Center for Health Statistics and Information, National Health and Family Planning Commission, Beijing, China. c

Transcription

1 Policy & prctice & prctice An integrted ntionl mortlity surveillnce system for deth registrtion nd mortlity surveillnce, Chin Shiwei Liu, Xioling Wu, b Aln D Lopez, c Lijun Wng, Yue Ci, b Andrew Pge, d Peng Yin, Yunning Liu, Yichong Li, Jingmei Liu, Jinling You & Migeng Zhou Abstrct In Chin, smple-bsed mortlity surveillnce systems, such s the Chinese Center for Disese Control nd Prevention s disese surveillnce points system nd the Ministry of Helth s vitl registrtion system, hve been used for decdes to provide ntionlly representtive dt on helth sttus for helth-cre decision-mking nd performnce evlution. However, neither system provided representtive mortlity nd cuse-of-deth dt t the provincil level to inform regionl helth service needs nd policy priorities. Moreover, the systems overlpped to considerble extent, thereby entiling dupliction of effort. In 2013, the Chinese Government combined these two systems into n integrted ntionl mortlity surveillnce system to provide provincilly representtive picture of totl nd cuse-specific mortlity nd to ccelerte the development of comprehensive vitl registrtion nd mortlity surveillnce system for the whole country. This new system incresed the surveillnce popultion from 6 to 24% of the Chinese popultion. The number of surveillnce points, ech of which covered district or county, incresed from 161 to 605. To ensure representtiveness t the provincil level, the 605 surveillnce points were selected to cover Chin s 31 provinces using n itertive method involving multistge strtifiction tht took into ccount the sociodemogrphic chrcteristics of the popultion. This pper describes the development nd opertion of the new ntionl mortlity surveillnce system, which is expected to yield representtive provincil estimtes of mortlity in Chin for the first time. Introduction Relible nd timely informtion on cuse-specific mortlity is fundmentl for informing the development, implementtion nd evlution of helth policy. 1,2 Chin hs yet to estblish complete vitl registrtion system for its 1.3 billion popultion. To dte, the essentil dt on the cuses of deth in Chin used for decision-mking nd performnce evlution hve come from smple-bsed mortlity surveillnce systems, including the ntionlly representtive disese surveillnce points system of the Chinese Center for Disese Control nd Prevention (CDC) nd the vitl registrtion system of the Chinese Ministry of Helth. 3 The disese surveillnce points system ws estblished in 1978 with pilot study t two surveillnce points in Beijing. 3 By 1990, the number of points hd incresed to 145 nd the popultion covered ws pproximtely 10 million. 3,4 In 2004, the system ws expnded gin to include 161 points nd popultion coverge incresed to 73 million. The smpling strtegy nd the chrcteristics of this system hve been described in detil elsewhere long with the qulity control mesures nd the procedures for collecting dt, coding the cuse of deth nd determining the underlying cuse of deth. 3 5 For deths in hospitl, doctors certified the cuse of deth nd trined coders determined the underlying cuse of deth by pplying the rules of the Interntionl Clssifiction of Diseses. 6 For hospitls without the cpcity to determine or code the underlying cuse of deth, these functions were crried out by the county or district CDC. For deths occurring outside hospitl, villge helth workers or township or community hospitl stff did verbl utopsy from which doctors in these hospitls determined the underlying cuse of deth. Since 2008 informtion on individul deths in ll popultion ctchment res hs been reported in rel time using n Internet-bsed reporting system. 7 In this system, informtion on ech deth is systemticlly vlidted by locl including county, prefecture nd provincil level Centers for Disese Control nd Prevention, which lso check the completeness, coding nd internl logic of the items reported on deth certifictes. Cuses-of-deth re subsequently reported to the ntionl CDC, where dt re consolidted. The Chinese vitl registrtion system ws estblished in the 1950s to collect mortlity dt in 13 cities. By 2000, the popultion under surveillnce ws round 110 million nd the system covered 15 lrge cities, 21 medium-sized or smll cities nd 90 counties drwn from 15 provinces nd municiplities. 8 By 2012, the system hd expnded to include 319 sites (138 counties nd 181 districts) in 22 provinces covering bout 230 million people, mostly in estern nd centrl res of the country. Forty-two counties in the vitl registrtion system overlpped with popultion ctchments in the disese surveillnce points system. The collection of cuse-of-deth dt in the vitl registrtion system ws similr to the disese surveillnce points system. Dt on deths were compiled ccording to predetermined ggregtion principles nd reported monthly by electronic file trnsfer to the Center for Helth Sttistics nd Informtion of the Ntionl Helth nd Fmily Plnning Commission (previously the Ministry of Helth). Qulity control meetings were held nnully nd regulr trining ws crried out to ensure dt qulity. Together, the disese surveillnce points nd vitl registrtion systems provided ntionlly representtive picture of Ntionl Center for Chronic nd Non-communicble Disese Control nd Prevention, Chinese Center for Disese Control nd Prevention, 27 Nnwei Rod, Xicheng District, Beijing, , Chin. b Center for Helth Sttistics nd Informtion, Ntionl Helth nd Fmily Plnning Commission, Beijing, Chin. c School of Popultion nd Globl Helth, University of Melbourne, Melbourne, Austrli. d School of Science nd Helth, University of Western Sydney, Sydney, Austrli. Correspondence to Migeng Zhou (emil: migengzhou@126.com). (Submitted: 23 Jnury 2015 Revised version received: 9 August 2015 Accepted: 10 August 2015 Published online: 28 October 2015 ) 46 Bull World Helth Orgn 2016;94:46 57 doi:

