Journals LibraryNHS NIHR - National Institute for Health Research
Understanding variations in outcomes in asthma and COPD: use of routine clinical data
Understanding variations in outcomes in asthma and COPD: use of routine clinical data
29 April 2015
01 June 2013
31 August 2014
COPD; Database; Retrospective; Cohort; Respiratory; Variation; Hampshire
- Professor Mike Thomas, Department of Primary Care and Population Sciences, University of Southampton
- Dr Lynn Josephs, Department of Primary Care and Population Sciences, University of Southampton
- Mr Matthew Johnson, NIHR CLAHRC Wessex, University of Southampton (Analyst: data management and extraction)
- Mr David Culliford, Department of Primary Care and Population Sciences, University of Southampton (Senior Statistician)
- Miss Rosanna Orlando, NIHR CLAHRC Wessex, University of Southampton (Statistician)
- Dr Hugh Sanderson (previous Clinical Lead, Hampshire Health Record)
- Professor Paul Roderick, Department of Primary Care and Population Sciences, University of Southampton
- Dr Borislav Dimitrov, Department of Primary Care and Population Sciences, University of Southampton (Medical Statistician)
A recent Department of Health document has highlighted the large variations in care and in outcomes of asthma and COPD in the UK, and there are a number of national initiatives aimed at improving the care of these chronic respiratory diseases. Most diagnosis and management occurs in the community, and there is a pressing need to define good care in chronic respiratory disease and to reduce geographical variation and social inequalities.
There have been high-profile calls for the better use of routinely collected NHS data in research. Clearly, randomised controlled trials are essential in defining the benefits (and costs) of interventions and eliminating possible bias, but the restrictive entry criteria and the constraints of study protocols provide limits to generalisability and external validity, and limit how useful these studies are in informing best quality care in the community. The use of observational data from routinely held datasets could provide complementary and important information on ‘real-world’ patterns of care and outcomes, looking at what actually occurs in respiratory disease management and the outcomes that are currently achieved in general practice.
We are fortunate in having access to the Hampshire Health Record analytical database (HHRA), which we will use to examine aspects of respiratory care in Hampshire. The HHRA is a well validated local resource to researchers, giving access to pseudonymised routinely coded clinical and prescribing data for over 1 million patients treated in Hampshire, spanning primary and secondary care. It consists of an electronic database designed to accept existing electronic patient records from a number of sources. Records relating to an individual patient are collected into a single Common Health Record for that patient. The main sources are from General Practice computer systems (coded clinical entries and prescribing data) and Hospital activity data, including Hospital Episode Statistics (HES).
We plan to use the HHR to improve our understanding of the care of COPD and asthma in Hampshire. There is considerable evidence of on-going health inequalities in COPD and asthma, with wide variations in outcomes (including hospitalisation rates) between localities and between GP practices in the UK. Disadvantaged groups bear much of the brunt of poor outcomes in respiratory diseases. Multi-morbidity with other long-term conditions is common and frequently not addressed.
Specific aims of our project are as follows:
- To use the HHRA data to describe the profiles, processes of care and outcomes for patients with COPD and asthma within different practices in Hampshire;
- To test the feasibility of using the HHRA data to assess utilisation of primary and secondary care resources and to identify “at risk” individuals.
In particular, wewillusetheHHRA to assess:
- Variations in healthcare provision, e.g. prescribing, QOF reviews, use of pathology, hospital outpatient and inpatient services;
- Variations in outcomes, e.g. emergency hospital admission, out of hours service use, accident and emergency attendance, exacerbation frequency, mortality;
- Variations in patient behaviour, e.g. consultation rate, adherence to medication, smoking, patterns of medication use;
- Co-morbidity/multi-morbidity: prevalence, influence on health resource use and outcomes.
Data will be able to be analysed at individual patient level, by GP practice or hospital, or by locality/CCG. Such evaluations of care and outcome would have academic interest, but are also of great interest to commissioners and providers aiming to re-design and re-configure services to provide best outcomes and value for money. We have formed links to local NHS organisations and plan to use these data to inform subsequent interventions to improve the standards of respiratory care, and will contribute to the Wessex Academic Health Science Network programme (AHSN).
