Temporal analysis of routine surveillance data uses graphical representation, usually plotted by time of onset. However, the time of onset may give a false impression of a decreasing trend in the most recent time units because of reporting delays. If delays are known to occur, time of notification should be preferred.

Appropriate graphs for time data includes histograms (or joint bars) and line charts. The use of histograms must be limited to enumerated data (count of cases) and not used for rates for which line charts are preferred.

Figure 1: Distribution of viral hepatitis in Lebanon, by weeks, as of week 2003-15

Counts of notifications are generally used. Alternatively rates (cases * 100,000 / population) are preferred when the population is not stable over time. Graphs are produced for the national level as well as each intermediate or peripheral level, usually on a weekly basis.

The visual review of the graph aims at detecting aberrations in the data, such as changes in trends, depicting increase in person-to-person transmission, or peaks related to potential point source outbreaks. For certain health events, thresholds can be set to detect areas in which notified cases or rates exceed some pre-defined values. However, for other diseases, testing for statistical significance of these changes in notification patterns requires statistical calculation to quantify the departure from historical values. These methods are detailed in chapter Methods for setting thresholds in time series analysis.

Smoothing techniques, such as averaging values over a rolling period (moving averages), are useful to characterize trends and seasonality. Using an averaging period of:

  • 5 to 15 weeks highlights the seasonal pattern of the event, by smoothing the variations occurring from one week to the next.
  • 52 weeks highlights the secular trend in the data by averaging the seasonal effect across averaging time period
  • See annex 1, page 7 for further information on the use of simple smoothing techniques for describing the components of time series.
  • Analysis of case-based surveillance time characteristics is subject to limitations due to its nature. Surveillance is an evolving process that can be affected by:
  • Change in case definitions as new tests become available
  • Enrolment of new sources of data
  • Increase in the completeness of reporting following a sensitization campaign
  • Enhancement of surveillance in the event of an outbreak,

Each of these approaches induces limitations in the use of historical data as a baseline against which alerts can be detected.