Associations between job demands, work-related strain and perceived quality of care: a longitudinal study among hospital physicians

Abstract

Objective

Drawing on a sample of hospital physicians, we attempted to determine prospective associations between three job demands, work-related strain and perceived quality of care.

Design

Longitudinal follow-up study with with a 1-year time lag.

Setting

Physicians of two acute-care hospitals in Germany (one general urban and one children's hospital).

Study participants

Ninety-five physicians filled out a standardized questionnaire.

Main outcomes measures

Physicians’ evaluations of quality of care at both waves.

Results

Our results support the hypothesis that job demands directly influence quality of care irrespective of strain. Specifically, high social stressors (β = −0.15, P = 0.036) and time pressure (β = −0.19, P = 0.031) were associated with decreased quality of care over time. We additionally observed reversed effects from quality of care at baseline to time pressure at follow-up (β = −0.35, P = 0.006). Contrary to expectations, physicians’ work-related strain did not mediate the job demands–quality of care-relationship, nor were strain-to-stressor effects observed.

Conclusions

Our results corroborate that hospital work environments with high demands have a direct impact on physician-perceived quality of care. In turn, poor care practices contribute to increased job demands. Our findings also emphasize that further understanding is required of how physicians’ workplace conditions affect job demands, well-being, and quality of care, respectively.

job demands, time pressure, social stressors, quality of care, strain, longitudinal, physicians

Introduction

Hospital physicians often report alarmingly high levels of mental health conditions [1, 2]. The impact of an adverse work environment in the hospital and associated outcomes is also considerable in terms of productivity or medical malpractice [3, 4]. Accordingly, the identification of determinants of physician well-being as well as their joint effect on health care quality spurs increased attention [2, 5, 6, 7].

The Job Demands-Resources (JD-R) model proposes that work-related strain partially mediates the relationship between job demands and organizational outcomes [8]. Job demands require sustained efforts from employees. Thus, excessive job demands lead to strain or exhaustion and ultimately performance and organizational outcomes suffer [8–10]. The central tenets of the JD-R model have been expanded by the notion that employee well-being might also reversely influence work characteristics, either through changed perceptions or actual job changes due to changes in mental well-being [9, 11]. This has been described as the reversed strain-to-stressor effect [12]. Moreover, reciprocal associations between job demands and job resources, and work-related mental well-being were observed [13].

This study contributes to the literature on physicians’ work life and quality of care in various ways. We investigate the prospective effects of three prevalent job demands on physicians’ work-related strain and perceived quality of care. This study also contributes to the growing evidence base on health care professionals’ work stress and quality of care outcomes [6, 14]. Whereas cross-sectional studies have provided preliminary confirmation, respective longitudinal research is scarce [15, 16].

Job demands, work-related strain and quality of care in hospital physicians

We hypothesized that job demands in a hospital have direct effects on physicians’ work-related strain and perceived quality of care. These relationships also hold true in the reverse direction from quality of care to strain and demands. Firstly, we assume a direct impact of job demands on quality of care outcomes (Hypothesis 1). Mounting evidence suggests that the hospital work environment has an immediate influence on the quality of care provided [14]. These effects may be attributed to various processing changes in individuals that occur under stressor exposure, e.g. attentional narrowing. These processes adversely affect individual decision making or behavior and consequently increase the risk of an inferior quality of care [4, 17]. Similarly, the systems approach to quality and safety in health care differentiates between immediate threats to high-quality performance, such as employee fatigue and strain, and systemic threats like staffing numbers. Inherent weaknesses of the care system thus manifest themselves in work stressors and promote errors and low-quality performance of employees [18, 19].

