Pharm D Year 4, 4.4 Biostatistics and Research Methodology — Pharm D Year 4

Unit 1: Research Study Designs

Syllabus scope

This unit covers case studies, observational studies, and interventional studies.

Research question and design

Research begins with a focused question, defined population, exposure or intervention, comparator, outcome, and timeframe. The design should fit the question. A descriptive case can identify an unusual event. An observational study estimates patterns without assigning treatment. An interventional study evaluates an assigned action.

Design selection, step by step

1. Define whether the aim is description, association, prediction, diagnosis, prognosis, or treatment effect.
2. Specify the population and eligibility criteria.
3. Define exposure, intervention, comparison, and outcome before data collection.
4. Choose a feasible design that limits bias.
5. Plan measurement, follow-up, analysis, ethics, and reporting.

Case reports are useful for signal generation but lack a comparison group. Cross-sectional studies estimate prevalence at a point or period but are weak for temporal sequence. Case-control studies compare prior exposure between people with and without an outcome and are efficient for rare outcomes. Cohort studies follow exposed and unexposed groups and can estimate incidence and relative risk.

Randomized trials allocate interventions by chance, reducing confounding when implemented well. Blinding, allocation concealment, follow-up, and intention-to-treat analysis affect credibility. Pragmatic trials ask how an intervention performs in routine care, while explanatory trials ask whether it works under controlled conditions.

Clinical correlation

The design determines what can reasonably be concluded. An association in a cross-sectional survey does not prove that an exposure caused the outcome. A case report can raise a safety signal but cannot estimate frequency.

Worked example

To study whether a dispensing reminder reduces omitted counselling, a controlled before-and-after or randomized design may be more informative than a case report. The outcome should be defined before implementation, such as documented counselling completion or patient understanding.

FAQs

Is a randomized trial always possible? No. Ethics, feasibility, rarity, or timing may require another design.

What is confounding? A third factor related to both exposure and outcome that distorts their association.

Takeaways

• Match design to the research question.
• Define variables before collecting data.
• Do not claim causation from a design that cannot support it.

Integrated practice note

The design should be justified in the protocol before data collection begins. If a question changes, the team should record whether the change is a refinement or a new analysis. A protocol also protects participants by limiting unnecessary procedures and clarifying who can access identifiable information. Researchers should state the limits of the design honestly so readers can use the result appropriately.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Hulley et al., Designing Clinical Research, latest available edition.

Unit 2: Research Methodology and Data Collection

Syllabus scope

This unit covers methodology design, variables, sampling, measurement, data quality, ethics, and preparation for analysis.

Methodology

A methodology converts a question into reproducible operations. It states population, setting, eligibility, sampling frame, sample size, exposure or intervention, comparator, outcomes, instruments, schedule, quality control, analysis plan, and ethical safeguards.

Method design, step by step

1. Define the primary outcome and its measurement scale.
2. Identify independent, dependent, confounding, and effect-modifying variables.
3. Select a sampling approach and recruitment pathway.
4. Pilot the questionnaire, case form, or procedure.
5. Train data collectors and standardize measurements.
6. Create a data dictionary and missing-data plan.
7. Protect confidentiality and obtain required approval.

Categorical variables represent groups. Ordinal variables have order but unequal intervals cannot be assumed. Continuous variables may take values on a scale. Operational definitions make abstract concepts measurable, such as defining adherence by a specific record or threshold.

Sampling and bias

Probability sampling gives a known selection process. Convenience sampling is easier but may reduce representativeness. Selection bias occurs when participation or retention differs systematically. Information bias occurs when measurement differs from reality or between groups. Training, blinding, standardized instruments, and quality checks reduce error.

Ethics and data governance

Ethical research requires scientific value, risk minimization, informed consent when applicable, privacy, confidentiality, fair selection, and independent review. Store only necessary data, restrict access, preserve an audit trail, and define retention. Follow the current approved institutional and jurisdictional requirements before collecting or sharing data.

Worked example

If adherence is measured only by patient memory, recall bias may be high. A stronger methodology combines a defined interview question with refill history or a validated scale, while recognizing that no measure is perfect.

FAQs

Why pilot a data-collection form? It reveals ambiguity, missing options, timing problems, and feasibility issues before the main study.

What is a data dictionary? A record defining each variable, format, coding, and allowed values.

Takeaways

• Make every variable and procedure operational.
• Plan bias control before data collection.
• Protect participants and data throughout the study.

