job strain and heart rate variability in resident physicians within a general hospital

AMERICAN JOURNAL OF INDUSTRIAL MEDICINE

Job Strain and Heart Rate Variability inResident Physicians Within a General Hospital

Sendy Isarel Hernandez-Gaytan, ScD,1 Stephen J. Rothenberg, PhD,2

Paul Landsbergis, PhD, MPH,3 Leonor Cedillo Becerril, ScD,4 Guillermo De Leon-Leon, ME,5

Sean M. Collins, ScD,6 and Francisco Javier Dıaz-Vasquez, MHSA5�

Objective To evaluate the association of heart rate variability with job strain in firstyear resident physicians.Methods We performed the study at the ‘‘Manuel Gea Gonzalez’’ General Hospital inMexico City. 54 resident doctors were studied over a period of 24 hr in their first yearof specialization. Two questionnaires were administered: the first on general demo-graphics, and the second, the Job Content Questionnaire. Heart rate variability wasevaluated through the frequency domain (low-frequency power, high-frequency power,and low-frequency power/high-frequency power ratio) and time domain (SDNN). Thedoctors wore a Holter monitor over a 24-hr period, which included a workday plustheir on-call time. They recorded their activities in a log.Results Compared to physicians in the ‘‘low strain’’ category, physicians working inthe ‘‘passive’’ category had lower overall peak-to-peak cardiac variability (standarddeviation of N–N intervals, SDNN), �9.08% (95% CI �17.97, 0.74), a �25% (95%CI �45.00, 0.22) lower high-frequency power, and �26.95% (95% CI �39.00,�12.53) lower low-frequency power. Physicians working in the ‘‘high strain’’ categoryhad lower low-frequency power, �17.85% (95%CI �32.34, �0.25), and lower low-frequency/high-frequency ratio �24.29% (95% CI 38.08, 7.42) compared to those inthe ‘‘low strain’’ category.Conclusions High job strain and low job control among medical residents wereassociated with several indicators of lowered heart rate variability. Thus, analysisof heart rate variability may be an informative marker for evaluating thephysiological impacts of workplace stressors. Am. J. Ind. Med.� 2012 Wiley Periodicals, Inc.

KEY WORDS: job strain; heart rate variability; demand–control model

Disclosure Statement: The authors report no conflicts of interests.1Instituto Nacional de Enfermedades Respiratorias,Me¤ xico,DF,Mexico2Instituto Nacional de Salud Publica, Cuernavaca,Morelos,Mexico3Department of Environmental and Occupational Health Sciences, School of Public

Health, State University of NewYork-DownstateMedical Center,NewYork,NewYork4Instituto Nacional de Ecolog�a,Me¤ xico,DF,Mexico5Comisio¤ n Coordinadora de Institutos Nacionales de Salud y Hospitales de Alta

Especialidad,Me¤ xico, DF,Mexico6Department of Physical Therapy, School of Health and Environment, University of

Massachusetts, Lowell,Massachusetts

Contract grant sponsor: Irving J. Selikoff International Scholar of the Mount Sinai Schoolof Medicine; Contract grant number:D43TW000640.

*Correspondence to: Francisco Javier D|¤ az-Va¤ squez, MHSA, Comisio¤ n Coordinadora deInstitutos Nacionales de Salud y Hospitales de Alta Especialidad, Perife¤ rico Sur 4118, 1erPiso, Col. Jardines del Pedregal, Delegacio¤ n A¤ lvaro Obrego¤ n, C.P. 01900, Me¤ xico, DF, Mexico.E-mail: [email protected]

Accepted 28 June 2012DOI10.1002/ajim.22098.Published online inWiley Online Library

(wileyonlinelibrary.com).

�2012WileyPeriodicals,Inc.

INTRODUCTION

Heart rate variability (HRV) measures are useful

descriptors and predictors of cardiac prognosis in the clin-

ically ill [Kristal-Boneh et al., 1995; Task Force of the

European Society of Cardiology and The North American

Society of Pacing and Electrophysiology, 1996; Stein and

Kleiger, 1999] and can predict cardiovascular morbidity

and mortality in healthy individuals [Tsuji et al., 1994,

1996; Dekker et al., 1997].

Several studios have shown that low HRV is associat-

ed with cardiovascular disease [Togo and Takahashi,

2009]. Lombardi et al. [2001] found that the parameters

most consistent with cardiovascular disease (CVD) risk

were the standard deviation of all NN intervals (SDNN),

high Frequency (HF), low Frequency (LF). However, the

parameter low-frequency/high-frequency ratio was less

consistently associated with CVD risk.

HRV is an indicator of heart rate regulation; it meas-

ures autonomic nervous system interactions, reflecting

all the physiological factors that modulate the normal

heart rate. In general, highly stressed people exhibit

reduced HRV as a result of the overactivity of the sympa-

thetic nervous system [Kang et al., 2004] and reduced ac-

tivity of the parasympathetic nervous system. HRV

analysis has been proposed to assess job strain and its re-

lation with cardiovascular diseases [Kristal-Boneh et al.,

1995].

One widely studied model [Belkic et al., 2004], evalu-

ating main job characteristics leading to stress is the de-

mand–control model [Karasek and Theorell, 1990; Schnall

et al., 1994], also known as the job strain model [Van Der

Doef and Maes, 1999].

The model describes two work dimensions: psycho-

logical job demands and job control [Schnall et al., 1994].

Job demands refer to work exigencies in terms of quantity,

promptness, effort, time [Karasek, 1985]. Job control, also

called job decision latitude, refers to those conditions

which make learning and skill development and their exer-

cise possible. Decision latitude includes two components:

skills and authority; these components are usually com-

bined in one measure [Van Der Doef and Maes, 1999].

Job strain (or high-strain jobs) occurs when the human

body is psychologically overloaded (high demands) and

the worker has little control over the job [Karasek and

Theorell, 2000]. Jobs strain increase the risk of cardiovas-

cular diseases [Van Der Doef and Maes, 1999].

There is a wide array of research on work stress using

the demand–control model and cardiovascular diseases,

but very few studies have examined the association be-

tween work stress and changes in the heart rate variability.

