Viral hepatitis in Israel: the effect of canvassing physicians on notifications and the apparent epidemiological pattern*

Bull. Org. mond. Sante 1972, 46, 457-464 Bull. Wid Hith Org. Viral hepatitis in Israel: the effect of canvassing physicians on notifications and the apparent epidemiological pattern * DANIEL BRACHOTT I & JAMES W. MOSLEY 2 Notifications of morbidity are frequently incomplete but they represent the only practicable way of carrying out large-scale surveillance. Studies of hepatitis in the Central District of Israel offered an opportunity of assessing the reliability of routine data. The regular canvassing ofphysicians resulted in a 2-3-fold increase in reports of viral hepatitis. Comparisons of rates before and after canvassing with those for an insured population within the Central District indicated that there was an improvement from 37.4 % in 1960/61 to 96.2 % in 1968/69. None the less, the lower (precanvassing) level of reporting adequately delineated year-to-year fluctuations in morbidity, the seasonal cycle, and age distribution. One consequence of canvassing was a distortion of the secular trend not only for hepatitis but also for some other communicable diseases. Communicable diseases, even those of consider- able public health importance, are consistently underreported to varying degrees in most countries. Because underreporting is so widespread, morbidity data are often assumed, especially by physicians aware of their own poor compliance, to have little value. Yet, morbidity reporting, even if it is incomplete, offers the only practicable system for the large-scale surveillance of most infectious disease. Because underreporting varies markedly from country to country, it is evident that attack rates based on routine notifications cannot validly be compared (Mosley & Kendrick, 1969). Relative incidence, however, is only one of several epidemio- logical comparisons that may be of interest. A study of secular trends, seasonal fluctuations, and patterns of age- and sex-specific attack rates may yield valuable clues concerning the behaviour of infectious agents under differing conditions. Unfortunately, there is little information about the distortions that * From the Hepatitis Project, Ministry of Health, Israel, and the Center for Disease Control, US Public Health Service, Atlanta, Ga., USA. 1 Assistant Director General, Ministry of Health, Jerusa- lem, and Chairman, Department of Preventive and Social Medicine, Tel-Aviv University, Israel. 2 Associate Professor of Medicine, University of Southern California, Los Angeles, Calif., USA. Formerly Chief, Hepa- titis Unit, Center for Disease Control, Atlanta, Ga., USA. irregularities in notification may impose upon epi- demiological parameters other than overall incidence. In a previous investigation, Reisler et al. (1970) estimated that the completeness of notifications in Israel averaged 36.3% from 1954 to the end of 1966. Therefore, one of the main objects of the Hepatitis Project was to obtain complete reporting of the disease in the Central District of Israel. Regular canvassing of physicians for notifications was fol- lowed by the submission of reports consistently more numerous than those previously received. It was necessary to evaluate the presumed improvement in reporting, and this evaluation also afforded an opportunity to determine whether the apparent epidemiological pattern would be changed if notifi- cations were more complete. METHOD OF STUDY Before canvassing began in July 1965, each physi- cian in the Central District received a letter re- questing the prompt and complete notification of cases of viral hepatitis in his practice. Physicians were also informed that they would be questioned periodically about the disease among their patients. The canvassing of physicians was usually made an additional responsibility of the public health nurse regularly assigned to a given area. The nurse was 2819 -457 3 D. BRACHOTT & J. W. MOSLEY instructed to inquire weekly about recent cases of viral hepatitis during the course of other contacts with practising physicians. If her usual duties did not cause her to visit all the physicians in her area within any 2-week period, she was required to make specific inquiries about viral hepatitis from those she had not seen. In localities where preventive services were not provided by the Ministry of Health, a special nurse was given the responsibility of making regular contact with physicians. The system of notification