Health Metrics for Helminth infections

Abstract

Health metrics based on health-adjusted life years have become standard units for comparing the disease burden and treatment benefits of individual health conditions. The Disability-Adjusted Life Year (DALY) and the Quality-Adjusted Life Year (QALY) are the most frequently used in cost-effect analyses in national and global health policy discussions for allocation of health care resources. While sometimes useful, both the DALY and QALY metrics have limitations in their ability to capture the full health impact of helminth infections and other ‘neglected tropical diseases’ (NTDs). Gaps in current knowledge of disease burden are identified, and interim approaches to disease burden assessment are discussed.

1. INTRODUCTION

1.1 Definition of Health Metrics

In a broad sense, health metrics are numbers assigned to quantify the impact of different diseases on personal health status. They are numerical judgments of the value of certain health states or treatment outcomes and, in the literature of the different disciplines that evaluate health outcomes, they can be referred to as ‘preferences’, ‘values’, or ‘utilities’ (Gold et al., 1996). In economics, the ‘utility’ of a transaction can be inferred from an informed, ‘rational’ consumer’s willingness to pay for goods or services. By contrast, in health care, many personal factors can influence a person’s valuation of a given health state. This makes it quite challenging to define health metrics that accurately reflect the value of individual health states to society-at-large.

The notion underpinning the use of such health ‘utilities’ is that they could allow us to compare the efficacy and effectiveness of different health interventions on a ‘like-is-like’ basis (Gold et al., 1996; Murray, 1996). If it is believed that if such metrics are applied to different diseases in a ‘fair’ manner, then the relative value of each intervention can be assessed in terms of the health utility gained, as captured by the health metric. The result is that decisions in favor of more ‘effective’ interventions can be made on a utilitarian basis across a broad range of health-related harms (Murray, 1996). The use of health metrics presupposes that the decision maker agrees with this unitization of health states, and that maximization of health utility is the right approach to allocation of limited health care resources.

1.2 Why try to quantify disease burden?

The move to quantify disease impact of different health states comes from the greater implementation of health economics in evaluation of disease control initiatives. It springs from the efforts of policymakers to improve the efficiency of health care investments and health care delivery in all settings, including less-developed countries.

In standard cost-effectiveness analysis (CEA), the final outcome that is typically assessed is ‘cost per health-unit gained’, while in cost-benefit analysis (CBA) the outcome is in the form of ‘cost spent per costs saved or averted’ through the intervention program. To conduct either CEA or CBA, there must be an identifiable, measurable, scalable, unitized consequence that results from the proposed treatment (or preventive care) given by the health program under study. In sum, the gain can either be measured as non-monetary health effects, as is done in CEA, or as associated monetary benefits, as is done in CBA. Those who pay for health care (payors) may be most interested in CBA, whereas patients, social programs, and healthcare providers may be more interested in the outcomes that are included in CEA.

Early evaluations of helminth control programs used ‘cases prevented’ or ‘cases cured’ as the outcomes that were measured in their CEA (Guyatt, 1998). The drawback to this disease-centric approach was the outcomes could not then be compared among the different parasite infections, because each infection had its own pathogen-specific morbidities. Similarly, without a generalizable health metric, there was no meaningful comparison between deworming interventions and the many competing disease control programs implemented elsewhere in the health sector.

The cost effectiveness approach is fundamentally utilitarian in philosophy, encapsulating the belief that program impact can be potentially maximized at a minimum of cost (Gold et al., 2002). The assumption is that the expected outcomes of the program are scalable (i.e., retaining the same value at all locations and at all levels of aggregation), additive (i.e., two units gained are twice as valuable as one unit gained), and are fungible (i.e., interchangeable among different individuals in the same or different locations) (Murray, 1996). As discussed later in this paper, human health perception and health-seeking behavior diverge significantly from this assumed unitization of health. However, the creation of such health ‘units’ is essential to performing the sort of comparisons favored by health economists in their ‘league table’ rankings of health investments (Jamison, 1996, 2006). It has long been recognized in economics that the value of a transaction is determined by the consumer’s (not the seller’s) perception of value. Because of this, the units of health or health burden employed in CEA must in some way incorporate patient preferences about the possible alternative health outcomes that are being compared (Gold et al., 1996).

The ultimate factor driving the move towards CEA is the knowledge that resources are limited for investment in health care and the efficient allocation of scarce resources would be an ethical ‘good’ for the world community. The salient policy discussion in setting disease control priorities has been, “How much health can we buy for 1 million dollars?” (Jamison, 1996) Unfortunately the basic cost- effect algorithm is an essentially linear approach that is often not well suited to the realities of health care, and not well suited to the non-linear economic features of low-income life (Banerjee and Duflo, 2011; King and Bertino, 2008).

1.3 The challenge of disease burden assessment for parasites

In defining the health burden of developing countries, there is a new appreciation of the role of chronic parasitic diseases in the perpetuation of disability, particularly in the setting of rural poverty (Engels and Savioli, 2006). In highly developed regions such as Europe and North America where parasites are now infrequent, we tend to conceptualize infectious diseases predominantly as ‘acute’ health problems that will respond rapidly to appropriate antimicrobial therapy and leave the treated patient with only minimal or no lasting disability. While there is an increasing realization that a number of major chronic diseases are caused by infection, including those caused by human papilloma virus (HPV), Helicobacter pylori, HIV, hepatitis B virus (HBV) and hepatitis C virus (HCV)), in high-income countries, the impact of communicable diseases is often considered to be minimal compared to the impact of chronic non-communicable diseases (Gwatkin et al., 1999; Murray et al., 2012c).

