Study on Reliability of Health
Statistics in Southeast Asian Countries
Seiji Ohsawa*
Department of child development and environmental studies,
Institute
of Human Living Sciences
1. LOCUS OF THE PROBLEM OF HEALTH STATISTICS IN SOUTHEAST ASIAN
COUNTRIES
In recent years, research concerning statistics on
health and medical
care in developing countries has been conducted in a variety of forms
by a wide range of bodies, including the World Health Organization
(WHO) and other international institutions, universities, and research
institutes. Behind this situation lies a budding realization that such
statistics are indispensable as basic data needed for the formulation
of health care planning, as well as that such statistics and the
related interpretative know-how are effective means of orienting public
administrative policy in such countries in a scientifically sound
fashion. Under the banner of improved primary health care (PHC), WHO
has been promoting a campaign to safeguard the health of, and assure at
least the minimum r~quisite health care for, populations in developing
countries or outlying regions. Such efforts must be under-pinned by
indicators enabling evaluation of the low level of health in such areas
as viewed from a PHC standpoint. Although WHO officials recognize the
importance of health indicators for PHC in the context of health
statistics, the reality is that the accuracy of health statistics tends
to be poorer in areas with a greater need for PHC. As a conse-quence,
it is more difficult to obtain reliable statistics to serve as the
basis for decisions on the level of PHC aid required and of the nature
of the PHC needs. This situation forms the background of the ongoing
search for effective, practical health indicators.
In recent years,
this field as well has seen much application of technique of
statistical (or numerical) analysis, including factor analysis, cluster
analysis, and other types of multi-variate analysis. As is well known,
this type of research approach involves the collection of great
quantities of health (or medical) care indicators and related
indicators over a wide range of countries (or regions). The data are
then subjected to matrix processing on a single-year or time-series
basis, and the findings are used to provide the basis for various
analyses of the current status of health care and for drafting of
administrative policy. In addition, the composite variable data derived
from multi-variate _ analysis are often advanced as effective health
indicators.
Generally speaking, multi-variate tests upon the use
of a
broad (high-quantity) sampling and numerous variables. These features
are the source of serious
problems in the execution of research based on this approach concerning
health and medical care in developing countries. This is because there
is a great difference between such countries and advanced
industrialized countries in regard to the handling of data. In the
first place, a broad sampling is by no means a simple operation in
de'veloping countries. The statisti-cal .system tends tC~ vary from
country to country, and behind each statistical item lies a unique set
of statistical systems.
For example, there is a gap of 16.6 years
between Indonesia and Japan in regard to female average life expectancy
at
birth. Nevertheless, it would be improper to conclude from this
statistic alone that the Japanese live 16.6 years longer than
Indonesians on the average. This is because a legitirnate comparison
cannot be drawn between the two sets of nurneri-cal data, which were
derived through completely different methods. However, extensive
sampling ignores such distinctions by, in effect, performing the same
data processing for Indonesia, Japan, and Bangladesh, for example. This
violates the basic axiom that data handled in a parallel fashion must
be capable of mutual comparison to begin with.
Table
1 and table
21) present
comparisons of statistics for medical facilities and personnel,
respectively, in Japan and various Southeast Asian countries. The
tables exemplify the great gap among the six countries in regard to the
items of statistical survey. Even items that appear to afford a common
basis of comparison, such as "health statistician" do not appear in
statistics, despite the presence of such personnel in the national
Ministry of Health and Welfare. This serves to illustrate the
difficulty of making valid intercountry com pansons.
There are also
serious problems associated with preparation of the numerous variables
needed for multi-variate analysis.
Obviously, omissions are not
desirable in constructing a data matrix. Generally, the quantity of
samples with omissions rises along with the number of countries
subjected to analysis. For this reason, the researcher attempts to
select mainly indices that are less susceptible to omission.
Unfortunately, most such indices concern population or economic
statistics ; very few are deeply related directly to health or medical
care. This kind of variable bias results in a particularly detrimental
omission in the case of multivariate statistical analysis involving the
extraction of common factor from variable relationships. The reason
is'that the components or factors derived are bound to be removed from
the actual health/medical realities, and the equations or expressions
constructed from thern will be nothing more than abstractions.
In other
words, use of multi- variate analysis as a method of intercountry
comparison and evaluation is apt to encounter two types of serious
difficulties : 1) data incapable of legitimate cornparison are
processed uniformly, in parallel, and 2) the data themselves do not
accurately reflect the health/medical realities. As a result, the value
of the findings of such analyses as the basis for effective action is
likely to be low.
