Observational evidence
Retrospective studies
|
Clegg 2016,2 UK |
Design: 5-year retrospective cohort (2008–2015).
Aim: Development and validation of an electronic frailty index (eFI) using primary care electronic health record data from the THIN database. | Individuals aged 65–95 years registered to a Research One or THIN database practice.
Development cohort (n = 207 814)
Internal validation cohort (n = 207 720)
External validation cohort (n = 516 007) | An FI was created using the cumulative deficit model in a randomly split sample of the Research One database. |
External validation cohort
Hazard ratios (with 95% CI) for risk of unplanned hospital admissions for each degree of frailty:
Year 1 Mild 2.03 (1.96 to 2.10) Moderate 3.50 (3.38 to 3.63) Severe 5.58 (5.34 to 5.84)
Year 3 Mild 1.89 (1.85 to 1.93) Moderate 3.03 (2.96 to 3.11) Severe 4.66 (4.51 to 4.80) |
Donate-Martinez 2014,12 Spain |
Design: 12-month retrospective (2008–2009).
Aim: To determine the viability of the implementation of two screening tools — namely, the Probability of Repeated Admission (Pra) and the Community Assessment Risk Screen (CARS) — to detect patients at risk of hospital readmission. | Community-dwelling patients in Spain aged ≥65 years. (n = 500) | Data related to the variables that comprise both Pra and CARS were collected with a reference date of December 2008. Pra was categorised as high, medium, and low risk of admission. CARs was categorised as high and low risk. |
Admissions, mean (SD):
Pra
P≤0.001 High 0.47 (0.86) Medium 0.25 (0.61) Low 0.12 (0.45)
CARS
P≤0.001 High 0.36 (0.76) Low 0.11 (0.40)
Length of stay (days), mean (SD):
Pra
P≤0.001 High 2.29 (7.72) Medium 0.98 (3.00) Low 0.43 (2.08)
CARS
P ≤0.001 High 1.76 (5.28) Low 0.44 (2.01)
ROC curve analysis, AUC: Pra: 0.67 CARS: 0.69 |
Cross-sectional studies
|
Dent 2016,17 Australia |
Design: Secondary cross-sectional.
Aim: To investigate specific health service provision among frail older people in the rural community of Port Lincoln, south Australia. | Participants aged ≥65 years who completed the LINKIN health study health census September–November 2010. (n = 1796) | An FI of cumulative deficits was used to classify frailty. Frailty guidelines were used to construct an FI of 23 variables falling into three categories: comorbidities, functional measures, and quality of life. Frailty was dichotomised into frail and non-frail. Participants with ≥6 accumulated deficits were considered to be frail. Participants with one or more missing FI variables were excluded from final dataset (n = 1501 for final dataset).
| Hospital admission as an inpatient over previous 12 months, by frailty category (n = 1490).
n (%):
Non-frail 73 (5)
Frail 55 (15) OR = 2.39 (95% CI = 1.74 to 3.29)
P<0.001
|
Ng 2014,19 Singapore |
Design: Cross-sectional with 2-year validation follow-up.
Aim: Development of a simple frailty risk index (FRI), and evaluation for use in primary care on an external validation cohort of community-living older persons. | Adults aged ≥55 years in the Singapore Longitudinal Ageing Studies.
Development cohort
n = 1685
Validation cohort
n = 2478 | The FRI was developed by identifying 13 salient and independent multisystem risk factors for the 5-criteria frailty phenotype used in the Cardiovascular Health Study (CHS): weakness, slowness, low physical activity, weight loss, and exhaustion. A risk score was assigned to each risk factor and the FRI was derived from summating risk scores associated with each risk factor. |
Validation cohort Association of frailty defined by the FRI (as continuous variable) with self-reported hospitalisations:
OR = 1.14 (95% CI = 1.05 to 1.24),
P = 0.002 ROC curve analysis, AUC: 0.63 |
Rochat 2010,18 Australia |
Design: Cross-sectional.
Aim: To describe the relationship between frailty and use of several health and community services in community-dwelling older men. | Men aged ≥70 years in the Concord Health and Ageing in Men Project. | Frailty was assessed using a modified version of the CHS criteria: Weight loss or shrinking, weakness, exhaustion, slowness, and low activity. Robust was categorised as meeting 0 criteria; pre-frail as ≤2 criteria; and frail as ≥3 criteria. | Participants admitted for ≥1 night in hospital during previous 12 months, n (%):
Robust 152 (18.2), AOR 1.00 (reference)
Pre-frail 174 (25.7), AOR 1.34 (95% CI = 1.03 to 1.74)
Frail 81 (51.6), AOR 3.29 (95% CI = 2.18 to 4.96) |
McGee 2008,11 Ireland |
Design: Cross-sectional.
Aim: To assess if those categorised as vulnerable by the Vulnerable Elders Survey (VES) were likely to use health services more frequently than others. | Randomly selected community-dwelling individuals aged ≥65 years living at a private residential addresses and able to participate in a research interview. (n = 2033) | The VES is a 13-item questionnaire developed from studying >6000 community-dwelling Medicare beneficiaries aged ≥65 years. | Health service use in previous 12 months (n = 2033), by VES score: High VES score (32.1% of sample) versus low VES score (67.9% of sample)
Emergency department visits: 17% versus 8% P<0.001
Scheduled hospital inpatient stay: 21% versus 12%, P<0.001 |
Longitudinal studies
|
van Kempen 2015,13 Netherlands | Design: 1-year longitudinal cohort (2010–2011).