was, however, unchanged; i.e., the phy- sician still submitted a formal report to the sub- district office. The effectiveness of canvassing was probably improved by the fact that households known to have cases of viral hepatitis were visited as promptly as possible. Children in the family were then enrolled, if eligible, in a study of immune serum globulin prophylaxis (Mosley & Brachott, 1970). The activ- ities involved in this phase of the project probably continuously reminded all physicians and nurses about case finding. In assessing the completeness of the increased number of notifications after canvassing began, data from the statistical office of the Workers' Sick Fund (WSF) were again used (Reisler et al., 1970). This prepaid health insurance scheme covered approxi- mately 70% of the population concerned, and each visit to a WSF physician can be accounted for in terms of a diagnosis. Morbidity data concerning viral hepatitis and other diseases were made available by the Division of Epidemiology of the Ministry of Health. The Table 1. Cases and epidemiological year (EY), rather than the calendar year, has been used whenever data were available in a suitable form. Use of the epidemiological year (i.e., the interval from July of one calendar year to the end of June in the next calendar year) enables the seasonal cycle to be taken into account in analyses. RESULTS Changes in the reported incidence of viral hepatitis after canvassing was initiated Table 1 compares cases and rates of reported viral hepatitis in the Central District with those for the rest of Israel during a 14-year period. For conve- nience, data have been summarized for two 5-year periods (EY 1955/66-1959/60, and 1960/61-1964/65) prior to the inception of canvassing, and for the first 4-year period (1965/66-1968/69) that the procedure was operated. Even before improved reporting began, reporting rates were somewhat higher in the Central District (83 and 116 per 100 000 persons in the two 5-year periods, respectively) than in the rest of Israel (68 and 82 per 100 000 persons in the two 5-year periods, respectively). As a result, the Central District, with 19.7% and 18.4% of the total popula- tion of Israel during the two periods, contributed 23.1 % and 24.1 % of all reported cases. In the 4-year period after canvassing began in the Central District, the average rate of notified hepatitis increased to 204 per 100 000 persons, while that in the rest of Israel (61 per 100 000 persons) was somewhat below that for the two previous periods. rates of reported viral hepatitis in the Central District and the rest of Israel Period No. of cases a ~Percentage RatePeriod of cases a of total per 100 000 persons a cases reported Type of Epidemio- Central Rest of from the Central Rest of reporting logical year Total District Israel Central District IsraelDistrict Ditit Ire 1955/56- 1 389 321 1 068 23.1 83 68 1959/60 routine 1960/61- 2068 498 1 570 24.1 116 82 1964/65 canvassing 1965/66- 2 357 998 1 359 42.3 204 61 in the 1968/69 Central District a Annual average. 458 NOTIFICATION OF VIRAL HEPATITIS IN ISRAEL Thus, canvassing in the Central District was asso- ciated with a notification rate more than three times that for the remainder of the country during the same period, and twice that in the Central District itself during the 10 preceding years. The Central District, with 18.0% of the population from 1965/66 to 1968/69, accounted for 43.2% of all notifications of viral hepatitis in Israel. These findings suggest that canvassing resulted in improved reporting in the Central District. To determine if there were variations in the level of notification within the Central District, a similar analysis was made with respect to the four subdis- tricts. The differences were minor and compatible with local variations in the intensity and the phase of the epidemic cycle of viral hepatitis. Completeness of improved notifications To evaluate the higher level of reporting in the Central District after canvassing began, it appeared desirable to make a comparison with the incidence of viral hepatitis registered by WSF physicians in the same area. As Reisler et al. (1970) have pointed out, neither the boundaries of WSF districts nor the basis of reporting (the location of the physician or clinic rather than the residence of the patient) are the same as for Ministry of Health data; however, the information should provide a rough index of the completeness of notifications to the Ministry of Health. Three WSF districts are