1 – INTRO TO PHARMACOECONOMICS

Economics: The study of allocating limited resources to TYPES OF PHARMACOECONOMICS ANALYSIS CALCULATING RESULTS OF COSTS & BENEFITS
satisfy unlimited wants. COST- COST-EFFECTIVE COST-UTILITY COST-BENEFIT COST-ILLNESS • If retrospective data are collected for >1 year / if
MINIMIZATION ANALYSIS (CEA) ANALYSIS (CUA) ANALYSIS (CBA) ANALYSIS the project inputs / outcomes are estimated for >1
Health Economics (HE): Applies economic principles to ANALYSIS (CMA) year into the future, it is important to adjust /
healthcare, evaluating the costs (inputs) & consequences Compares: Costs of Compares: Costs ($) Compares: Costs ($) Compares: Measures: The discount these costs to one point in time
(outcomes) of interventions (drugs, devices, services). interventions. vs. outcomes measured vs. outcomes total economic • CBA can be presented in 3 formats:
in natural health measured in utility burden (direct
Pharmacoeconomics (PE): A specialized subset of HE units (e.g., mmHg units, most commonly and indirect • Difference between the
focusing specifically on evaluating the costs & outcomes reduced, life-years Quality-Adjusted Life costs) of a total costs & benefits
of pharmaceutical interventions. It's a type of outcomes gained, cases cured). Years (QALYs), which specific disease • Net benefits = total benefits
research. incorporate both on society. – total costs
NET BENEFIT
quantity & quality of CALCULATIONS • Net cost = total costs – total
life. benefits
PURPOSE & IMPORTANCE OF PE Outcomes: Assumed Outcomes: Measurable Outcomes: Can be Outcomes: Benefits Not a • Intervention is beneficial
• To inform resource allocation decisions within to be equivalent. but not necessarily different across are assigned a dollar comparative when: Net Benefit > 0 or Net
healthcare institutions and systems. identical. interventions. value (e.g., via analysis: Doesn't Cost < 0
• Helps determine coverage policies, formulary Willingness-to-Pay compare
management, drug pricing, drug approvals, treatment or Human Capital interventions but • Sum of the total benefits &
guidelines, and disease management programs. approach). quantifies the dividing by the total costs
• Answers key questions like: Is a drug/service worth its overall cost BENEFIT-TO-
• Can be expressed as
cost? Should it be reimbursed? Should it be added to a impact of an COST /
COST-TO- benefit-to-cost ratio // cost-
formulary? illness. to-benefit ratio
BENEFIT
RATIOS • Benefit-to-cost ratio > 0 or
Use: Choose the Use: Determine the Use: Compare Use: Determine if Use: Indicate Cost-to- benefit ratio < 0
BASIC PE EQUATION cheapest option "value for money" (cost interventions with total benefits resource
• PE studies fundamentally compare the COSTS ($) when efficacy and per unit of effect), often diverse health outweigh total costs magnitude
• Rate of return that equates
associated with providing a pharmaceutical safety are identical. using the Incremental benefits (e.g., (Net Benefit > 0 or needed, compare
the present value (PV) of
product/service (Rx) to the OUTCOMES achieved. Cost-Effectiveness screening vs. surgery); Benefit/Cost Ratio > economic impact
benefits to the PV of costs
• Outcomes can be: Ratio (ICER). allows comparison 1); allows between
• Goal is to find the rate of
o Clinical: Measurable changes in disease state. across different comparison with diseases, assess
return that would make the
o Humanistic: Patient's perception, quality of life, disease areas. Uses non-health market potential.
INTERNAL RATE costs & benefits equal
functioning. Incremental Cost- programs. Ethically
Utility Ratio (ICUR). challenging due to
OF RETURN • The decision rule for IRR is
o Economic: Costs of care. (IRR)
valuing life/health in to accept all projects with
dollars. an IRR > Hundle rate
COMPLEX EVALUATION CONCEPTS • Difficult to calculate
manually & requires a
Adjusting future costs & benefits
computer programme
DISCOUNTING to their present value.

