in order of their Cox-Snell r2 from the preliminary logistic regression and kept if they were statistically significant (P < .05). Where more than 1 candidate model was found, final models were selected by creating receiver operating characteristic curves and comparing areas under the curve. The final set of test data, which were not used in any earlier stage of the model selection process, were then used to determine an unbiased estimate of sensitivity, specificity, negative predictive value, and positive predictive value. Sensitivity describes the model's ability to correctly determine discharge “not home” (true positive rate) whereas specificity refers to the model's ability to correctly determine discharge “home” (true negative rate).25 The positive predictive value estimates the chance of a patient who is predicted not to go home actually not going home.25 Conversely, the negative predictive value estimates the chance of a patient who is predicted to go home actually going home.25 The areas under the curve provides an overall measure of the model's ability to separate patients who are discharged home from those who are not.25
Over the 12-month study period, 1528 patients were admitted to the general medicine unit, 1344 of whom were referred to physical therapy. The Figure outlines the recruitment of patients and reasons for exclusion. A total of 483 patients consented to the study. Fifty participants had significant missing data and were therefore excluded. Patients admitted from residential care were also excluded, as 10 of the 16 patients (62%) admitted from residential care were discharged directly back to their residential care facility. Therefore, data analysis was completed on 417 patients. Data collection commenced a median of 2 days (interquartile range: 1-3) after hospital admission.
Patient characteristics are presented in Table 1. Two hundred forty-five (59%) patients were discharged directly home from hospital. One hundred seventy-two (41%) patients were discharged “not home,” of whom 140 (34%) patients were discharged to subacute care, 21 (5%) patients did not survive their hospital admission, and 11 (3%) patients had other destinations (such as transfer to a private hospital or psychiatric facility). Diagnostic-related groups (derived postdischarge to classify main reason for admission) were similar between the study sample and all patients admitted to general medicine during the study period (Table 1). Main system categories were selected rather than specific diagnostic-related groups due to the large number present (almost 150). These were not considered for exploratory logistic regression as there were approximately 20 broad categories making analysis impractical. Patients discharged home had higher functional, mobility, and cognitive scores (see Table 1) as measured by the AlphaFIM (mean difference: 8.1, 95% CI: 6.7-9.6, P < .001), DEMMI (mean difference: 16.2, 95% CI: 13.3-19.2, P < .001), and RUDAS (mean difference: 3.2, 95% CI: 2.1-4.2, P < .001), respectively. Patients in the discharged “home” group had a shorter length of stay than those in the “not home” group (mean difference: 5.5 days, 95% CI: 3.1-7.8, P < .001) and a lower incidence of falls in the past 6 months (46.5% vs 59.3%, P = .01). Charlson Comorbidity Index did not differ between the groups. Readmission rates were slightly higher in those patients who went home than in those who did not go home (20.3% vs 16.3%). However, this was similar to the readmission rates for all patients admitted during the study period (19.7%) and similar to previously published readmission rates for this patient population (17%-18%).26,27
Table 1. -
Patient Demographics Presented as Median (Interquartile Range) or Count (Percentage)a
||All General Medical Patients (n = 1528)
||All Patients Included in Study (n = 417)
||Patients Discharged Home (n = 245)
||Patients Not Discharged Home (n = 172)
|English as a first language
|Number of falls
|DRG (disorders of):
|1 or 2 services
|3 or more services
||n = 346
||n = 214
||n = 132
||n = 1528
||n = 386
||n = 245
||n = 141
Abbreviations: AlphaFIM, Alpha Functional Independence Measure; BRASS, Blaylock Risk Assessment Screening Score; CCI, Charlson Comorbidity Index; DEMMI, de Morton Mobility Index; DRG, diagnostic-related group; ED, emergency department; ICU, intensive care unit; LOS, acute hospital length of stay; RUDAS, Rowland Universal Dementia Assessment Scale.
aPatients discharged “not home” include patients who went to subacute, died, or had other discharge destinations.
bED presentations: at least 1 previous presentation to the emergency department in the past 3 months; falls: at least a fall in the past 6 months; number of falls: median number of falls in the past 6 months; total services: median number of home services; readmission: at least 1 readmission in the 28 days after discharge.
