HR (heart rate) is the simplest physiological variable to measure for estimation of exercise intensity (EI) and energy expenditure (EE). The recent rapid development of wearable technology incorporating HR and motion sensors facilitates measurement of EE over extended periods of time and this information may assist individuals in making healthy life-style changes.

In both a home and work setting, technology has resulted in decreased levels of physical activity. Within manufacturing industries, automation, mechanisation and robotics have reduced the physical demands of OPA resulting in increased sedentary behaviour [1] considered to be associated with an adverse health risk [2].

The use of HR to estimate EI is reliant on the linear relationship of HR to oxygen uptake (VO₂) [3]. However cardiorespiratory fitness (CRF) and tobacco smoking will influence this relationship [3] and to use HR without individual calibration of the HR-VO₂ limits its use. When using HR to estimate EE at near sedentary levels of activity where the HR-VO₂ relationship is less precise, the flex HR (HR_flex) method is frequently used [4]. Individual HR-VO₂ calibration is a prerequisite and in defining HR_flex as the average of the highest resting HR (HR_rest) and the lowest level of activity HR (HR_activity), time spent below HR_flex is considered to be 1 MET (1 MET being the VO₂ at rest, namely 3.5mL O₂·kg^-1·min^-1)[5].

The ability to correct for variables such as CRF and smoking that influence the HR-VO2 relationship was demonstrated in the recently published HR index (HRI) equation [6]. The equation relies on two variables to estimate EE, namely HR_rest and HR_activity, the equation being METs = 6 × HRI – 5, with HRI being HR_activity/HR_rest. The equation was determined from data from 60 studies involving 11,257 subjects and the relation of HRI to VO₂ (expressed as METs) was shown to be independent of common factors such as age, gender, CRF, smoking and beta-blocker effect. The original data are shown in Figure 1.

The utility of this equation is that it provides a simple independent surrogate method for estimating VO₂ using only two HR measurements, HR_rest and HR_activity, either submaximal or maximal. Whilst there has been no analysis to date to determine its predictive accuracy for individual assessment of VO₂, a recent study used HRI to assess EE in soldiers performing loaded runs from 10-20km in full combat equipment in a field environment [7]. An extremely close agreement was found between the use of HRI to assess EE when compared with equations previously validated by indirect calorimetry. The study is important for at sustained high work level (>50-75% of maximum VO₂), “cardiovascular drift” may occur [8]. This altered HR-VO₂ relationship is largely attributed to a reduction in stroke volume due to inadequate fluid replacement during extended work, particularly in work conditions involving high levels of ambient temperature and humidity [8,9].

With the world-wide trend towards increasing obesity [10], measurement of the EE of both OPA and LTPA is paramount for determining the potential contribution of decreased EE to obesity. In using both HR and HRI the impact of both smoking and exercise habits on OPA was assessed.

Subjects ranged in age from 17-60 years, males 35.9 ± 11.2 (mean±SD), females 33.9 ± 10.2 years. The BMI for males was 24.9 ± 3.3 and for females 22.1 ± 3.2. At the time of the study in 1984, 39 males (46%) and 28 females (30%) smoked with rates of smoking in Australia at that time being reported as 40% and 29% for males and females respectively [14]. A total of 81 (46%) subjects stated that they engaged in no regular exercise or sport. Summary data is shown in Table 1A (males) and 1B (females).

For both sexes a gradation in HRI_mean is seen from the lowest to the highest intensity occupational category. For males, the HRI_mean for sedentary workers was 1.19 (estimated 2.04 METs) increasing to 1.37 (estimated 3.06 METs) for the heavy category. For females, comparable values of HRI_mean were 1.17 (estimated 1.90 METs) increasing to 1.30 (estimated 2.67 METs). If HR_mean is considered for both sexes, a gradation for the five occupational categories is not seen. In fact, the HR_mean of 91.0 beats·min^-1 for males in the heavy work category is similar to the ambulant sedentary category (89.6 beats·min^-1) and light category (90.2 beats·min^-1) and less than that of moderate category (97.2 beats·min^-1). However, once relative HR in the form of HR indices is considered, the low HR_rest observed in the heaviest work category impacts on HRI_mean allowing for better discrimination of work intensity. For females HR_mean was quite similar for the five occupational categories with the sedentary and heavy groups having virtually identical HR_mean, sedentary 93.3 beats·min^-1 and heavy 93.2 beats·min^-1. However, the gradation in work intensity was again consistently shown by use of HRI_mean. A similar gradation by occupational category for HRI_peak is seen for both sexes which suggests that as HRI_mean increases, HRI_peak tends to parallel this change. Male subjects in the heavy work category had both a higher HRI_mean and HRI_peak than females; HRI_mean males, 1.37 and HRI_mean females 1.30 (p = 0.009), HRI_peak males, 1.73 and females, HRI_peak 1.59 (p=0.004).