The aim in the search for health indicators useful
for PHC is, on the other hand, indices that can be readily applied in
area/district activities and understood
by the PHC personnel on the grass roots level to the point that such
personnel could even prepare such statistics themselves, if necessary.
WHO has asked researchers in the field of
international health
statistics for proposals of such indicators, but research in this
aspect is limited, particularly in Japan, where such researchers are
extremely rare and proposals of health indicators for PHC have been
virtually non=existent.
2. RELIABILITY OF HEALTH STATISTICS IN
DEVELOPlNG COUNTRIES
The reliability of values derrved from measurement
can be expressed as "T=M+E", where "T" rs the true value "M" Is the
measured value and "E" is the error of the measured value. Mimnnzatron
of "E" Is directly Imked to an mcrease m the reliability of the
measured value. "E" stems from a mixture of various factors as far as
health and medical statistics are concerned. The following is a
description of some such factors common in the case of developing
countries.
l) Generally speaking, the statistical system in
developing
countries is a relic of policy of instruction, aid, or colonial
administration imposed by the former suzerains. At the present time,
developing countries are the subjects of vigorous programs of
instruction implemented by various UN-affiliated agencies as well as of
aid extended by advanced industrialized countries. In the countries
behind such programs, censuses and surveys regarding their respective
populations are conducted through statistical bureaus set up within the
apparatus of the central government. By contrast, health statistics in
various Southeast Asian countries are under the jurisdiction of
health-related ministries, as is the case in Japan.
Specialists are
sent from WHO, other assistance organizations, and the governments of
the various advanced industrialized countries to these ministries, to
which they provide advice and aid concerning health statistics. Despite
efforts extending over several dec-ades, however, this approach has not
led to measures for fundamental improvement in the system of health
statistics. Part of the problem is that, by its nature, this approach
has bred a chronic dependence on foreign countries for expertise and
technology as well as for financial and material assistance. In
addition, the very presence of such foreign personnel in the country
has tended to reduce inclinations of self-help. More recently, the
introduc-tion of computers from foreign countries has acted to further
widen the already existing lack of coordination between the hurnan and
systemic organization.
The statistical bureaus do contain specialized
statisticians, and in most cases, these personnel have been promoted
and are salaried in a manner cotnmensurate with their training.
However, personnel specialized in the handling of health statistics are
not to be found within the health ministries on the national level ; in
most cases, these personnel have been trained as doctors. And since
most of thern prefer promotion
in another department or capacity to promotion as a statistician in a
statistical one, they tend to be apathetic about remedy of the
contradictions in the statistical system. This problem is therefore
associated with the machinery of the central government, but it goes
without saying that there are many similar problems related to the
personnel who should be supporting the statistical system on the
provincial level as well.
In governmental administrative agencies,
there is apparently a tendency to use the statistical levels for the
work of the agency unit in question as a reference for the performance
of personnel in it. That is, a worsening of performance as appears in
statistics acts to delay the promotion of the personnel. For this
reason, it is fairly common for provincial personnel to "improve upon"
statistics or to neglect to report "unfavorable" statistics in the
interests of their own promotion. In a certain Southeast Asian country,
such practices have become virtually routine, and official statistics
are almost devoid of reliability. Such statistics on the provincial
level are collected and totalized on the national (central governrnent)
Ievel before submission to UN-related agencies, which consequently are
basically unable to determine where (central, provincial, or district)
the national statistics were juggled, even if they detect impropriety.
Last year, a citizen of Japan died from cholera "imported" from a
developing country. The incident received great play in the Japanese
mass media, which led to a dramatic declir}e in the number of Japanese
tourists visiting the country in question. In response, the government
of the country suppressed the submission of cholera-related statistics
to UN-related agencies and other parties. Such statistical reports have
been resumed since, but the official figures announced by the
government are thought to be underestimates. In such cases, the central
government itself is taking a hand in the statistical distortion.
Abuses of this type are longstanding, and are
associated with the fact
that the preparation of accurate statistics in the final analysis
depends on the conscientiousness of the preparers. As such, they are
not likely to be eliminated without a wider diffusion of statistical
traning, fundamental improvement of the morale of the preparers, and
other measures that cannot be implernented overnight. In other words,
'however well-organized the statistical system and related legal
framework, such abuses will be difficult to correct as long as the
morale of the preparers is low.
2) Another barometer of statistical
reliability is the statistical dispersion. More specifically, the
reliability of health statistics may be regarded as suspect when
figures for a certain statistical item as viewed over a period of time
exhibit jumps in certain years or extremely great variation, provided
that there have been no commensurate outbreaks of contagious diseases,
etc., in the years in question.