Aim: To determine the predictive value of EASY-Care Two Step Older Persons Screening (EASY-Care TOS) for negative health outcomes within 12 months from assessment. | Patients aged ≥70 years from participating GP practices. (n = 520) | Participants were assessed with the complete EASY-Care TOS procedure. All subsequent assessment steps were finished, irrespective of the outcome (usually a two-step process), during routine primary care visits. Patients were assigned as frail or not-frail. | Hospital admission during previous 12 months for participants classified as frail (n = 195) or not frail (n = 325), n (%):
Frail 39 (22.0)
Not frail 41 (12.9) Absolute difference = 9.1 (95% CI = 2.0 to 16.2), P = 0.01. |
Sha 2005,20 US |
Design: Cross-sectional.
Aim: To describe the patterns of physical symptoms in older adults and to examine the validity of symptoms in predicting hospitalisation and mortality. | Individuals aged ≥60 years completing a screening for self-reported symptoms during a routine primary care visit. (n = 3498) | Self-reported symptoms were collected using an abbreviated primary healthcare evaluation of mental disorders screening instrument (PRIME-MD). | Hospitalisations in the year following screening according to medical records by number of reported symptoms, n (%):
0–2 symptoms 171 (16.2) OR = 1.0 (reference)
3–4 symptoms 191 (20.9) OR = 1.2 (95% CI = 0.9 to 1.5)
5–7 symptoms 218 (22.3) OR = 1.2 (95% CI = 0.9 to 1.5)
8–12 symptoms 154 (27.9) OR = 1.4 (95% CI = 1.0 to1.9)a
|
Coelho 2014,16 Portugal |
Design: 10-month longitudinal.
Aim: To compare how different frailty measures predict short-term adverse outcomes, namely Frailty Phenotype (FP), Groningen Frailty Indicator (GFI), and Tilburg Frailty Indicator (TFI). | Individuals aged ≥60 years based in the community (n = 252) | Part A of TFI was used to assess life-course determinants of frailty and comorbidity, while FP, GFI, and part B of TFI were used to measure frailty. | Hospitalisations in the previous 12 months χ2:
FP P = 0.29b
Non-frail, n (%) Yes: 6 (54.5) No: 61 (72.6)
Frail, n (%) Yes: 5 (45.5) No: 23 (27.4)
GFI P = 0.08
Non-frail, n (%) Yes: 3 (27.3) No: 46 (54.8)
Frail, n (%) Yes: 8 (72.7) No: 38 (45.2)
TFI P = 0.09
Non-frail, n (%) Yes: 3 (27.3) No: 46 (54.8)
Frail, n (%) Yes: 8 (72.7) No: 38 (45.2) |
RCT and nRCT evidence
|
|
Study design and aims
|
Eligibility criteria or population
|
Frailty assessment or group assignment
|
Trial outcome
(secondary care use)
|
Ruikes 2016,14 Netherlands |
Design: Two-arm, 12-month nRCT (September 2011–September 2012).
Aim: To evaluate the effectiveness of a GP–led extensive, multicomponent programme for the prevention of functional decline.
Risk of bias: High.
| GP practices with sufficient numbers of patients aged ≥70 years and adequate facilities to enable programme implementation. Exclusion criteria: Admission to a residential or nursing home and/or critical or terminal illness (n = 536). | Six intervention practices were informed about the programme and six control practices were not (usual care). After Easy-Care TOS assessment, GP and practice nurse or research assistant made a final decision on the presence of frailty. Intervention: The Care Well primary care programme, consisting of four key elements: multidisciplinary team meetings, proactive care planning, case management, and medication review. | Data collected at baseline and 12 months through a home visit by either a trained nurse (in the intervention arm) or a research assistant (in the control arm). Hospital admissions during follow-up, n (%):
Intervention: 52 (18.1)
Control:
57 (22.9) OR = 0.74 (95% CI = 0.48 to 1.14),
P = 0.17. |
Imhof 2012,15 Switzerland |
Design: RCT.
Aim: To evaluate the effects of a 9-month advanced practice nurse (APN) in-home health consultation programme (HCP) on quality of life and health.
Risk of bias: Low.
| Inclusion criteria: German-speaking, community-dwelling individuals aged ≥80 years, cognitively able to understand study and consent. Exclusion criteria: End of life, with a major psychiatric diagnosis, or severe cognitive impairment. Intervention: n = 231 | Participants were randomly assigned to the intervention or control group following two baseline assessment visits with an APN, which included a standardised comprehensive geriatric assessment (CGA). All participants received health care as usual. Intervention: In addition to usual care, a complementary 9-month in-home HCP was delivered by one of four APNs. The HCP comprises a CGA and consultations that identify and consider the health problems and concerns of the participants. The intervention included four home visits (mean length 46 ± 6 minutes) after 4, 12, 24, and 36 weeks, and three telephone calls (mean length 17 ± 4 minutes) after 8, 18, and 30 weeks. Total intervention time per participant averaged 4 hours. | Outcomes were collected at 3, 6 and 9 months. With regard to hospitalisations, multilevel analysis showed no evidence of effect for intervention (P = 0.86), time course (P = 0.33), or interaction (P = 0.24). However, hospitalisations (number of 3-month periods with a planned or unplanned hospital admission or emergency department visit) were lower in the intervention group, n (%):
Intervention versus control 47 (23) versus 68 (33) Relative risk = 0.70 Numbers needed to treat = 10.0, P = 0.03. |