Exploratory logistic regression was completed on 70 individual factors (5 tools; 65 subcomponents of these tools or other demographic factors). Upon completing this preliminary exploratory analysis, 54 individual factors were found to be significantly associated with discharge destination (see Supplemental Digital Content Appendix 1, available at: http://links.lww.com/JGPT/A43). Despite having an association with discharge, the RUDAS was removed from model development as data were missing for a large number of patients (n = 71, 17%). The DEMMI and AlphaFIM tests had the strongest correlations with discharge destination and, therefore, these 2 tests were chosen for the basis of candidate models. Factors were added into a candidate model in order of their Cox-Snell r2 value. Candidate models were assessed on the basis of their performance on the validation data set. As identified in Table 2, 8 candidate models were found, 5 based on the DEMMI and 3 based on the AlphaFIM. The models with the best trade-off between sensitivity and specificity were fitted to the test data and are described in Table 2. The DEMMI and a dichotomous version of the AlphaFIM's toilet transfers (ie, independence with getting on and off the toilet) made up the first model. The second model included the AlphaFIM and a dichotomous version of the DEMMI's walking independence (ie, independence in walking, with or without an aid). The model equation can be found in Supplemental Digital Content Appendix 2, available at: http://links.lww.com/JGPT/A44.
Table 2. -
Table Showing Specificity, Sensitivity, Negative Predictive Value, Positive Predictive Value, and Area Under the Curve With 95% Confidence Intervals for the 8 Candidate Models (Fitted to Validation Data) and 2 Final Models That Best Predicted Discharge Destination (Fitted to Test Data)a
||AUC (95% CI)
|Candidate models (validation data)
|DEMMI and bowel management
|DEMMI and toilet transfer
|DEMMI, toilet transfer and appropriate behavior
|DEMMI, toilet transfer, appropriate behavior, and medication
|AlphaFIM and walking independence
|AlphaFIM, walking independence, and BRASS
|Final 2 models (test data)
|DEMMI and toilet transfer (DEMMI Plus)
|AlphaFIM and walking independence (AlphaFIM Plus)
Abbreviations: AlphaFIM, Alpha Functional Independence Measure; AUC, area under the curve; BRASS, Blaylock Risk Assessment Screening Score; CI, confidence interval; DEMMI, de Morton Mobility Index; NPV, negative predictive value; PPV, positive predictive value; Sens, sensitivity; Spec, specificity.
aBowel management: independence (with or without a device) with managing bowels (as measured on the AlphaFIM); toilet transfers: independence (with or without a device) with toilet transfers (as measured on AlphaFIM); appropriate behavior: appropriate behavior prior to hospital admission (as measured on BRASS); medication: independent (with or without a device) in medication management (as measured on the BRASS); walking independence: independence with walking as measured on the DEMMI.
Predicting discharge destination for patients following an acute general medical admission is an important but challenging task. These patients are complex; in addition to their medical problems, they often present with social issues that can complicate discharge planning. This observational study investigated factors associated with discharge destination (“home” or “not home”). Fifty-four factors, assessed a median of 48 hours from admission, were associated with discharge destination. In keeping with previous research, patients who went home generally had higher functional, cognitive, and mobility scores than those patients who did not go home.13,21,28 Two models were identified that were best associated with discharge destination: the “DEMMI Plus” (DEMMI and toilet transfers) and the “AlphaFIM Plus” (AlphaFIM and walking independence). It is worth noting that walking features within the motor component of the AlphaFIM; retention of walking independence as an individual item within the AlphaFIM Plus model indicates that walking independence is important in determining discharge destination. Further research is needed to establish generalizability of these models to other health services and determine their ability to predict probable discharge destination following hospital admission in this complex patient group.
Although 2 models have been proposed, the practicality of completing assessments in the clinical setting should also be considered. Both the AlphaFIM and the DEMMI were feasible to complete early in the admission, take a similar time to administer, and required minimal equipment.21,29 The DEMMI is freely available via a web site and requires minimal training, whereas the AlphaFIM requires formal training (at a cost to the hospital).30 The higher sensitivity of the “DEMMI Plus” indicates that it may be better at identifying patients unable to be discharged home, whereas the higher specificity of the “AlphaFIM Plus” indicates that it may be better at determining patients able to go directly home. Given the similarities in statistical measures of validity, clinicians should consider other properties of the 2 models when selecting which to use in clinical practice. This may include whether it is more clinically relevant to correctly identify patients who are likely to go directly home, at the risk of misclassifying people who are unlikely to be discharged directly home, or vice versa. Other considerations that may impact model selection include the available resources (including training, time, and costs) and clinical expertise. Creation of decision trees may also assist clinicians in the practical classification of likely outcomes. Further research is needed to explore these statistical methods.
