Assignement introduction Strategies for Automating Pharmacovigilance Adverse Event .pdf

1

Section 1: Foundation of the Study

The goal of pharmacovigilance (PV) is to protect patients from unnecessary harm.

When a medicine is marketed, the identification and monitoring of adverse events (AEs)

and reactions is a part of a post-marketing safety surveillance program as required by

federal regulations (Kothari, Shah, & Patel, 2018; Tan et al., 2016). Health regulatory

agencies have regulations to ensure that drug products are safe for human consumption.

Companies that are licensed marketing authorization holders are required by regulation to

can adopt an automation-enabled digital PV path for drug safety surveillance and redirect

the savings from PV operations toward investing in the actual PV tasks like benefit-risk

assessments of products, thereby making the products more efficacious and safer for

human use.

Background of the Problem

The safety of medicines is a global responsibility. A drug product or medicine

will gain marketing authorization after a regulatory agency conducts comprehensive

reviews of the drug product to evaluate the benefits and risks (Schurer, Bam, & de Kock,

2017). When the officials of the agency find a favorable benefit-risk balance, they

2

approve the drug product to be marketed within the jurisdiction of the relevant agency

(Kothari et al., 2018). When the product is marketed and consumed by a large population,

adverse reactions or side effects may occur (Pitts, Louet, Moride, & Conti, 2016). When

a product is reported to have caused serious AEs, national authorities analyze the reports

of the product and weigh the risks and benefits. They then decide on a course of action,

which may include issuing an alert or, in extreme cases, taking the product off the market

(McNaughton, Huet, & Shakir, 2014).

hospitals, contributing to 2.6 million deaths annually due to unsafe care (World Health

Organization, 2019c).

Biomedical informatics including artificial intelligence (AI) and robotics-based

solutions can accelerate efficiencies and can be used to identify and process AEs faster

and contribute to reduction of costs (Beninger, 2018; Donzanti, 2018). Despite multiple

successful experiments, not many managers in pharmaceutical companies have developed

and/or implemented innovative technology to realize the benefits.

3

Problem Statement

Business leaders who fail to implement innovative technology solutions in their

companies face economic distress in these organizations (Bal & Erkan, 2019). Failure to

implement innovative technology solutions result in high pharmacovigilance (PV)

operational costs related to the 2.2 million adverse events (AEs) documented by the U.S.

Food & Drug Administration (2019) that could cost pharmaceutical companies $900

million for drug recalls (Hall, Stewart, Chang, & Freeman, 2016). The general business

PV systems managers in the pharmaceutical industry use to implement innovative

technology solutions to automate AE case processing. The target population consisted of

four PV systems managers working in a pharmaceutical company located in the Boston

area of Massachusetts, United States, who had used successful strategies to implement

innovative technology solutions to automate AE case processing. The implications for

positive social change include the potential to improve patient safety by identifying

successful strategies that PV systems managers can use or adapt to implement innovative

technology solutions to automate AE case processing.

4

Nature of the Study

Qualitative studies are based on interpretive philosophy. Using an inductive

approach, researchers use a smaller sample of participants’ data to develop rich data and

thick descriptions through a variety of methods for data collection to identify patterns

associated with phenomena (Fusch, Fusch, & Ness, 2018; Saunders, Lewis, & Thornhill,

2015). In quantitative and mixed methods studies, researchers use quantitative measures

and inferential statistics to examine relationships among variables (Saunders et al., 2015).

Qualitative case studies are flexible in nature, focusing on a phenomenon, and

evolve over the course of the study (Yin, 2018). Using a case study strategy enables

researchers to perform in-depth inquiry into a topic to generate insights in a real-life context

5

design was more appropriate for this study than a phenomenological or ethnographic design.

Research Question

The primary research question for this study was: What strategies do PV systems

managers use to implement innovative technology solutions to automate AE case

processing?

Interview Questions

1. What strategies did you use to identify automation solutions for AE case

processing?

2. What strategies did you use for implementing the identified automation solution?

3. What, if any, kinds of operational changes did the automation solution require to

6. How did your organization overcome the key difficulties to implementing

automation technology solution?

7. How did your organization assess the implementation strategies’ effectiveness?

6

8. What additional information can you add about the development and

implementation of the strategies your organization used for automating AE case

processing?

The conceptual framework that I selected for my doctoral study was based on the

information systems theory’s task-technology fit (TTF) model. The TTF model focuses

on the degree to which systems characteristics match user task needs. Goodhue (1995)

researchers have improved the TTF model by incorporating additional characteristics and

constructs or by integrating other theories such as the appropriation theories (Park, 2019).

I compared the expected results from the TTF model with the actual results from my

study to understand the study’s findings.

Adverse events or reactions: Adverse events or reactions are unexpected harm

caused by the normal use of medication at the normal dosage (Karimi, Wang, Metke-

Jimenez, Gaire, & Paris, 2015).

7

Adverse event reporting: Adverse event reporting refers to the activities that

marketing authorization holders perform to collect, and report AE cases reported by

patients as required by a 2012 PV legislation (Kothari et al., 2018).

Digitalization: Digitalization involves using digitized data to make decisions and

generate change (Gobble, 2018).

Healthcare information technology: Healthcare information technology is the

application of computers for the storage, retrieval, sharing, and use of health care

Assumptions

Assumptions are unverified beliefs pertaining to a study (Locke, Spirduso, &

Silverman, 2014). I made the following assumptions in this study: (a) the participants

provided honest, truthful responses to the interview questions, (b) the participants had

relevant information and experiences, (c) the responses provided by the participants

assisted in understanding the strategies used to implement innovative technology

solutions to automate AE case processing, and (d) the strategies provided by the

participants improved efficiencies in processing AEs. Following the conclusion of my

interviews with the participants, I found all these assumptions to be true.

8

Limitations

Limitations are potential weaknesses which are beyond the researcher’s control

(Singh, 2015). One of the limitations for this study was that the participants were

employed in one pharmaceutical company. Restricting the study population to one

company may limit the transferability of the methods used in my study for future

research. Another limitation was that the member checking was conducted by emails

since most of the follow-ups occurred during the actual interviews.

eligible, were not a part of this study.

PV is a study of the safety and effectiveness of drugs including the activities

relating to detection, assessment, understanding, and prevention of the AEs associated

with the use of drugs (Beninger & Ibara, 2016; World Health Organization, 2019b).

