Docstoc
EXCLUSIVE OFFER FOR DOCSTOC USERS
Try the all-new QuickBooks Online for FREE.  No credit card required.

Appendix Methods for estimating disability burden septicaemia

Document Sample
Appendix Methods for estimating disability burden septicaemia Powered By Docstoc
					     Appendix: Methods for estimating disability burden
     In this section, we provide a brief overview of the methods for calculating disability for each of the major disease
     and injury groups covered in this study. We had adopted methods used in the burden of disease and injury in
     Australia 2003 study (Begg et al. 2007) for most conditions in general, but modified them where necessary,
     based on the availability and nature of detailed local data, and in consultation with local disease experts. While
     this list is extensive, it is not exhaustive, and explicit models were not developed for many conditions. Table A1
     lists the abbreviations of the data sources underlying our models commonly used in this section and their full
     names for ease of reference.

                      Table A1: List of abbreviated and full names of commonly used data sources
      Abbreviated name                     Full name

      ABD2003                              The burden of disease and injury in Australia 2003 (Begg et al. 2007)

      Australian derived weight(s) from    Weights derived from regression model of Dutch disability weights which
      disability weight regression model   requires inputs of health state description based on the six domains of the
                                           EuroQol EQ5D+ (p. 158 of AIHW: Mathers et al. 1999)
      GBD study                            Global burden of disease and risk factors, 2000 (Lopez et al. 2006)

      NHS2004                              2004 National Health Survey, Singapore (MOH 2005a)

      NHSS2001                             2001 National Health Surveillance Survey, Singapore (MOH 2003)

      Singapore birth defect data          2003 National Birth Defects Registry (NBDR, personal communication)

      Singapore ESRD dialysis and
                                           2004 Singapore Renal Registry (SRR, personal communication)
      transplant data

      Singapore hospital data              Singapore hospital discharge database with information on discharge
                                           diagnoses and type of procedures performed. Data for the year 2004 used
                                           unless specified otherwise.
      Singapore mortality data             Cause of Death dataset (RBD). Data for the year 2004 used unless specified
                                           otherwise.

      Singapore notification data          2004 notification data from infectious diseases reporting & investigation
                                           system, which includes electronic notifications and reports; For HIV/AIDS,
                                           data are from the HIV/AIDS National Registry (MOH 2005b)
      Singapore primary care survey        2005 Primary Medical Care Survey, Singapore (MOH 2006)

      Singapore Tanjong Pagar Survey       Causes of blindness, low vision, and questionnaire-assessed poor visual
                                           function in Singaporean Chinese Adults: The Tanjong Pagar Survey (Saw et
                                           al. 2004)

      SMHS2003                             2003 National Mental Health Survey on Anxiety and Depression in Adult
                                           Singaporeans, and on Dementia in Elderly Singaporeans (Lim et al. 2005;
                                           Chiam et al. 2004)

      SOHS2003                             2003 Singapore Adult Oral Health Survey (HPB, personal communication)



1A Infectious and parasitic diseases
Tuberculosis
Incidence estimates for tuberculosis were based on Singapore notification data. We assumed that these were
reasonable approximation of all active cases in Singapore. Average duration for tuberculosis was assumed to be
8 months, reflecting 6 months for the shortest treatment cycle available and another 2 months of symptoms prior
to treatment.

Sexually transmitted diseases (excluding HIV/AIDS)
Incidence estimates for syphilis, chlamydia and gonorrhoea were derived using Singapore notification data. We
assumed that notifications for syphilis and gonorrhoea would represent all incident cases. Syphilis was modelled
using a staged approach applying proportionate distributions for primary, secondary, and tertiary syphilis from
the GBD study. For chlamydia, we adjusted the number of notifications (other than those aged 0-4 years) upwards
to account for under-reporting due to asymptomatic infections and the reluctance of some patients to consult
general practitioners about sexually transmitted diseases. We assumed that our chlamydia notifications would
represent approximately only 30% and 70% of all male and female incident cases respectively. We further
assumed all chlamydia notifications for children aged 0-4 years were for neonatal conjunctivitis
We estimated the incidence of pelvic inflammatory disease (PID), a complication of both chlamydia and
gonorrhoea in women using Singapore hospital data and adjusted the figure upwards following assumption used
in the ABD2003 that one quarter of cases did not present to hospital. We attributed 60% of the PID incident
estimates to chlamydia and another 1% to gonorrhoea. Common sequelae of both chlamydia- and gonorrhoea-
related PID include ectopic pregnancy, chronic pelvic pain, infertility and tubo-ovarian abscess. We based our
rates of complications following PID on GBD assumptions. We adjusted our incident estimates of infertility
resulting from PID for women who did not wish to have a child and therefore did not experience disability, using
findings from a German study (Stobel-Richter et al. 2005).
We used the assumptions from the GBD study to model disability weights and durations for syphilis, chlamydia
and gonorrhoea and their sequelae.

HIV/AIDS
We modelled HIV as a progressive condition with four stages: (1) asymptomatic HIV; (2) symptomatic HIV; (3)
AIDS prior to terminal phase; and (4) terminal AIDS. We used the Singapore notification data assuming that the
annual number of new HIV diagnoses would represent all incident cases of HIV in the year. We used the Dutch
weights for each of the stages but adjusted the weight for stage 1 to account for the proportion of undiagnosed
asymptomatic HIV cases (estimated) to whom we assigned a disability weight of 0 (Aalen et al. 1997). Using
available local information on the progression from HIV infection to development of AIDS and from AIDS to
death, we estimated average durations of 10.5 years for the combined stages 1 and 2 and 3.2 years for stage 3. We
adjusted our duration estimates for stages 1 and 2 based on the assumption that an approximately equal amount
of time was spent in each stage. We assumed that stage 4 lasted an average of 0.5 of a year.

Diarrhoeal diseases and gastroenteritis
We derived the incidence of diarrhoea based on data on acute diarrhoeal illnesses attendances at all polyclinics
in 2004, and adjusted the figures upwards to account for diarrhoeal related attendances at private general
practitioners clinics. The adjustment was made based on the distribution of diarrhoeal related attendances seen
between the private general practitioners (GP) clinics and the polyclinics estimated from the one-day Singapore
primary care survey. We applied duration of 2 days and used age-specific weights for uncomplicated diarrhoea
(average weight of 0.093) from the GBD study. We used Singapore hospital data to estimate the incidence of

                                                                                                                    
     complications, and assumed one week of severe disability and a further week with disability equivalent to that
     of uncomplicated cases.

     Childhood-cluster diseases
     We did not model diphtheria, whooping cough, tetanus and poliomyelitis for 2004. This was because there were
     no notifications for diphtheria, tetanus and poliomyelitis from 1997 to 2004, 1995 to 2004 and 1991 to 2004
     respectively; and only one notification of whooping cough in 2004.

     Measles
     We based the incidence of measles on Singapore notification data, assuming annual notifications would represent
     all cases of measles in Singapore in the year. For acute measles episodes, we applied the GBD duration (2 weeks)
     and disability weight (0.152). We used Singapore hospital data to estimate the incidence of measles sequelae:
     measles encephalitis and sub-acute sclerosing panencephalitis.

     Rubella
     We estimated the incidence of rubella using Singapore notification data. We used the measles disability weight
     with duration of one week since there was no GBD or Dutch weight for rubella. Postnatal acquired rubella infection
     is generally not associated with significant complications. However, congenital acquired rubella infection may
     be associated with severe complications. The classic triad of complications associated with congenital rubella
     infection includes cataract, heart disease and deafness. We did not model for congenital rubella complications
     since there were no notifications for congenital rubella in 2003 and 2004, and only one notification in 2002.

     Haemophilus influenzae type b (Hib)
     There has been a lack of population based data on Hib in Singapore but hospital-based study has supported
     the notion that invasive Hib disease is relatively uncommon in Singapore (Lee et al. 2000). We estimated the
     incidence of Hib in children aged 0-4 years using findings from a Singapore hospital-based retrospective study
     over a period 1990-1995 (Lee et al. 2000). We only modelled the disability associated with the following sequelae
     found in the study, namely, epiglottitis, meningitis and pneumonia manifestations. For epiglottitis, we used the
     GBD weight for meningitis (0.616 for 0-14 years) assuming duration of 1.5 weeks. We applied the Australian
     derived weights from disability weight regression model of 0.913 (for an acute meningitis episode) and 0.373 to
     pneumonia respectively, with duration of 1 month for meningitis and 2 weeks for the latter. We did not estimate
     YLD attributable to Hib among those aged 5 years and above.

     Meningitis
     We estimated the incidence of meningitis based on the number of meningococcal infections notified in
     Singapore. Meningitis was modelled as a progressive condition with acute episodes of 1 month, after effects
     lasting up to 6 months and subsequent lifelong effects, in some, for a range of conditions (including hearing loss,
     ventriculoperitoneal stunt, seizure disorder, less severe developmental problems, mental retardation and motor
     deficit and physical deformities). We followed the assumptions in the ABD2003 with regard to proportion of
     meningitis progressing to sequelae and their associated disability.

     Septicaemia
     We estimated incident cases of septicaemia using Singapore hospital data. We did not adjust our estimates to
     account for meningitis-related septicaemia assuming that the potential of double-counting was insignificant.
     Data from Australia had suggested that the proportion of septicaemia cases due to meningitis was low (<2%).
     We used the Australian derived weight from disability weight regression model for an acute meningitis episode
     (0.913) with an average duration of 1 month.


