Ill-health, work, and productivity are interrelated. The pro-longed ill-health due to chronic diseases has a higher chance of premature mortality [1], increasing the chance of disability [2], higher use of medical services and exerts greater economic burden to household and nation. At the households level, economic burden can be both direct and indirect [3]. The high out-of-pocket spending, catastrophic health spending and impoverishment are direct consequences of increasing chronic diseases [4]. Indirect burden of chronic diseases includes work absenteeism, voluntary retirement from work [5], and reduced propensity to work [6]. The cascading effect of ill-health reduces individual income [7] and may lead to poor physical and mental health [8] and may lead to gradual loss of productivity and welfare.
Productivity loss reduces the income and well-being of individuals and households. Ill-health often reduces the work participation as it affects the prime working age group. Productive time forgone due to ill-health cost both, to the household and the nation as well. Productivity loss is measured using multiple indicators; work absenteeism, presenteeism, permanent withdrawal from the workforce, and job interruption [9]. While work absenteeism refers to absence due to illness, presenteeism is low work performance during sickness [10]. Permanent withdrawal from the workforce includes voluntary retirement due to impairment or other health problems. Work-related injuries or accidents and success and failure also add to productivity loss [11].
Most of the studies on the consequences of chronic diseases on work productivity were carried out in developed countries [12,13,14]. People with poor health are more likely to spend a considerable time in seeking healthcare and that may lead to work absenteeism [15]. Among respondents who experienced symptoms related to health conditions in Germany, the average number of workdays lost due to absenteeism and presenteeism was 27 days per respondent annually [16]. Results from a study in Australia shows that the full-time workers with mental disorders lost an average of one day due to absenteeism and three days due to presenteeism in one month reference period [17]. In USA, the weekly absenteeism costs US$1685/employee per year and about 71% of the total productivity loss was contributed by reduced performance at work [18]. Asthma, cancer, heart disease, and respiratory disorders were estimated to have presenteeism costs of more than US$200 per person annually in USA [19]. Presenteeism represents the largest component and leading driver to the medical costs, specifically among the patients with migraine/headache, allergies, and arthritis [20]. Depression ranked third among health conditions with an annual productivity loss of US$878 per person [21]. A higher number of health risks is associated with lower on-the-job productivity [22]. Adults with multiple chronic diseases may have high chance of reduced productivity [23] In India, nearly a quarter of the companies lose approximately 14% of the total working days annually due to sickness [24].
Older adults in India are vulnerable to chronic diseases and, that may affect their work temporary or permanently [25]. The country has achieved the replacement level of fertility and nearing completion of demographic transition, resulting increasing share of older adults and elderly in the country and increasing burden of non-communicable disease (NCD). The share of middle aged and elderly population (45+) has increased from 18.9% to 2001 to 25.1% by 2020 [26]. The median age of onset of NCDs was also declining from 57 years in 2004 to 53 years by 2018 [27]. Though large number of studies estimated the OOP and catastrophic health spending, socio-economic inequality and determinant of OOPS and CHE [28], there is no nationally representative studies on productivity loss due to health problems. Present study explores the pattern and prevalence of limiting paid work and productivity loss among middle-aged and elderly in India and their association with chronic diseases. Figure 1 presents a schematic presentation of productivity loss. It depicts the pathways how economic burden of ill-health lead to loss of income and welfare through various medical and non-medical components. The non-medical component includes absenteeism, presenteeism and job-interruption.
Data and methods
Data
The study utilizes data from the first wave of Longitudinal Ageing Study in India (LASI), collected during April 2017 to December 2018. The survey was conducted by International Institute for Population Sciences (IIPS) in collaboration with Harvard T.H. Chan School of Public Health (HSPH), University of Southern California (USC) and other national institutions. Using multistage sampling method, a total of 42,949 households and 72,250 individuals aged 45 years and older and their spouses were successfully interviewed. Among these individuals, a total of 3,213 ever stopped working for a year or more due to health problem and 6,300 had limiting paid work. The data is publicly available for all states except Sikkim at the time of drafting this paper. The household and individual response rate was 95.8% and 87.3% respectively. Detailed about the survey and the findings are available in national report [29].
Variable description
Outcome variables
In LASI survey, a detailed module on ever work, current work, stopped work and limiting paid work due to health issues were collected. The questions on stopped work begins with “have you ever stopped working for one year or more at a time due to reasons of family, health, education, economic recession, natural disasters, etc.?” and the question on limiting work reads as “Do you have any impairment or health problem that limits the kind or amount of paid work you can do?”. We used ever stopped work (1 = yes, 0 = no) for one year or more due to health problem and whether health problem had limit the paid work (1 = yes, 0 = no) as two outcome variables.