Processing of an AE is an expensive resource-intensive process, which includes high risk

of errors and operational inefficiencies (Ghosh et al., 2020). In 2019, the U.S. Food &

Drug Administration (2019) received 2.2 million AEs which is a 450% increase since

2009, which increased the derivative AE-related operational costs for pharmaceutical

9

companies. Biomedical informatics including AI and robotics-based solutions can

accelerate efficiencies, identify and process AEs faster, and contribute to costs’ reduction

(Beninger, 2018; Donzanti, 2018). I explored the strategies that PV systems managers in

the pharmaceutical industry use to implement innovative technology solutions to

automate AE processing.

Contribution to Business Practice

Information technology (IT) offers a wide range of opportunities to organizations

be effective in detecting what may be otherwise invisible to the human eye, increasing

productivity, and more likely to identify important public health information within large

data bases (Lu, 2009). Data mining, machine learning, AI, and robotics could result in

new operating models for PV (Beninger & Ibara, 2016; Lai, 2017), improve efficiencies

and reduce PV operations costs (Karimi et al., 2015; Lu, 2009; Schmider et al., 2018),

better allocation of resources, and improve patient outcomes (Lewis & McCallum, 2019).

Implications for Social Change

The implications for positive social change include the potential to identify

strategies to assist in making pharmaceutical medicines safer for human consumption. PV

offers a stronger model because the variable constructs overlap in a certain way (Sombat

et al., 2018). Dishaw and Strong (1999) devised the integrated model based on study data

from questionnaire responses from three established firms from financial services,

aerospace manufacturing, and insurance industries. The questionnaire contained items for

each TAM variable: intention to use, attitude toward use, perceived ease of use, and

perceived usefulness (Dishaw & Strong, 1999). The integrated TAM/TTF model

combines an attribute/behavior model (TAM) with models of task-technology fit with a

systems tend to self-organize to adapt to continuing challenges to overcome chaos

(Faggini & Parziale, 2016). As a contrasting theory for this study, chaos theory can assist

in understanding of the unpredictability of a system’s behavior. To demonstrate the chaos

theory, Lorenz (1963) used the butterfly effect, which is a phenomenon that generates

small changes in the input but distributes dramatic consequences in the output. The chaos

theory helps researchers understand how complex systems work and what causes

ambiguity regarding the behavior of the dynamic systems.

In the perspective of technological innovation, Li and Wang (2019) found that

chaos of the system is visible when the technological content is larger. Therefore, Li and

Wang (2019) summarized that companies use optimal technological contents to obtain

maximum profits. Since the innovative technology solutions are a recent development for

PV, the pharmaceutical companies implement and operate with optimal solutions to

balance the profit margins with the stability of the market (Li & Wang, 2019) despite the

need of using technological innovation to create competitive advantage.

Based on the literature review of original models and integrated models for

technology implementation, as described in this paper, I chose the TTF model for my

needs technology solutions. Ali, Romero, Morrison, Hafeez, and Ancker (2018) applied

the TTF framework to evaluate how a patient portal helped improve the portal's fit to

patient needs. The study concluded that patients lack a clear understanding of tasks that

would help them accomplish personal health information management (Ali et al., 2018).

Because my research topic included new strategies PV systems managers use to

implement innovative technology, an unstructured inductive qualitative study based on

the constructs of the TTF model assisted in achieving my study objectives.

PV and Healthcare Information Technology

In this section, I will describe the findings of the integrative review to understand

the business of PV and IT in the pharmaceutical industry. This section includes PV-

related topics such as the function of PV, mandated AE reporting, and related challenges

like product recalls and AE detection and reporting. This section of the literature review

also includes the perspectives of technology on PV and AE reporting, digital PV,

automation, innovative technology implementation trends and risks, and implementing

product’s safety, efficacy, and quality to obtain a product license to market it (Lewis &

McCallum, 2019). A drug product or medicine will gain a marketing authorization after a

regulatory agency conducts comprehensive reviews of the drug product to evaluate the

benefits and risks (Schurer, Bam, & de Kock, 2017), finds a favorable benefit-risk

balance, and approves the drug product to be marketed within jurisdiction of the relevant

agency.

When a drug product is marketed, and the drugs are used by large populations,

unexpected adverse reactions or AEs may occur, which changes the risk-benefit ratio of

the drugs, requiring a regulatory action in post marketing surveillance (Pitts et al., 2016).

AE is unexpected harm caused by the normal use of medication at the normal dosage

(Karimi et al., 2015). No medicine is safe at the time of approval for sale or marketing

authorization (Pitts et al., 2016). Only when the benefit of taking the drug outweighs the

risks, consuming a drug can be justified (Kothari et al., 2018). Since clinical trials are

conducted in a controlled environment with a small set of human patients, the adverse

reactions and AEs collected are not representative of the large population that will

consume the drug. AE from consuming drugs could remain undetected, exposing

pharmaceutical industry. In the European Union, about 197,000 deaths are caused by AEs

per year and the total cost to society of AEs is €79 billion (Mohiuddin, 2018). AEs are

unexpected effects occurring from normal use of drugs taken in normal dosage (Basile et

al., 2019). In 2013, medical errors including AEs were identified as the third leading

cause of death in the United States (Makary & Daniel, 2016). Health authorities and

agencies have implemented multiple mechanisms to monitor the safety of the drug

products across the discovery, development, and marketed phases like the Center for

Drug Evaluation and Research (CDER), the New Drug Application (NDA), and Post-

marketing Surveillance (PMS) program to assess and monitor drug safety risks by the

U.S. FDA (Tan et al., 2016).

Unsafe healthcare causes harm and leads to other forms of collateral damage such

as loss of trust in the system and loss of reputation and credibility in health services

(World Health Organization, 2017). The occurrence of AE is one of the top 10 causes of

death and disability across the world (World Health Organization, 2019a). Officials of the

World Health Organization reported that 134 million adverse events occur each year in

2016; Liu, Zhao, & Zhang, 2016). A review of U.S. FDA recalls showed that AE was one

of the top three reasons for drug recalls, which could cost a large pharmaceutical

company $900 million (Hall et al., 2016). 644 products were withdrawn from the market

because of AEs between 1953 and 2013 (Onakpoya, Henegan, & Aronson, 2016). The

withdrawal of anti-inflammatory rofecoxib (Vioxx™) occurred only after millions of

patients had been exposed, highlighting the need for earlier, more comprehensive data

reporting (Bailey et al., 2016). Evidence from spontaneous reporting of AE and clinical

research studies are used to support the decisions for revoking marketing authorization

and withdrawal of products (McNaughton et al., 2014). Therefore, pharmaceutical

companies and MA holders must identify AEs at the earliest to avoid the expenses from

recalls due to AEs and assist with patient safety.