By contrast, in the developing world, parasitic infections are common, recurrent, and long-lasting health problems that represent an ongoing inflammatory challenge and a significant health threat to the populations who are at continuing daily risk for reinfection (Jia et al., 2012; Satayathum et al., 2006; Wang et al., 2012). Also, in the context of health burden assessment, it is important to realize that parasite-related disease outlasts the period of parasitic infection (Giboda and Bergquist, 1999)—in the typical endemic setting, active infection represents only a major risk factor for parasite-associated disease. Although past Global Burden of Disease Program assessments (Mott, 2004; Murray and Lopez, 1996) have assumed that patients with helminth infections are ‘mostly asymptomatic’, this is not true--meta-analysis of available clinical evidence has shown that helminth infections are associated with many significant morbidities and chronic/permanent disabilities (Carabin et al., 2011; Chan, 1997; Furst et al., 2012a; King et al., 2005; Quattrocchi et al., 2012).

There is a severe lack of information on the long-term outcomes of patients exposed to chronic parasitic infections during their childhood or young adulthood. Risk of advanced pathology over time has often only been inferred via cross-sectional studies (e.g., (King et al., 1988), in which patient age serves as a proxy for duration of exposure among long-term residents of an endemic area.

Formal decision analysis for health care resource allocation is often based on Markov- type models that project the expected outcomes of an intervention based on the probabilities of disease, its proper diagnosis, and its response to different treatments over time (King et al., 2011; Petitti, 2000). For helminth diseases, such a life-path analysis approach has suffered from a lack of well-measured longitudinal data inputs. Conditional probabilities for transition from mild to more severe health states are mostly unknown (Kirigia, 1997, 1998). It has been demonstrated that both patients and ‘disease experts’ have a poor ability to prognosticate on the risk of transition between health states associated with the various conditions associated with helminth infection (Kirigia, 1997, 1998). By the same token, most formal treatment studies have typically provided outcomes and follow-up of therapy for only 1–2 years’ duration (Richter, 2003).

While it is true that there are many highly effective antihelminthics for treating these parasitic infections (e.g., praziquantel, albendazole, mebendazole (Anonymous, 2010)), within a resource-challenged region, access to effective treatment is still not generally available. Furthermore, even after successful therapy, environmental factors can strongly favor the process of reinfection (Jia et al., 2012; Satayathum et al., 2006). As a result, for the majority of local residents, the worm infections will frequently recur, and disease will persist for most of their lives. In fact, worm infections are so common that chronic manifestations of infection-associated disease are often mistaken as normal for health status in endemic areas (Amazigo et al., 1997; Danso-Appiah et al., 2004; Mekheimar and Talaat, 2005; Ukwandu and Nmorsi, 2004). This makes it difficult for people in endemic areas to provide a valid comparison of their infected health state to a ‘full-health’ state, which is understood to be the expected norm in the cultures of high-income, highly developed economies, such as Europe, Japan, or North America. As discussed later in the article, this places limitations on the use of standard health metric valuation approaches for diseases of developing countries.

2. Commonly used health metrics: DALYs and QALYs

This section reviews the approaches most commonly used to quantify disease-related health burden in policy discussions. The metrics used most often are generically referred to as Health-Adjusted Life-Years (HALYs) (Carabin et al., 2005). Disability-Adjusted Life Years, or DALYs, and Quality-Adjusted Life Years, or QALYs, are the two formulations most frequently seen in the literature (Gold et al., 2002).

2.1 The DALY

In international circles, the Disability-Adjusted Life Year, or DALY, is probably the most frequently used metric for policy planning. The DALY is a time-based unit that sums two disease-related components, person-years lost to disability (YLD) and person-years of life lost (YLL), linked to having any individual health condition (Gold et al., 2002; Murray, 1996). The intent is to reflect the disability burden for that prevailing health condition, and then allow comparisons of disease-specific burdens across a large number of different disease states.

The DALY was first used in the 1990s as part of the World Bank-WHO Global Burden of Disease Project (GBD) (Murray and Lopez, 1996). To calculate DALYs for a given condition and year, it is necessary to know the age- and sex-specific incidence and death rates for the disease, its typical duration, and the ‘disability weight’ (Dw, see below) ascribed to having the condition. The Dw is allowed to vary if the person with the condition is under treatment or not. Assuming a stable epidemiological situation, the death rates, incidence and duration of a disease can be approximated from information on current age- and sex-specific prevalence and on known remission or cure rates. The DALY is basically one person’s life-year weighted by the GBD-assigned Dw for his or her health state, which varies from zero (perfect health) to one (death). Of note, DALYs are assigned and counted for an individual for only one disease health condition, presumably their most ‘disabling’ condition. Late disease complications that occur with chronic infection, such as cancer, epilepsy, anemia, and undernutrition, have often been treated as separate diseases in the DALY system because of limited data on their pathogen-related attributable risk. This serves to minimize the perceived impact of communicable diseases, and leads to systematic underestimation of their DALY scores.