Testing how results change when

SENSITIVITY key assumptions are varied.
ANALYSIS

STRATIFICATION Analyzing results for specific

subgroups of patients.
2 – MEASURING & ESTIMATING COSTS

Cost: Represents the value of resources (inputs) ESTIMATING DIRECT MEDICAL COSTS TIMING ADJUSTMENTS FOR COSTS
consumed to produce a good or service. Data Sources: Measured directly (patient logs), collected
retrospectively (medical records, claims data), or STANDARDIZATION DISCOUNTING
Opportunity Cost: The true economic cost; the value of estimated from standard lists.
the best alternative use of those resources that was Adjusting past costs to a Adjusting future costs and
forgone. Estimating Specific Costs: common point in time benefits back to their
Use sources like Average (usually the present) to present value because
Costs vs. Charges: Crucially different. Wholesale Price (AWP - list price) account for inflation. money today is worth
• Cost: The actual expense incurred by a provider to or Average Manufacturer's Price Methods include using more than money in the
MEDICATIONS
deliver a service. (AMP - closer to actual cost, but unit costs or inflation future. A standard
• Charge: The price billed to a payer (often includes often proprietary). Source must be indices (like MCPI/CPI). discount rate (e.g., 3-5%)
profit, may be higher than actual cost). stated. is typically applied.
• Reimbursement: The amount actually paid (often MEDICAL Use provider charge lists or payer
lower than the charge). SERVICES fee schedules. COST COMPARISON TERMS
Calculated based on time spent
PERSONNEL (estimated from logs or time MARGINAL COST INCREMENTAL COST
COST CATEGORIZATION studies) multiplied by wage/salary.
HOSPITALIZATION Estimated with varying precision: Cost of producing one The difference in total
DIRECT Costs associated with medical additional unit of costs between two
MEDICAL care (e.g., medications, physician outcome/product. competing interventions.
COSTS visits, hospitalizations, labs). Most MICRO- Detailed tracking of every specific Used extensively in
commonly measured. COSTING service/resource used (Most CEA/CUA/CBA (as part of
DIRECT NON- Costs incurred by precise, but resource-intensive). ICER/ICUR).
MEDICAL patients/families directly related to DIAGNOSTIC- Average cost/reimbursement for a
COSTS treatment but not medical in RELATED bundle of services related to a
nature (e.g., travel, special diets, GROUPS (DRGs) specific diagnosis/procedure
childcare). group (Commonly used by
INDIRECT Costs related to lost productivity payers).
COSTS due to illness (morbidity) or DISEASE Average cost per day for a specific
premature death (mortality). SPECIFIC PER disease (More precise).
Non-financial costs reflecting DIEM
INTANGIBLE pain, suffering, anxiety, or grief PER DIEM Average cost per day for any
COSTS associated with hospitalization (Least precise).
disease/treatment. Difficult to
quantify monetarily.

PERSPECTIVE IN PE STUDIES
Determines for which costs & benefits are included.
Common perspectives:
• Provider/Institution: Focuses on actual costs to the
hospital/clinic.
• Payer: Focuses on reimbursed amounts (e.g.,
insurance company, government).
• Patient: Focuses on out-of-pocket expenses and
indirect costs.
• Societal: Broadest view includes all direct (medical
& non-medical), indirect, and potentially intangible
costs across all parties. Often ideal but complex to
measure fully.
3 – MEASUREMENT OF HEALTH OUTCOMES