HRV studies have been carried out in samples of shipyard

workers [Kang et al., 2004], manufacturing workers [Van

Amelsvoort et al., 2000], nurses [Riese et al., 2004], and

other workers [Collins et al., 2005], but physicians in

training have not previously been studied using the

demand–control model and HRV.

The demand–control model proposes that professio-

nals such as physicians are classified as having active jobs

[Karasek et al., 1998]; that is, they are exposed to high

psychological demands mixed with high decision latitude.

We expect to find a much lower percentage of resident

physicians in the active category, due to a high work load

combining extensive training and patient treatment activi-

ties [Michie and Sandhu, 1994; Butterfield, 1998]. In

Mexico, first year residents have little decision-making lat-

itude; they basically carry out medical procedures ordered

by more advanced residents and staff physicians.

Studies of resident physicians have evaluated sleeping

hours and their consequences in car accidents, resident

physician’s satisfaction [Perry and Osborne, 2003] and

contrasting the quality of life between resident physicians

and staff physicians, including depression, drug abuse,

loneliness, marriage breakdown and cognitive damage

[Sargent et al., 2004]. But there are no studies regarding

resident physicians’ stress using the demand–control mod-

el and HRV. The purpose of the study was to evaluate the

association of HRV with job strain in first year resident

physicians at a general hospital. The hypothesis of this

study is that a reduction will be observed in the param-

eters of heart rate variability associated with stress in resi-

dent doctors who report high job strain compared to those

who report low job strain.

METHODS

We performed the study at the ‘‘Manuel Gea

Gonzalez’’ General Hospital in Mexico City. A total of 54

resident doctors with the following attributes were includ-

ed: (1) free of diagnosed cardiovascular diseases; (2) not

under treatment with medications that act on the autono-

mous nervous system; and (3) entered their first year of

specialization within various medical specialties in the

month of March in the years 2007 and 2008. The medical

specializations were anesthesiology, general surgery, gyne-

cology and obstetrics, internal medicine, ophthalmology,

otolaryngology, orthodontics, orthopedics, pathology, pedi-

atrics, radiology and imaging, and medical emergencies.

The Research and Ethical Boards of the Manuel Gea

Gonzalez General Hospital and the National Institute of

Public Health approved the study. Each medical resident

signed an informed consent. Between the months of

August 2007 and February 2008, resident physicians were

given two types of questionnaires during their working

hours: the first questionnaire collected data on general

background and personal characteristics, health habits, and

medical history; the second questionnaire was the Job

Content Questionnaire (JCQ) [Karasek, 1985], which

2 Hernandez-Gaytan et al.

assesses psychological job demands (five items) and job

decision latitude (nine items).

Residents responded on a 5-point Likert scale as

follows: strongly agree (5), agree (4), neither agree nor

disagree (3), disagree (2), and strongly disagree (1). We

used the median of the psychological demands scale

and the median of the job decision latitude scale as

the cutoff point to classify these variables as high or low.

The resulting combinations of the two dimensions were:

(a) low strain (low demands–high latitude); (b) passive

(low demands–low latitude); (c) active (high demands–

high latitude); and (d) high strain (high demands–low

latitude).

Holter Monitoring and Heart RateVariability

Heart rate variability was evaluated throughout a

24 hr period. The Holter monitor was a 3 channel recorder

with 7 leads (CSG-HP3VER306 model, Cardiosolutions

Group INC, 1000-North Washington Blvd Sarasota, FL),

with a 200 Hz sampling rate, and 8 bit resolution. Skin

preparation and electrode placement were according to the

international standard [Task Force of the European Society

of Cardiology and The North American Society of Pacing

and Electrophysiology, 1996]. The ECG signal was

recorded from the standard lead configurations (V2, V4,

V5). Only 24 hr reading records were used. The data was

processed with Holter Plus III software version 3.06. Only

those segments without noise were analyzed. All the

ectopic beats were automatically and manually identified

and were excluded from the analysis. A cardiologist certi-

fied by the Mexican Council of Cardiology reviewed

all electrocardiographic records. Each resident wore the

Holter-type recorder during the entire 24-hr shift at

the hospital, including their regular daily work activities.

The residents recorded their own activities throughout

their shift every hour in a logbook. The seven activity

groups were: medical assistance, medical visit, rest (they

were also asked to include the time devoted to sleeping, in

hours and minutes), surgical activities, food ingestion, aca-

demic activities and anesthesiology activities. In this hos-

pital, all resident doctors work an 8-hr daytime shift.

Nevertheless, each third day the resident is asked to fulfill

a 24-hr (on-call) period and carry on with their 8-hr day-

time shift. Upon completion of the second day, the resi-

dent has worked a total of 32 hr, then rests and the next

day comes back to work their normal daytime shift plus

the on-call time. This hospital offers 12 distinct residency

programs. Each medial specialty has its unique academic

program with different times devoted to theoretical and

practical classes, presentations before the specialty group,

medical visits to hospitalized patients, surgical procedures,

and emergency room service. Thus, the work load of

residents in different specialties varies in both timing dur-

ing working hours and in tasks performed.

Frequency domain HRV variables were calculated

through spectral analyses using fast Fourier transforma-

tion. We calculated (a) low frequency (LF, 0.04–0.15 Hz)

is an indicator of sympathetic autonomic response [Task

Force of the European Society of Cardiology and The

North American Society of Pacing and Electrophysiology,

1996]. Yeragani et al. [1993] reported lower LF among

panic disorder patients; (b) High Frequency (HF, 0.15–

0.4 Hz) is an indicator of parasympathetic autonomic

response [Malliani et al., 1991, 1994]. Friedman and

Thayer [1998] found lower HF among patients with

generalized anxiety disorder; and (c) the low-frequency/

high-frequency ratio reflects the global sympatho-vagal

balance [Sztajzel, 2004] and can be used as a measure of

this balance, in the frequency domain over the total

period (24 hr) of the monitoring and for the 24 one-hour

periods. The HRV measurement in the time domain was

calculated by the standard deviation of all the NN inter-

vals (SDNN) for the total period (24 hr) of monitoring

and per 1-hr period. SDNN is a global HRV index that

reflects the long term circadian components and rates of

the variability within the recording period of 24 hr

[Sargent et al., 2004].