Integrated practice note

Good methodology anticipates practical failure. Plan how participants will be contacted, how missing observations will be recorded, how adverse events will be escalated, and how data will be checked before analysis. A small pilot can test recruitment, timing, instrument clarity, and staff workload. These steps improve feasibility without pretending that a pilot has the power of the final study.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Council for International Organizations of Medical Sciences, ethical guidance.

Unit 3: Sample Size and Power

Syllabus scope

This unit covers sample size for simple comparative experiments, confidence intervals of specified width, and power of a study.

Core concepts

Sample size depends on design, outcome type, expected effect, variability or baseline risk, significance level, desired power, allocation ratio, and anticipated loss to follow-up. A larger sample generally narrows a confidence interval, but precision is not the same as validity.

Planning sequence

1. Define the primary outcome and clinically meaningful difference.
2. Estimate baseline risk or variability from credible evidence.
3. Select a significance level and desired power.
4. Choose the appropriate formula or validated software.
5. Allow for attrition, clustering, or design effects when relevant.
6. Document assumptions and perform sensitivity analysis.

Power is the probability of detecting a specified effect when that effect truly exists under the planning assumptions. A study can be underpowered because of a small sample, high variability, small effect, poor measurement, or attrition.

Confidence-interval width

For estimating a mean, required sample size increases as variability and the desired confidence level increase, and decreases as the tolerated margin of error increases. Proportion estimates similarly depend on expected proportion and precision. The final number must match the analysis method and design.

Worked example

Suppose a team wants a narrower confidence interval for a mean outcome. If all other assumptions stay constant, reducing the desired margin of error requires a larger sample because sampling uncertainty must be reduced. The exact number cannot be guessed without variability, confidence level, and design assumptions.

Clinical correlation

An unrealistic sample-size calculation can waste resources or expose participants without a reasonable chance of answering the question. The pharmacist should check whether the proposed outcome is measurable and whether the assumed difference matters to patients.

FAQs

Does statistical power prove clinical importance? No. It concerns detection probability, not patient value.

Can a larger sample fix bias? No. It may make a biased estimate more precise.

Takeaways

• Define a clinically meaningful effect before calculating.
• Report assumptions transparently.
• Account for loss, clustering, and design.
• Use validated software or statistical consultation when needed.

Integrated practice note

Power calculations should be revisited when assumptions are uncertain. Present a range of plausible baseline risks, variability, effect sizes, and attrition rather than one falsely exact number. A statistician can help identify whether the proposed primary analysis, repeated measurements, cluster structure, or missing-data strategy changes the calculation. The final protocol should preserve the approved assumptions.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Chow, Shao and Wang, Sample Size Calculations in Clinical Research, latest available edition.

Unit 4: Report Writing and Data Presentation

Syllabus scope

This unit covers report writing and presentation of data.

Scientific report structure

A report usually includes title, abstract, introduction, methods, results, discussion, limitations, conclusion, references, and supplementary material where needed. The methods must permit reproducibility. The results should present analyses planned or clearly label exploratory analyses. The discussion should distinguish data from interpretation.

Reporting sequence

1. State the question and rationale.
2. Describe population, design, variables, methods, and ethics.
3. Present participant flow and missing data.
4. Report estimates with appropriate uncertainty.
5. Explain clinical meaning without overstating causation.
6. State limitations, conflicts, and funding.
7. Provide a conclusion supported by the results.

Tables should have clear titles, units, denominators, and footnotes. Graphs should use honest scales, readable labels, and a display appropriate to the data. Avoid decorative three-dimensional effects and unnecessary precision.

Data integrity

Keep a versioned analysis record, preserve raw data, document exclusions, and distinguish prespecified from post hoc analyses. Do not remove inconvenient observations without a defensible rule. Protect identifiable data in reports and presentations.

Worked example

A report says treatment “reduced risk by 50%” but gives no baseline risk. The pharmacist requests absolute event rates, follow-up, confidence interval, and harms before presenting the result to clinicians.

Clinical correlation

Poor presentation can change decisions. A truncated axis may exaggerate a small difference, while a missing denominator can make a proportion misleading. A clear table may be safer than a visually impressive but ambiguous graph.

FAQs

Should every result be displayed? Include results relevant to the question and prespecified analyses, with transparent reporting of important deviations.

Why report limitations? Readers need to judge confidence and applicability.

