Discussion
This systematic review found that COPD poses large economic impact on countries across continents. We found that there were limited data on the costs of post-TB sequelae from the literature. There were only two studies on the cost of COPD-like symptoms on patients with a previous diagnosis of TB. These studies indicated a high economic impact after TB treatment. In urban Blantyre, Malawi, around two-thirds of patients sought healthcare during the year after TB treatment and one-fifth lost the work they had prior to TB, while some households faced financial difficulties and interruption of children’s schooling.42 In Thailand, patients with diagnosed TB accounted for 2.3% of the total economic burden of hospital admissions of COPD and 41% of bronchitis and emphysema.43
Yet, these studies did not clinically define post-TB disease but stated that patients had completed TB treatment for 1 year or that patients had a TB diagnosis. Some studies showed evidence that TB is associated with the presence of COPD,6–9 11 whereas some scholars argued that the pathophysiologic process in COPD and post-TB obstructive airway diseases are different.10 47 The features of COPD due to smoking can be different from COPD due to post-TB disease, and whether chronic airflow limitation due to other diseases such as TB should be defined as COPD needs further research.48 Although the cause of COPD, whether it is smoking or previous TB, might make a difference in terms of the symptoms experienced, it may not make a substantial difference in terms of resources spent. Therefore, as there is no better proxy at the moment, the cost of COPD (even if caused by smoking) will have to be the most appropriate estimate of COPD caused by TB. Overall, TB adds substantial risk and burden to chronic lung and respiratory diseases.11 49–51 The lifetime burden of post-TB disease was estimated to be 58 million DALYs (95% uncertainty interval, 38–83), globally accounting for 47% of the overall burden caused by TB.12 More work should be conducted to investigate the economic impact of post-TB or the lifetime costs of TB.
There were also a limited number of cost analysis studies conducted in low-income countries and in Africa (three and five studies, respectively), which bear a large proportion of the disease burden of TB12 and chronic respiratory disease.19 This suggests that more efforts and priorities should be put to collect or generate quality data in resource-limited countries to inform local governments and international organisations to take action to reduce the global disease burden of COPD.
Similar to a previous systematic review that comprehensively searched the literature published between 2003 and 2013,24 variation in country of study, data source, perspective of cost, disease diagnosis, disease severity, comorbidity and patient characteristics generally exists and makes it challenging to compare costs across different studies and countries. However, both reviews indicate COPD contributed a considerable burden to society.
The costs of COPD in LMICs seemed to be lower than in HICs when comparing to the evidence from HICs collected by other reviews.25 27 28 However, the cost range is large in both LMICs and HICs, and it could be due to differences in labour costs, the costs of medicines and other medical supplies and the healthcare services provided. For example, our review found that hospitalisation costs of COPD ranged from $59 per patient in Vietnam52 to $3240 per year per patient in Mexico with the highest annual costs—$5775 per patient—occurring among patients with severe disease53; the costs of AECOPD ranged from $46 per exacerbation54 to $5732 per year per patient55 while reaching $15 442 per patient among those with non-invasive positive-pressure ventilation failure.56 Anees ur Rehman et al28 found that the hospitalisation costs of COPD ranged from $92557 to $692958 in Europe and $7242 in the USA with costs due to severe exacerbation reaching $20 757.59
We also found lower costs in lower MICs in comparison to upper MICs. Direct medical costs of COPD ranged from $26 per hospitalisation in India60 to $1720 per patient in private facilities in Kenya,61 while among upper MICs, the minimum direct medical cost was $428 per year per patient in Thailand62 and the maximum was $2694 per year per patient with severe disease in Mexico.53 Similarly, the range of direct medical costs of AECOPD in lower MICs was $631–$3232 per patient with both costs occurring in Iran,63 while in upper MICs, it ranged $137–$4207 per exacerbation, both estimates from studies conducted in Turkey.54 This finding is also supported by the subgroup analysis, but there were high heterogeneities across populations and countries.
An advantage of the present systematic review compared with previous ones is that we explored the possibility of performing meta-analyses while taking into account heterogeneity across studies and subgroups. Heterogeneity is high within and between subgroups, possibly due to differences in study setting, disease severity and comorbidity between studies, and labour costs between countries. Meta-regression was not performed because cost data were sparsely distributed across disease and participant characteristics, resulting in limited data to explore the sources of heterogeneity using meta-regression. However, we conducted subgroup analyses to examine differences in costs between countries, regions, income groups and cost units (such as costs per event or per patient).
Demographic factors, such as sex and age, and health factors, such as comorbidities and disease severity, might have an impact on the costs (online supplemental tables S2.8–S2.11). Costs were higher among women, the elderly, patients with comorbidities and those with more severe disease. Future research should further investigate the factors that contribute to the cost burden.
There were no data on the cost of invasive or non-invasive mechanical ventilation, but there were studies reporting costs on oxygen64 or home oxygen therapy.22 Home oxygen costs ranged from $91 in Brazil to $168 in Russia. There were limited data on indirect costs from cost analysis studies.
On the other hand, high economic burden attributable to air pollution and smoking was documented for several countries. Direct medical costs of COPD due to smoking ranged from 15.5 million in Honduras (0.86% of HE) to 4004.2 million in Brazil (2.9% of HE),65 and the economic burden of premature mortality attributable to PM₂.₅ pollution was $717.3 million (correspondence with authors) in Nigeria (where the GDP per capita was $3204 and the population was 184 million)66 and $6903.9 million in Thailand (where the GDP per capita was $6487 and the population was 70.6 million).67 The estimates could differ due to author’s different methods of estimation.
One strength of our study is the breadth of literature that we examined. The review covers publications over 9 years and across LMICs without language restrictions. However, both the quality of the studies and the heterogeneity between the studies warrant caution at the time of interpretation.
The present review also faced some limitations. First, although this review covers the literature from 2013 up to 2022, the data may already be outdated at the time of publication. Future research should update the literature search and data accordingly. Second, our data were not extracted in duplicate due to time constraints, but instead we chose an approach of validation with a peer. Third, while our quality assessment was performed as consistently as possible, there is no conventional standard on which to judge all studies, and, therefore, we chose an assessment of the presence and absence of the characteristics mentioned in the JBI checklist, CHEC, GHCC and CHEERS. Moreover, the data presented challenges at the time of synthesis, as there were inconsistencies in the summary statistics that were presented; nonetheless, the package we used was able to approximate means and SDs when there was a measure of central tendency (mean or median), spread (IQR or range) and the sample size.
Meta-analyses were focused on hospitalisations, medications and outpatient visits, but the data might support a wealth of other subgroup analyses of interest. Heterogeneity is potentially biased upward68 due to the small number of studies in the subgroup analysis and variation in disease and participant characteristics across studies.
Meta-regression was not performed due to the sparse distribution of cost data across disease and patient characteristics and the limited data available on factors that may influence costs. Costs can vary widely between health systems and countries. Costs are influenced by several factors including the economic development of the country, the functioning of the health system, the income level of patients, the availability and coverage of health insurance, the severity of the disease, the presence of comorbidities, the perspective of the study and, therefore, the types of costs included, and other relevant factors. Apart from the level of economic development of the countries, data on most of the other factors are very limited, as presented in the CD-CPTB database. This makes it challenging to conduct and interpret meta-regression to further explore sources of heterogeneity.
Finally, it was often hard to categorise the costs per patient or per event (hospitalisation, outpatient visit or exacerbation), but with chronic conditions, a more appropriate appraisal of the costs would be all the hospitalisations and doctor’s visits over a certain period of time.