Low values of SDNN, HF power represents an

increase of cardiovascular risk and psychological stress

and job strain.

Statistical Analysis

Descriptive statistics summarized the socio-demo-

graphic and physiological features of the resident physi-

cians’ sample. All four dependent variables (a) low-

frequency power; (b) high-frequency power; (c) low-fre-

quency power/high-frequency power ratio; and (d) HRV

(measured with SDNN) were natural log transformed. We

used mixed linear models [Diggle, 1996].

Each measurement was modeled as Yij ¼ Bo þBiTij þ B2Gi þ V0j þ V1iTij þ eij where i represents each

individual 1. . .54 (each physician), j represents each group

1. . .12 (specialty), Yij corresponds to those variables

which depend on the HRV parameters (low-frequency

power, high-frequency power, low-frequency/high-

frequency ratio, SDNN) of each subject, Bo is the

intercept, Tij is the time variable for individual i in time

j, Gi are the fixed explanatory variables in the subject

throughout the study including categories of the demand–

control model (low strain, passive, active, and high strain),

sex, age, body mass index (BMI) with BMI (weight

in kg/height in m2) as a categorical variable where

low BMI < 18.5, normal BMI ¼ 18.5–24.9, overweight

BMI ¼ 25–29.9, and obese BMI > 30 [WHO, 1995], the

waiting time variable is the time in days that each

Job Strain Influences Heart Rate Variability in Residents Physicians 3

physician waited from the start of their specialty training

until the day on which they started their electrocar-

diographic record (Holter)), V0j is the random intercept

(specialty and physicians), V1i are the variables with ran-

dom effects (physicians’ activities and time of day) within

each specialty. The model assumes that random effects

have a normal distribution with a mean ¼ 0 and a con-

stant variability; Bo, Bi, B2 are the adjusted effect

coefficients, V is the random coefficient vector and e is therandom error term which follows the following distribu-

tion: eij � N(0, s i2 þ s d

2) where: si2 ¼ variability be-

tween subjects; sd2 ¼ variability within subjects. Akaike’s

information criteria (AIC) was used to select the model

with best fit [Akaike, 1974]. To determine if the addition

of a random intercept improved model adjustment, the

Maximum Likelihood test was used at a ¼ 0.05. The

quality of the final model adjustment was verified by nor-

mal quantile–quantile plot of residuals. The confounding

effect of the age, sex, and BMI, variables was evaluated.

Associations with a P < 0.05 value in the main effects

were considered, with a confidence interval of 95%. The

Stata statistical package version 9.0 (StataCorp LP, Col-

lege Station, TX) was used.

In addition, we tested for significant association be-

tween job demands and job decision latitude and the four

dependent variables, in linear mixed models. Models V

and VI (see Apendix) present the main effects of decision

latitude and demands each entered as dichotomous varia-

bles split at the median score.

In the Results Section, coefficients and intervals for

categorical variables are shown in percentages for the four

multivariable models, converted using the formula:

(100 � (exp(coefficient)�1))[Halvorsen and Palmquist,

1980; Kennedy, 1981], for natural log dependent variables

whereas continuous variables are naturally read as percent

change.

RESULTS

Table I describes the demographic, health and occupa-

tional characteristics of the sample of 54 first year resident

physicians. There were no statistically significant differen-

ces between men and women on these characteristics.

Table II presents descriptive statistics for each HRV

variable. Women had lower heart rate variability than

men. Women have lower values of low-frequency and

high-frequency power compared to men; values for men

and women. are within normal levels for all subjects [Task

Force of the European Society of Cardiology and The

North American Society of Pacing and Electrophysiology,

1996]. The low-frequency/high-frequency ratio of women

is within the 1.5–2 interval, which corresponds to the gen-

eral population. However, this same ratio for men is above

this interval (2.1) [Task Force of the European Society of

Cardiology and The North American Society of Pacing

and Electrophysiology, 1996].

Table III shows SDDN and high-frequency power de-

rived from the models in resident physicians, correspond-

ing to percentages calculated according to the formula

described in methods. Physicians reporting passive jobs

were �9.08% (95% CI �17.97, 0.74; P ¼ 0.069) lower in

TABLE I. General Characteristics of First Year Resident Physicians bySex,Me¤ xico City, 2007^2008

Males(n ¼ 36)number (%)

Females(n ¼ 18)number (%)

Fisher’sexact test(Pvalue)

Age23^25 6 (16.6) 2 (11.1) 0.19726^27 23(63.8) 8 (44.4)28^36 7 (19.4) 8 (44.4)

Marital statusSingle 33(91.6) 13 (72.2) 0.100Married 3 (8.3) 5 (27.7)

IMC26

Malnutrition 0 (0.0) 1 (5.5) 0.179Normal 20(55.5) 13 (72.2)Overweight 14 (38.8) 3 (16.6)Obesity 2 (5.5) 1 (5.5)

Averageexercise in aworkingweek<1 hr 13 (36.1) 4 (22.2) 0.4581 3 hr 4 (11.1) 4 (22.2)4^6 hr 2 (5.5) 0 (0.0)None 17 (47.2) 10 (55.5)

Health conditionExcellent 8 (22.2) 7 (38.8) 0.633Good 21 (58.3) 8 (44.4)Regular 6 (16.6) 3 (16.6)Bad 1 (2.7) 0 (0.0)

Howmanydaysdidyougo toworkwithpain lastmonthMostdays 1 (2.7) 1 (5.5) 0.46950%ofdays 3 (8.3) 3 (16.6)Acouple ofdays 18 (50.0) 10 (55.5)Never 14 (38.8) 4 (22.2)

Jobstrain categoriesLowstrain 8 (22.2) 3 (16.6) 0.958Passive 9 (25.0) 6 (33.3)Active 11 (30.5) 5 (27.7)Highstrain 8 (22.2) 4 (22.2)

JobdemandLow 17 (47.2) 9 (50.0) 0.346High 19 (52.7) 9 (50.0)

Decision latitudeLow 17 (47.2) 10 (55.5) 0.387High 19 (52.7) 8 (44.4)

(Continued )


SDNN and �25.75% (95% CI �45.00, 0.22; P ¼ 0.052)

lower in the high-frequency power component (parasym-

pathetic) compared to physicians reporting low job strain.

