Published 12th August 2019

By Paul M. Henery, Research Assistant in the Inequalities programme.

Over the last few years, the Scottish Government has been integrating health and social care provision under one joint body. One of the aims of this integration process is to provide home-based, joined-up care for older people – primarily as a response to rising NHS costs and hospital bed-days for this demographic, as well as a drive  to reduce emergency admissions. The majority of over-65s have multiple conditions or “multimorbidity” – this by definition means that social care cannot use a one-size-fits-all approach.

How important is multimorbidity in terms of integrated care?

Understanding multimorbidity is key to providing effective person-centred home care, as the impact of having multiple conditions on the health of the individual is greater than the sum of its parts. There is evidence of potentially adverse interactions between medications prescribed for different conditions (Guthrie et al 2015), and the mental load of maintaining appointments and managing prescriptions is another factor, particularly in deprived areas (O’Brien et al 2011) where mental co-morbid conditions are more common (Barnett et al 2012).

Historically treatment of multimorbidity has followed the “disease-focused” approach, which treated each disease individually. Integrated care represents a shift to a “person-focused” approach which considers the individual as a whole, recommending for example a reduction in medication if the costs (both mental and physical) outweigh the benefits.

How can we measure multimorbidity?

Multimorbidity is a particularly difficult concept to measure as what defines a “disease” is in itself subjective, and subject to diagnosis by a physician. In the context of integrated care, it is important to determine the most precise way of measuring multimorbidity – in terms of who is and is not multimorbid, as well as how “severe” the multimorbidity is – in order to best target care to those who need it. Little research has been conducted using a Scottish population into determining which kinds of multimorbidity measures predict health outcomes, and no studies have looked at the best multimorbidity measure for predicting social care.

When conducting research using administrative datasets, how well a particularly multimorbidity measure predicts an outcome can vary primarily according to three aspects:

  • The study population (particular conditions may have a greater impact on health in a general population compared to an older one)
  • The outcome measure (specific conditions or measures may perform better at predicting different outcomes, such as admissions or mortality – the same measure cannot always be used)
  • The data available (admissions, prescriptions, or self-reported)

The latter is where the researcher’s judgement has the biggest impact: multimorbidity “indices” generally consist of flags which identify particular conditions, and can be weighted (where particular conditions have a higher score) or non-weighted. Measures are typically derived from either diagnosis data (such as ICD codes – standardised worldwide identifiers for specific conditions, recorded at physician appointments or on admission to hospital) or medication data (from prescription dispensation records). These indices are often compared to “proxy” measures (such as a count of prescription classes or ICD codes) in order to ascertain whether the additional complexity improves precision.

What I did

Part of my PhD research focused on how well six multimorbidity indices – three diagnosis-based and medication-based measures apiece, with one proxy measure each – performed in predicting transition into social care. I analysed a linked health and social care panel dataset consisting of the entire over-65 population in Scotland from 2011-16, whom were not in receipt of social care at the beginning of the financial year (the “index date”).

I calculated each case’s multimorbidity “score” from the conditions identified in each index added together, with data for each score taken a year prior to the “index” date (i.e. the date at which the individual was first recorded in the dataset), with outcome being if the individual chose to use social care having not used it at the start of the year. I ran separate regression analyses for each multimorbidity measure, and following this determined which measure performed best at correctly predicting who did and did not transition into social care.

What I found

My research found that all six multimorbidity measures performed adequately at predicting transition into social care. One of the medication-based measures, the Chronic Disease Score 2 (CDS-2) performed better than all others. The CDS-2 was derived from American medication data and has never before been adapted for use with the UK prescribing system.

These results indicate that multimorbidity, however measured, can be used to predict transition into social care in older people in Scotland. In addition, the performance of the CDS-2 provides health and social care providers with a new tool to predict who is likely to transition into social care if they have access to medication records.

First published: 12 August 2019

<< 2019