Possible Effect of Implementing a National Query Program on Site-Specific Cancer Mortality Rates in Taiwan

Examining site-specific cancer mortality across years or regions could help to generate epidemiological hypotheses or to evaluate cancer prevention programs. Given that site-specific cancer mortality is derived from information reported on the death certificate, if medical certifiers report poor-quality cancer-related cause-of-death statements, the validity of site-specific cancer mortality rates is threatened. For example, if a medical certifier reported only “tumor of the lung” on the death certificate, the coder would classify this case as ICD-10 code D38.1 “neoplasm of uncertain or unknown behavior of the lung” instead of ICD-10 code C34 “malignant neoplasm of the lung”. In other words, if a country or region has more death certificates reporting “neoplasm of uncertain nature” by medical certifiers, the quality of site-specific cancer mortality data is less assured. One remedy to solve the above-mentioned problem is to query medical certifiers for more detailed information for coding purposes (National Center for Health Statistics, 2009). Because the availability of staff and resources used for querying varies across regions, the National Center for


Introduction
Health Statistics of the US recommends different priority levels in the query program for regional vital statistics programs. The priority level of query ranges from level 1, indicating cases that should always be queried, to level 6, which is an optional category (Table 1). Priority level 1 is further subdivided into six categories, and neoplasmrelated query is in the second category in level 1, which says "Always query neoplasm for a primary site and to determine if benign or malignant. When a malignant neoplasm is stated to be the underlying cause of death, it is important to determine the primary site." Several studies have examined the possible effects of querying on cause-specific mortality (Hopkins et al., 1989;Swerdlow, 1994;Hanzkuck, 1996;Lahti et al., 2003;Hoyert et al., 2005); however, only one study assessed the possible effects of querying on site-specific cancer mortality rates (Swerdlow, 1994). One limitation of this study was that no uniform systematic program was implemented for querying. The Department of Statistics, Ministry of Health and Welfare of Taiwan initiated a national query program since January 1, 2009, in which all medical certifiers who report "neoplasm of uncertain nature" on death certificates were queried for Priority level 1 contains the minimum level of querying that all State vital statistics programs should use to promote basic integrity of State and national mortality data. Level 1a: Always query if an infrequent or rare cause appears anywhere in the medical certification section. Level 1b: Always query neoplasm for a primary site and to determine if benign or malignant. When a malignant neoplasm is stated to be the underlying cause of death, it is important to determine the primary site. Level 1c: Always query trivial conditions, symptoms and signs, ill-defined conditions, or mechanisms of death are reported alone or as the underlying cause of death on the death certificate. Level 1d: Always query for the reason for "surgery or medical care" when the underlying disease or condition is not reported anywhere on the death record. Level 1e: Always query for an external cause when only nature of injuries, that is, codes classifiable to ICD-10 codes S00-T98, are reported alone on the death certificate. Level 1f: Query when the sequence arrangement of the reported entries is questionable. Level 1g: When any of the selected HIV-related conditions is reported, whether in part I or part II on the death certificate, and there is no mention of HIV (Human immunodeficiency virus) infection, query for HIV. Priority level 2 includes conditions not usually considered as the underlying cause for which querying will help classify the underlying cause of death more specifically. (For example querying whether 'peritonitis' was due to ruptured appendix, ruptured peptic ulcer or others.) Priority level 3 provides more detailed information that would enable the cause of death to be classified more accurately and to a more detailed ICD category. (For example querying whether 'chronic liver disease' was alcoholic or biliary cirrhosis or chronic hepatitis.) Priority level 4 includes those cases in which the certifier may already provide a logical chain of events leading to death but determining the site or location of stated diseases or conditions will lead to a more precise code. (For example querying whether the specific site of 'embolism' was brain or lung or coronary arteries.) Priority level 5 contains queries which would enable the cause of death to be coded to a more precise subcategory within the threedigit category. This level of detail is frequently required for specified special studies or research projects within a defined reporting area, but may not be necessary for general querying. (For example querying whether the specific site of 'cancer of pancreas' was body or head or duct.) Priority level 6 reflects the most thorough recommended level of querying. The conditions in this category are queried for the purpose of obtaining even more explicit statements, thus eliminating the necessity of using the assumption which are allowed under ICD rules. (For example querying whether 'tuberculosis' was pulmonary.) . Thank you very much for your cooperation. Doctor signature: Date: . Notes: 1. Cause of death statistics are an important reference for health policy making and are based on the information reported by medical certifiers. 2. To improve the quality of cause of death statistics, the Office of Statistics will routinely query medical certifiers for clarification or further information for better coding. We want to assure you that the information you provide is confidential and will be handled accordingly. 3. Malignant neoplasm is the number one leading cause of death in Taiwan and accurate cancer mortality data are important in designing cancer prevention programs. If you reported "malignant neoplasm of the liver" or "liver cancer" on the death certificate, the case will be coded ICD-10 code C22. However, if you reported "liver tumor" on the death certificate, the case will be coded ICD-10 code D376 and would not be counted as the leading cause of death. more detailed information for coding purposes. This study sought to examine possible effect of the implementing of national query program on site-specific cancer mortality rates.

