Observational and Single-Subject Research Designs in Occupational Therapy

Design Choice: Explain the criteria used to choose between observational designs and single-subject designs.
Core Features: Identify the fundamental design features, including sampling, phases, and data analysis methods.
Validity Risks: Recognize common threats to study validity and the methods used to address them.
NBCOT Application: Apply research concepts to National Board for Certification in Occupational Therapy (NBCOT) style exam scenarios.

Definition of "Observational": These studies involve no manipulation of variables. Data is collected within natural contexts rather than controlled environments.
Categorization:
- Descriptive: Asks "what is happening?"
- Exploratory: Asks "what relates to what?"
Utility: Highly effective for early-stage research questions or in scenarios where experimental manipulation would be unethical or impractical.

Prospective (Forward):
- Mechanism: Collects data forward in real time.
- Strengths: More control over variables, clear temporal sequence of events.
- Weaknesses: High risk of attrition (participants dropping out).
Retrospective (Backward):
- Mechanism: Mines existing records or databases (looking back at historical data).
- Strengths: Cheaper, faster to execute, and allows for much larger sample sizes ( $N$ ).
- Weaknesses: Threat of missing or inconsistent data within records.

Prospective Case: A study by Pineda et al. (2020) compared a new assessment to the Neonatal Oral Motor Assessment Scale (NOMAS). This was used to establish concurrent validity.
Retrospective Case: Yokoi et al. (2025) conducted "Profile of Independence in Activities of Daily Living Among Patients With Parkinson's Disease: A Retrospective Observational Study."
- Setting: Hospital in Japan.
- Participants: $N = 209$ ; 75 men and 134 women with a mean age of $73.3\text{ yr}$ (SD = $7.7$ ).
- Measures:
- Disease Severity: Hoehn and Yahr (H-Y) stage.
- ADL Independence: Barthel Index (BI).
- Detailed Results (BI Scores Across H-Y Stages):
- H-Y Stage 2 ( $n = 55$ ): Total BI Median = $90$ . High independence across most ADLs except mobility and stairs.
- H-Y Stage 3 ( $n = 98$ ): Total BI Median = $85$ . Limitations appear in bathing, mobility, and stairs.
- H-Y Stage 4 ( $n = 40$ ): Total BI Median = $55$ . Significant assistance required except for feeding and bowel control.
- H-Y Stage 5 ( $n = 16$ ): Total BI Median = $45$ . High dependence, though feeding ( $5$ ) and grooming ( $0-5$ ) may show relative preservation of some ability.
- Statistical Significance: Results across stages showed significant differences (p < .001) via Kruskal-Wallis H test and Dunn\'s test with Bonferroni correction.

Cross-Sectional: Captures data in one snapshot. It is efficient but carries a high risk of selection bias.
Longitudinal: Involves repeated measures over time. This design can suggest causal order but faces threats from attrition and testing effects.
Correlational Study: Measures the strength and direction of association between variables, typically denoted by the correlation coefficient ( $r$ ).
Predictive Study: Utilizes regression analysis to forecast outcomes (e.g., using FIM scores to predict Length of Stay/LOS).
Important Caveat: Correlation does not equal causation; establishing causation requires specific theory and further controlled testing.

Case-Control Design:
- Start Point: Outcome is present (e.g., they have the disease).
- Direction: Looks backward to identify exposures.
- Primary Use: Studying rare diseases.
Cohort Design:
- Start Point: Exposure is present (e.g., they were exposed to a risk factor).
- Direction: Looks forward to see if outcomes develop.
- Primary Use: Studying rare exposures or multiple outcomes from one exposure.

Establishing Causality: Researchers must establish a temporal sequence, biological plausibility, and a dose-response relationship.
Common Biases:
- Selection Bias: Inequities in how participants are chosen.
- Recall Bias: Inaccuracies in participant memory (common in retrospective studies).
- Interviewer Bias: Influence of the researcher on participant responses.
- Confounding: Other variables affecting the outcome.
Solutions/Fixes:
- Matching participants.
- Blinding chart reviewers.
- Statistical adjustments.
- Triangulation (using multiple data sources).

Purpose: Patient-level investigation design meant to inform Evidence-Based Practice (EBP) for clinicians.
Nomenclature Note: The name is slightly misleading; SSD refers to studies with very small sample sizes (typically $n = 1$ to $n = 10$ ).
Ideal Scenarios: Low-incidence diagnoses or highly individualized clinical goals.
Key Methodology:
- Repeated, within-person comparisons.
- Independent Variable (IV): The intervention.
- Dependent Variable (DV): The target behavior (measured by frequency, duration, or magnitude).

Repeated Measures: Requires at least $3-4$ data points per phase to establish a trend.
Phases:
- Baseline (A): Data collected before the intervention.
- Intervention (B): Data collected during the intervention.
- Visualization: All data points must be graphed to observe trends and changes across phase lines.

A-B Design: Moving from Baseline (A) to Treatment (B).
- Pro: Provides quick clinical feedback.
- Con: Weak internal validity; impossible to claim causation because there is no control comparison.
- Strengthening: Repeat phases after a return to baseline, test on more than one subject, or compare multiple interventions.
Withdrawal Designs:
- A-B-A: A reversal design that adds a second baseline to see if behaviors reverse without treatment.
- A-B-A-B: The "gold standard" for reversible behaviors. It provides two separate opportunities to see the treatment effect.
Alternating and Multiple Treatment Designs:
- Alternating-Treatments: A rapid switch between treatments (e.g., sensory diet vs. video modeling). Often used to compare two different interventions to see which is more effective.
- Multiple Treatment (A-B-C-A): Compares two interventions sequentially after a baseline. This involves applying one treatment (B), withdrawing it, and introducing one or more additional treatments (C).

Direct Replication: Repeating the same design with the same type of participants. At least $3$ successful replications are needed to build confidence.
Systematic Replication: Varying one element of the study (e.g., a different setting or a different therapist).
Goal: Builds evidence to show that findings can generalize beyond the original specific case.

Design Fit: Always match the research question to the design.
Threat Recognition: Identify specific threats (e.g., attrition is a primary threat in longitudinal studies).
Graph Analysis: In SSD graphs, look for distinct changes occurring at the phase lines to attribute effects to the intervention.
Dual Utility: Observational studies are used to inform broad hypotheses, while SSDs are used to inform individual care plans.