Study design will be a retrospective observational cohort study using individual patient-anonymised routine data in the HHRA database. We will refine and update Read (Read Clinical Classification) code lists to define patient cohorts with COPD or asthma and develop Read code sets to study consultations, prescriptions and use of secondary care to provide useful estimates of treatment patterns, adherence to prescribed medication, exacerbation frequency, hospital admission and re-admission data and prevalence and management of recorded co-morbidity.
Changes to project objectives
Our project plans remain unchanged but, so far, we have restricted our study to patients with a primary care diagnosis of COPD rather than studying a second cohort of patients with a diagnosis of asthma in parallel to our COPD study. We experienced unforeseen delays in accessing the HHRA data and found the need to develop new Read code lists for identifying COPD, processes of care, outcomes and co-morbid diseases; we believe that our new coding strategies and methods for interrogating the HHRA database will improve the reliability of our findings, but this has slowed our progress. We are continuing our evaluation of COPD and clinical outcomes and this ongoing work has formed the basis for a “respiratory theme” NIHR (National Institute of Healthcare Research) CLAHRC (Collaboration for Leadership in Applied Health Research and Care) Wessex collaboration, aimed at implementing our findings for the benefit of Hampshire patients. We have now finalised a Read code list to identify patients with a primary care diagnosis of asthma, which will be the starting point for studying asthma in a similar way and with the benefit of our experience from studying COPD.
We performed a retrospective observational cohort study, using individual patient-anonymised routine data held in the Hampshire Health Record Analytical database (HHRA, an electronic NHS database holding coded clinical data for over one million patients living in Hampshire, UK) described above. We compiled lists of diagnostic Read codes (Version 2 codes) for chronic obstructive pulmonary disease (COPD) and asthma, and developed an Access database to manage code lists for subsequent analysis. We defined an initial cohort of patients diagnosed with COPD prior to 01/01/2011 from pseudonymised patient data held in the HHRA for whom we had follow-up records for the following 2 years. We used Read codes to describe patient demographics (age, sex, ethnicity, deprivation indices) and characterised patients in terms of smoking status, pulmonary function (FEV1 levels, FEV1 %predicted and FEV1/FVC ratio) breathlessness scores and processes of care (flu and pneumococcal vaccination, smoking cessation advice, referral for pulmonary rehabilitation). We studied outcomes in these patients over a two year period (01/01/11 to 31/12/12). We devised code lists to define emergency secondary care use (respiratory A+E attendance, hospital admission and readmission) using ICD-10 codes from the Secondary Uses Service (SUS) and Accident and Emergency Clinical Codes. We then identified a COPD cohort with follow up records available for a 3 year period (1/1/11 to 31/12/13) in order to study COPD outcomes (exacerbations, A+E attendances and respiratory-cause hospital admissions) over 3 years. In this cohort, we have studied variation in COPD practice prevalence rates (% of patients in each practice diagnosed with COPD) and variation in respiratory-cause hospital admission rates across these HHRA practices.
For the initial cohort for whom we had 2 years follow-up data, we identified 21243 patients with COPD among 140 practices, mean age (SD) 71.5 (11.7) years, 55.0% male. FEV1 values were available in 19085 patients (89.8%) and % Predicted FEV1 in 10236 (48.2%); median (IQR) %predicted FEV1 was 58 (43–72)%. FEV1/FVC ratios were available in 17536 (82.5%); median (IQR) FEV1/FVC was 58.1 (47.0–69.0) %. Smoking status was recorded in 21068 (99.2%): 37.8% ‘current smokers’, 51.0% ‘ex-smokers’, 10.4% ‘never-smokers’. Over the two years, 2777 patients (13.1%) had one or more respiratory hospital admission; in addition, 1285 (6.0%) attended A+E with a respiratory complaint. 2446 (11.5%) patients died (12.2% of men, 10.7% of women, p<0.001). Comparing those who died with those who survived, mean (SD) age was greater (79.2 (9.8) versus 70.5 (11.6) years, p<0.001) and median (IQR) FEV1 was lower (1.04 (0.73-1.49) versus 1.39 (0.99-1.88) litres, p<0.001). Death occurred in 856 of 2777 who were hospitalized (30.8%), compared to 1590 of 18466 (8.6%) of those who were not (p<0.001).