These differences approached statistical significance. No

significant difference was found for either SDNN or the

high-frequency power component when comparing high

job strain to low job strain physicians.

Table IV shows that physicians reporting passive jobs

were �26.95% (95%CI �39.00, �12.53) lower in low-fre-

quency power compared to those reporting low-strain jobs.

Physicians reporting high job strain was �17.85% (95%CI

�32.34, �0.25) lower in low-frequency power (sympa-

thetic) compared to those reporting low job strain.

The low-frequency/high-frequency ratio was �24.29%

(95%CI �38.08, �7.42) lower in physicians reporting high

job strain than in those reporting low job strains.

When the components of job strain (job demands and

job decision latitude) were examined separately, job

demands were not significantly associated with any HRV

parameter (see Appendix in Supporting Information).

However, job decision latitude was significantly associated

with 25.84% higher low-frequency power (95%CI 9.39,

44.77) and 18.62% higher low-frequency/high-frequency

ratio (95% CI 2.44, 37.34). Thus, the associations we ob-

served between job strain and lower low-frequency power

and lower low-frequency/high-frequency ratio, were due

to the effect of job decision latitude and not job demands.

DISCUSSION

The present study provides evidence of lower heart

rate variability over four of its parameters (standard devia-

tion of all of the NN intervals, high-frequency power, low-

frequency power and the ratio of low-frequency power

over high-frequency power), in residents reporting passive

jobs or high-strain jobs compared to those reporting low-

strain jobs.

Our hypothesis that high-strain jobs would demon-

strated reduced high-frequency power was not confirmed.

However, we found lower high-frequency power in the

TABLE I. (Continued )

Males(n ¼ 36)number (%)

Females(n ¼ 18)number (%)

Fisher’sexact test(Pvalue)

SpecialtiesEmergency care 5 (13.89) 2 (11.11) 0.246Radiology 1 (2.78) 1 (5.56)Pediatrics 3 (8.33) 4 (22.22)Orthopedics 3 (8.33) 0 (0.00)Orthodontics 0 (0.00) 2 (11.11)Ophthalmology 3 (8.33) 1 (5.56)Internalmedicine 4 (11.11) 2 (11.11)Otolaryngology 2 (5.56) 0 (0.00)Pathologic anatomy 1 (2.78) 1 (5.56)Anesthesiology 4 (11.11) 2 (11.11)Gynecologyandobstetrics 3 (8.33) 3 (16.67)General surgery 7 (19.44) 0 (0.00)

TABLE II. Descriptive Statistics of Heart RateVariability in First Year Resident Physicians by Sex,Mexico City, 2007^2008

Na

Men1 (n ¼ 36) Women1 (n ¼ 18)

PvaluebMeanc 95%CIofmean Meanc 95%CIofmean

Heart rate (beats/min) 54 84.3 83.4,85.3 88.8 87.4,90.0 0.127Heart ratevariability (SDNN) (ms)2 54 87.2 84.6,89.8 74.6 71.5,77.6Low-frequencypower (ms)2 54 652.5 621.7,683.2 363.9 340.3,387.5High-frequencypower (ms)2 54 676.0 609.2,742.7 443.4 387.2,499.7Low-frequencypower/high-frequencypower ratio 54 2.1 1.9,2.1 1.5 1.4,1.6

Na Geometricmean 95%CIofgeometricmean Geometricmean 95%CIofgeometricmean Pvalue

Heart ratevariability (ms)2 54 80.1 77.9,82.3 68.9 66.4,71.6 0.046Low-frequencypower (ms)2 54 517.4 493.1,542.9 293.9 275.7,313.2 0.001High-frequencypower 54 322.0 296.5,349.7 242.48 218.2,269.3 0.092Low-frequencypower/high-frequencypower ratio 54 1.6 1.5,1.6 1.2 1.1,1.3 0.018

aN refers to number of subjects; each subject value is themean of 24 consecutive1-hr electrocardiogram samples.bP values are for the difference in meansbetweenmen andwomen.cDistributions of all heart rate variables except ‘‘heart Rate’’ were highly right-skewed; thus,means of these variables and their 95% confidence intervals are not accuratemeasuresof central tendency and uncertainty andP-value of sexdifferences of arithmeticmeans are not shown. Arithmeticmeans are shown only for comparisonswith published literature,where they are frequently used.More accurate measures of central tendency and their uncertainty in this data set are shown for the geometric means and 95% CI.Pvalues of sexdifferences for geometric means are statistically valid.


group of doctors reporting passive jobs (P ¼ 0.06). Eller’s

study et al. [2010 reported a reduction of high-frequency

power in men (nursery school teachers, teachers, police-

men, engineers, technical personnel and salesmen) with

lack of support (P ¼ 0.05). Vrijkotte et al. [2000] found

high-frequency power to be lower in men exposed to

greater job stress (as measured by the effort-reward imbal-

ance model). However, Van Amelsvoort et al. [2000]

found no association between the high-frequency power

parameter and the categories of the demand–control

model.

Our hypothesis was also not confirmed for variability

of heart rate (SDNN). However, we did observe lower

SDNN among residents reporting passive jobs (P ¼ 0.06).

Kangs study et al. [2004 reported reduction in SDNN

(P ¼ 0.06) in the high-strain group. Collins et al. [2005]

found no association between heart rate variability and

high job strain (P ¼ 0.46). Riese et al. [2004] found no

effect on heart rate variability in job strain by itself or in

interaction with social support. Van Amelsvoort’s study

et al. Van Amelsvoort et al., 2000[2000] found no associa-

tion between heart rate variability and high demands plus

low control (job strain), but reported a reduction in SDNN

(17%, P ¼ 0.05) in the category of high demands and

high control. In addition, Hintsanen et al. [2007] found

that high effort–reward imbalance (ERI) was associated

with low heart rate variability in women, but found no

association in men.

