report

ww.sciencedirect.com

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 7

Available online at w

Public Health

journal homepage: www.elsevier.com/puhe

Review Paper

Challenges to the census: international trends and a need to consider public health benefits

R.T. Wilson a,*, S.H. Hasanali b, M. Sheikh c, S. Cramer d, G. Weinberg e, A. Firth f, S.H. Weiss g, C.L. Soskolne h,i

a Department of Public Health Sciences, Pennsylvania State University College of Medicine, 500 University Drive,

Hershey, PA, 17033, United States b Department of Sociology and Criminology, Program in Demography, Pennsylvania State University, State College,

PA 16801, United States c Department of Economics, Carleton University, Ottawa, Ontario K1S 5B6, Canada d Royal Society for Public Health, John Snow House, 59 Mansell Street, London, E1 8AN, United Kingdom e Bureau of Epidemiology, Pennsylvania Department of Health, Health and Welfare Building, 625 Forster Street,

Harrisburg, PA, 17120, United States f Royal College of Paediatrics and Child Health, 5-11 Theobalds Road, London, WC1X 8SH, United Kingdom g Department of Medicine, Rutgers New Jersey Medical School, 30 Bergen Street, Suite 1614, Newark, NJ, 07103,

United States h University of Alberta, 11405 87 Ave, Edmonton, Alberta T6G 1C9, Canada i Health Research Institute, University of Canberra, University Drive, Bruce ACT 2617, Australia

a r t i c l e i n f o

Article history:

Received 23 March 2016

Received in revised form

10 May 2017

Accepted 21 May 2017

Available online 29 July 2017

Keywords:

Census

Epidemiology

Disease surveillance

Standardized incidence ratio

Disease cluster

Population survey

* Corresponding author. E-mail addresses: rwilson@psu.edu (R

(M. Sheikh), gweinberg@pa.gov (G. Weinberg ualberta.ca (C.L. Soskolne). http://dx.doi.org/10.1016/j.puhe.2017.05.015 0033-3506/© 2017 Published by Elsevier Ltd o

a b s t r a c t

The Canadian government decision to cancel the mandatory long-form census in 2010

(subsequently restored in 2015), along with similar discussions in the United Kingdom (UK)

and the United States of America (USA), have brought the purpose and use of census data

into focus for epidemiologists and public health professionals. Policy decision-makers

should be well-versed in the public health importance of accurate and reliable census

data for emergency preparedness planning, controlling disease outbreaks, and for

addressing health concerns among vulnerable populations including the elderly, low-

income, racial/ethnic minorities, and special residential groups (e.g., nursing homes).

Valid census information is critical to ensure that policy makers and public health prac-

titioners have the evidence needed to: (1) establish incidence rates, mortality rates, and

prevalence for the full characterization of emerging health issues; (2) address disparities in

health care, prevention strategies and health outcomes among vulnerable populations; and

(3) plan and effectively respond in times of disaster and emergency. At a time when budget

and sample size cuts have been implemented in the UK, a voluntary census is being

debated in the US. In Canada, elimination of the mandatory long-form census in 2011

resulted in unreliable population enumeration, as well as a substantial waste of money and

resources for taxpayers, businesses and communities. The purpose of this article is to

.T. Wilson), stephaniehowe14@gmail.com (S.H. Hasanali), munir_sheikh@hotmail.com ), Alison.Firth@rcpch.ac.uk (A. Firth), weiss@njms.rutgers.edu (S.H. Weiss), colin.soskolne@

n behalf of The Royal Society for Public Health.

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 788

provide a brief overview of recent international trends and to review the foundational role

of the census in public health management and planning using historical and current

examples of environmental contamination, cancer clusters and emerging infections. Citing

a general absence of public health applications of the census in cost-benefit analyses, we

call on policy makers to consider its application to emergency preparedness, outbreak

response, and chronic disease prevention efforts. At the same time, we call on public

health professionals to improve published estimates of monetary benefit (via either cost-

benefit or cost-effectiveness analysis) to a given public health intervention.

© 2017 Published by Elsevier Ltd on behalf of The Royal Society for Public Health.

Public health impact statement

Accurate census data are relied upon by epidemiologists and

public health professionals to establish standardized inci-

dence rates, mortality rates, and prevalence for the reliable

identification and characterization of population groups for

targeted public health intervention. The census is the chief

source of information for denominators used for deriving and

comparing rates of illness or conditions and contributing to

the calculation of social deprivation scores and wider de-

terminants of health within a locality. In general, cost-benefit

analyses of the census do not consider public health benefits.

Public health professionals should improve documentation of

the actual and projected benefits of public health action. Pol-

icy makers considering cuts to the census should consider a

full cost-accounting of hampered, disjointed, and ill-focused

public health responses in relation to potential and current

disease threats.

Background and purpose

Over the past few years, trends to cut or severely hamper

census and public health information have occurred inter-

nationally. In Canada, response rates to the long-form census

dropped substantially (as low as 25% in a number of com-

munities with some communities providing zero responses)

after it was made voluntary in 2011. Canadian census esti-

mates among the poor and under-served are now considered

to be unreliable by experts.1

In the United Kingdom (UK), the Office for National Sta-

tistics (ONS) is a key UK governmentefunded provider of na-

tional statistics. Census data are used in the UK to inform local

and national public health intelligence for the commissioning,

design and delivery of community services, and for the

investigation of and handling of disease clusters, outbreaks,

and vaccination. Between 2008 and 2015, the ONS budget was

cut by 25% in real terms.2 The sample size of the UK General

Lifestyle Reports (GLR) was cut by one-third. The GLR includes

national UK data on smoking, drinking, and chronic health

conditions.

Citing the right to privacy as the justification, a 2015 US

House of Representatives legislative amendment proposed

eliminating penalties for non-response to the US Census' American Community Survey (ACS), previously known as the

‘long-form’ of the census. In 2012, the US House of Repre-

sentatives voted to prohibit the use of funds for conducting

the ACS. The ACS uses a sample of approximately 1% (3

million households nationwide), and the data collected facil-

itates the distribution of federal assistance, including the

Bureau of Economic Analysis' per capita income series and the Census Bureau's annual population estimates.