2.1.1. The DALY component from Years of Life Lost (YLL)

For those individuals who die from their disease, the resulting disease DALYs are scored based on the years of life lost (YLL) for that person. The expectation is that if that person had lived, they would have had the life expectancy of a person living in the best-off communities within high-income countries, e.g., 86 years for men and women living in Japan (Gold et al., 2002; Murray, 1996; Murray et al., 2012b). As an example, an infant girl who dies at age 2 of malaria adds a basic 84-DALY burden (non-weighted, non-discounted DALYs, see below) assigned to the disease ‘malaria’ in the year that she died. Because the DALY system does not account for co-morbidities or for serial health problems, an important, hidden assumption in the YLL DALY calculation is that the years of life lost would have been spent in perfect health. YLL calculations are now based on a new reference standard life table using the lowest observed death rate for any age-group in countries with more than five million in population (Supplementary appendix to Murray et al. (2012b)). The 86-year life expectancy for all humans was chosen by the GBD Program as an aspirational value, and does not reflect the actual life-expectancy of people living in less well-off countries. Where cause(s) of death are complex or not well diagnosed, YLL DALYs are undoubtedly missed or mis-assigned.

2.1.2. The DALY component from Years Lost to Disability (YLD)

For non-lethal conditions (and for the ante-mortem period of lethal diseases) one calculates a ‘years lost to disability’ (YLD) score for each person affected by that disease. YLD is determined as (disability weight) times (time spent in that disease condition) (Figure 1). Net total DALYs for any specific health condition are then summed, overall, as DALY = YLL +YLD.

Figure 1.

Schematic comparison of DALY and QALY health metric approaches to assessing disease impact.

For many short-term illnesses such as acute malaria, disability is significant for a period of only a few weeks, then quite minimal after that. If say, someone is 25% disabled for 2 weeks (1/26^th of a year), the resulting DALY score would be 0.01, based on YLD. For chronic conditions such as helminth infections, which persist for an indefinite period of time but have less profound disabling impact, perhaps a 1% disability, the DALY score would be 1% × 1 year = 0.01 DALY. Thus, according to the DALY system, a year of helminthic infection is the same in terms of YLD as for having a malaria attack once a year.

2.1.2.1 How disability weights were determined for the 1990–2009 era

The Dw values for the DALY program were initially determined by a health economics exercise performed in Geneva as part of the first GBD rankings. Panels of non-patients, drawn from a number of different countries were asked to rank order 21 ‘anchoring’ conditions over the DALY disability range from 0 (full health) to 1 (death) (Murray, 1996). To elucidate their preferences, the panel were asked to use a ‘person trade-off (PTO)’ approach, in which they were asked if they would rather buy an extra year of life for 1000 health persons, or an extra year of life for some number (> 1000) of disease affected persons (PTO1). Later they were asked if they would trade a year of healthy life for 1000 persons in order to eliminate a disabling condition for one year for some number (>1000) of persons with one of the ranked diseases (PTO2). Numerical answers to these exercises then had to be reconciled by ‘guided deliberation’ of the panel, and the results were ranked in assigning numeric disability scores. Once the 22 ‘benchmark’ conditions were established, all other health conditions to be included in the GBD were ranked relative to this ladder of benchmarked conditions, and disability weights assigned within the benchmarked interval range (Murray, 1996). Of note, these additional health conditions were not formally rated by PTO, and so cannot be considered ‘utilities’ in a formal economics sense. Nevertheless, the DALYs derived from these disability ranking are often treated as utilities in public policy CEAs.

2.1.2.2 How disability weights were determined for the latest GBD 2010 estimates

In order to address the concerns raised about the 1990s approach to disability weight estimation, the new Global Burden of Disease 2010 (GBD 2010) project implemented a new approach to disability weight assessment (Salomon et al., 2012). In this component of GBD 2010, brief vignettes of people in each of 108 health states were paired and presented to people in areas of Bangladesh, Indonesia, Peru, Tanzania, and the USA. They were asked to identify which of the two hypothetical people being compared they thought was the healthier person. The subjects interviewed for this ranking exercise were selected through a multi-stage, stratified household-based sampling scheme, and were felt to be representative of the national populations surveyed. A supplemental web-based survey was also performed based on 220 health states (Salomon et al., 2012). Following rank ordering of the survey preferences for the tested health states, the study authors “…used numerical integration to obtain mean estimates of disability weights on the natural zero-to-one scale.” (Salomon et al., 2012) The health vignettes had to be fewer than 35 words long, and could contain information on the severity of the condition. Relevant to helminthes, only two diseases were included in the health state comparisons. These were: i) Symptomatic Intestinal nematode infections, described as “…has cramping pain and a bloated feeling in the belly” which was ultimately assigned a Dw of 0.03 (equivalent to 3% disability); and ii) Symptomatic lymphatic filariasis, described as “has swollen legs with hard and thick skin, which causes difficulty in moving around,” which was given a Dw of 0.11 (Salomon et al., 2012). Of note, other helminth infections were not included in the descriptions presented in the surveys. In addition, in the subsequent GBD calculations, most persons with worm infections were considered to be in a generic, minimally symptomatic category of ‘acute mild infection’ having Dw = 0.005, or one-half of one percent disability. It is concerning that, although the GBD 2010 ranked 291 conditions, ranking inputs were collected on only 220 diseases, including only a limited number of the 1160 sequelae that were ultimately scored the overall GBD 2010 report. As for all previous GBD rankings, this information gap regarding disability inputs raises concerns about the fairness of the rankings for many of the NTDs.

2.1.2.3 Current DALY estimates from the GBD 2010 Project

Table 1 shows the global DALY estimates for the nine helminthic infections included in the GBD 2010 Project (Murray et al., 2012c; Vos et al., 2012). Notably, over 95% of the helminth-attributable DALYs are from years lost to disability (YLD), and not from premature mortality. By contrast, for malaria, 98.5% of DALYs are due to years of life lost (YLL), reflective of its high mortality rates in early childhood.

Table 1.