Second, we assume a mediated effect of job demands on quality of care through physicians’ work-related strain, operationalized as irritation. Excessive job demands require individual coping processes that exceed physicians’ mental or physical capacities and thus trigger psychological strain (Hypothesis 2). Work-related strain, in turn, adversely affects organizational performance, i.e. quality of care (Hypothesis 3). Irritation thus acts as a mediator between job demands and physician-perceived quality of care [20]. Third, we propose reciprocal associations between job demands, irritation and quality of care [20]. In addition to direct causal effects in Hypotheses 1–3, additional reversed strain–stressor relationships are assumed where strained physicians respond more irritable towards their occupational environment causing perceptions of higher patient demands, more time pressure and deteriorated social interactions at work (Hypothesis 4) [21]. The same mechanism is assumed for physician-perceived quality of care and job demands (Hypothesis 5) [12, 20].

Methods

Design and procedure

A prospective study with two waves over a 1-year time lag was conducted as part of a larger improvement project (cf. [21]). The study was approved by the Ethics Committee of the Medical Faculty, Munich University (124/07).

All study variables were assessed via standardized questionnaires. At baseline (T1), letters of informed consent and return envelopes were distributed to 300 physicians from two public German hospitals (one children's and one general hospital). Around 159 participated at T1 (response rate of 53.0%). A total of 142 (47.3%) returned surveys at follow-up (T2) which were matched to baseline questionnaires using identification codes.

Sample

Finally, n = 95 physicians completed questionnaires at both waves (with 49 men, 51.6%). Average age was M = 39.83 years (SD = 9.05) and mean organizational tenure was M = 8.90 years (SD = 7.59). Fourteen physicians (14.7%) were working part-time with less than 40 h per week. Fourteen were heads of department (14.7%), 16 were senior physicians (16.8%), 30 were specialists (31.6%), and 33 were junior physicians in specialty training (34.7%). The self-reported average number of patients seen per day was 15.31 (SD = 9.91) and the amount of time spent in patient contact per day ranged from 15 to 600 min with an average of 205.50 min (SD = 114.32).

Measures

Three job demands were assessed with an established tool for the assessment of work stressors in hospitals [TAA-KH-S; 22, 23]. This tool has been validated for use by physicians [24]. A record of the used items in German language can be found here [24]. Social stressors were measured with three items which assess conflicting relationships with direct colleagues, supervisors, and coworkers in other wards (e.g. ‘Collaboration with coworkers within this department is a frequent burden’). Patient demands were measured with a five-item scale whereby each item describes common emotional challenges when dealing with patients (e.g. ‘Working with dying patients is a frequent burden’). The time pressure scale included five items (e.g. ‘One frequently has to hasten and yet cannot complete the work tasks’). All response formats ranged from 1 = not at all to 5 = to a great extent.

Physicians’ strain

Work-related irritation is defined as short-term mental strain comprising a cognitive and emotional dimension and was assessed with an eight-item scale [25]. Example items are ‘Even at home I can't stop thinking about problems at work’ (cognitive irritation) and ‘I sometimes get grumpy when others approach me’ (emotional irritation). The answer format ranged from 1 = strongly disagree to 7 = strongly agree.

Quality of care

A three-item scale measured physicians’ perceptions of the impairments in clinical work and quality of care due to work conditions in hospitals [4, 17]. An example item is ‘Adverse working conditions frequently lead to errors’. A frequency scale was applied from 1 = not at all to 5 = to a great extent.

Control variables

Type of contract (1 = full-time, 2 = part-time) and time spent in direct patient care were included as control variables as they showed significant associations with study variables (cf. Table 1).

Table 1

Means (M), standard deviations (SD), internal reliabilities (Cronbach's Alpha on the diagonal) and intercorrelations (Pearson's r) of study variables