TABLE III. Job strain and Heart RateVariability (SDDNandHigh-Frequency Power),Mixed LinearModel, in Resident Physicians,Mexico City, 2007^2008

Variable

SDNNa,b High-frequencypowera

% 95%CIc Pvalue % 95%CIc Pvalue

Jobstrain categoriesLowstraind

Passive �9.089 �17.97,0.74 0.069 �25.75 �45.00,0.22 0.052Active 0.180 �8.57,9.77 0.968 �10.42 �31.51,17.12 0.421Highstrain �0.240 �10.21,11.94 0.965 �7.70 �21.52,49.60 0.626

SexWomend

Men 11.54 3.66,20.02 0.003 23.13 �0.64,52.63 0.057Age23^25d

26^27 �4.18 �12.72,5.20 0.370 �9.73 �31.41,18.80 0.46528^36 �19.21 �27.48,�10.00 0.001 �48.04 �62.07,�28.82 0.001

IMCe

<18.5d

18.5^24.9 �33.45 �48.94,�13.27 0.003 �80.44 �90.88,�58.02 0.00125^29.9 �29.17 �45.58,�7.82 0.010 �78.92 �90.16,�54.82 0.001>30 �41.69 �56.94,�21.04 0.001 �90.86 �96.18,�78.11 0.001

ActivityRestd

Medical visit �14.73 �24.22,�4.05 0.008 �40.78 �55.32,�21.51 0.001Medical assistance �9.65 �18.65,0.33 0.058 �18.52 �36.20,4.03 0.101Surgical activities �5.67 �17.10,7.31 0.375 �17.56 �39.43,12.19 0.220Food ingestion �15.50 �24.72,�5.16 0.004 �35.05 �50.59,�14.64 0.002Academic activities �2.43 �13.74,10.32 0.694 �10.96 �33.56,19.32 0.437Anesthesiologyactivities �23.63 �38.01,�5.89 0.011 �47.19 �68.00,�12.83 0.013

Waitingperiod (days) 0.03 �0.009,0.09 0.145 6.18 �0.08,0.21 0.146

aN refers to sample size.The target number of1-hr measurements for the 54 resident physicianswas1,296.bSDNNstandard deviation of all theNN intervals.cConfidence intervals at 95%.dReference category.eCategories according toWHO (normal18.5^24.9, overweight 25^29.9, and obese>30 kg/m2).


We found lower values in the parameter of low-

frequency power among residents reporting high job strain

(P ¼ 0.047) and passive jobs (P ¼ 0.001), in comparison

to residents reporting low job strain; and for low-frequen-

cy/high-frequency ratio among residents reporting high

job strain (P ¼ 0.007). The results provide evidence of an

association between exposure to job stress, and the physi-

ological processes known to be associated with cardiac

deregulation. Dishman et al. [2000] reported that healthy

people with high psychological stress exhibit low values

on the parameters of heart rate variability, irrespective of

age, sex, heart rate, cardio-respiratory capacity, blood

pressure and respiration rate. It was also reported that

those who displayed levels of stress and fatigue had lower

low-frequency and high-frequency power. In turn, Collins

et al. [2005] found an increase in low-frequency power

during working hours in the high-strain group

(P ¼ 0.002). Van Amelsvoort et al. [2000] found an in-

crease in low-frequency power in the group reporting high

demands plus low control (P ¼ 0.05). Collins et al. [2005]

reported an increase in low-frequency/high-frequency ratio

during working hours in the high-strain group (P ¼0.009). Our results do not coincide with these studies,

because there was no consistency within the HRV para-

meters in these studies, with one possible explanation

being the differences in the criteria for inclusion, as well

as in evaluation of the variables. Collins et al. [2005] se-

lected a heterogeneous group of workers, with an average

age of 45 years, which leads one to speculate that they

had already experienced a prior (and perhaps chronic) job

TABLE IV. Job Strain and Heart RateVariability (Low-Frequency Power and Low-Frequency/High-Frequency Ratio),Mixed LinearModel, in ResidentPhysicians,Mexico City, 2007^2008

Variable

Low-frequencypowera Low-frequency/high-frequency ratioa

% 95%CIb Pvalue % 95%CIb Pvalue

Jobstrain categoriesLowstrainc

Passive �26.95 �39.00,�12.53 0.001 �1.95 �18.67,18.18 0.836Active �5.29 �19.42,11.29 0.509 5.68 �10.55,24.89 0.516Highstrain �17.85 �32.34,�0.25 0.047 �24.29 �38.08,�7.42 0.007

SexWomenc

Men 64.08 44.19,86.73 0.001 33.70 16.97,52.83 0.001Age23^25c

26^27 �10.04 �23.76,6.13 0.210 �0.85 �16.42,17.58 0.92128^36 �36.70 �47.59,�23.54 0.001 21.22 �0.35,47.49 0.054

IMCd

<18.5c

18.5^24.9 �64.38 �77.43,�43.80 0.001 79.67 11.63,189.18 0.01625^29.9 �57.13 �72.83,�32.34 0.001 101.89 25.57,224.65 0.004>30 �79.10 �87.60,�64.79 0.001 127.48 31.92,292.19 0.003

ActivityRestc

Medical visit �13.75 �26.24,0.83 0.064 51.19 24.95,82.92 0.001Medical assistance �2.84 �14.88,10.88 0.668 23.17 4.45,45.23 0.013Surgical activities 6.49 �10.21,26.30 0.470 34.63 9.26,65.88 0.005Food ingestion �14.90 �26.90,�0.93 0.037 37.01 13.63,65.20 0.001Academic activities �2.93 �17.61,14.35 0.721 12.23 �8.20,37.23 0.261Anesthesiology activities �22.30 �41.31,2.83 0.078 55.79 10.77,119.13 0.011

Waitingperiod (days) 0.16 0.06,0.25 0.001 0.09 0.0000,0.19 0.046

aN refers to sample size.The target number of1-hr measurements for the 54 resident physicianswas1,296.bConfidence intervals at 95%.cReference category.dCategories according toWHO (normal18.5^24.9, overweight 25^29.9, and obese>30 kg/m2).


stressor exposure. Likewise, the subjects were asked to re-

cord their activities in a diary, which permitted tracking

variations in diurnal rhythm. Van Amelsvoort et al. [2000]

selected persons from specific populations who were start-

ing a new job, with the average age of this group being

30 years. Young people have less time on the job, and

thus fewer years of exposure to job stressors.