To assess the individual effect of smoking and exercise participation on HR_rest, group data for males and females were combined. The group mean HR_rest for smokers (n = 67) was 78.8 beats∙min^-1 and 71.3 beats∙min^-1 for nonsmokers (n = 109), a difference of 7.5 beats∙min^-1 (p<0.001). The group mean HR_rest for exercisers (n = 95) was 69.1 beats∙min^-1 and 80.1 beats∙min^-1 for non-exercisers (n = 81), a difference of 11.0 beats∙min^-1 (p<0.001). When considered by gender, similar differences were observed between males and females.

The effect of smoking and exercise habit on HR_rest is shown in Table 2 where HR_rest is displayed in quartiles. In quartile 1 (HRrest 60.2 beats·min^-1), 84% had a regular exercise habit and 84% were non-smokers. A gradient of a decreasing rate of exercise habit and increasing rate of smoking is seen from quartile 1 to quartile 4, such that in quartile 4 (HR_rest 86.3 beats·min^-1) only 18% had a regular exercise habit and only 32% were non-smokers.

The study design attempted to discriminate between sedentary and ambulant-sedentary categories assuming that self-reported activity would identify individuals who undertook a higher level of incidental activity. However, a similar HRI_mean for sedentary and ambulant-sedentary categories was observed for males, the HRI_mean for both groups being 1.19 (estimated 2.0 METs). For females, the corresponding HRI_mean values for sedentary and ambulantsedentary were 1.17 and 1.22 (estimated 1.9 and 2.2 METs respectively, p=0.016). To identify subjects within the sedentary and ambulant-sedentary groups with higher levels of either incidental or work-related activity, a separate analysis of the combined group of sedentary and ambulantsedentary (S+AS) was undertaken. The intention was to discern the effect of smoking and exercise habit on HRI_mean. Data for both males and females for the two groups were pooled and divided into two roughly equal groups, a low range HRI_mean (males 1.14-1.18, average HRImean 1.16) and females (1.10-1.17, average HRI_mean 1.14) and a high range HRI_mean (males 1.18-1.32, average HRI_mean 1.23) and females (1.18-1.34, average HRI_mean 1.24). Subject characteristics for both sexes are shown in Table 3.

For both sexes, the observed differences of age and BMI between the two activity groups were not significant. However, for males the HR_rest and the HR_mean for the least active S+AS were significantly higher than the more active S+AS (p<0.001 and p=0.009, respectively); the HR profile for least and most active S+AS females was similar to the males (HR_rest, p<0.001and HR_mean, p<0.001).

Whilst the HR results for low and high range A+AS for both males and females were unexpected, this observation is likely related to the smoking and exercise habits of both groups. For males, 62% of the least active S+AS smoked compared with 0% for the more active S+AS (p<0.001). For females, 58% of the least active S+AS smoked compared with 6% of the more active S+AS (p=0.001). Fewer males in the least active S+AS exercised when compared with the more active S+AS, 37% and 80% respectively (p=0.029). For females, the least active S+AS exercise participation rate was 26% compared with the more active S+AS exercise participation rate of 74% (p=0.003).

The purpose of this study was to examine the individual effects of smoking and a regular exercise habit on HR and therefore HRI, and the ability of HRI to determine activity patterns of both sedentary and manual workers within a large metropolitan hospital. Though the concept of displaying occupational HR in the form of a histogram has previously been used, there has been no published method of processing histogram data to estimate EI in METs [15- 17]. The HRI method used in this study simplifies the use of HR to determine EI and estimate EE for a defined work period. However, at the time of the study, the limited memory capacity of the R-R interval recorder did not allow for estimates of accumulated time above defined MET thresholds e.g. light intensity 1.6 – 2.9 METs, moderate intensity 3.0 – 5.9 METs and heavy intensity >6 METs [5].