This study utilized tools that have been validated in the general medical population and have been validated for use in other languages or are largely free from cultural sensitivities.13,21,22 The study also included patients with a cognitive impairment and those who do not speak English, which increases the generalizability of results. Only patients referred to physical therapy were included in this study. This represents the majority of patients admitted during this time period (88%). However, referral to physical therapy may depend on the threshold of the referrer, physical therapy resources, and health service culture. Patients not referred to physical therapy are less likely to have complex physical needs or may require palliation; this may be reflected in the shorter hospital length of stay for the “all general medical patients” group (5 days vs 7 days, Table 1). Diagnostic-related groups, which reflect the primary admission diagnosis, were not included in the exploratory logistic regression as there were more than 20 categories, making analysis impractical. Frailty was not assessed. There is no consensus as to how frailty should be examined in this population; however, our selection of measures included many domains considered important in frailty identification.31,32 It is, therefore, unlikely that the addition of a frailty tool will change the overall outcome.
While we differentiated between patients discharged “home” and “not home,” we were unable to further differentiate between discharge destinations for the “not home” cohort. When patients are identified as unlikely to have a direct discharge home, alternate destinations should be considered. This study also considers only discharge directly from the acute setting. It did not consider long-term outcomes such as discharge destination from subacute or readmission rates. While long-term outcomes are important, this was beyond the scope of this study. Finally, the study was conducted only at 1 site. Further research is required to determine whether these findings are applicable to the wider general medical population.
This study found a variety of factors spanning mobility, functional, and cognitive domains that were associated with discharge destination following an acute general medical hospital admission. From these factors, 2 models were created, the “DEMMI Plus” and the “AlphaFIM Plus,” which were able to discriminate between patients being discharged home and “not home.” These models may have the potential to quickly determine discharge destination, but further research is required to validate these tools before use in the clinical setting.
The authors acknowledge the participants and the general medicine physical therapy team at The Royal Melbourne Hospital for their contribution and time.
1. Levant S, Chari K, DeFrances C. Hospitalizations for patients aged 85 and over in the United States, 2000-2010. https://www.cdc.gov/nchs/data/databriefs/db182.pdf
. 2015 Accessed November 5, 2018.
2. Australian Institute of Health and Welfare. Australia's hospitals 2015-2016 at a glance. https://www.aihw.gov.au/getmedia/d4e53b39-4718-4c81-ba90-b412236961c5/21032.pdf.aspx?inline=true
. 2017 Accessed November 5, 2018.
3. Jenkins P, Thompson C, MacDonald A. What does the future hold for general medicine
? Med J Aust. 2011;195(1):49–50.
4. Shepperd S, Lannin NA, Clemson LM, McCluskey A, Cameron ID, Barras SL. Discharge planning from hospital to home. Cochrane Database Syst Rev. 2013(1):CD000313.
5. Cameron ID, Kurrle SE. 1: Rehabilitation and older people. Med J Aust. 2002;177(7):387–391.
6. Sanford AM, Orrell M, Tolson D, et al. An international definition for “nursing home.” J Am Med Dir Assoc. 2015;16(3):181–184.
7. Selker HP, Beshansky JR, Pauker SG, Kassirer JP. The epidemiology of delays in a teaching hospital: the development and use of a tool that detects unnecessary hospital days. Med Care. 1989;27(2):112–129.
8. Tan MZ, Mackett A, Ratcliff C, Niruban A. 38 Analysis of timelines from “Medically Fit” to discharge in older patients in the Norfolk and Norwich University Hospital. Age Ageing. 2014;43(suppl 2):ii9.
9. Moore G, Hartley P, Romero-Ortuno R. Health and social factors associated with a delayed discharge amongst inpatients in acute geriatric wards: a retrospective observational study. Geriatr Gerontol Int. 2018;18(4):530–537.
10. Slade A, Fear J, Tennant A. Predicting outcome for older people in a hospital setting: which scales are appropriate? Int J Ther Rehabil. 2004;11(1):25–30.
11. Wachtel TJ, Fulton JP, Goldfarb J. Early prediction of discharge disposition after hospitalization. Gerontologist. 1987;27(1):98–103.
12. Lakhan P, Jones M, Wilson A, Gray LC. The Higher Care At Discharge Index (HCDI): identifying older patients at risk of requiring a higher level of care at discharge. Arch Gerontol Geriatr. 2013;57(2):184–191.