largely or entirely con- tained within three subdistricts of the Central District. The WSF statistical office made special compilations of data from these three districts for three epidemiological years selected by us. The periods were 1960/61, an EY of high incidence before canvassing began; 1965/66, the first EY of canvassing; and 1968/69, the most recent EY for which complete information is available. Table 2 shows that for 1960/61 the rate of hepatitis reported to the Ministry of Health from the three subdistricts was 37.4% of the rate in the three WSF districts. In 1965/66 and 1968/69, however, the rates reported to the Ministry were 85.9% and 96.2%, respectively, of the WSF rates. Although data for the individual pairs are not given, this comparison was made and showed essentially similar increases in each. These findings support the assumption that the higher rate of hepatitis in the Central District after canvassing was attributable to greatly improved reporting. They also suggest that most cases re- ceiving medical attention in the Central District were Table 2. Completeness of Ministry of Health notifica- tions in 3 subdistricts of the Central District as indicated by WSF registrations for 3 geographically similar WSF districts Complete- Ministry ness of Year nof Health Registrations MinistryYear notin icthree in 3 WSF of Healthsubdistrects a districts b notifica-subdistricts a ~tions (% ofWSF rate) epidemiological cases 409 cases 770 year prior to 37.4 canvassing rate c 133 rate d 356 (EY 1 960/61 ) first year of cases 700 cases 604 canvassing 85.9 (EY 1965/66) rate c 197 rate d 229 fourth year of cases 990 cases 752 canvassing 96.2 (EY 1968/69) rate c 260 rate d 270 a Sharon, Petah Tiqwa, and&Ramlah subdistricts. b Hefer, Petah Tiqwa, and Rishon Le Ziyyon districts. c Per 100 000 persons in the total population. d Per 100 000 persons insured. reported to the Ministry of Health as a result of canvassing. The effect of canvassing on seasonal distribution of notified cases Improved reporting in the Central District pro- vided a means for assessing the adequacy with which the " true" seasonal distribution had previously been delineated. Fig. 1 is a comparison of the J A S 0 N D J F M A M J I ~~~~~~~~~~~~~~~~~~~4-W"O 20252 Fig. 1. Seasonal distribution of reported viral hepatitis before and after an improvement in reporting in the Central District (semilogarithmic scale). Continuous line, after canvassing 1965/66-1968/69; broken line, before canvassing 1955/56-1964/65. 459 D. BRACHOTT & J. W. MOSLEY average monthly case total (adjusted to 30 days) during 4 years of canvassing with those during the 10 preceding years. With improved notifications, the rise in the number of cases in autumn was more gradual, and the decline in late winter much less abrupt, than previous reports had shown. With better reporting, the total number of cases in the 3 months of highest incidence was 1.7 times that for the 3 months of lowest incidence; before the improvement, the ratio was 2.7. Similar analyses for the rest of Israel show seasonal increase ratios of 2.4 and 2.3 for the later and earlier periods, respectively. Thus, the change in the Central District presumably does not result from a change in the epidemiological behaviour of the disease itself. Assuming that the level of notifications achieved with canvassing represents complete reporting, Fig. 2 Gn Fig. 2. Seasonal fluctuations in completeness of reporting for viral hepatitis. shows the percentage completeness of reports by month of the seasonal cycle. Under usual circum- stances, reporting seems to have been most complete during the autumn and early winter rise in number of cases, and poorest during the subsequent decline. The effect of improved reporting on apparent age distribution Improved reporting also provides a mechanism for determining whether some age groups were dispro- portionately represented in routine notifications. Unfortunately, data concerning age distribution by district are available for the period since 1963 only; this restriction limits the extent of the comparison. Table 3 shows age-specific attack rates in the Central District and the rest of Israel for 2 periods: the 2 calendar years (1963 and 1964) before the institution of canvassing; and the 3 calendar years following (1966-68). As noted previously, attack rates in the Central District during the early 1960s were already somewhat higher than in the rest of Israel, presumably on account of better reporting. With canvassing