diabetes research and clinical practice 105 (2014) 382–390
Contents available at ScienceDirect
Diabetes Research and Clinical Practice jou rnal hom ep ag e: w ww.e l s e v i er . c om/ loca te / d i ab r es
Do physicians with diabetes have differences in dialysis use and survival than other patients with diabetes Shang-Jyh Chiou a,b, Pei-Tseng Kung b, James M. Naessens c, Kuang-Hua Huang d, Yu-Chia Chang b, Yueh-Hsin Wang d, Wen-Chen Tsai d,* a
Department of Health Care Management, National Taipei University of Nursing and Health Sciences, No. 89, Nei-Chiang Street, Taipei 10845, Taiwan, ROC b Department of Healthcare Administration, Asia University, 500, Lioufeng Road, Wufeng, Taichung 41354, Taiwan, ROC c Division of Health Care Policy and Research, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States d Department of Health Services Administration, China Medical University, No. 91, Hsueh-Shih Road, Taichung 40402, Taiwan, ROC
article info
abstract
Article history:
Aims: To assess whether the increased knowledge and resources available to physicians led
Received 5 February 2014
to differences in dialysis and survival rates between physicians and non-physician patients
Received in revised form
with diabetes.
10 April 2014
Methods: All newly diagnosed (1997–2009) type 2 diabetes patients aged 35 years from the
Accepted 5 July 2014
National Health Insurance Program of Taiwan database were included. After propensity
Available online 19 July 2014
score matching (1:10), we estimated the relative risk of dialysis and death using Cox proportional hazards model adjusted for demographic characteristics and comorbidities.
Keywords:
Results: Physicians with diabetes were more likely to start dialysis than general patients,
Diabetes
with a 48% increased hazard risk (HR) (P = 0.006). Physicians with diabetes had significantly
Dialysis
lower risk of death (HR: 0.88; P = 0.025). However, those requiring dialysis had a non-
Physician with diabetes
significant increased risk of death (HR: 1.19). There was an increased HR for death in older
Survival analysis
physicians (HR: 1.81; P < 0.001) and those with cancer or catastrophic illness. The HR of dialysis (7.89; P < 0.0001) increased dramatically with increasing Charlson Comorbidity Index scores. Conclusions: Physicians with DM survived longer than other patients with diabetes, likely benefiting from their professional resources in disease control and prevention. Nonetheless, they displayed no advantage from their medical backgrounds compared with the general patients if they developed end stage renal disease. # 2014 Elsevier Ireland Ltd. All rights reserved.
* Corresponding author. Tel.: +886 422073070; fax: +886 422028895. E-mail addresses:
[email protected],
[email protected] (W.-C. Tsai). http://dx.doi.org/10.1016/j.diabres.2014.07.004 0168-8227/# 2014 Elsevier Ireland Ltd. All rights reserved.
diabetes research and clinical practice 105 (2014) 382–390
1.
Introduction
There are numerous studies related to health seeking and illness behaviors, the majority of which focus on specific types of patients. The Andersen behavioral model [1], the health belief model [2], and the general theory of help-seeking [3] describe the basic foundation for the determinants of different diseases. A limited number of studies have examined the behavior of disease-affected health care providers to determine whether they are influenced by their medical knowledge. With regard to physicians as patients, most studies focus on economic or social theories to explain the determining factors, medical utilization, and optimal structures for accessibility, affordability, and accountability of healthcare. These studies have employed qualitative methods or small group surveys to gather information and have shown that physicians may be more aware of their physical situation [4] and have more sources of help than the general population of patients. In general, the slightly lower mortality rate or the below-average utilization [5,6] of health care by physicians reflects their personal health choices [7,8], which is most likely attributable to their high socioeconomic status (SES) and knowledge. However, a number of studies have found that there is no protective effect for health care workers who are frequently exposed to ill patients [9], and studies have emphasized the high prevalence of mental health issues [10], such as suicide [11] and substance abuse [12], among doctors. Whether physicians have different health-seeking behavior because of their professional status and whether that type of behavior affects their health status have not been systematically studied. Although doctors prefer to be treated by other physicians and the role conflict inherent in being a doctor-patient and a doctor’s doctor has been discussed [13], studies in Norway found that 75% of physicians treat themselves. This self-treatment behavior may lead to a number of adverse consequences, such as a delayed diagnosis or a worsening illness [14]. How physician health behaviors and illness affect relevant factors warrant further study. Diabetes is highly prevalent worldwide and is the primary cause of end-stage renal disease (ESRD), accounting for 20–40% of cases requiring for peritoneal dialysis (PD) or hemodialysis (HD) [15–17]. Studies have shown that patients with diabetes can have a higher quality of life without suffering from high health care expenditures if they follow medical advice concerning diet, medication, and lifestyle modification [18,19]. We were interested in determining whether physicians with more healthcare resources and professional knowledge have better disease outcomes. This study compared physicians and general patients to determine whether the survival rate and the risk of ESRD are different between the two groups. This information may provide a foundation for a future discussion regarding the occupational health in health promotion, prevention and protection.
