While there are several methods for modeling in time series forecasting, including SARIMA and exponential smoothing methods, SARIMA models have shown better performance in predicting road traffic injuries in relation to gender [19]. Advantages of the SARIMA model include autoregressive, moving average and seasonal functions for trend, auto-correlation, smoothing and season [34]. Previously acknowledged disadvantages of the SARIMA model include the requirement of longitudinal data with a large sample size [19]. The number of data points necessary to develop a time series model has been determined to be at least 52 [38] or 80 data points in which 7 years of data is to be collected given the availability of monthly data [39]. With a data collection period of 2012 to 2018, our study has sufficient data points that would allow the SARIMA model to be used. Comparing SARIMA models in the monthly and quarterly analysis, the monthly analysis yielded a low R2 of 0.238 while the quarterly iteration had a higher R2 of 0.747. This suggests that the quarterly SARIMA model is a better predictor of hand injuries.

When comparing for citizenship status, in the quarterly analysis, multiple linear regression for non-Singaporeans suffering from hand injuries yielded a higher R2 (0.898) than Singaporeans (0.430). A possible explanation could be the rise in work permits given out in the period of 2012 to 2018. In particular, the total number of work permits issued for semi- and low-skilled jobs in 2014 was 991,300, however in the short span of 6 months in January to June 2015, 993,900 work permits for blue collar jobs (construction workers and factory workers) were issued [40]. Another contributory factor is the large proportion of foreigners that make up ground constructions teams, with figures as high as 90% being reported and a 78% proportion of foreign workforce relative to total employment in construction [41]. The Singapore government has also predicted that two-thirds of Singaporeans will hold white-collar jobs by 2030, up from half the workforce in 2013 [42].

In both construction and fishing industries, regardless of gender, there is a higher incidence of hand injuries. In the monthly comparison, the R2 for CI and AgriFish were 0.171 and 0.128 respectively for males; while in females this was 0.531 and 0.138. In the quarterly analysis, the R2 values were 0.204 (CI) vs. 0.157 (AgriFish) for males and 0.516 vs. 0.297 for females. This finding is likely attributed to the larger workforce in Singapore’s construction industry as opposed to the fishing industry. In 2013, it was reported that the main source of labour in Singapore’s fishing industry are the fish farm owners and his or her family members. Such a fish farm would employ less than 20 workers, even when seasonal workers are included. Many of these firms are labour-intensive with minimal automation [43].

Since men often work in more dangerous jobs than women and appear to have higher overall injury rates [44], we expected that the multiple linear regression model would fit better for men than women. However, our findings were contrary to this expectation. In both the monthly and quarterly analysis, the female subgroup analysis had a higher R2 than males. A previous study amongst aluminum smelter workers found that the injury rate was higher for females compared to males, arriving at the conclusion that women receive less on-the-job safety mentoring from supervisors and coworkers than men [45]. Male workers also tend to have more autonomy and control at work [46, 47], leading to a greater agency in the duration of work that they are exposed to compared to their female counterparts. It has been demonstrated that employees working long hours are more vulnerable to diverse types of occupational health problems [48] including hand injuries. These factors are likely more evident in industrial workforces traditionally dominated by men and could account for female workers having significantly higher risks of all injuries compared to male colleagues [49]. Our findings thus suggest a greater need for occupational workplace precautions and education amongst women to prevent the incidence of hand injuries.

With 17 to 62 being the legal working ages in Singapore [31, 32], we expected that the subgroup analysis for age would yield a higher R2 than SARIMA. This hypothesis held true in the monthly multiple linear regression, where R2 for the age subgroup yielded a higher R2 of 0.275 than SARIMA (0.238). However, this trend was not present in the quarterly analysis with the age subgroup analysis yielding a lower R2 of 0.477 than SARIMA (0.747). The rate of accidents and occupational injuries has been reported to be higher among blue-collar compared to white-collar workers [50–52]. The contracts for blue-collar work are relatively short-term relative to white-collar contracts [53], possibly explaining how the monthly analysis is better able to reflect the hand injuries suffered by blue-collar workers than the quarterly analysis.

After controlling for citizenship, we found that manufacturing index (MI) remained a significant dependent variable in the univariate analysis. The R2 for MI in the monthly analysis was 0.079, p value = 0.010, for Singaporeans and 0.305, p value = < 0.001 for non-Singaporeans, while this was 0.148, p value = 0.044 vs. 0.508, p value = < 0.001 in the quarterly analysis. This suggests that the higher the MI, the higher the rate of hand injuries regardless of nationality. The proportion of the foreign workforce relative to total employment is about 56% in manufacturing [41], indicating an almost equal percentage of Singaporean and non-Singaporean workforce in the manufacturing industry. We also found that TransAir was significant regardless of citizenship. While we were unable to find publicly available figures describing citizenship demographics in Singapore’s air transport cargo industry, we know that advances in automation in cargo operations industry could lead to a decrease in hand injuries. For example, previous cargo operation assistants previously had to manually sort and lift mail bags, which can be as heavy as 35 kg. However, since the introduction of automation in the form of automated tilt-tray sortation systems for intelligent processing capabilities and assisted loading devices for lifting mail bags, workers have undergone training to operate automated systems as technical specialists [54].