in the Central District, rates in all age groups increased dramatically; in the rest of Israel there was no consistent or notable change. In the Central District, the increase was greatest in the rate for children under 5 years of age, and least for younger adults (persons aged 20-44 years). Rela- tively more underreporting in the former group and relatively less underreporting in the latter, who are likely to be hospitalized, would not be surprising. Table 3. Age-specific attack rate for viral hepatitis before and after canvassing began in the Central District, and comparable data for the rest of Israel Central District Rest of Israel Age group Age-specific Age-specific pe t Age-specific Age-specific p rate a for rate a for ercentage rate a for rate a for ercentage 1963-64 1966-68 change 1963-64 1966-68 change 0-4 307 862 +181 242 270 + 12 5-9 244 444 + 82 174 159 - 9 10-14 114 196 + 72 76 61 - 20 15-19 51 115 +125 39 54 + 38 20-44 30 48 + 60 19 16 - 16 45-64 15 30 +100 8 7 - 12 >65+ 12 25 +108 7 6 - 14 all ages b 100 216 +116 67 67 no change a Per 100 000 persons based on 1965 estimates. b Age-specified cases only. 460 r M A 11aI r NOTIFICATION OF VIRAL HEPATITIS IN ISRAEL Epidemiological year 68/69 1- WHO 20253 Fig. 3. Incidence of repcrted v ral hepatitis in Israel as a whole, the Central District, and Israel excluding the Central District; EY 1955/56-1968/69. The effect of improved reporting in the Central District on the apparent pattern of viral hepatitis Fig. 3 shows the secular trend of reported hepatitis for all of Israel, for the Central District, and for Israel excluding the Central District. Interpretation of the trend in the Central District would be quite erroneous if it were not known that an augmentation of reporting began in EY 1965/66. Furthermore, the procedural change in this area with slightly less than 20% of the country's population had a considerable effect on the trend for the whole country. In 1965/66, a moderate decrease in incidence in the rest of the country, which may have occurred also in the Central District, was largely obscured by improved reporting in the latter. A moderate increase in the incidence of viral hepatitis in the rest of the country in 1966/67 was almost cancelled by a modest decrease in the Central District; the converse was true in 1967/68. The effect of improved reporting of viral hepatitis on the reporting of other diseases It seemed possible that the canvassing procedure for viral hepatitis could indirectly remind physicians of their responsibility to report other infectious diseases. This possibility was examined for three reportable diseases of childhood. Table 4 indicates an increase in the proportion of measles, whooping cough, and scarlet fever notifications contributed by the Central District after canvassing began. DISCUSSION Evaluation of the completeness of morbidity reporting is difficult. Unless some measure of morbidity independent of routine notifications is available, reliance must be placed on improving the notification performance of practising physicians, and then judging the earlier level in terms of the later one. Controlled observations are, therefore, impos- sible. The situation in Israel is sufficiently distinctive to have permitted these rather detailed observations. WSF procedures provide an independent index of the completeness of Ministry of Health notifications. In the present study, as in a previous one (Reisler et al., 1970), it was assumed that the WSF rates for the insured population are true rates for the entire population. The completeness of the Ministry of Health reports was assessed by that means, although 461 D. BRACHOTr & J. W. MOSLEY Table 4. Possible effect of canvassing for hepatitis notifications upon reporting of other infectious diseases in the Central District Measles Whooping cough Scarlet fever 1960-64 1966-68 1960-64 1966-68 1960-64 1966-68 total no. of cases in Israel a 7 567 3 941 1 428 420 1 055 687 no. of cases in the Central District 2 134 1 473 411 158 224 176 percentage of total contributed by the Central District 28.2 37.4 28.8 37.6 21.2 25.6 a Annual average. it was recognized that the WSF data were simply an approximation to the true rate. The finding that canvassing brought the incidence rate according to Ministry of Health reports up to approximately the same level as WSF registrations serves to check the latter as well as the former. This is in agreement with previous assessments by the same method of completeness and reliability of