Estimation of economic burden of COVID-19 using Disability-Adjusted Life Years (DALYs) and Productivity Losses in Kerala state, India

Background: The COVID-19 pandemic has had a huge impact on the global economy and stressed the health care systems worldwide. Measuring the burden of disease on health and economy is essential for system preparedness by way of allocation of funds and human resources. Methods: The present study estimates Disability-Adjusted Life Years (DALYs), Years of Potential Productive Life Lost (YPPLL) and Cost of Productivity Lost (CPL) due to premature mortality and absenteeism, secondary to COVID-19 in Kerala state, India. The impact of disease on various age-gender cohorts has been analyzed. Sensitivity Analysis has been conducted by adjusting six variables with a total of 21 scenarios. Results: Severity of infection and mortality were higher among older sub-group of patients, and male were more susceptible than female in most of the age groups. DALY for the baseline scenario was 15,924.24 and 8,669.32 for males and females respectively. The CPL due to premature mortality was 26,80,36,179 and 42,510,946 for males and females respectively. Conclusion: People aged more than 50 were disproportionately affected by the disease, with presence of comorbidities further raising vulnerability.


INTRODUCTION
Since the first case of COVID-19 was reported in December 2019, COVID-19 has spread across 218 countries infecting 58,099,381 people with 1,380,859 deaths and 40,263,305 recoveries worldwide as on November 21, 2020 [1]. The majority of cases have been reported from Low-and Middle-Income Countries (LMICs) with a significant proportion of people not having access to quality health facilities, being placed at a higher risk [2]. India ranks second with 9,065,301 and 8,491,462 recoveries as on November 21, 2020 [1]. Higher infection rates of the disease has increased the burden on healthcare systems and thus increased the fatality rate [3]. The major impact globally due to COVID-19 has been through attributed mortality. Estimating the mortality due to COVID-19 helps to understand the dynamics of the pandemic. Role of socio-demographic, social determinants and geography are important to assess the differential risk levels to the disease with age, gender and geography [4].
Governments have imposed strict measures to curtail the morbidity and mortality caused by .Individual level measures such as self-isolation and social distancing and population level lockdowns are widely adopted to limit spread [2,5,6]. Quantifying the health and economic impact of COVID-19 would reflect on the consequences of the policy decisions taken [7].
In India, the first case of COVID-19 was reported in Kerala. Timely interventions such as contact tracing, testing, quarantine, isolation and treatment had reduced the spread of the disease in Kerala [8]. The number of cases in Kerala has now reached 551,669 with 1997 deaths and 481,718 recoveries as on November 21, 2020. The present study estimates the economic burden and productivity loss using Disability-Adjusted Life Year (DALY), Years of Potential Productive Life Lost (YPPLL) and Cost of Productivity Lost (CPL) [9,10] and the effect of age and gender [4,11] for the state of Kerala

Data Collection
Publicly available data from various sources have been used to gather sociodemographic details, information about incidence, death due to COVID-19, information on quarantine, per capita income, etc., for the state of Kerala [9,12,13]. The first reported case of COVID-19 in Kerala dates back to January 30, 2020, from when on the estimates for the present study have been collected [9,13].. The 5-year age-gender population of Kerala, working population in each age-gender cohort and the corresponding life expectancies were obtained from 2011 Census of India [12]. Incidence data documented by Team Collective for Open Data Distribution-Keralam (CODD-K) was till August 20, 2020 was used to classify the incidence of age-gender cohort. Out of 52,199 reported cases till August 20, 2020, data of 7645 patients were excluded as they did not contain information on age and gender [13]. Recovery time documented by the team for 1,012 patients in Kerala was used from CODD-K [13].
Data for number of deaths were extracted from the Government of Kerala dash board and CODD-K [9,13]. The infections were categorized as mild, moderate and severe/ critical [9,14]. Table I presents the information on various parameters used for the study. Sensitivity Analysis (SA)* Number of cases 10% (S12), 20% (S13) and 30% (S14) increase in infected cases [3,13] 20-Aug-20 Actual 52199 Valid records 44554 Number of deaths 10% (S18), 20% (S19) and 30% (S20) increase in deaths [3,9,13] 15-Nov-20 Actual 1869 Valid records 1841 Quarantine 10% (S15), 20% (S16) and 30% (S17) increase in mild cases [9,13] [3,7,25]. DALYs take into account the disability caused by the disease (YLD) and the premature mortality (YLL) [7]. Determining YLLs and YLDs and DALYs would allow us to measure the shortfall of the deceased and life years lost. YLLs explain the loss incurred as a result of death due to an event by comparing with the years that they would have lived otherwise. YLLs become high in case of either higher mortality or mortality of younger people or both [4]. Years of Potential Productive Life Lost (YPPLL) and Cost of Productivity Lost (CPL) are widely adopted measures that majorly explain the economic burden due to an event [11].
DALYs were calculated using an incidence based approach [3,18]. DALY estimates were obtained for age-gender split to identify the more vulnerable groups [7].Though incidence based approaches do not consider the severity of diseases, owing to the diverse impact of COVID-19, severity was considered to assign the Disability Weights (DW) [3].
where r = discount rate; D = disability duration (years); I = number of incident cases where L = life expectancy at age of death (years); N = number of deaths.
Most of the Burden of Disease (BoD) studies do not consider multimorbidity, which might produce inaccurate estimates [25]. Three methods for calculation of Combined Disability Weights (CDW) for multimorbidity as reported by Hilderlink et al (2016) viz. additive, multiplicative and maximum limit methods, were employed in this study [25].
where 'i' and 'j' indicate the DWs of 'i' th and 'j' th disabilities.