2 Shiwei Liu et l. Policy & prctice Tble 1. Vitl registrtion nd disese surveillnce points mortlity surveillnce systems, Chin, surveillnce system Vitl registrtion system Disese surveillnce points system Estblishment Development Smpling method During the 1950s in 13 cities, including Beijing, Shnghi nd Nnjing In 1978, in Dongcheng district nd Tong county, Beijing By 1985, 28 lrge nd medium-sized cities nd 70 counties were covered; by 2000, 36 cities (i.e. municiplities nd prefecturelevel cities, which included mny districts) nd 90 counties in 15 provinces were covered. By 2012, the system included 319 surveillnce points in 22 provinces By 1989, 71 surveillnce points cross 29 provinces were covered In 1990, 145 surveillnce points cross 31 provinces were covered In 2004, 161 surveillnce points cross 31 provinces were covered Voluntry Voluntry Sites covered Cities (including counties nd districts) nd counties Counties nd districts Popultion coverge By 1985, 60 million; by 2000, 110 million By 2012, 230 million Representtiveness Mostly cities nd estern rurl res of the country Mostly estern nd centrl res of the country Voluntry ND ND Mostly lrge cities nd more welthy rurl res Multistge, strtified, cluster smpling (smpling probbility proportionl to popultion size) Multistge strtifiction, selection, evlution nd djustment 1 or 2 townships in ech county or 1 or 2 subdistricts in ech district for ech surveillnce point The whole popultion covered by ech surveillnce point 10 million (pproximtely 1% of the totl Chinese popultion) 73 million (pproximtely 6% of the totl Chinese popultion) Both ntionlly nd regionlly (i.e. estern, centrl nd western; urbn nd rurl) representtive Both ntionlly nd regionlly (i.e. estern, centrl nd western; urbn nd rurl) representtive Dt reporting Locl CDCs mde monthly reports by electronic file trnsfer to the CHSI of the NHFPC Locl CDCs mde monthly reports by electronic file trnsfer to the CHSI of the NHFPC Locl CDCs mde monthly reports by electronic file trnsfer to the ntionl CDC Locl CDCs mde monthly reports by electronic file trnsfer to the ntionl CDC Since 2008, hospitls nd locl CDCs hve reported to the ntionl CDC using rel-time Internet-bsed system CDC: Center for Disese Control nd Prevention; CHSI: Center for Helth Sttistics nd Informtion; ND: not determined; NHFPC: Ntionl Helth nd Fmily Plnning Commission (previously the Ministry of Helth). Ech surveillnce point corresponded to one county or district. mortlity in Chin. 9 The vitl registrtion system, while not representtive, ws ble to give more ccurte estimtes of the proportion of deths due to specific cuses nd lrger smple of deths thn the disese surveillnce points. The disese surveillnce points system reflected totl mortlity, the brod cuse-of-deth distribution nd the geogrphic distribution of deths more ccurtely, becuse the smpling strtegy employed ensured ntionlly representtive smple. 10 However, neither system ws ble to provide representtive dt on mortlity or the cuses of deth t the provincil level. Differences between the two systems nd their development re described more fully in Tble 1 nd Tble 2. In 2013, the Ntionl Helth nd Fmily Plnning Commission combined the vitl registrtion system nd disese surveillnce points system to crete n integrted ntionl mortlity surveillnce system. The gols were to integrte nd rtionlize the helth resources expended on these systems nd to ccelerte the development of complete Bull World Helth Orgn 2016;94:46 57 doi: 47

3 Policy & prctice Shiwei Liu et l. Tble 2. Reporting cuse-of-deth through vitl registrtion nd disese surveillnce points systems, Chin, Period, by plce of deth In erlier yers, in hospitl In recent yers, in hospitl In erlier yers, outside hospitl Deth informtion collection Fmily members reported to locl vitl registrtion offices in hospitls, hospitls reported to locl CDCs nd locl CDCs prepred summries Stff in the hospitl s disese prevention unit collected deth certifictes nd reported to the locl CDC, which provided summries Fmily members reported to locl vitl registrtion offices in locl hospitls, hospitls reported to locl CDCs nd locl CDCs prepred summries Deth certificte Stff in the hospitl s vitl registrtion office completed the certificte using informtion from fmily members nd ny vilble medicl records nd documents Clinicl doctors in the hospitl completed the certifictes Stff in locl hospitls vitl registrtion offices completed the certifictes using informtion from fmily members nd ny vilble medicl records nd documents Vitl registrtion system Disese surveillnce points system Coding cuse of deth Coding ws done by stff in locl CDCs; ICD-10 clssifiction used since 2002 Coding ws done either by: doctors in the hospitl s medicl records unit or by stff in the hospitl s disese prevention unit, nd locl CDC stff checked the coding; or locl CDC stff when the locl hospitl did not hve the cpcity. ICD-10 clssifiction used since 2002 Coding ws performed by stff in locl CDCs; ICD-10 clssifiction used since 2002 Determining the underlying cuse of deth Determined by stff in locl CDCs Determined either by doctors in the hospitl s medicl records unit or by stff in the hospitl s disese prevention unit, nd checked by locl CDC stff; or by locl CDC stff when the locl hospitl did not hve the cpcity Determined by stff in locl CDCs Deth informtion collection Stff in the hospitl s disese prevention unit collected deth certifictes nd reported to the locl CDC, which provided summries Stff in the hospitl s disese prevention unit collected deth certifictes nd entered the detils onto rel-time, Internetbsed system; locl CDC stff checked the informtion Villge helth workers nd disese prevention unit stff in township hospitls in rurl res nd disese prevention unit stff in CHs in urbn res collected informtion using household surveys; the informtion ws checked with bodies such s the locl police, the civil ffirs deprtment nd the mternl nd child deprtment nd then reported to the locl CDC, which prepred summries Deth certificte Clinicl doctors in the hospitl completed the certifictes Clinicl doctors in hospitls completed the certifictes Stff in locl hospitls disese prevention units completed the certifictes using informtion from fmily members nd ny vilble medicl records nd documents Coding cuse of deth Initilly, coding ws done by stff in the ntionl CDC (previously the Chinese Acdemy of Preventive Medicine); then there ws grdul trnsition to the procedures used in recent yers. ICD-10 clssifiction used since 2004 Coding ws mostly done by locl CDC stff, but lso by doctors in hospitls medicl records units or stff in hospitls disese prevention units, nd checked by locl CDC stff. ICD-10 clssifiction used since 2004 Coding ws performed by stff in the ntionl CDC (previously the Chinese Acdemy of Preventive Medicine); ICD-10 clssifiction used since 2004 Determining the underlying cuse of deth Initilly determined by stff in the ntionl CDC (previously the Chinese Acdemy of Preventive Medicine); then there ws grdul trnsition to the procedures used in recent yers Mostly determined by locl CDC stff, but lso by doctors in hospitls medicl records units or stff in hospitls disese prevention units, nd checked by locl CDC stff Determined by stff in the ntionl CDC (previously the Chinese Acdemy of Preventive Medicine) (continues...) 48 Bull World Helth Orgn 2016;94:46 57 doi:

4 Shiwei Liu et l. (...continued) Vitl registrtion system Disese surveillnce points system Determining the underlying cuse of deth Coding cuse of deth Deth certificte Deth informtion collection Determining the underlying cuse of deth Coding cuse of deth Deth certificte Deth informtion collection Period, by plce of deth Mostly determined by locl CDC stff, but lso by stff in hospitls disese prevention units, nd locl CDC stff checked the informtion Coding mostly done by locl CDC stff, but lso by stff in hospitls disese prevention units, with locl CDC stff checking the informtion; ICD-10 clssifiction used since 2004 Certifictes were completed by clinicl doctors in chrge of emergency tretment or by disese prevention unit stff in chrge of household surveys or of checking field reports Villge helth workers nd disese prevention unit stff in township hospitls in rurl res nd disese prevention unit stff in CHs in urbn res collected informtion using household surveys; the informtion ws checked with gencies such s the locl police, the civil ffirs deprtment nd the mternl nd child deprtment nd then entered onto rel-time, Internetbsed system; locl CDC stff checked the informtion Determined by either: (i) the locl hospitl s disese prevention unit stff, nd checked by locl CDC stff; or (ii) locl CDC stff when the locl hospitl did not hve the cpcity Coding ws done either by the locl hospitl s disese prevention unit stff, nd checked by locl CDC stff; or locl CDC stff when the locl hospitl did not hve the cpcity. ICD-10 clssifiction used since 2002 Certifictes were completed by clinicl doctors in chrge of emergency tretment or by disese prevention unit stff in chrge of household surveys or of checking field reports Villge helth workers in rurl res nd disese prevention unit stff in CHs in urbn res collected informtion using household surveys; the informtion ws checked with bodies such s the locl police, the civil ffirs deprtment nd the mternl nd child deprtment nd then reported to the locl CDC, which prepred summries In recent yers, outside hospitl CDC: Center for Disese Control nd Prevention; CH: community hospitl or community helth centre or sttion; ICD-10: Interntionl sttisticl clssifiction of diseses nd relted helth problems, 10 th revision. Exct time cnnot be given since this ws grdul process nd different sites hve different strting time points. Note: Locl CDC refers to county nd district levels. Policy & prctice vitl registrtion nd mortlity surveillnce system covering the entire popultion of Chin. Initilly, the ntionl mortlity surveillnce system covered popultion of million (24.3% of the totl popultion of the country) nd comprised 605 surveillnce points, with ech point covering n entire county or district. This pper describes the development nd opertion of this new mortlity surveillnce system. Development of new system The ntionl mortlity surveillnce system ws estblished using the sme generl principles pplied in developing the disese surveillnce points system. 3 5 First, the Ntionl Helth nd Fmily Plnning Commission determined tht the surveillnce popultion should be not less thn 5 million in ny province tht hd popultion greter thn 10 million nd ws economiclly well developed; for other provinces, the popultion smple hd to be t lest 20% of the totl popultion. These criteri were used to estblish the number of surveillnce points required in ech province. Second, we divided ll counties nd districts in ech province into eight strt ccording to their degree of urbniztion, popultion size nd the crude mortlity rte (totl number of deths per 1000 people per yer). Third, we selected counties nd districts in ech strtum s cndidte surveillnce points for ech province in ccordnce with the number of surveillnce points required. We then determined how representtive the cndidte surveillnce points were of the whole province using dt from the 2010 census. 11 The finl surveillnce points for ech province were selected using n itertive process tht ensured the combintion of points ws representtive of the popultion of the province (Fig. 1). Surveillnce popultion The finl size of the trget surveillnce popultion in ech province ws bsed not only on the two criteri used by the Ntionl Helth nd Fmily Plnning Commission s described bove but lso on decisions mde by the ntionl CDC in consulttion with provincil CDCs, which took into considertion differences between provinces in popultion nd in the cpbility nd trining Bull World Helth Orgn 2016;94:46 57 doi: 49