13. de Morton NA, Davidson M, Keating JL. Validity, responsiveness and the minimal clinically important difference for the de Morton Mobility Index (DEMMI) in an older acute medical population. BMC Geriatr. 2010;10(1):72.
14. Zureik M, Lang T, Trouillet JL, et al. Returning home after acute hospitalization in two French teaching hospitals: predictive value of patients' and relatives' wishes. Age Ageing. 1995;24(3):227–234.
15. Pittock S, Meldrum D, Dhuill C, Hardiman O, Moroney J. The Orpington Prognostic Scale within the first 48 hours of admission as a predictor of outcome in ischemic stroke. J Stroke Cerebrovasc Dis. 2003;12(4):175–181.
16. Simonet M, Kossovsky M, Chopard P, Sigaud P, Perneger T, Gaspoz J. A predictive score to identify hospitalized patients' risk of discharge to a post-acute care facility. BMC Health Serv Res. 2008;8(1):154. doi:10.1186/1472-6963-8-154
17. Mistiaen P, Duijnhouwer E, Prins-Hoekstra A, Ros W, Blaylock A. Predictive validity of the BRASS index in screening patients with post-discharge problems. J Adv Nurs. 1999;30(5):1050–1056.
18. Frenkel WJ, Jongerius EJ, Mandjes-van Uitert MJ, van Munster BC, de Rooij SE. Validation of the Charlson Comorbidity Index in acutely hospitalized elderly adults: a prospective cohort study. J Am Geriatr Soc. 2014;62(2):342–346.
19. de Morton NA, Davidson M, Keating JL. Reliability of the de Morton Mobility Index (DEMMI) in an older acute medical population. Physiother Res Int. 2011;16(3):159–169.
20. Trøstrup J, Andersen H, Kam CA, Magnusson SP, Beyer N. Assessment of mobility in older people hospitalized for medical illness using the de Morton Mobility Index and Cumulated Ambulation Score-Validity and Minimal Clinical Important Difference. J Geriatr Phys Ther. 2019;42(3):153–160.
21. Hinkle JL, McClaran J, Davies J, Ng D. Reliability and validity of the adult Alpha Functional Independence Measure instrument in England. J Neurosci Nurs. 2010;42(1):12–18.
22. Storey JE, Rowland JT, Conforti DA, Dickson HG. The Rowland Universal Dementia Assessment Scale (RUDAS): a multicultural cognitive assessment scale. Int Psychogeriatr. 2004;16(1):13–31.
23. Pang J, Yu H, Pearson K, Lynch P, Fong C. Comparison of the MMSE and RUDAS cognitive screening tools in an elderly inpatient population in everyday clinical use. Intern Med J. 2009;39(6):411–414.
24. Cheung G, Clugston A, Croucher M, Malone D, Mau E, Sims A, Gee S. Performance of three cognitive screening tools in a sample of older New Zealanders. Int Psychogeriatr. 2015;27(6):981–989.
25. Portney L, Watkins M. Foundations of Clinical Research: Applications to Practice. 3rd ed. Upper Saddle River, NJ: Prentice Hall Health; 2009.
26. Allaudeen N, Vidyarthi A, Maselli J, Auerbach A. Redefining readmission risk factors for general medicine
patients. J Hosp Med. 2011;6(2):54–60.
27. Hasan O, Meltzer DO, Shaykevich SA, et al. Hospital readmission in general medicine
patients: a prediction model. J Gen Intern Med. 2010;25(3):211–219.
28. Gambier N, Simoneau G, Bihry N, et al. Efficacy of early clinical evaluation in predicting direct home discharge of elderly patients after hospitalization in internal medicine. South Med J. 2012;105(2):63–67.
29. de Morton NA, Davidson M, Keating JL. The de Morton Mobility Index (DEMMI): an essential health index for an ageing world. Health Qual Life Outcomes. 2008;6(1):63.
30. Quinsey K, Findlay C, Willmott L. FIM Information and Procedures Manual. https://ro.uow.edu.au/cgi/viewcontent.cgi?article=1003&context=aroc
. 2005 Accessed November 5, 2018.
31. Rodríguez-Mañas L, Féart C, Mann G, et al. Searching for an operational definition of frailty: a Delphi method based consensus statement. The Frailty Operative Definition Consensus Conference Project. J Gerontol A Biol Sci Med Sci.2013;68(1):62–67.
32. Lee L, Patel T, Hillier LM, Maulkhan N, Slonim K, Costa A. Identifying frailty in primary care: a systematic review. Geriatr Gerontol Int. 2017;17(10):1358–1377.