Dengue
We observed a cyclical pattern in the number of dengue notifications over the past 15 years. We estimated
the incidence of dengue using Singapore notification data averaged between 2000 and 2005 (the most recent
epidemic cycle). We assumed that dengue haemorrhagic fever accounted for 1% of total dengue notifications.
For dengue fever, we used the GBD weight for malaria with duration of 6 days. For dengue haemorrhagic fever,
the Australian derived weight from disability weight regression model for a meningitis acute episode for just
over 1 week was applied.

Hepatitis
Hepatitis A
We estimated the total incidence of hepatitis A using Singapore notification data with an upward adjustment by
a factor of 2 to account for under-reporting following expert advice. We assumed that 10% of incident cases
would represent prolonged hepatitis A and that hospitalisation data on hepatitis A would represent all cases of
complicated hepatitis A. We calculated number of incident cases of uncomplicated hepatitis A by deducting the
prolonged and complicated cases from our total estimate. For uncomplicated hepatitis A, we used the average
GBD weight of 0.093 with duration of 3 weeks (Amin et al. 1999). For complicated hepatitis A, we assumed an
average duration of 4 and 6 weeks for children and adults respectively (Melnick 1995), with half of this time
at severe disability and half at the same weight as used for uncomplicated cases. We assumed that prolonged
hepatitis A cases experienced depression or fatigue for 6 months at a provisional weight equivalent to the Dutch
weight for mild depression (McIntyre 1990; Willner et al. 1998).

Hepatitis B
We estimated the incidence of acute hepatitis B using Singapore notification data. Following expert advice, we
adjusted our estimates upwards by a factor of 2 to account for under-reporting. We derived the incidence of
chronic hepatitis B infection in DisMod 2, based on an estimated overall hepatitis B carrier prevalence rate of
2.7%, a remission of 0.5%, and an overall relative risk of mortality of 1.5. We assumed that the average duration
for an acute symptomatic episode was 4 weeks (Lee et al. 1997), and used the Dutch weight for acute hepatitis
infection (0.21). For chronic hepatitis B infection, we adopted the adjusted Dutch weight of 0.002 applied in the
ABD2003. This adjusted weight was derived using the Dutch weight for chronic hepatitis B infection with active
viral replication (0.36) taking into account expert advice that only 15% of chronic cases had a symptomatic
episode for 2 weeks each year.

Hepatitis C
Due to the asymptomatic nature of hepatitis C infection, we assumed that all YLD were a result of hepatitis C
sequelae, which were cirrhosis of the liver and liver cancer.
The method used to derive estimates of hepatitis-related cirrhosis is described in the section on cirrhosis.

Malaria
We derived the incidence of malaria using Singapore notification data. We modelled episodes of malaria as well
as neurologic sequelae in children aged 0-4 years and used the GBD assumptions for disability weights and
durations.




                                                                                                                    
     1B Acute respiratory infections
     Lower and upper respiratory tract infections
     We based our incidence estimates for episodes of lower respiratory infections (influenza, acute bronchitis and
     pneumonia) and upper respiratory infections (acute nasopharyngitis, acute sinusitis and pharyngitis/tonsillitis)
     on polyclinics and private GP clinics attendance data from the one-day Singapore primary care survey. There
     were apparently inconsistencies in diagnosis coding between the polyclinics and private GP clinics. We therefore
     redistributed the total attendances for lower and upper respiratory infections recorded in the survey (other than
     influenza and acute nasopharyngitis) to the specific respiratory conditions based on the distribution seen at
     the private GP clinics only, since this was deemed more plausible than the distribution at the polyclinics. We
     assumed that about 1% of cases reported under other upper respiratory tract infections were influenza cases and
     the remaining were attributed to acute nasopharyngitis. For pneumonia, further adjustment was done to take into
     account attendances seen at the emergency department of public hospitals in the year 2005.
     For both lower and upper respiratory infections, we used the same assumptions for disability and durations as
     in the ABD2003. GBD duration estimates were halved to 3.5 days for acute bronchitis, and remained at 1 and
     2 weeks respectively for influenza and pneumonia. For acute nasopharyngitis, we applied an average duration
     of 1.5 days, and for tonsillitis and pharyngitis and sinusitis, we used the GBD duration of 3.5 days. Australian
     derived weights using disability weight regression model for these conditions were applied (influenza 0.047;
     acute bronchitis 0.132; pneumonia 0.373; acute nasopharyngitis 0.014; tonsillitis and pharyngitis and sinusitis
     0.061).

     Otitis media
     We modelled the following stages of otitis media; acute infection, bilateral chronic infection, and life-long
     deafness. We estimated the overall incidence rate of acute episodes for each gender using data from the one-day
     Singapore primary care survey but adopted the age-specific incidence pattern observed in the ABD2003 instead,
     arising from comments by local expert that the age distribution from our survey did not appear plausible. We
     assumed that those with relatively low disability did not seek treatment and based our YLD estimates on treated
     numbers. We assumed duration of 1 week and used the Australian derived weight of 0.090 from disability weight
     regression model.
     We estimated the incidence of bilateral chronic infection using the ratio of acute infections to chronic infection
     episodes recorded in the Singapore primary care survey. We assumed an average duration of 1 year and used the
     Dutch weight for early acquired mild to moderate hearing loss of 0.11.
     Permanent deafness as a result of otitis media is very rare. Infections resulting in deafness were based on GBD
     assumptions for established market economies. The GBD assumed that 5 in 100,000 episodes of acute otitis
     media among children aged 0-14 years resulted in lifelong deafness. We derived the duration using DisMod 2,
     and used the Dutch weight for early acquired severe hearing loss (0.233).

     1C Maternal conditions
     We based our incidence estimates for maternal haemorrhage, maternal sepsis, hypertension in pregnancy,
     obstructed labour, abortion and other maternal conditions on Singapore hospital data. We adopted the methods
     applied in the ABD2003, which followed the GBD methods in general except in the following instances. It
     assumed that hypertension in pregnancy resulted in restricted activity for 2 months at a derived weight of 0.117,
     with 1 in 2,500 cases developing neurological sequelae. The sequelae caesarean section was modelled with
     2 weeks of disability at the Australian derived weight of 0.349 from disability weight regression model. For



0
abortions, the disability of infertility resulting from the sequelae pelvic inflammatory disease was modelled.
It was assumed that 20% of hospitalised cases of pelvic inflammatory disease following abortion experienced
infertility from age at infection to post-reproductive age, which we assumed to be 44 years. Our incident estimates
were adjusted for infertility, in the abortion and maternal sepsis models, for women who did not wish to have
a child and who therefore did not experience disability, using findings from a German study (Stobel-Richter et
al. 2005). Unlike in the GBD study, disability of stress incontinence as sequelae of obstructed labour was not
estimated here since most stress incontinence occurred in the absence of such a history.

1D Perinatal conditions
Birth trauma & asphyxia
We estimated the incidence of mild, moderate and severe birth asphyxia using Singapore hospital data. We
separated the mild and moderate incident cases using data from the GBD study, and based our sequelae estimates
of neurological disability on the same study, which assumed that 0% of mild, 25% of moderate and 100% of
severe cases would have neurological disability.
We further assumed that the distribution of incidences of disability into the various sequelae of birth trauma and
asphyxia (i.e. severe hearing loss-congenital or early acquired, seizure, cerebral palsy with intellectual disability,
intellectual disability (mild, moderate, severe and profound)), and the reduction in life expectancy (in years)
associated with each sequelae followed those in the ABD2003.

Low birth weight
We estimated the incidence of low birth weight (>=1500g and <2500g) and very low birth weight (<1500g) in
neonatal survivors using birth registration data from the RBD. We adopted the probability of disability among
low birth weight survivors from the GBD study: 25% for very low birth weight (<1500g) and 5% for low birth
weight (>=1500g and <2500g). This corresponded to a total of 204 incident cases (97 males, 107 females) of
disability in low birth weight survivors in 2004.
We distributed the incident cases of disability in low birth weight survivors to disability type considered in the
GBD study. Distribution of intellectual disability to the varying severity categories was assumed to follow that in
the ABD2003. In addition, we attributed 60% of total incident cerebral palsy cases (assuming at 2.25 per 1,000
live births) to low birth weight and followed the distribution of low birth weight related cerebral palsy with and
without intellectual disability in the ABD2003.
Just over one-half of the low birth weight survivors with permanent disability did not have severe neuron-
development disability. In the absence of a defined disability weight for this health state, we assumed that these
cases had a level of disability similar to the Dutch weight for permanent childhood acquired moderate hearing
loss. For all other sequelae, we applied the relevant Dutch weight. For each sequelae, we assumed the same
reduction in life expectancy (in years) in the ABD2003.

Neonatal infections
We estimated the incidence of neonatal infections from Singapore hospital data. We assumed 1 month of disability
using the Dutch weight of 0.894 (same as meningitis) for acute episodes. Main long-term sequelae are deafness,
motor deficit and intellectual disability. From the ABD2003, it was estimated that there were approximately
20 cases with intellectual disability following neonatal infections (27.8 in males and 12.5 in females for every
1000 infection episodes). We applied these rates and the distribution of intellectual disability severity following
neonatal infections from the ABD2003 to our data. For every case of intellectual disability following neonatal
infection, we assumed approximately 1 case of deafness and 1/3 case of motor deficit.