Variable	M	SD	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
1 Gender (male/female)	–	–
2 Age (years) T1	39.83	9.05	−0.19
3 Tenure (years) T1	8.90	7.59	−0.12	0.81**
4 Type of contract T1 (full-/part-time)	–	–	0.38**	0.11	0.08
5 Patient contact (minutes) T1	205.50	114.32	0.09	0.14	0.08	0.02
6 Number of patients T1	15.31	9.91	−0.23*	.07	0.11	−0.27*	0.09
7 Social stressors T1	2.46	0.68	0.03	0.11	0.02	0.14	−0.15	0.06	0.63
8 Social stressors T2	2.49	0.79	0.03	0.06	0.03	0.18	−0.05	−0.11	0.47**	0.78
9 Patient demands T1	3.42	0.77	0.09	−0.05	0.07	−0.12	−0.11	0.00	0.24*	0.05	0.80
10 Patient demands T2	3.31	0.73	0.01	−0.11	−0.05	−0.29**	0.04	0.18	0.06	0.07	0.55**	0.79
12 Time pressure T2	3.56	0.67	−0.07	0.01	−0.02	−0.20	−0.08	−0.01	0.26*	0.28**	0.29**	0.35**	0.53**	0.77
13 Irritation T1	3.63	1.26	0.07	−0.05	−0.12	0.02	−0.06	−0.01	0.32**	0.19	0.31**	0.25*	0.38**	0.25*	0.89
14 Irritation T2	3.63	1.18	0.07	−0.08	−0.18	0.04	0.03	−0.08	0.22*	0.24*	0.24*	0.28**	0.29**	0.28**	0.73**	0.88
15 Quality of care T1	2.74	0.94	0.05	0.22*	0.15	0.21*	0.31**	−0.10	−0.42**	−0.19	−0.51**	−0.41**	−0.67**	−0.53**	−0.38**	−0.27**	0.85
16 Quality of care T2	2.95	0.89	0.08	−00.13	0.09	0.21*	0.11	−0.12	−0.37**	−0.37**	−0.40**	−0.52**	−0.43**	−0.62**	−0.28**	−0.32*	0.64**	0.85

Notes: T1: baseline assessment; T2: follow-up assessment; n = 95, *P < 0.05, **P < 0.01; variables 7–12 and 15–16: 5-point scale ranging from 1: ‘not at all’ to 5: ‘to a great extent’; variables 13–14: 7-point scale ranging from 1: ‘strongly disagree’ to 7: ‘strongly agree’.

Statistical analyses

We used cross-lagged path models (maximum-likelihood estimation) with manifest variables to determine the hypothesized pathways; using Lavaan (0.5–14) package, R environment [26]. Chi-square (χ²) and relative chi-square (χ²/df) statistics were examined to determine goodness-of-fit. Correcting for sensitivity of the χ²-value towards sample size, further fit indices were applied: levels of >0.90 for Tucker–Lewis Index (TLI) and Comparative Fit Index (CFI) and levels of <0.05 for Root Mean Square Error of Approximation (RMSEA) indicate good model fit.

In order to test a potential mediation in a two-wave longitudinal design, we followed recommendations by Taris and Kompier [27]. Separate analyses were conducted to test prospective relationships between predictor at T1 and outcome at T2, between predictor at T1 and mediator (i.e. work-related strain) at T2 and, finally, mediator at T1 and outcome at T2. A step-wise approach of structural models was used to identify cross-lagged associations over time:

• Baseline model: The baseline model included temporal stabilities between corresponding variables from T1 to T2. All consecutive models included autoregressive paths.

• Direct effect model: This model extended the baseline model with a path from job demands at T1 to quality of care at T2.

• Reversed effect model: To test reversed effects, a path from quality of care at T1 to the respective demands variable at T2 was added to the baseline model.

• Prospective effect model: It extended the baseline model with lagged paths from the respective demand at T1 to irritation at T2 and from irritation at T1 to quality of care at T2.

• Reversed effect model: A reversed prospective path from irritation at T1 to the respective demand at T2 was added to the baseline model.

• Reciprocal effect model: This model was established by adding a path from job demands at T1 to irritation at T2 and the reversed path from irritation at T1 to job demands at T2 to the baseline model.

Finally, regression paths from the two control variables at T1 were added to the respective predictor, mediator and outcome variable at T2.

