Relationship between Frailty and Risk
This analysis explores the relationships between:
- High degrees of frailty
- Risk of emergency admission in the next 12 months
- Risk of being in the top 2% of most expensive patients in the next 12 months
The risk data is derived from three of the predictive models calculated by the Johns Hopkins ACG® System.
Segmentation Analysis 1
The Venn diagram analysis looks at the overlap between three patient cohorts:
- The top 2% of people at risk of an emergency admission in the next 12 months (count = 3,313).
- The top 2% of people at risk of having high costs in the next 12 months (count = 3,313).
- Those identified by the electronic Frailty Index (eFI) as being moderately or severely frail (count = 3,064).
The total number of unique patients is 5,998.
The Venn diagram above helps to illustrate the degree of overlap between the patient cohorts. It serves as a useful reminder that it is important to use the correct predictive model to identify specific patient cohorts. Using the sample data above, if you are looking to reduce emergency admissions, by focusing only on patients with frailty you would exclude 2,081 at risk patients (783 + 1,298) and erroneously include 1,832 patients with frailty but who are not at risk (1,608 + 224).
The table below breaks down the average total cost per patient and average activity for each segment of the Venn diagram. It is colour-coded as a heat map.
| Measure / Segment | Patient Count | Avg. Total Cost (£) | Avg. No. Emer Admissions | Avg. No. OP 1st Attends | Avg. No. OP Follow-up Attends | Avg. No. GP Visits | Avg. No. Distinct Drug Count |
|---|---|---|---|---|---|---|---|
| Frail Only | 1,608 | 1,545 | 0.1 | 1 | 2.4 | 3.5 | 12.3 |
| Top 2% Emer Admission Only | 783 | 4,641 | 0.9 | 2 | 9.7 | 5.3 | 11.6 |
| Frail and Top 2% Emer Admission | 294 | 3,133 | 0.5 | 1.3 | 3.2 | 6.4 | 14.9 |
| Top 2% Prob High Cost Only | 853 | 8,279 | 0.7 | 2.5 | 9.2 | 5.6 | 15.3 |
| Frail and Top 2% Prob High Cost | 224 | 5,219 | 0.5 | 1.7 | 5.1 | 5.5 | 19.5 |
| Top 2% Emer Admission and Top 2% Prob High Cost | 1,298 | 11,090 | 1.7 | 3.1 | 11.7 | 7.1 | 18.4 |
| All three | 938 | 9,053 | 1.6 | 2.3 | 6.1 | 8.5 | 22.8 |
The heat map above shows that frailty by itself impacts less on cost and activity than being at risk of emergency admission or being at risk of high cost.
The cohort in both the top 2% at risk of an emergency admission and the top 2% at risk of incurring high costs have a higher average total cost (£11,090) than patients in all three segments (£9,053). This may be due to their slightly higher number of emergency admissions and that those with frailty had significantly fewer outpatient attendances.
At risk patients with frailty have fewer outpatient first attendances than at risk patients who are not frail.
Segmentation Analysis 2
This segmentation analysis is similar to the previous one, except that it casts a wider net to capture more patients who are at risk of an emergency admission.
The three patient cohorts are:
- The top 5% of people at risk of an emergency admission in the next 12 months (count = 8,281).
- The top 2% of people at risk of having high costs in the next 12 months (count = 3,313).
- Those identified by the eFI as being moderately or severely frail (count = 3,064).
The total number of unique patients is 9,513.
Of the 8,281 patients identified as being in the top 5% of those at risk of an emergency admission in the next 12 months, 25% (2,067) have moderate or severe frailty.
Segmentation Analysis 3
The final segmentation analysis is adjusted to include those patients falling into the top 3%-5% at risk of emergency admission (thus excluding the most at-risk patients). This is a slightly smaller cohort who may be more impactable by interventions or programmes designed to prevent those patients becoming higher risk.
The three patient cohorts are:
- The top 3%-5% of people at risk of an emergency admission in the next 12 months (count = 4,971).
- The top 2% of people at risk of having high costs in the next 12 months (count = 3,313).
- Those identified by the eFI as being moderately or severely frail (count = 3,064).
The total number of unique patients is 8,732.
Here you can see a very small overlap of patients who belong to all three segments.
When using risk models for case finding and patient selection, it makes sense to use the most appropriate predictive model. For example, to identify those at risk of an emergency admission, use the risk of emergency admission model. As we’ve seen, using less appropriate models can reduce the impact of interventions or programmes aimed at specific cohorts.
STP Population Health Analyses
Download the complete PDF of analyses and insights into the population profile of a Sustainability and Transformation Partnership in the South of England.
For more information about our population health management and analytics tools, see Population Health Analytics.
For more details about the ACG System and how it uses patient data, see our FAQ.