Example of analysis

This page presents an example intended to show a interpretation of the statistics in this service.

• Introduction
• General
• Table: Year and month statistics per county
• Map: Cases per 100.000 population per county and year
• Map: Cases per county
• Graph: Number of cases per week
• Graph: Trend (moving average)
• Country of infection facts

Introduction

All statistical visualisation methods, of the web service, are used in this example, which gives generic as well as disease specific tips and advice for the interpretation of data. Attention has also been drawn to some common statistical "pitfalls" to give the user a greater understanding of, and ability to use, the system.
The example given here is based on Campylobacter data, taken from the service in August 2002. Campylobacter is chosen because:

• There is a huge amount of data available for analysis.
• The data varies in a way that lends itself very well to illustration.

General

The statistics show the county from which the notification originated, ie: the county where the doctor saw and diagnosed the patient or the laboratory diagnosed the patogen. The source of the notification may be the same county where the disease was contracted, but this is not always so. For example a student may contract a disease in their university town, but not seek medical attention until they go home during a vacation.

In some cases the disease may have been contracted in a different country (for example when on holliday abroad). These circumstances differ from the case with the student; the doctor must ask which country the patient feels the most probable source of infection.

Table: Year and month statistics per county

The table shows the total number of cases in red and the incidence (cases per 100.000 pop. and year) in blue during complete calender years for 1999, 2000 and 2001. An incidence table is a good indicator of disease trends and tendencies, which may be quite hard to spot from a table. It is primarily this problem the maps are intended to solve, to visualise the tabled information.

By considering the table we can observe the following about Campylobacter:

• Halland has the same incidence in 1999 and 2000, but a different numbers of cases. This is probably due to an increase in the population in proportion to the number of cases. This could also be caused by the rounding of errors in the calculation, because the incidence is presented as an integer and not as a decimal number.
• Örebro illustrates the same phenomenum, but inverted, in 1999 and 2000, which is probably caused by a decrease in the population during these years.
• Jämtland shows a significant increase in both the incidence and the number of cases between 2001 and 1999-2000.
• Västra Götaland also shows a significant increase 2000-2001 in comparison with 1999.

Map: Cases per 100.000 population per county and year

The maps illustrate the comparison between the incidence (cases per 100.000 pop. and year) for complete calender years 1999, 2000 and 2001. By combining the maps with data from the table above, a more detailed analysis can be preformed (this is done by clicking on a county in the normal service – but does not work on this help page). The maps in general do not enable any further conclusions apart from the analysis of the table, but it makes it easier to see trends and to get a general impression.

From the figure above, we can observe the following:

• In all years Gotland shows a high incidence.
• In general, the northern part of Sweden, especially Norrbotten and Västernorrland, tend to show low values of incidence. The increase in Jämtland in 2001 breaks this trend and is therefore conspicuous.
• None of the major cities in Sweden (Stockholm, Göteborg and Malmö) differs from other Swedish cities.
• In a larger perspective, the incidence tends to show an increasing trend in the south and middle parts of Sweden, while a decreasing trend is shown in the north.

The variation in incidence between counties cannot be totally explained by the fact that one population is infected more, or less, than another population. For example some doctors have a higher index of suspicion and request tests more frequently, different populations may have a higher or lower threshold for presenting to the medical services because of geographical distance to the health centre.

Map: Cases per county

The maps show the total number of cases in each county for the years 1999, 2000 and 2001. As before we can use the maps to find trends and tendencies that later we can try to verify with the data of the table.

Information extrapolated from maps which show either the incidence of a disease or the total number of cases is not comparable. For example the map showing the total number of cases, above, only reflects the distribution of densely populated areas, because the incidence of Campylobacter in Sweden is quite evenly distributed. Stockholm, Malmö and Göteborg are the three most populated cities in Sweden and the maps show this very clearly.

Maps showing total numbers of cases are valuable in budget planning or for the allocation of preventive resources.

Graph: Number of cases per week

The graph shows an untreated plot of the number of notified cases of Campylobacter infection, per week, in Sweden, for the period of August 1998 to July 2002. The raw data graph does not always look as clear as this one and so it is not always possible to draw conclusions from such a graph. It is often better to begin by considering the trend graphs (see below) and then use the raw datagraph to try and verify possible trends and tendencies.

In the graph above Campylobacter appears to have a constant rate. Yet, during the summer barbecue season (June/July/August/September), the number of cases, per week, doubles compared with the rest of the year.

At Christmas and the New Year the number of cases seems to decrease. This is probably due to the fact that the notifications are not reported until after the Christmas New Year vacation.

Graph: Trend (moving average)

The graph shows two moving averages for the period August 1998 to February/June 2002. A long trend 52 week plot and a short trend 6 week plot. The 52-week plot describes a rough trend of the development over several years. Possible seasonal trends disappear and a general trend can be observed. The six-week plot, on the other hand, simplifies the observation of possible seasonal trends and can be used to predict a future trend.

As previously shown in the raw data graph, a seasonal trend between June/July and August/September can clearly be seen in the six-week plot.

The 52-week plot shows that the number of cases per week for Campylobacter seems to continue steadily with only minor fluctuations.

Country of infection facts

Summery:

For the years 1999 to 2002 the summary table shows the proportion of Swedish cases of Campylobacter which are contracted domestically and internationally. Occasionally the information regarding the country of origin is missing. Often this is because there is only a laboratory notification and the clinical details are missing.

Because a large percentage of the cases are contracted abroad, other information may be extrapolated from the tables and maps about the Swedish population. For example the countries most frequently visited. Also because Stockholm has a higher level of incidence than the average, it may suggest that the population in the county of Stockholm travel more commonly than the average population.

From the table above 57% of Swedish Campylobacter infections in 2001, are contracted abroad. The proportion of cases with missing information seems to be increasing (4%-17%). This may be due to the presenting clinical doctor omitting the ‘country of origin’ information on the clinical notification, or, more commonly, failing to complete a clinical notification form at all.

In 1999 the information concerning the country of origin of infection, is missing in only 4% of cases. In 2002 this has risen to 17%, indicating that the figures for 1999 are more reliable than the figures for 2002. Generally speaking the number of cases missing information, regarding the country of infection, is greater in the current year than previous years. This is because when the information is missing, it is requested; a reply necessarily takes some time.

1: Thailand - 743
2: Spanien - 364
3: Turkiet - 134
4: Indonesien - 83
5: Indien - 76
6: Frankrike - 61
7: Tunisien - 51
8: Portugal - 48
9: Tjeckien - 45
10: Marocko - 37
11: Bulgarien - 34
12: Storbritannien - 31
13: Grekland - 30
14: Kina - 30
15: Egypten - 24

The list above shows the countries most frequently implicated as being the source of infection in 2001. It is hard to draw concrete conclusions from this list, about the relative risk of the individual countries, because the total number of Swedish tourists to each country is not known.

Thailand and Spain (Spanien) are top of the list for all years 1999-2002. Both countries are well visited, but compared with Greece (Grekland: about 7th place 99-02, also well visited), it is easy, but not necessarily true, to think that Campylobacter is more easily contracted in Thailand and Spain than Greece. A similar comparison can be done with Indonesia (Indonesien: about 4th place 99-02, not well visited)

This type of information can be significant. When Sweden's Salmonella Control Program, which stated that domestically produced and imported food must be Salmonella free, was evaluated, the evaluation showed that 85% of salmonella cases in Sweden were contracted abroad indicating that the program had been a success.