OUTCOME RESEARCH CORE METHODOLOGIES HEALTH-RELATED QUALITY OF LIFE (HRQoL) & UTILITY
• research that is concerned with the effectiveness of 2 main approaches to evaluating health interventions: • This is the foundation for measuring patient-centered outcomes.
public health interventions & health services • Health-Related Quality of Life (HRQoL): A multi-dimensional
• study of the effectiveness of a treatment in the real- METHODOLOGY WHAT IT IS KEY OUTCOME MEASURED concept that captures an individual's / group's perceived physical
world setting A comparison of alternative options The main types are: & mental health over time. It's about the impact of a health state on
• measures result of various medical treatments &/ (e.g., two different drugs) in terms of • Cost-Benefit analysis a person's life.
interventions in patient populations ECONOMIC their costs (inputs) and consequences • Cost-Effectiveness: Consequences • Utility: The preference or value that a person places on a particular
• outcomes that can be measured: ANALYSIS (outputs). in natural units (e.g., life-years gained, health state. It is the core component for calculating QALYs.
▪ survival blood pressure reduction). • Two Theories of Utility:
▪ costs • Cost-Utility: Consequences in TYPE MEASUREMENT EXAMPLE
▪ physiological measures "healthy years" (usually QALYs). CARDINAL Quantitative. Assigns a "My health state
▪ QoL UTILITY specific numerical value has a utility of 0.8."
Involves modelling a clinical problem A recommended course of action
DECISION to guide decision-making when there is based on the model's output. (e.g., 0 to 1 scale).
IMPORTANCE Qualitative. Ranks "I prefer Health
ANALYSIS uncertainty. It uses data like disease
• to provide better information to inform patient prevalence, intervention effectiveness, ORDINAL preferences without State A over Health
decisions costs, and patient utility values. UTILITY assigning a specific State B."
• to guide health providers value.
• to inform health policy decision
GENERIC vs. DISEASE-SPECIFIC INSTRUMENTS HOW TO MEASURE HRQoL WEIGHTS (HEALTH UTILITY)
BENEFITS These are the tools (questionnaires) used for indirect measurement. DIRECT ELICITATION METHODS
• ↑ participation in decision You ask people directly to value a health state. 3 main methods are:
making INSTRUMENT ADVANTAGES DISADVANTAGES EXAMPLE
• ↑ choice regarding treatment TYPE VISUAL A simple "feeling thermometer" where a patient
CONSUMER options May not be sensitive to EQ-5D ANALOGUE marks their health on a scale from 0 (worst
• Broadly applicable.
• Assurance regarding GENERIC • Allows comparison across small but important SCALE (VAS) imaginable health) to 100 (best imaginable
effectiveness of interventions different diseases and changes in a specific health).
• Assessment & development of populations. disease. Asks a person to choose between 2 options:
interventions DISEASE- More relevant and responsive to Cannot compare BPH Impact Index (for • Option 1: A certain, intermediate health
• Protection from malpractice SPECIFIC changes in a specific condition. results across prostate issues) STANDARD state (e.g., living with chronic pain).
suits different diseases. GAMBLE (SG) • Option 2: A "gamble" with a probability of
• Greater certainty regarding the achieving perfect health & a probability of
HEALTH-CARE benefit of an intervention FOCUS ON EQ-5D immediate death.
PROVIDER • Standards / guidelines to guide This is the most common generic instrument. • The most important to know. A person is
clinical practice • What it is: A concise, self-reported health measure. asked to trade years of life for better health
• Shared responsibility in TIME TRADE- quality.
decision-making • What it measures: 5 Dimensions (5D): Mobility, Self-Care, Usual Activities, Pain / Discomfort & Anxiety OFF (TTO) • The Choice: Live for t years in a poor health
/ Depression. state OR live for x years (where x < t) in
• Greater use of effective
HEALTH-CARE interventions • Why it's important: It's internationally compatible, widely used in research & health policy & provides perfect health.
ORGANIZATION • Discontinuation of ineffective the utility values needed for economic evaluations. • Calculation: The utility is calculated when
MANAGEMENT interventions / practice the person is indifferent between 2 choices:
• Cost savings Health State Score = x / t. (e.g., if someone is
ULTIMATE OUTCOME: QUALITY-ADJUSTED LIFE YEAR (QALY) indifferent between 10 years with diabetes &
• Greater ability to plan health
This is the single most important concept to understand. 8 years in perfect health, the utility for that
services
• Target research in areas of • Definition: A generic measure of disease burden that combines both the quality (utility) and diabetic state is 8/10 = 0.8).
GOVERNMENT greatest potential impact based the quantity (years) of life into a single number.
on examination of database • Calculation: INDIRECT ELICITATION METHODS
• Only effective pharmaceuticals You use a standardized questionnaire where the answers correspond
& services are subsidized to pre-determined utility scores derived from a general population
survey.
• Cost savings • The QALY Scale:
o 1 QALY = 1 year of life in perfect health. DATA SOURCES IN OUTCOME RESEARCH
o 0 QALY = Represents death. • Administrative databases
o < 0 QALY = Represents health states considered worse than death. • Clinical databases
• Purpose: QALYs are the primary outcome measure used in Cost-Utility Analysis, allowing for a • Disease registers
standardized comparison of the value for money of different health interventions. • Clinical trial databases
How does pharmacoepidemiology research data contributes towards
policy making?

