SOURCES OF

PHARMACOEPIDEMIOLOGY DATA
lecture 5

Dr. Rania Hussein

Fellow, Alexandria University Hospitals
Doctor of Public Health Sciences, HIPH
Pharm D
CPHQ
Learning objectives

● Data sources in pharmacoepidemiology.

● Methods for drug exposure measurement in
Pharmacoepidemiology.
● Guidelines for a good pharmacoepidemiology
practice (GPP).
 pharmacoepidemiologic methods

The science that applies epidemiologic methods to the area of

clinical pharmacology to study the use, effectiveness, and safety
of pharmaceuticals.
Epidemiologic methods include:
● Study designs.
● Sources of bias and confounding and techniques to reduce
them.
● Data sources.
● Measurement of medication use and outcomes.
 Types of data collected
Primary and secondary data
Primary data is collected directly from the original source
(patients and healthcare providers).
e.g., Follow a cohort of patients and collect specific research-
quality data on these patients. However, this is expensive,
takes longer time, affected by biases such as selection bias.
Secondary data
● Data collected for another purpose.
● Automated databases containing healthcare data from
routine clinical practice are the paradigm of the sources of
data collection.
● Data sources should be based on International Health
Terminology Standards.
● In pharmacoepidemiologic studies, the value of a database
must be evaluated by accuracy, dependability, completeness,
legibility, timeliness, accessibility, granularity, the size of the
data source (population coverage), and cost.
 Secondary data
Secondary data are easily accessible, lower cost, large number of
individuals, and limitations of biases such as recall biases, interviewer
biases, and reporting biases.
Medico-administrative databases are increasingly used for
pharmacoepidemiology studies. These databases offer preregistered data.
Examples include electronic health record databases, administrative
databases used for billing and payment, third-party payers health insurance
companies databases (e.g., Medicaid), electronic medical records, patient
registries, and health surveys.
These databases can be used to evaluate drug use patterns.
 Secondary data
1. Electronic health records (EHRs): The most commonly used data source in
pharmacoepidemiologic studies. This refers to electronic versions of a patient’s
medical history and health-related information. They can provide a wealth of
information on medication prescribed, diagnoses, allergies, lab results, radiology
images, and health outcomes.
2. Practice setting data : Clinical data (collected in healthcare settings)
– Prescribing data (outpatient and inpatient prescription forms) – It provides information
on patient demography, diagnosis, drug name, dosage form, strength, dose, frequency of
administration, and duration of treatment. The prescriptions are a good source for
evaluation of WHO indicators of drug use, which include the following: average number
of drugs per prescription, % of prescriptions with an antibiotic, % of prescriptions with an
injection, % of drugs prescribed from the essential drugs list, average drug cost per
prescription. It can be used to identify patterns of medication use and potential drug
interactions. However, in patients with poor adherence to treatment, such information
would not be equivalent to the medications taken by the patient.
 Secondary data