5 Policy & prctice Shiwei Liu et l. Fig. 1. Selection of points in the ntionl mortlity surveillnce system in ech province, Chin, 2013 Minimum surveillnce popultion in the province determined Number of surveillnce points required in the province determined Counties nd districts in the province divided into eight strt b Cndidte surveillnce points in ech strtum selected Cndidte surveillnce points reselected Not representtive of the province Representtiveness of selected surveillnce points evluted Not representtive of the province Representtive of the province Representtiveness of selected surveillnce points re-evluted Representtive of the province Surveillnce points finlized An itertive process ws used to determine whether the popultion covered by the combintion of selected surveillnce points in province ws representtive of the popultion of the whole province. If the combintion ws not representtive of the whole province, different points were selected within ech strtum until it ws representtive; however, the number of points in ech strtum ws not chnged. b Counties nd districts were divided into eight strt ccording to their degree of urbniztion, popultion size nd crude mortlity rte. Tble 3. Surveillnce points in the ntionl mortlity surveillnce system, Chin, 2013 Province No. of counties nd No. of surveillnce points b (% of counties districts nd districts) Anhui (22.9) Beijing 18 7 (38.9) Chongqing (27.5) Fujin (23.8) Gnsu (23.0) Gungdong (22.8) Gungxi (19.1) Guizhou (22.7) Hinn 24 8 (33.3) Hebei (17.4) Heilongjing (20.5) Henn (22.6) Hubei (21.4) Hunn (23.0) Inner Mongoli (19.8) Jingsu (25.5) Jingxi (20.2) Jilin (25.0) Lioning (22.0) Ningxi (45.5) Qinghi (21.7) Shnxi (12.1) Shndong (22.1) Shnghi 18 7 (38.9) Shnxi (16.8) Sichun (17.1) Tinjin 16 7 (43.8) Tibet 73 8 (11.0) Xinjing (15.3) Yunnn (19.4) Zhejing (24.4) Totl (21.1) The number of counties nd districts in ech province ws the sme s the number used in the 2010 ntionl census. 11 b Ech surveillnce point covered n entire county or district. 50 Bull World Helth Orgn 2016;94:46 57 doi:

6 Shiwei Liu et l. Policy & prctice Fig. 2. Surveillnce points, ntionl mortlity surveillnce system, Chin, 2013 Former 158 disese surveillnce points 447 ntionl mortlity surveillnce points Not pplicble No dt N km Note: Drk green lines represent 447 counties with ntionl mortlity surveillnce points; light green lines represent 158 counties which lso hve disese surveillnce points. The mp ws produced using QGIS version (QGIS Development Tem). Source: Mp dt for Chin from the Nturl Resources nd Geosptil Bse Informtion Dtbse (Ntionl Development nd Reform Commission, Chin). Mp dt for neighbouring countries from WHO. Fig. 3. Urbniztion, popultion size nd mortlity rte, Chin, 2013 U vlue 1.4 to to to to to 6.4 Not pplicble No dt Other countries N km Notes: The county or district representtion ws clculted s follows: first ech U-vlue for the urbniztion index, popultion size nd mortlity rte ws obtined by using the eqution U = (x μ)/σ, where x is the observed vlue of the prticulr index in the country or district, μ is the men vlue of the index in ll counties nd districts in the province nd σ is the stndrd devition. We then clculted the men U-vlue for the three indices for ech county or districts. The mp ws produced using QGIS version (QGIS Development Tem). Source: Mp dt for Chin from the Nturl Resources nd Geosptil Bse Informtion Dtbse (Ntionl Development nd Reform Commission, Chin). Mp dt for neighbouring countries from WHO. Bull World Helth Orgn 2016;94:46 57 doi: 51

7 Policy & prctice Fig. 4. Strtifiction of counties nd districts in ech province for the development of the ntionl mortlity surveillnce system, Chin Popultion Urbniztion Popultion of helth-cre stff nd their bility to implement the new surveillnce system. For exmple, the surveillnce popultion chosen for Beijing ws lrge becuse the city hs good infrstructure, well-trined workforce nd lrge popultion. In contrst, the surveillnce popultion in western provinces ws smller, prtly becuse the locl cpcity nd resources vilble for crrying out relible mortlity surveillnce were limited. Following detiled considertion of the verge popultion of ll counties nd districts nd fter consulttion with helth uthorities in ech province, we determined tht the ntionl mortlity surveillnce system required totl of 605 surveillnce points the 605 counties nd districts covered by these points comprised 21.1% of ll counties nd districts in Chin (Tble 3; Fig. 2). Strtifiction In the previous two djustments to the disese surveillnce points system in 1990 nd 2004, urbn res (i.e. districts) nd rurl res (i.e. counties nd county-level cities) were used s the primry units for strtifiction nd the urbn-to-rurl popultion rtio ws lso tken into ccount. 3 5 However, with the rpid socioeconomic development of the lst decdes, this rtio is no longer pproprite for defining county or county-level city s rurl re or district s n urbn re. In ddition, per-cpit gross domestic product (GDP) ws used only in the further strtifiction of rurl All counties nd districts in province Popultion Urbniztion Popultion Counties nd districts were divided sequentilly into high nd low ctegories ccording to their urbniztion index, popultion size nd mortlity rte (see min text for detils). res becuse there ws lck of dt on how urbn per cpit GDP vried by district. 3 5 Given the incompleteness of these dt nd the potentil positive correltion between urbniztion nd per cpit GDP, we decided to use the urbniztion index (i.e. the frction of the popultion residing in n urbn re) s strtifying index insted of the urbn-to-rurl popultion distribution or per-cpit GDP. Following consulttions with experts, popultion size ws retined s n importnt strtifying index in the selection of surveillnce points for the Ntionl Surveillnce System, s ws the crude mortlity rte. These three strtifying indices were used s descriptors for ech surveillnce point nd were clculted for ech province. To obtin grphicl illustrtion of the chrcteristics in ech county or district bsed on the three strtifiction indices, we first clculted the representtion (U) for ech index s follows: U = (x μ)/σ, where x is the observed vlue of the prticulr index in the country or district, μ is the men vlue of the index in ll counties nd districts in the province nd σ is the stndrd devition. We then clculted the men U-vlue for the three indices for ech county or district (Fig. 3). The multistge strtifiction process included three steps: (i) counties nd districts in ech province were divided into two strt bsed on the medin urbniztion index for ech province (i.e. high or low urbniztion); Shiwei Liu et l. (ii) counties nd districts with high or low urbniztion index were further subdivided into two strt ccording to the medin popultion size in ech of the two urbniztion strt in ech province (i.e. high or low popultion size); nd (iii) counties nd districts in these four strt (i.e. two urbniztion strt two popultion-size strt) were subdivided into two further strt using the medin totl mortlity rte in ech of these four strt in ech province. This process yielded eight strt in ech province (Fig. 4) nd