Current DALY estimates from the Global Burden of Disease 2010 Project^a

Health Condition	2010 DALYs due to Years- Lost-to- Disability (excluding mortality)	2010 DALYs due to Years-of-Life- Lost	Total 2010 DALYs (due to disability plus mortality)b
GBD 2010 Included Helminths:
Hookworm	3,230,800	0	3,230,800
Schistosomiasis	2,986,200	323,000	3,309,000
Lymphatic filariasis	2,774,700	0	2,774,700
Food-borne trematodiasis	1,875,400	0	1,875,400
Ascaris infection	1,110,600	204,000	1,314,800
Trichuriasis	638,200	0	638,200
Onchocerciasis	494,000	0	494,000
Taeniasis/cysticercosis	456,900	47,000	503,400
Echinococcosis	109,900	34,000	143,900
Not included:
Loiasis	--	--	--
Toxocariasis	--	--	--
Trichinosis	--	--	--
Strongyloidiasis	--	--	--
Enterobiasis	--	--	--
Helminths subtotal	13,676,700	608,000	14,284,200
Other NTDs and non-malarial vector- borne diseases	4,472,800	7,296,000	11,769,200
Malaria	4,070,000	78,615,000	82,685,200
All NTDs, VBDs, and malaria	22,219,500	86,519,000	108,738,600
All 291 causes included in GBD 2010	777,401,500	1,712,983,000	2,490,384,900

^aAbbreviations used: DALY, Disability-Adjusted Life Year; GBD 2010, Global Burden of Disease 2010 Project; NTD, neglected tropical diseases; VBD, vector-borne diseases.

^bDALY estimates are taken from (Lozano et al., 2012; Murray et al., 2012a; Vos et al., 2012). Note that the disease-specific totals may not sum exactly across the rows because of rounding off of numeric values reported by these references in their GBD 2010 tables.

The GBD 2010 Project worked with Expert Groups to develop the new DALY estimates. The experts’ inputs included extensive literature reviews regarding the prevalence, incidence, mortality, remission or cure rates, and duration of their disease of interest in different sex and age groups and regions. It also was meant to help analyze the different stages of severity of the disease, and their impact on personal health burden. For example, for the Schistosoma parasites, these were then taken by the GBD 2010 staff to calculate DALYs related to the different stages of schistosomiasis. Because disability weights were not specifically developed for each Schistosoma infection-related condition, the GBD team used Dw values developed for analogous conditions in the vignette ranking exercise described in section 2.1.2.2. Table 2 indicates the breakdown of the prevalence estimates for different known complications of Schistosoma infection and their corresponding DALY estimates developed by the GBD 2010 core team (Jasrasaria et al., 2012). The estimated case numbers are higher than WHO estimates of active infections (WHO, 2009), because Schistosoma infection-related disease, particularly the severe complications, persist into adult life after active infection can be over (Giboda and Bergquist, 1999). The total number of schistosomiasis-attributed deaths was low (11,654) compared to previous WHO estimates (van der Werf and de Vlas, 2001).

Table 2.

Breakdown of DALY burden for Schistosoma infection and its complications according to the GBD 2010 Project working estimates (Jasrasaria et al., 2012)

Schistosoma- associated condition	Disability Weight	Number affected	YLDa in 2010	YLLa in 2010
Diarrhea	0.0586	111,690	6540	--
Mild Anemia	0.004725	32,100,000	151,673	--
Severe Anemia	0.164	2,182,853	357,824	--
Hepatic inflammation	0.01081	13,924,407	150,562	--
Hematemesis	0.3262	395,491	128,998	--
Ascites, moderate	0.1208	1,197,777	144,699	--
Dysuria, mild	0.01081	26,700,000	289,074	--
Bladder pathology, mild	0.01081	33,600,000	364,174	--
Hydronephrosis	0.01081	28,800,000	311,247	--
Active infection, NOSa	0.00427	221,000,000	911,717	--
Total YLD			2,816,508b
Deaths		11,654
Total YLL				323,000
Total DALYs		3,139,508b

^aAbbreviations used: YLD, years lost to disability; YLL, years of life lost; NOS, not otherwise specified

^bNB: The numbers of total YLD and consequently also total DALYs in this table do not exactly match with the respective published DALY numbers in Table 1 because only unofficial preliminary GBD 2010 YLD numbers were made available to the author during manuscript preparation. Final YLD numbers for Schistosoma infection-related complications have not yet been released.

There are two other significant changes in the current GBD 2010 DALY estimates--Age weighting and time-discounting were eliminated from the calculations. Their controversial inclusion (Anand and Hanson, 1997) was strongly influential in creating the original DALY scores: Specifically, age-weighting was used to assign much greater burden estimates to diseases affecting persons 20–30 years of age, and time discounting reduced the ‘importance’ of events far in the future. The justification for age-weighting is that there was a ‘social consensus’ that persons of peak productive potential are more valuable to the community (Murray, 1996). Time discounting was included as a standard aspect of health economics analysis to avoid undue deferral of health interventions with a fixed budget. This latter aspect was sufficiently controversial, however, that discounted and non-discounted DALYs were both provided in the 1996 and subsequent GBD tables (Murray and Lopez, 1996).

