English - Home page

ISS
Istituto Superiore di Sanità
EpiCentro - L'epidemiologia per la sanità pubblica
Istituto Superiore di Sanità - EpiCentro


Sensitivity and representativeness of an HIV surveillance system in the Veneto Region

Federica Michieletto, Cinzia Piovesan, and Giovanni Gallo

Epidemiology and Public Health Service

Office of Prevention Veneto Region

 

Acquired immunodeficiency syndrome (AIDS) represents the full-blown form of the human immunodeficiency virus (HIV) infection.  Until recently, monitoring of the HIV epidemic has been based on AIDS surveillance.  Surveillance based on AIDS cases provides a delayed picture of the AIDS situation because of the long and variable period of incubation of the disease (median of approximately 10 years) and even further lengthening of the observed incubation period following the introduction of combination treatment with anti-retroviral agents.  These developments have encouraged the development of surveillance based on the diagnosis of HIV.

 

The Veneto Region, which has participated in the national AIDS registry since 1982, instituted its own system of HIV surveillance beginning in 1988. After a regional ruling on 2/2/1998, all public structures that conducted HIV tests on the general population (including centers for drug treatment, anonymous counselling and testing centers, infectious disease units, and public health services, for a total of 150 different sites) were requested to collect data according to a standardized procedure.  The information, gathered by physicians in each site during pre-test counselling using a specific form, contains an identifier code, risk behaviors for HIV, the date, location, and results of testing, and the Local Health Agency at which the individual is registered. The results of the ELISA serologic tests are considered positive only after confirmation with a Western-blot test. The paper data forms are send to the public health offices of the Local Health Agencies where they are entered into the computer and transferred via diskette to the regional registry, which is located in the Office of Prevention for the Veneto Region. Each year, a descriptive report is produced on the trends in new diagnoses of HIV infection, which is posted in the Internet and is also sent to all sites that provide data to the system.

 

To evaluate the extent to which the system is ale to identify all new infections (sensitivity) and if the probability of being included in the system differs by risk category in the population (representativeness), a study was conducted based on a comparison of the data obtained from the HIV surveillance system and that of the AIDS registry for the region.

 

Methods

As of December 31, 2000, 3116 cases of AIDS and 6516 diagnoses of HIV infection had been reported in Veneto. The analysis was limited to cases diagnosed since 1989 (when the HIV surveillance system was fully activated) who were residents of the region and were older than 14 years. A total of 1915 AIDS cases and 5416 cases from the HIV register were therefore included in the analysis.

 

A cross-matching of the two systems revealed that 1245 cases were notified in both systems. A total of 670 cases were found in the AIDS surveillance system but not the HIV system, but 94 of these were residents of Veneto who were receiving treatment outside the regions and likely had undergone HIV testing outside the region as well. Thus, 576 cases remained in the AIDS system who were not present in the HIV system.

 

Sensitivity

The sensitivity of the HIV system was estimated as the ratio between persons included in both systems and the total number of persons included in the AIDS registry. The sensitivity of the system was thus 68% (1245/1821), with an under-notification therefore estimated at about 32%.

 

Representativeness

To determine the representativeness of the persons included in the HIV surveillance system, logistic regression was performed to determine if certain groups were had a greater or lesser probability of being included in both the AIDS and HIV surveillance systems. HIV risk factors (homosexual, heterosexual, injecting drug use, and unknown), sex, age, and year of diagnosis of AIDS were independently evaluated using the model. The table shows the odds ratios (OR) and relative 95% confidence intervals (CI) for each of the variables studied.

 

From the results, it is apparent that older individuals are under-represented, as are those who became infected via sexual transmission and those whose AIDS diagnosis was made prior to 1993. In 1995, the AIDS surveillance form was modified to include the date on which the diagnosis of HIV infection was made, and it was therefore possible to examine the time between the two dates. An analysis of these data demonstrated that cases for whom the diagnosis of AIDS was made within 6 months of the diagnosis of HIV infection were also under-represented (OR = 0.2; 95% CI = 0.1 –0.3).

 

A second analysis was performed to evacuate agreement (using the Kappa statistic) regarding risk factors reported in the two systems, such that the accuracy of the HIV system could be evaluated regarding this important variable. The analysis was stratified by date of AIDS diagnosis (1992, 1993-1995, and <1996). The Kappa value remained essentially constant at approximately 70%; for 1992, the Kappa was 73%, for 1993-1995 it was 67%, and for the subsequent period it was 69%. It should be noted that the cases in which there was non-agreement had been classified as “non-determined risk factor” in the HIV surveillance system but as “sexual transmission” in the AIDS system.