It is important to note that the first-year residents in

our study work 24 hr straight, and that the following

morning they work an 8-hr shift and then rest but this is

when they carry out their academic tasks and prepare for

exams and class presentations. Previous studies have dem-

onstrated evidence that long working hours modify HRV

parameters [Kageyama et al., 1997]. Since long working

hours may be correlated with job demands, that suggested

an analysis of job demands and HRV.

With respect to the analysis of the parameters of low-

frequency power and high-frequency power by sex, we

found lower values in comparison to the values in the gen-

eral population [Task Force of the European Society of

Cardiology and The North American Society of Pacing

and Electrophysiology, 1996]. We also found that men

have higher low-frequency/high-frequency ratio in com-

parison with the value for the population at large. These

results demonstrate that there is a difference by sex with

respect to heart rate variability, and to its low-frequency

power, high-frequency power and low-frequency/high-

frequency ratio components. We stratified the multi-level

analysis by sex for each of the four parameters, and found

significantly different associations by sex in the categories

of the demand–control model, but these findings are not

presented in this article due to the small sample size and

uneven distribution of the sample by sex. An analysis by

sex should be replicated in a larger and more evenly bal-

anced sample. This analysis should be replicated in a larg-

er and more homogeneous sample. It is known that there

are other factors—like sex [Pomeranz et al., 1985; OBrien

et al., 1986; Umetani et al., 1998], age [Stein et al., 1997;

Ramaekers et al., 1998], BMI [WHO, 1995], exercise [De

Meersman, 1993], circadian cycle [Furlan et al., 2000],

and atmospheric pollution [Schneider et al., 2010], which

may modify the HRV and ought to be controlled for in

study design or analysis. In our study, we adjusted for sex,

age, and BMI. We also recorded use of medications that

may have altered the HRV, but this was not included in

the final analysis because it was not statistically significant

in preliminary analyses.

Limitations of the Study

The small sample size of 54 medical residents, and

the imbalance between men (36) and women (18), made it

difficult to use interactive models and reduced the

statistical power of the stratified models. There was also a

disproportional distribution of medical residents by spe-

cialties certain specialialties like radiology, orthodontics

and pathological anatomy had two members each in the

sample, while other specialties had as many as seven, such

as emergency care and general surgery. We evaluated

only heart rate variability and its parameters over the

working day (8 hr) plus the on-call time (16 hr), for a

total of 24 hr. It would have been useful to evaluate the

working day by itself, in order to be able to make compar-

isons. In addition, no questions were included regarding

exhaustion. The study began after 3 months of the aca-

demic cycle for the medical residents, and bio-markers of

stress (such as cortisol levels) were not available. The

questionnaires were administered 3 months following

the beginning of the residency, which suggests that the

doctors had already adapted to the various activities, job

demands, and decision-making. This last factor was

tracked through a waiting-time variable (which is the

time, in days, that each doctor waits, from their entry into

the specialty, up to the day on which their Holter electro-

cardiographic recording began). The strengths of this

study are the repetition of observations for each doctor,

and the extensive tracking for each of the doctors, through

a log where they wrote down each one of the activities

that they carried out.

CONCLUSIONS

Medical residents in their first academic year report-

ing low demand–low control (passive) and high demand–

low control (high strain) jobs had some lower parameters

of heart rate variability, as compared with their colleagues

reporting low demand–high control (low strain) jobs,

over a 24-working hour period. These results suggest that

medical residents work in stressful conditions that may

increase their risk of heart disease, and it is important to

note that they still have three more years ahead of them

before they complete their specialty.

Twenty-four-hour cardiac monitoring is an informa-

tive marker for evaluation of the impact of work stressor

and alteration in the equilibrium of the autonomic nervous

system due workplace factors. Additional research is need-

ed concerning the response of heart rate variability to

acute or chronic stressors, given that the pathophysiologi-

cal mechanisms are not yet clear.

ACKNOWLEDGMENTS

This study was financed by Irving J. Selikoff Interna-

tional Scholar of the Mount Sinai School of Medicine.

The project described was supported in part by Award


Number D43TW000640 from the Fogarty International

Center ITREOH program. The Content is solely the re-

sponsibility of the authors and does not necessarily repre-

sent the official views of the Fogarty International Center

or the National Institutes of Health. Likewise, a great

gratitude to Dr. Moises Selman, Research Director of the

National Institute of Respiratory Diseases Ismael Cosio

Villegas because with his valuable management the reali-

zation of this project was possible. Thanks to our transla-

tor Ken Hansen.

APPENDIX

TABLE A1. Job Demands, JobDecision Latitude and Heart RateVariability Parameters: (SDDNandHigh Frequency Power) in Resident Physicians,Mixed LinearModel,Mexico City, 2007^2008

Variable

SDNNa,b Highfrequencypowera

% 95%ICc Pvalue % 95%CICc Pvalue

DemandsLowd

High 4.81 �1.81,11.94 0.158 13.06 �6.78,37.16 0.213Decision latitudeLowd

High 5.02 �3.02,13.88 0.227 6.32 �15.91,34.46 0.609SexWomenMend 12.17 4.28,20.65 0.002 27.02 2.48,57.44 0.029

Age23^25d

26^27 �4.24 �12.80,5.12 0.362 �10.02 �31.77,18.63 0.45428^36 �18.73 �27.03,�9.48 0.001 �46.24 �60.78,�26.30 0.001

IMCe

<18.5d

18.5^24.9 �36.13 �50.69,�17.27 0.001 �83.95 �92.42,�66.04 0.00125^29.9 �31.89 �47.38,�11.84 0.004 �82.62 �91.79,�63.21 0.001>30 �44.34 �58.61,�25.15 0.001 �92.71 �96.91,�82.79 0.001