                                                                                                                         
     1E Nutritional deficiency
     Only iron deficiency anaemia, the most common form of nutritional deficiency in Singapore was included in this
     study.

     Iron deficiency anaemia
     We modelled the following levels of severity for iron deficiency: mild anaemia, moderate anaemia and severe
     anaemia, and used the definition for anaemia (in terms of blood haemoglobin levels) as per the GBD study.
     Estimates of any anaemia in the age group 15-69 years for males and females were from the 1998 National Health
     Survey (using data for the age group 18-69 years) (MOH 2000), and we followed the GBD assumption that 60%
     of anaemia would be due to iron deficiency (Rastogi & Mathers 2003). We applied the distribution by severity in
     the ABD2003 for the age group 15-69 years to our data. We adopted the prevalence estimates for mild, moderate
     and severe iron deficiency anaemia for those aged 0-14 years and 70 years and above from the ABD2003 in
     the absence of local and comparable Asian population data for these ages. Interpolation and smoothing of the
     prevalence estimates across the age groups for both males and females were done within DisMod 2. We derived
     our incidence estimates in DisMod 2 assuming an average duration of 1.0 year, a remission rate of 1.0 and used
     the GBD weight for mild, moderate and severe anaemia.

     2F Malignant neoplasms
     The basis of YLD estimation for malignant neoplasms was a series of models of disease progression developed
     by the Dutch burden of disease study team for 26 cancers for which they determined disability weights (Stouthard
     et al. 1997).
     The general disease model started with an initial phase of diagnosis and primary therapy, with duration of
     up to 12 months. Thereafter, cases were classified as would be cured or would not be cured. Those who were
     considered cured proceeded to a phase of up to 5 years after which they were considered cured with no further
     cancer-related disability (except for cancer with long term sequelae as discussed below). Those who would
     not be cured generally entered a phase (of variable length) of remission followed by a phase of disseminated
     carcinoma (lasting 12 months or less), then a terminal phase (lasting 1 month) and death depending on their
     mean length of survival.
     A Dutch weight was allocated to each of these phases. Where no Dutch weights were available for a specific
     cancer site, we extrapolated weights based on the cancer that it most resembled. The Dutch study did not derive
     a weight for the terminal phase of any of the cancers, so we used instead the Dutch weight for general end-stage
     disease.
     We modified this general disease model to each cancer site in consultation with local clinicians to reflect local
     treatment practices. For some cancers, the ‘5 year survival’ point were extended to 10 years. We also included
     long term sequelae for colorectal cancer, breast cancer, female reproductive cancers, male genitourinary cancers,
     larynx cancer, eye cancer, bone cancer and brain cancer.
     The cancer incidence data were obtained from the Singapore Cancer Registry. The SBoD 2004 Study used data
     from 1991 to 2004 in the calculation of mean survival time and the 5 year or 10 year cure rate. The durations
     of the initial treatment, disseminated and terminal stages for each cancer were based largely on the Dutch study
     assumptions and the ABD2003.

     2G Other neoplasms
     Benign neoplasms are not notifiable in Singapore. We considered three categories in estimating the YLD –
     uterine myoma, benign brain tumors and other benign neoplasms. For uterine myoma and benign brain tumors,
     the incidence estimates were based on Singapore hospital data.

Specifically, for uterine myoma, we used the numbers of myomectomies and hysterectomies for fibroids. We
assumed that surgical treatment was undertaken for all cases of rapidly growing or large tumours and myoma-
related symptoms. We assumed a six month pre-operative state equivalent to the GBD weight for chronic
pelvic pain and an additional three-week postoperative state equivalent to laparotomy. We also assumed that
reproductive disability occurs in 3% of hysterectomy cases to whom we applied the GBD weight for infertility.
Our model for benign brain tumour was based on the model for malignant brain tumours where we modelled the
disease in stages for survivors (diagnosis and initial treatment, and post-curative treatment) and non-survivors
(diagnosis and initial treatment, pre-terminal and terminal). Readmissions were excluded. We based our survival
estimates on mortality data and assumed that successfully treated cases recovered normal efficiency (Steiner et
al. 1998) with a period of ‘worry’ after treatment of 2 years. In the absence of specific disability weights, we used
those for malignant brain tumours.
For the ‘other’ residual category, we applied the YLD:YLL ratio for leukaemia to observed mortality because
most deaths in this category were from causes (such as polycythemia vera, idiopathic thrombocythemia and
chronic lymphoproliferative disease) with a disease progression assumed to be similar to that of leukaemia.

2H Diabetes
Diabetes cases
We modelled insulin dependent diabetes mellitus (Type 1) and non-insulin dependent diabetes mellitus (Type 2)
together in a single model. Prevalence of diabetes (both Type 1 and Type 2 combined) among persons aged 15
years and above, by sex and age, were estimated from the NHS2004 (covering resident population aged 18-74
years). Prevalence of diabetes in elderly persons aged 75 years and above was assumed to follow those in the
age group 70-74 years. We derived the prevalence of Type 2 diabetes in children (0-14 years) using information
on ratio of Type 2 to Type 1 diabetes from the International Diabetes Federation-Western Pacific Region (IDF-
WPR) Childhood and Adolescent Diabcare 2001 study, and estimated prevalence of Type 1 diabetes in children
in Singapore by the IDF. IDF’s estimation of prevalent Type 1 diabetes cases in children in Singapore was based
on a local 1992-1994 study (Lee et al. 1998). We used the relative risk of all-cause mortality for diabetes obtained
from the Asia Pacific Cohort Studies Collaboration (Woodward et al. 2003), and no remission in DisMod 2.
We subtracted out those with diabetic nephropathy to avoid double-counting as the Dutch weight for diabetic
nephropathy includes the disability associated with diabetes per se. We used the Dutch weight for an uncomplicated
diabetes case (0.070).
We calculated YLD for complications from diabetes and they include vision loss due to diabetic retinopathy,
cataract and glaucoma; renal failure, neuropathy, peripheral vascular disease, diabetic foot, amputations,
ischaemic heart disease and stroke.

Retinopathy
We derived the incidence and duration of vision loss due to diabetic retinopathy in DisMod 2 using prevalence
estimates of vision loss (bilateral low vision and blindness) from the Singapore Tanjong Pagar Survey, assuming
no remission and twice the excess risk of mortality as for all diabetes. GBD weight for low vision treated (0.227)
was applied.

Cataract and glaucoma
We estimated the proportion of YLD from cataract and glaucoma attributable diabetes using population
attributable fractions. Relative risk estimates were from the Blue Mountain Eye Study (Mitchell et al. 1997) and
severity distributions from the Melbourne Visual Impairment Project (Weih et al. 2000).


                                                                                                                        
     Renal failure
     Incidence of diabetes-related renal failure estimated using Singapore ESRD dialysis and transplant data. Only
     cases where the primary cause of end-stage renal disease was diabetes, were included in the incidence estimates.
     We used DisMod 2 to estimate the average duration for people on dialysis, assuming a case-fatality rate reflecting
     observed deaths from the register, and based our annual remission estimates on observed transplant data. We
     applied the Dutch weight for diabetic nephropathy (0.29). We estimated YLD for transplant patients assuming
     a case-fatality ratio reflecting observed deaths from the register and ‘remission’ due to graft failure (as these
     patients returned to the pool of dialysis cases). We used a disability weight of 0.29 for the first 6 months following
     the transplant and a GBD weight of 0.11 thereafter.

     Neuropathy
     We followed the ABD2003 which used data on diabetic neuropathy prevalence since diagnosis from Tapp and
     colleagues (Tapp et al. 2003), and estimated an annual increment in prevalence by linear regression which was
     then applied to survivors of incident cases of diabetes by age as they progressed to other age groups. On expert
     advice, we used the Dutch weight of 0.074 for Type 2 diabetic mellitus.

     Peripheral vascular disease
     Similar to the approach for diabetic neuropathy, we estimated by linear regression an annual increment in the
     prevalence of diabetes-related peripheral vascular disease to derive incidence. Estimates of peripheral vascular
     disease incidence and prevalence were also from Tapp and colleagues (Tapp et al. 2003) and it was assumed that
     only those with claudication were symptomatic. Australian derived weights from disability weight regression
     mode of 0.246 for males and 0.286 for females were used. Remission from surgery by vascular grafts was
     assumed to be 20%.

     Amputation and diabetic foot
     The incidence estimates for diabetes-related toe and leg or foot amputation were derived using Singapore hospital
     data. GBD weights for these conditions were applied. We derived the duration from DisMod 2 assuming no
     remission and twice the excess risk of mortality as for all diabetes.
     For diabetic ‘foot’, we followed the assumption that for every amputee, there would be 20 cases of diabetic
     ‘foot’. This was based on an amputation rate of 5.3% for diabetics with foot ulcers from 1994-2005, from the
     Diabetes Research Foundation. We applied the GBD weights to episodes of diabetic ‘foot’ (of which 80% were
     assumed to be treated), with an average duration of 8.9 months.

     Ischaemic heart disease and stroke
     We estimated the proportion of ischaemic heart disease YLD and stroke YLD attributable to diabetes using a
     population attributable fraction based on prevalence and the relative risk of dying from ischaemic heart disease
     and stroke amongst diabetics from the Asia Pacific Cohort Studies Collaboration (Woodward et al. 2003). IHD
     and stroke sequelae were not included in diabetes as of our main burden of disease categories.