Sources of AE. MA holders are required to collect, and report AE cases reported

by patients by 2012 PV legislation (Kothari et al., 2018). Each national drug agency

collects its own AEs in a database, so do the MA holders. Companies use relational

databases to collect, store, and process AEs, which are linked to data warehouses that are

used for reporting AEs to the health authorities (Lewis & McCullum, 2019). Sources of

important source for AEs, which has increased recently despite challenges with reporting

systems and confusions with reporting (Inacio, Gomes, Airaksinen, & Cavaco, 2018;

Mohiuddin, 2018). Patients, manufacturers, and healthcare professionals report AEs

(Schurer et al, 2017) in variety of ways by phone calls, e-mails, and report directly to the

manufacturers and the health authorities or discuss on social media such as blogs and

social networking sites. Typical process of collecting, processing, and reporting AEs

within a MA holder company is shown in Figure 2.

Figure 2. Typical process for AEs case processing and reporting within a marketing

authorization holder. Adapted from “Utilizing advanced technologies to augment

2016; Pitts et al., 2016; Schurer et al., 2017). Traditional AE data sources such as

spontaneous reporting data, electronic health records, pharmaceutical databases, and

biomedical literature are limited by under-reporting ratio, privacy issues, high cost, or

long publication cycle (Yang & Yang, 2018). Despite being obligated to participate in the

PV process and spontaneous AEs reporting, pharmacists do not fully comply (Kopciuch

et al., 2019). Only 16% of pharmacists were found to be trained in PV in Poland

(Kopciuch et al., 2019). Nurses generally report AEs since doctors are not engaged and

do not report AEs, which causes bias and missed critical diagnosis (Mitchell et al., 2016).

Under-reporting is a result of non-standard reporting process, inadequate knowledge of

reporting, negligence, time constraints, and gaps in safety information systems

(Mohiuddin, 2018). The median under-reporting rate was 94% (Lexchin, 2015). Only 5-

10% of AEs are reported and about 95% of health care professionals do not report AEs

(Mohiuddin, 2018). Under-reporting of AEs could cause delays in making decisions

reduce operational costs.

The information about the type and variety of AE data elements collected within

AE reporting systems varies internationally. Timely communication of AEs is a challenge

though the AEs are recognized sooner (Lexchin, 2015). Bailey et al. (2016) identified 108

reporting systems, containing 1,782 unique data fields mapped to 33 reporting concepts.

A minimum required dataset for a valid AE case report included an identifiable patient,

an AE, a suspect medicinal product, and an identifiable reporter. However, the absence of

standardized AE data collection and reporting data elements and variations across

systems rendered aggregate reporting a challenge, which directly affected safety

surveillance (Bailey et al., 2016; Schurer et al., 2017).

With the increase in awareness about AE reporting, regulatory agencies have

improved their safety surveillance guidance for the MA holders (Stergiopoulos, Fehrle,

Caubel, Tan, & Jebson, 2019). Since the spontaneous reporting of AEs is biased with

challenges, PV function leaders from the industry and the regulatory agencies are

exploring other sources and advanced computational methods to supplement existing

considered a viable source of AE for PV. The user posts on social media contain

colloquial language and misspellings, which disrupts the balance between sensitivity and

specificity in identifying AEs (Raschi et al., 2019).

The current process of collecting and processing AE in the pharmaceutical

industry is manual, complex, and poses high risk of errors. The challenges on the

AEs, globally diverse and changing regulations, complex and evolving systems, and

sparse PV expertise and talent (Price, 2018). In 2019, the U.S. Food & Drug

Administration received 2.2 million AEs (“FDA adverse event”, 2019), which is a 450%

increase since 2009 and increased AE-related operational costs for pharmaceutical

companies. Pharmaceutical companies must invest in PV to avoid risks and frequently

manage PV as a cost-center. Processing of an AE is an expensive resource-intensive

process, which includes high risk of errors and operational inefficiencies (Ghosh et al.,

2020). The increasing costs of PV have led to outsourcing of PV activities to service

providers operating in low-cost locations (Price, 2018). A continuous, digitized,

automated data management within a computerized system that leverages AI, machine

contemplated in the algorithms, and that contributes enormously to the establishment of

the causal relationship between a product and AE. However, Marante (2018) concluded

that the clinical judgment method poses difficulties in standardization of evaluation; and

therefore, a need for a universally accepted high-quality assessment tool.

Recently, innovative technologies are impacting the way companies process and

evaluate AEs. Beninger and Ibara (2016) and Donzanti (2018) summarized that the

rapidly developing innovations in informatics pose challenges for PV, however, offer

opportunities for close collaboration. Innovative technologies like robotics and AI have

permeated the pharmaceutical industry and assist in enhancing the drug discovery process

Digital PV. Digitalization involves using digitized data to make decisions and

generate change (Gobble, 2018). Gobble (2018) further explained that the digitalization

process leads to digital transformation, which is the re-structuring and re-defining of the

business organizations considering opportunities from emerging technology. Digitization

and innovations like Big Data and analytics have led to new operating models in multiple

functions within healthcare industry, but, not PV (Beninger & Ibara, 2016). Karimi et al.

(2015) summarized that drug safety had not been comprehensively surveyed from the

overlap between the two despite the increase in awareness of innovative technology in

PV. Beninger and Ibara (2016) presented a detailed analysis of the PV adverse event case

processing process and the corresponding technology used starting from relational

databases and faxes to the recent technology like Blockchain and artificial neural

networks. Beninger and Ibara (2016) identified advanced technology like AI for PV is

emerging and growing and concluded with an emphasis for implementing informatics

and developing structures to facilitate an approach to adopt informatics. Prior to

deploying automation, organizations must embark on a business process re-engineering

journey (Mishra, Devi, & Narayanan, 2019).