totl of 248 strt (i.e provinces) cross the whole country. Ensuring representtiveness After strtifiction, counties nd districts in ech province were selected s surveillnce points using inclusion criteri nd principles developed in consulttion with helth dministrtors nd disese surveillnce experts. First, the number of selected counties nd districts in ech strtum of ech province should be pproximtely n/8, where n is the number of counties nd districts in ech province included in the new surveillnce points system (Tble 3). Second, the existing 161 surveillnce points in the current disese surveillnce points system were considered s priority for inclusion in the ntionl mortlity surveillnce system. Then, other counties nd districts with experience in mortlity surveillnce were lso considered for inclusion, s fr s possible. The vitl registrtion system sites were given lower priority during the selection process thn disese surveillnce points system sites becuse the vitl registrtion system did not provide continuous, longitudinl, mortlity dt nd the sites did not ll hve the sme qulity control mesures in plce. In ddition, counties nd districts in which locl stff expressed strong desire to prticipte nd where there ws demonstrble locl government support were lso considered. Finlly, the counties nd districts selected hd to be evenly distributed cross different geogrphicl res with different chrcteristics nd ll prefecture-level cities hd to be included. At ech stge of the selection process, cndidte surveillnce points were evluted to determine how representtive they were of ech province. Reselection nd re-evlution were repeted until the finl popultion smple ws 52 Bull World Helth Orgn 2016;94:46 57 doi:

8 Shiwei Liu et l. considered sufficiently representtive of the province. We employed n itertive process to ensure the representtiveness of the cndidte surveillnce points (Fig. 1). First, the prmeters used to evlute representtiveness were similr to those used in the two previous djustments to the Disese Surveillnce Points system: 8,9 (i) the urbniztion index; (ii) the rtio of the size of the popultion ged 65 yers or more to the size of the totl popultion; (iii)the rtio of the size of the popultion younger thn 15 yers to the size of the totl popultion; nd (iv) the crude mortlity rte. Second, there hd to be no significnt difference between the vlue of given prmeter in the smple popultion nd the corresponding vlue for the whole province, s indicted by sttisticl test with the threshold of n α-level greter thn For vribles tht met the conditions for prmetric tests, t tests were performed on log-trnsformed vribles; otherwise, non-prmetric tests were used. Although we imed to select similr number of counties or districts in ech strtum of ech province, this ws not lwys possible. Becuse of the inclusion criteri nd the vrition in popultion size between counties nd districts, inevitbly the counties nd districts in province did not ll hve the sme probbility of being selected. Consequently, during the sttisticl evlution of representtiveness, we weighted ech selected county nd district ccording to its popultion Finl surveillnce points After severl rounds of representtiveness evlution nd djustment, we found tht there ws no significnt difference in prmeter vlues between the counties nd districts chosen s surveillnce points in ech province nd the entire province for ll provinces in the ntionl mortlity surveillnce system (Tble 4). In totl, 605 surveillnce points were selected cross Chin (Fig. 2): the number of counties nd districts selected in ech province vried from 7 to 36. Three of the 161 former sites in the disese surveillnce points system were excluded becuse of poor dt qulity nd limited locl government support nd cpcity. Of the existing 319 vitl registrtion system sites, 113 were retined in the ntionl mortlity surveillnce system. In 2013, the ntionl mortlity surveillnce system covered million people. At the provincil level, the smple popultion rnged from in Tibet to in Gungdong. In five provinces (Hinn, Ningxi, Qinghi, Tibet nd Xinjing), the surveillnce popultion ws less thn 5 million. The popultion covered by the surveillnce points s proportion of the totl popultion in ech province rnged from 14.5% in Shnxi to 54.2% in Ningxi; it ws over 20% in ech of Policy & prctice the five provinces with surveillnce popultion less thn 5 million (Tble 5). In 2014, the budget llocted by centrl government to run the ntionl mortlity surveillnce system included two types of cost: (i) the cost of bsic deth registrtion procedures (i.e. collection, registrtion, reporting, qulity control, supervision nd trining 6444 United Sttes dollrs, US$, per surveillnce point) nd of periodic surveys of Tble 4. Representtiveness of points in the ntionl mortlity surveillnce system, by province, Chin, 2013 Province Rtio 65 Rtio < 15 Urbniztion rte f index c yers d yers e Anhui Beijing Chongqing Fujin Gnsu g 0.31 Gungdong Gungxi Guizhou Hinn g 0.74 Hebei Heilongjing Henn Hubei Hunn Inner Mongoli Jingsu Jingxi 0.67 g g 0.55 Jilin Lioning g Ningxi Qinghi Shnxi Shndong Shnghi Shnxi Sichun Tinjin Tibet g 0.15 g 0.11 Xinjing Yunnn Zhejing Whether or not the popultion covered by the selected surveillnce points in province ws representtive of the popultion of the whole province ws determined using the four prmeters listed in the tble. b We used either t-test or the Wilcoxon signed-rnk test to clculte if there ws significnt difference between the prmeter vlue cross the surveillnce points in province nd the corresponding vlue for the whole province. c Represents the frction of the popultion residing in n urbn re. d Rtio of the popultion ged 65 yers or more to the totl popultion. e Rtio of the popultion younger thn 15 yers to the totl popultion. f Totl number of deths per 1000 people per yer. g A non-prmetric test ws performed for this prmeter. P b Bull World Helth Orgn 2016;94:46 57 doi: 53