Singapore’s labour force has fluctuated across the years. The Report on Labour Force in Singapore by the Ministry of Manpower details that between 1994 to 1996, the percentage change year on year was 4.5 to 5.04% to 15.75% respectively [11]; and from 2012 to 2018, the percentage change was 3.71 to 2.38% to 2.47 to 2.21% to 1.69% to − 0.43% to 0.51% respectively [55]. In 2018, Singapore’s Ministry of Manpower’s reported that the total labour force was 3,675,600, with the employment rate being 97.3% (3,575,300) [11]. With a high employment rate of 97.3%, economic indicators such as the Straits Times Index (STI) and Gross Domestic Product (GDP) performed well for that year, with the 2018 economy expanding by 1.9% from 2017 [56].

In Singapore, hand injuries make up a significant portion of acute injuries [57], many of which are sustained in the industrial workplace. In 2020, there were 46 out of 463 (9.34%) cases where workers lost their hands or fingers in amputation accidents [58, 59]. The number of hand injuries would fluctuate with fluctuations in the total number of labour force in a country.

PMI is a measure of the prevailing direction of economic trends in manufacturing, summarising whether market conditions from the perspectives of managers are expanding, staying the same or contracting [60–62]. CI indicates the level of economic activity in the Construction sector [63]. Labour Productivity Change indicates whether output is increasing or decreasing per worker [64, 65]. Greater productivity describes being able to do more in the same amount of time, this in turn frees up resources to be used elsewhere.

Qualitatively, for example, as Purchasing Manager Index (PMI) is based on a monthly survey of supply chain managers across various industries, if these managers feel they expect a higher demand from customers for their goods, they will increase their orders. This provides a favorable outlook to the overall economic activity. Correspondingly, based on our equation, the number of hand injuries would increase. In like manner, if Labour Productivity Change for Manufacturing increases, which means that the hourly manufacturing economic output produced by an hour of labour rises, then there will also be a reciprocal increase in the number of hand injuries.

Regression models could be of value in resource utilisation by targeting the individuals in occupations or workplaces that need the most intervention [26]. Thus, prediction of hand injuries can provide a useful tool for occupational health safety policymakers by simulating changes in economic variables when applying new workplace or manpower interventions and regulations in the future.

To our knowledge, this is the first study to analyse the correlation of economic factors against the number of hand injuries. In our data, there is a relatively high proportion of foreign patients (41.21%) that undergo hand surgeries. As of 2018, the total foreign workforce in Singapore stood at 1,386,000 (24.58% of the entire population). However, they are usually under-represented in studies. In this study, we present a large cohort of foreign patients compared to the national proportion.

The clinical implications and findings of this study may allow us to forecast the clinical resources required to treat hand injuries from the construction, manufacturing and transport industries in relation to economic indicators.

We identified a few limitations in our study. Firstly, we used the incidence rate for hand surgeries as a proxy for the incidence rate for hand injuries. These hand surgeries included both elective and emergency operations and therefore may not be reflective of patients who presented with hand injuries but did not undergo hand surgeries. However, our large sample size of 20,764 patients is likely to account for exceptions pertaining to these cases. Hand injuries are typically emergency injuries such as lacerations and crush trauma; but our data combines both. Fortunately, elective operations are not known to fluctuate much. For example, trigger finger operations are not known to fluctuate from month to month [66–68]. Since we have found in our results that there is a significant correlation between the incidence of hand injuries and economic indicators, an extrapolation between two time points would effectively nullify the electives, leaving the emergencies to account for the cause of fluctuations.

Secondly, our tertiary institution covers referrals from the south and western regions of Singapore where a higher proportion of construction and manufacturing industries are located. Hence, there may be a higher proportion of hand injuries that may present to our institution as opposed to other local tertiary institutions. However, this allows us to have an increased sensitivity to the fluctuations in hand surgeries done emergently in response to fluctuations in the labour force.

Lastly, some of the economic markers used in our study were not reported in a timely fashion. For example, PMI is reported monthly, because companies use it to purchase equipment but others such as CI were published 1 year later. Therefore, the reporting time lag of the economic variables may dampen the usage of our equation as a timely method of forecasting.