For both male and female occupational categories, HR_mean did not show any consistent trend to suggest increasing work intensity in progressing from sedentary to manual work categories, being influenced by exercise participation and smoking rates within occupational categories. However, with the use of HRI_mean, the trend of increasing work intensity was clearly apparent for both males and females across the five occupational categories (Tables 1A Males, 1B Females). The correction resulting from the use of HRI accounts for (and negates) the influence of the two common factors of smoking and CRF on the physiological modulation of HR. Other factors that can affect HR response include extreme environmental work conditions (heat, cold, high humidity) but were not relevant to the work conditions observed within this study at this hospital [9].

The significant difference in HR_rest observed in this study between smoking and non-smoking workers of 7.5 beats∙min^-1 (p<0.001) is similar to that observed in previous studies [18,19]. Exercise participation had an even greater effect on HR_rest when comparing exercisers with nonexercisers, the difference being 11.0 beats∙min^-1 (p<0.001). The impact of smoking and non-exercise participation is shown in the representation of HR_rest as quartiles in Table 2, with quartile 1 (predominantly non-smoking with a regular exercise habit) having a group mean HR_rest of 60.2 beats·min^-1, compared with quartile 4 (predominantly smoking and non-exercise), which had a group mean HR_rest of 86.3 beats·min^-1. Population studies have consistently shown that higher levels of CRF and physical activity are associated with lower rates of smoking [20,21]. Within this study, rates of smoking in exercising and non-exercising males were 27% and 67% respectively. A similar trend was noted with females, with smoking rates for exercising and non-exercising females being 22% and non-exercising 57% respectively.

The independent value of HR_rest as a prognostic variable is well established with lower levels of HR_rest being associated with reduced all cause and cardiovascular mortality [22]. The methods used for determining HR_rest are variable, two principal factors being position (supine, sitting or standing) and the time at rest before recording [22]. However, the HR_rest of male and female non-smoking subjects in this study is remarkably similar to Center of Disease Control (CDC) data (study males 69.2 beats∙min^-1, CDC males 68.5 beats∙min^-1, study females 72.7 beats∙min^-1, CDC females 72.5 beats∙min^-1) [23]. HR_rest could have been measured independently of the HR histogram recorded data but the study design relied on the inclusion of a minimum 20-30 minute rest period (during refreshment and lunch breaks) for determination of HR_rest. As HR_rest is a crucial measurement for determining HRI, it is highly desirable that its measurement is standardized.

Whilst the essential component of sedentary behaviour is that of prolonged sitting, the current consensus is that sedentary behaviour be defined by activities < 1.5 METs [24]. Other studies e.g. Current Population Survey and the American Time Use Survey have used a higher cut-off point of 2 METs [1,25]. Within a sedentary occupation there will be components of prolonged sitting interspersed with periods of standing and walking. The sitting activity itself may entail various physical levels of upper limb activity. This variation in incidental activity associated with sedentary work will affect the average EI during a defined work period.

Whereas it is necessary to individually calibrate the HR-VO₂ relationship with exercise testing when using the HR_flex method [4], the HRI method does not have this requirement being independent of the various factors that influence the HR-VO2 relationship [6]. HR_flex is frequently in the range of 8%-16% above HR_rest [26,27], i.e., equivalent to an HRI of 1.08-1.16, (estimated 1.5-2.0 METs), and that HR_flex increases significantly with ageing [28]. Studies that have used HR_flex as a cut-off point advocate that time spent below HR_flex be considered as equal to resting metabolism i.e. 1 MET [28,29]. For the 68 subjects within the combined sedentary and ambulant-sedentary categories, mean predicted MET levels for the entire work period ranged from 1.61 to 3.05 METs, with not one subject having a mean MET level of < 1.5 METs. For the 25 subjects that had a HRI of < 1.16 (approximating the upper limit of HR_flex), the estimated mean MET level was 1.84 METs, noticeably higher that what would have been accepted by the HR_flex method.