Results

Panel attrition

Sociodemographic characteristics between the final sample and panel dropouts showed no significant differences. Moreover, no differences between the two groups were found for patient demands, time pressure, irritation and quality of care. Only, social stressors were significantly lower for the panel compared to dropouts at T1 [M = 2.20, SD = 0.78; t(156) = 2.20, P = 0.03].

Descriptive statistics

Study variables were stable over time except mean quality of care which significantly increased over time [t(93) = −2.35, P = 0.02] (cf. Table 1). Control variables were also stable over the 1-year time lag. Job demands and irritation were positively related, whereas quality of care was negatively associated with job demands and irritation.

Hypotheses testing

Table 2 presents indices for the baseline and competing cross-lagged path models, subdivided by the three job demands.

Table 2

Structural models of social stressors, patient demands, time pressure, irritation and quality of care

	χ²	df	χ²/df	CFI	TLI	RMSEA [CI]	Δχ²	Δdf
Social stressors
M0i: Baseline model	6.60	6	1.10	1.00	0.99	0.033 [0.000–0.141] n.s.	To M0i
M1i: Direct effect model	2.34	5	0.47	1.00	1.04	0.000 [0.000–0.091] n.s.	4.26*	1
M2i: Reversed direct effect model	6.37	5	1.27	0.99	0.98	0.054 [0.000–0.161] n.s.	0.23 n.s.	1
M3i: Prospective effect model	5.60	4	1.40	0.99	0.97	0.065 [0.000–0.180] n.s.	1.00 n.s.	2
M4i: Reversed effect model	6.47	5	1.29	0.99	0.98	0.056 [0.000–0.162] n.s	0.13 n.s.	1
M5i: Reciprocal effect model	5.63	4	1.41	0.99	0.97	0.066 [0.000–0.180] n.s	0.97 n.s.	2
Patient demands
M6i: Baseline model	5.53	6	0.92	1.00	1.01	0.000 [0.000–0.128] n.s.	To M6i
M7i: Direct effect model	2.88	5	0.58	1.00	1.03	0.000 [0.000–0.106] n.s.	2.65 n.s.	1
M8i: Reversed direct effect model	3.74	5	0.75	1.00	1.02	0.000 [0.000–0.123] n.s.	1.79 n.s.	1
M9i: Prospective effect model	4.48	4	1.12	1.00	0.99	0.036 [0.000–0.164] n.s.	1.05 n.s.	2
M10i: Reversed effect model	5.34	5	1.07	1.00	1.00	0.027 [0.000–0.148] n.s.	0.19 n.s.	1
M11i: Reciprocal effect model	4.63	4	1.16	1.00	0.99	0.041 [0.000–0.166] n.s.	0.90 n.s.	2
Time pressure
M12i: Baseline model	12.30	6	2.05	0.97	0.93	0.106 [0.000–0.191] n.s.	To M12i
M13i: Direct effect model	7.75	5	1.55	0.99	0.97	0.077 [0.000–0.176] n.s.	4.55*	1
M14i: Reversed direct effect model	4.94	5	0.99	1.00	1.00	0.000 [0.000–0.143] n.s.	7.36**	1
M15i: Prospective effect model	9.44	4	2.36	0.97	0.91	0.121 [0.000–0.223] n.s.	2.86 n.s.	2
M16i: Reversed effect model	12.29	5	2.46	0.96	0.91	0.125 [0.035–0.216] n.s.	0.01 n.s.	1
M17i: Reciprocal effect model	9.57	4	2.39	0.97	0.91	0.122 [0.010–0.224] n.s.	2.73 n.s.	2

Notes: n = 95; *P < 0.05, **P < 0.01; df = degrees of freedom; CI = 90% Confidence Interval; Δχ² = scaled chi-square difference; Δdf = difference in degrees of freedom.

