Methods and Applications

We have used various methods to interrogate, analyse and assess data held in primary care databases. Listed below are just a selection. For further information please contact the corresponding author of the relevant article below or contact Kelvin Jordan.

Assessment of Quality of Medical Records

We have assessed the quality of medical records, both through a systematic review and within our own regional database, CiPCA.

  • Jordan K, Porcheret M, Croft P. Quality of morbidity coding in general practice computerized medical records: a systematic review. Family Practice, 2004;21:4:396-412. More information
  • Porcheret M, Hughes R, Evans D, et al. Data quality of general practice electronic health records: the impact of a program of assessments, feedback and training, Journal of the American Medical Informatics Association, 2004;11:1:78-86. More information

Sickness Certificate and Fit Note Data

Within primary care medical records, GP certified absence from work may be identified. Work has been undertaken to establish the validity of GP recording of sickness certificates. Prior to 2010 the issue of a sickness certificate and the type of certificate issued may be located within the medical record. To establish the reason for issue of a sickness certificate the date of the certificate needs to be matched to a consultation date and the problem recorded at the consultation used as a proxy of the reason for certification.

After 2010 Fit Notes may be identified along with whether the patient was advised that they are not fit for work or may be fit for work (i.e. the GP has ticked one of the may be fit for work options: phased return to work, altered hours, amended duties, workplace adaptations). Recorded with the Fit Note is the reason for issue and the duration of issue.

  • Wynne-Jones G, Dunn KM, Mallen CD, Main CJ. Sickness certification in general practice: A comparison of electronic records with self-reported absence, Primary Health Care Research and Development, 2008:9:2:113-118.

Symptom Prognosis

We have an increasing interest in the prognosis of symptoms presented to primary care for which a diagnosis is not initially made. For example, in rheumatoid arthritis (RA), there is evidence that treatments are most effective if begun early (i.e. within three months of symptom onset). However, when patients do present to primary care, GPs often find it difficult to identify those patients with likely RA. This can result in further delays in diagnosis, referral and the commencement of appropriate therapies. Current work, using data from the Clinical Practice Research Datalink, aims to identify “at risk” symptom patterns based on primary care consultations that could be used to improve the recognition of patients with early RA in primary care. In order to do this, a definition of RA has been updated and is currently under-going peer-review. Work on the identification of signs and symptoms preceding a diagnosis of RA is ongoing. Contact Sara Muller for more information (

Other work is looking at the long term outcomes of undiagnosed joint pain, chest pain, and breathlessness symptoms recorded in primary care. Contact Kelvin Jordan for more information (

Comparison of Self-Reported and Recorded Morbidity

We have assessed how well self-reported morbidity, and self-reported consultation, matches what is recorded in the medical records.

  • Barber J, Muller S, Whitehurst T, Hay E. Measuring morbidity: self-report or health care records? Family Practice. 2010; 27:1:25-30. More information
  • Jordan K, Jinks C, Croft P. Health care utilization: measurement using primary care records and patient recall both showed bias, Journal of Clinical Epidemiology, 2006;59:8:791-797. More information

Use of Records to Assess Non-Response in Surveys

Non-response and attrition are common problems in longitudinal surveys. We have used medical records to assess the extent of bias this may cause.

Lacey RJ, Jordan KP, Croft PR. Does attrition during follow-up of a population cohort study inevitably lead to biased estimates of health status? PLoS ONE 2013; 8(12): e83948. doi:10.1371/journal.pone.0083948. More information 

Joinpoint Regression

Joinpoint regression is a useful approach to analysing trends. We used joinpoint regression to identify time of changes (the ‘joinpoints’) in the underlying trend in the incidence of new prescriptions for analgesia.

  • Bedson J, Belcher J, Matino OI, et al. The effectiveness of national guidance in changing analgesic prescribing in primary care from 2002 to 2009: an observational database study. Eur J Pain 2013;17:434-443. More information.

Propensity Scores

Propensity scores are an efficient approach to adjusting for many potential confounders by amalgamating them into a single variable (the propensity score). We have used propensity scores to assess the association of primary care management with outcome. In this case, the propensity score reflects the probability of an individual receiving that treatment or management given their baseline characteristics.

  • Ashworth J, Green DJ, Dunn KM,  Jordan KP. Opioid use among low back pain patients in primary care:  Is opioid prescription associated with disability at 6 month follow-up? Pain 2013;154:7:1038-1044. More information.
  • Blagojevic M, Jinks C, Jordan KP. The influence of consulting primary care in knee pain in older people: a prospective cohort study, Annals of the Rheumatic Diseases, 2008;67:12:1702-1709. More information.

Latent Class Analysis

Latent class analysis is an effective method for grouping people together based on observed characteristics. When used on the same variable measured repeatedly over time, it can be used to determine trajectories of morbidity, an approach we have used within primary care databases.

  • Strauss VY, Jones PW, Kadam UT, Jordan KP. Distinct trajectories of multimorbidity in primary care were identified using latent class growth analysis, Journal of Clinical Epidemiology, 2014;67:10:1163-1171. More information.