3) In any country, population statistics are a vital
item
of concern in the context of the overall statistical administration.
Even in Southeast Asian countries, population statistics tend to be
more accurate that health statistics, and are contained in considerable
detail in national statistical annuals and demographic reports.
Nevertheless, the reliabil-ity of population statistics as well can by
no means be termed high in absolute terms. In the developing countries,
much birthsand deaths, go unreported. In addition, there is a great
movement of population within the countries. Particularly, for example,
notable is the influx of population into districts of Bangkok (the
respective rates of population influx and efflux and over the years
1965-1970 were 12.4 and 9.8 percent for Phra Nakhon and 20.O and 5.0
percent for Thonburi3)). Moreover, the influx is concentrated
in Bangkok's slums, making it more difficult to ascertain the actual
population. The job is further complicated by the minority and
semi-nomadic ethnic groups living on the borders with Burma, Laos,
Cambodia, and Malaysia, and the population living on boats.
Research in
the area of health statistics requires accurate statistics concerning
death and disease in order to ascertain the actual health situation.
Unfortunately, the process of accurately counting cases of death and
disease by cause/type is far more difficult than that of counting
population. The process depends upon the subjective judgement of a
third party (the doctor's diagnosis), and in some cases statistics are
prepared without a proper diagnosis. Such cases are so numerous in the
Philippines that the government has released mortality statistics
classified into categories "medically certified" and "non-medically
certified"'). (In the case in question, deaths in the latter category
accounted for roughly half of the total.)
Even in the case of
"medically certified" deaths, there is a great variation in factors
bearing upon the accuracy of the certification, including the
qualification of the certifying doctor (specialist/non-specialist),
Ievel of.- clinical test expertise, experience, depth of diagnosis
before death, and whether or not an autopsy was performed. The level of
knowledge concerning classification of causes of death is also an
important influence. Moreover; in certain districts, a single doctor is
accorded responsibility for certification of several hundreds of deaths
daily in statistics, as will be related below. In such cases, it is
thought that a considerable portion of the "certification" was in fact
preformed by nurses, folk doctors, or monks.
For such reasons,
statistics concerning death and disease tend to be extremely unstable
and to harbor many errors.
These circumstances notwithstanding, the
health situation of populations in such areas must be gauged by some
method. Various medical indicators have been proposed in this
connection. A particularly favorable rating has been garnered by the
proportional mortality index (PMI ; the ratio of the number of deaths
of those aged 50 or over to the total number of deaths), which is
regarded as both simple to apply and highly effective. (It being
difficult to obtain accurate statistics on causes of death in
developing countries, use of PMI offers a reading of the prevalency of
adult diseases. Generally, the higher the PMI, the longer the average
life span, and by extension, the better the health situation of the
population in question. The major virtue of this indicator is the fact
that it can be easily applied as long as age information is available.)
Summary and conclusion
The following can be cited as the main factors
contributing to error and bias in health statistics in developing
countries.
1) The system for the pfeparation of statistics is
underdeveloped, and
there are serious problems related to the morale and quality of the
corps of personnel handling statrstics.
2) Statistical items and units are not entirely
uniform,
making it difficult to conduct international and historical
(time-series) comparisons.
3) There are many omissions to form official
notifications and registrations.
4) The reliability of those health
statistics related to cause of death and disease type is particularly
low, making them unusable in many cases.
REFERENCES
1) SEAMIC Health Statistics 2002, Southeast
Asia
Medical Information Center, 2003, Tokyo.
2) Statistical Reports of
Changwat (Ubonrachatani), National Statistical Office, Office of the
Prime Minister, Thailand, 1979, 1985.
3) Internal Migration of
Thailand, 33 p., Asia Economics Research Institute, 1981.
4)
Philippines Health Statistics, Ministry of Health, Philippines, 1974.
摘要
筆者は発展途上国特に東南アジア諸国の保健統計を過去20年間にわたって編集してきたが本論では以下の問題点，反省点を明らかにした．
1）統計を扱う要員
の訓練をはじめモラルの向上や彼ら自身への統計教育が不可欠で緊急の課題である．
2）統計標識，統計単位，用語の定義が域内各国で不統一のため比較可能性
が低い．
3）届出，中告漏れ等調査上，届出上の誤差が大きい．
4）WHOの統計概念が末端まで徹底しておらず一国内においてさえも統計標識，概念の翻齢が
うかがえる．年次比較でさえ困難をきたしている．
Table1
Comparative Table on Methods Establishments
Table2
Comparative Table on Medical and Allied Health Personnel