     2I Endocrine disorders
     We did not derive specific morbidity models for endocrine disorders. We derived the YLD estimates for endocrine
     disorders by applying the YLD/YLL ratio estimates by sex and age group estimated by the WHO for the year
     2002 for Singapore (WHO 2004b) for this disease category to our corresponding YLL estimates for the year
     2004.




2J Mental disorders
Very few epidemiological studies have been carried out in Singapore over the past 25 years to estimate the
prevalence of mental health disorders in the general population. The landmark population-wide SMHS2003
conducted between February 2003 and March 2004 in adults Singaporeans was set up to determine the prevalence
of depressive and anxiety disorders in those aged 20-59 years; and prevalence of depressive disorders and
dementia in elderly Singaporeans aged 60 years and above . It was our only population-based data source for
estimates of anxiety and depressive disorders’, and dementia’s prevalence. For other mental conditions such as
alcohol use disorders, schizophrenia, bipolar disorder, we derived incidence estimates extrapolated from other
studies or from available hospitalisation and clinics records.
For alcohol use disorder, anxiety & depression and bipolar disorder, we adopted the comorbidity-adjusted
disability weights for these conditions in the ABD2003, derived using data from the Australian National Survey
of Mental Health and Well-being (MHS) 1997 (ABS 1997). Major mental disorders included in the Australian
MHS 1997 were: anxiety disorders (panic disorder, agoraphobia, social phobia, generalized anxiety disorder,
post-traumatic stress disorder), affective disorders (depression, dysthymia, mania, hypomania, bipolar disorder)
and substance use disorders (alcohol and drugs). Details on the derivation of the comorbidity-adjusted disability
weights can be found in the ABD2003. In short, a unit change in disability weight for each unit change in
mental component score (MCS) of the SF-12 was estimated and applied to all disorders. A disability weight
for each respondent in the survey was then calculated, and adjusted for the number of comorbid mental health
diagnoses.

Alcohol use disorder
Population level prevalence or incidence information on alcohol use disorder is absent in Singapore. From the
WHO Global Status Report on Alcohol, the total recorded alcohol per capita consumption of pure alcohol (in
those aged 15 years and above) for Singapore for the year 2000 was 2.73 litres per capita. This was about one-
third the amount reported for Australia at 9.19 litres per capita. The estimated 12-month prevalence rate of any
alcohol use disorder among males and females in Australia using data from its MHS 1997 were 9.4% and 3.6%
respectively. Due to the absence of better data, we assumed that the 12-month prevalence rate of alcohol use
disorder in Singapore men followed that of the Australia men, adjusted by a factor of 2.73/9.19 (ratio of total
alcohol per capita consumption of pure alcohol for Singapore to that for Australia). For the 12-month prevalence
rate of alcohol use disorder in Singaporean women, we applied the ratio of binge drinking prevalence in females
to that in males (age-specific) derived using the NHS2004 to the set of prevalence estimates for Singapore
men.
Instantaneous remission rate of 23.7% calculated from a two-year follow-up study (Booth et al. 2001), and
an elevated mortality risk of 1.8 in males and 3.84 in females (Harris and Barraclough 1998) were used in
the DisMod 2 model. We used the proportion of one-year prevalent cases reporting symptoms of alcohol use
disorder in the previous two weeks from the Australia MHS 1997, as an approximation of the proportion of time
symptomatic.

Schizophrenia
Information on schizophrenia’s prevalence in the population is similarly not available in Singapore. Prevalence
estimates were derived linking Singapore hospital data of all inpatient discharges with a recorded diagnosis
of schizophrenia between 1991 and 2004, and Singapore mortality data for the period 1991 to 2003. Cases in
the discharge dataset found in the mortality database were removed, and any readmissions found within the
Singapore hospital data were subsequently excluded to estimate the pool of schizophrenia survivors in 2004 by
age and sex. In an attempt to improve our prevalence estimates (since not all patients would be admitted), we
linked the prevalent ever-hospitalised schizophrenics’ information to specialist outpatient clinics’ information

                                                                                                                    
     from the Institute of Mental Health/Woodbridge hospital and its 2 satellite clinics over one-year period Oct-04
     to Sep-05 (data for the period Jan-04 to Sep-04 were unavailable) to add on patients missing from the discharge
     dataset but present themselves at outpatient settings for schizophrenia treatment. We inflated the number of
     outpatients added to the ever-hospitalised prevalent pool of schizophrenics by 30% to account for outpatient
     records with missing diagnosis. Kohn et al (2004) reviewed community-based psychiatry epidemiology studies
     that used standardised diagnostic instruments and included data on the proportion of individuals receiving care
     for schizophrenia, and estimated that the median treatment gap for schizophrenia (including other non-affective
     psychosis) in the Western Pacific region to be 35.9%. We further bumped up our prevalence estimates by this
     same proportion. The estimated prevalence rate of schizophrenia in Singapore resident population aged 15 years
     and above in 2004 was 1.0% (1.0% for males, 0.9% for females).
     Annual remission was based on a number of longer term studies and was set at the median of the reported rates
     (1.5%) (Ciompi 1980, Huber et al. 1980, Harding et al. 1987, Helgason 1990, Harrison et al. 2001). We assumed
     that the mortality experience of people with schizophrenia in Singapore was similar to those in Australia and
     used the derived age-pattern relative risks in the ABD2003. The average time spent in psychosis was assumed to
     be 30% (Leff et al 1992). We used a composite weight (0.434) based on 30% of the GBD weight for psychoses
     corresponding to the estimated time spent in this state and 70% of the treated weight.

     Anxiety & depression
     We followed the approach in the ABD2003 where estimates were made for a combined state of anxiety and
     depression. Anxiety disorders that were considered in our study i.e. generalized anxiety disorders, panic,
     agoraphobia, social phobia, and unipolar depressive disorders (major depression and dysthymia) were grouped
     and modelled as a single disease category in our study. Obsessive-compulsive disorder, post-traumatic stress
     disorder and separation anxiety disorder included in the ABD2003 were not considered due to the lack of local
     data. In any case, these conditions were deemed to be uncommon in our Singapore population. The advantage of
     the combined approach is that it takes away some of the difficulties of dealing the frequent comorbidity between
     all these disorders. Because of the high degree of comorbidity and the similarity in treatment approaches, some
     mental health researchers considered all these disorders as part of the same entity with a continuum between
     mostly depressed to mostly anxious (e.g. Andrews 1990 and Andrews 2002).
     One-month prevalence of any of the included conditions by age and sex were obtained from the SMHS2003
     - adults aged 20-59 years. Anxiety disorders were not covered in the SMHS2003 for the elderly aged 60 years
     and above. We therefore estimated a correction factor, using one-month prevalence information for anxiety
     with depression and depression among those in the age group 20-59 years and applied it to the one-month
     depression prevalence estimate in the elderly to account for anxiety disorders. These were then extrapolated to
     one-year prevalence estimates using information on proportion of one-year prevalent cases reporting symptoms
     in the previous one-month, from the Australia MHS 1997. In DisMod 2, we used a remission of 0.043, a pooled
     estimate from follow-up studies for people with various anxiety disorders (Steketee et al. 1999; Wewetzer et al.
     2001; Yonkers et al. 2003) and an increased relative risk of 1.5, a value in between the range of meta-analysis
     estimates reported for anxiety and depressive disorders (Harris and Barraclough, 1998). We used the proportion
     of one-year prevalent cases reporting symptoms in the previous two weeks from the Australia MHS 1997, as an
     approximation of the proportion of time symptomatic and thus assumed that all these conditions had a chronic
     nature with periods of remission in between.

     Bipolar disorder
     The approach to estimate bipolar disorders’ prevalence followed exactly that for schizophrenia where a
     combination of Singapore hospital data, Singapore mortality data and specialist outpatient clinics records was
     used. Median treatment gap for bipolar disorders estimated for Western Pacific region was 50.2% (Kohn et al.
     2004). We hence adjusted our estimates up by a factor of 1/0.498. Based on this method, the estimated prevalence


rate of bipolar disorders in our resident population aged 15 years and above was 0.3% in 2004 (0.2% for males,
0.3% for females). Our final estimates showed an increase in prevalence with age which does not seem plausible
according to the disease expert we consulted. In DisMod 2, due to the absence of better data, we assumed that
the age-specific prevalence pattern of bipolar disorders for each gender in Singapore was similar to that observed
in the ABD2003, constrained to the estimated overall prevalence rate of bipolar disorders for Singapore males
and females.
Bipolar disorder was modelled as a life-long condition with an elevated risk of dying based on the standardised
mortality ratio for all-cause mortality reported in the meta-analysis of Harris and Barraclough (1.96 in males
and 1.76 in females). Bipolar disorder is associated with high suicide risk (Bijl 2002), more than in people with
depression or anxiety. In DisMod 2, we applied an overall relative risk of mortality of this magnitude and used
a remission rate of 0.035 calculated from a follow-up study (Angst & Preisig, 1995).

