31.2 The Physician as Researcher I: The Research Process and Proposal, Research Ethics, Study Designs and Meta-Analysis, Demography and Vital Statistics, Evidence-Based Medicine and Quality Assurance

Why this area matters

The TOS researcher competencies are to:

• identify the basic research process;
• compute rates and ratios;
• describe the demographic status of a community;
• apply evidence-based medicine to research findings;
• create a research proposal from a scenario.

Questions ask you to choose the right design for a research question, recognize bias and confounding, compute vital rates, or rank evidence.

The research process and proposal

1. Identify the problem and state the research question. A useful framework is PICO: population, intervention or exposure, comparison, outcome.
2. Review the literature to see what is known and justify the study.
3. Set the objectives (general and specific) and the hypotheses (null and alternative).
4. Choose the methods:
   • the study design;
   • the setting and population;
   • sampling (probability methods such as simple random, systematic, stratified and cluster sampling, or non-probability methods such as convenience and purposive sampling);
   • the sample size (from the expected effect, alpha, power and variability);
   • variables and their operational definitions;
   • data collection tools;
   • the data management and analysis plan.
5. Obtain ethics review and informed consent.
6. Collect and analyze the data, interpret it, and disseminate the results.

The proposal also includes a budget, timeline (Gantt chart) and references.

Research ethics in the Philippines:

• Research with human participants must be reviewed by an ethics review committee accredited by the Philippine Health Research Ethics Board (PHREB).
• Researchers follow the National Ethical Guidelines for Research Involving Human Participants.
• The principles:
  • respect for persons, through informed consent and protection of vulnerable groups;
  • beneficence and non-maleficence;
  • justice;
  • privacy and confidentiality, consistent with the Data Privacy Act (RA 10173).

The Hierarchy of Clinical Evidence and Study Designs

Medical evidence is structured in an established hierarchy reflecting internal validity and resistance to systematic bias.

Observational Study Architectures

1. Ecological (Correlational) Studies:

   • Unit of Observation: Entire populations or geographic regions, rather than individual patients (e.g., comparing country-wide dietary fat consumption against national breast cancer incidence).
   • Major Pitfall (Ecological Fallacy): The invalid inference that an association observed at the aggregate population level necessarily holds true for individual members of that population.

2. Cross-Sectional Studies (Prevalence Surveys):

   • Design: Exposure status and disease presence are assessed simultaneously at a single cross-section in time.
   • Metrics: Quantifies point or period prevalence and odds ratios.
   • Limitations: Cannot establish temporal sequence (whether exposure preceded disease onset); susceptible to Neyman bias (prevalence-incidence bias), in which rapidly fatal or rapidly resolving cases are systematically missed.

3. Case-Control Studies:

   • Design: Observational and retrospective. Subjects are identified and enrolled based on the presence of disease (Cases) or absence of disease (Controls). Investigators then look backward in time to determine past exposure histories.
   • Metrics: Quantified by the Odds Ratio (OR); cannot calculate true population incidence or Relative Risk directly.
   • Advantages: Highly efficient and cost-effective for investigating rare diseases (e.g., rare neoplasms) or diseases with long latency periods.
   • Vulnerabilities: Highly prone to recall bias (cases recall past exposures differently than healthy controls) and selection bias in choosing control groups.

4. Cohort Studies:

   • Design: Observational, prospective or retrospective. Subjects are identified and enrolled based on exposure status (Exposed vs. Unexposed), must be verified free of the disease outcome at study entry, and are followed forward through time to measure the emergence of incident disease.
   • Metrics: Quantifies Incidence Rates, Relative Risk (RR), and Attributable Risk (AR).
   • Advantages: Ideal for rare exposure factors (e.g., occupational chemical exposures); establishes definitive temporal sequence between exposure and disease.
   • Limitations: Inefficient for rare outcomes; prolonged follow-up is resource-intensive and vulnerable to attrition bias (loss-to-follow-up).

Experimental Designs: Randomized Controlled Trials (RCTs)

The Randomized Controlled Trial represents the gold standard for establishing causal therapeutic efficacy:

• Randomization: Balances known and unknown baseline prognostic confounders equally between experimental and control cohorts.
• Blinding (Masking):
  • Single-blind: Participant is unaware of treatment assignment (prevents placebo effect).
  • Double-blind: Participant and treating clinician/assessor are unaware (prevents observer and reporting bias).
  • Triple-blind: Participant, clinician, and data monitoring/statistical committee are unaware.
• Intention-to-Treat (ITT) vs. Per-Protocol Analysis:
  • Intention-to-Treat (ITT): Analyzes all randomized participants in the group to which they were originally allocated, regardless of non-compliance, protocol deviations, or premature treatment discontinuation. Primary virtue: Preserves the baseline prognostic balance generated by randomization, prevents attrition bias, and provides a conservative estimate of real-world clinical effectiveness.
  • Per-Protocol (PP): Restricts analysis exclusively to participants who fully adhered to the prescribed study protocol. Virtue: Evaluates biological efficacy under perfect compliance; fatal flaw: destroys randomization and introduces selection bias.