Productivity Losses (YPPLL and CPL)
YPPLL defines the number of productive years an average person would have lived otherwise. Working population proportion of each cohort was multiplied with the YPPLL to estimate the CPL lost due to morbidity and absenteeism. Recovery days for the severe cases were extended by 8 days to account for the ICU stay. Productivity losses were estimated using the Human Capital Approach considering the absenteeism and premature mortality for temporary and permanent losses respectively [7,28].
For calculation of productivity losses, people from age groups 15 to 60 were chosen considering the employment age [29] and retirement age of Kerala [22]. [18,30].
where 'i' represents 'n' age-gender cohorts; D i = deaths at age; w i = productive years remaining at age  (8) and (9).
where S = average salary per day considering the number of paid working days per week as six; L j = average recovery time; N = Number of incident cases; P = proportion of working population, in cohort 'j'. For computation of productivity losses, the proportion of working population was considered along with an extended disability period for severe cases to account for the ICU stay [20,21].

Sensitivity Analysis (SA)
A spectrum of scenarios (table 5) was considered to analyze the effect of each parameter on the DALY estimates. The key idea of performing SA is to assist policymakers to anticipate the effects brought in by each of the driving variables. As most of the deceased cases had underlying health conditions that reduce the life expectancy, Scenario 1 (S1) and Scenario 2 (S2) have been developed [4,33]. Increasing the number of mild cases, overall cases and deaths are conservative analysis to help healthcare fraternity and policymakers [3].

RESULTS
From the distribution of cases and deaths (figure S1), it is clear that the older adults are disproportionately vulnerable to be severely affected by the disease.   Figure S2 presents the DALYs calculated for the nine different scenarios (refer To measure the impact of the disease on productivity of the state, YPPLL and CPL were calculated considering the productive age group to be 15-60 years. It is to be noted that the age groups 25-49 comprise of more working population, which makes them higher contributors of productivity.
Fortunately, there is a relatively lesser impact of the infection in terms of prolonged illenss and mortality for these age groups which has reduced the losses to some extent. Recovery days mentioned in table S III are exclusive of the ICU stay of severe cases.    S2 S3 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 19, 2020 [9,34]. Considering the effect of asymptomatic cases that act as latent spreaders , DWs due to presenteeism, and unpaid work could increase the losses [35]. These facts are concerning as the actual DALYs might be way higher than those estimated using official figures.

Change in DALYs for all scenarios
In addition, the productivity losses in terms of YPPLL and CPL for mortality and absenteeism for the productive population (15 to 60 years of age) were estimated. Multimorbidity was taken into account for the calculation of CDW. CPL values (for premature mortality) for younger population less than 25 years of age just 1.97% and 2.37% for male and female respectively. This is because of the lesser proportion of people working in this age group and also lesser impact of the disease in terms of mortality. Considering the CPL due to absenteeism, the values for younger people less than 25 years of age are 8.16% and 8.83% for male and female respectively. This is higher compared to that of the deaths because of relatively lower proportion of people dying due to infection.

What is already known on this topic
Several researchers have been conducting researches to estimate the economic burden and productivity losses of various diseases around the globe such as the estimation of YLLs due to COVID-19 in the US [4], India [26] & Swiss [5], DALYs due to COVID-19 in Korea [3]& Italy [7], YPPLL due to Cancer in Brazil, Russia, India, China, and South Africa (BRICS) [36], YPPLL due to five leading causes of deaths in Iran [11], productivity loss due to Cardiovascular disease and mental illness in India [35], etc. and 121449 DALYs, productivity losses of EUR 300 million and EUR 100 million due to premature mortality and absenteeism respectively, in Italy [7].

What this study adds
As evident from this study, the disease has impacted the older population to a greater extent. or more [27].
SA has been vastly adopted by the researchers to depict the influence of one or more variables on the outcome(s) [3,4]. A total of 21 scenarios by adjusting six variables were analyzed in the study. The increase in number of deaths highly increased the DALYs whereas the reduction in life expectancy reduced the DALYs. Reducing life expectancy could be related to practical findings of researchers.

Limitations of this study
Exclusion of incidence of cases post August 20, 2020 for estimation of DALYs due to unavailability of open data has led to a certain underestimation. Also, psychological impacts of the mitigation strategies are a potential risk that could increase mortality and are not in the scope of present study [5]. Most of the policies have not considered the mental illness and allied problems [5,38,39]. Though YLDs contribute a minor proportion in DALY, including factors such as unpaid work and presenteeism might improve the accuracy. In Kerala, about 30 and 10 percent of recovered patients have experienced post recovery illness and long-term effects, which have not been considered in the study [40].

CONCLUSION
The present study aimed at analyzing the economic burden and productivity losses using common estimates such as DALYs (YLLs and YLDs), YPPLL and CPL (for mortality and absenteeism) for Kerala, India. Public domain data from various resources were merged to form the age-gender cohort data required to calculate the above-mentioned. The study could be integrated with simulation models to project the economic burden and productivity loss using the estimates of simulation [41].

DATA AVAILABILITY STATEMENTS
All data are incorporated into the article and are openly available from the references mentioned.