This document discusses record linkage and its use in pharmacovigilance. Record linkage involves combining records from different data sources that relate to the same individual to create a single longitudinal record. It allows rapid access to a patient's complete medical history across different data sources. This reduces the time needed to study relationships between drug exposure and health outcomes. Challenges include ensuring data quality and completeness when integrating records from various sources.

1. Vineetha Menon
5th Pharm.D

2.  Pre-marketing clinical trials are not able to
define many aspects of drug safety because:
1. Small no of patients
2. Large no of patients receiving the drug for
small durations
3. Doses and formulations of the drug may
change during drug development
4. Exclusion of special population from the
clinical trials

3.  PEM is a non-interventional, observational
cohort form of pharmacovigilance
 PEM was developed by Professor Bill Inman
at the Drug Safety and Research Unit (DSRU)
at Southampton in 1981

4.  In UK, all the patients are registered with
NHS-GP
provides the primary care
and act as a gateway to specialist and
hospital care
 File notes in general practice contains
information about primary care, secondary
and tertiary care (life long record)

5.  GP issues prescription for medications he
considers medically warranted
 Patient
takes the prescription to the
pharmacist, who dispenses the medication
and sends the prescription to the PPD
(which is a part of NHS-BSA), for
reimbursement
 PPD provides DSRU with electronic copies
of all the prescriptions issued throughout
UK, for the drugs being monitored

6.  Products that are selected for study by PEM
1. New drugs, expected to be used widely
2. Established
products, used
indication/ new population
for
new
 Collection of exposure data begins soon
after the new product is launched
 These arrangements operate for a length of
time necessary for the DSRU to collect first
50,000 prescriptions, that identify 20,00030,000 patients given the new drug being
monitored

7.  For each patient in the study, DSRU prepares
a computerized longitudinal record in the
date order of drug use
 After 3-12 months from the date of first
prescription for each patient, the DSRU
sends the prescriber a green form
questionnaire
 This is done on an individual patient basis
 Doctor receives maximum of 4 green forms in
a month

8. Request information
on:










Age
Sex
Indication for Rx
Dose
Start date
Stop date
Concurrent diseases
Concomitant
therapy
All events that have
occurred since Rx
Cause of death

9.  Each green form is reviewed by a medical/
scientific officer monitoring the study, to
identify possible serious ADRs or events
requiring action
 Events are coded and entered in database
using a hierarchical dictionary arranged by
system-organ class with specific lower terms
grouped under broader higher terms

11. 1. PEM is non-interventional
2. The method is national in scale and thus
provides real world data
3. Exposure data is derived from dispensed
prescriptions
4. Method can detect adverse reactions or
syndromes that none of the reporting
doctors suspected to be due to the drug
5. Method allows close contact between the
research staff and reporting doctors

12. 6. ADR reporting is more complete by this
7.
8.
9.
10.
method
Method is found to be successful in regularly
producing data in 10,000 or more patients
given newly marketed drugs
Method identifies patient with ADRs who
can be studied further
Allows comparison of safety profile of drugs
belonging to the same therapeutic group
Evaluate signals generated by other systems
or databases

13. 1. Not all green forms are returned
2. PEM depends upon reporting by doctors.
Underreporting is possible
3. PEM is currently restricted to general
practice
4. Its not known whether the patient took the
dispensed medication
5. Detection of rare ADRs is not always
possible

14. 1. Searching for signal
2. Assessment of important AE
3. Medically important events
4. Reason for stopping the drug
5. Analysis of events during the study while on
drugs
6. Ranking of ID and reason for withdrawal
7. Automated signal generation
8. Long latency adverse reactions

15. 9. Comparison with external data
10. Outcomes of pregnancy
11. Studies to examine hypothesis generated by
other methods
12. Studies of background effects and diseases

16.  A study was carried out to assess the sedation
properties of 4 anti-histaminic in the market
loratadine, cetrizine, fexofenadine and
acrivastine
 Objectives: To investigate the frequency with
which sedation was reported in post
marketing surveillance studies of four second
generation
antihistamines:
loratadine,
cetrizine, fexofenadine, and acrivastine
 Design: Prescription event monitoring
studies.

