====== CUSUM Analysis for Intracranial Pressure Monitor Placement ======

===== Introduction =====

[[Cumulative Sum]] ([[CUSUM]]) analysis is a well-established [[statistical method]] used in [[surgical performance monitoring]]. It provides a sensitive and continuous [[assessment]] of [[deviation]] from a predefined [[outcome target]]. In neurosurgery, CUSUM is particularly useful for [[learning curve evaluation]], detecting changes in [[complication rate]]s, and auditing [[surgical quality]].

This article presents a practical example of applying CUSUM analysis to [[intracranial pressure monitor placement]], focusing on [[identifying trends in procedural success and complication rates over time]].

===== Objective =====

To assess [[performance consistency]] and detect [[potential learning curve]]s or [[learning curve deviation]]s in complication rates during ICP monitor (e.g. bolt or EVD) placement procedures.

===== Methods =====

==== Dataset ====

Twenty consecutive cases of ICP monitor placements were analyzed. The [[primary outcome]] was binary:  
  * ''0'' = [[Success]] (no complication or malposition)  
  * ''1'' = [[Failure]] (complication, including [[infection]], [[hemorrhage]], or [[malposition]])

^ Case # ^ Date       ^ Outcome ^
| 1      | 2025-01-03 | 0       |
| 2      | 2025-01-04 | 1       |
| 3      | 2025-01-06 | 0       |
| 4      | 2025-01-07 | 0       |
| 5      | 2025-01-09 | 0       |
| 6      | 2025-01-10 | 1       |
| 7      | 2025-01-11 | 0       |
| 8      | 2025-01-12 | 0       |
| 9      | 2025-01-13 | 0       |
| 10     | 2025-01-14 | 0       |
| 11     | 2025-01-16 | 0       |
| 12     | 2025-01-17 | 1       |
| 13     | 2025-01-18 | 0       |
| 14     | 2025-01-19 | 0       |
| 15     | 2025-01-20 | 0       |
| 16     | 2025-01-21 | 0       |
| 17     | 2025-01-23 | 0       |
| 18     | 2025-01-24 | 0       |
| 19     | 2025-01-25 | 0       |
| 20     | 2025-01-26 | 0       |

==== CUSUM Formula ====

For binary outcomes, CUSUM was calculated using:

''Cₙ = max(0, Cₙ₋₁ + (Xₙ - k))''

Where:  
  * ''Cₙ'' = [[cumulative sum at case]] ''n''  
  * ''Xₙ'' = outcome (0 or 1)  
  * ''k'' = [[reference value]] (target complication rate = '''0.1''')

A [[threshold value]] of ''2.5'' was used as a [[visual alert level]] in the chart.

==== How It’s Determined (Why 2.5?) ====

The [[threshold]] [[value]] in [[CUSUM]] is essentially a decision limit: it defines the point at which accumulated deviations from [[expected outcome]]s are considered [[statistically significant]] or [[clinically significant]].

In our example, we used a **threshold of 2.5**, which is a commonly used **empirical cutoff** in a [[clinical quality control study]]. This choice is based on:

  * **[[Sensitivity to change]]**: A value like 2.5 balances the need to detect real performance issues without overreacting to random variation.
  * **Visual simplicity**: In graphical CUSUM charts, 2.5 provides a practical visual alert without requiring complex statistical modeling.
  * **Historical precedent**: Many published studies in surgery (e.g., on appendectomy, cholecystectomy, aneurysm clipping) use fixed decision limits like 2.5 or 3 as informal thresholds for further review.

Formally, threshold values can also be calculated using statistical parameters:
  * **h0**: acceptable failure rate (e.g. 10%)
  * **h1**: unacceptable failure rate (e.g. 25%)
  * **α (Type I error)**: false alarm rate, e.g. 0.05
  * **β (Type II error)**: missed detection rate, e.g. 0.10

Using these, you can compute optimal thresholds based on log-likelihood ratios or decision interval tables.

However, in practice — especially in early learning curve assessment or small data sets — **a fixed threshold of 2.5 is often sufficient** to guide reflection and discussion without overcomplicating the analysis.


===== Results =====

The CUSUM chart showed an initial upward trend with three complications within the first 12 cases. After case 12, the cumulative value gradually declined and stabilized, suggesting a reduction in complications and improvement in procedural performance.

{{::cusum_chart_for_icp_monitor_placement.png?600|}}



===== Interpretation =====

  * **Early phase (Cases 1–12):** Small cluster of complications, causing an upward slope.
  * **Late phase (Cases 13–20):** Stable or downward trend, indicating performance improvement and consistency.
  * **Threshold not breached:** No critical alarm was triggered during the observation period.

===== Conclusion =====

CUSUM analysis provides a simple yet powerful method to monitor surgical outcomes in ICP monitor placements. It enables early identification of performance trends and potential training needs. When used routinely, it may enhance surgical safety, quality assurance, and self-assessment in neurosurgical practice.

===== Recommendations =====

  * Apply risk-adjusted CUSUM models if patient variability is high.
  * Integrate CUSUM into quality dashboards for neurosurgical departments.
  * Use CUSUM charts during residency to track individual learning curves.