---

Systematic Reviews and Meta-Analyses

Meta-analyses statistically combine quantitative data across multiple independent primary studies to generate a pooled, high-power summary effect estimate.

Forest Plot Anatomy

A Forest Plot visually synthesizes meta-analytic data:

• Individual Studies: Represented by horizontal lines with a central square.
  • The square's location denotes the study's point estimate of effect (RR, OR, or mean difference).
  • The area of the square is proportional to the study's statistical weight (dictated by sample size and inverse variance).
  • The horizontal line represents the 95% Confidence Interval (95% CI).
• Line of No Effect: A vertical reference line placed at 1.0 for ratio metrics (Relative Risk, Odds Ratio, Hazard Ratio) or 0.0 for difference metrics (absolute risk difference, mean differences). If a study's horizontal 95% CI crosses this vertical line, its finding is not statistically significant at $p < 0.05$.
• Summary Estimate (The Diamond): Positioned at the bottom of the plot. The center of the diamond represents the pooled summary effect estimate, and its lateral tips represent the 95% CI of the pooled estimate. If the diamond does not intersect the line of no effect, the overall meta-analytic result is statistically significant.
• Heterogeneity: Quantified by the $I^2$ statistic. An $I^2 > 50\%$ indicates substantial between-study statistical heterogeneity, warranting a random-effects model rather than a fixed-effects model.

Funnel Plots and Publication Bias

• A Funnel Plot is a scatter plot displaying treatment effect estimates on the horizontal axis against a measure of study precision (sample size or standard error) on the vertical axis.
• In the absence of bias, the plot resembles a symmetrical inverted funnel, with larger, highly powered studies clustering tightly at the apex and smaller studies scattering widely at the base.
• Asymmetry (Missing Studies): An asymmetric funnel plot—specifically showing a void or absence of small studies in the quadrant of unfavorable or non-significant results—signals publication bias (the tendency of investigators, journals, and sponsors to leave negative or null studies unpublished in file drawers). Begg's adjusted rank correlation and Egger's linear regression tests provide formal mathematical quantification.

Bias and confounding

• Selection bias: systematic differences in how subjects are chosen. Examples are hospital-based controls, loss to follow-up and the healthy worker effect.
• Information (measurement) bias:
  • recall bias in case-control studies, where cases remember exposures better;
  • interviewer bias;
  • misclassification.
• Confounding: a third variable is linked to both exposure and outcome and distorts the association. For example, an apparent link between coffee and lung cancer is confounded by smoking.
  • At the design stage: randomization, restriction and matching.
  • At the analysis stage: stratification and multivariable regression.
• Effect modification: the association differs across levels of a third variable. It is reported, not removed.
• Validity:
  • internal validity means the results are true for the study population;
  • external validity (generalizability) means they apply to other populations.

Demography and vital statistics

• Demography is the study of the size, composition and distribution of populations and of changes through births, deaths and migration.
• Sources of data:
  • the census: the 2020 Census counted about 109 million Filipinos;
  • civil registration (births, deaths, marriages);
  • surveys such as the National Demographic and Health Survey;
  • health facility records.

• Population pyramids:
  • an expansive (broad-based) pyramid shows high fertility and a young, growing population;
  • a constrictive pyramid shows falling fertility;
  • a stationary pyramid shows low birth and death rates.
  • The Philippine pyramid is becoming narrower at the base as fertility falls, so the population is aging.
• Demographic transition: populations move from high births and deaths, through falling deaths, to falling births and slow growth. The epidemiologic transition shifts the disease burden from infections to NCDs.

Evidence-based medicine

• Evidence-based medicine integrates the best research evidence with clinical expertise and patient values and circumstances.
• The steps (5 A's):
  1. ask a focused question (PICO);
  2. acquire the evidence;
  3. appraise it for validity, importance and applicability;
  4. apply it to the patient;
  5. assess the outcome.
• The hierarchy of evidence for therapy, from strongest to weakest:
  • systematic reviews and meta-analyses of RCTs;
  • RCTs;
  • cohort studies;
  • case-control studies;
  • case series;
  • expert opinion.
  • GRADE rates the certainty of evidence (high, moderate, low or very low) and the strength of recommendations.
• Critical appraisal of an RCT asks:
  • was allocation randomized and concealed?
  • were groups similar at baseline?
  • was it blinded?
  • was follow-up complete?
  • was it analyzed by intention to treat?
  • Report results as absolute and relative risk reduction, NNT and confidence intervals.

Quality assurance

• Donabedian's framework:
  • structure: facilities, equipment, staff;
  • process: what is done, and whether care follows guidelines;
  • outcome: results, such as mortality, infection rates and patient satisfaction.
• Methods:
  • clinical audit, which compares practice with a standard, changes it and re-audits;
  • continuous quality improvement with the plan-do-check-act (PDSA/PDCA) cycle;
  • licensing and accreditation of facilities (DOH licensing; PhilHealth accreditation);
  • clinical practice guidelines;
  • mortality and morbidity reviews, including maternal and perinatal death reviews;
  • patient safety programs.