The purpose of this article is to: (1) provide an overview of

recent proposed changes to the census in three different

countries; (2) discuss the role of the census for effectively

addressing public health threats and developing public health

interventions; and (3) call on policy makers to consider costs

to taxpayers of inadequate public health response to chronic

disease prevention, emerging disease threats, and emergency

preparedness in the absence of accurate census information.

What is a census?

A census is the official enumeration of a population defined

according to administrative boundaries which includes

enumeration by age, sex, race/ethnicity, and occupation,

among other demographics. The term derives from the Latin

word c�ens�ere meaning ‘to assess’ as governments historically

have used the census as a means of taxation. Debates about

the census are not new. The census of Quirinius sparked a tax

rebellion in Roman Syria and fueled the Zealot movement in

the year 6 CE.3

As required in article 1, section 2 of the United States (US)

Constitution, the census is central to political, economic, and

public health planning. The US Supreme Court has referred to

the census as the ‘linchpin of the federal statistical system … col-

lecting data on the characteristics of individuals, households, and

housing units throughout the country.’4

There are three major approaches used internationally for

the collection of census data:

1. Traditional (used by 115 countries, including, e.g., Canada,

the UK, and Australia);

2. Population register (used by about 20 countries, particu-

larly in Scandinavia); and

3. Rolling census/continuous measurement (used by the US

and France).

Table 1 provides a summary of the approaches to the

census and brief overview of the strengths and limitations.

Table 1 e Summary of strengths and limitations of various approaches to the population census.

Approach Strengths Limitations

Population register 1. Lower cost relative to a decennial

census

2. Can conduct randomized sampling for

targeted assessment of specific issues

3. If local registration required, might in-

crease local control over population

statistics

1. Does not include group units such as

household, housing unit and the

workplace. Linkage between group and

individual data not necessarily

available

2. Data protection measures may be

beyond the capacity of small localities

to protect

3. Infrastructure and educational ex-

penses to ensure individual reporting

4. May not be accepted in countries or

localities with a cultural suspicion of

government

5. Quality of statistics may be poorer

Administrative record files

from governmental programmes

1. Low cost

2. Ease of access for electronic records

1. May require legislative changes (e.g.

amending Title 13 of the US Code)

2. Unknown/unpublished population

coverage of administrative files with

respect to a gold standard

3. Use of data for a purpose other than

initially collected may reduce trust in

government

4. May require integration at multiple

levels (local/national/regional)

Rolling census 1. More frequently conducted than a

decennial census

2. Focuses on data collection at a finer

geographic scale

1. Not a snapshot at one point in

timedshort-term variations have to be

interpreted with caution and under

certain conditions

2. Loss of comparability between small

areas due to temporal differences

Traditional decennial/quinquennial

census

1. Generally highest quality information

at a specific time point in small areas.

2. Provides benefit to reducing sample

size and costs of other national survey

and polling efforts

1. High expense relative to use of a pop-

ulation register and/or administrative

data

2. Online delivery has been subject to

cyber attack

3. Public willingness to respond has

declined significantly

4. Works under a mandatory requirement

for the census in order to cover all

segments of the population

Redfern P (1989), Population registers: some administrative and statistical pros and cons. Journal of the Royal Statistical Society Series A, Volume

152, Part 1, pages 1e41.UK ONS (2003) Information paper Census strategic development review. Alternatives to a Census: Review of previous UK

studies (https://www.ons.gov.uk/file?uri¼/census/2011census/whywehaveacensus, Date last accessed 11 November 2016). National Research Council (NRC). 2010. Envisioning the 2020 Census. Washington, DC: The National Academies Press. doi: 10.17226/12865. National Institute of

Statistics, France (INSEE) (2008) Census Quality of French Rolling Census. (http://unstats.un.org/unsd/censuskb20/KnowledgebaseArticle10692.

aspx, Date last accessed 11 November 2016). BBC (2016) Australian census attacked by hackers. August 10, 2016 (http://www.bbc.com/news/

world-australia-37008173, Date last accessed 11 November 2016).

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 7 89

The traditional approach collects data at a specific point in

time on a 5-year and/or 10-year cycle. In Canada, the census

questionnaire (i.e., short-form) is supplemented by informa-

tion from a representative sample of more questions (i.e., long

form). The first census in Canada was conducted in 1666 and

included only age, sex, marital status, and occupation and was

expanded by 1871 (after confederation) to 211 questions

(including land holdings, vital statistics, religion, education

and so forth).

In the UK, there are separate coordinated decennial cen-

suses conducted by agencies located in England and Wales,

Scotland and Northern Ireland, including the ONS for England

and Wales, the General Register Office for Scotland, and the

Northern Ireland Statistics and Research Agency. The most

recent (2011) census contained 14 basic household questions,

and 42 socio-economic status (SES) questions for each member

of the household. The census provides an unparalleled source

of consistent data used by government departments, local

authorities and public sector organizations to underpin local

funding allocation decisions for the distribution and provision

of public services, including education and health care.

The population register approach uses population and

building/dwelling registries and may link them with existing

administrative registries/data sources. Each time someone

moves to a different residence, they are required to register

their change of address with a local magistrate. Additional

information may be obtained through survey sampling

questionnaires.

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 790

A population register is formally defined as

‘an individualized data system, that is, a mechanism of contin-

uous recording, and/or of coordinated linkage, of selected infor-

mation pertaining to each member of the resident population of a

country in such a way to provide the possibility of determining up-

to-date information concerning the size and characteristics of that

population at selected time intervals’.5

The rolling census/continuous measurement approach is

conducted continuously with a rotating sampling frame.

France uses a rolling census approach and schedules

enumeration annually for select small population units. This

provides more up to date information on smaller localities,

although estimates for the entire country are not point in

time, but rolling averages over a specified period of time.

