Author: The NHS Information Centre, Prescribing Support and Primary Care Services. Responsible Statistician: David Lloyd
Access to Healthcare for People with Learning Disabilities
Analysis of Primary Care Data to Support the Work of the Independent Inquiry into Access to Healthcare for People with Learning Disabilities
Copyright © 2010, The Health and Social Care Information Centre, Prescribing Support
and Primary Care Services. All Rights Reserved.
1
The NHS Information Centre is England’s central, authoritative source of health and social care information. Acting as a ‘hub’ for high quality, national, comparative data, we deliver information for local decision makers, to improve the quality and efficiency of care.
www.ic.nhs.uk Author: The NHS Information Centre, Prescribing Support and Primary Care Services Responsible Statistician: David Lloyd, Senior Service Manager Version: 1 Date of Publication: February 2010
2
Copyright © 2010, The Health and Social Care Information Centre, Prescribing Support and
Primary Care Services. All Rights Reserved.
Contents Contents
3
Background
4
Aims and Objectives
4
Study Design
5
Study Population
5
Data Analysis
5
Identification of Patients with Learning Disabilities
5
Demographics of Patients with Learning Disabilities
6
Ill-health Prevention
7
Healthcare Management
8
Discussion
9
Conclusion
11
Table Index
12
Appendices
18
Appendix 1 Read Codes for Identifying Patients with Learning Disabilities
18
Appendix 2 Read Codes Considered for Identifying Patients with Learning Disabilities
24
Copyright © 2010, The Health and Social Care Information Centre, Prescribing Support
and Primary Care Services. All Rights Reserved.
3
Background The Independent Inquiry into Access to Healthcare for People with Learning Disabilities (LD) published in July 2008, found convincing evidence that people with LD have higher levels of unmet need and receive less effective treatment, despite the fact that the Disability Discrimination Act and Mental Capacity Act set out a clear legal framework for the delivery of equal treatment. The inquiry team contacted the NHS Information Centre to commission a piece of supporting research analysing primary care data relating to healthcare for people with LD. The aim was to undertake an analysis to identify the domains most likely to be useful in relation to recommendations concerning service and equality monitoring, particularly at a local level. Comparative analysis was to be undertaken to identify health care needs using the non-LD population as a comparator group. The tender was awarded to the General Practice Research Database (GPRD) group of the Medicines and Healthcare Products Regulatory Agency. The GPRD is an anonymised primary care database containing data for approximately 10 million patients from 500 practices. It has been used in analysis for peer reviewed papers since the 1990s and is widely accepted as valid and representative. The multi-disciplinary team in the GPRD division are uniquely placed to analyse primary care data. This study analyses the GPRD to examine how GPs are recording LD on their computer systems, and compares the demographics of patients with LD to the rest of the population. It then compares the provision of ill-health prevention advice and interventions and healthcare management between patients with LD and the patient population as a whole.
Aims and Objectives 1. To describe the extent to which different Read codes are being used to identify patients with LD and to indicate what future work on standardisation is needed. 2. To describe the demographics of patients with LD compared to the rest of the population. 3. To compare the provision of recorded ill-health prevention advice and interventions between patients with LD and the general patient population in six key areas: (a) cholesterol check for people in a high risk subset, (b) body mass index measurement, (c) influenza immunisation for people in a high risk subset, (d) smoking status for people in a high risk subset, (e) contraceptive advice for patients aged 18-49, stratified by gender and (f) cervical screening stratified by age (25-49, 50-64). 4. To compare the healthcare management of patients with LD with the rest of the patient population for those with the following health problems: (a) obesity, (b) smoking, (c) asthma, (d) epilepsy, (e) urinary tract infection (UTI), (f) diabetes, (g) mental illness. The study specific aims stem from the measures used in the NHS Quality and Outcomes Framework (QOF). The measures are indicators of the quality of care for patients. Analysing differences between patients with LD and the rest of the patient population has the potential to identify areas where patients with LD experience a poorer quality of care. 4
Copyright © 2010, The Health and Social Care Information Centre, Prescribing Support and Primary Care Services. All Rights Reserved.
Study Design A cohort study design was selected to maximise the power of each analysis, as LD is rarer than many of the outcomes of interest.
Study Population The study window ran from 1st January 2002 until 31st December 2007. The LD cohort consisted of any acceptable patient within GPRD with a clinical or referral record of a Read code for LD at any age, with at least one day of “up to standard” (UTS) follow-up as an adult (aged 18 or over) in the study window. The index date for each patient was set to be the latest of the LD diagnosis date (or randomly selected date for the unexposed cohort), the 1st January of the year the patient became aged 18, the registration date, or the study start date (1st January 2002). Patients with indeterminate gender were excluded. The unexposed cohort for objective (2) consisted of a sample of 500,000 patients from the rest of the GPRD population. Patients with indeterminate gender were excluded. For analyses (3) and (4), patients with LD in the subgroup of interest (e.g. for (4a) patients with LD who were obese) were identified and then matched to up to four patients from the unexposed cohort by gender, age at index date (within five years) and practice using observation window matching.
Data Analysis Data was managed and analysed using Stata v10. All proportions were compared using either the standard chi-square test or conditional logistic regression (suitable for comparing proportions with matched data).
Identification of Patients with Learning Disabilities The first objective of the study was to describe the extent to which different Read codes are being used to identify patients with LD and to indicate what future work on standardisation is needed. In order to achieve this objective, experts in the field were consulted to determine a list of Read codes for LD. Read codes were included if at least three of the four experts agreed on their inclusion. The resulting code list can be found in Appendix One. Appendix Two shows all of the suggested Read codes and the level of agreement. The frequency of use of each Read code was then summarised by year and overall and is shown in Table 1. The number of records of LD increased over the study period from about 1,500 records in 2002 to about 2,500 in 2007. The code used most frequently in the study period, and in four of the six years, was “Learning disability NOS”. However, the number of records of this code remained relatively stable throughout the study period, and hence the proportion of records Copyright © 2010, The Health and Social Care Information Centre, Prescribing Support and Primary Care Services. All Rights Reserved. 5
attributable to this code fell. Autism, included in our code list due to the strong relationship with LD, was frequently recorded, with “Autism” and “Autistic disorder” making up the remaining two of the top three codes each year and overall. There was a reduction in the use of the code “Down’s syndrome – trisomy 21” from approximately 150 events (11%) in 2002 to 100 (4%) in 2007. Down’s syndrome is the largest known cause of learning disabilities, but there are no codes specifying Down’s syndrome in the QOF for LD. There was a significant increase in the number and proportion of records relating to procedures and administration relating to LD (Read codes starting with 8 or 9).
Demographics of Patients with Learning Disabilities The second objective of the study was to describe the demographics of patients with LD compared to the rest of the patient population. A sample of 500,000 patients without LD from the GPRD population was randomly selected and chi-squared and anovas were conducted with the a priori null hypothesis of no difference in each case. Since only the year of birth is available in GPRD, age was calculated as the difference between the year of the index date and the birth year. Body mass index (BMI) was calculated from weight and height information. The closest height and weight measurements to index date were used. BMI was calculated using weight (kg) / (height(m)*height(m)). BMI measurements of 70 were set to missing. Patients were categorised as being underweight (BMI < 20), healthy (BMI 20-25), overweight (BMI 25-30) or obese (BMI > 30). Alcohol consumption and smoking status were calculated from records of drinking and smoking status, and from searching for alcohol consumption and smoking records in the patient’s history using Read code lists used in previous studies by GPRD. Only records from before or on the index date were included. Due to the large number of patients, all statistical tests resulted in p-values of