ActivityRestd

Medical visit �5.55 �15.53,5.60 0.316 �27.07 �43.86,�5.25 0.018Medical assistance 10.41 �0.54,22.60 0.064 21.32 �4.78,54.60 0.118Surgical activities 4.32 �7.68,17.90 0.498 1.09 �24.15,34.75 0.941Food ingestion �6.56 �15.99,3.92 0.211 �20.45 �37.63,1.58 0.067Academic activities 7.50 �4.60,21.15 0.235 7.44 �18.87,42.31 0.616Anesthesiology activities �14.99 �30.72,4.27 0.119 �33.31 �59.00,8.46 0.103

Waitingperiod (days) 0.04 �0.009,0.09 0.119 0.07 �0.06,0.22 0.348

aN refers to sample size.The target number of1-hr measurements for the 54 resident physicianswas1296.bSDNNstandard deviation of all the NN intervals.cConfidence intervals at 95%.dReference category.eCategories according toWHO (normal18.5^24.9, overweight 25^29.9, and obese>30 kg/m2).


REFERENCES

Akaike H. 1974. A new look at the statistical model identification.IEEE Trans Automat Control AC-19:716–723.

Belkic KL, Landesbergis PA, Schanall P, Baker D. 2004. Is job straina major source of cardiovascular disease risk? Scand J Work EnvironHealth 30(2):85–128.

Butterfield PS. 1998. The stress of residency. A review of the litera-ture. Arch Intern Med 148(6):1428–1435.

Collins S, Karasek R, Costas K. 2005. Job strain and autonomic indi-ces of cardiovascular disease risk. Am J Ind Med 48:182–193.

De Meersman RE. 1993. Heart rate variability and aerobic fitness.Am Heart J 125:726.

Dekker JM, Shouten EG, Klootwijk P, Pool J, Swemme CA,Kromhout D. 1997. Heart rate variability from short electo-cardio-graphic recordings predicts mortality from all causes in middle-agedand elderly men. The Zuptphen Study. Am J Epidemiol 145:899–908.

Diggle PJ, Liang KY, Zeger SL. 1996. Analysis of longitudinal data.Oxford, UK: Claredon Press. p 126–129.

Dishman RK, Nakamura Y, Garcia ME, Thompson RW, Dunn AL,Blair SN. 2000. Heart rate variability, trait anxiety, and perceivedstress among physically fit men and women. Int J Psychophysiol37(2):121–133.

Eller NH, Kristiansen J, Hansen AM. 2010. Long-term effects ofphychosocial factors of home and work on biomarkers of stress. Int JPsychophysiol 79(2):195–202.

TABLE A2. JobDemands, Job Decision Latitude andHeart RateVariability Parameters: (LowFrequency Power and LowFrequency/High FrequencyRatio) in Resident Physicians,Mixed Linear Model,Mexico City, 2007^2008

Variable

Low frequencypowera Low frequency/highfrequency ratioa

% 95%CIb Pvalue % 95%CIb Pvalue

DemandsLowc

High 2.80 8.42,15.43 0.638 �9.01 �19.05,2.61 0.124Decision latitudeLowc

High 25.84 9.39,44.77 0.001 18.62 2.44,37.34 0.022SexWomenc

Men 65.91 45.89,88.66 0.001 31.11 14.71,49.85 0.001Age23^25c

26^27 �10.11 �23.84,6.09 0.208 �0.66 �16.34,17.93 0.93928^36 �35.88 �46.88,�22.60 0.001 18.61 �2.54,44.39 0.089

IMCd

<18.5c

18.5^24.9 �66.71 �78.56,�48.10 0.001 103.39 27.12,225.43 0.00325^29.9 �59.90 �74.33,�37.25 0.001 128.55 43.13,263.27 0.001>30 �80.60 �88.35,�67.69 0.001 163.39 54.09,348.16 0.001

ActivityRestc

Medical visit �11.07 �22.85,2.48 0.105 22.28 2.29,46.19 0.027Medical assistance 2.64 �9.90,16.94 0.695 �18.12 �30.71,�3.41 0.018Surgical activities 9.65 �6.17,28.17 0.246 9.46 �10.06,33.25 0.367Food ingestion �12.42 �23.28,�0.03 0.049 11.41 �5.91,32.00 0.211Academic activities �0.62 �14.88,16.00 0.936 �7.65 �23.95,12.14 0.422Anesthesiologyactivities �19.05 �38.21,6.03 0.125 23.68 �11.52,72.92 0.214

Waitingperiod (days) 0.16 0.07,0.25 0.001 0.08 �0.0005,0.18 0.067

aN refers to sample size.The target number of1-hr measurements for the 54 resident physicianswas1296.bConfidence intervals at 95%.cReference category.dCategories according toWHO (normal18.5^24.9, overweight 25^29.9, and obese>30 kg/m2).


Friedman BH, Thayer JF. 1998. Autonomic balance revisited: Panicanxiety and heart rate variability. J Psychosom Res 44:133–151.

Furlan R, Barbic F, Piazza S, Tinelli M, Seghizzi Malliani A. 2000.Modifications of cardiac aunotomic profile associated with a shiftschedule of work. Circulation 102(16):1912–1916.

Halvorsen R, Palmquist R. 1980. The interpretation of dummy varia-bles in semilogarithmic equations. Am Econ Rev 70(3):474–475.

Hintsanen M, Elovaino M, Puttonen S, Kivimaki M, Koskinen T,Raitakari OT, Keltikangas-Jarvinen L. 2007. Effort–reward imbal-ance, heart rate and heart rate variability: The Cardiovascular Risk inYoung Finns Study. Int J Behav 14(4):202–212.

Kageyama T, Nishikido N, Honda Y, Kurokawa Y, Imai H, Kobaya-shi T, Kaneko T, Kabuto M. 1997. Effects of obesity, current smok-ing status, and alcohol consumption on heart rate variability in malewhite-collar workers. Int Arch Occup Environ Health 69:447–454.