9 Policy & prctice deth underreporting (US$ 4833 per surveillnce point); nd (ii) the cost of work tht vried with the estimted number of deths, such s the printing, distribution nd storge of registrtion crds (US$ 0.97 per deth), the mngement nd nlysis of cuse-of-deth dt (US$ 0.25 per deth) nd interdeprtmentl comprison nd verifiction of dt (US$ 0.25 per deth). Discussion In Chin, the ultimte im is to estblish comprehensive vitl registrtion nd mortlity surveillnce system. However, in the interim, smple-bsed mortlity surveillnce system 15 is the only vible option for generting vlid nd relible informtion on totl nd cuse-specific mortlity in the country. The estblishment of the ntionl mortlity surveillnce system with 605 surveillnce points covering lmost one qurter of the Chinese popultion is highly significnt step towrds the gol of chieving the vitl registrtion of ll births nd deths cross the country by Perhps the gretest dvntge of the new system is tht it will yield nnul dt on deth rtes nd the cuses of deth for ll provinces. The inclusion of most of the existing disese surveillnce points system surveillnce points in the new system ensures the continuity of mortlity dt from these points without ffecting the ntionl or regionl representtiveness of the dt overll. The ntionl mortlity surveillnce system is now the only mortlity surveillnce system in Chin covering ll cuses of deth in people of ll ges. Dt from the 605 surveillnce points will be reported t the time of deth registrtion to the ntionl CDC, 7 which is responsible for the opertion nd mintennce of the informtion system. The Ntionl Helth nd Fmily Plnning Commission is responsible for overll project mngement, policy-mking nd informtion dissemintion. One of the min objectives of the ntionl mortlity surveillnce system is to relibly monitor specific cuses of deth t both ntionl nd provincil levels. Over the long term, surveillnce dt will become incresingly importnt for describing chnges in mortlity, for identifying emerging helth-cre priorities nd for informing helth policy development. Although knowing the extent to which the surveillnce system reflects mortlity ptterns is useful for interpreting dt, representtiveness should not be the only fctor considered when constructing system nd should not be overemphsized t the expense of prcticlity. Mny leding epidemiologists hve rgued tht representtiveness is not impertive, especilly when investigting cusl inference or exmining ssocitions between diseses nd their component cuses Our view is tht building cpcity nd qulity control should be the min priorities in implementing the new surveillnce system in ddition to ensuring representtiveness. The completeness of deth registrtion nd ccurte coding of the cuse of deth nd of identifiction of the underlying cuse of deth re key issues for ny Shiwei Liu et l. mortlity surveillnce system. Previous surveys of the disese surveillnce points system found underreporting of 12 to 17% the proportion ws even higher mong children younger thn five yers nd in rurl res. 3,19,20 Moreover, in 2012, report indicted tht 2.73% of cuses of deth in Chin were coded inccurtely. 21 One of the min chllenges for the new system is the high proportion of deths occurring outside hospitl. Trditionl buril customs, including the desire to return to one s plce of originl residence before deth, men tht pproximtely 70% of deths in rurl res (s much s 90% in some plces) occur t home nd medicl records re limited or nonexistent. Second, stff t most new surveillnce points lck relevnt experience, especilly Tble 5. Ntionl mortlity surveillnce system surveillnce popultion, by province, Chin, 2010 Province Totl popultion Surveillnce popultion Proportion of totl popultion (%) Anhui Beijing Chongqing Fujin Gnsu Gungdong Gungxi Guizhou Hinn Hebei Heilongjing Henn Hubei Hunn Inner Mongoli Jingsu Jingxi Jilin Lioning Ningxi Qinghi Shnxi Shndong Shnghi Shnxi Sichun Tinjin Tibet Xinjing Yunnn Zhejing Totl Popultion dt re from the 2010 ntionl census Bull World Helth Orgn 2016;94:46 57 doi:

10 Shiwei Liu et l. Policy & prctice with stndrdized workflow procedures nd mechnisms for interdeprtmentl collbortion. Third, it my be difficult to recruit enough professionl helth workers, prticulrly t the locl level. Fourth, there is n enormous difference in locl cpcity between the provinces due to lrge vritions in economic development. To meet these chllenges the integrtion nd ppliction of new utomted methods for collecting informtion on the cuse of deth identified by verbl utopsy should be priority. 22,23 Also, uniform trining mterils should be used by ntionl nd provincil triners to strengthen trining; supervision nd informtion-shring should be enhnced nd dditionl technicl nd finncil support should be offered to underdeveloped provinces when necessry. In 2014, centrl government gencies, including the Ntionl Helth nd Fmily Plnning Commission, the Ministry of Public Security nd the Ministry of Civil Affirs, issued n updted officil document imed t strengthening deth registrtion. However, pproprite legisltion to ensure ll deths re registered nd properly certified is lso essentil. Relible informtion on mortlity nd the cuse of deth is essentil for the development of ntionl nd interntionl helth policy nd of progrmmes for preventing nd controlling disese nd preventing injury. Dt from the disese surveillnce points nd vitl registrtion systems hve been extensively used to ssess the burden of disese both regionlly nd ntionlly in Chin nd globlly 9,10,24 26 s well s for other reserch purposes For the future, there re plns to use the ntionl mortlity surveillnce system smple popultions to crry out periodic ntionl surveys of chronic disese, nutrition nd injury. Electronic linkge of dt is becoming esier in Chin nd it my soon be possible to convert these periodic surveys into prospective cohort studies. The ntionl mortlity surveillnce system will not only ply unique nd criticl role in providing helth metrics for Chin but will lso serve s n essentil resource for evluting helth-cre policy t provincil, ntionl nd interntionl levels, prticulrly for the prevention nd control of chronic diseses. Acknowledgements The uthors thnk stff t the 31 provincil Centers for Disese Control nd Prevention. Competing interests: None declred. ملخص نظام مراقبة الوفيات الوطني املتكامل لتسجيل ومراقبة الوفيات بالصني لقد تم استخدام أنظمة مراقبة الوفيات املستندة إىل العينات مثل نظام نقاط مراقبة األمراض التابع للمركز الصيني ملكافحة األمراض والوقاية منها ونظام السجالت املدنية اخلاصة بوزارة الصحة هناك لعدة قرون