Eating disorders (anorexia and bulimia)
Estimates for anorexia were based on a review by Hoek et al. (2003) for an average prevalence rate of 0.3%
found among young females, and on remission rate of 0.11 calculated from a follow-up study (Strober et al.
1997). DisMod 2 was used to derive incidence and duration estimates for females from these figures (prevalence
rate of 0.3% was applied to females aged 14-29 years), assuming an increased annual risk of mortality of 0.58%
(Sullivan 1995). We assumed that the incidence in males was 10% of the rate in females.
For bulimia, estimates were based on a prevalence rate of 0.7% among Swiss females aged 14-17 years
(Steinhausen et al. 1994). This is the mid point in the range of prevalence between 0.5% and 1.0% reported
from more rigorous epidemiological studies (Gilchrist et al. 1998). We used a remission rate of 0.21 from figures
reported in a review of follow-up studies (Keel et al. 1999). Incidence and duration estimates for females were
derived in DisMod 2 using these figures (prevalence rate of 0.7% was applied to females aged 14-29 years) and
we assumed no increased risk of mortality. We used the Dutch weight of 0.28 for both types of eating disorder.

Attention-deficit hyperactive disorder
We adopted the prevalence rate of attention-deficit hyperactive disorder in Australia, estimated from its Child
and Adolescent Component of the 1997 National Survey of Mental Health and Well-being (Sawyer et al. 2000).
The survey defined attention-deficit hyperactive disorder to include children with a diagnosis on the survey
and whose parents report the child having more emotional or behavioural problems than other children of the
same age. Our estimates of burden of attention-deficit hyperactive disorder were derived from DisMod 2 using
prevalence rates of 6% and 3% in male and female children (<12 years) respectively, 2% and 1% in male and
female adolescents (>12 years) respectively; remission rate of 0.15% in children aged less than 10 years, 0.25%
in those aged 10-19 years and 0.3 thereafter to model the change in prevalence between children and adolescents;
and no increased risk of mortality. A composite weight (0.082) was used based on the Dutch weights for both
mild and moderate-to-severe attention-deficit hyperactive disorder (0.02 and 0.15), and weighted by the severity
distribution found in the 1997 Australian’s survey.

Autism spectrum disorders
Incidence of autism spectrum disorders (inclusive of Asperger’s syndrome) was based on the average number of
new cases (425) of autistic spectrum disorders among preschoolers diagnosed at the Child Development Units
(CDUs) of Singapore Kandang Kerbau Women’s and Children’s Hospital and Singapore National University
Hospital per year between the period 2004-2006. This works out to be an average of 15.8 per 10,000 children
aged 1-6 years. Distribution of new cases of autism spectrum disorders by sex, age and specific conditions
however, was not readily available. We approximated the incidence rate ratio of Asperger’s disorder & pervasive
developmental disorder not otherwise specified to autism by 1.3:1 (the average incidence rate ratio between 3
studies: Kartina Williams et al. 2005, Powell 2000, Lauritsen 2004) and the sex ratio by 4.5:1. We assumed no

                                                                                                                     
     remission and an elevated risk of mortality as reported by Shavelle (2001) in DisMod 2. We used the Dutch
     disability weight of 0.55 for autism, and for Asperger’s disorder & other pervasive developmental disorders, an
     estimated weight of 0.25, somewhere between Dutch weights for moderate/severe attention-deficit hyperactive
     disorder and for autism.

     2K Neurological and sense disorders
     Alzheimer’s and other dementias
     We based our estimates of dementia prevalence on a meta-analysis of population-based studies by Wancata et al.
     (2003). These estimates were within the ranges estimated for the Western Pacific and Southeast Asian regions in
     a more recent Delphi consensus study by Ferri et al (2005). We did not use prevalence data from the SMHS 2003
     as there were large differences by sex especially among the older age groups. Some studies showed a slightly
     higher prevalence rate of Alzheimer’s disease in women. However, many studies showed no sex differentials in
     the overall prevalence of dementia.
     We used relative risks of mortality of 2.38 up to age 89 years (Jagger el al. 2000) then declining to 1.8 and
     1.6 in females and males aged 90 years above respectively (Henderson et al. 1998). We derived incidence
     and duration using DisMod 2, based on aforementioned prevalence and relative risks estimates, assuming no
     remission. We used a composite weight of 0.479, which combined the Dutch weights with a severity distribution
     from a European population-based cohort study.

     Epilepsy
     We estimated incidence for primary epilepsy using the 1980-84 Rochester Epidemiology Project medical record
     linkage system (Zarrelli et al. 1999). We used these incidence estimates, assuming no gender differences, with
     age-specific remissions (Annegers et al. 1979) and an overall standardised mortality ratio of 1.3 (Tomson 2000)
     to derive estimates of incidence and duration in DisMod 2. We applied the Dutch weight for epilepsy (0.110).

     Parkinson’s disease
     We based our estimates of Parkinson’s disease prevalence on a community-based prevalence survey of neurologic
     diseases (which includes stroke, Parkinson’s disease, epilepsy and dementia) conducted from 2001-2003 in
     adults aged 50 years and above in Singapore (Tan et al. 2004).
     Our estimates for the relative risk of mortality were based on adjusted age-specific risk of mortality from the
     ABD2003 using data from the meta-analysis of prevalent cases of Parkinson’s disease and survival undertaken
     by the Neurologic Diseases in the Elderly Research Group (Berger et al. 2000), and smoothing for irregular
     pattern by age. We used these in DisMod 2 together with the prevalence estimates, assuming no remission to
     derive incidence and duration.
     Disability weights were derived based on an Australian patient-based cohort study (Hely et al. 1999) reporting
     on the distribution of Hoehn and Yahr stages (which corresponds with the descriptions of the severity states
     of Parkinson’s disease for which Dutch disability weights are available) and survival at each 2-year interval.
     Parkinson’s disease was modelled assuming that all cases started with mild symptoms and progressed over
     time to moderate and then severe symptoms over time. From simple linear regression lines, an annual increase
     in those with moderate and severe symptoms was derived. The proportion of cases over time in the moderate
     category would be the balance between those moving from mild to moderate and those exiting moderate by
     shifting to the severe category. For each age group, the average disability weight during the estimated average
     duration was calculated. As severity progresses with time since incidence and younger age groups have longer
     durations, disability weights are highest in the younger age groups.



Migraine
Estimates for migraine were based on a community-based study of headache diagnosis and prevalence for ages
12 years or above in Singapore (Ho et al. 2001). The study reported prevalence estimates for migraine without
aura only. Lyngberg (2005) reported that prevalence and incidence studies on the two migraine sub-types showed
30-40% of people with migraine reporting migraine with aura. We assumed that migraine with aura accounted
for 30% of migraineurs in Singapore and there was no coexistence of migraine with aura, with migraine without
aura. We therefore adjusted upwards the estimates for migraine without aura from the Singapore study using this
factor to obtain estimated migraine prevalence for Singapore.
We estimated the incidence and duration of migraine in DisMod 2 using prevalence, incidence estimates from
international data (Stewart et al. 1991) and a case fatality rate of zero. Within DisMod 2, we used manual
smoothing to extrapolate incidence to older ages, and to smooth prevalence, especially in older age groups
where variations and zero prevalence were assumed to be a result of sampling. We assumed that 20% of cases
received treatments in developed countries and that the average duration for untreated and treated episode was
24 hours and 6 hours respectively. We used the average weights estimated for treated and untreated models using
frequency, severity, and disability weight data from the Global burden of migraine in the year 2000 (Leonardi
& Mathers 2003).

Vision disorders
We modelled bilateral low vision and blindness due to the following causes: glaucoma, cataract, age-related
macular degeneration and the category ‘other causes of low vision and blindness’. Our prevalence estimates
were based on results for causes of low vision and blindness among adult Chinese Singaporeans aged 40-79
years from the Singapore Tanjong Pagar Survey. We extrapolated the prevalence rate of bilateral low vision and
blindness for each cause in those aged 80 years and above using results from the Melbourne Visual Impairment
Project (Weih et al., 2000). For bilateral low vision and blindness (combined) due to glaucoma, age-related
macular degeneration and ‘other causes’ category, we derived incident cases using DisMod 2. We assumed
a relative mortality risk of 1 for all three causes; a remission of 1 for both glaucoma and age-related macular
degeneration related bilateral low vision and blindness, and a remission of 0.25 for the ‘other causes’ category.
We used GBD weights for low-vision treated and blindness to derive a composite disability weight from the
severity pattern across all ages.
Treatment for cataract is readily available in Singapore. Prevalent cases of cataract with bilateral low vision
were assumed to have waited on average of 2 years for cataract surgery. For cataract with bilateral blindness, the
duration was assumed to be 1 year for surgery.