• Identifies Health Burdens: Shows high prevalence (nearly 80%) of

prehypertension/hypertension → Justifies national health priority.
• Enables Targeted Interventions: Focuses on high-risk groups (e.g., rural,
older adults, men) → Enables customized programs.
• Guides Clinical Guidelines: Strong link to diabetes supports BP
monitoring in diabetic patients.
• Evaluates Policy Impact: Acts as a baseline to assess the success of
health campaigns (e.g., salt reduction, screenings).
• Informs Drug Policies: Supports inclusion of safe, effective
antihypertensives in national formularies.
• Promotes Health Equity: Highlights socioeconomic disparities → Guides
policies to support vulnerable populations.
4 – DECISION ANALYTIC MODEL

• A quantitative approach to decision-making under MARKOV MODEL OTHER MODEL TYPES & SOFTWARE STEP-BY-STEP PROCESS OF BUILDING A MODEL
uncertainty, where at least two options are compared • Best for: Modeling chronic diseases or conditions • Discrete Event Simulation (DES): A more complex • Frame the Question: Clearly define the problem
by evaluating their potential consequences. that progress over a long period, where patients can "micro-simulation" model that tracks individual using a framework like PICO (Population,
• Purpose: Models are used to synthesize data from move between different health states over time. patients and their unique event histories, offering Intervention, Comparator, Outcome) and establish
multiple sources (clinical trials, literature, expert • Structure: greater flexibility than Markov models. the perspective (e.g., healthcare system or societal).
opinion) to simulate long-term scenarios and • Health States: A set of mutually exclusive • Software: Common tools used to build these models • Structure the Problem: Choose the appropriate
estimate costs and outcomes more comprehensively conditions a patient can be in (e.g., "Well," include TreeAge, Microsoft Excel, and R. model type (Decision Tree or Markov Model) based
than a single study can. "Recurrence," "Dead"). on the nature of the disease.
• Contrast with Randomized Controlled Trials (RCTs): • Cycles: The model runs in discrete time steps • Estimate Probabilities and Outcomes: Gather data
Models overcome the limitations of RCTs, such as (e.g., months or years). In each cycle, a patient from clinical trials, epidemiological studies, and
short duration, inability to compare all relevant can either remain in their current state or other sources to assign values for probabilities,
alternatives, and a study population that may not transition to another. costs, and health outcomes (utilities/QALYs) to every
reflect the real world. • Transition Probabilities: The probability of part of the model.
• Core Elements: Every decision model is built using moving from one state to another during a single • Analyze the Model:
four key inputs: cycle. These are often presented in a "transition o For a Decision Tree, this involves "folding back"
matrix." the tree from right to left, calculating the
MODEL The framework of choices and • Process: The model simulates a cohort of patients expected costs and benefits at each chance
STRUCTURE events (e.g., a decision tree). moving through the health states cycle by cycle, node.
PROBABILITIES The likelihood of different accumulating costs and health benefits along the o For a Markov Model, this involves running the
events occurring. way. cohort simulation to calculate the total costs
COSTS The resources consumed for and benefits over the model's time horizon.
each event or health state.
BENEFITS / The health effects, often • Calculate the ICER: The Incremental Cost-
OUTCOMES measured in QALYs or life- Effectiveness Ratio (ICER) is the primary result. It is
years. calculated as:
o ICER = (Cost of Intervention - Cost of
2 MAIN TYPES OF DECISION MODELS Comparator) / (Effect of Intervention - Effect of
DECISION TREE Comparator)
o This ratio represents the additional cost for
• Best for: Simple, short-term problems where events
each additional unit of health benefit (e.g., cost
happen in a clear sequence and do not repeat.
per QALY gained). An intervention is considered
• Structure: dominant if it is both more effective and less
o Decision Node (□): Represents a choice costly.
between different options (e.g., Treatment A vs.
Treatment B). • Address Uncertainty (Sensitivity Analysis): Since
o Chance Node (○): Represents a point of model inputs are uncertain, sensitivity analysis is
uncertainty where different events can occur, crucial.
each with a specific probability. o One-Way SA: Changes one input variable at a
o Terminal Node (△): The final outcome of a time to see its impact on the result.
pathway, with an associated cost and health o Probabilistic SA (PSA): Randomly samples all
benefit. input variables from statistical distributions to
• Limitation: Not suitable for chronic diseases with generate a range of possible ICERs, providing a
recurring events or conditions where time is a critical more robust measure of confidence in the
factor. result.
• Discounting: For long-term models (like Markov),
future costs and benefits are "discounted" (usually
at a rate of 3% per year) to reflect their lower value in
present-day terms.
5 – INTRO, ROLE & APPLICATION OF PHARMACOEPIDEMIOLOGY