The US Census is a hybrid of the traditional and continuous

measurement approach, making an enumeration of the US

population every 10 years (on April 1st) with a restricted set of

10 questions.6 Since the first US Census in 1790, there have

been 22 federal enumerations of the US population.7 After

2000, the US Census long-form became the ACS, collecting not

only the 10 basic short-form questions, but also questions

about demographic, housing, social, and economic data. The

ACS reports at one-, three-, and five-year intervals, with the

most accurate data at five-year intervals for all communities,

and data for communities of 65,000 or more at one-year in-

tervals.8 The ACS is used to determine distribution of more

than $400 billion in federal and state funds.

Threats to the census

Cost and privacy

In general, threats to the census have come about through a

number of concerns surrounding cost and privacy. Issues

surrounding cost of the census are not without meritdthe

Committee on National Statistics of the National Academies

of Sciences, Engineering, and Medicine has provided an

extensive list of recommendations to substantially increase

efficiency and reduce costs in the US.9 While the compulsory

nature of the census heightens privacy concerns, census

employees in several countries are required to take an oath

and are subject to steep fines and penalties for breach of

confidentiality.10,11 Nonetheless, it is expected that issues of

cost and privacy will remain paramount as countries move to

online formats.12,13

Canada

In June 2010, the long form of the census (LFC) was replaced

with the voluntary National Household Survey (NHS) when

Prime Minister Stephen Harper and the Conservative Party

majority in parliament decided to let Canadians opt out of the

quinquennial census, arguing that the mandatory LFC was too

invasive.14 In contrast, other surveys (such as all business

surveys and the agricultural census) remained mandatory.

To offset the anticipated drop in response rate, the sample

size was increased from one in five to one in three. Indeed the

overall response rate tumbled by about 30% points to less than

70%, and the self-selected sample, despite increasing the num-

ber of respondents, was not representative of the population.

In addition, the population count excluded a higher pro-

portion of citizens from low-income and minority groups.15

As a consequence, reliable population counts could not be

provided for a quarter of all places, including some key socio-

economic indicators. The quality of data for the voluntary

NHS among low-income, immigrants, aboriginals, the less

educated, and mobile students is particularly poor.1 In addi-

tion, the increased sample size resulted in $22 million more in

expendituresdfor a total cost of $652 million.16 In November

2015, the mandatory long-form census was reinstated by the

newly elected Canadian government for implementation in

2016. However, concerns remain that some Canadians may

not realize that the change has occurred, affecting response

rates.16

England and Wales

Between 2008 and 2014, the ONS was subject to funding cuts of

approximately 25%. Proposals made following consultation

resulted in a new focus on producing annual population sta-

tistics derived from online delivery and use of other admin-

istrative data.18 Although online delivery with use of

administrative data will likely produce more timely popula-

tion statistics, it is of concern whether online data collection

will lead to a loss in precision and accuracydespecially for

populations with reduced ability/willingness to access the

internet. Heavy reliance on administrative data is not viewed

as a viable option as there is not a population register to un-

derpin such an approachdother countries that have moved to

reliance on administrative data also have population

registers.19

United States of America

In the US, the fiscal year 2016 Commerce, Justice, and Science

Appropriations Bill (H.R. 2578) cut the ACS budget by $42

million. The 2015 Poe amendment, written by Ted Poe, a

conservative former judge representing the Houston suburbs,

wrote the amendment that House Republicans included in the

annual spending bill that covers the Census Bureau with the

goal of making the ACS voluntary. The ACS, an ongoing part of

the decennial census, provides irreplaceable, high-quality

annual data on key social and economic characteristics for

every community in the country. The ACS not only targets

households, but also people living in special groupsdsuch as

nursing homes, college campuses, and military installations.

The specific objections cited by Representative Poe included

questions about whether or not a household has a refriger-

ator, and the number of faucets, and toilets in the house.17

The European Union

In the mid-1990s, the European Union (EU) was faced with an

appeal from data protectionists whose main arguments were

rooted in the individual's right to privacy. They persuasively argued that the linkage of records from disparate data sys-

tems should not be permitted. In the absence of a counter

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 7 91

position in the development of the ensuing EU directive, a 45-

page law was about to be passed that would have required all

EU member countries to harmonize legislation consistent

with it. Such legislation would have disallowed any form of

record linkage, including for public health purposes. A

groundswell of appeal to exclude public health research from

the EU directive followed and a half-page exclusion at the end

of the 45-page directive resulted.20 The need for vigilance on

the part of researchers (i.e., data accessionists) is a lesson

learned; data accessionists and data protectionists need to

find a forum to discuss their distinct positions.

Cost-benefit analysis of the census

Cost-benefit analyses of the census are generally confined to

the allocation of funding from the central to local govern-

ments in the USA and Canada. In the UK, the cost-benefit

analysis for the 2011 census assumed six main uses:

resource allocation, investment targeting, service planning,

policy making and monitoring, research, and as a benchmark

for other national statistics.21 In 2007, the benefit of the UK

census was projected to be £945 million over a 10 year period

(2012/13e2021/22), compared with costs of £482 million;