Adult-onset hearing loss
We followed the approach used in the Global burden of hearing loss in the year 2000 (Mathers et al. 2003)
which estimated YLD for adult-onset hearing loss with moderate (41-60 dBHL in the better ear) and severe
(>60 dBHL in the better ear) stages only. Slight impairment (26-40 dBHL in the better ear) was assumed to
have negligible disability weight and therefore would not contribute to the YLD for adult-onset hearing loss.
We derived prevalence estimates for moderate and severe hearing impairment (>41 dBHL in the better) among
adults in our population based on self-reported data to the question: ‘Have you suffered from hearing loss?’ in
the NHSS2001. To use these population data, we further assumed that prevalence of congenital hearing loss or
loss as sequelae of specific diseases and injuries was small as compared to the component arising from noise/
age. Local ENT experts consulted commented that the prevalence estimates in the elderly (60 years and above)
derived from the survey were under-reported (8.8%). We therefore adjusted upwards the prevalence rate of
hearing loss in those aged 60 years and above to 15.6% based on findings from the Blue Mountain Hearing Study
(Sindhusake et al. 2001), which was thus far the best available comparable study. We split our age-sex specific

                                                                                                                     
     prevalence estimates for Singapore by severity (moderate vs. severe) using the ratio of overall age-standardised
     prevalence rate for severe-moderate hearing loss to severe hearing loss for Australia estimated in the Global
     burden of hearing loss for the year 2000.
     We modelled severe hearing loss in DisMod 2, using prevalence, no remission and a relative risk of 1. We similarly
     assumed no remission, and used incidence of severe hearing loss from the DisMod 2 output as ‘mortality’ in the
     moderate hearing loss DisMod 2 model; this took the cases of severe hearing loss out of the pool of susceptible
     cases for moderate hearing loss and hence gave more accurate average durations than if remission were used as
     remitted cases in the DisMod 2 model, as the cases continued to be subject to the hazard of incidence.
     Assuming that all cases progressed from mildest to most severe category, from the cross-sectional data on
     prevalence, it was not possible to estimate these progression times exactly. However to be consistent with other
     disease models where subsequent severity levels for same health state was discounted back to first incidence, we
     applied a 25-year lag for severe hearing loss and 15 years for the moderate category.
     Local ENT expert opined that the rate of hearing aid usage in Singapore among those with hearing impairments
     would be lower than that in the US (≈10%, Gates GA et al.1990), UK (≈14%) and perhaps even Hong Kong due
     to costs and social stigma. We assumed that the prevalence of hearing aid use among those with hearing loss in
     Singapore was one half of that estimated for the region EURO A by age and sex in the Global burden of hearing
     loss for the year 2000. We used disability weights from the GBD study adjusted for estimated prevalence of
     hearing aid use in Singapore.

     2L Cardiovascular diseases
     Ischaemic heart disease
     We considered three health states for ischaemic heart disease: angina pectoris, acute myocardial infarction
     and heart failure, and modelled them separately. Incidence of angina pectoris was modelled as the number of
     admissions for angina to hospital in the year 2004 without any mention of angina in any previous admission in 14
     years of linked Singapore hospital data. We modelled angina pectoris pre- and post-myocardial infarct together.
     Duration was estimated in DisMod 2 using remission estimated from the number of revascularisation procedures
     from Singapore hospital data, and case-fatality rates calculated over the period 1999-2004 in ‘prevalent cases’
     of angina pectoris (defined as anyone with an admission for angina pectoris since 1990 and still alive over the
     follow-up period). In DisMod 2, we included a trend of -2% per year in both incidence and case-fatality based
     on trends in ischaemic heart disease mortality over the past 10 years. We used a composite weight of 0.105 for
     angina pectoris assuming 95% of angina was experienced at the mild-moderate level with the corresponding
     Dutch weight of 0.08, and the remaining 5% with a weight of 0.57.
     For acute myocardial infarction, we based our incidence estimates on the number of new and recurrent cases
     (excluding persons who died within 28 days of onset) of acute myocardial infarction from the Singapore Acute
     Myocardial Infarction Registry for the year 2002. We assumed an average duration of 3 months and used the
     GBD treated disability weight of 0.395.

     Heart diseases resulting in heart failure
     Population-level prevalence or incidence information on heart failure is absent and scarce elsewhere. Our model
     for heart failure was based on hospitalised heart failures. We identified people presenting to hospital with heart
     failure (either as a primary diagnosis or as an associated condition) at any time between 1990 and 2004. To derive
     case-fatality, we linked these data to Singapore mortality data in the period 1999 to 2004 and calculated the
     number of years lived between 1999 and 2004 by anyone who had ever been admitted with a diagnosis of heart
     failure since 1990. The case-fatality rate was then taken as the number of deaths over person-years of follow-up
     in 5-year age group after subtracting out the background mortality. The complete descriptive epidemiology in

0
this group was then derived in DisMod 2 from incidence, case-fatality and assuming no remission. A declining
trend in case-fatality of 3% per year, and a 2% decline in incidence per year for both males and females over the
last 10 years were included in the model.
We then identified the underlying causes for all heart failure cases – rheumatic heart & fever disease, ischaemic
heart disease, inflammatory heart disease, hypertensive heart disease pulmonary heart disease or non-rheumatic
valvular heart disease – from the linked hospitalisation database between 1995 and 2004, if any of these were
mentioned as a cause in the 10 years of hospital admission data. For cases with multiple causes found, the
underlying cause was decided based on the following the priority order: rheumatic heart & fever disease followed
by non-rheumatic valvular heart disease, pulmonary heart disease, ischaemic heart disease, inflammatory heart
disease, hypertensive heart disease, ill-defined and others. A significant number of the cases appeared to have
no or an ill-defined underlying cause. We then redistributed these cases by age and sex so that the proportions
for all underlying causes of heart failure added up to 100%. We used the duration, together with the incidence
and prevalence estimates obtained from the heart failure DisMod 2 model described above, and the derived
proportion of heart failure cases for each of the above six underlying causes, to calculate the YLD for each of
these conditions (including ischaemic heart disease).

Stroke
We modelled stroke in terms of the following health states: a short period of disability for those who died in the
first 28 days, survival beyond 28 days with no permanent impairment at one year after onset, and survival beyond
28 days with permanent impairment. We used data on first-ever strokes admissions to public sector hospitals in
the year 2004. To get an approximation of total first-ever stroke incidence, we adjusted these figures upwards
to account for admitted strokes to the private hospitals (an estimated 4.3% of all first-ever admitted strokes
based on linked hospitalisation records). We further assumed that about 5% of all first-ever strokes in Singapore
were not admitted to hospitals. Next, from the estimated total first-ever strokes in Singapore, we subtracted a
proportion of cases that died, using information on 28-day case fatality rate in first-ever strokes at public sector
hospitals by stroke subtype (we assumed no significant difference in the 28-daycase fatality rate between total
first-ever strokes and those admitted to public sector hospitals).
For those surviving beyond 28 days, the case-fatality rate of stroke were derived linking Singapore hospital data
and Singapore mortality data using a similar approach described above for heart failure. The case-fatality rate for
DisMod 2 was the excess mortality in prevalent cases, which we defined as anyone still alive at the beginning
of the follow-up period (1999-2004) with a mention of stroke during any admission between 1990 and 1998 as
well as new cases of admitted stroke during follow-up. Follow-up time and number of deaths were analysed for
each 5-year age group and the overall case-fatality rate reduced by the relevant background mortality. We then
modelled duration in years in DisMod 2 assuming no remission. We included in the model a declining time trend
of 2% per year for both case fatality and incidence over the last 14 years.
We adopted the disability weights used in the ABD2003 based on a one-year follow-up data of stroke survivors
in the Perth Community Stroke comparing health status information before and at 4 months and 1 year after the
stroke event.

Other cardiovascular diseases
Heart failure is the main disability from rheumatic heart & fever disease, inflammatory heart disease, hypertensive
heart disease and non-rheumatic valvular heart disease. The proportions of heart failure cases for each of these
causes were derived as described above for all heart failure. For rheumatic heart disease and non-rheumatic
valvular heart disease, a separate model was done using DisMod 2 based on heart failure prevalence and remission
through surgical interventions using Singapore hospital data.
We estimated the incidence of aortic aneurysm using Singapore hospital data. We assumed a one-month period of

                                                                                                                       
     disability during treatment and no residual disability for those who survived treatment. We applied the Australian
     derived weight using disability weight regression model for laparotomy (0.349).

     2M Chronic respiratory diseases
     Chronic obstructive pulmonary disease
     There is a lack of local population-level prevalence information on COPD with spirometry measurements. For
     this study, we had modelled symptomatic COPD individuals ever-hospitalised instead. This approach would
     however miss those who were asymptomatic as well as persons with symptomatic COPD but had never been
     admitted to hospital. Prevalence estimates for the year 2004 were derived linking Singapore hospital data on
     inpatient discharges with a diagnosis of COPD between 1991 and 2004, and Singapore mortality data for the
     period 1991 to 2003. Cases in the discharge dataset found in the mortality database were removed, and any
     readmissions found within the Singapore hospital data were subsequently excluded to estimate the pool of COPD
     survivors in 2004 by age and sex. To derive case-fatality, we calculated the number of years lived between 1999
     and 2004 by anyone who had ever been admitted with a diagnosis of COPD since 1991. The case-fatality rate
     was then taken as the number of deaths over person-years of follow-up in 5-year age groups after subtracting out
     the background mortality. We then used DisMod 2 assuming no remission to model incidence, age of onset and
     duration. We derived a composite average disability weight for males (0.288) and females (0.284) using the Dutch
     weight for mild/moderate and severe COPD and the proportional distribution by level of disability (moderate
     and severe disability considered here only) from the Busselton study. We added the proportion of heart failure
     cases attributed to ‘pulmonary heart disease’ on the basis that chronic lung disease was the underlying cause.