Kang M, Koh S, Cha B, Park J. 2004. Association between Job stresson heart rate variability and Metabolic Syndrome in Shipyard maleworkers. Yonsei Med 45(5):838–846.

Karasek RA. 1985. Job content instrument: Questionnaire and usersguide. Los Angeles: University of South California.

Karasek RA, Theorell T. 1990. Healthy work: Stress, productivity,and the reconstruction of working life. New York, NY: Basic Books.

Karasek RA, Theorell T. 2000. The demand–control–support modeland CVD. In: Schnall PL, Belkic K, Landesbergis P, guest editors.The workplace and cardiovascular disease. Philadelphia, PA: Hanley& Belfus, Inc., p 78–83.

Karasek R, Brisson C, Amick B. 1998. The Job Content Question-naire (JCQ): An instument for internationally comparative assess-ments of psychosocial job characteristics. J Occup Health Psych3(4):322–355.

Kennedy P. 1981. Estimation with correctly interpreted dummy vari-ables in semilogarithmic equations. Am Econ Rev 71(4):801.

Kristal-Boneh E, Raifal M, Froom P, Ribak J. 1995. Hear rate vari-ability in health and disease. Scand J Work Environ Health 21:85–95.

Lombardi F, Makikallio T, Myerburg R, Huikuri H. 2001. Suddencardiac death:role of heart rate variability to identify patients at risk.Cardiovasc Res 50:210–217.

Malliani A, Pagani M, Lombardi F, Cerutti S. 1991. Cardiovascularneural regulation explored in the frequency domain. Circulation84:428–492.

Malliani A, Lombardi F, Pagani M. 1994. Power spectrum analysisof heart rate variability: A tool to explore neural regulatory mecha-nisms. Br Heart J 71:1–2.

Michie S, Sandhu S. 1994. Stress management for clinical medicalstudents. Med Edu 28:528–533.

OBrien IA, OHare P, Corral RJ. 1986. Heart rate variability inhealthy subject: Effects of age and the derivation of normal rangesfor tests of autonomic function. Br Heart J 55:348–354.

Perry M, Osborne W. 2003. Health and wellness in residents whomatriculate into physician training programs. Am J Obstet Gynecol7:670–683.

Pomeranz B, Macualay RJ, Caudill MA, Kutz I, Adam D, GordonD, Kilborn KM, Barger AC. 1985. Assessment of autonomic func-tion in humans by heart rate spectral analysis. Am J Physiol 248:H151–H153.

Ramaekers D, Ector H, Aubert AE, Ruben A, Van de Werf F.1998. Heart rate variability and heart rate in healthy volunteers. Isthe female autonomic nervous system cardioprotective? Eur Heart J19:1334–1341.

Riese H, Van Doornen LJ, Houtman IL, De Geus E. 2004. Job strainin relation to ambulatory blood pressure, heart rate, and heart ratevariability among female nurses. Scand J Work Environ Health30(6):477–485.

Sargent C, Sotile W, Sotile MO, Rubash H, Barrack R. 2004. Stressand coping among orthopaedic surgery residents and faculty. J BoneJoint Surg 86(7):1579–1586.

Schnall PL, Landesbergis PA, Baker D. 1994. Job strain and cardio-vascular disease. Ann Rev Public Health 15:381–411.

Schneider A, Hampel R, Ibald-Mulli A, Zareba W, Schmidt G,Schneider R, Ruckerl R, Couderc JP, Mykins B, Oberorster G, WolkeG, Pitz M, Wichmann HE. 2010. Changes in deceleration capacity ofheart rate and heart rate variability induced by ambient air pollutionin individuals with coronary artery disease. Part Fibre Toxicol 7:29.

Stein PK, Kleiger RE. 1999. Insights from the study of heart ratevariability. Annu Rev Med 50:249–261.

Stein PK, Kleiger RE, Rottman JN. 1997. Differing effects of age onheart rate variability. Am J Cardiology 80:302–305.

Sztajzel J. 2004. Heart rate variability: A noninvasive electrocar-diographic method to measure the autonomic nervous system. SwissMed Wkly 134:514–522.

Task Force of the European Society of Cardiology and the NorthAmerican Society of Pacing and Electrophysiology. 1996. Heart ratevariability: Standards of measurement, physiological interpretation,and clinical use. Circulation 93:1043–1065.

Togo F, Takahashi M. 2009. Heart rate variability in occupationalhealth: A systematic review. Ind Health 47:589–602.

Tsuji H, Venditti FJJ, Manders ES, Evans JC, Larson MG, FeldmanCL, Levy D. 1994. Reduced heart rate variability and mortality riskin an elderly cohort. The Framingham Heart Study. Circulation 90:878–883.

Tsuji H, Larson MG, Venditti FJJ, Manders ES, Evans JC, FeldmanCL, Levy D. 1996. Impact of reduced heart rate variability on riskfor cardiac events. The Framingham Heart Study. Circulation 94:2850–2855.

Umetani K, Singer DH, McCraty R. 1998. Twenty four hour timedomain heart rate variability and heart rate: Relations to age andgender over nine decades, JACC 31:593–601.

Van Amelsvoort LGPM, Schouten EG, Maan AC, Swenne CA,Kok FJ. 2000. Occupational determinants of heart rate variability, IntArch Occup Environ Health 73:255–262.

Van der Doef M, Maes S. 1999. The job demand–control(–support)model and phychological well-being: A review of 20 years of empir-ical research. Work Stress 13:87–114.

Vrijkotte TG, Van Doornen LJ, de Geus EJ. 2000. Effects of workstress on ambulatory blood pressure, heart rate, and heart rate vari-ability. Hypertension 35(4):880–886.

WHO. 1995. Physical status: The use and interpretation of anthro-pometry. Geneva: World Health Organization.

Yeragani VK, Pohl R, Berger R, Balon R, Ramesh C, Glitz D,Srinivasan K, Weinberg P. 1993. Decreased heart rate variability inpanic disorder patients: A study of power-spectral analysis of heartrate. Psychiatry Res 46:89–103.


job strain and heart rate variability in resident physicians within a general hospital

Documents