however, reassessment of these benefits in 2014 identified

total projected benefits of Author: £4.9 billion? over 10 years.21

While the ONS in the UK includes case studies of benefits of

the census placed on their website and highlights the uses by

local authorities and emergency services and governmental

entities, including the electoral commission and the National

Archives, public health benefits are not specifically

mentioned.22

There is recognition of an increased need for well-

documented rigorous cost-effectiveness analyses (e.g., costs

per life saved or disability-free life years gained, etc.) related to

public health interventions,23 and cost-benefit analyses for

the census itself.9 With as many as one-third of deaths in

developed countries classified as preventable, there remains

large potential margin of benefit for a given public health

intervention.24 Cost-benefit analysts of the census have clas-

sified the determination of benefit for non-allocative uses of

the census (i.e., such as public health interventions) as ‘highly

uncertain.’25 Consequently, quantification of benefit for the

census has typically considered benefits relating to direct

allocation of tax dollars. For example, statistics Scotland

estimated the allocative benefit of the census to exceed the

cost nine-fold, without even accounting for public health

benefits.26

A comprehensive cost-benefit analysis of the census as

conducted in the UK21,26 and New Zealand27 has not been

repeated in the US or Canada. Instead, in the latter countries,

smaller-scale analyses have focused on certain aspects, e.g.,

cost savings using the internet,28 federal funding alloca-

tions,29 implications of eliminating the census long form on

accuracy of population statistics,8 and cost-benefit of certain

questions on the ACS,30 to name a few. Major conclusions

from these published reports are that the cost per household

of the decennial census is increasing, whilst the benefits to be

accrued from the census (costing more) have not been

calculated.31

In one of the most recent cost-benefit analyses, Spencer

et al.25 found that it was more cost-effective for South Africa

to improve postcensal population estimates than to conduct

the 2016 quinquennial census. Their analysis, and one con-

ducted for New Zealand,27 indirectly considers the public

health consequences of a census through the federal alloca-

tion formula which considers medical needs by location.

However, despite widespread use of these formulas, they have

received little formal scrutiny.32 Other important aspects of

public health benefits of the census have not entered into

formal cost-benefit analyses.

The multiple uses of census information in public healthdconsiderations for cost- effectiveness and cost-benefit analyses

Cancer control

The ability to measure disease rates and identify variation

between areas as small as neighborhoods is a critical aspect of

not only investigating suspected disease clusters, but also to

efficiently and effectively address local resident concerns.

Surveillance of cancer rates by geographic area are also useful

for testing and generating hypotheses about disease occur-

rence, causes, and outcomes and may be used to target public

health programs. Such an example of a current-day applica-

tion started with the collection of population-based cancer

incidence data by the National Cancer Institute for the Pitts-

burgh Standardized Metropolitan Statistical Area during the

three-year period from 1969 to 1971 for the third national

cancer survey.33 Employing 1970 US Census findings, Alle-

gheny County was partitioned into 30 areas, with 15 as

neighborhoods inside the City of Pittsburgh. This was per-

formed by aggregating contiguous census tracts demon-

strating similar socio-economic and/or race characteristics

utilizing the information captured by the US Census. Geo-

coded cases were allocated to the respective areas and inci-

dence rates were calculated. Rates were examined by gender,

age, race, years of education, income, occupation, and

geography.34

Lung cancer rates for men followed the pattern predicted,

demonstrating a two-fold higher risk in low socioeconomic

status (SES) populations often close to industrial sites,

compared to high SES areas.35 The observation led to studies

of exposures to carcinogens from air pollutants and tobacco

smoke. More recent studies of rates have utilized 1990, 2000,

and 2010 census information in combination with Pennsyl-

vania Cancer Registry (PCR) cases for 2000 to 2010.36 The

design, sampling, and administration of detailed population

surveys to assess carcinogenic exposures to hydrocarbons in

these populations relied on data from the census.

Similar studies of breast cancer incidence and mortality

rates throughout Allegheny County revealed elevated inci-

dence rates associated with women residing in neighbor-

hoods of higher income and educational level, while death

rates showed an inverse relationship. The highest death rates

appeared in neighborhoods composed of households with

low-income black women. This led to an examination of fac-

tors associated with mortality for the years 2000e2010,

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 792

including healthcare coverage, disease stage at diagnosis,

histology, and hormone receptor status of tumors. PCR re-

cords for this period suggested that receptor status for black

women is a significant determinant of survival.37

Asthma

Since asthma is a growing, serious, and sometimes fatal

condition, it is essential that populations within urban envi-

ronments that demonstrate the greatest need for acute, well-

trained responders be identified. A recent study by Raun et al.

examined census tract characteristics of high treatment

areas. Emergency medical services response rates and the

demographic characteristics of these areas were compared to

other urban area populations.38 Census block groupings were

used to characterize the neighborhoods and to calculate rates.

The greatest intervention needs were identified for adult

males from poor black neighborhoods most frequently after

doctors' office hours.

Environmental threats

Concerns about potential environmental health problems

frequently emerge from perceived disease clusters or origi-

nate with a documented public exposure to a disease agent.

When studies require determining whether environmental

factors contribute to clustering in small populations, infor-

mation on populations residing in areas as small as a census

block group or zip code is of key importance.

In 2000, the Pennsylvania Department of Health (PADOH)

conducted a descriptive cancer incidence study of the popu-

lation residing near the site of a 50,000 gallon underground

gasoline spill.39 The study examined the types of cancer, as

well as the incidence rates based on cases diagnosed between

1985 and 2000. Population counts and age-distribution in the

exposed region were determined by a review of US Census

block-group data, and by a PADOH door-to-door enumeration

of households not defined by US Census groupings. The ex-

pected number of new cancer cases (based on the age-specific

incidence rate for the entire state and the population age-

distribution for the population residing in the exposed re-

gion) were compared with the observed number of cases for

each cancer site among the exposed population.

More recently, following resident concerns, the Missouri

Department of Health and Human Services (DHHS) conducted

a 2014 follow-up study of cancers diagnosed in North St. Louis

within a population spanning eight zip codes that were

potentially exposed to radioactive waste from old Manhattan

Project dumping sites. Using cases diagnosed from 1996 to

2011 and the 2000 and 2010 census Zip Code Tabulation Area

data, the DHHS found a significantly decreased incidence of

cancers of the thyroid and cervix, but an increased incidence

of leukemia as well as an increased incidence of cancers of the

colon, female breast, prostate, bladder, and kidney/renal

pelvis, when compared with rates for the rest of Missouri.40

This study allowed the DHHS to seek and obtain federal sup-

port for an ongoing investigation into the possible links to

cancer and other health effects by of exposure to radioactive

waste leaking into Coldwater Creek and surrounding

neighborhoods, citing radium-226, thorium-230, and

uranium-238 as particular contaminants of concern.41,42

In addition to the descriptive rates of fixed populations in

one time point are studies of cohorts over a specified time

period. Cohort studies have examined exposed residents from

the Three-Mile Island (TMI) nuclear power plant accident in

1979. The TMI cohort studies examined the incidence of

health problems longitudinally for residents within a five-mile

radius of the TMI facility. These included cancer incidence,

major underlying causes of death, birth outcomes, and psy-

chological stress.43 Census data and enumerators trained for

these studies were utilized to define the study cohorts for

follow-up.