Covariates
We have used a set of demographic, economic, behavioural and health covariates in the analyses. These includes age (45–54, 55–64, 65–74, 75+), sex (male/female), educational attainment (illiterate, less than 5 years, 5–9 years completed, 10 years or more), monthly per capita expenditure quintile (MPCE), place of residence (rural/urban), caste (scheduled caste, scheduled tribe, other backward classes, others), religion (Hindu, Muslim, Christian, others), marital status (currently married, widowed, others) and regions (north, central, east, northeast, west, south) were used as the predictors in this study. The MPCE was used to depict the living standard of the household. In addition, the number of chronic diseases (hypertension, diabetes, chronic lung disease, chronic heart diseases, stroke, arthritis, neurological or psychiatric problems), health insurance coverage (yes/no), practicing exercise (yes/rarely/never) and smoking tobacco (yes/no) are included to examine their association with the limiting paid work or ever stopping work for one year or more among older adults.
Treatment variable for PSM
In LASI, respondents were asked if they were diagnosed with chronic disease such as hypertension, diabetes, cancer, chronic lung disease, chronic heart disease, stroke, arthritis, and neurological problem. The individuals who had reported being diagnosed with any chronic diseases (1 = yes, 0 = no) have been considered as treatment group and those not being reported any of the chronic diseases have been treated as control group in the study. The treatment and control group did not overlap as they were mutually exclusive in nature.
Statistical analysis
Descriptive statistics, age-sex adjusted estimates, propensity score matching and logistic regression model were used in the analysis.
Prevalence of ever stopped work and limiting paid work
We estimated age-sex adjusted prevalence of ever stopped working and limiting paid work using the nationally representative full sample age-sex composition as reference using logistic regression.
Propensity score matching analysis
The propensity score matching (PSM) considers the potential selectivity in the sample. PSM is a statistical technique that estimates the effect of an intervention or a treatment by adjusting for covariates that predicts the results of receiving the treatment [30]. The advantage of using PSM model is that it compares the treated and controlled group on the basis of similar observed characteristics [31, 32]. The PSM has been used for evaluating various programme in a number of research studies [31,32,33,34]. For determining the average treatment effect (i.e., the effect of having any chronic disease), a counterfactual model is estimated.
Propensity score
The PSM is the probability of the middle aged and elderly population who had chronic diseases with certain characteristics, may be written as,
$$\mathrm P(\mathrm X)\:=\:\Pr\;(\mathrm D\:=\:1\vert\;\mathrm X)$$
(1)
Where, D = 1 if the population had any chronic diseases D = 0, otherwise.
And X is the vector of all the covariates used in the model.
Generally, PSM model estimated three probabilities, such as, Average Treatment Effect on the Treated (ATT), Average Treatment Effect on the Untreated (ATU) and Average Treatment Effect (ATE).
ATE is the average treatment effect of the intervention variable on the outcome variable and can be explained by using following equation
$$\mathrm{ATE}\;=\;\mathrm E\;(\mathrm\delta)\;=\;\mathrm E\;({\mathrm Y}_1-{\mathrm Y}_0)$$
(2)
where E (.) means average and Y1 represents potential outcome for those having any chronic disease and Y0 represents potential outcome for the population having no chronic diseases.
With the help of counterfactual model, the ATT can be written as
$$\mathrm{ATT}\;=\;\mathrm E\;({\mathrm Y}_1/\mathrm D=1)-\mathrm E\;({\mathrm Y}_0/\mathrm D=1)$$
(3)
The counterfactual model is the potential outcome that would have been obtained in case of not having any chronic disease and vice versa.
Where, E (Y1/D = 1) is stopping work who have any chronic disease.
E (Y0/D = 1) is the expected outcome for the individuals having any chronic disease if they would not have any of the diseases.
Similarly, the average treatment effect on the untreated (ATU) is defined as:
$$\mathrm{ATU}\;=\;\mathrm E\;({\mathrm Y}_1/\mathrm D=\;0)\;-\;\mathrm E\;({\mathrm Y}_0/\mathrm D=0)$$
(4)
Where E (Y1/D = 0) is the expected outcome if the individuals without any chronic disease were to have any chronic disease.
E (Y0/D = 0) is the counterfactual model predicts the outcome for the individuals who would have had any chronic disease but earlier they had not any.
The average treatment effect (ATE) is the difference between the expected outcome for those with any chronic disease and those without any chronic disease.
We used psmatch2 command in the STATA 16 which provides all the estimates using Mahalanobis matching technique.
Logistic regression
We used the multivariate logistic regression as a robustness check in support to our PSM model. We used three different models to understand the impact of each covariate on ever stopping work and limiting paid work separately. In the Model 1, we adjusted only for the number of chronic diseases. In model 2, socio-demographic variables were considered (age, sex residence, caste, religion, marital status and region). Finally, the socioeconomic variables along with smoking/substance abuse, exercise, health insurance and other predictors were adjusted in Model 3 to assess the adjusted effect of all the covariates on ever stopping work for one year or more. The following regression equation has been used.
$$\mathrm{Logit}\;({\mathrm Y}_{\mathrm i})\:=\:\ln(\mathrm p/1-\mathrm p)\;=\;\mathrm\alpha\:+{\:{\mathrm\beta}_{\mathrm i}\mathrm X}_{\mathrm i}$$
Where, Y is the probability of outcome event of the ith individual. The model estimates the log odds of ever stopped work and limiting paid work adjusted for a set of explanatory variables (Xi).
STATA version 16 was used for cleaning, standardizing data (to adjusted form), and for analysing data. Independent variables included individual level variables.
