Resource use measurement in trials conducted in care homes: a study of level-of-agreement between data collected from GP records and care home records

Sach, T, Desborough, J, Houghton, J and Holland, R (2015) Resource use measurement in trials conducted in care homes: a study of level-of-agreement between data collected from GP records and care home records. Value in Health, A689. (doi:10.1016/j.jval.2015.09.2555).

Abstract

Objectives
Methodological research focused on resource use measurement has been limited in comparison to the amount of research focussed on measuring outcomes within the economic evaluation context. This study was designed to assess the level-of-agreement between two different sources of health and social care resource use data collected on care home residents.

Methods
The methods were informed by a review of level-of-agreement studies concerned with resource use in older people. In the base case, resource use data collected from both GP medical records (electronic records) and care home records (paper-based records) on 362 care home residents were obtained as part of the CAREMED cluster randomised controlled trial. Descriptive statistics were explored before assessing level-of-agreement through percent agreement, 95% limits of agreement, and Lin’s concordance correlation coefficient (CCC). Sensitivity analyses excluded non-users and tested timeframe. Factors affecting the magnitude of difference were explored using multi-level modelling.

Results
Several resource items (number of GP, out of hours GP and podiatrist contacts) were found to have substantial agreement (0.61 to 0.80) between the GP records and care home records according to the CCC. The number of total visits, dietician, paramedic and SLT contacts showed moderate agreement (0.41 to 0.60). Most resources showed a poor (less than 0.00) or slight (0.00 to 0.22) level-of-agreement either due to care home records (for chiropodist, music therapy, and social worker contacts) or GP records (for phlebotomist and practice nurse) recording a greater number of visits. Patient classification (residential/nursing), number of falls, number of STOPP criteria met, number of medications and comorbidities significantly affected the magnitude of differences observed.

Conclusions
This research suggests that both sources of data are reliable for some resources but not others, indicating dual sources may be necessary where a wider perspective is important and feasible in terms of costs of data collection.

This record has no associated files available for download.

Contributors

Author: T Sach

Download statistics