Epidemiologic methods and mapping

Census data have also been useful for identifying study par-

ticipants and controls for a number of other studies including

those with case-control designs. These are constructed to

estimate odds ratios for individuals with a characteristic or

risk factor believed to be associated with an exposure of

concern. Last, census data are essential for planning and

conducting population-based studies to measure disease

prevalence and population characteristics and understanding

exposureedisease relationships. Several countries produce

atlases which map rates of disease, for example the Environ-

ment and Health Atlas for England and Wales, an independent

publication produced by the Small Area Health Statistics Unit.

Emerging disease threats

In 2005, the Joint Committee on Vaccination and Immunisa-

tion revised guidelines for bacille Calmette-Guerin (BCG)

vaccination in the UK. The new guidelines discontinued the

requirement for all school children to have received BCG

vaccination, but focused intervention on infants (0e12

months) living in areas where the annual incidence of tuber-

culosis (TB) is 40 per 100,000 or greater.44,45 This targeted

intervention would not be possible without accurate local

census numbers used to determine the annual incidence rate.

To our knowledge, the resulting savings with respect to vac-

cine costs and expected/actual number of TB cases prevented

have not yet been considered in a cost-benefit analysis of the

census.

In 2016, the New York City Department of Health and

Mental Hygiene (DOHMH) assessed Zika virus preparedness

by census tract.46 Census projections from the ACS

(2010e2014) by census tract allowed epidemiologists and

public health professionals to track the rate of Zika virus

testing per 100,000 among females of child-bearing age (Fig. 1)

and to identify census tracks within the city with low rates of

coverage and possibly at higher risk of virus transmission due

to a higher proportion of persons born in regions of the world

with active-Zika transmission (Fig. 2). The DOHMH investi-

gation revealed that the prevalence of Zika virus testing was

highest in census tracts with the lowest proportion of immi-

grants from active-Zika transmission countries. In response,

DOHMH immediately distributed multilingual educational

materials to healthcare providers in at-risk communities,

Fig. 1 e Zika virus testing rate per 10,000 among females aged 15e44 years, by US Census tract of residencedNew York City,

JanuaryeFebruary 2016. (Reprinted with Permission).46

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 7 93

provided public service advertisements to Spanish language

media outlets, and handed out informational cards, as well as

travel warning flyers for pregnant women in at-risk commu-

nities. In addition, DOHMH conducted hundreds of pre-

sentations at social, community, and religious gatherings

throughout the city. By May 2016, the DOHMH had increased

the Zika testing rate in census tracts with the highest quartile

of immigrants born in active-Zika transmission countries to

65 per 100,000dup from 29 per 100,000 just months before.

Avoiding the long-term consequences and health costs of

microcephaly among children born to Zika-infected mothers

should be included in cost-benefit analyses of the census.

Over the past decade, Lyme disease has increased in inci-

dence more than two-fold in the US and Europe. Mapping the

most recent case data from 2009 to 2012, Vandenesch et al.

(2014) identified an overall incidence of 42 per 100,000 popu-

lation.47 However, regionally, using age-specific census data

for the denominators by region, they identified that incidence

was highest in the eastern and central regions of France,

ranging from 184 (95% confidence interval [CI]: 31e356) per

100,000 population in the Limousin region and 157 (95% CI:

34e279) in Alsace, compared with Nord-Pas-de-Calais and

Bourgogne with no cases. The use of this type of information

in geographically targeted prevention programs has not been

considered in cost-benefit analyses of the census.

Population threats, including threats to public health

Any analysis of vulnerable populations requires information

on likely determinants such as incomes, education, ethnicity,

labor market experience, and access to services. In Canada,

this information could be found in the mandatory LFC, but is

no longer reliable from the voluntary NHS.16 The in-

consistencies in how the Canadian census has been carried out

leads to other problems of linking the data for historical trends.

Statistics Canada cautions that such linkages are not feasible

because of the very different methods underlying the NHS

compared to the LFC. While population projections may be able

to cover some of the holes in enumeration, in the short-term,

key information on SES and trends has been lost for Canada.

Public health agencies in the UK have expressed concern

over potential impacts of altering the traditional census.48 For

improving public health, it is essential that the UK retain the

ability to produce analyses for small areas and groups of

people because of the growing health inequalities that occur

both between and within local communities in the UK. In

addition, census data have a number of additional important

direct national uses, including the allocation of the adult so-

cial care budget and setting payment levels for general

Fig. 2 e Number of persons born in Mexico, the Caribbean, Central America, and countries in South America with active-

Zika virus transmission, by US Census tract of residencedNew York City, JanuaryeFebruary 2016. (Reprinted with

permission).46

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 794

practitioners. It provides accurate age and sex population

denominators for premature mortality statistics which are

used to better understand inequalities in the catchment area

of a health service and what the demand for each service is

likely to be.

Crucially, the census is the only source of data in the UK that

identifies all groups within the population at the local scale,

including minority groups, those from overseas countries, and

those residing in pockets of deprivation. Reduced accuracy to

identify these groups may impede delivery of necessary health

services and public health interventions in locations where

they are most needed. Short-term financial savings in census

expenditures could lead to greater public spending on poorly

targeted health care and public health initiatives.