We have now studied prevalence rates (% of patients in each practice diagnosed with COPD) and hospital admission rates in those 16149 COPD patients in whom we have follow-up records for 3 years. We have shown a very large variation in COPD prevalence rates across those 137 practices who have contributed data to the HHRA over the 3 year period, with rates ranging from 0.1% to 2.8% (average rate=1.3%). Regarding emergency respiratory hospital admissions over the 3 years, one fifth of patients (3276) were admitted to hospital on ≥1 occasion with 132 patients (0.8%) being admitted on ≥6 occasions. In 75% of cases, patients were admitted to hospital having presented themselves at the A+E department, rather than being referred by a GP or out-of-hours doctor. We have demonstrated a 13-fold variation in respiratory-cause unplanned hospital admission rates (ratio of number of unplanned admissions to number of COPD patients in each practice) across practices. There was greater variation across practices when the analysis was confined to those patients who referred themselves to A+E, compared with those patients admitted by a GP or out-of-hours doctor.
We have also studied smoking status and lung function measurements in 16185 COPD patients with 3 years’ follow-up data (male 54.5%) mean (SD) age 70.9 years (11.4). FEV1 %predicted values were available in 50% of these patients: median (IQR) 59 (45-72). Smoking status was recorded in 96.2% of patients: 37% current smokers, 57% ex-smokers and 3% never smokers. We have found that in patients aged <65 years, 57% (n=2521) were current smokers, compared to 29% (n=3411) of those aged >65 years (<0.0001). In those patients with FEV1 ≥50%, the proportion of currently active smokers was highest in the youngest patients: 72.7% (n=341) in those aged <55 years and 55.4% (n=572) in those aged 55-64 years, whereas in the older patients only a minority were active smokers: 38.8% (n=698) in those aged 65-74 years and 24.8% (n=476) in those aged >75 years.
Conclusions from 2 year data
Outcome data for the original 21,243 COPD patients followed for 2 years has highlighted the high percentage of patients still smoking (over 1/3) and the poor prognosis of COPD: 1 in 10 patients died over the 2 year follow-up period, respiratory hospitalisation occurred in over 1 in 6, and almost 1 in 3 of those with a respiratory admission were dead at the end of the 2 year observation window.
Conclusions from 3 year data
We have shown a very wide range in prevalence rates for COPD diagnosis across HHR practices in Hampshire. We are currently examining age demographics across these practices, as some practices (including some with the lowest COPD prevalence rates) have very high proportions of young patients under the age of 30 years who would be at minimal risk of COPD. However, our findings suggest that the diagnosis of COPD may be being missed more often in some practices than others and this requires further evaluation.
We have also shown a thirteen-fold variation in respiratory-cause unplanned hospital admission rates (hospital admissions in relation to practice COPD populations) across HHR practices and we believe that factors other than disease severity are likely to be important in explaining this variation. Hospital admissions may represent poor care in the community and are a heavy financial burden to the NHS and more work needs to be done to address this.
Our smoking data have shown that the proportion of currently active smokers was highest in the youngest patients and this was marked in those patients with mild or moderate airflow obstruction. We conclude that while smoking cessation is important in all patients with COPD, it is especially important that we focus attention on younger patients and those with better preserved lung function, in whom there is the greatest scope for improving the prognosis. We are currently looking for evidence of smoking cessation support in the records of our COPD cohort to try and shed light on the current situation to know how best to improve care.
Plain English summary
Chronic obstructive lung disease “COPD” is a common lung disease. It is progressive and disabling, and often undiagnosed. Our study assessed variation in care and outcomes for patients with COPD in Hampshire, and factors underlying variation to inform interventions to improve care.
The Hampshire Healthcare Record (HHRA) is a well-established database of NHS information on >1.3 million patients in Hampshire and includes hospital and GP records. We have used anonymised information from the HHRA to look at variations in care and outcomes for COPD. Our findings will inform future projects aimed at improving care in underperforming areas.