2.1.3. Assumptions of the DALY process

The objective of the DALY framers was to develop a ‘fair’ estimate of the impact of a given disease in all settings around the world (‘like is like’ was the term used (Murray, 1996)). That is, the context of disease was meant to be totally excluded—the focus was solely on the individual impact of disease, as determined by the preferences of the ‘average person’ panels (Murray, 1996; Salomon et al., 2012). By programmatic policy, ‘handicap’, or disease impact within a social setting, was not considered. As the GBD team states, ‘…our study’s focus [is] on the construct of health loss rather than welfare loss…’ (Salomon et al., 2013). Disease impact on earnings or livelihood is excluded from the disability scoring. Likewise growth stunting and mild cognitive impairment caused by infection are not considered disabilities. Lastly, for zoonoses, the impact of a pathogen on household livestock and earnings is also ignored (Carabin et al., 2005). For the most part, co-disability with two or more conditions is not considered in the DALY formulations. In addition, patient self-reported illness impact or health-related quality of life is not considered. This aspect was deemed by the GBD team to be too variable or unreliable (lacking ‘criterion validity’) to provide standardized disease impact values (Murray, 1996).

2.2. The QALY

The QALY is an alternative means of assessing health burden based on Health Related Quality of Life (HRQoL) that has been developed over the last several decades. It is also a time-based unit of HALY, in which the impact of health status is derived from patient preference regarding a specific level of HRQoL related to their health condition. HRQoL determination is typically based on using questionnaires to inventory patient status in different performance domains, then asking patients and non-patients to rate that performance constellation on a visual analogue (thermometer) scale, a Likert scale, or time trade-off scale. This latter component is needed to incorporate patient preference into the valuation of the health state under study, and hence render the HRQoL assessment a ‘utility’ in the economic sense (Gold et al., 1996). Less commonly, QALYs have been estimated using standard gamble approaches to determine preferences to different health states (Gold et al., 1996). From the economist’s standpoint, this latter approach would provide more orthodox health state valuation for use as ‘utility’ in CEA. The domains in a HRQoL health status profile typically include social function (communication, social relations, intimacy), psychological function (cognition, emotion, mood), physical function (mobility, activity, self-care), sensory impairment (vision, hearing or other), and the individual’s overall perception of their health. This approach evokes, in more systematic fashion, the multiple potential impacts of a particular disease, while also incorporating an estimate of personal perception or preference regarding the overall disease state. The net score is translated into an estimate of how the subject’s health state compares to full function on a HRQoL scale of 0 (dead) to 1 (full health), and the Quality-Adjusted Life Year (QALY) impact of the disease calculated by multiplying the HRQoL valuation weight times the duration of time spent in that health state (Figure 1). Different from DALYs, life-path QALY summary scores do not include years of life lost (Gold et al., 2002). Note that because of the differences in the directionality of the scales used, the desired health outcomes of a program would be to maximize QALYs through intervention, whereas one would wish to minimize DALYs or avert them altogether.

2.2.1. Current evidence on the QALY impact of helminthic infections

Since the advent of the DALY as a dominant metric for allocating disease control priorities in developing countries (Jamison et al., 2006), researchers and health workers have questioned the validity of the low DALY disability weights (0.4 – 3%) attributed to chronic, non-lethal disease such as helminth-related illnesses. In response to the DALY’s brief ‘scenario’ approach to quantifying disability due to parasitic infection, researchers have begun the time-consuming task of formally quantifying health performance status of people with these chronic helminth-associated conditions (Furst et al., 2012b; Jia et al., 2007; Jia et al., 2011; Terer et al., 2013). Their studies have utilized standardized HRQoL instruments, such as the Euro-QoL 5, WHOQoL-Bref, and the Peds-QL to more fully assess patient physical performance and psychosocial status, and linked those to QALY-based estimates of disease impact for schistosomiasis and soil-transmitted helminths. In related patient health assessments, the disability weights derived from these descriptions for people living in parasite endemic areas ranged from 4% to 51%, increasing in magnitude depending on: i) the infecting species (Schistosoma versus STH); ii) the subject’s age, sex, and socioeconomic status; and iii) the presence of advanced complications of infection, such as severe S. japonicum-related hepatic fibrosis. Notably, these disability weights are 1.3-fold to > 20-fold higher than those estimated by the GBD DALY Dw approaches (Murray, 1996; Salomon et al., 2012).

2.2.2. Assumptions of the QALY process

The assignment of a quality weight or score for a given condition presupposes that the value reflects community, or at least patient, preferences about the health state being scored. The health-related performance issues evoked in the QoL questionnaires are primarily generic, and not specifically related to one disease. This potentially allows for a more generalizable assessment of health status by each individual, but would not discriminate the specific impact of an individual disease on their health, if more than one condition were present. Perceived quality-of-life is likely to vary based on individual age and socio-economic standing, and could be affected by local environment and the individual’s ability to adapt to a particular impairment. Compared to the DALY approach, the QALY approach could be said to be more comprehensive in prompting assessment of disease impact, yielding a more informed patient preference decision, but its use may prove less disease-specific and more location- and population- dependent.

3. Making the choice of metrics for disease burden calculations - DALY vs. QALY

3.1. Concerns about the DALY

Despite the widespread use of the DALY metric in public policy discussions, is it the right metric for low mortality conditions that occur uniquely in certain sub-populations of the world? There is no doubt that the DALY has become a ‘norm’ for ranking health states (Gold et al., 2002), and many policymakers and funding agencies unquestioningly accept the DALY values as accurate and valid for all ranked health conditions. As a result it is important to consider what the DALY actually can and cannot do.

The GBD programs have put extensive efforts into accurately estimating the number of people affected by the ranked health conditions, and to accurately attributing causes of death to specific diseases (Murray et al., 2012c). This work is highly commendable and affords the best estimates of disease prevalence on a global and regional scale. However, a good deal of controversy persists on aspects of the disability weight determinations used in the DALY calculations, so that the interpretation of the overall DALY values must be viewed with caution (King and Bertino, 2008).