This study suggests that both smoking and exercise habit influence OPA particularly in sedentary occupations. Individuals with higher levels of incidental activity in the A +AS group as shown by a higher HRI_mean had significantly higher participation rates of LTPA (Table 3A and 3B). In measuring sitting time and step counts of office workers, Clemes et al showed that office workers with lower sitting time had both higher step counts at work and on nonworkdays, suggesting higher levels of LTPA, a similar observation to this study [30]. The importance of identifying the more active sedentary worker is supported by the results a recent publication from the Maastricht study. Reallocation of 30 mins of sedentary time per day to standing or stepping was significantly associated with 7-25% lower odds for the metabolic syndrome and 6-21% lower odds for type 2 diabetes [31]. In our study, a trend towards a lower BMI was noted in both males and females in the more active ambulant and ambulant-sedentary groups, with these subjects having higher rates of LTPA and lower rates of smoking. Whether a higher HRI_mean within a defined occupational category is associated with increased work productivity is a matter of conjecture. However, a known consequence of smoking and exercise habit is that higher rates of absenteeism have been observed with both smoking and lower levels of CRF [32,33].

Three categories were used to define manual work (light, moderate and heavy). For these three categories, the estimated mean MET levels for males were 2.3, 2.4 and 3.1 respectively, and for females 2.2, 2.3 and 2.7 METs respectively. In using the commonly accepted international reference range for PA (sedentary 1.0-1.5 METs, light 1.6- 2.9 METs, moderate 3.0-5.9 METs and heavy > 6 METs), 93 (86.1%) of the manual 108 workers within the hospital would have been classed as ‘light’ OPA, the remaining 15 (18.9%) of workers being within the ‘moderate’ OPA classification. Whilst this internationally accepted classification may be an appropriate expression of EI over a short period of continuous PA e.g. 15-60 minutes, it may be less appropriate for expressing the EI intensity over a prolonged period of discontinuous PA, with varying work intensity interspersed with rest breaks that are likely to occur during an 8-hour work schedule.

The utility of using a defined level of peak activity (as METs or alternatively as an HRI_peak) is demonstrated in a study determining daily EE from ambulatory electrocardiographic monitoring of 843 patients at Beaumont Health in Royal Oak, Michigan. Mean and peak METs were estimated from the HRI equation [34]. Their study showed that peak activity greater than 5 METs was associated with significantly lower health cost (approximately 50%) than with subjects having less than a 5 MET peak activity. With epidemiological studies it is clear that the method used to assess OPA is crucial, particularly if studies are to be compared, a comprehensive review of the strengths and limitations of PA assessment tools having been recently published by Ainsworth [35].

The rapid expansion of wearable technology has focussed on measurement of energy expenditure (EE). By using different sensors, commonly optical HR and motion sensors, proprietary algorithms utilize recorded data in conjunction with anthropometric variables to estimate EE. Currently, errors of measurement of EE with these devices are large (20-40% error) when compared with the accepted comparators of indirect calorimetry or doubly labelled water [36]. However, the recent study by Colosio demonstrated that in using only HRI to determine EE, the results closely approximated those from equations validated by indirect calorimetry [7]. Data used in this study was acquired in 1983-84 using used a purpose-built microprocessorbased recorder to measure HR and generate and store HR histograms. Use of a histogram with a bin width of 5 beats·min^-1 was necessary to compress the data to fit into the miniscule available memory capacity at that time of only 256 bytes of data memory. Currently available HR recording devices have extremely large memory capacities (up to gigabytes) enabling recording of every heart beat over extended time periods (more than 24 hours). Our study suggests that the use of HR data alone, in the form of HRI may well provide a more accurate estimate of EE than that attained from proprietary algorithms incorporated in the presently available wearable devices. Data downloaded from these HR recording devices can be easily processed using commercially available spreadsheet programs enabling histogram construction and comprehensive data analysis. Time spent above defined MET thresholds can be determined and the extended recording time offered by these devices allows for separate interrogation of OPA and LTPA which assists in determining temporal trends of PA and health risk associated with both low and high levels of PA.

The results of this study suggest that smoking and CRF may influence the occupational behaviour of sedentary workers. By continuous recording of HR and the use of a HR histogram, HR indices derived from HR_rest, HR_mean and HR_peak allow for estimates of EI during a period of OPA. The intrinsic value of the HRI is that it negates the effect of factors that may influence HR such as level of fitness, smoking habit and the effect of medication e.g. beta-blocker therapy. The observed levels of the three HR measurements (HR_rest, HR_mean and HR_peak) and HR indices (HRI_mean and HRI_peak) have the potential to stratify health risk resulting from the effect of smoking and CRF on HR. This method appears to be particularly suited to evaluating sedentary occupations where the contribution of incidental activity to EE can be determined.