We identified 21,243 patients with a diagnosis of COPD as at 1/1/2011 and examined their progress over two years. We highlighted the high percentage of patients still smoking (over 1/3) and the poor prognosis of COPD: 1 in 10 patients died over the 2 years, unplanned respiratory hospital admissions occurred in over 1 in 6, and almost 1 in 3 of those with an admission died during the 2 years, making the prognosis worse than that following a heart attack or most cancers. We have also studied COPD prevalence rates (numbers of patients in each practice diagnosed with COPD) and hospital admission rates in 16,149 COPD patients in whom we have follow-up records for 3 years, showing a very large variation in COPD diagnosis rates across 137 practices, with rates ranging from 0.1% to 2.8% (average rate=1.3%). 1/5 were admitted to hospital on ≥1 occasion with 1% being admitted on ≥6 occasions. In ¾ of cases, patients were admitted to hospital having presented themselves at the A+E department, rather than being referred by a GP or out-of-hours doctor. We demonstrated a 13-fold variation in hospitalisation rates. Hospital admissions may represent poor care in the community so work needs to be done to address this.
We found that in patients aged <65 years, 57% (2521 patients) were still smoking, compared to 29% (3411) of those aged >65 years. Smoking cessation is especially important in younger patients and those with better preserved lung function when there is the greatest scope for improving the prognosis.
- Abstract IPCRG conference Athens, 2014: Understanding variations in outcome in COPD: Early results from an observational study using routine clinical data. L. K. Josephs, M. Johnson, P. J. Roderick, M. Thomas.
- Abstract IPCRG conference Athens, 2014: Understanding variations in outcome in COPD: Use of routine clinical data. L. K. Josephs, P. J. Roderick, D. M. Thomas.
- Poster presentation NIHR SCPR Showcase 2014 (Abstract 0061) Understanding Variations in Outcome in COPD: Early Results of an Observational Study using Routine Clinical Data. L.K. Josephs, M. Johnson, P.J. Roderick, D.M. Thomas.
- Abstract accepted for oral presentation at 4th IPCRG Scientific Meeting, Singapore, May 2015: Age-related smoking status in patients with COPD stratified by severity of airflow obstruction: an observational cohort study. Lynn Josephs, Matthew Johnson, Rosanna Orlando, David Culliford, Paul Roderick, Mike Thomas.
- Abstract accepted for oral presentation at 4th IPCRG Scientific Meeting, Singapore, May 2015: Variation in Respiratory Hospital Admissions in patients with COPD: an observational cohort study using the Hampshire Health Record Analytical database (HHRA). Lynn Josephs, Matthew Johnson, Rosanna Orlando, David Culliford, Paul Roderick, Mike Thomas.
As this is a retrospective database study, there are limited opportunities for PPI involvement in study design or conduct. However, we have met with the Lead Respiratory Patient Representative on several occasions (he also regularly attends CLAHRC research group meetings) and he has offered support in reinforcing the importance of accurate clinical coding in primary care (which is important in improving the quality of database research, as well as being fundamental to good clinical care). One lesson learned is that patient representatives attend many different meetings and they can often make their voices heard above those of the medical profession and academics, which gives them an important role as a mediator between groups with different interests and responsibilities.
Early results have confirmed the marked variation in COPD outcomes (respiratory-cause hospital admissions) across Hampshire, which supports the findings of the 2012 NHS Atlas of Variation in Respiratory Disease in England. Factors other than disease severity are likely to be important in explaining this variation and more work needs to be done to investigate this. We are currently exploring the relationship between practice respiratory skill-mix (based on questionnaire data examining COPD training among primary care practitioners) and clinical outcomes in our COPD cohort. We will use our findings to highlight areas where outcomes are worse and we believe this will help identify practices most in need of intervention.
During the development of our study methodology, it has become apparent that clinical coding lists are very rarely available to the reader of published database research. This makes it impossible to compare studies or to know how generalizable the results of published studies are to other populations. We are developing Read code lists, agreed by consensus, for codes used in diagnoses (especially important for multi-morbidity studies), disease monitoring/processes of care and clinical outcomes and we intend to make these web-based and freely available for use or discussion.
Articles published :
Patient coping strategies in COPD across disease severity and quality of life: a qualitative study. Mike Thomas, George Lewith, Sarah Brien. 2016 Doi:10.1038/npjpcrm.2016.51
This project was funded by the National Institute for Health Research School for Primary Care Research (project number 195)
Department of Health Disclaimer
The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR School for Primary Care Research, NIHR, NHS or the Department of Health.