Since the initial DALY formulations in the 1990s, the age-weighting and time discounting factors have been removed from the DALY scores in the GBD 2010 listings. Age-weighting was severely criticized for, in effect, valuing health interventions for 20–30 year olds over interventions for the very young and the older segments of the population (Anand and Hanson, 1997; Gold et al., 2002). The DALY framers countered that this, in fact, reflected societal preferences regarding the value of health in different age ranges; however, age-weighting is no longer used.

Critics also pointed out that the use of time-discounting minimized the long-term impact of diseases prevalent in childhood (e.g., helminthic infections) that have severe post-infection consequences much later in life (Anand and Hanson, 1997; Carabin et al., 2005).

The DALY framers arbitrarily assumed that years of life lost to any disease should reflect years lost from the longest average lifespan in the world (i.e., 86 years). Hidden in this formulation is the assumption that all years of life lost would be ones of perfect health, which clearly does not correspond with reality. In comparing DALY impact for different parasitic infections, malaria generates a significantly higher score than the ranked helminths (82.6 million total DALYs vs. 14.3 million DALYs, see Table 1) primarily because of malaria’s lethality in early childhood. Every child who dies of malaria is assumed to lose ~84 years of perfect health, even though he or she lives in conditions for which this is logically not possible. In practice, this DALY-based bias has resulted in an undue preference for public health funding for control of infectious diseases that are lethal in early childhood, e.g., HIV, TB, and malaria.

For non-lethal conditions, the DALY system ignores the reality of co-morbidities, although we know that co-morbidities become increasingly common as people age (Gold et al., 2002). For helminths, this factor becomes significant, because over 20% of children in endemic areas carry co-infections, yet their joint disease burden is not accounted for. In essence, under the DALY scheme, each person can only have one disease at any given time, and in the case of multiple disease prevalence, preference is given to the more ‘disabling’ condition (e.g., they are only counted as ‘malaria’ or ‘HIV’ and not also as ‘worm’ infection) (Murray et al., 2012b). The greatest DALY impact for an individual will always be his or her cause of death. The serial switching of imputed DALY causes over a lifespan tends to further discount the DALY impact of conditions such as helminthic infection that have a lifetime impact on health, but have few severe or lethal outcomes (Guyatt, 2000).

Estimation of disability weights in the DALY system has been based on comparison of health states based on ‘scenarios’ or vignettes presented to a survey sample of the general public (see section 2.1.2.2 for details). Creation of these vignettes has been the sole purview of the GBD core team, with consultation (but not final approval) by health experts (Taylor et al., 2013). The public’s perception of these (rather superficial) disease vignettes (see 2.1.2.2) is used to rank the 291 disease in the DALY system. The strength of this approach is that the evaluators are asked to provide their preferences in terms of importance for investment for disease prevention or control. The problem is that popular (or even experts’) understanding of a disease’s disabling impact is often limited or even biased (Kirigia, 1997, 1998; Taylor et al., 2013), or culturally pre-determined. As a result, the DALY disability weights do not accurately reflect the impact of a disease in a continuous scalar system. DALY proponents argue that ranking of conditions had been fairly consistent around the world (Gold et al., 2002; Salomon et al., 2012). However, where Dw values are low and affected individuals are many (e.g., 100s of millions for helminth infections), small inaccuracies in the disability weights result in large changes in the cumulative DALY scores for non-lethal diseases. Where ‘cost per DALY averted’ is the decision point for health care decision making, then inaccurate DALY estimates will have a perverse influence on prioritization of public health care investments.

An overarching issue about preference measurements is context. Contemplation of a disease or health state in the abstract, when one is healthy, may result in one set of rankings. However, when one is significantly ill, priorities often quickly change, such that consumer prioritization in the pre-illness setting may have nothing to do with actual demand in the clinical practice setting. Economic theorists generally have not included irrationality in their models, and time-trade-off exercises are unlikely to effectively model consumer preferences, particularly where information presented or consumer knowledge is incomplete (Gold et al., 2002). An adequate description of disease severity is essential for proper ranking of disease states (Hansson et al., 1994; Taylor et al., 2013). For many helminthic infections, a proper cataloguing of the milder, infection-attributable sub-clinical morbidities has not been done, and because of a lack of cohort studies, risk for longitudinal progression to more severe health states is unknown, even by experts (Carabin et al., 2005; Kirigia, 1997, 1998). This has led to the false impression that ‘the majority of helminth infections are asymptomatic’, which is not the case, but which often translates to a public indifference to their impact on community health and concomitantly low DALY scores.

3.2 Issues about the use of QALYs

The QALY, compared to the disease-focused DALY, is not disease specific, and so reflects a person’s overall health state from now into the future. QoL-based preference weighting for disability, performed in conjunction with a performance inventory, helps to make explicit the physical and non-physical impacts of any condition. This approach adopts the view that an impairment of HRQoL is a harm or an inherent disability when regarded from the perspective of the WHO’s definition of health as “a state of complete physical, mental and social well-being-- and not merely the absence of disease or infirmity.” (WHO, 1946) However, other groups’ definitions of ‘disability’ may vary significantly from this approach, and based on their own philosophical or legal frameworks, they may reject the QALY in favor of other health outcomes measures.

HRQoL may decline during life, independent of any particular diagnosis, and it may not be possible to define a ‘disease-attributable’ deficit due to a specific condition (Gold et al., 1996; Gold et al., 2002). Context and patient background can influence the HRQoL scores, and different QoL scoring instruments may yield different quantitative estimates of disease impact (Murray, 1996). Whereas the DALY attempts to assess a generic Dw that is valid in all locations, variations in QALY systems can result in large differences in health burden assessments for the same disease in different populations. Unlike the DALY, the QALY was developed in the context of evaluation of medical interventions in a given clinical setting. It may better capture the nuances of disease impact, but may not offer a generalized standard metric that would apply to all locations.

