Calculation of ‘expected’ frailty prevalence and diagnosis rate

The calculation of ‘expected' frailty prevalence and hence the diagnosis rates used in the analysis has been based on assumptions from the development of the electronic Frailty Index and its application to the Kent Integrated Dataset.  An overview of the approach is set out below and the calculation for England is also shown.

Electronic Frailty Index

The index was developed and validated by the University of Leeds in partnership with TPP/SystmOne and the University of Birmingham[1].   Available across the main primary care clinical systems, it uses routinely collected data to produce a frailty score for everyone aged 65 and over based on the presence or otherwise of 36 different age related long term conditions, disabilities signs or symptoms.

The research[2] to develop and validate the tool identified that around 15% of population aged 65 and over are living with moderate or severe frailty and have around 3-5 times higher risk of adverse outcomes such as hospitalization, nursing home admission or mortality than someone aged 65 and over who is considered to be living without any level of frailty.

Kent Integrated Dataset

Research into Current & Future Cost of Frailty to Health and Care by Kristin Bash MFPH Public Health Specialty Registrar NHS England included applying the eFI to the Kent Integrated Dataset (KID).  This provided more detailed information about the prevalence of different levels of frailty by age band (see chart).  This was presented at National Frailty Conference 28 September 2017.  It is these assumptions that have been used to generate ‘expected’ levels of frailty for each practice population, based on the practice-specific demographic profile.  Whilst these expected levels are also likely to be influenced by the level of deprivation (as a proxy for determinants of health), no adjustment has been made for this in the current analysis of the diagnosis rate.

Calculation of average (mean) diagnosis rate for England

The prevalence assumptions from the KID analysis are applied to the population within each age band to estimate the number of people living with no or different levels of frailty (i.e., fit, mild, moderate, severe).  Nationally, the age band with the most people living with moderate and severe frailty is the 80 – 84 year old cohort, although prevalence continues to increase in the higher age bands.  This increase in prevalence with age is in part due to the construct of the Frailty Index – using ‘deficits’ that increase in prevalence with age but do not 'saturate' too early.

The estimated number of ‘expected’ moderate and severe frailty diagnoses is then compared with the numbers reported in the GP Contract data.  These are indicators CCDCMI12 (moderate) and CCDCMI13 (severe).

In the analysis above, the expected number is adjusted down to reflect the GP Practices not included in the GP Contract Data. The sum of indicators CCDCMI12 (moderate) and CCDCMI13 (severe) is then divided by the adjusted expected number to produce the Estimated Diagnosis rate (expressed here as a percentage).  

The same approach has been used for each GP Practice population.  CCG and STP estimated have been calculated by summing up the individual GP practice ‘expected’ and actual numbers for the practices within their area.


  • No adjustment has been made for deprivation in the current analysis of the diagnosis rate.
  • Confidence intervals have not been calculated.
  • The results of applying the eFI at practice level will be influenced by the quality of coding for each patient.  Practices with a high quality and depth of coding are likely to produce higher Frailty Index scores for equivalent patients when compared to practices with poorer or incomplete coding.
  • Patients who have changed practices are particularly at risk of ‘under scoring’, particularly if the new practice is on a different clinical system, as not all the historic information may have been recoded.



Other links