country-wide com- municable disease reports to the Ministry of Health and the WSF (Kallner, 1955). Data for seasonal and age distribution are not available for WSF registrations; thus changes appar- ently attributable to canvassing could be validated only by comparing the pattern in the Central District with that in the rest of Israel during the same years, and that in the district during the preceding years. The general uniformity of hepatitis in Israel and the shortness of epidemic cycles support the applicability of this procedure, at least with respect to gross differences. The " true " and " apparent " seasonal distribu- tions shown in Fig. 1 indicate that the true distribu- tion revealed by relatively complete reporting differs moderately from the apparent pattern at a lower level. As the number of cases seen by a physician increases during the seasonal rise in incidence, the physician evidently becomes more aware of the disease and notifies the Ministry of Health about a larger proportion of cases (" recruitment " of report- ing). " Reporting fatigue ", however, sets in imme- diately after the seasonal peak, and the completeness of reporting during the fall in incidence is less than at the same absolute level of incidence during the rise. Neither " recruitment " nor " fatigue " could be demonstrated in the previous study (Reisler et al., 1970) with respect to a secular trend. Since that study was based on calendar years rather than on epidemiological years, the failure to observe these phenomena is not necessarily conclusive. The 2-year cycle in Israel could conceal distinctions between high and low years by combining relatively more underreporting during the decline in one seasonal cycle with relative less underreporting on the rise in the next cycle. Nevertheless, when both the seasonal distribution and the age distribution are examined differences between the " apparent " and " true" rates are seen not to be very great. The occurrence of a seasonal cycle and the time of its peak were indicated by the incomplete notifications. Similarly, even if the rela- tive incidence in young children had been understated previously, Table 4 indicates that there was no change in the order ofrank for age-specific rates, and suggests relative under- or overreporting in only 2 of 7 groups. This evidence, in conjunction with that presented by Reisler et al. (1970) showing that secular trends are also adequately depicted despite incomplete notification, suggests that several important epi- demiological features can be assessed from routine reports. Although canvassing appears from our study to have been quite effective over a period of 4 years, it cannot necessarily be concluded that it would be equally effective under all conditions. In the Central District, the report of a case leads to definite action with respect to children who are household contacts. Continuing emphasis was placed on case finding in order that the maximum number of household contacts would be available for our study of globulin prophylaxis. 462 NOTIFICATION OF VIRAL HEPATITIS IN ISRAEL If it is desired to check levels of reporting or to estimate true incidence in relation to other popula- tions, canvassing of a sample of physicians may be preferable to soliciting complete notifications from all physicians. It is evident that our emphasis on complete notifications for an entire district had some consequences that were not entirely desirable. The two-fold increase of notified hepatitis in the Central District resulted in a sudden apparent increase of some 35% for the entire country (Fig. 3). Further, although the epidemiological behaviour of the disease in the Central District did not vary markedly from that in the rest of the country during the 4 years reviewed, it is obvious that an aberration in the Central District would have had a disproportionate effect on the national data. The sudden improve- ment in reporting in 1965/66, as well as the slight difference in phase in 1966/67 and 1967/68, resulted in a flattening of the trend in Israel as a whole, even though the usual alternation of high and low incidence occurred in the rest of the country. Even a moderately large common-vehicle epidemic in the Central District would have an unrepresentatively great effect on the short-term trend of viral hepatitis in the whole country, and on the apparent age and sex distribution of notified cases. Furthermore, improved reporting of hepatitis appears to have influenced the reporting of other