In 2014, the FDA received 528,192 new AE reports indicating a serious or fatal

outcome, 4.7% directly from health professionals and consumers, and 95.3% from drug

manufacturers (Moore, Furberg, Mattison, & Cohen, 2016). Majority of the AEs received

by the FDA are from the manufacturers and MA holders since most AEs are reported

directly to the manufacturers (Stergiopoulos, Brown, Felix, Grampp, & Getz, 2016). A

study on AE quality and completeness by Moore et al. (2016) concluded that the report

completeness from drug manufacturers was poor compared with direct submissions.

full advantage of the rapidly developing electronic health records and technologies to

automate voluntary reporting of incidents (Mitchell et al., 2016).

A study by Ribeiro, Motta, Marcondes-Fonseca, Kalil-Filho, and Giavina-Bianchi

(2018) revealed that in a hospital setting, there were a total of 94 adverse drug reactions

in 75 patients and most reactions were predictable and of moderate severity. For each

medication introduced during hospitalization, there was a 10% increase in the rate of AEs

with the probability of observing an AE was 1 in 104 patients per day. 134 million AEs

occur each year in hospitals, contributing to 2.6 million deaths annually due to unsafe

care (World Health Organization, 2019c).

Proactive prevention of risks can be achieved using AI for analysis of data from

post marketing studies, electronic medical records, and the Internet among other trends

(Fermont, 2019). Organizations must understand and identify the opportunities and

implications of implementing innovative technology to add value at a conceptual level

(Lewis & McCallum, 2019) as shown in Figure 3. Despite multiple successful

experiments with using innovative technologies to improve performance and efficiency,

the primary challenges are transparency and traceability (Spasic, Uzuner, & Zhou, 2020).

Figure 3. PV systems automation: Conceptual diagram showing the different entities

relating to PV systems automation. Adapted from “Utilizing advanced technologies to

augment pharmacovigilance systems: Challenges and opportunities,” by D. J. Lewis and

J. F. McCallum, 2019. Therapeutic Innovation & Regulatory Science, p. 2.

Leaders must adopt a systems thinking approach and focus on systems, processes,

and measures to build a learning organization which is both reliable and resilient

(Gossett, Padgett, Pierce, & Scott, 2019). Emerging technology and intelligent

automation in the PV business systems function is complex, and are not readily

understandable by the PV business professionals who are trying to implement them

(Lewis & McCallum, 2019). Eseryel (2019) summarized three strategies for non-IT

managers to achieve digital business transformation: create and procure endorsement for

and improved engagement of patients and healthcare professionals. Streefland (2018)

highly recommended the PV system to focus on the patient rather than on regulations and

recognized a need to work toward optimal analysis of real-time medicine use insights,

instead of AE creation, reporting, and adherence to procedures.

Innovative technology trends in PV. Mesko (2017) recognized that medicine

requires a variety of disruptive technologies to be implemented into developing

treatments, practicing medicine, and delivering care. Recent developments in the

innovative technologies include text analytics, robotic process automations, cognitive

technologies including machine learning and deep learning. Since rule-based

technologies are traceable, the applications are abundant. Despite multiple successful

experiments with using these innovative technologies to improve performance and

efficiency, the primary challenges are transparency and traceability (Spasic et al., 2020).

Ghosh et al (2020) considered the innovative automation technologies like natural

language processing (NLP), AI, machine learning, optical character recognition, and

robotic process automation (RPA) as “intelligent” since they are more advanced. Ghosh

et al. (2020) also predicted that the risk of failure or malfunction of these technologies

savings, reducing errors, and improving efficiency (Dey & Das, 2019). Carden,

Maldonado, Brace, and Myers (2019) listed accuracy, audit trail, consistency, flexibility,

location agnostic, low-risk and non-invasive technology, productivity, reliability,

availability, and productivity as the value proposition for RPA. RPA software robots

(bots) only work on the user interface, which are simple to develop and do not require

complex programming (Phillips & Collins, 2019). Phillips and Collins (2019) compared

bots, which follow a set of rules with AI-based solutions, which are neural networks

designed to replicate human brain process. Since bots are straight forward and rule-based,

they can be built to replace any repetitive and mundane tasks and applied in any industry.

RPA can be applied successfully for PV; however, the PV systems managers and the

business users must work closely to identify highest potential processes and identify the

most beneficial automation approach (Carden et al., 2019; Phillips & Collins, 2019).

Innovative technology must be used intelligently. For example, processes that involve

decision making, the RPA solutions must be blended with AI-based cognitive solutions

(Mishra et al., 2019)

Multiple researchers have identified a variety of applications for text analytics in

extraction of case information by training the machine learning algorithms (Schmider et

al., 2018). NLP can structure and customize automatically extracted AE information and

store in a way that healthcare professionals can specifically share AEs that are of the

individual patients’ interest (Kusch, Zien, Hachenberg, Haefeli, & Seidling, 2020). NLP

can be used to detect healthcare associated infections in hospital facilities and NLP chain

processing can facilitate a standardized AE detection process across multiple hospitals

(Tvardik et al., 2018).

Social media is a potential rich source of AEs. Considering the median under-

reporting rate of AE to be 94% in spontaneous reporting systems, social media and

crowd-sourced data can be potential sources to detect AE related to health products

including pharmaceuticals, medical devices, biologics and natural health products

(Tricco, Zarin, Lillie, Pham, & Straus, 2017). Despite the possibility of mining AE data

quicker from social media, poor quality of information makes it a challenge for social

media to be considered a viable source of AE (Convertino et al., 2018). Data from social

media is not rich because of the use of slangs and colloquial posts, which cause a

negative impact on the AE extraction program (Arnoux-Guenegou et al., 2019). In their

content. Nikfarjam, Sarker, O’Connor, Ginn, and Gonzalez (2015) devised a machine

learning-based approach that was scalable and suitable for social media mining, as it

relied on large volumes of unlabeled data, which removed the need for large, annotated

training data sets. Yang and Yang (2018) developed a framework to facilitate drug safety

signal detection by harnessing online health community data, a timely, informative, and

publicly available data source like MedHelp, which collected patient-contributed content.

Advanced NLP techniques like feature-based and kernel-based methods that can

effectively identify and separate AEs and non-AEs from the informal text in the social

media (Liu et al., 2016).

Cognitive technologies have proven to successfully detect AEs in experimental

setting. Based on multiple pilot projects, Chen et al. (2016) summarized that IBM Watson

can assist the drug safety process through faster recognition and coding of AE from text.

Watson can be used to improve the efficiency of existing drug safety personnel and

support timely reporting of AEs to regulatory agencies.