Pharmacoepidemiology: The study of the use and STUDY DESIGN INSURANCE RECORDS MAJOR ISSUES / CHALLENGES
effects of drugs in large numbers of people. It combines
pharmacology (study of drugs) and epidemiology (study of Follow groups (exposed vs. POLICYHOLDER INFO CLAIMS TRANSACTION • Quality & Availability: Data can be
determinants, distribution, and control of unexposed) forward in time DATA incomplete, inaccurate, or biased.
diseases/health factors in populations). COHORT (prospective) or analyze past demographics: age, claim details: date, type DATA • Linkage: Difficulty linking different
STUDIES data (retrospective) to assess gender, location, etc. (e.g. medical, accident), RELATED data sources (e.g., pharmacy
Goals: outcomes. Good for examining status ISSUES records with clinical outcomes).
• Quantify and understand drug use patterns long-term effects/risk factors. (approved/denied), & • Variability: Inconsistent coding for
(prescribing habits, appropriateness, adherence). Start with outcome (cases vs. payment amounts diagnoses and medications across
• Identify predictors for medication use. CASE-CONTROL controls) and look back for past risk factors: health claim frequency & different systems.
• Assess beneficial and harmful effects (risks/benefits) STUDIES exposures. Efficient for rare conditions & lifestyle severity: no. of claims & • Confounding Variables: Other
of drugs in real-world populations. diseases/outcomes. their financial impact factors (like lifestyle or
• Optimize drug use, often requiring individualized CROSS- Assess exposure and outcome at comorbidities) can distort the
therapy. SECTIONAL a single point in time. Good for relationship between a drug and an
STUDIES prevalence, hypothesis DATA DERIVED FROM INSURANCE RECORDS METHODO- outcome.
INDIVIDUALIZED THERAPY generation. LOGICAL • Bias: Selection bias (how
• Tailoring drug therapy involves considering Gold standard for efficacy; LOSS DATA EXTERNAL DATA ISSUES participants are chosen) and
risk/benefit ratios, potential drug interactions, participants randomly assigned INTEGRATION information bias (errors in data
patient clinical status, and specific needs. to intervention/control. Less • claim payouts: total • regional health recording) can skew results.
• Older Adults: A key focus due to higher drug use, RANDOMIZED common for post-approval amounts paid out for trends: data on • Causality: It is difficult to prove a
polypharmacy risk, altered sensitivity, potential non- CONTROLLED safety/effectiveness studies in claims including prevalent health drug caused an effect in
adherence, and increased vulnerability to side TRIALS (RCTs) pharmacoepidemiology due to reserve estimates & conditions in specific observational studies.
effects (e.g., anticholinergics, BZNs, NSAIDs). cost / ethics / generalizability recovery amounts regions • Studies are often expensive and
issues compared to • loss ratios: the ratio • economic indicators: time-consuming.
OBJECTIVES & APPLICATIONS observational designs. of claims paid to economic factors RESOURCE • Requires interdisciplinary
premiums earned, that might influence & collaboration (clinicians,
Identify adverse drug events used for risk healthcare utilization LOGISTICAL statisticians, policymakers).
SAFETY (ADEs), frequency, and risk DATA SOURCES modelling & / claim rates ISSUES • Translating complex findings into
factors. profitability analysis clear, actionable guidance is often
Assess how well drugs work in PRIMARY DATA Collected specifically for the difficult.
EFFECTIVENESS real-world settings (vs. efficacy in study (e.g., surveys, trial data).
controlled trials). SECONDARY Existing data collected for other FUTURE DIRECTIONS
PATTERN OF Describe how drugs are actually DATA purposes (e.g., administrative • Big Data & AI: Using large datasets and artificial
USE used in clinical practice (dosing, claims, EHRs, registries). intelligence to identify patterns, improve data quality,
duration, patient types). Data from routine clinical and predict outcomes.
Drug Safety Monitoring, Post- practice (EHRs, claims, patient • Real-World Evidence (RWE): Increasing focus on using
Marketing Surveillance (detecting REAL-WORLD registries, wearables, social real-world data to provide more applicable insights for
KEY rare ADEs), Pharmacovigilance DATA SOURCES media). Insurance records patient care.
APPLICATIONS (detecting, assessing, preventing (RWDS) provide claims details, payment
• Global Collaboration: Creating international networks
ADEs), and informing info, risk factors, and can be
to share data and expertise to address global health
Policymaking (guidelines, linked to health trends.
challenges.
regulations).