National query program in Taiwan
Staffs of the Department of Statistics, Ministry of Health and Welfare have routinely queried all medical certifiers who report "neoplasm of uncertain nature" on death certificates since January 1st, 2009. The format of the query letter was modified from the recommended query letter designed by the National Center for Health Statistics. We designed a specific query letter for specific neoplasm sites (please see the example of liver tumor in Table 2). The query letter was printed on the opposite side of a copy of the death certificate issued by the medical certifier. Besides the query regarding whether the nature of the neoplasm was begin or malignant, we further DOI:http://dx.doi.org/10.7314/APJCP.2014.15.2.793 Possible Effect of Implementing a National Query Program on Site-Specific Cancer Mortality Rates in Taiwan   asked medical certifiers to outline the evidence on which they based the diagnosis. We also added notes on the importance of the query in terms of the quality of cancer mortality data. If the medical certifier had left the hospital, the nosologists in the hospital were asked to provide the information according to the information recorded in the medical chart.

Analysis
We first computed the response rates by year, the demographic characteristics of the deceased, the clinical settings and the major neoplasm sites. We then calculated the percentages of malignancy, benign or uncertain nature among responses by neoplasm site. Third, we determined the change in the number of deaths after medical certifiers had responded to the query letter by cancer site.

Results
The number of death certificates with mention of "neoplasm of uncertain nature" was 1040 in 2009 and decreased to 963 in 2010 and 871 in 2011. Of 2874 query letters sent by the Department of Statistics, Ministry of Health and Welfare from January 1st 2009 to December 31st 2011, 2571 letters were responded to, a response rate of 89.5%. The response rate did not differ by the demographic characteristics of deceased and was lowest in clinics, the Northern region and cases of kidney tumor (Table 3).
Among the 2571 responses, in 1398 cases (54%) medical certifiers were not able to determine the nature of the neoplasm (Table 4). There were four neoplasm sites of which more than 50% of the responses changed in nature into a malignant nature, i.e., the gastrointestinal system (73%), urinary system (60%), stomach (55%) and rectum (53%). On the contrary, there were two neoplasm sites with a noteworthy number of death certificates on which the nature of the neoplasm changed from an uncertain nature to a benign nature, the pituitary (56%) and brain (14%). Table 5 illustrates the number of cancer deaths added by cancer site after medical certifiers responded to the query letter. The liver was the cancer site that showed the largest absolute increase in the number of deaths after query, an increase of 363 deaths between 2009 and 2011. However, the brain exhibited the largest relative increase, a 12% increase from 1293 to 1474 cases.

Discussion
The findings of this study indicate that among responses to the query letter, in more than half of cases medical certifiers were still not able to determine the nature of the neoplasm. The gastrointestinal system, urinary system, stomach and rectum were the four sites showing the largest change from an uncertain nature to a malignant nature, and the pituitary and brain were the two sites presenting the largest change from an uncertain nature to a benign nature. The liver was the cancer site that showed the largest absolute increase in the number of deaths after query; however, the brain exhibited the largest relative increase.
Five empirical studies have assessed the effects of the query program, three of which covered all causes of death (Hopkins et al., 1989;Lahti et al., 2003;Hoyert et al., 2005) . Of 1033 death certificates queried Finland in 2005, 53 (5.1%) asked for specification of the nature (malignant or benign) of the neoplasm and 54 (5.1%) asked for specification of the primary site of a malignant neoplasm (Lahti et al., 2003). In Finland, the most often queried cancer sites were the liver, lung and peritoneum. However, the authors in the Finnish study did not present the sites most often queried in neoplasms of an uncertain nature. In Taiwan, the most often queried sites were the liver, brain and lung. In other words, medical certifiers are prone to report an uncertain nature for particular neoplasm sites (Lahti et al., 2003).
Regarding the fact that only in half of cases were medical certifiers able to provide detailed information on the nature of the neoplasm, one possible explanation was the high percentage of people who die at home in Taiwan. District public health physicians are assigned to issue death certificates in these cases, and on many occasions the information provided by the families is not specific enough for detailed certification .
In terms of the effect of querying on cancer mortality rates, the increase was 2% in Oregon State (Hopkins et al., 1989) and only 0.7% in Washington State (Hoyert et al., 2005). However, the two US studies did not illustrate the effect of querying on site-specific cancer mortality rates. A study in England and Wales suggested that the percentage increase in the number of deaths after enquiry was trivial for some cancer sites with a large number of deaths and was substantial for cancer of the eye at 35% and cancer of the body of the uterus at 31%, cancers of the thymus, heart and mediastium at 18%, and pleural cancer at 17% (Swerdlow, 1994). Similarly, we found that the greatest absolute increase in the number of deaths was for liver cancer (an increase of 363 cases), but this only constituted a 1.8% increase. In contrast, there was an increase of 181 brain cancer deaths, which represented a 12% of increase because of the small denominator.
The strength of this study is that it was the first study to assess the possible effects of implementation of a nationwide systematic query program on national site-specific cancer mortality. One of the limitations of this study was that we were unable to verify information provided by medical certifiers. Further studies are needed to link cause of death mortality data with the cancer registry in order to better estimate the underreporting of site-specific cancer mortality.
In conclusion, different neoplasm sites showed different magnitudes of change in the nature of the neoplasm after query. Brain cancer mortality showed the largest increase after query. The query program should be persistently implemented to ensure the comparability of site-specific cancer mortality across years or regions.