Machine learning models pose risks when PV managers must justify the decisions

made by the system in the event of a health agency inspection or an audit. PV managers

safety issues (Wang, Coiera, Runciman, & Magrabi, 2017).

Negi, Pavuri, Patel, and Jain (2019) proposed a novel method of combining

machine learning and NLP technology to assist with automating AE and suspect drugs

extraction from medical reports. The first method tried to establish a causal relationship

between the drug and the medical condition. The second method classified a drug into a

suspect drug or a non-suspect drug using NLP features and machine learning for

classification. Deep learning-based models and multi-task learning-based models can

assist NLP in significantly improving the probability of identifying and extracting AEs

from electronic health records (Li, Liu, & Yu, 2018). Most methods use annotated

datasets to predict AEs, neural nets, attention mechanisms, and multitask learning are

popular choice for NLP applications for PV (Basile et al., 2019). Basile et al. (2019)

noted that the use of machine learning and deep learning are common in academic

research for PV but are not used in practice in the industry yet.

Blockchain technology stores data in a list of securely linked blocks. Blockchain

technology is currently widely used in the financial industry (Lewis & McCallum, 2019).

Blockchain can be adopted for PV where the PV system is contained within a single

into PV, thereby enabling predictive analytics and reasoning. However, due to the lack of

regulatory guidance and problems with system validation, implementing and adopting

dynamic intelligent systems is a challenge (Lewis & McCallum, 2019).

Strategies for implementing innovative technology in PV. Regulatory agencies

worldwide have improved and changed their safety monitoring processes, which has led

to increased reporting requirements (Stergiopoulos et al., 2019). Due to the changing

regulations for safety surveillance, MA holders in the pharmaceutical industry create new

internal processes or include new data sources to detect AEs (Stergiopoulos et al., 2019).

Multiple factors determine the successful implementation of innovative technology

including the approach and strategy, types of data, and change management. Alotaibi and

Federico (2017) presented that the same technology used in different companies produce

varied outcomes. Different kinds of data can be handled with different technology

solutions like machine learning for unstructured data (Karimi et al., 2015).

In their study focused on PV operational costs and drivers, Stergiopoulos et al.

(2019) reported that several companies out of the 12 that participated in the survey had

plans for including automation technology like machine learning. But none had

Beninger (2018) emphasized on the new skillsets, which will define the roles of the next

generation of professionals who perform PV activities. Mesko et al. (2018) identified

human resource crisis like accessibility and shortages in healthcare. Although Mesko et

al. (2018) did not directly associate the resource problems to PV in the article, the authors

concluded that the gaps in the human resource could be filled using advanced

technologies like AI. Mesko (2017) strongly recommended that data analysis solutions

must be implemented to support the skills of physicians and must not replace the

traditional physician–patient relationship. Organizations must expect to supplement the

semistructured interviews, documents and presentations, and the reflective journal to

facilitate triangulation.

I used semistructured interviews to understand the strategies that the PV managers

use to implement innovative technologies. Interviews provide a more complete picture of

the phenomenon by reconstructing and interpreting the past (Cypress, 2018). Good and

actionable questions are used to gather rich and deeper understanding of the issues and

events (Jonsen, Fendt, & Point, 2017). In order to regulate the interview procedure and

I used the Belmont Report to guide my ethical behavior during this study. The

Belmont Report focused on three ethical principles and their applications to conduct

research. Informed consent, risk/benefit assessment, and selection of subjects of research

requirements cover the principles-respect for persons, beneficence, and justice (National

Commission for the Protection of Human Subjects and Biomedical and Behavioral

Research, 1979). I adhered to the three principles by using confidentiality procedures to

protect the participants and maintain privacy. By explaining the study objective, the

interview process, the use of data collected, and obtaining informed consent, I mitigated

ethical issues as recommended by Cypress (2018). I will store the data collected in a

password-protected flash drive and duly discard after 5 years.

Participants

I purposively selected a pharmaceutical company in the Boston area of

Massachusetts, United States as the single case for this study. Qualitative studies that aim

to explain a phenomenon in a specific context tend to use a purposive sampling design

(Sovacool, Axsen, & Sorrell, 2018; Tobi & Kampen, 2017). The primary criteria for

participate in the study (Peticca-Harris, deGama, & Elias, 2016). I obtained authorization

from Walden University’s Institutional Review Board (IRB). Walden University’s

approval number for this study is 04-24-20-0972741 and it expires on April 23, 2021.

The number of participants depend on the characteristics of population from which they

are chosen (Saunders & Townsend, 2016). I identified eligible and qualified employees

for this study prior to inviting them to participate.

The relationship between the researcher and the participant is crucial. The

participants must be at ease to respond honestly to the questions, therefore, building trust

and establishing a respectable relationship are imperative (Malterud, Siersma, &

Guassora, 2016; Yin, 2018). To build the relationship, I followed the best practices

suggested by Yin (2018) for credible qualitative research. I explained the objective and

the process of the study to the participants and obtained their willingness to participate. I

established a professional working relationship with the participants by talking to them

over the phone and maintaining confidentiality.

Research Method and Design

The research method and design define the intention of the study and represent

solutions for AE case processing automation.

Research Method

A research method guides the investigators to obtain answers to the research

question while ensuring rigor in the research and reliability of results (Dresch, Pacheco

Lacerda, & Cauchick Miguel, 2015). There are three types of research: qualitative,

quantitative, and mix-methods research (Yin, 2018). During the research for a

methodology, I considered all the three types of research. Upon detailed research, I

decided to conduct a qualitative study to explore the strategies managers use to

implement innovation technologies.

The primary purpose of my study was exploration, with a pursuit of identifying

strategies to address my research question. Exploratory studies allow researchers to

understand an issue, a problem, or a phenomenon (Saunders et al., 2015). Qualitative

studies focus on phenomena based on ideas related to human knowledge, including an

objective to share knowledge and extend research data (Ward et al., 2018). Using a

qualitative method, the researcher can understand a phenomenon through a participant’s

perspective based on rich data and thick descriptions (Azungah, 2018; Barnham, 2015).

Developing generalizable rules and verification or falsification of hypotheses are

the focus of quantitative research methods (House, 2018). Researchers collect numerical

data in order to test theories or hypotheses using measurement of specific variables

(Saunders et al., 2015). The research problem and question of my study did not require

testing correlations, hypotheses, or existing theories using statistical analysis or

mathematical models. Therefore, a quantitative approach was not suitable for this study.