HISTORICAL CONTEXT
• The field existed before being formally named
"Pharmacoepidemiology" in 1984.
• ISPE (International Society for
Pharmacoepidemiology) formed in 1989.
• Acknowledges significant historical contributions
from non-European civilizations, particularly the
Islamic Golden Age (e.g., Ibn Sina/Avicenna's "Canon
of Medicine," Ibn al-Baythar's pharmacopeia),
building on Greek/Roman knowledge. Early records
also exist from Mesopotamia, Egypt, China.
TUTORIAL

Instruction: Using the decision analytic model, craft a decision-making STEP 3: ESTIMATE THE PROBABILITIES Prevalence and Factors Associated with Prehypertension and
process and show how the decision can be made based on the information At each chance node, we must assign a probability. Hypertension Among Adults: Baseline Findings of PURE Malaysia Cohort
provided. Provide justifications. • Clinical Trials and Meta-Analyses: To find the probability of achieving Study
Background: A large urban hospital is developing a Clinical Decision Support blood pressure control with each drug class.
System (CDSS) aimed at optimizing the treatment of hypertensive patients • Pharmacovigilance Data & Cohort Studies: To find the probability of Background
with type 2 diabetes mellitus (T2DM). The goal is to select the most suitable specific adverse events. For this problem, we would need to find: Although prehypertension and hypertension can be detected at the primary
antihypertensive therapy for individual patients based on a balance of o The probability that Thiazide diuretics worsen glycemic control. healthcare level and low-cost treatments can effectively control its
efficacy, safety, cost, and patient-specific factors. o The probability that a patient on Beta-blockers experiences a severe complications, hypertension is still the world's leading preventable risk factor.
Clinical Challenge: There are several classes of antihypertensive hypoglycemic event due to masked symptoms. Therefore, the present study aimed to determine its prevalence and its risk
medications available, including: o The probability of slowing the progression of diabetic nephropathy factors among Malaysian adults.
• ACE inhibitors (e.g., Lisinopril) (kidney disease) with ACE inhibitors/ARBs versus other classes. Methods
• Angiotensin II receptor blockers (ARBs) (e.g., Losartan) A cross-sectional study involving 7585 adults was performed covering the
• Calcium channel blockers (CCBs) STEP 4: ESTIMATE THE VALUES (UTILITIES) OF THE OUTCOME rural and urban areas. Respondents with systolic blood pressure (SBP) of 120-
• Thiazide diuretics We need to assign a value, or "utility," to each terminal node on a scale (e.g., 0 139 mmHg and/or diastolic blood pressure (DBP) of 80-89 mmHg were
• Beta-blockers for worst outcome, 1 for best outcome). The value reflects the desirability of categorized as prehypertensive, and hypertensive categorization was used for
Additional information: that health state. respondents with an SBP of ≥140 mmHg and/or DBP of ≥90 mmHg.
• ACE inhibitors and ARBs - renal protective effects. • Highest Value (e.g., Utility = 1.0): An outcome of "BP Controlled, No Results
• Thiazide diuretics may worsen glycemic control. Adverse Events, with Renal Protection." This would be associated with Respondents reported to have prehypertension and hypertension were 40.7%
• Beta-blockers may mask hypoglycemia symptoms the ACE inhibitor and ARB paths. and 38.0%, respectively. Those residing in a rural area, older age, male, family
• High Value (e.g., Utility = 0.9): "BP Controlled, No Adverse Events" history of hypertension, and overweight or obese were associated with higher
(achieved with a CCB). This is good but lacks the specific long-term benefit odds of prehypertension and hypertension. Unique to hypertension, the
STEP 1: FRAME THE QUESTION for this patient group. factors included low educational level (AOR: 1.349; 95% CI: 1.146, 1.588),