     Asthma
     Asthma is a dynamic disease. Its true prevalence is difficult to determine due to the lack of a single objective
     diagnostic test, different methods of classification of the condition, differing interpretation of symptoms
     in different countries, as well as the uncertain influence of increasing public and professional awareness of
     asthma.
     For the prevalence of asthma in adults, we used self-reported data from NHS2004 due to a lack of data on airway
     hyper-responsiveness test (AHT) plus current wheeze, which was largely proposed as a ‘gold standard’ for the
     diagnosis of asthma. Those who answered ‘Yes’ to the question “During the last 12 months, have you had an
     episode of asthma or an asthma attack?”, were considered as having asthma.
     For children, we used data from the International Study of Asthma and Allergies in Childhood (ISAAC) conducted
     in 2001. We used the prevalence of current wheeze and adjusted the prevalence by a factor of 0.42 based on studies
     where both wheeze and AHT were measured (Peat et al., 1994, Peat et al., 1995, Toelle et al., 2004). As ISAAC
     only conducted the studies on children aged 6-7 years and 12-15 years, we obtained the prevalence of other age
     groups by fitting a curve on available data and interpolating. We derived incidence estimates from DisMod 2,
     assuming remission rates from a follow-up study in the United States (Bronniman & Burrows, 1986).
     We did not use the Dutch weight for this health state (0.36) as we considered it to be for a more severe health
     state than the average for symptomatic asthmatics in the population. Instead, we used a derived weight of 0.229
     based on the severity distributions found in the 1998 Australian disability survey (ABS 1998) and the disability
     weight regression model. The remainder of the time was assumed to be spent in a state equivalent to the Dutch
     weight for asthma controlled by treatment (0.03). To derive the combined weight, we need to calculate the
     proportion of time asthmatics were symptomatic. For children, we used the number of wheezing episodes they
     had in a year. For adults, we used the number of days per month they had difficulty sleeping because of asthma
     symptoms. We calculated the respective disability weights for the children and adults and derived the average
     disability weight.


2N Digestive diseases
Peptic ulcer
There is an absence of population data on the frequency of peptic ulcer disease. We modelled incident cases
of symptomatic peptic ulcer disease as the number of individuals to hospital (either as inpatient or day surgery
patient) in the year 2004 for esophagoscopy, gastroscopy or duodenoscopy, with/without biopsy, and with a
recorded discharge diagnosis of peptic ulcer disease using Singapore hospital data. We assumed that 100%
treatment rate for peptic ulcer in Singapore and that 83% of cases were treated by Helicobacter pylori eradication
therapy, which had a cure rate of 90% (Mollison et al. 1999). We modelled those who were cured using eradication
therapy as being symptomatic for one month with no residual disability, and that the remainder of those who
were treated but not cured (including those receiving alternative treatments) received relief from their treatment
but remained with the condition for the GBD duration. We used derived weights from the Dutch study for both
symptomatic (0.064) and treated states (0.009).

Cirrhosis of the liver
There is a paucity of information on the incidence and prevalence of cirrhosis at the population level. This is
not so surprising given that diagnosis can only be made through liver biopsy, which is an invasive investigation.
Currently, diagnosis can also be done by ultrasound but there are no population studies in a representative sample
using this technique. We estimated the prevalence of cirrhosis in the year 2004 linking Singapore hospital data
of all inpatient discharges with a diagnosis of cirrhosis between 1991 and 2004, and Singapore mortality data
for the period 1991 to 2003. Cases in the discharge dataset found in the mortality database were removed, and
any readmissions found within the Singapore hospital data were subsequently excluded to estimate the pool of
cirrhosis survivors in 2004 by age and sex. For case-fatality, we calculated the number of years lived between
1999 and 2004 by anyone who had ever been admitted with a diagnosis of cirrhosis since 1991. The case-fatality
rate was then taken as the number of deaths over person-years of follow-up in 5-year age groups after subtracting
out the background mortality. We assumed no remission in DisMod 2.
In all first-ever hospitalised cirrhosis cases between 1995 and 2004 with a stated underlying cause, 35.5% were
alcohol related and 60.5% were non-alcohol related. Based on expert opinion in the ABD2003, we attributed
5% of non-alcohol related cirrhosis to other causes and the remainder to hepatitis (i.e. 59.5%). We assumed that
the number of hepatitis B related cirrhosis was about 300 times that for hepatitis C related cirrhosis, based on
estimated hepatitis B to hepatitis C carriers’ rates in the population. Hence, we estimated that 59.3% of cirrhosis
cases were due to hepatitis B and 0.2% due to hepatitis C. We gave a disability weight for the last 3 years lived
with cirrhosis at 0.31 (minus 2 months) and 0.84 for the last 2 months when encephalopathy occurs (effectively
interpreting the Dutch weight for compensated cirrhosis as relevant for the time spent in decompensated cirrhosis
and the decompensated cirrhosis weight of the Dutch as the weight for terminal liver failure). We then attributed
YLD due to cirrhosis associated with hepatitis B, hepatitis C and alcoholism to hepatitis B, hepatitis C and
alcohol use disorder respectively.

Appendicitis
The incidence of appendicitis was based on the number of individuals with appendectomy performed from
Singapore hospital data. GBD weight for appendicitis was used and we adopted duration of 2 weeks.




                                                                                                                      
     2O Genitourinary diseases
     Nephritis & nephrosis
     We modelled nephritis & nephrosis into 3 different categories: ESRD on dialysis, ESRD with transplant and
     untreated ESRD, excluding diabetic nephropathy which was treated separately as sequelae of diabetes. Our
     incidence estimates for dialysis and transplant cases were based on Singapore ESRD dialysis and transplant
     data, from which we derived durations for both categories of patients using DisMod 2. For dialysis patients, we
     used case-fatality rates to match observed deaths and remission through transplant, and applied the Dutch weight
     for diabetic nephropathy (0.290). In the first 6 months after transplant, we assumed a health state equivalent
     to the Dutch weight for diabetic nephropathy while for the remaining period with the transplant, a weight of
     0.11 was used, which was equivalent to both the GBD weight for treated renal failure and the Dutch weight for
     ‘uncertain prognosis’. Untreated ESRD was derived from the difference between dialysis or transplant deaths
     and total renal deaths, to which we apply duration of 1 year prior to death at the GBD weight for untreated renal
     failure (0.104). Based on the Singapore ESRD dialysis and transplant data on underlying distribution of ESRD
     incidences (excluding diabetic nephropathy), the YLD from analgesic nephropathy was attributed to the injury
     category of medical misadventure, congenital dysplasia and polycystic kidney disease to congenital urogenital
     disease and we retained only those for primary renal disease in the ‘nephritis & nephrosis’ category.

     Benign prostatic hypertrophy
     We based the incidence of benign prostatic hypertrophy on Singapore hospital data on the number of individuals
     with benign prostatic hypertrophy coded as the discharge diagnosis and transurethral resection of prostrate as
     the procedure performed. We followed expert advice adopted in the ABD2003: (a) number of benign prostatic
     hypertrophy cases was adjusted upwards to account for the proportion of cases that received medical instead of
     surgical treatment; (b) half of all benign prostatic hypertrophy cases received surgical treatment, a proportion
     of whom experienced complications or continuing symptoms following surgery (1% with lifelong incontinence
     at a derived weight of 0.157, 15% with lifelong impotence at the GBD weight of 0.195, and 5% with urethral
     stricture for 4 weeks at the GBD weight of 0.151); (c) of those opting for medical treatment, 70% was assumed
     to use alpha-blocker drugs, of which half were cured and the other half may then try surgery; (d) none of those
     receiving drugs other than alpha-blockers were cured.
     GBD weight for symptomatic benign prostatic hypertrophy was applied to each of these intervention pathways
     with the following duration assumptions: 1.5 years for surgery, 1 year for successful medical treatment, 2 years for
     unsuccessful medical treatment then surgery, and lifelong for unsuccessful medical treatment but no surgery.

     2P Skin diseases
     Specific morbidity models for skin diseases were not modelled. We used the age-specific YLD rates by sex for
     skin diseases estimated by the WHO for Singapore for the year 2002 (WHO 2004b).

     2Q Musculoskeletal diseases
     Rheumatoid arthritis
     There is a lack of local population-based data on rheumatoid arthritis. We used the GBD OAI estimates of
     1.5/1,000 in males and 2.5/1,000 in females. The higher prevalence of rheumatoid arthritis in women was a
     consistent finding across the world. In DisMod 2, we assumed a remission rate of 0.04 (based on Prevoo et al.
     (1996) who reported 15% remission over 4 years of follow-up) and relative risk of mortality of 1.6 (Pincus et
     al, 1994) which we tapered off to 1.1 in the oldest age group. We assumed the distribution by level of severity
     from a Finnish study (Hakala et al, 1994) and calculated a compound weight of 0.271 from the Dutch weights by
     severity level: 25% no disability, 40% mild at 0.21, 25% moderate at 0.37 and 10% severe at 0.94.

Osteoarthritis
We based our estimates for this condition on reported findings of radiographic osteoarthritis (grade 2 and above)
by affected joint, age and sex from a large-scale study in Massachusetts, USA (Oliveria et al. 1995) because of
the lack of local population-based studies on osteoarthritis. We modelled hip and knee osteoarthritis. We used
DisMod 2 to derive average durations, assuming a slightly increased risk of mortality (1.1) and then using the
observed remission rate from joint replacement surgery. Because osteoarthritis is a relatively slow progressive
disease, with few patients showing symptomatic progression over an 11-year period (Ahern & Smith 1997), we
applied an average of the relevant Dutch weights, assuming a severity distribution based on the Framingham
study (Guccione et al. 1990).