4. General concerns about the use of health-adjusted life years (HALYs)

The impetus for the development of time-based health metrics has been their use in cost-utility and CEA of health care interventions (Gold et al., 1996). The goal of CEA is to find means to maximize the ‘good’ for populations in a cost-efficient manner, in which expenditures can have the greatest impact. The approach is based on utilitarian philosophy, but the implementation can be flawed, leaving individuals or sub-populations significantly disadvantaged in the analysis (Gold et al., 2002).

The DALY system was developed as anti-social welfare, laissez-faire economics were ascendant. The DALY focus is on generic impact of specific diseases, independent of location, because that is ‘more fair’ (Murray, 1996). Impact of disease on earnings, physical performance, and schooling were specifically excluded from the DALY disability deliberations, and mild effects on cognition (leaving IQ > 85) were ignored. The influential aspects of local environment and social context were excluded (Anand and Hanson, 1997) as was the impact of endemic diseases on community-level factors.

Was this a scientific decision, convenience based, or doctrinaire or opinion-based? If the latter, then the DALY is a cultural construct that might apply to the developed economies, but not to the less-developed world. Exclusion of inconvenient ‘externalities’ thus prevents prioritization of those who are worst off (‘welfarism’) but leads to a serious underestimation of the impact of infection in the resource-limited world, and hence an undervaluation the impact of helminth control programs. In effect, the DALY overvalues diseases of the developed world, and undervalues diseases of developing areas. As Anand and Hanson (1997) point out, “It is clear that many non-health circumstances will also need to change for life expectancy to rise to the level used in the DALY calculations. These interventions would have to address the socio-economic determinants of health. They would include raising incomes, increasing female education, improving water supply and sanitation conditions, improving workplace safety, and reducing accidents and violence. Hence the burden that is measured by DALY s is the burden of disease and underdevelopment, and not that of disease alone.” [Authors’ emphasis]

Given their imperfections, are DALYs and QALYs appropriate for health burden assessment and for resource allocation? It is apparent that they are rough, but sometimes useful models of disease impact on human health. As such they serve a purpose in comparisons that are limited in scope, say disease estimation in a specific region, or assessment of the differential impact of a limited number of health care interventions, Neither really serves as a universal metric for policy planning on a global scale, so that any broad-scale ranking of diseases, such as presented by the GBD 2010 Program (Murray et al., 2012c; Salomon et al., 2012; Vos et al., 2012), must be viewed with a high degree of skepticism because of the inherent flaws of the metrics in ranking overall health impact (Anand and Hanson, 1997).

5. The Neglected Tropical Diseases movement

Since the promulgation of the initial DALY assessments in 1996 (Murray and Lopez, 1996), many workers who were active in control of diseases of developing countries, including helminth infections, have responded to the apparent undervaluation of these diseases in the DALY system. The Neglected Tropical Disease (NTD) initiatives began by drawing attention to the flaws of the DALY system, while aiming to mobilize increased efforts for NTD control. Beginning in 2002, the Global Fund to Fight AIDS, Tuberculosis and Malaria began to focus its massive resources for ‘infectious disease control’ only on three ‘big-DALY’ diseases, while infections caused by helminths, protozoa, and many tropical bacterial species did not receive any priority in international public health initiatives. This NTD neglect occurred despite the fact that NTDs were and are more prevalent and provide a significant share of the burden of disease both in terms of chronic disability and mortality in developing countries. The donor countries’ nearly exclusive interest in HIV, TB, and malaria framed the NTD problem-- their donation programs, which exclusively focused on these ‘big three’ infections helped to indicate (through the gaps in their results) the relative ‘neglect’ of many other diseases of underdevelopment (LaBeaud, 2008).

In part, the underestimation of the NTD impact on health arose from a lack of accurate information about their impact on health over a lifetime of exposure. Long-term longitudinal studies of NTDs are very few, with most data coming from cross-sectional studies of short term intervention trials. For example, skeptics in the GBD2010 program and elsewhere have often fallen into the trap of “…confus[ing] the measured effect of a deworming program on child growth with the effects of a lifetime hookworm infection on child growth. [They confuse] a one-year effect in one modern study with the effects of years of infection….” (McGuire and Coelho, 2011). With this false impression, their final ‘societal’ estimates of the health impact of NTDs (i.e., the DALY Dw) tend to be unrealistically low, with a resulting low priority for intervention based on aggregate DALY scores (Table 1).

The reductionist focus on individual diseases found in the DALY system obscures the joint impact of multiple helminth and protozoan infections on childhood health (Bustinduy et al., 2013; Ezeamama et al., 2005a; Ezeamama et al., 2005b). The 8–12 year follow-up of communities included in a long-term ‘deworming’ campaign begun in 1998 (Miguel and Kremer, 2004) has shown significant long-term effects in terms of improvements in cognitive function (Ozier, 2011) for children who were infants at the time the deworming program was initiated. A follow-up of adults in the same communities indicates that 2 or 3 extra years of treatment in childhood is associated in significant increases in the number of hours worked, a > 20% increase in wage earnings, and net increase in the number of meals consumed (Baird et al., 2011). These important ‘externalities’ of NTD control are missed if one relies solely on the DALY system.