diseases in the Central District. Unless this effect is recognized, serious misinterpretations of the distri- bution and epidemiological behaviour of the other diseases could result. In fact, even if it is known that reporting did improve, it may still be difficult to determine the extent to which this influenced changes in trends. It should not be concluded from these last remarks that incomplete reporting is favoured. Even though incomplete notifications appear to present a fairly reliable picture, confidence in morbidity data in- creases in proportion to their completeness, and the possibilities for serious sampling error are also reduced. Greatly improved reporting in one field of public health, especially if it is suddenly established, may however cause serious problems in the interpreta- tion of morbidity data. If the only aim is to increase the reliability of notifications, gradual improvement on a broad scale should be sought, completeness being assessed by a sampling procedure. ACKNOWLEDGEMENTS WSF data, including special tabulations, were kindly provided by Mrs S. Zelikson-Singer and her staff. The study was supported by a grant (contract No. 06-327-2) from the US Department of Health, Education, and Welfare. RESUMt HIPATITE VIRALE EN ISRAP-L: INFLUENCE D'UNE ACTION DE PROPAGANDE PARMI LES MEDECINS SUR LES NOTIFICATIONS ET SUR LA SITUATION tPIDEMIOLOGIQUE APPARENTE Afin d'obtenir une meilleure estimation de l'incidence de l'hepatite virale au sein de la population du District central d'Israel, on a mene, de 1965 a 1969, une action de propagande parmi les m6decins de cette collectivit6 visant a accroitre l'efficacit6 du systeme de notification de la maladie. Durant les quatre ans d'application du programme, le nombre des cas d'hepatite virale d6clar6s dans le District central a et trois fois plus eleve que dans le reste du pays, et deux fois plus 61ev6 que pendant les 10 annees prec6dentes. En 1960/61, l'incidence de la maladie estimee d'apres les notifications ne representait que 37,4% de l'incidence calculee d'apres les statistiques de l'assurance- maladie; en 1968/69, le pourcentage atteignait 96,2% et 6tait considere comme 6quivalant virtuellement a la notification de tous les cas. Les progres de la notification ont fait apparaitre des variations saisonnieres moins accus6es de l'incidence de I'affection. Avant l'application du programme, tout accroissement de l'incidence avait apparemment pour effet d'augmenter la proportion des d6clarations, alors que les periodes de declin de la morbidit6 s'accompa- gnaient d'un certain relachement et d'une chute du taux de notification. Cependant, les fluctuations d'ann&e en annee, le cycle saisonnier et l'epoque d'incidence maxi- male de la maladie tels qu'ils ressortaient des taux * apparents * obtenus en periode de d6claration insuffisante n'etaient pas essentiellement differents de ceux fournis par les taux r6els * aprbs amelioration de la notification. Les taux d'incidence specifiques selon l'age sont de- meur6s sensiblement les memes avant et apres la mise en aeuvre du programme. I1 semble cependant qu'avant 463 464 D. BRACHOTT & J. W. MOSLEY 1965 les lacunes de la notification aient e plus grandes en ce qui concerne les enfants de moins de 5 ans et moins nombreuses en ce qui concerne les sujets ages de 20 a 44 ans que dans les autres groupes d'age. Une notification insuffisante, meme si elle ne porte que sur 50% en moyenne du total des cas reels #, ne donne donc pas une idee tres erronee des principaux parametres 6pid6miologiques. Par suite de l'action de propagande menee au sein du corps medical, on a enregistre dans le District central - dont la population represente a peine 20% de la popu- lation totale d'Israel - 42,3 % de tous les cas d'hepatite virale declares dans le pays. L'appreciation des tendances epidemiologiques de l'affection sur le plan national en a e faussee. En outre, on a assiste concomitamment a un accroissement des notifications d'autres maladies ce qui a rendu difficile l'evaluation de leurs tendances. Les auteurs sugg&rent de recourir A des methodes plus progressives et d'application plus etendue si on desire ameliorer la notification de certaines maladies. REFERENCES Kallner, G. (1955) Acta med. orient. (Tel-Aviv), 14, 174-178 Mosley, J. W. & Brachott, D. (1970) Immunoglobulins, Washington, D.C., National Academy of Sciences, pp. 234-243 Mosley, J. W. & Kendrick, M. A. (1969) Bull. N.Y. Acad. Med., 45, 143-163 Reisler, D. J. et al. (1970) Amer. J. Epidem., 92, 62-72