Modelling global, regional and national prevalence of asthma: projections from 2018 to 2040

Abstract

INTRODUCTION: Asthma is a common chronic condition that results in considerable morbidity, healthcare utilisation and economic burden, both nationally and globally. While there has been encouraging progress made in reducing asthma mortality in some countries, there is still work to do in several countries where limited progress has been made. To inform policy deliberations for sustainable and effective global and national asthma control programmes, it is crucial to have a robust evidence base for future trends of the global, regional, and national prevalence of asthma. However, there is a lack of such important information in the published literature. The aims of this PhD were to: (i) systematically identify, describe and critique existing models for estimating and/or projecting the global, regional and national prevalence and disease burden of asthma; (ii) develop a critical appraisal checklist for assessing the quality of models for estimating and projecting prevalence and burden of asthma; and (iii) generate projections of global, regional and national prevalence of asthma from 2018 to 2040.

METHODS: This PhD was conducted in three consecutive phases. In Phase-I, I undertook a systematic review of models for estimating and projecting global, regional, and national prevalence and disease burden of asthma. I searched Medline, Embase, World Health Organization Library and Information Services (WHOLIS) and Web of Science databases from 1980 to 2017 for modelling studies on prevalence and burden of asthma. Data were descriptively and narratively synthesised. The identified models were appraised critically in relation to their strengths, limitations and reproducibility. In Phase-II, I developed a critical appraisal checklist for assessing the quality of models for estimating and projecting prevalence and burden of asthma by reviewing the existing critical appraisal checklists, risk of bias tools, reporting guidelines and other guidelines for good practice in modelling studies and consulting with a panel of experts in the field of asthma and disease modelling. Then, I applied this critical appraisal checklist to the models identified through the systematic review. Based on the findings, I determined the best quality models for projecting the prevalence and disease burden of asthma. In the final phase (Phase-III) of this PhD, I conducted a modelling study to generate projections of global, regional and national prevalence of asthma from 2018 to 2040. Drawing on the learning from Phase-II, I developed dynamic models by using regression models with Auto Regressive Integrated Moving Average (ARIMA) errors to generate the projections of the prevalence of asthma. Input data for this modelling study were extracted from the Global Burden of Disease (GBD) Study 2017 (GBD 2017) because GBD was the only comprehensive source of estimates of asthma prevalence.

RESULTS: In Phase-I, I identified 108 eligible studies. These studies employed a total of 51 unique models. Logistic and linear regression models were mostly used in national estimates and projections. Bayesian meta-regression models such as DisMod-MR and Cause Of Death Ensemble models (CODEm) were most commonly used in international estimates. Most models for prevalence and burden of asthma suffered from several methodological limitations – in particular, suboptimal reporting, poor quality and lack of reproducibility.

The critical appraisal checklist that I developed in Phase-II included the following quality criteria for models: (i) statement of objectives and scope, (ii) model structure, (iii) model assumptions, (iv) underlying theory, (v) model appropriateness, (vi) description of input data, (vii) data representativeness, (viii) outcome measure, (ix) model fitting and parameter estimation, (x) quantification of uncertainty, (xi) goodness of fit, (xii) model performance, (xiii) model presentation, (xiv) user manual, and (xv) replication and usability. Application of this critical appraisal checklist to the models that I identified through the systematic review undertaken in Phase-I determined regression models with ARIMA error as the best quality models for projecting prevalence and burden of asthma.

The modelling study in Phase-III projected that there would be an expected 337.9 (95% CI: 336.8-339.0) million people with clinician-diagnosed current asthma in the world in 2040. The global prevalence of clinician-diagnosed current asthma was forecasted to be 3.7% in 2040, which is a 0.03% decrease from 2017. However, the global asthma cases were projected to increase by 65.0 million (24.0%) during this period due to population growth. The global prevalence of clinician-diagnosed current asthma in women (185.4 million; 95% CI: 184.6-186.2) was projected to be higher than those in men (152.5 million; 95% CI: 151.7-153.3) in 2040. The regional prevalence of clinician-diagnosed current asthma for the year 2040 was forecasted to be highest in Polynesia (11.5%; 95% CI: 9.3 to 13.7); followed by Australasia (10.5%; 95% CI: 10.2-10.8), Micronesia (9.0%; 95% CI: 6.8-11.2) and Northern Europe (7.6%; 95% CI: 7.4-7.8); and lowest in Southern Africa region (1.6%; 95% CI: 1.5-1.7). Among 22 Sustainable Development Goals (SDG) regions, the prevalence of clinician-diagnosed current asthma was forecasted to decrease in around two-thirds of regions (n=15; 68.2%) from 2017 to 2040. There was substantial variation in the national projections of the number of people with clinician-diagnosed current asthma ranging from 4,800.0 (95% CI: 640.0-8,960.0) in Andorra to 36.1 (95% CI: 35.7-36.4) million in India in 2040. The highest national prevalence of clinician-diagnosed current asthma for the year 2040 was forecasted to be in Tonga (14.1%; 95% CI: 8.1-20.2); followed by Vanuatu (12.5%; 95% CI: 9.5-15.5) and Australia (11.2%; 95% CI: 10.9-11.6); and the lowest in South Africa (1.4%; 95% CI: 1.3-1.5). The trends in the number of people with clinician-diagnosed current asthma from 2017 to 2040 were projected to increase in 113 countries (62.1%).

CONCLUSIONS: Various models have been used to estimate and project the prevalence and burden of asthma, but almost all suffer from methodological limitations - in particular, suboptimal reporting and lack of reproducibility. There is a need for reporting sufficient details of models and making data and code available to ensure their reproducibility. The critical appraisal checklist that I developed contains essential quality criteria that a good model should possess, ranging from model structure to reproducibility. It will enable model users to assess the quality of a particular model of interest. Moreover, this checklist will be able to serve as the foundations for a future quality appraisal tool for the systematic reviews of modelling studies on prevalence and burden of diseases. The modelling phase projected that global asthma prevalence will decrease over the coming two decades. However, the global number of asthma cases is projected to increase due to population growth. Moreover, prevalence of asthma is projected to increase in two-thirds of countries. To my knowledge, this is the first study to generate comprehensive and reproducible projections of the global, regional and national prevalence of asthma. These projections of asthma prevalence can be useful inputs for governments, national and international funders, and inter-governmental organisations to make informed decision on future asthma policy worldwide.