Researchers use mixed method research when their research question can be best

answered by an integrated qualitative and quantitative research than either of them alone

(House, 2018; Saunders et al., 2015). Challenges of mix method include the large amount

of time the study requires, size of samples, and quality of data and related interpretation

during analysis (House, 2018). Mixed method is complicated and is discouraged for

students (Fusch et al., 2018). Though a mixed method approach may have fit my research

problem, required sample size, large amount of time, and potential for discrepancies

during analysis were major deterrents.

Research Design

A research design is a plan to collect and analyze data in order to obtain answers

how they were implemented which yielded certain results (Yin, 2018). Using a case study

strategy enables researchers to perform in-depth inquiry into a topic to generate insights

in a real-life context experiences (Saunders et al., 2015; Yin, 2018). A case study focuses

on how things are and how they behave (Dresch et al., 2015). Case study facilitates

collecting richer information pertaining to a phenomenon (Dasgupta, 2015).

experiences related to the study topic (McIntosh & Morse, 2015).

A researcher selects a case study design to explore a smaller sample using

multiple approaches to gather data (Dresch, Pacheco Lacerda, & Cauchick Miguel, 2015)

and conducts investigation and builds theories where the phenomenon is newer or poorly

understood (Runfola et al., 2017). A case study design can be conducted on a single case

or multiple cases. A single case study design is appropriate when researching a unique

case (Dasgupta, 2015). Multiple case design allows researchers to build theories by

proving a strong base for cross case analysis (Yin, 2018). A single case study design was

deemed more appropriate to explore strategies used by some PV managers to implement

innovative solutions for automation.

automation solutions, and not on lived experience of an employee, phenomenology was

not an appropriate design. The purpose of ethnography is exploration of a phenomenon in

a cultural context (Lewis, 2015). Because cultural context was not critical for my study,

ethnography was not a suitable design.

Data saturation is an important aspect of qualitative case studies. Data saturation

is reached when no new information or themes are observed and the information can be

used to replicate the study, making it capable of generalization (Fusch & Ness, 2015;

Saunders et al., 2017). The depth and richness of data will determine the data saturation

and not solely the sample size of the population (Boddy, 2016; Fusch & Ness, 2015). I

ensured I collected in-depth information and rich and thick descriptions and continued the

interview process until no new information was discovered. I used member checking as

recommended by Fusch and Ness (2015) to achieve data saturation.

The study population is the total number of people within the organization eligible

for sampling consideration in the interview study (Yin, 2018). To fit the purpose of the

study and collect relevant knowledge, researchers following qualitative approach use

Because this study is set in a specific function within a pharmaceutical industry, a

purposive sampling design was used to select participants for this study. A purposive

sampling design is used when a particular phenomenon is studied in a specific context

(Tobi & Kampen, 2017; Yin, 2018). The participant sample must be representative of the

population (Boddy, 2016). The sampling technique did not yield a large and diverse

selection of participants from the PV function of a small pharmaceutical company.

Therefore, I invited four PV systems managers to participate in the study.

To decide the number of participants to be interviewed for this study, I followed

the principles of data saturation. Data saturation is reached when no new information or

themes are observed and the information can be used to replicate the study, making it

capable of generalization (Fusch & Ness, 2015; Saunders et al., 2017). The number of

participants for a study is contingent on the population chosen (Saunders & Townsend,

2016). The depth and richness of data will determine the data saturation and not solely

the sample size of the population (Boddy, 2016; Fusch & Ness, 2015).

I interviewed four PV systems managers, selected using the purposive sampling

method. In their qualitative study, Palaya, Pearson, and Nash (2018) conducted eight

managers working in a single pharmaceutical company.

consent from potential candidates to participate. Walden University’s approval number

for this study is 04-24-20-0972741 and it expires on April 23, 2021. I contacted the study

participants by phone or email requesting their participation in the study and their

response indicating their willingness to participate was documented. Quantitative

interviews provide rich descriptions of complex organizational realities and offer

ecological validity (Saunders & Townsend, 2016). The interview protocol and the

interview questions as shown in the appendix A were administered to the participants via

telephonic interviews. I digitally audio recorded and transcribed the interviews to

facilitate reliability of the data. Reliability is also ensured through member checking,

which is the process of involving the participants in the interpretation of data to enhance

the credibility of the results (Birt, Scott, Cavers, Campbell, & Walter, 2016). The

participants were asked additional questions if needed to confirm accuracy and obtain any

additional information or potential correction of data (Birt et al., 2016; Morse, 2015).

Member checking follow-up interviews were originally planned to be conducted if

participants, and understanding of contextual risks contribute to ethical concerns

(Wallace & Sheldon, 2015). I used confidentiality procedures to protect the participants

and maintain their privacy by following the protocols defined in the Belmont report

(National Commission for the Protection of Human Subjects and Biomedical and

Behavioral Research, 1979). To ensure adherence to the ethical standards, I obtained

study is 04-24-20-0972741 and it expires on April 23, 2021.

Consent is a combination of the form or document and communication with the

participants as a part of the conversation (Wall & Pentz, 2016). I used the consent form

and held an informed consent conversation prior to the interview to ensure the

participants understood the key elements and their rights pertaining to informed consent

as a recommended best practice by Koyfman, Reddy, Hizlan, Leek, and Kodish (2016).

The consent process must include the privacy and confidentiality protection (Ennever,

Nabi, Bass, Huang, & Fogler, 2019). I provided a student participant invitation and

informed consent form to all participants and obtained their consent to participate prior to

data collection. The participants had the right to withdraw from the study at any point

the rights of the participants, I will store the data collected as password-protected files on

a flash drive for 5 years. After 5 years, I will duly destroy the data collected including the

recording files and any documents. To ensure confidentiality, I did not use the

participants’ names, organization name, or any identifiable information as a best practice

recommended by Ennever et al. (2019) and instead, assigned codes such as P1 and P2 to

the participants to ensure participant privacy. This study was voluntary. There was no

payment for participation.

and triangulate the data I collected from the interviews, internal and publicly available

company and project documentation as evidence of strategies the PV systems managers

used to implement innovative technology solutions for automation. The constructs of the

TTF model and the research question assisted in focusing the data analysis on the themes

identified from the literature review.