In the US, the Census Bureau plans to eliminate the ACS

three-year estimates beginning in 2015.49 Instead, ACS esti-

mates based on a five-year interval could reduce the ability of

smaller and mid-sized communities to compete for federal

grant funding because they will be less able to demonstrate

impacts in a timely fashionda disadvantage when compared

with larger sized communities which will still be getting

annual data from ACS. This impacts a total of 55 million res-

idents or 17.5% of the US population, and demographers argue

this lack of timely data in more rural locations could result in

heightened ruraleurban income inequality.50

Questions and special residential groups that the ACS

covers are often of particular importance for public health.

State health departments, for example, must know the

number of individuals in nursing homes if they are to

adequately characterize annual influenza mortality risks. The

Housing Assistance Council (HAC) has formally registered its

opposition to the Census Bureau's plan to eliminate the ‘flush toilet’ question, recommending that the plumbing facilities

questions in the ACS be retained in their current form.51 HAC

has argued that the flush toilet, in particular, is an important

indicator of basic housing quality, and is integral to informing

policies and strategies to eliminate substandard hou-

singdparticularly, in the rural US where the number of homes

lacking basic plumbing may be 20-times the national rate. In

2014, the US County Health Rankings added ‘Severe Housing

Problems’ tracking for the first timednoting that 19% of the

US households had at least one or more housing problems

including overcrowding, high housing costs, or lack of kitchen

or plumbing facilities.52 The US Census Bureau's proposal to alter the ACS may result in an under-estimation of housing

problems in the US.

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 7 95

Conclusions

Census data are relied upon by epidemiologists and public

health professionals to establish standardized incidence rates

and prevalence within populations by age, sex, and race/

ethnicity; to understand the role of individual and

community-context influences (e.g., SES and education) in

disease transmission, incidence, stage at diagnosis and prog-

nosis; and to use demographic information in order to assess

whether participants in clinical trials are representative of the

general population. Accurate and timely monitoring and

reporting by the census is critical to ensure that policy makers

and public health practitioners have the evidence needed to:

(1) establish reliable incidence rates in the general population,

as well as in demographically defined sub-groups in order to

rapidly characterize the nature of the public health problem;

and (2) efficiently address disparities in health care and health

outcomes among socio-economically vulnerable, traditionally

disadvantaged, and racial/ethnic minority populations.

Threats to the census are threats to public health surveil-

lance, disaster preparedness, policy-making, and evidence-

based decision-making; they reduce our ability to track and

understand health disparities by SES, education and race/

ethnicity.

In considering the true costs of the census, we call on

policy makers to consider the absence of accurate information

on susceptible populations in relation to emergency pre-

paredness, outbreak response and chronic disease prevention

efforts. Currently, public health benefits of the census are not

directly considered in any cost-benefit analysis.

However, even in the examples we cited, the benefit (e.g.,

number or range of expected cases averted for a given public

health intervention) is not routinely published. Therefore, we

call on public health as a discipline to consider estimating and

publishing a range of likely benefit (either via cost-benefit or

cost-effectiveness analysis) of a given intervention or action.

For example, unrealized benefits to targeted Zika surveillance

could be substantial as lifetime medical costs for one case of

microcephaly is estimated at $10 million US dollars.53 Public

health professionals could also benefit from working more

closely with health economists in coming up with total dollar

values, cases averted, and/or disease-free years of life.

It falls to those of us in public health to identify fair and

reasonable estimates of the benefits of public health action so

that the tools which we rely heavily upon, such as the census,

are not withdrawn without full understanding and assessment

of their benefit. Those in public health must remain vigilant of

policiesthatwillerode the censusand therebyits abilitytoserve

the public interest by providing the evidence base for rational

public health policy formulation and related interventions.

Author statements

Acknowledgments

The authors would like to thank the International Joint Policy

Committee of the Societies of Epidemiology (www.ijpc-se.org)

for their generous discussion and suggestions for the

manuscript.

Ethical approval

None sought. This manuscript reviews publicly-available

information.

Funding

None declared.

Competing interests

None declared.

Contributors

All authors have participated in the article preparation. All

authors have approved the final article.

Authorship

All authors made substantial contributions to the following:

(1) the conception of the manuscript idea; (2) interpretation of

information; and (3) drafting and/or critical revision of the

manuscript for intellectual content.

r e f e r e n c e s

1. Freeman S. National household survey that replaced long-form census is unreliable: experts. Huffington Post; 2013; 6 May.

2. Flanagan O. ONS responds to consultation on statistical output cuts. StatsLife 2014; 13 February. https://www.statslife. org.uk/news/1238-ons-responds-to-consultation-on- statistical-output-cuts [last accessed January 23, 2017].

3. Freeman C. A new history of early christianity. New Haven: Yale Press; 2009.

4. Department of Commerce v. US House of Representatives. 1999. 525 US 316, US Supreme Court.

5. United Nations. Department of Economic and Social Affairs, Statistical Office. Methodology and evaluation of population registers and similar systems, New York: United Nations. 1969. Series F: Number 15.

6. Mather M, Rivers KL, Jacobsen LA. The American community survey. Popul Bull 2005;60. http://www.prb.org/pdf05/60.3the_ american_community.pdf [last accessed 23 January 2017].

7. United States Census Bureau. Measuring America: the decennial censuses from 1790 to 2000. Washington, DC: US Department of Commerce; 2002. Contract No. OL/02/MA, www.census.gov/ prod/2002pubs/pol02-ma.pdf [last accessed 23 January 2017].

8. Griffin DH, Waite PJ. American community survey overview and the role of external evaluations. Popul Res Policy Rev 2006;25:201e23.

9. National Research Council. Committee on National Satistics, Division of Behavioral and Social Sciences and Education. In: Brown LD, Cohen ML, Cork DL, Citro CF, editors. Envisioning the 2020 census: panel on the design of the 2010 census program of evaluations and experiments. Washington, DC: The National Academies Press; 2010.

10. United Kingdom, Office for National Statistics. Census confidentiality protected by law. https://www.ons.gov.uk/census/ 2011census/confidentiality/censusconfidentialityprotectedb ylaw [last accessed 23 January 2017].