 

Conclusions

This review of the HIV surveillance system in Veneto allowed us to conclude that the level of under-notification in the system was approximately 30% and that those who acquired their infection through sexual transmission or who were diagnosed with HIV infection shortly before being diagnosed with AIDS were less likely to be notified to the HIV system.  A more detailed analysis of completeness may be possible using other data sets, such as the regional hospital discharge registry or death registry.

  • Scheer S., McQuitty M., Denning P., Hormel L., et al. Undiagnosed and Unreported AIDS Deaths: Results from the San Francisco Medical Examiner. JAIDS. 2001

  • Klevens M.R., Fleming P.L., Li J. et al. The Completeness, Validity, and Timeliness of AIDS Surveillance Data. Elsevier Science Inc. 2001

  • Teutsch S.M., Churchill R.E. Principles and Practice of Public Health Surveillance. Oxford University press, 2000

  • Regione del Veneto - Giunta Regionale. Sistema di sorveglianza regionale AIDS. Veneto documenti, serie U.L.S.S.; novembre 1988

  • Pezzotti P., Piovesan C., Rezza G. et al., Combining HIV and AIDS surveillance: an experience from an Italian region, Int J Epidemiol. 1997 Dec;26(6):1352-8

  • Pezzotti P., Piovesan C., Rezza G., HIV/AIDS surveillance in Italy: the potential misinterpretation of time trends. Am J Public Health. 1997 Mar;87(3):469-70

  • Declich S., Carter A.O. Sorveglianza di sanità pubblica: origini storiche, metodi e valutazione. Ann. Ist. Super. Sanità, vol.32 n. 3 1996

Comment:

Patrizio Pezzotti

AIDS and Sexually Transmitted Diseases Unit, Laboratory of Epidemiology and Biostatistics

 

Following the introduction of effective treatments that reduce morbidity and has resulted in further increases the incubation period for AIDS, the surveillance of AIDS cases has become less relevant, while HIV surveillance has now become essential. In Italy, notwithstanding the recommendations of the World Health Organization (1), there is no national-level HIV surveillance and only a limited number of regions and provinces have initiated such surveillance.

The cross-matching of the AIDS and HIV surveillance systems in Veneto provides important data on the functioning and sensitivity of an HIV surveillance system. The analysis demonstrates that more than 30% of AIDS cases were not reported to the HIV system. In particular, it showed that those whose serostatus was diagnosed at or near the time of their AIDS diagnosis and those who acquired the infection through sexual transmission were less likely to be notified to the HIV system. The under-notification of the sexually transmitted cases was also underlined by the fact that many of those who were reported as having an unknown risk factor for HIV infection indeed were reported in the AIDS surveillance system to have acquired the infection through sexual transmission.

 

These results have two implications: 1) efforts are needed to understand the reasons behind lack of HIV notification by reporting centers and 2) that there is clearly an under-estimation of the frequency of HIV infection, especially that due to sexual transmission.

 

With these data, one can estimate the extent of HIV infection in Italy. First, if the sensitivity of the system is 68%, this means that an additional 3000 cases should be added to the more than 6500 cases reported in Veneto as of December 31, 2000, which thereby would bring the total to 9500 cases of HIV infection since the beginning of the epidemic. Given that Veneto has reported approximately 6.5% of the AIDS cases in Italy (2), and assuming that the relationship between HIV and AIDS would be constant across regions, the estimate of HIV infections for Italy would be on the order of 146,154 (9500/0.065). This estimate of cumulative incidence is nonetheless a conservative since it does not take into account infections that have not yet been diagnosed.

 

The validity of the above-reported estimate is conditioned by two basic assumptions (correct sensitivity and homogeneity between Veneto and the rest of Italy). The estimated sensitivity can be biased if the notification of cases in the two systems is not independent (3). It would thus be highly desirable to cross-match other existing data such as the nominal death registries and hospital discharge records maintained at regional level, thereby obtaining estimates that take into account the dependence between the various data sources (3). With respect to uniformity, these data are consistent with those obtained with the surveillance systems in the province of Modena (Emilia-Romagna) and the Lazio Region (unpublished data).

 

References

1. European regional consultation on the introduction of second-generation HIV surveillance guidelines. World Health Organization 2001.

2. Aggiornamento dei casi di AIDS notificati in Italia al 30 Giugno 2001. Notiziario ISS, volume 14. Ottobre 2001.

3. Chao A et al. The applications of capture-recapture models to epidemiological data. Statist Med 2001; 20:3123-3157.

TOP