These ‘societal’ improvements, not captured in the GBD system, had often been tacitly assumed by policy planners, but without firm evidence. The adverse valuations of the DALY system have caused them to move away from helminth control as a priority. Nevertheless, over the last 10 years, the stronger advocacy of workers familiar with the problems of helminthic infections and other NTDs has created number of national programs for widespread control (and even elimination) of selected NTDs (Molyneux et al., 2005; Molyneux et al., 2009; Rollinson et al., 2012; WHO, 2006, 2007). Despite what the health metric analysis may tell us, there is overwhelming evidence that investments in communicable disease control has a disproportional beneficial impact for developing countries (Gwatkin et al., 1999).

6. Remaining problems with health burden assessments for helminthic infections

Helminthic infections are mostly diseases of poverty in resource-poor areas, and consequently, resources have often been quite limited for their study. In areas where vital statistics are not collected, populations at risk can only be estimated. Our standard diagnostics for helminths (eggs counts) are insensitive (Colley et al., 2013; Knopp et al., 2008; Knopp et al., 2009), particularly for young children (Verani et al., 2011). This leads to serious misclassification bias in health outcomes assessments (Table 3), and suggests we should completely re-study and re-evaluate our assessment of the lifetime burden of helminthic diseases. Many experts confuse active infection (egg output) with disease prevalence, without considering the long-term impacts of infection. By strict definition, infection is not the same thing as disease, although many use the terms interchangeably. Teen-agers and adults who have grown up with helminth infections continue to carry disease from their childhood infections even though the parasites are gone. These individuals should be enumerated and included in the disease burden assessments (Giboda and Bergquist, 1999). With better data about the life history of diseases related to helminthic infections, including information on the economic spillover benefits of treatment, better programmatic decisions will be made.

Table 3.

Priorities for better disease burden assessment for helminth infections

Priority	Rationale
I. More accurate case counting	Infection status is frequently underestimated by standard parasitological testing
--Active infection	Active infection is the immediate cause of acute morbidity
--Previous infection	Chronic disease persists after active infection ends
II. Use of patient-based measurements of disease impact on disability status	Much of helminth disease research has focused on organ- specific pathologies, while not accurately defining the systemic and performance impacts of helminth-associated diseases. As a result, the disease scenarios of the DALY projects have provided quite minimal views of the impact of infection.
--Overall impact	e.g., Quality-of-life score provides a global assessment of disease impact in the context where it occurs.
--Nutritional status	Identified links between helminth infection, growth stunting and nutritional wasting
--Physical performance	Association between short stature, reduced physical fitness, and reduced earnings
--Cognitive level	An important long-term deficit associated with chronic disease. Accurate testing requires proper adaptation to local language and culture, and test reliability and consistency
--Scholastic achievement	School attendance and grades are usually not reliably standardized, and not good markers of performance. Better tools are needed to capture this important human capital metric
--Employment	Need careful measures of wage and non-wage activity and productivity in affected areas, before and after intervention
--Economic impact	Need measures of improvements at the individual, household, and community level.
III. Use of randomized trials with long-term follow-up to define the benefits of intervention	Intervention trials have to be of sufficient size (adequately powered) to detect small but significant changes in quantitative outcome measures. Studies should be of sufficiently long duration to detect full benefits of preventive chemotherapy during childhood. Analysis should account for concurrent community-level, household-level, and individual effects. Adjustment for secular trends may be required. Repeated studies will determine generalizability of findings.

The kinds of research that will ‘define’ the beneficial outcomes of helminth control should include studies of physical, scholastic, social, economic, and psychological impacts. Details of a proposed research agenda are listed in Table 3, which indicates some of the priorities for disease burden studies taken from the issues raised in this paper. Although these projects will require larger resources and will take longer time to perform, their high value in terms of evidence for policy planning will be worth the effort (Banerjee and Duflo, 2011). The remaining challenge is that poverty is not the same in all locations, and the ecology of helminth infections will vary depending on the parasite species, on local climate and environment, and on local governmental and societal factors, including culture and caste. Programs should be prepared to adapt to such local circumstances to get the most value out of helminth control.

7. Conclusion

Health metrics are a commonly used system of estimating the impact of health care interventions in policy planning. While they serve a useful role in CEA, they are flawed in their estimation of the impact of diseases on individual and community well-being. While many decision-makers have assumed that the DALY is an accurate measure of the impact of parasitic diseases on global health, the limitations of the DALY system mean that major aspects of helminthic disease are overlooked in most burden assessments. More work is needed for long-term longitudinal assessment of the outcomes of helminth control, including outcomes specifically relevant to ‘disability’ as viewed by social planners and health economists. Health planners are not obliged to accept current DALY or QALY estimates as face value. If there is sufficient reason to believe that the disability from a given disease is greater than that used by the GBD2010, one may choose to ‘re-calibrate’ the disability weight and use the published GBD2010 prevalence and incidence numbers to recalculate ‘revised’ DALYs for any suite of competing conditions under consideration for control.

Highlights.

• Health-adjusted life years are standard units for comparing disease burden

• The DALY and the QALY are the most frequently used in cost-effect analyses

• While sometimes useful, both the DALY and QALY metrics have significant limitations

• This is particularly true for ‘neglected tropical diseases’ (NTDs)

• Interim approaches for disease burden assessment are discussed

Abbreviations

QALY

Quality-Adjusted Life-Year

DALY

Disability-Adjusted Life-Year

GBD 2010

the Global Burden of Disease 2010 Project

Dw

disability weight

YLD

years lost to disability

YLL

years of life lost

NTD

neglected tropical diseases

VBD

vector-borne diseases

CEA

cost-effectiveness analysis

CBA

cost-benefit analysis