Takeaways

• Make methods reproducible.
• Show denominators, units, uncertainty, and missing data.
• Separate findings from interpretation.

Integrated practice note

Readers should be able to trace a conclusion back to a table or analysis. Use consistent denominators, define exclusions, and avoid reporting more decimal places than the measurement supports. If an analysis was not planned, label it exploratory. Conclusions should match the population studied, the follow-up completed, and the uncertainty shown by the data.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. EQUATOR Network reporting guidelines.

Unit 5: Data Distributions and Descriptive Statistics

Syllabus scope

This unit covers introduction to biostatistics, types of data distributions, mean, median, mode, range, variation, standard deviation, variance, coefficient of variation, and standard error of the mean.

Describing data

Descriptive statistics summarize observed data. The mean uses every value but is sensitive to extreme observations. The median is the middle ordered value and is more robust to skew. The mode is the most frequent value. Range is maximum minus minimum and is highly influenced by extremes.

Variance is the average squared deviation according to the chosen sample or population convention. Standard deviation is the square root of variance and describes spread around the mean in the original units. The coefficient of variation is standard deviation divided by the mean, commonly expressed as a percentage, and helps compare relative variability when a meaningful zero exists. Standard error describes uncertainty in an estimated mean, not the spread of individual observations.

Worked example

For values 2, 3, and 7, the mean is 4, the median is 3, and the range is 5. The mean is pulled upward by 7, which illustrates why the median may better represent a skewed small sample. Exact variance depends on whether a sample or population formula is used.

Distributions

A distribution describes how values occur. Symmetry, skew, spread, outliers, and multimodality influence summaries and test choice. The normal distribution is continuous, symmetric, and described by its mean and standard deviation, but real clinical data may not be normal.

Clinical correlation

Reporting only a mean can hide unequal patient experiences. A pharmacist reviewing treatment duration should inspect spread and outliers, not just average duration.

FAQs

Is standard error a measure of patient variability? No. Standard deviation describes data spread; standard error describes precision of an estimated mean.

When is the median useful? For skewed data or data with influential extreme values.

Takeaways

• Choose summaries that fit the distribution.
• Distinguish variability from estimation precision.
• Report units and sample size with summaries.

Integrated practice note

Before summarizing, inspect the raw values and the measurement process. An extreme value may be a true clinical observation, a transcription error, or a unit mismatch. Do not delete it because it is inconvenient. Verify the source, correct documented errors, and report the rule used. Descriptive statistics are only as trustworthy as the data they summarize.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Motulsky, Intuitive Biostatistics, latest available edition.

Unit 6: Data Graphics

Syllabus scope

This unit covers construction and labelling of graphs, histograms, pie charts, scatter plots, and semilogarithmic plots.

Choosing a graph

Bar charts compare categories. Histograms display the distribution of continuous or grouped numerical data using adjacent bins. Pie charts show parts of a whole only when categories are mutually exclusive and the number of categories is small. Scatter plots display the relationship between two numerical variables. Semilogarithmic plots use a logarithmic scale on one axis and may help show exponential change or concentration-response patterns.

Construction sequence

1. Define the data type and question.
2. Select a graph that preserves the relationship.
3. Label axes, units, categories, and time.
4. Choose a scale that does not mislead.
5. Add a clear title and legend where needed.
6. Check readability, denominators, missing values, and outliers.

Interpretation

A scatter plot can show association but not causation. A trend line summarizes a model and should not imply that every observation follows it. A histogram’s appearance changes with bin width. A pie chart can become unreadable when slices are numerous.

Worked example

To examine whether adherence score relates to blood pressure, use a scatter plot with adherence on one axis and blood pressure on the other, include units and patient-level points, and avoid concluding causation from a visible trend alone.

Clinical correlation

Graph design can alter clinical interpretation. Truncated axes exaggerate differences, log axes can confuse readers if not explained, and missing denominators can turn a small sample into a misleading percentage.

FAQs

Why use a histogram instead of a bar chart? Histogram bins represent continuous intervals and touch; bars represent separate categories.

When is a logarithmic scale useful? When values span orders of magnitude or a multiplicative pattern is being examined.

Takeaways

• Match graph to data and question.
• Label units, scale, denominators, and uncertainty.
• Use visual displays to clarify, not decorate.

Integrated practice note

Every graph should answer a question. Decide whether the reader needs comparison, distribution, trend, relationship, or composition. Use a caption to explain the population, timeframe, units, and important exclusions. Check the final display at the size in which it will be read, because labels that are clear on a large monitor may disappear in a report or presentation.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Kirkwood and Sterne, Essential Medical Statistics, latest available edition.