Based on Yin’s (2018) recommended five-step data analysis, I used the ATLAS.ti

software to compile, disassemble, reassemble, interpret, and conclude the findings. I

recorded in my reflective journal for coding in ATLAS.ti software for data analysis. This

process assisted in successfully performing bracketing and the researcher bias was

included while performing content analysis. ATLAS.ti software assisted in organizing the

collected data, maintaining a list of codes and keywords to maintain consistency

(Bengtsson, 2016; Yin, 2018), and identifying themes. I established a case study database

as recommended by Yin (2018) to track all the data collected and maintain a chain of

evidence.

Reliability and Validity

Reliability

Researchers must be methodical in order to establish rigor and ensure reliability

of their findings. Reliability refers to replication or repeatability to achieve consistent

findings (Bengtsson, 2016; Yin, 2018). To ensure dependability, repeatability, and

consistency during data collection, I used an interview protocol with a set of interview

questions (Appendix A) and administered the same protocol to all the participants.

included using a coding system, member checks, thick description, and triangulation

(Morse, 2015).

Validity

Validity refers to appropriateness of measures, accuracy of analysis, and

generalizability (Saunders et al., 2015) that truthfully and accurately describes the

phenomena (Bengtsson, 2016). Validity refers to the credibility, transferability, and

confirmability of the findings. A study is credible when the results presented with the

context are recognizable to people who share the experience, which refers to truthfulness

or the internal validity of the study (Hammarberg, Kirkman, & de Lacey, 2016).

Transferability is the extent to which the study findings are useful to other people, which

is achieved by researchers being transparent (Connelly, 2016). Confirmability refers to

the degree to which a study is consistent and can be repeated (Connelly, 2016). Validity

is one of the prime concerns of researchers using qualitative methodology (Fusch et al.,

2018). The subjective nature of unstructured data collection poses a threat to validity

(Morse, 2015). To determine validity, Morse (2015) identified prolonged engagement,

peer review, negative case analysis, using a coding system, member checking,

of the study (Fusch et al., 2018; Joslin & Müller, 2016). (b) member checking or

participant validation: the process where the participant in involved in the interpretation

of the collected data and is questioned during the interview process to confirm accuracy

and obtain any additional information or potential correction of data (Birt et al., 2016;

Morse, 2015). I used both triangulation and member checking as methods to reach data

saturation. Data saturation is reached when no new information or themes are observed

and the information can be used to replicate the study, making it capable of

generalization (Fusch & Ness, 2015; Saunders et al., 2017). Data saturation refers to the

depth of the data (Fusch & Ness, 2015). I ensured I collected in-depth information and

rich and thick descriptions and continued the interview process until no new information

was discovered. I used member checking as recommended by Fusch and Ness (2015) to

achieve data saturation.

I gathered rich data and thick descriptions and documentation during the

interviews (Sovacool et al., 2018) to allow for transfer of the study findings to another

context, thereby increasing credibility. I shared the interview transcripts and my

interpretation with the participants to review and confirm the accuracy of the transcripts

information emerged. However, I did not conduct follow up interviews. I triangulated the

information gathered from the interviews with the documentation to further confirm data

saturation. To increase the reliability of the study, I established a case study database as

evidence. I addition, I maintained an audit trail and recorded the approach for the data

collection, interpretation, and analysis to enable retracing of the steps and duplication,

thereby increasing transferability.

The purpose of this qualitative single study was to explore the strategies used by

PV systems managers to implement innovative technology solutions to automate AE case

processing. In section 2, I provided justification for selecting qualitative single case

design for this study and focused on the strategies and approaches used for key

components of this study. I explained the research method and design, population and

sampling of participants, my role as the research instrument, and my approach for ethical

Section 3: Application to Professional Practice and Implications for Change

Introduction

The purpose of this qualitative single case study was to explore the strategies that

PV systems managers in the pharmaceutical industry use to implement innovative

technology solutions to automate AE case processing. I used a purposive sampling

technique to select four PV systems managers working in a pharmaceutical company

located in the Boston area of Massachusetts, United States, who have used successful

model changes, and (c) communication and training strategies. Internal drivers like

vision, objectives, timelines, cost, and compliance; ensuring solution is fit for purpose;

open communication, collaboration, and engagement; strong project management; and

simplified, standardized, and efficient business processes emerged as the key strategies

from these themes. The organization had contracted an external vendor to build a

strategic long-term partnership for their PV database, which was their foremost selection

and implementation strategy. The findings supported the existing findings on strategies

from the academic literature review. In the next sections, I will present the study findings,

application to professional practice, implications for social change, recommendations for

action, recommendations for further research, reflections, and conclusion.

In this study, I focused on the strategies that PV systems managers use to

implement innovative technology solutions to automate AE case processing. I conducted

semistructured interviews to collect data from four PV systems managers from a

pharmaceutical company in the Boston area of Massachusetts, United States. I used the

analyzed the documents collected from the participants for document review. For initial

analysis, I used word cloud for a visual representation of word frequency to get the first

look and summarize the interview transcripts. The word cloud showed system, project,

vendor, automation, implementation, communication, change, issues, and users as the

key and most frequently used words by participants in the interviews as shown in Figure

4. The words mapped to the themes and strategies that were identified during detailed

analysis.

Figure 4. Word cloud showing the most frequently used words in the interview data.

Created from ATLAS.ti.

The organization had contracted an external vendor to build a strategic long-term

partnership for their PV database, which was their foremost selection and implementation

strategy. The following key themes emerged with detailed analysis of the interview data:

solution and project. Business operational model changes theme included two strategies:

Simplified, standardized, and efficient business processes by supplementing and reducing

operational workforce and changes to organizational structure and job profiles.

Communication and training strategies included open communication, collaboration, and

engagement and training. The three main themes and the nine key strategies are shown in

Table 2 below. The references coded refers to the number of times the data references

were coded to the strategies.

Table 2

Themes and Strategies

coded

Theme 1: Selection and Implementation Strategies

Automation solution selection and implementation strategies was one of the key

themes that emerged from the interview data. I analyzed the data in two steps, initial and

detailed analysis. Initial analysis of the word cloud created by ATLAS.ti included vendor,

system, project, time, implementation, automation, and communication as the key words

as shown in Figure 5.

increasing efficiencies and consistency, thereby enabling reporting compliance. Beninger

(2018) emphasized on the new skillsets, which will define the roles of the next generation

of professionals who perform PV activities. Tomita (2019) added that organizations must

expect to supplement the human workforce with expert systems.

