Our study showed that electronic databases from primary care networks can be used to estimate the incidence of chlamydia diagnoses. However, we observed variability in the incidence rate reported by the four networks; different registration rules/habits and socio-demographic characteristics of the patient populations in the networks contributed to this variability. Diagnosed cases of chlamydia infection were recorded in several ways in GP electronic records. While most GPs used ICPC codes for recording details of their diagnosis, others resorted to free text fields from which extracting standardized diagnoses is difficult and time-consuming. The application of ICPC codes in general as well as the use of specific (sub)codes differed from network to network hampering comparability and the analysis of overall trends.
The estimated incidence of chlamydia diagnoses varied from 103 to 590 per 100,000 in the four networks. Despite the large difference in incidence rates, similar trends were observed. Chlamydia incidence was highest in the youngest age groups (15 to 30 years), where incidence rates in women were approximately twice as high as in men. This is in agreement with what is found in Dutch STI clinics, where the numbers of young women diagnosed with chlamydia are higher than in young men . The highest incidence rates of diagnosed cases were found in the AMC network. Patients included in the AMC network, in the south-eastern part of Amsterdam have on average a lower socioeconomic status and are more often of non-western origin. Based on a classification of Statistics Netherlands and The Netherlands Institute for Social Research 92% of all patients in these practices belonged to the lowest of five socioeconomic status levels while 53% were of non-western descent, in 2007. Other studies revealed that the risk of acquiring chlamydia is associated with ethnicity and social economic status [6, 15]. In addition, the southeast of Amsterdam is a highly urbanized area and as in the IPCI and LINH network we observed a general increase in chlamydia incidences with higher degree of urbanization. In contrast, the relatively high incidence in rural areas in the two national networks is remarkable. Although it cannot be excluded that this reflects a chance effect due to low numbers, further investigations into testing practices are needed to explore this finding.
Our results regarding the trends in incidence of chlamydia diagnoses by age, sex and degree of urbanization compare well with a large population-based chlamydia screening pilot study conducted in 2002/2003. In that study, the prevalence of chlamydia was found to be 2% among 15-29 years old participants (1.5% in men and 2.5% in women). Prevalence was lower in rural areas (0.6%) compared to very highly urbanized areas (3.2%) . Assuming an (untreated) chlamydia infection lasts on average one year, these prevalence rates can be roughly interpreted as annual incidence rates [16–18], which would mean an incidence rate of 1500 and 2500 per 100,000 in young men and women. Our findings in the nationwide networks (an incidence of 300 and 500 per 100,000 in young men and women) indicate that about 20% of the total incident chlamydia cases in that age group is diagnosed by GPs.
Because chlamydia was not recorded under one code and reporting laboratory results was not standard, uncertainty remains about whether all of the extracted cases were actually chlamydia cases and about how many cases of chlamydia were recorded in a way that they were not identified by our extraction rules. The definition of a chlamydia infection that uses fixed main-codes in combination with the code for a specific antibiotic treatment in the LINH network could potentially lead to an overestimation of chlamydia incidence especially in higher age groups. For example, the antibiotic treatment for epididymitis in older men is also prescribed for other pathogens than Chlamydia trachomatis. Information about chlamydia testing results were lacking for the majority of patients in these networks. Finally, the years of study were not the same for every network, which could have influenced the observed incidence rates.
A study of morbidity rates in general practice registration networks in the Netherlands identified potential reasons for the observed variability that apply to our findings as well . We also perceived differences in the methodological characteristics of the networks, such as definitions and registration rules used. Another reason for variability pertains to the variety of testing practices used among GPs. A qualitative study exploring the reasons for variation in diagnostic testing found major differences between high and low chlamydia testers . Possibly the rate of testing is higher in an urban setting compared to a rural area, explaining at least partly the difference in incidence rates in both areas. The last reason these authors mentioned was related to the socio-demographic characteristics of the catchment population of the networks, which is certainly true for the differences between the national and regional networks in our study, as described above.