Timeliness of recording in the Clinical Practice Research Datalink (CPRD) – an initial step in the implementation of near real-time vaccine safety surveillance IJPDS (2017) Issue 1, Vol 1:317 Proceedings of the IPDLN Conference (August 2016)

Main Article Content

Andreia Leite
Nick Andrews
Sara Thomas
Published online: Apr 19, 2017


ABSTRACT

Objective

Near real-time vaccine safety surveillance (NRTVSS) using electronic health records (EHR) is an option for post-licensure vaccine safety assessment. NRTVSS requires timely recording of outcomes in the database used. Our study aimed to examine recording delays in the Clinical Practice Research Datalink (CPRD) to inform the feasibility of implementing NRTVSS in England using these data. 


Approach
To examine delays we selected 4 outcomes of interest for NRTVSS: Guillain-Barre syndrome (GBS), Bell’s palsy (BP), optic neuritis (ON), and seizures for the period January 2005 to July 2015. Timeliness of CPRD records was assessed in two ways: 1) Using linked CPRD-hospital episode (HES) data to compare the hospital diagnosis date with the date the record was entered in CPRD (system date), 2) Looking at delays in recording (e.g. due to feedback from specialist referral) in stand-alone CPRD. For the latter the event date was compared with the system date. However, system dates can be changed when practice software is updated or there is mass transfer of a patient’s records. After investigation, we excluded these uninformative system dates by excluding records from patients who had more than 100 records with the system date on the same day.


Results
67813 patients were identified in CPRD (GBS:n=1081, BP:n=15835, ON:n=2236, seizures:n=48866), 64527 in HES (GBS:n=1680, BP:n=8468, ON:n=1746, seizures:n=53080) and 14104 in both databases (GBS:n=356, BP:n=1511, ON:n=226, seizures:n=12036). For the CPRD-HES comparison, 11843 patients with a diagnosis of interest both in CPRD and HES were included (GBS:n=321, BP:n=1374, ON:n=190, seizures:n=9976). Of these, the majority had a record in CPRD before or within 1 month of the HES record (GBS:49.5%, BP:83.8%, ON:66.8%, seizures:69.8%). For 6 months the corresponding percentage was more than 85% for all conditions examined (GBS:85.4%, BP:92.9%, ON:90.0%, seizures:86.6%). For stand-alone CPRD 57317 patients were included (GBS:n=972, BP:n=14275, ON:n=1958, seizures:n=40327). The majority had a record within one month of the event date (GBS:67.9%, BP:89.3%, ON:71.8%, seizures:83%). More than 87% of records occurred within 6 months of the event date (GBS:87.9%, BP:94.4%, ON:91.6%, seizures:94.9%).


Conclusion
This work shows that most diagnoses examined were recorded with a delay of ≤30 days, making NRTVSS possible. The distribution of the delays was condition-specific and the weekly delay distribution could be used to adjust for delays in the NRTVSS analysis. CPRD can be a viable data source to use in this kind of analysis; next steps will include trial implementation of the system using these data.


Objectives

Evidence on the economic impact of heart failure (HF) is vital in order to predict the cost-effectiveness of novel interventions. The objective was to estimate the healthcare costs of HF during the last five years of life.

Approach

Adults who died with HF in 2012/3 were identified through linked English Office of National Statistics mortality data and Clinical Practice Research Datalink (CPRD) primary care data. CPRD and linked Hospital Episode Statistics admissions data were used to estimate the cost of primary care prescriptions and primary care and hospital admission healthcare with 95% confidence intervals (CI). Generalized least squares regression was used to estimate the relationship between costs, HF diagnosis and patient characteristics.

Results

In the last 90 days of life of 1,555 identified patients, healthcare costs were £8,912 (95% CI £8,436-9,388)per patient, more than 90% of which were for inpatient or critical care. In the last 90 days, patients spent on average 17.8 days (95% CI 16.8-18.8) in hospital and had 8.8 (95% CI 8.4-9.1) primary care consultations. Most (59%) patients were in hospital on the day of death. Mean quarterly healthcare costs were significantly higher after diagnosis than preceding diagnosis (by \textsterling1,479, 95% CI {\textsterling}1,286-1,671). Younger patients and patients with higher comorbidity had higher costs.

Conclusion

Healthcare costs increased sharply at the end of life and were dominated by hospital care. There is potential to save money by implementation and evaluation of interventions known to reduce HF hospitalisations, particularly at the end of life.

Article Details