11. United States Census Bureau. Data protection and privacy. http://www.census.gov/about/policies/privacy/data_

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 796

protection/our_privacy_principles.html [last accessed 23 January 2017].

12. Huffadine L. Australia Census myths busted: how you can legally avoid filling out the form (but giving a fake name or going camping won't save you from a massive fine). Daily Mail. 2016; 3 August. http://www.dailymail.co.uk/news/article-3721113/Census- 2016-myths-busted-legally-avoid-filling-form. html#ixzz4PznQSEtk [last accessed 23 January 2017].

13. United Kingdom, Office for National Statistics. The census and future provision of population statistics in England and Wales: privacy impact assessment for the initial research stage. 2015; March. https://www.ons.gov.uk/census/censustrans formationprogramme/beyond2011censustrans formationprogramme/privacyandconfidentiality [last accessed 23 January 2017].

14. Government of Canada, Standing Committee on Industry, Science and Technology. Minutes of Proceedings. 2010; 27 July. http://www.parl.gc.ca/HousePublications/Publication.aspx? DocId¼4648017&Language¼E [last accessed 23 January 2016].

15. Statistics Canada. National household survey: final response rates for Canada, provinces and territories. 2011. http://www12. statcan.gc.ca/NHS-ENM/2011/ref/about-apropos/nhs-enm_ r012.cfm?Lang¼E [last accessed 23 January 2016].

16. Sheikh M. Good data and intelligent government. In: Gorbet F, Sharpe A, editors. New directions for intelligent government in Canada: papers in honour of Ian Stewart. Ottawa: Centre for the Study of Living Standards; 2011. http://www.csls.ca/ stewartfestschrift.asp [last accessed 23 January 2017].

17. Berman R. Republicans try to curtail the census. Atlantic monthly 2015; 9 June. http://www.theatlantic.com/politics/ archive/2015/06/republicans-try-to-rein-in-the-census- bureau/395210/ [last accessed 23 January 2017].

18. United Kingdom Office for National Statistics. The census and future provision of population statistics in England and Wales: recommendation from the National Statistician and Chief Executive of the UK Statistics Authority. 2014; 27 March. http:// webarchive.nationalarchives.gov.uk/20160105160709/http:// www.ons.gov.uk/ons/about-ons/who-ons-are/programmes- and-projects/beyond-2011/beyond-2011-report-on-autumn- 2013-consultation-and-recommendations/index.html [last accessed 23 January 2017].

19. United Kingdom, Central Government submission to the ONS' public consultation. The census and future provision of population statistics in England and Wales. 2014; 28 March. https://www. gov.uk/government/publications/response-to-ons- consultation-on-the-census [last accessed 23 January 2017].

20. Soskolne CL. Population health research wins “reprieve” in Europe. Epidemiology 1996;7:451e2.

21. Cope I. The value of census statistics in England and Wales. Newport, South Wales: Office of National Statistics; 2015. https://www.ukdataservice.ac.uk/media/455474/cope.pdf [last accessed 23 January 2017].

22. United Kingdom Office for National Statistics. How others use census data https://www.ons.gov.uk/census/2011census/ 2011censusbenefits/howothersusecensusdata [last accessed 23 January 2017].

23. Russell LB, Sinha A. Strengthening cost-effectiveness analysis for public health policy. Am J Prev Med 2016;50(5 Suppl 1):S6e12. http://dx.doi.org/10.1016/j.amepre.2015.11.007.

24. Teutsch SM, Fielding JE. Applying comparative effectiveness research to public and population health initiatives. Health Aff (Millwood) 2011 Feb;30(2):349e55. http://dx.doi.org/10.1377/ hlthaff.2010.0593.

25. Spencer BD, May J, Kenyon S, Seeskin Z. Cost-benefit analysis for a quinquennial census: the 2016 population census of South Africa. Institute for Policy Research Working paper series WP- 15e06. Evanston, IL: Northwestern University; 2015. http://

www.ipr.northwestern.edu/publications/papers/2015/ipr- wp-15-06.html [last accessed 23 January 2017].

26. Aldridge J, General Register Office for Scotland. 2011 census business case. July 2006. Prepared by John Aldridge, Consultant. (General Register Office for Scotland), https:// www.whatdotheyknow.com/request/8345/response/20302/ attach/3/business%20case.pdf [last accessed 23 January 2017].

27. Bakker C. Valuing the census. Statistics New Zealand: Wellington; 2014. http://www.stats.govt.nz/Census/census- transformation-nz/census-transformation-papers.aspx [last accessed 23 January 2017].

28. Castro D. e-Census unplugged: why Americans should be able to complete the census online. Washington, DC: The Information Technology & Innovation Foundation; 2008:February. www. itif.org/files/eCensusUnplugged.pdf [last accessed 23 January 2017].

29. Seeskin ZH, Spencer BD. Effects of Census Accuracy on apportionment of congress and Allocations of Federal Funds. Institute for Policy Research Working Paper Series. Evanston: Northwestern University; 2015. WP-15e05, www.ipr. northwestern.edu/publications/papers/2015/ipr-wp-15-05. html [accessed 23 January 2017].

30. US Census Bureau. The american community survey supporting statement. OMB Control No. 0607e0810. Washington, DC: Government Accountability Office; 2014. https://www2. census.gov/programs-surveys/acs/operations_admin/2014_ content_review/methods_results_report/2016_ACS_OMB_ Supporting_Statement.pdf [last accessed 23 January 2017].

31. United States Government Accountability Office. Decennial census: additional actions could improve the census bureau's ability to control costs for the 2020 census. Washington, DC. 2012. http:// www.gao.gov/assets/590/587898.pdf [accessed 23 January 2017].

32. Buehler JW, Holtgrave DR. Who gets how much: funding formulas in federal public health programs. J Public Health Manag Pract 2007;13(2):151e5.