17.  Setting: Prescriptions were obtained for each
cohort in the immediate post marketing
period.
 Subjects: Event data were obtained for a
total of 43,363 patients.
 Main outcome measure: Reporting of
sedation or drowsiness.
 Results: The odds ratios for the incidence of
sedation were 0.63 (95% confidence interval
0.36 to 1.11; P = 0.1) for fexofenadine; 2.79
(1.69 to 4.58; P < 0.0001) for acrivastine, and
3.53 (2.07 to 5.42; P < 0.0001) for cetrizine
compared with loratadine. No increased risk
of accident or injury was evident with any of
the four drugs.

18. Incidence density of ADRs in first moth of
treatment with 4 anti-histaminics
Incidence density of events related to sedation in first
month of treatment with 4 anti-histaminics

20.  Record linkage is the process of bringing
together two or more records relating to the
same individual (person), family or entity (e.g.
event, object, geography, business etc).
 It is the process of assembling the outcomes of
drug exposure into a single database

21.  Record linkage can be considered as part of
the data cleaning process
 Provides rapid access to records of thousands
of patients and thus reduces the time
required for exploring the relationship
between drug exposure and outcomes

23.  An ideal database would include records from
inpatient, outpatient, emergency care,
mental health care, laboratory and
radiological tests, prescribed and over-thecounter medications as well as alternative
therapies
 All the parts should be easily linked by a
unique patient identifier
 It should be updated regularly

24. Researchers and the community‘s demand for detailed statistical information
In response to increasing business and health needs.
Improving data quality and timeliness
In reducing the complexity of data
Reducing respondent burden and costs
International collaborative
works

25.  The objective of the linking process is to
determine whether two or more records refer
to the same person, object or event

26. Probabilistic
Types of
Record
Linkage
Strategies
Deterministic

27.  A pair of records is said to be a link if the two
records agree exactly on each element within
a collection of identifiers called the match
key.
 For example, when comparing two records on
last name, street name, year of birth, and
street number, the pair of records is deemed
to be a link only if the names agree on all
characters, the years of birth are the same,
and the street numbers are identical.

28.  Pairs of records are classified as links,
possible links, or non-links.
 Here, we consider the probability of a match
in the given observed data.
 In probability matching, a threshold of
likelihood is set (which can be varied in
different circumstances) above which a pair
of records is accepted as a match, relating to
the same person, and below which the match
is rejected

30.  Patient goes to pharmacy
drug gets
dispensed
pharmacy bills the insurance
carrier for cost of that medication
 Should specify which drug was dispensed,
amount dispensed, etc.
 Patient goes to hospital/physician for medical
care
bills the insurance carrier for cost
of the medical care
 Should justify the bill with diagnosis
 Common patient identification no:
link
pharmacy and medical care claims

31.  Recent development with increased use of
computerization in medical care
 Computers are used to record medical
information

32.  Provide
large sample size, esp. for
pharmacoepidemiological studies
 Inexpensive
 Data will be complete
 Population based
 Include information on outpatient drugs and
diseases
 Avoid recall and interviewer bias

33.  Uncertainty of diagnosis data
 May not contain information regarding
smoking, alcohol, date of menopause, etc.
 May not contain data of medications
obtained without prescription or outside
insurance carriers prescription plan
 Instability of population due to job changes,
changes in insurance plans, etc.
 Include illnesses severe enough to come to
medical attention

34. 1. Data Quality
2. Bias
3. Coverage
4. Tracing Tool
5. Benchmarking/Calibration
6. Building New Data Sources (e.g., Registries)
7. Creation of patient-oriented, rather than
event-oriented statistics
8. Reducing costs and respondent burden

35. Thank you!