The search identified 4298 studies and covered publications from 1 March 1990 to 11 April 2022. After title and abstract and full-text screening, 281 publications were included for data extraction (Fig. 1). The most common exclusion criterion was the study omitting a clear explanation of methods used to estimate productivity losses (27%, 43/160). A complete list of the included studies and their characteristics can be found in Supplementary Material 3. Most studies were from Asia (45%, 125/281) and Africa (32%, 89/281) (Table 1). The number of eligible studies increased over time, with only one study from before 1991 [33], while 83% (232/281) of included studies were published after 2010. The majority were cost-of-illness studies (80%, 225/281), while the rest were economic evaluations (20%, 56/281).

Many studies clearly stated the use of the societal perspective (49%, 138/281). A further one-fifth of the studies were assumed to be from a societal perspective despite not explicitly stating so, given their inclusion of both provider (or payer) and patient (or household) perspectives (20%, 56/281). In addition, 10% (28/218) of studies used a patient perspective with another 12% (33/281) of studies taking the household perspective, including lost productivity for caregivers. Overall, 26 studies (9%) presented results using more than one perspective (e.g. societal and provider perspective).

All included studies satisfied the quality assessment criteria for study scope by clearly stating the research question, target population, study design, and where the data came from (Table 2). In total, 85% (239/281) of the included studies performed primary cost data collection, 10% (28/281) cited cost values from other studies and 5% (14/281) used a mix of primary data collection and citation of previous studies.

Some studies did not fulfil the quality assessment criteria for data analysis and results. For example, only 48% (134/281) stated the total time losses in the result section. Most studies (96%, 269/281) clearly specified the year for which the costs were reported. In addition, 94% (265/281) of studies explicitly mentioned the total cost of productivity losses in the results, and 72% (202/281) provided the mean or median cost value of productivity losses. There were 61 studies that considered interventions or costs over multiple years, of which 80% (49/61) reported applying a discount rate to the results. Moreover, 8 out of those 61 studies were economic evaluation studies; all of these stated that discounting was applied in the analysis.

Overall, only half (51%, 143/281) of studies performed at least one type of sensitivity analysis. In total, 39% of the cost-of-illness studies (88/225), and almost all economic evaluation studies (98%, 55/56) completed a sensitivity analysis to assess the uncertainty of parameters or models to the result. Among the economic evaluation studies, 11% (6/56) evaluated whether the use of a different study perspective and the inclusion of productivity losses impacted the incremental cost-effectiveness ratio. All six of those studies found that the inclusion of productivity losses resulted in a higher probability of the intervention being cost-effective [34,35,36,37,38,39]. We generally found that the quality of conduct and reporting increased over time for the data analysis criteria while decreasing or stagnated over time for the results criteria (Table 2).

Half of the studies estimated only the productivity losses of patients (141/281), while 13% (37/281) reported caregiver losses and 37% (103/281) reported both patient and caregiver losses (Table 3). The majority of papers reported productivity losses in adults (230/281) with a further 8% (21/281) of papers reporting productivity losses for retirees. All studies that included retirees reported their time losses owing to illness. Among the 30 studies (11%) focused on diseases in children, only 23% (7/30) reported the time losses of children, while the rest only reported the time losses of caregivers. No studies looked at productivity losses across all age groups.

The most reported type of productivity loss was absenteeism (67%, 187/281). Three studies (1%, 3/281) reported only presenteeism, 24 studies (9%) reported both absenteeism and presenteeism, and 15 studies (5%) reported only long-term disability (premature death or permanent disability). Only 16% (46/281) of studies included both temporary and long-term disability to consider the dual impact of the disease on morbidity and premature mortality.

Over half (58%, 164/281) of the studies did not explicitly mention what overall approach was used to estimate productivity losses (Table 3). With slightly more than one-third (39%, 109/281) of studies using the human capital approach, it was the most frequently used approach of those that were specified. Other methods less frequently employed were friction cost (1%, 3/281), willingness to pay (1%, 3/281), and the application of both human capital and friction cost approaches (0.7%, 2/281).