Unit 7: Hypothesis Testing and Parametric Methods

Syllabus scope

This unit covers null hypothesis, significance, power, p value, confidence intervals, paired and unpaired t tests, chi-square test, and one-way and two-way analysis of variance.

Testing logic

The null hypothesis states a defined absence of difference or association. A p value describes how compatible the observed data are with the null model and assumptions. It is not the probability that the null hypothesis is true. A confidence interval describes estimation uncertainty under the method. Statistical significance does not establish clinical importance.

Test-selection sequence

1. Define outcome scale and comparison.
2. Identify independence or pairing.
3. Check distribution and variance assumptions.
4. Choose a prespecified test.
5. Report estimate, confidence interval, p value, and clinical context.

The paired t test analyzes differences within matched pairs or repeated measurements. The unpaired t test compares means between independent groups under its assumptions. Chi-square tests association between categorical variables when expected counts and independence conditions are reasonable. One-way ANOVA compares means across more than two groups; two-way ANOVA examines two factors and possible interaction.

Worked example

If the same patients have blood pressure measured before and after an intervention, the observations are paired. A paired analysis uses each patient’s change rather than treating the two measurements as unrelated. The clinical conclusion still requires the size, confidence interval, harms, and relevance of the change.

Errors and multiplicity

A type I error rejects a true null hypothesis. A type II error fails to detect a specified effect. Multiple testing increases false-positive risk unless addressed in the design and analysis. Post hoc test selection can produce misleading certainty.

Clinical correlation

A small p value can accompany a trivial effect in a large sample. A non-significant result can reflect imprecision. The pharmacist should ask what decision would change and whether the result meets a clinically meaningful threshold.

FAQs

Does p less than 0.05 prove effectiveness? No.

Why distinguish paired and unpaired tests? The dependence structure changes the uncertainty calculation.

Takeaways

• Select tests from design and data structure.
• Report effect size and confidence interval, not p value alone.
• Interpret statistical results alongside clinical importance.

Integrated practice note

A statistical test is part of a reasoning chain, not the endpoint. State the clinical question, estimate the difference, show uncertainty, explain assumptions, and describe how the result affects a decision. If several outcomes or subgroups were examined, disclose that fact. A transparent negative or uncertain result is more useful than a confident claim unsupported by the design.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Altman, Practical Statistics for Medical Research, latest available edition.

Unit 8: Nonparametric Methods, Correlation and Regression

Syllabus scope

This unit covers sign test, Wilcoxon signed-rank test, Wilcoxon rank-sum test, Mann Whitney U test, Kruskal-Wallis test, linear regression, Pearson and Spearman correlation, and correlation coefficients.

Nonparametric tests

Nonparametric methods may be useful for ordinal, skewed, or small-sample data when parametric assumptions are unsuitable. The sign test uses direction of paired differences. Wilcoxon signed-rank uses ranks and directions of paired differences. Wilcoxon rank-sum and Mann Whitney U compare independent groups through ranks. Kruskal-Wallis compares more than two independent groups, followed by an appropriate post hoc approach when indicated.

Selection mechanism

1. Determine whether observations are paired or independent.
2. Define the outcome’s measurement level.
3. Examine distribution and outliers.
4. Choose a method before inspecting outcome significance.
5. Report the effect and uncertainty in a clinically understandable form.

Correlation and regression

Pearson correlation measures linear association between numerical variables under assumptions. Spearman correlation uses ranks and can describe monotonic association. Neither proves causation. Linear regression models an outcome as a function of one or more predictors and estimates an average change, conditional on the model.

Worked example

If dose and response show a curved relationship, a Pearson correlation may understate association and a simple linear model may fit poorly. Plot the data, check residuals, consider a justified transformation or alternative model, and avoid forcing a straight line.

Clinical correlation

A strong association can be confounded, and a weak association can still matter if the relationship is nonlinear or clinically important in a subgroup. Always inspect the data and study design.

FAQs

Are nonparametric tests assumption free? No. They still require appropriate independence, measurement, and interpretation assumptions.

Does correlation predict an individual patient exactly? No. It summarizes a population relationship and may have substantial individual error.

Takeaways

• Match nonparametric tests to pairing and group number.
• Plot data before fitting correlation or regression.
• Never equate association with causation.