All three baseline models (M_0i, M_6i and M_12i) showed acceptable fit. Significant auto-regression paths indicating high stabilities over time of job demands, irritation and quality of care were observed. Hypothesis 1 proposed a lagged effect from job demands to quality of care. The model M_1i was significantly superior to the baseline model [∆χ²(1) = 4.26, P = 0.04] showing that high social stressors at T1 were associated with decreased quality of care at T2 (β = −0.15, P = 0.036). No prospective effects of patient demands on quality of care were identified. With regard to time pressure, the direct effect model M_13i was superior to the baseline model [∆χ²(1) = 4.55, P = 0.03]. Thus, high time pressure at T1 was associated with diminished quality of care at follow-up (β = −0.19, P = 0.031) (cf. Fig. 1).

Figure 1

Comprehensive model of all individually significant pathways between physicians’ job demands (social stressors/time pressure/patient demands), work-related strain and quality of care (n = 95 physicians, 1-year time lag).

Next, potential reversed effects of impaired quality of care on job demands were analyzed. No effects were observed for quality of care T1 to social stressors T2 as well as to patient demands T2 (M_8i). In regard to time pressure, the reversed effect model (M_14i) showed a significant better fit than the baseline model (M_12i) [∆χ²(1) = 7.36, P = 0.007]. High physician-perceived quality of care negatively predicted time pressure at follow-up (β = −0.35, P = 0.006; cf. Fig. 1).

Next, we tested the mediation hypothesis, postulating that job demands have a direct effect on irritation (H2), whereas irritation exerts a direct negative effect on quality of care (H3). All mediation models for the three job demands (M_3i, M_9i  and M_15i) showed no superior fit, thus rejecting hypotheses 2 and 3. Then, we tested if irritation at T1 has positive lagged effects on the three job demands at T2. All reversed effects models (M_4i, M_10i and M_16i) showed no superior model fit compared to the baseline model, thus rejecting hypothesis 4. Finally, we tested reciprocal associations between job demands and irritation (Hypothesis 5). For all three demands, the reciprocal effect models did not yield increased model fit compared to the baseline model (M_5i, M_11i and M_17i).

In an additional step we added two control variables to the above reported models. After including time spent in patient contact, the pathways from social stressors T1 to quality of care T2, and from time pressure T1 to quality of care T2 became nonsignificant in the respective direct models. Quality of care T1 still negatively determined time pressure T2 (β = −0.31, P = 0.012). In regard to the inclusion of the type of contract as a control variable, paths from social stressors T1 (β = −0.15, P = 0.036) and time pressure T1 (β = −0.19, P = 0.031) to quality of care T2 in the direct models as well as the reversed path from quality of care T1 to time pressure T2 (β = −0.34, P = 0.008) in the reversed model remained significant. Additionally, working part-time at T1 positively determined an increase in social stressors at T2 (β = 0.06, P = 0.02).

Discussion

Physician well-being and performance are of concern to various stakeholders [6]. We tested prospective effects between job demands, work-related strain and quality of care across time in hospital physicians. We observed that high social stressors and time pressure contributed to inferior quality of care over time, while high quality of care was associated with lower time pressure at follow-up.

Effects of job demands on quality of care

A direct effect of high time pressure and social stressors on inferior quality of care was observed. Our results thus corroborate the detrimental effects of clinical systemic factors on quality of care, irrespective of physicians’ work-related strain. Hence, psychosocial working conditions influence physician-perceived quality of care directly [14]. The systems design approach maintains that various structural determinants (e.g. staffing levels and work overload) and processes (e.g. procedural protocols) facilitate or hinder health professionals from performing effectively. They provide conditions within which mistakes and lapses occur [28]. According to the systems design approach, alongside individual factors such as knowledge and experience of employees, the deliberate design of the work environment enables employees to perform their tasks responsibly and effectively [28–30].

The nonsignificant results of the proposed mediation mechanisms between job demands, irritation and quality of care deserve consideration. Temporal stability of irritation over time was high (β = 0.74). High stability and ‘multifactorial determination’ of strain variables [31] often result in the predictor's failure to account for additional variance in the outcome strain variable [27]. Furthermore, nonsignificant mediation results may be explained by the 1-year time lag which is, however, consistent with time lags in the majority of similar studies [31]. Theoretically, an interval that is too short may not allow enough time for stressor effects to be reflected in a change of work-related strain. When applying overlong time lags, employees may adjust to stressors, thus negating any previous effects on mental well-being [12].