– Pharmacy dispensing data: Number of drugs per prescription, Drug names

with dosing schedule, Quantity of drugs dispensed, Cost of each drug,
Dispensing of OTC medications.
they confirm that the patient has acquired the medication, without
guaranteeing that the medication has been taken. Therefore, it is a vague
indicator of an ambulatory patient’s exposure to a drug.
3. Disease registries: systematically collected data on specific populations
or conditions that can be used to study the effects of drugs on these
populations.
 Secondary data
4. Pharmacovigilance databases: These contain information on suspected
ADRs, suspected drugs, and patient outcomes, which are collected from
a variety of sources, including healthcare providers, national authorities,
pharmaceutical companies, medical literature, and directly from patients.
Spontaneous reporting is the most common method used in
pharmacovigilance to generate signals on new or rare adverse events not
discovered during clinical trials.
It comprises the primary data source upon which post-marketing
surveillance depends, combined with data from formal clinical studies and
from medical and scientific literature.
It is very useful in hypothesis generation, with the need to explore possible
explanations for the adverse event in question.
Secondary data
5. National health surveys and community studies (field data):
Large-scale surveys that collect information on the health status,
to assess compliance to the treatment regimen and medication
use of individuals in a population (real-world environment).
6. Data from drug regulatory authorities: It provides information
on approved and banned drugs.
 Secondary data
7. Economic assessment: This calculates the costs of medical care,
including costs of treating ADRs (such as length of stay), and the
economic consequences of the benefits of a drug.
8. Patient-generated health data: Due to the digitization of the
population with wearable devices and mobile apps. Data can be
transmitted to clinicians and researchers. Some examples of these
data are glucose blood levels, heart rate, stress level, time and type
of physical activity, and hours and quality of sleep per day.
9. Social media: Recent evidence has shown that data from social
networks such as Facebook or Twitter provide useful information
for drug safety analysis.
Insurance claims databases: Records of
billing for all services covered (Inpatient,
Outpatient).
These contain information on medical claims
and can be used to identify patterns of
medication use and adverse events.
• Care needs to be taken because data
were not collected for the purposes of
research.
• Important information may be missing.
Studies in such databases may not reflect
the breadth of exposure in the target
population.
 Biases in studies using Secondary data
With the use of secondary databases, biases have to be taken into
account, such as the immortal-time bias and the immeasurable time bias.
The immortal time bias when a period of time during which a subject
cannot experience the event is incorrectly included in the analysis.
The immeasurable time bias refers to a period of time where certain
information, interventions, or health events are not measured or recorded
in the dataset.
 Methods for drug exposure measurement in Pharmacoepidemiology

Drug exposure measurements are issued from studies,

drugs sales, or medico-administrative databases.
Number of prescriptions is another measure frequently
used.
Biomarker measurements is considered the most objective
for measuring drug exposure in pharmacoepidemiology
(measuring drug level in blood or urine).
 DDD / PDD
The DDD system has been developed by WHO in order to measure and
compare drug use at an international level, across countries, regions, and
health facilities.
DDD “the assumed average maintenance dose per day for a drug used for its
main indication in adults.” to compare drug consumption between countries
or regions without being affected by local dosing variations.
PDD is defined as the “average daily dose prescribed, as obtained from a
representative sample of prescriptions.”
Only one DDD is assigned per ATC code.
PDD can vary depending on the type of disease, severity of disease, and
country of study, body weight, interethnic differences in drug metabolism,
and the prescribing culture of the health provider, and even between or
health facilities within the same country.
 Specific measures for drug exposure:

Indicators of compliance (adherence and persistence)

Adherence refers to whether a patient takes a
prescribed drug.
Persistence indicates if a patient stays on therapy.
 Indicators of compliance (adherence and persistence)
The medication possession ratio (MPR): it is the proportion of the number
of days of medication supplied during a specific time period to the number of days
the patient should have been on the medication.
In other words: MPR is the ratio of the number of days a patient is stocked for their
medication to the number of days a patient should be stocked for their medication.
˃1 =1 ˂1

It indicates over adherence, It indicates perfect It indicates under adherence,

which may suggest early adherence, meaning the suggesting the patient did
refills. patient has enough not have medication for
medication to cover every every day in the period,
day of the observation leading to gaps in therapy.
period.

Refill interval is the time period between when a patient fills one prescription and
the next refill of the same medication. It is another indicator that can help identify
the pattern of medication use: adherence, overuse, underuse.
example
Suppose a patient is prescribed a medication for 60 days. If they pick up refills twice
during a 60-day observation period, the MPR would be calculated as follows:
Total Days' Supply of Medication = 30 days (first supply) + 30 days (second supply) = 60
days
Total Number of Days in the Observation Period = 60 days

This indicates that the patient had medication available for the entire 60-day period,
and therefore their MPR is 1.0 (or 100% adherence).
However, if the patient only picked up the medication for 45 days during the same
observation period, the MPR would be:

This would indicate that the patient had medication available for 75% of the
observation period, suggesting non-adherence.
 Specific measures for drug exposure: indicators of compliance

