Overview

Explain importance of identifying frailty

​

Introduce the Hospital Frailty Risk Score (HFRS)

• Lancet 2018
• Lancet Health Longevity 2021

​

Apply HFRS to all adults admitted to hospital

​

Show preliminary results applied to primary care data in Catalunya

​

Frailty – clinical relevance

Frailty identification – micro level

Frailty identification – why?

• Micro level

Frailty identification – meso level

CGA allows a care plan to be generated that can modify trajectories

• At 6 months, NNT of:
• 17 (1 unnecessary death or deterioration);
• 20 (1 institutionalisation)
• NNT 25 at 12 months

Frailty identification – macro level

AFS Counting Activity - SDEC Collaboration Platform - FutureNHS Collaboration Platform

https://www.youtube.com/watch?v=sP846CSZOV8

Frailty registries?

Frailty registry

Primary care

eFI(2)

Clinical validation

CFS

Opportunistic

CFS scores

Secondary care

HFRS

Holistic assessment

Examine or develop Urgent Care Plan

Determine treatment paradigm – palliative or curative?

Address admission or readmission risk

What matters to me (PROM)? Shared decision making

Plan

Emerging ideas - PROMs

https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(18)30668-8/fulltext

Hospital Frailty Risk Score (HFRS) – Lancet 2018

• weighted set of 109 3-character ICD-10 codes recorded during the current admission and any previous emergency admissions occurring in the prior two years

​

• values range from 0 (no frailty risk) to 173·2

​

• Patients 75+ categorised into three frailty risk groups:
• low (HFRS <5),
• intermediate (5≥ HFRS <15)
• high (HFRS ≥15)

National validation cohort

In the national validation cohort (n=1 013 590)

• Low risk - 42·4%
• Intermediate risk - 37·6%
• High risk - 20·0%

​

Compared with those low HFRS, patients with high HFRS had increased odds of:

​

• Long hospital length of stay (OR 6·03, 95% CI 5·92–6·10)
• 30-day mortality (OR 1·71, 95% CI 1·68–1·75)

​

Econometric Model

Outcomes – Long LoS, in-hospital mortality

Patient and pathway characteristics, temporal effects

Hospital fixed-effects

Intermediate risk DV

High risk DV

1183 citations as of 8/1/25

Inconsistent applications

1. weighted set of 109 3-character ICD-10 codes recorded during the current admission and any previous emergency admissions occurring in the prior two years

​

• Inconsistencies in application
• Any previous admission
• patients might be censored by when data window opens

Internal inconsistency

​

• Prior two years
• some studies only current adm … one study prior 15 years

External inconsistency

https://www.thelancet.com/journals/lanhl/article/PIIS2666-7568(21)00004-0/fulltext

Trade-off between …

1. Value of information: more information increases likelihood that the HFRS captures the true effect of frailty risk on patient outcomes
2. Costs of obtaining information: obtaining historical data might incur costs which might not be worth incurring if the extra data do not offer substantial additional information

​

When we have a stable estimate, we have enough information

​

How much diagnostic information is enough?

As we account for more ICD10 codes …

​

1. More people will be classified into the high risk category

​

• There may be a change in size of effect of frailty risk on the outcomes

​

• Those newly classified as high risk will be marginally less frail than those originally classified

• Hence, effect size of high frailty risk reduced

​

How much diagnostic information is enough?

HFR Score

n

Low

Int

High

Adding ICD10 information increases the HFRS

HFRS(a) – use diagnostic information from current admission only

HFRS(a+1,1) – add info from previous admission in past year

HFRS(a+2,1) add info from previous two admissions in past year

…

HFRS(a+1,2) – add info from previous admission in past two years

…

HFRS(a+n,t)

Various constructions of HFRS

Data

• Setting: Yorkshire and Humble region

​

• 4 linked datasets

NHS 111 & 999 calls

Ambulance

Emergency Department

Hospital admissions

​

• Population 75+

​

• 282 091 patients had 675 155 emergency hospital admissions between April 1, 2013, and March 31, 2017

​

​

​

​

​

% in each HFRS category

Econometric Model

Outcomes – Long LoS, in-hospital mortality

Patient and pathway characteristics, temporal effects

Hospital fixed-effects

Intermediate risk DV

High risk DV

High risk group: impact of additional diagnostic information

High frailty risk based on ICD in current adm only

How much diagnostic information is enough?

Key message

​

Previously: ICD-10 codes recorded during the current admission and any previous emergency admissions occurring in the prior two years

​

Now: ICD-10 codes recorded during the current admission and previous two emergency admissions occurring in the prior two years

​

LoS more than 10 days

Frailty risk is the biggest predictor of long LoS.

​

People with HIGH frailty risk were 4 times more likely to have LoS>10d than those with LOW frailty risk

In-hospital deaths

Charlson comorbidy index biggest predictor of in-hospital death

Frailty risk is also a very important predictor

Readmissions

Previous admissions most important predictor

Costs – HRG tariff (1,000 pounds)

Frailty risk important predictor of cost

Hospital adm most important predictor

Data

​

• 5% random sample from the Hospital Episode Statistics (HES) of adults admitted to hospital as emergencies in England.

​

• Analytical time period: 1 April 2013 until 31 March 2019.

​

• 1,478,554 emergency admissions for 653,294 patients

​

​

​

HFRS – 75+ v 18+

• Patients 75+ categorised into three frailty risk groups:
• low (HFRS <5),
• intermediate (5≥ HFRS <15)
• high (HFRS ≥15)

​

• Patients 18+ categorised into four frailty risk groups:
• Zero (HFRS = 0)
• low (0> HFRS <5)
• intermediate (5≥ HFRS <15)
• high (HFRS ≥15)

HFRS distribution by age band (n=1,478,554)

As age increases, we have higher proportions in the high HFRS category.

LoS (count data)

Compared to those�with Zero frailty risk, the three HFRS categories have significant explanatory power in all groups except�for those aged 95+

In-hospital deaths

Hospital costs

HFRS applied to Spanish data

HFRS applied to Catalan data

Hospital patients 55+ categorised into four frailty risk groups (n=118k):

• Zero (HFRS = 0) – 67.7%
• low (0> HFRS <5) – 17.6%
• intermediate (5≥ HFRS <15) – 12%
• high (HFRS ≥15) – 2.7%

​

Primary care patients 55+ categorised into four frailty risk groups (n=156k):

• Zero (HFRS = 0) – 11.4%
• low (0> HFRS <5) – 38.8%
• intermediate (5≥ HFRS <15) – 42.7%
• high (HFRS ≥15) – 7.1%

​

Conclusion

• The HFRS has become a widely used tool to identify frailty risk

​

• Different forms of the HFRS are used capture diagnostic information from previous admissions

​

• High frailty risk is associated with longer LoS, higher costs and greater likelihood of in-hospital mortality

​

• This is association is evident across all adult age groups, not just those 75+

​

• Exploring applicability to primary data in contexts where ICD10 used to code diagnoses

Conclusion

• Identifying frailty is important at micro, meso and macro levels

​

• The HFRS has become a widely used tool to identify frailty risk

​

• Different forms of the HFRS are used capture diagnostic information from previous admissions

​

• High frailty risk is associated with longer LoS, higher costs and greater likelihood of in-hospital mortality

​

• This is association is evident across all adult age groups, not just those 75+

​