لتوفري بيانات متثيلية وطنية حول احلالة الصحية ألغراض اختاذ القرارات املتعلقة بالرعاية الصحية وتقييم األداء. إال أن أيا من هذه األنظمة مل يقدم بيانات متثيلية حول الوفيات أو أسباب حدوثها عىل مستوى املقاطعات حتى يوضح احتياجات اخلدمة الصحية اإلقليمية وأولويات السياسة. وعالوة عىل ذلك حدث تداخل بني األنظمة إىل درجة ملموسة مما يؤدي بدوره ملضاعفة اجلهود. وقامت احلكومة الصينية يف عام 2013 بدمج هذين النظامني يف نظام وطني واحد متكامل ملراقبة الوفيات هبدف تقديم صورة متثيلية للوفيات بشكل عام وألسباب الوفيات عىل مستوى املقاطعات وللتعجيل بتطوير سجالت مدنية شاملة ونظام مراقبة شامل للوفيات جلميع أنحاء الدولة. وقد أدى هذا النظام اجلديد التساع مدى املراقبة من 6% من الرشحية السكانية إىل 24%. وقد زاد عدد نقاط املراقبة والتي تغطي كل منها منطقة أو مقاطعة من 161 إىل 605 نقطة. وللتحقق من مصداقية متثيل البيانات عىل مستوى املقاطعات فقد تم اختيار ال 605 نقطة مراقبة بحيث تغطي 31 مقاطعة يف الصني باستخدام طريقة تكرارية تنطوي عىل تصنيف متعدد املراحل وتضع يف احلسبان اخلصائص االجتامعية الديموغرافية للسكان. ترشح هذه الورقة كيفية تطوير وتشغيل نظام مراقبة الوفيات الوطني اجلديد والذي من املتوقع أن يقدم تقديرات متثيلية للوفيات عىل مستوى املقاطعات ألول مرة يف الصني. 摘要适用于中国死亡登记和死亡率监测的全国死亡监测综合系统在中国, 基于样本的死亡率监测系统, 如中国疾病预防亡率监测综合系统的发展 这个新系统将监测人口从控制中心的疾病监测点系统和卫生部的生命登记系统, 中国总人口的 6% 提升至 24% 涵盖区或县的各监控几十年一直用于提供有关健康状态的全国代表性数据, 点数量从 161 个增至 605 个 为了确保省级的代表性, 以进行准确的医疗保健决策和绩效评估 然而, 这两使用迭代法, 其中包括基于人口社会人口学特征的多个系统都无法在死亡率和死因方面提供省级代表性数层分级法, 选择 605 个监控点, 覆盖中国 31 个省 本据, 以告知地区卫生服务需求和政策优先事项 此外, 文介绍了全国新型死亡率监测系统的形成和运行, 预这两个系统在相当大的程度上互相重叠, 因此导致重计将首次得出中国死亡率的省级代表性估计值 复劳动 2013 年, 中国政府将这两个系统纳入全国死亡率监测综合系统, 在总体死亡率和死因特定死亡率方面提供省级代表性现象, 并加速全国生命登记和死 Bull World Helth Orgn 2016;94:46 57 doi: 55

11 Policy & prctice Shiwei Liu et l. Résumé Chine: un nouveu système ntionl intégré pour l enregistrement des décès et l surveillnce de l mortlité En Chine, des systèmes de surveillnce de l mortlité fondés sur des et de l mortlité pr cuse et d ccélérer l crétion d un système échntillons (comme le système à points de surveillnce des mldies exhustif d enregistrement des données d étt civil et de surveillnce de du Centre chinois pour le contrôle et l prévention des mldies et le l mortlité pour tout le pys. Ce nouveu système permis d ugmenter système d enregistrement des données d étt civil du ministère de l l couverture de l surveillnce (de 6% de l popultion chinoise Snté) sont utilisés depuis plusieurs dizines d nnées pour obtenir des couverte uprvnt à 24%). Le nombre de points de surveillnce données représenttives à l échelle ntionle de l sitution snitire, en (chcun couvrnt un district ou un comté) est pssé de 161 à 605. vue d éclirer les processus décisionnels en mtière de snté et d évluer Pour grntir une bonne représenttivité à l échelle provincile, les 605 les résultts. Or, ucun de ces systèmes trditionnellement utilisés ne points de surveillnce ont été sélectionnés de mnière à couvrir les 31 proposit de données représenttives sur l mortlité et les cuses de provinces chinoises à l ide d une méthode itértive impliqunt une décès à l échelle provincile, pourtnt nécessires pour correctement strtifiction à plusieurs degrés qui tenu compte des crctéristiques définir les priorités et les besoins régionux en mtière de politiques sociodémogrphiques de l popultion. Cet rticle décrit l élbortion de snté. Pr illeurs, ces systèmes étient lrgement redondnts entre et le fonctionnement de ce nouveu système ntionl de surveillnce eux, ce qui impliquit donc une dupliction inutile des efforts. En de l mortlité, qui devrit permettre d obtenir pour l première fois 2013, le gouvernement chinois fusionné ces deux systèmes dns un des estimtions représenttives à l échelle provincile de l mortlité système ntionl intégré de surveillnce de l mortlité fin d obtenir en Chine. une imge représenttive à l échelle provincile de l mortlité totle Резюме Единая национальная система контроля за смертностью для регистрации случаев смерти и контроля за смертностью, Китай В Китае системы выборочного контроля за смертностью, распределением по провинциям и смертности с разбивкой такие как система пунктов контроля заболеваний Китайского по причинам смерти. Это было сделано также для ускорения центра по профилактике болезней и борьбе с ними и система разработки всеобъемлющей системы регистрации естественного регистрации естественного движения населения Министерства движения населения и контроля за смертностью для всей страны. здравоохранения, используются на протяжении десятилетий Благодаря этой новой системе количество жителей, входящих для получения национально-репрезентативных данных о в сферу контроля, увеличилось с 6 до 24% от всего населения состоянии здоровья, на основе которых принимаются решения Китая. Количество пунктов контроля, каждый из которых и оцениваются результаты в сфере здравоохранения. Однако ни привязан к округу или району, было увеличено с 161 до 605. В одна из этих систем не позволяла получить репрезентативные целях обеспечения репрезентативности на уровне провинций данные о смертности и причине смерти на уровне провинций, 605 пунктов контроля были отобраны для курирования которые позволили бы определить потребности и приоритеты 31 провинции Китая с помощью метода итерации, включающего политики региональных служб здравоохранения. Кроме того, многоэтапную стратификацию, которая учитывала социальнодемографические особенности населения. В данной статье сферы деятельности двух систем во многом пересекались, вследствие чего происходило дублирование работы. В описывается развитие и работа новой национальной системы 2013 году Правительство Китая объединило эти две системы в контроля за смертностью, которая, как ожидается, впервые единую национальную систему контроля за смертностью для позволит осуществить репрезентативную оценку смертности в получения репрезентативной картины общей смертности с Китае на уровне провинций. Resumen Un sistem ncionl integrdo de vigilnci de l mortlidd pr el registro de defunciones y l vigilnci de l mortlidd en Chin En Chin, los sistems de vigilnci de l mortlidd bsdos en muestrs, todo el pís. Este nuevo sistem umentó l poblción de vigilnci de tles como el sistem de puntos de vigilnci de ls enfermeddes del un 6 un 24% de l poblción chin. El número de puntos de vigilnci, Centro de Prevención y Control de