33. United States National Institutes of Health. The third national cancer survey: incidence data. Natl Cancer Inst Monogr 1975;i- x:1e454.

34. Redmond CK, Breslin PB, Lerer TJ, Emes J, et al. Cancer incidence in the Pittsburgh metropolitan area, 1969-1971. Department of Biostatistics Technical Report. Pittsburgh: University of Pittsburgh; 1975.

35. Weinberg GB, Kuller LH, Redmond CK. The relationship between the geographic distribution of lung cancer incidence and cigarette smoking in Allegheny County, Pennsylvania. Am J Epidemiol 1982;115:40e58.

36. Zhen-Qiang Ma, Weinberg GB. Geographic distribution of lung cancer incidence in Allegheny County, Pennsylvania, 2000-2010. Harrisburg: Pennsylvania Department of Health; 2013.

37. Zhen-Qiang Ma. Descriptive rates of breast cancer incidence and mortality for Pennsylvania black and white females. Harrisburg: Pennsylvania Department of Health; 2014.

38. Raun L, Ensor KB, Campos LA, Persse D. Factors affecting ambulance utilization for asthma attack treatment; understanding where to target interventions. Public Health 2015;129:501e8.

39. Pennsylvania Department of Health. Tranguch gasoline spill report Hazleton, Pennsylvania. Harrisburg: Pennsylvania Department of Health; 2001. http://www.health.state.pa.us/ pdf/tranguch/tranguchspill.pdf [date last accessed 23 January 2017].

40. Yun SS, Schmaltz CL, Gwanfogbe P, Homan S, Wilson J. Analysis of cancer incidence data in eight ZIP code areas around Coldwater Creek, 1996-2011. St. Louis: Missouri Department of Health and Senior Services (DHSS); 2014. http://health.mo. gov/living/healthcondiseases/chronic/cancerinquiry/pdf/ ccanalysisSept2014.pdf [date last accessed 23 January 2017].

p u b l i c h e a l t h 1 5 1 ( 2 0 1 7 ) 8 7 e9 7 97

41. Vasterling G. St. Louis FUSRAP sites e radiation dose reconstruction request. Letter to Dr. Robin Ikeda Acting Director, Agency for Toxic Substances and Disease Registry. Missouri Department of Health and Human Services; 2014. http:// health.mo.gov/living/healthcondiseases/chronic/ cancerinquiry/pdf/ATSDRSept2014.pdf [date last accessed 23 January 2017].

42. Barker J. Army Corps finds more radioactive contamination on properties near Coldwater Creek. St Louis Post-Dispatch 2015:7 December. http://www.stltoday.com/news/local/ army-corps-finds-more-radioactive-contamination-on- properties-near-coldwater/article_b87e9209-c922-56bb-9230- cad422d01042.html [date last accesses 23 January 2017].

43. Goldhaber MK, Tokuhata GK, Digon E, Caldwell GG, Stein GF, Lutz G, Gur D. The three mile island population registry. Public Health Rep 1983;98(6):603e9.

44. Teo SS, Shingadia DV. Does BCG have a role in tuberculosis control and prevention in the United Kingdom? Arch Dis Child 2006;91(6):529e31.

45. United Kingdom, Public Health England. Tuberculosis (TB) and other mycobacterial diseases: diagnosis, screening, management and data. 2017;18 January. https://www.gov.uk/government/ collections/tuberculosis-and-other-mycobacterial-diseases- diagnosis-screening-management-and-data [date last accessed 24 January 2017].

46. Lee CT, Vora NM, Bajwa W, Boyd L, Harper S, Kass D, Langston A, McGibbon E, Merlino M, Rakeman JL, Raphael M, Slavinski S, Tran A, Wong R, Varma JK. Zika virus surveillance and preparedness e New York city, 2015-2016. Morb Mortal Wkly Rep 2016;65(24):629e35. http://dx.doi.org/10.15585/ mmwr.mm6524e3.

47. Vandenesch A, Turbelin C, Couturier E, Arena C, Jaulhac B, Ferquel E, Choumet V, Saugeon C, Coffinieres E, Blanchon T,

Vaillant V, Hanslik T. Incidence and hospitalisation rates of Lyme borreliosis, France, 2004 to 2012. Euro Surveill 2014;19(34). pii: 20883, http://www.eurosurveillance.org/ images/dynamic/EE/V19N34/Blanchon_fig1.jpg [date last accessed 24 January 2017].

48. United Kingdom Office for National Statistics (ONS). The census and future provision of population statistics in England and Wales: report on the public consultation. 2014; March. http:// www.ons.gov.uk/ons/about-ons/who-ons-are/programmes- and-projects/beyond-2011/beyond-2011-report-on-autumn- 2013-consultation-and-recommendations/report-on-public- consultation.pdf [last accessed 4 March 2016].

49. US Census Bureau. Census bureau statement on american community survey 3-year statistical product. US Census Bur Bull http://content.govdelivery.com/accounts/USCENSUS/ bulletins/eeb4af. [last accessed 04 March 2016].

50. Scardamalia R. Census reduces data for mid-sized places. Daily Yonder 17 March 2015. http://www.dailyyonder.com/ census-data-cut-will-hurt-mid-sized-places/2015/03/17/7770/ [last accessed 5 March 2016].

51. Loza M. Housing Assistance Council Letter to the US Census Bureau Re: The American Community Survey. Changes in 2016 ACS Content Concerning the Flush Toilet Section of the Plumbing Facilities Question: OMB Control Number: 0607- 0810. 80 FR 30655 e Document Number: 2015-13061. www. ruralhome.org [last accessed 4 March 2016].

52. University of Wisconsin Population Health Institute. County health rankings & roadmaps. 2014. http://www. countyhealthrankings.org [last accessed 4 March 2016].

53. Centers for Disease Control and Prevention. In: Transcript for CDC Telebriefing Zika Summit Press Conference; 2016: 1 April. https://www.cdc.gov/media/releases/2016/t0404-zika- summit.html [last accessed 24 January 2017].