Reversed effect of quality of care on time pressure

High quality of care was associated with reduced time pressure at follow-up. One post-hoc explanation may lie within increased levels of fulfillment gained from successful performance [32]. When physicians are successful at what they do, they may be better able to cope with future demands. An alternative explanation may lie in physicians’ ambition for high standards in which low quality of care is viewed as unacceptable. In a deficient work environment, characterized by high work overload, physicians are required to cover up for shortcomings. Adding to an already tight schedule, system failures may intensify physicians’ time pressure. However, these assumptions need to be examined in future studies.

We controlled for potential confounders with the consequence that direct effects of job demands on irritation and quality of care became nonsignificant. Results also indicate that working part-time increases levels of social stressors. Part-time physicians may not be able to maintain close social ties that full-time employed coworkers possibly do.

Our study contributes to the growing literature on the multiple sources of system-related stress that physicians face in the hospital environment. Our attempt to elicit single and shared effects of job demands and work-related strain emphasize that adverse working conditions have a direct effect on impaired quality of care. Moreover, our study provides a starting point for further investigations into the balance of system stress versus physician's work-related stress. However, further studies need to elaborate the role of individual stress or adverse well-being on the job in the context of systemic and work-related stressors in hospitals [14].

Limitations

Firstly, we disregarded potentially buffering job resources [9]. Physicians may apply effective job resources to cope with high demands and thus reduce the adverse effects on their mental well-being [8]. The choice of a 1-year time lag between study waves is debatable and may have been too long to observe immediate changes in physicians’ strain levels. However, a 1-year time lag has been shown to combine practicality and adequacy to capture significant time-lagged effects [13]. Our sample size is rather small. We therefore used manifest variables to test the hypotheses, which is methodologically inferior to latent-variable modeling. Although we used an established procedure to test for potential mediation between study variables, a more thorough test of mediation effects requires a three-wave study [27]. We cannot exclude potential selection bias such that burned-out physicians refused to participate at follow-up. Future studies should therefore include burnout measures to gain further insights into the interplay of physicians’ job demands, work-related strain, and quality of care. Although the dropout analysis did not yield significant differences in study variables, we account for a possible non-response bias. The external validity of our findings concerning other clinical environments might be limited. Two job demands differed between hospitals, such that pediatricians reported significantly less social stressors (M = 2.27, SD = 0.60) but more patient demands (M = 3.60, SD = 0.81). All other main variables did not differ between both institutions. Our outcome measure does not equate to conventional (objective) quality of care evaluations. In fact, our measure reports physicians’ evaluation of the impact of adverse contextual conditions on compromised quality of care. Therefore, we cannot entirely exclude a potential confounding between determinant variables and the outcome.

Practical implications

Firstly, hospitals need to expand their understanding of workplace characteristics that promote physician performance and quality of care [6, 10, 14]. Interventions to reduce time pressure include decreasing competing demands, reducing bureaucratic duties, and increasing staffing levels. Our observations on social stressors implicate that once they become an actual burden to physicians, their quality of care is affected [17]. Weaknesses within the hospital system which promote errors and poor work quality need to be addressed accordingly [28, 29].

Conclusions

We found direct and reversed effects between job demands and quality of care over time in a sample of hospital physicians. Social stressors and time pressure are important determinants of physician-perceived quality of care. Additionally, physicians who indicated a high quality of care also perceived reduced time pressure at follow-up. Taking a critical look at and adapting the working environment of physicians is able to reduce the level of job demands and can significantly improve the quality of care.

Acknowledgments

The results reported were part of the Master thesis requirements of Tanya Krämer (Munich University).

Funding

This work was supported by the Munich Centre of Health Sciences (MC-Health).