MPR PDC

It looks at the total days' supply of Measures the number of days during the
medication versus the total number of observation period when the patient had
days in the observation period. enough medication available to cover their
prescribed regimen.
● Scenario 1:
The patient receives their 30-day supply at the start of the month and takes the
medication consistently throughout the month without missing any doses.
Proportion of Days Covered (PDC): the patient has access to the medication for all
30 days of the month. PDC = 30/30=1 (or 100%)
Medication Possession Ratio (MPR):The patient receives a 30-day supply, so they
have a 30-day supply for the month. MPR = 30/30=1 (or 100%).
Both PDC and MPR would be 100%, indicating perfect adherence.
● Scenario 2:
The same patient receives the 30-day supply at the start of the month but does
not use the medication every day, missing 5 days during the month.
PDC: The patient had access to the medication for the full 30 days, but they only
used it for 25 days. PDC = 25/30=0.83 (or 83.3%)
MPR: The patient still has a 30-day supply for the month (even though they missed
doses), meaning they possess enough medication to cover all 30 days.
MPR = 30/30=1 (or 100%)
● Scenario 3:
The patient has the 30-day supply but runs out of medication after
20 days and does not get a refill for another 10 days, so they miss 10
days of treatment.
PDC: The patient had access to medication for 20 days but missed
10 days (due to running out of medication and not getting a refill).
PDC = 20/30=0.67 (or 66.7%)
MPR: The patient received the full 30-day supply at the beginning,
so the total amount dispensed is 30 days' worth.
MPR = 30/30=1 (or 100%)
 Indicators of persistence

The discontinuation of the drug : time from when the patient starts
the medication until he stops it; a longer duration before
discontinuation indicates better persistence.
Switching: it is defined as a delivery of a different drug within the
same therapeutic class or for the same condition. If a patient
switches to another drug and continues treatment without a
significant gap, he is still be considered persistent with the therapy.
Medication gaps: tracks time between refills to identify gaps that
suggest non persistence.
 Indicators of misuse

Indicators of misuse or drug abuse have been developed in order to identify

the relative abuse liability of several psychoactive drugs in real-life setting:
The doctor shopping quantity (DSQ): defined by the quantity obtained by
overlapping prescriptions from several prescribers.
The doctor shopping indicator (DSI): it measures the proportion of the drug
obtained by doctor shopping (DSQ) among the overall quantity of the drug
reimbursed or dispensed in a given setting.
A signal of abuse by doctor-shopping is considered meaningful when the DSI
is superior to 1%. Below this value, one considers that there is no clear signal
of abuse.
In one country:
Total dispensed of Oxycodone = 10,000 tablets
Tablets obtained through doctor shopping = 500
500
𝐷𝑆𝐼 = × 100 = 5%
10000
This means 5% of the dispensed oxycodone may be linked
to doctor shopping.
 Guidelines for good pharmacoepidemiology practice (GPP)
The ISPE Guidelines for a
Good Pharmacoepidemiology
Practice (GPP) are intended to
assist investigators with Essential
practices and procedures
pertaining to the planning,
conduct, interpretation, and
documentation of
pharmacoepidemiologic research
to help ensure its quality and
integrity.
Steps of a GPP
 Research question

The research question may be phrased by

using the PICOT template:
Population : the relevant people in
relation to the clinical problem.
Intervention: the management strategy,
exposure, or test.
Comparator: an alternative or control
strategy, exposure or test for comparison.
Outcome: the result.
Timing: time frame.
 Research question
At a routine immunization visit, Lisa, the mother
of a six-month-old, tells you that her baby
suffered a nasty local reaction after her previous
immunization. Lisa is very concerned that the
same thing may happen again this time.
Recently, a colleague told you that needle
length can affect local reactions to
immunization in young children but you cannot
remember the study details.
Develop a clinical research question using PICO
to help you find the information you need.
 Research question
Susan is expecting her baby in 2 months. She
has been reading about the potential benefits
and harms of giving newborn vitamin K
injections. She is alarmed by reports that
vitamin K injections in newborn may cause
childhood leukemia. She asks you if this is true
and what is the risk for her baby.
 How to write a research protocol

A protocol should be drafted as

one of the first steps in any
research project, and it should be
amended or updated as needed
throughout the course of the
study.
the protocol should include the
following elements:
How to write a research protocol

The title:
● The title should be descriptive and concise. It
should include the study design, drug or device
name, the main objective, study population.
● The names, titles, degrees, addresses, and
affiliations of all responsible parties should be
written in the protocol title page.
 introduction
Background

● Summarize the published literature that supports the research idea.