The TTF model matches the tasks as the events occur with the ability of a

technology to address these events. TTF focuses on the application of technology

The third key theme that emerged from the data was communication and training

strategies. This theme included change management-related strategies. Initial analysis of

the word cloud created by ATLAS.ti included communication, vendor, training, project,

team, and management as the key words as shown in Figure 7.

communication and collaboration was critical. Multiple touch points like steering

committee and core team meetings were built to ensure the information is flowing

through the organization and timely escalation and resolution of issues. P1 and P2 also

indicated collaborating externally with other companies to explore the opportunities for

automation, which can be practically implemented.

Training. Since the introduction of automation changed the way certain business

processes were managed, training was a critical aspect. The participants noted that

initial stages of the project. Considering varied stakeholders, both internal and external,

and the vendor implementation team located remotely added to the communication

challenges. The team used online collaboration tools like Google and Smart Sheets to

help track tasks and keep the teams aligned.

Findings regarding communication and training strategies matched previous

studies. Training and re-learning initiatives assist in preparing and guiding the in-house

staff for digital transformation (Mishra et al., 2019). The human and technology

relationship should also include user training, interaction, and collaboration sub-goals in

order to be effective (Oyekan et al., 2017). PV systems managers realized the importance

of communication and collaboration early during the implementation project and ensured

adequate measures were included to create open communication channels assisting in

timely escalations and change management. Ghosh et al (2020) also recommended

collaboration across the pharmaceutical industry between regulators and companies to

build a generic process flow to fuel third-party innovations.

In the TTF model, Goodhue and Thompson (1995) linked the technology solution

with individual performance. Wang et al. (2020) used the user task technology fit model

I included the researcher biases and assumptions that I had identified during data

analysis. My biases and assumptions were confirmed to be a practical reality. PV systems

managers I interviewed confirmed that they made the decisions to identify and implement

innovative technology solutions. Participant P1 confirmed that among the many strategies

available the PV organization managers decided to pursue the few above mentioned

strategies for successful implementation. All participants confirmed that PV business

operational model had an impact of introducing automation. And finally, the participants

confirmed that innovative automation technology was used by their organization to

automatically read and create database entries, thereby eliminating redundant manual

tasks, improving efficiencies, and over a period, achieving cost savings at the same time.

Applications to Professional Practice

Processing of an AE is an expensive resource-intensive process, which includes

high risk of errors and operational inefficiencies (Ghosh et al., 2020). A detailed review

of literature concluded that innovative technology such as data mining, machine learning,

AI, and robotics could result in new operating models for PV (Beninger & Ibara, 2016;

technology solution to automate AE processing. The organization had a larger vision to

automate PV business operations and this project was a part of that vision. Leaders and

managers in the PV function decided to partner with a vendor to source and implement

the solution, having realized that the organization lacked the necessary expertise on the

technology. Multiple internal drivers such as objectives, timelines, cost, and compliance

shaped the strategies the PV systems managers used to execute the project. Considering

this being their first automation solution, the team ensured that the solution is fit for

purpose, both task (functionality) and technology (architecture) perspectives. Strong

project leadership and steering committee were established for robust governance and

monitoring. The managers established an open communication, collaboration, and

training model to engage with the relevant stakeholders across multiple functions within

the organization and the vendor. Measures were established to capture metrics in terms of

volume, time, and quality of the business AE processing tasks before and after

automation. Regular project governance and monitoring measures ensured the automation

solution was implemented on time. The metrics defined the successful outcome of this

implementation.

the PV organization managers have started defining a sustainable model for PV business

operations. Other pharmaceutical companies can use the strategies identified in this study

to realize the benefits of implementing innovative automation technology solutions to

build a resilient and sustainable PV business operations organization.

The successful strategies identified in this study can be used by organizations to

adopt a digitized PV future. The study findings show a positive outcome and realization

of savings from PV business operations because of strategically implementing and

embedding automation. PV leaders can redirect the savings from PV operations toward

investing in the actual PV tasks like benefit-risk assessments of products, thereby

improving patient outcomes and making the products more efficacious and safer for

human use.

management of the high volume of AEs, globally diverse and changing regulations,

complex and evolving systems, and sparse PV expertise and talent (Price, 2018).

Automation technologies can detect, database, and report AEs quicker, and manage high

talent.

Recommendations for Action

This study revealed multiple strategies used by PV managers to implement AE

case processing automation solution. The three themes that emerged included automation

solution selection and implementation strategies, business operational model changes,

and communication and training strategies. The results revealed that the PV systems

managers can follow their company’s internal processes for solution selection and

implementation, while paying close attention to the strategies identified in this study. By

successfully implementing automation technology for AE case processing, PV managers

can realize time and effort savings and improve efficiencies with PV operations.

Other stakeholders in the pharmaceutical industry who could benefit from this

study include PV organization leaders and PV system database vendors. PV organization

leaders can include the strategies identified in this study as they start defining their PV

automation vision and objectives. PV system database vendors can build automation

solutions which fit the demand from the end users and build a mutually benefitting,

managers employed in one pharmaceutical company. This study can be extended in a

future research to include additional perspectives like end user feedback and IT

technology strategies. Due to the ever-increasing volume of AE cases to be detected,

processed, and reported, more companies are building visions for digitalization and

sustainability by incorporating more automation for their PV business operations and

case management. This opens multiple avenues for future research. This study can be

replicated with other study population to explore additional successful strategies. This

study focused on successful strategies to implement automation and not on the actual

automation solution that was implemented. A future study can delve deeper into the

technology that was used in addition to exploring successful implementation strategies.

One of the participants stated that their implementation was a success and that many

others are not. A future study can explore strategies that were unsuccessful, which can

serve as learnings for other companies starting on this journey.

technology solutions and implementations. I selected this research topic based on an

The doctoral study process seemed a daunting task when I started. Soon, I started

enjoying the process of conceptualizing the research, conducting, and concluding it. I did

not face any challenges with the selection of participants and interviewing them. The

participants took pride in their successful implementation as they are seeing the benefits

now. I conducted my doctoral study as the small-scale project in the DBA Qualitative

Research process course. I could effectively use the learnings from the pilot run for my

study.

I used an application on my mobile phone for the initial transcript of the interviews. The