Type of chart | Type of outcome data with example |
X-bar S-charts | For continuous data. Under only common cause variation, the means will follow the normal distribution, such as length of stay32 or time to antibiotics. These are used in combination where the X-bar chart will investigate whether the intervention has improved the mean, and the s-chart whether it has improved the SD (ie, less variability) of the outcome. Both are needed as the X-bar control limits are calculated under the assumption that the SD is in control, and one may go out of control independently from the other. |
P-Charts | For discrete event data. Under only common cause variation, the proportions follow the binomial distribution, such as percentage of patients prescribed a new sedative2 or percentage of patients not seen in the last year.15 |
C-Chart or U-chart | For count data. Under only common cause variation, the counts follow the Poisson distribution. c-Charts are used when the variable counts the total number of an event such as the total number of patient falls in a week (provided a stable ‘area of opportunity’, eg, consistent number of patients on the ward across weeks). u-Charts are used when the variable is an average rate of an event, adjusted for a denominator size (or ‘area of opportunity’), such as the number of opioid-related oversedation events per 1000 patient-days.14 |
G-Chart | For variables that follow the skewed geometric distribution under only common cause variation. Can be used to investigate time (or number of patients) between outcomes, particularly useful if the outcome occurs infrequently, such as the number of deliveries between newborns with an Apgar score <7 after 5 min.17 |
(EW)MA chart | For different types of data, plots moving averages (MA) over time and thereby includes more measurements in every plotted data point. MA charts give equal weight to all previous measurements within a rolling window of a specified size, whereas exponentially weighted (EW)MA charts place the greatest weight on the most recent observation. Particularly useful to detect small or gradual shifts in the outcome, such as small increases in surgical site infections and thereby earlier detection of an outbreak.33 34 |
CUSUM chart | For different types of data, the cumulative sum chart statistic summarises the extent to which an outcome shifts away from a baseline rate with higher values corresponding to stronger evidence of a change. Particularly useful for earlier detection of small increases in event rates, such as a worsening in 1-year revision rates after total hip and knee replacements.35 |
CUSUM, cumulative sum.