Enfermeddes de Chin y el sistem donde cd uno cubrí un distrito o conddo, subió de Con de registro civil del Ministerio de Slud, se hn utilizdo durnte décds el objetivo de grntizr un representción nivel provincil, los 605 pr proporcionr dtos ncionlmente representtivos del estdo de puntos de vigilnci se seleccionron pr cubrir ls 31 provincis de slud pr tomr decisiones médics y evluciones de rendimiento. Chin medinte l utilizción de un método itertivo que consistí Sin embrgo, ningún sistem ofrecí dtos representtivos en cunto en un estrtificción de etps múltiples que tení en cuent ls defunciones y ls cuss de ls defunciones un nivel provincil con el crcterístics sociodemográfics de l poblción. Este rtículo describe objetivo de informr de ls necesiddes de servicios snitrios regionles el desrrollo y funcionmiento del nuevo sistem ncionl de vigilnci y ls prioriddes de l polític. Asimismo, los sistems se solpbn hst de l mortlidd, el cul se esper que umente ls estimciones un punto considerble, lo que suponí un duplicción de los esfuerzos. provinciles representtivs de mortlidd en Chin por primer vez. En 2013, el gobierno chino combinó estos dos sistems en un sistem ncionl integrdo de vigilnci de l mortlidd pr proporcionr un imgen provincilmente representtiv de l mortlidd totl y de l mortlidd por cuss específics y pr celerr el desrrollo de un registro civil completo y un sistem de vigilnci de l mortlidd pr 56 Bull World Helth Orgn 2016;94:46 57 doi:

12 Shiwei Liu et l. Policy & prctice References 1. Ruzick LT, Lopez AD. The use of cuse-of-deth sttistics for helth sitution ssessment: ntionl nd interntionl experiences. World Helth Stt Q. 1990;43(4): PMID: Strengthening civil registrtion nd vitl sttistics for births, deths nd cuses of deth: resource kit. Genev: World Helth Orgniztion; Yng G, Hu J, Ro KQ, M J, Ro C, Lopez AD. registrtion nd surveillnce in Chin: history, current sitution nd chllenges. Popul Helth Metr Mr 16;3(1):3. doi: PMID: Yng G. [Selection of DSP points in second stge nd their presenttion]. Zhonghu Liu Xing Bing Xue Z Zhi Aug;13(4): Chinese. PMID: Zhou M, Jing Y, Hung Z, Wu F. [Adjustment nd representtiveness evlution of ntionl disese surveillnce points system.] Disese Surveillnce. 2010;25(3): Chinese. 6. Interntionl sttisticl clssifiction of diseses nd relted helth problems, 10th revision. Genev: World Helth Orgniztion; Wng L, Wng Y, Jin S, Wu Z, Chin DP, Kopln JP, et l. Emergence nd control of infectious diseses in Chin. Lncet Nov 1;372(9649): doi: PMID: Ro C, Lopez AD, Yng G, Begg S, M J. Evluting ntionl cuse-of-deth sttistics: principles nd ppliction to the cse of Chin. Bull World Helth Orgn Aug;83(8): PMID: Murry CJL, Lopez AD. The globl burden of disese: comprehensive ssessment of mortlity nd disbility from diseses, injuries, nd risk fctors in 1990 nd projected to Genev: World Helth Orgniztion; Lopez AD, Mthers CD, Ezzti M, Murry CJL, Jmison DT, editors. Globl burden of disese nd risk fctors. New York: Oxford University Press; doi: Ntionl dt. Beijing: Ntionl Bureu of Sttistics of Chin; Avilble from: [cited 2015 Sep 28]. 12. Hu N, Jing Y, Li Y, Chen Y. [Weighting method for Chinese chronic disese surveillnce in 2010.] Zhongguo wei sheng tong ji. 2012;29(3): Chinese. 13. Hou XY, Wei YY, Chen F. [Weighted estimtion methods for multistge smpling survey dt]. Zhonghu Liu Xing Bing Xue Z Zhi Jun;30(6): Chinese. PMID: Pfeffermnn D, Skinner CJ, Holmes DJ, Goldstein H, Rsbsh J. Weighting for unequl selection probbilities in multilevel models. J R Stt Soc Ser A Stt Soc. 1998;60(1): doi: Begg S, Ro C, Lopez AD. Design options for smple-bsed mortlity surveillnce. Int J Epidemiol Oct;34(5): doi: org/ /ije/dyi101 PMID: Rothmn KJ, Gllcher JE, Htch EE. Why representtiveness should be voided. Int J Epidemiol Aug;42(4): doi: org/ /ije/dys223 PMID: Nohr EA, Olsen J. Commentry: Epidemiologists hve debted representtiveness for more thn 40 yers hs the time come to move on? Int J Epidemiol Aug;42(4): doi: dyt102 PMID: Richirdi L, Pizzi C, Perce N. Commentry: Representtiveness is usully not necessry nd often should be voided. Int J Epidemiol Aug;42(4): doi: PMID: Wng L, Wng LJ, Ci Y, M LM, Zhou MG. [Anlysis of under-reporting of mortlity surveillnce from 2006 to 2008 in Chin]. Zhonghu Yu Fng Yi Xue Z Zhi Dec;45(12): Chinese. PMID: Guo K, Yin P, Wng L, Ji Y, Li Q, Bishi D, et l. Propensity score weighting for ddressing under-reporting in mortlity surveillnce: proof-of-concept study using the ntionlly representtive mortlity dt in Chin. Popul Helth Metr. 2015;13(1):16. doi: PMID: Report of cuse-of-deth surveillnce in Chin Beijing: Chinese Center for Disese Control nd Prevention; Ro C, Lopez AD, Yng G, Begg S, M J. Evluting ntionl cuse-of-deth sttistics: principles nd ppliction to the cse of Chin. Bull World Helth Orgn Aug;83(8): PMID: Murry CJL, Lozno R, Flxmn AD, Serin P, Phillips D, Stewrt A, et l. Using verbl utopsy to mesure cuses of deth: the comprtive performnce of existing methods. BMC Med. 2014;12(1):5. doi: dx.doi.org/ / PMID: Yng G, Wng Y, Zeng Y, Go GF, Ling X, Zhou M, et l. Rpid helth trnsition in Chin, : findings from the Globl Burden of Disese Study Lncet Jun 8;381(9882): doi: org/ /s (13) PMID: Wng H, Dwyer-Lindgren L, Lofgren KT, Rjrtnm JK, Mrcus JR, Levin-Rector A, et l. Age-specific nd sex-specific mortlity in 187 countries, : systemtic nlysis for the Globl Burden of Disese Study Lncet Dec 15;380(9859): doi: org/ /s (12)61719-x PMID: Lozno R, Nghvi M, Foremn K, Lim S, Shibuy K, Aboyns V, et l. Globl nd regionl mortlity from 235 cuses of deth for 20 ge groups in 1990 nd 2010: systemtic nlysis for the Globl Burden of Disese Study Lncet Dec 15;380(9859): doi: S (12) PMID: Chen Y, Ebenstein A, Greenstone M, Li H. Evidence on the impct of sustined exposure to ir pollution on life expectncy from Chin s Hui River policy. Proc Ntl Acd Sci USA Aug 6;110(32): doi: PMID: Zhou M, Liu Y, Wng L, Kung X, Xu X, Kn H. Prticulte ir pollution nd mortlity in cohort of Chinese men. Environ Pollut Mr;186:1 6. doi: PMID: Xu G, Sui X, Liu S, Liu J, Liu J, Li Y, et l. Effects of insufficient physicl ctivity on mortlity nd life expectncy in Jingxi province of Chin, PLoS ONE. 2014;9(10):e doi: pone PMID: Bull World Helth Orgn 2016;94:46 57 doi: 57