Table 2

Impact of chart design parameters on chart performance

ParameterChange or unit amountSensitivitySpecificityYouden’s index
ShewEWMAMAShewEWMAMAShewEWMAMA
Baseline data (DICON vs local hospital)Switch to DICON baseline+0.111+0.121+0.114NSNSNS+0.109+0.123+0.121
Control limit widths
(0.5–4 σ)
Per σ−0.217−0.213−0.193+0.239+0.207+0.122+0.094−0.006−0.071
Baseline size
(3–24 months)
Per month+0.002+0.002+0.002−0.002−0.003−0.003NSNSNS
Baseline lag
(0–24 months)
Per month+0.002+0.002+0.003+0.002+0.002+0.001+0.004+0.002+0.001
Baseline type
(fixed vs rolling)
Switch to fixed baselineNS−0.017−0.020+0.048+0.046+0.022+0.038+0.029NS
Supplementary rules (yes vs no)Switch to also use rules+0.084N/AN/A−0.028N/AN/A+0.057N/AN/A
EWMA weight
(0.2–0.8)
N/A+0.089N/AN/A−0.381N/AN/A−0.292N/A
MA span
(3–12 months)
Per monthN/AN/A−0.008N/AN/A+0.016N/AN/A+0.008
  • Regression analysis results, conducted separately for each chart type. Values given are unstandardised regression coefficients. Change or unit amount denotes the change or increment size associated with each coefficient.

  • σ, estimated SD; DICON, Duke Infection Control and Outreach Network;EWMA, exponentially weighted moving average chart; MA, moving average chart; N/A, not applicable; NS, coefficient not statistically significant at the 0; Shew, Shewhart chart.