Table 4 reports methods used to assign a monetary value, including how these methods varied across the following populations: formal workers, informal workers, unpaid workers, school-aged children, and retirees. The methods used by each study to assign a monetary value can be found in Supplementary Material 3. Since some of the included studies used more than one method to assign a monetary value given the different population categories, there were more methods of assigning a monetary value than there were studies (N = 312). Only one cost of illness study completed a sensitivity analysis with more than one method for estimating the monetary value by applying minimum market wages, median market wages and GDP per capita [40].

While a range of methods were applied to assign a monetary value to productivity losses, half of the included studies used market wages (157/312). Of these, local market wages were applied in 4% of studies (13/312) as a way of accounting for different wage rates across provinces or districts within a country. Among those countries that used local market wages, 54% (7/13) were from Africa, while the rest were from Asia. Other methods used to assign a monetary value to productivity losses included self-reported income (28%, 86/312), macroeconomic measures (15%, 48/312), daily production value (2%, 6/312), willingness to pay for work replacement (1%, 3/312), proportion of the market wages for formal workers (3%, 9/312), conversion from annual tuition cost per student into a daily rate (1%, 2/312) and conversion from pension income into a daily rate (0.3%, 1/312).

Two approaches were used for self-reported income (28%, 86/312), including the estimated last income (27%, 84/312) and income changes (1%, 2/312). The self-reported last income asks what the payment amount was for the last period of income, while the self-reported income changes focused on the changes in cash earnings before and after the disease episode.

Macroeconomic measures included applying either the gross domestic product (GDP) or the gross national income (GNI) per capita per day to the reported time losses. While the GDP is a standard measure to capture a country’s economic activity during a particular year, the GNI is calculated by adding a country’s GDP to its residents’ income from abroad [41,42,43]. While GDP per capita per day was used in 11% (34/312) of studies, GNI per capita per day was applied in 4% (14/312) of studies.

In the six studies (2%) using daily production value, overall household income was estimated, including the monetary value of a previous harvest and sales of any products or services over a certain period of time [35, 43,44,45]. For example, for farmers, the amount of sales in a year or in a month was converted into a daily value in alignment with their time lost due to illness [46].

For the three studies (1%) that used the willingness to pay approach, the patient or caregiver provided the maximum daily amount of money that they would be willing to pay to hire a substitute worker to replace the sick household member [43, 47, 48].

Among 30 studies that included school-aged children, the majority did not value their time losses (77%, 23/30), even though they reported the total time losses due to illness. In those studies, only caregivers’ time was valued. The remaining studies divided the annual student tuition cost by the number of school days to calculate a daily value (7%, 2/30) or by using a proportion of the market wages of formal workers (16%, 5/30). The proportion of market wages of formal workers method was used in 16% (5/30) of studies that focused on children and 10% (2/21) of studies focused on retirees. In these studies, it was assumed that individuals received a specified percentage of national market wages (e.g. 50% of minimum wage) [40, 49]. Similarly, most studies that included retirees (76%, 16/21) reported their time losses but only valued caregivers’ time. For the studies which estimated the productivity losses for retirees, the monetary value was estimated through self-reported last income (10%, 2/21), a proportion of the market wages of formal workers (10%, 2/21), and by dividing the pension income into a daily value (4%, 1/21).

Different methods were used to assign a monetary value for formal and informal workers in 21 studies (7%, 21/281). One additional study assigned a monetary value for informal workers only (Table 4). While the self-reported income method was used in the majority of studies with formal workers (71%, 15/21) it was applied to only two of the studies with informal workers (9%). The market wage method was used in the remaining 29% of studies with formal workers (6/21), and most studies with informal workers also used this method (67%, 14/22). The rest of the studies including informal workers used either the daily production value (5%, 1/21), the willingness to pay for work replacement (5%, 1/21) or a proportion of the market wages of formal workers (9%, 2/22).