2.5 Data Collection, Presentation, Visualization & Statistical Analysis

2.5 Data Collection, Presentation, Visualization & Statistical Analysis

Transforming raw surveillance counts into actionable clinical knowledge requires mastery of data visualization tools and statistical analysis. Infection Preventionists must present data clearly to multidisciplinary quality committees, distinguish random process variation from true clinical degradation, and apply inferential statistics to determine whether observed changes in infection rates are statistically meaningful.

---

1. Data Presentation & Visualization Tools

Selecting the correct chart type depends on the underlying measurement scale and analytical purpose:

Run Charts vs. Control Charts

• Run Chart: A line graph displaying continuous data points plotted over time (e.g., monthly infection counts over 12 months). A horizontal center line representing the median or mean is included. Run charts display temporal trends, shifts, and cycles without formal statistical control limits.
• Control Chart (Statistical Process Control / SPC): A run chart enhanced with mathematically calculated upper and lower statistical boundaries:
  • Center Line (CL): Represents the historical process mean or median.
  • Upper Control Limit (UCL): Positioned $3 \text{ standard deviations } (+3\sigma)$ above the center line.
  • Lower Control Limit (LCL): Positioned $3 \text{ standard deviations } (-3\sigma)$ below the center line.

  UCL  -------------------------------------------------- (+3 Sigma)

  CL   ────────────────────────────────────────────────── (Process Mean)

  LCL  -------------------------------------------------- (-3 Sigma)

Types of Control Charts in Infection Prevention

• p-Chart (Proportion Chart): Used for attribute data expressed as proportions or percentages (e.g., monthly hand hygiene compliance rates, proportion of surgical patients receiving timely prophylactic antibiotics).
• u-Chart (Rate Chart): Used for rate data where the denominator varies (e.g., monthly CLABSI rates per $1,000$ central line-days).
• c-Chart (Count Chart): Used for absolute defect counts when the sample size/denominator remains constant (e.g., monthly needle-stick injury counts).

Interpreting SPC Charts: Variation Types

1. Common Cause Variation: Inherent, natural background noise expected within a stable healthcare process. Points fluctuate randomly between the UCL and LCL without distinct patterns. Requires systemic process redesign to improve.
2. Special Cause Variation: Non-random, assignable variation caused by an external factor or process failure. Indicated by SPC rule breaches:
   • Any single data point falling outside the UCL or LCL boundaries.
   • Shift: 8 consecutive data points falling entirely above or below the center line.
   • Trend: 6 consecutive data points continuously increasing or continuously decreasing.

Bar Graphs vs. Histograms

• Bar Graph: Displays categorical/discrete variables (e.g., infection counts by clinical specialty: ICU, Stepdown, Med-Surg). Distinct spaces exist between bars to reflect discrete, non-continuous categories.
• Histogram: Displays continuous quantitative numerical variables (e.g., patient age distribution, duration of mechanical ventilation). Bars touch with no spaces, illustrating continuous frequency distributions.

---

2. Descriptive Statistics: Central Tendency & Dispersion

Descriptive statistics summarize key characteristics of a dataset:

Measures of Central Tendency

• Mean (Arithmetic Average): Sum of all values divided by total count ($N$). Sensitive to extreme outliers; best suited for symmetric, normally distributed data.
• Median (Middle Value): The 50th percentile value when data are ranked sequentially. Robust against extreme outliers; preferred for skewed distributions (e.g., post-operative length of stay).
• Mode: The most frequently occurring value in a dataset.

Distribution Skewness Pearl:

• Symmetrical (Normal) Distribution: $\text{Mean} = \text{Median} = \text{Mode}$.
• Positively (Right) Skewed: $\text{Mean} > \text{Median} > \text{Mode}$ (tail extends right toward higher values).
• Negatively (Left) Skewed: $\text{Mean} < \text{Median} < \text{Mode}$ (tail extends left toward lower values).

Measures of Dispersion

• Standard Deviation (SD / $\sigma$): Quantifies the spread of data points around the mean. In a normal distribution:
  • $\pm 1\text{ SD}$ encompasses $68.27%$ of data points.
  • $\pm 2\text{ SD}$ encompasses $95.45%$ of data points.
  • $\pm 3\text{ SD}$ encompasses $99.73%$ of data points (foundation of SPC control limits).

---

3. Inferential Statistics & Significance Testing

Inferential statistics allow IPs to draw conclusions about a broader population based on sample data.

The Hypothesis Testing Framework

• Null Hypothesis ($H_0$): Assumes no true difference, association, or effect exists between groups (any observed difference is due purely to random chance).
• Alternative Hypothesis ($H_1$): Assumes a true, non-random difference or association exists.
• p-Value: The probability of obtaining a sample result as extreme as (or more extreme than) the observed data, assuming the null hypothesis is true. By convention, $\alpha = 0.05$.
  • $p < 0.05$: Reject $H_0$. Finding is statistically significant.
  • $p \ge 0.05$: Fail to reject $H_0$. Finding is not statistically significant.

95% Confidence Intervals (95% CI)

A Confidence Interval provides a range of plausible values for the true population parameter with a specified level of certainty ($95%$).

Critical CBIC Rules for Interpreting 95% CIs:

1. Ratio Metrics (Relative Risk, Odds Ratio, SIR):
   • If the 95% CI includes 1.00 (e.g., $\text{RR} = 1.80$, $95% \text{ CI: } 0.90 - 3.40$), the result is NOT statistically significant ($p \ge 0.05$).
   • If the 95% CI excludes 1.00 (e.g., $\text{RR} = 2.40$, $95% \text{ CI: } 1.30 - 4.20$), the result IS statistically significant ($p < 0.05$).
2. Difference Metrics (Difference between two rates or means):
   • If the 95% CI includes 0.00, the difference is NOT statistically significant.

Statistical Tests Matrix