Interpretation should state the population, direction, magnitude, uncertainty, and important assumptions.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Kirkwood and Sterne, Essential Medical Statistics, latest available edition.

Unit 9: Statistical Software and Epidemiology

Syllabus scope

This unit covers SPSS, Epi Info, SAS, incidence, prevalence, relative risk, and attributable risk.

Software workflow

Software performs calculations but cannot repair a poorly defined question, miscoded variable, biased sample, or inappropriate test. A reproducible workflow includes data dictionary, import check, cleaning log, analysis script or syntax where available, output review, and secure storage. SPSS, Epi Info, and SAS have different interfaces and capabilities. Check the currently supported version and institutional access before beginning a project.

Analysis sequence

1. Inspect raw data and coding.
2. Check ranges, duplicates, missingness, and impossible values.
3. Describe the sample.
4. Run the prespecified analysis.
5. Check assumptions and sensitivity.
6. Export tables and figures with labels.
7. Preserve syntax, output, and interpretation.

Epidemiological measures

Incidence describes new events in a population at risk over time. Prevalence describes existing cases at a point or period. Relative risk compares risk in exposed and unexposed groups. Attributable risk is the excess risk associated with exposure under the comparison assumptions.

Worked example

If 10 of 100 exposed people develop an outcome and 5 of 100 unexposed people do, risk is 0.10 versus 0.05, relative risk is 2, and attributable risk is 0.05, or five excess events per 100 over the study period. This does not prove causation and depends on valid measurement and comparison.

Clinical correlation

Prevalence may rise because incidence increases, survival improves, or case detection changes. A relative risk can be large while absolute excess risk is small. Explain both when supporting a health-system decision.

FAQs

Does software output decide the method? No. Study design and data determine the method.

Why preserve syntax or an analysis log? It improves reproducibility and auditability.

Takeaways

• Validate data before analysis.
• Interpret incidence, prevalence, and risk measures precisely.
• Preserve a reproducible analysis trail.

Always inspect output against the original question before sharing it.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. Rothman, Greenland and Lash, Modern Epidemiology, latest available edition.

Unit 10: Computer Applications in Pharmacy

Syllabus scope

This unit covers computer systems in hospital pharmacy, community pharmacy, drug-information retrieval and storage, patient records, medication orders, labels, intravenous admixtures, inventory, reports, dispensing, accounting, and pharmaceutical care.

Hospital applications

Hospital systems may support patient-record databases, computerized medication-order entry, clinical alerts, drug labels and lists, intravenous admixture records, medication profiles, inventory, management reports, and statistics. Safe implementation requires user identification, access control, data validation, downtime procedures, audit trails, interoperability, training, and review of alert burden.

Digital medication workflow

1. Verify patient identity and encounter.
2. Enter or receive the medication order.
3. Apply decision support and pharmacist review.
4. Generate preparation, dispensing, and administration information.
5. Record the event and monitoring data.
6. Audit access, changes, errors, and outcomes.

Community pharmacy applications

Computerized dispensing can support prescription entry, stock checks, labels, interaction screening, patient profiles, accounting, general ledger, and pharmaceutical-care documentation. Automation increases consistency only when data are accurate and the pharmacist reviews exceptions.

Drug-information retrieval and storage

Electronic resources improve speed, searching, updating, and storage, but access does not guarantee quality. Record the source, version or date, search terms, answer, and patient-specific interpretation. Protect confidential queries and restrict access.

Worked example

An interaction alert appears for two medicines. The pharmacist checks the alert’s evidence, dose, timing, patient factors, and clinical significance rather than accepting or dismissing it automatically. The final decision and rationale are documented.

Clinical correlation

Poor interface design can create wrong-patient selection, alert fatigue, copy-forward errors, and unsafe defaults. Computerization is a safety tool only when combined with human review and downtime planning.

FAQs

Can software replace pharmacist review? No. It detects patterns but may lack patient context and can generate false alerts.

Why are audit trails important? They support accountability, security, investigation, and improvement.

Takeaways

• Validate data and patient identity at every digital step.
• Review alerts clinically.
• Protect privacy, access, and continuity during downtime.

Review digital systems regularly for safety today.

References

1. PharmacoLearn, Pharm D Year 4 syllabus: https://pharmacolearn.fyi/syllabus/pharm-d/year-4/
2. Pharmacy Council of India, Pharm.D. Regulations, 2008.
3. World Health Organization, digital health and medication-safety resources.