For a patient with co-existing hypertension and type 2 diabetes, which first- unemployment (1.350; 1.16, 1.572), comorbidity of diabetes (1.474; 1.178,
• Lower Value (e.g., Utility = 0.6): "BP Controlled, but with Worsened 1.844), and inadequate fruit consumption (1.253; 1.094, 1.436).
line antihypertensive medication class (ACE inhibitor, ARB, CCB, Thiazide
Glycemic Control." This is a possible outcome on the Thiazide
diuretic, or Beta-blocker) provides the optimal balance of efficacy, safety, and Conclusions
diuretic path.
long-term benefits? As the prehypertensive state may affect the prevalence of hypertension,
• Very Low Value (e.g., Utility = 0.2): "BP Controlled, but with Severe proactive strategies are needed to increase early detection of the disease
STEP 2: STRUCTURE THE CLINICAL PROBLEM Hypoglycemia Event." This is a risk on the Beta-blocker path. among specific group of those residing in a rural area, older age, male, family
• Decision Node (Square): The tree begins with a single decision node • Lowest Value (e.g., Utility = 0.0): "BP Uncontrolled with Severe Adverse history of hypertension, and overweight or obese.
representing the choice of initial therapy. Event."
• Branches (Lines): Five branches extend from this node, one for each Discuss the strength and limitation of this study.
medication class: STEP 5: ANALYZE THE TREE (CALCULATE EXPECTED VALUE)
o ACE inhibitor (e.g., Lisinopril) The analysis involves "rolling back" the tree. For each medication branch, we STRENGTHS EXPLAINATION
o ARB (e.g., Losartan) calculate the Expected Value (EV) by multiplying the probability of each
o Calcium Channel Blocker (CCB) outcome by its assigned utility and summing the results. Large & 7585 adults from both rural and urban areas
o Thiazide Diuretic The CDSS would perform this calculation for all five drug classes. The optimal representative improve generalizability.
o Beta-blocker decision is the one with the highest calculated Expected Value.
sample
• Chance Nodes (Circles): Each branch leads to chance nodes representing Clear, standard Used internationally accepted BP criteria for
key clinical events and outcomes. The primary outcomes to consider are: STEP 6: DEALING WITH UNCERTAINTY & HETEROGENEITY (SENSITIVITY
definitions prehypertension and hypertension.
o Efficacy: BP Controlled vs. BP Not Controlled. ANALYSIS)
Identified multiple Included demographic, clinical, socioeconomic,
o Adverse Events: Presence vs. Absence of significant The model's recommendation is based on estimates. A sensitivity analysis is
risk factors and lifestyle associations.
side effects. performed to test how robust the conclusion is. We would ask:
Quantitative analysis Provided adjusted odds ratios for clearer
• Terminal Nodes (Triangles): Each path ends in a terminal node that • Uncertainty: How much would our estimate for the probability of renal (AOR & CI) understanding of risk strength (e.g. diabetes AOR
describes the final health state (e.g., "BP controlled, no adverse events, protection have to decrease before an ACEi/ARB is no longer the top 1.474 > unemployment AOR 1.350).
with renal protection"). choice? LIMITATIONS
• Heterogeneity: The "best" choice may differ by patient. For a patient with
brittle diabetes and frequent hypoglycemia, the negative utility of a Beta- Cross-sectional Cannot establish causality — only associations.
blocker would be even lower, making it a strongly contraindicated choice. design
Conversely, for a patient with a history of angioedema (a severe side effect Potential recall bias Self-reported data (e.g., fruit intake, family
of ACE inhibitors), the model would need to exclude that option. history) may be inaccurate.
Limited Sociocultural differences may affect
STEP 7: INTERPRET THE RESULT generalizability applicability in other countries.
For this patient with hypertension and Type 2 Diabetes, the recommended outside Malaysia
first-line therapy is an ACE inhibitor or an ARB. Unmeasured Factors like salt intake, physical activity, or
confounding factors healthcare access not assessed.