● Identify the gaps in existing literature that the study is intended to fill.
● Justify why the research would create valuable and useful knowledge, and the
potential impact of the research findings.
● Additionally, this section should be written in a way that a person who is not an
expert in the field can understand.

Previous studies have demonstrated the safety and efficacy of drug A for outcome B in
population C. This study investigates the efficacy of drug A for the treatment of outcome B
in population D.

Purpose (objectives)
Describe the specific aims and objectives of the study.
Aim of the study

 General objectives are broad statements of what the protocol hopes to accomplish.
They are the same statement as the title.
 All specific objectives should meet the SMART Criteria:
▪ Specific: each objective has a single key result.
▪ Measurable: each objective relates to one or more parameter(s) that can be
measured.
▪ Attainable: each objective is realistic in terms of the available resources.
▪ Relevant: each objective is linked to general objective.
▪ Timely: each objective should be able to be accomplished within the time frame
established for the study.
 Plan of the study
 Study setting
 Study design
 Study population
The population is defined in terms of person, place, time, and selection criteria, in addition to
Sample size (number of subjects expected to participate).
Demographics of subjects
Describe the projected distribution of gender, age range, race, and ethnicity of subjects.
Justification for the inclusion or exclusion criteria should be included.
Inclusion criteria: Define subject eligibility based on scientific rationale and safety considerations.
➢ Subjects between the ages of X-Y
➢ Subjects with a diagnosis of outcome B
Exclusion criteria: Define exclusion criteria to limit the study population and prevent subjects
from being exposed to excess risk by the study; certain vulnerable populations.
➢ Subjects with co-morbidity E
➢ Pregnant, nursing, or child-bearing women
 Plan of the study

 Sampling design
 Data collection methods and tools
 Ethical considerations
Ethical framework for conducting research
 Ethical principles
Guidelines on the general conduct of research on
human beings to ensure that interventions do not
have serious unintended or harmful consequences.
All epidemiological studies involving human subjects
must be approved and reviewed by ethical review
committees as an Institutional Review Board (IRB),
Independent Ethics Committee (IEC), or other
appropriate body.
Studies using commercially or publicly available de-
identified secondary data sources, or which meet
certain other criteria, are not considered research
involving human subjects in some countries and may
be exempt from IRB review.
 Informed consent

Informed consent: it must be

obtained from participants, and they
must retain the right to withdraw at any
time.
The patient has the right to understand
three important aspects: nature (i.e.,
what), purpose (i.e., why), and
consequence.
 Ethical principles
The guidelines published by the European Epidemiology Federation were structured
by four generally recognized ethical principles:
● Respect for Autonomy: autonomy is an individual patient’s right to self-determination in
making decisions for himself, to choose, and the right to know about the personal
consequences of joining a study.
● Beneficence (Do good): Participants in research should be treated well. The research
should aim at producing results beneficial for humankind (physical and psychological
benefits).
● Non-maleficence (Do no harm): Participants in research should not be subject to
unjustified or avoidable burdens. Personal integrity should not be harmed.
● Justice and veracity: To ensure that all patients have the same opportunity for good
health. The same ethical standards apply to every subject and in every country. It is
unacceptable to export risky research activities to disadvantaged countries and to carry
out hazardous or burdensome research activities with vulnerable individuals to the benefit
of others.
● Confidentiality: preservation of confidentiality of
information obtained through the study.
● Scientific integrity: the adherence to ethical and
professional standards in the conduct of scientific
research.
 Plan of the study
➢ Statistical analysis
Specify statistical and analytical methods (statistical power, sample size, significance level,
etc.).
This study will use a t-test, with the null hypothesis set at the standard significance level (p–value <
0.05), to determine if drug A significantly reduced mortality due to outcome B. A sample size of X was
calculated based on an assumption of 80% power.
➢ Data storage, security, and confidentiality
Describe where data will be stored and how data will be secured during the study,
confidentiality measures. Indicate who will have access to the data and how the data will be
used, and describe to whom and for what purpose the data will be released.
Data will be stored on a password-protected computer and maintained for 7 years after
study completion. Access to subject data will be limited to the Principal Investigator, Sub-
Investigators, and study staff. Data will be only used for research purposes.
➢ Data Sharing
Describe plans for data sharing, particularly for publication purposes.
➢ Bibliographic references