Low back pain
We only considered acute back pain cases as we did not have sufficient data to identify chronic cases. This would
underestimate the burden of back pain. Our estimates for back pain were based on self-reported prevalence of
pain located on the back from the NHSS2001. Only those with pain of at least mild intensity and limited in their
usual activities were considered. We used the Dutch weight for low back pain (0.06) as the disability weight for
recent episodes of back pain, which applied to an average health state involving some problems in walking about
and in usual activities, as well as moderate pain or discomfort. We used an average duration of 4.6 days and 5.5
days from the NHSS2001 for painful and limiting episodes of back pain for male and female respectively.

Gout
Gout is a common form of arthritis or inflammation of a joint. It most commonly affects the big toe joint (first
metatarsophalangeal joint), but can affect any joint. There is not much local data on gout. Our estimates for gout
were based on self-reported prevalence of arthritis from the NHSS2001. We assumed that gout accounted for
approximately 5% of all cases of arthritis. We also made adjustment on the gender-specific prevalence as gout
is often seen as a male disease. We assumed a slight increased risk of mortality associated with gout (relative
risk=1.1) and no remission, based on information that at 1 year 62%, at 2 years 78% and at 10 years 93% had at
least one repeat attack (Alamo Family Foot and Ankle Care 2005). Adopting the approach used in the ABD2003
in estimating the time symptomatic, we assumed that 10% had chronic symptoms and the remaining 90% had an
attack of 1 week every 0.44 years. Given that people might suffer gout at varying levels, from acute attacks of
a short duration to chronic gout, we assumed that on average, one attack per 2 months lasting 1 week in 90% of
people and the remaining 10% suffered chronic ongoing disease at the GBD weight of 0.061.

2R Congenital anomalies
Anencephaly
We estimated the incidence of anencephaly using Singapore mortality data, assuming deaths were equivalent to
incident cases. We assumed duration of 1 week with a disability weight of 1.

Spina bifida
We estimated the incidence of spina bifida using Singapore birth defects data. We assumed the same average
length of life for a liveborn spina bifida case as in the established market economies from the GBD study and
used the GBD weight (0.593).

Congenital heart disease
We modelled the disability associated with 4 types of congenital heart disease for live-born infants: surgically
treated atrial or ventricular septal defect, surgically treated Fallot’s tetralogy or transposition of great vessels,


                                                                                                                        
     surgically treated pulmonary stenosis, and complex but not curatively operable congenital heart disease. For
     each of the first three conditions, we assumed that a curative procedure from Singapore hospital data would
     represent an incident case with disability. We applied a duration of 1 year prior to operation with disability
     equivalent to the Dutch weight for moderate heart failure (0.35), while for post surgery, we used the relevant
     Dutch weights and reduced life expectancy was assumed except for those with septal defects. We assumed that
     lifelong underlying disability started at birth and discounted YLD back to birth. We estimated the incidence of
     other congenital heart malformations based on all other heart procedures performed on live-born infants with
     congenital malformations from the Singapore hospital data. We assumed that 50% of these cases were complex
     but not curatively operable. We reduced life expectancy of these cases to half that of those with surgically treated
     conditions and used the relevant Dutch weight (0.72).

     Cleft lip and/or palate
     We estimated the incidence of cleft lip or/and cleft palate using Singapore birth defect data. We assumed that
     100% were surgically treated; and that all received surgery within the first year. We assumed disability equivalent
     to the ‘treated’ GBD weights (0.016 for cleft lip with/without cleft palate, 0.015 for cleft palate without cleft
     lip).

     Digestive system malformations
     We modelled the disability for anorectal and oesophaheal atresia and other digestive system malformations. We
     estimated the incidence of these conditions using Singapore birth defect data. For both types of atresia cases,
     we assumed 26 weeks of disability from birth at the GBD weight for anorectal atresia (0.85). After this period,
     we assumed a proportion of both types of atresia cases had long-term disability (15% and 20% respectively)
     and decreased life expectancy (by 10 and 5 years respectively) and adopted the Australian derived weight from
     disability weight regression model (0.037). For other digestive system malformations, we assumed no lifelong
     problems, and a one-month period of disability from birth equivalent to the GBD weight for anorectal atresia.

     Renal agenesis
     We estimated the incidence of unilateral and bilateral renal agenesis using Singapore birth defect data. For
     unilateral cases, we assumed that 20% of survivors had ongoing problems, with a life expectancy of 70 years and
     used the GBD weight of 0.037 for symptomatic urethritis. For bilateral cases, we assumed an average duration of
     3.5 days and used the GBD weight for renal agenesis (0.85). We calculated YLD for end-stage renal failure due
     to renal dysplasia based on attribution from the Singapore ESRD dialysis and transplant data.

     Other urogenital tract malformations
     We modelled the disability associated with cystic kidney disease, obstructive defects of renal pelvis and ureter,
     and other urinary tract malformations. Incidence of cases of other urogenital tract malformations was estimated
     using Singapore birth defect data. We assumed that 30% of cases resulted in chronic lifelong problems, with a
     life expectancy of 50 years and a disability weight of 0.037. YLD were calculated for end-stage renal failure due
     to cystic kidney disease based on attribution from the Singapore ESRD dialysis and transplant data.

     Abdominal wall defect
     We estimated incidence of abdominal wall defects (exomphalos and gastroschisis) from Singapore birth defect
     data. We followed the assumptions used in the ABD2003 where duration of 4 weeks was used, and an estimated
     20% of cases had lifelong problems. For those cases with long term disability, a shortened life expectancy by 20
     years was assumed and the disability weight of 0.20 (the Dutch weight for young adult in permanent stage after
     surgical repair to Fallot’s tetralogy) was used.



Down syndrome and ‘Other chromosomal disorders’
For both Down syndrome and ‘other chromosomal disorders’, our estimates were again derived using the Singapore
birth defect data. We assumed the same distribution of intellectual disability by severity level (mild, moderate,
severe and profound) and sex in the ABD2003 for these two conditions due to paucity of local information, and
also followed the amount of reduction in life expectancy for mild, moderate, severe, and profound intellectual
disability estimated in the study. Dutch weights for these intellectual disability severity levels were used.

2S Oral diseases
Dental Caries
We estimated the incidence of dental caries using decayed (D), missing (M), and filled (F) primary or permanent
teeth (T) prevalence data from the 2003 National Oral Health Survey of School Children (HPB, personal
communication) and the SOHS2003. Fitting linear regression lines to these prevalence data by age group gave
slopes in children (0-19 years) of 0.24 and in adults (20 years and above) of .026. These would represent the
annual caries increment in children and adults. We adopted the durations of symptomatic caries of 28 days
in children and 55 days in adults from the ABD2003, at the Australian derived weight from disability weight
regression model of 0.057, and a proportion of 32.4% of people with symptomatic caries.

Periodontal disease
We estimated the prevalence of periodontal pockets larger than 6 mm using data from the SOHS2003 for adults
aged 20-74 years, and a 2001 school dental survey consisting of subjects at age 18 years selected from 4 junior
colleges (HPB, personal communication). We used DisMod 2 to interpolate prevalence from examined age
groups to age groups used in the SBoD 2004 Study, and extrapolated for older age groups assuming increased
prevalence with age. Incidence and duration were derived in DisMod 2 based on remission rates that reflect 15
years average duration and no case-fatality. We followed expert advice adopted in the ABD2003 that people
with periodontal pockets larger than 6 mm were symptomatic for around 1% of the time, and used the Australian
derived weight from disability weight regression model for periodontal abscess (0.056).

Edentulism
We estimated the prevalence of edentulism from the SOHS2003, and derived incidence and duration using
DisMod 2, assuming no remission and no excess case-fatality. We included in this model a 10% declining time
trend to reflect the observed decline of the prevalence over the last 12 years. We assumed no edentulism was
untreated in Singapore and adopted the Australian derived weight from disability weight regression model for
treated edentulism (0.004).

3 Injuries
We modelled the disability from non-fatal injuries where a person has an injury severe enough to warrant hospital
treatment but not leading to death. It was assumed that injuries treated outside the hospital system did not result
in significant disability. We obtained non-fatal incident injuries using Singapore hospital data and attendance
data from the emergency department of public hospitals. We classified incident cases according to a matrix of 16
‘external cause of injury’ categories (13 unintentional and three intentional) and 32 ‘nature of injury’ categories
(for example fractures, burns, wounds, brain injury, spinal cord injury). We excluded admissions for the same
ICD-9 code within 90 days, on the assumption that these were re-admissions, as well as those resulting in death.
Given that it was not uncommon for multiple sites of the body to be damaged from a single accident, we estimated
disability for only the most disabling ICD-9 code associated with each incident case, on the assumption that the
disability for the other ICD-9 codes was captured in the weight for the more severe injury. We redistributed ill-


                                                                                                                      
     defined injuries. We used disability weights, durations and the risk of mortality as per the GBD study.
     Our estimates for non-admitted emergency cases were based on data from the emergency department of public
     hospitals. However, not all hospitals had data for both ‘nature of injury’ and ‘external cause of injury’. Hence,
     extrapolations were made to obtain nation wide incidence of hospital-treated injury.






				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:1/20/2011
language:English
pages:23