2.
Subjects and methods
2.1.
Study background
This study was conducted based on data from the National Health Insurance Program (NHI) of Taiwan and used a specific
383
diabetes mellitus (DM) database that included all nationwide DM patients with a comprehensive medical record from 1997 to 2009. The NHI claims database contains the healthcare data of more than 99.3% of the population of Taiwan and includes comprehensive information, such as demographic data, dates of clinical visits, diagnostic codes, details of prescriptions, and medical expenditures. The comprehensiveness and accuracy of the NHI database have been confirmed by the Department of Health and the Bureau of NHI, and the database has been used in numerous studies [20,21]. This study was approved by China Medical University’s Institutional Review Board (IRB No. CMU-REC-101-012).
2.2.
Study populations with covariates
We selected newly diagnosed type 2 DM patients from 1998 to 2005 to calculate the likelihood of dialysis and survival hazard. The recently diagnosed DM patients were defined as having at least three outpatient visits or one hospital admission (ICD: 250.xx or A181) within one year [22]. This study excluded other types of DM patients and those who were on dialysis prior to the DM diagnosis. For the physician identification, we used the registry for medical personnel (PER) from the NHI from 1997 to 2009, which included physician registry information for Western and Chinese medicine and enabled us to classify the DM patients as general patients or physicians by linking the two databases. The physicians who have a record in the PER file would not belong to the general patient group. We excluded individuals in the PER file who had any type of medical professional license, including dentists, nurses, nutritionists, and physical therapists. We chose subjects who were diagnosed with DM age 35 years. There were 1,064,005 newly diagnosed DM patients registered from 1998 to 2005; we included 1,006,972 cases after applying the above exclusion criteria and excluding patients with missing data. For the dialysis patients, we included the records from the patients with a chronic kidney disease (CKD) diagnosis with an ICD-9 code of 585 and at least three months of continuous dialysis claims records within the study period, and we excluded the cases with dialysis for acute renal disease. The parameters included in the analysis were gender, age, residential urbanization, the Charlson Comorbidity Index (CCI) of comorbid conditions, the presence of cancer, and other catastrophic situations (e.g., illness or injury). These determining factors were used to identify factors related to the relative risk of dialysis and survival.
2.3.
Study participants
For comparison, we used the propensity score matching (PSM) method with a ratio of 1:10, which is applied widely in the health care field [23,24], to account for selection bias and obtain better participation effects on outcome compliance. This result was calculated by logistic regression using the 6 covariates listed in Table 1. Using greedy methods for 1:10 matches, we chose a PS of 0.00000001 and the closest (smallest) number as the matching group. In Table 1, we display the variables used for matching and compared the outcome prior to and after matching. Before matching, approximately 2024 physicians and 955,448 general patients
384
diabetes research and clinical practice 105 (2014) 382–390
Table 1 – The characteristics of general patients with diabetes and physicians with diabetes over 35 years old (prior and after matching). Variable
Unmatched Total
Total Gender Female Male Age (years) at DM diagnosed 35–44 45–54 55–64 65–74 75 Mean age of DM diagnosed Urbanization of residence area Level 1 Level 2 & 3 Level 4 & 5 Level 6 & 7 Charlson Comorbidity Index 0 1–2 3–4 5 Cancer No Yes Other catastrophic illness/injury No Yes
General patients
1:10 matched Physicians
N
%
N
%
N
%
957,472
100.00
955,448
99.78
2024
0.22
Total
General patients
Physicians
N
%
N
%
N
%
20,251
100.00
18,410
90.91
1841
9.09
769 19,482
3.80 96.20
692 17,718
3.76 96.24
77 1764
4.18 95.82
451,812 505,660
47.19 52.81
451,735 503,713
47.28 52.72
77 1947
3.80 96.20
0.430
<0.001
116,785 12.20 247,829 25.88 239,076 24.97 217,900 22.76 135,882 14.19 59.84 12.66
116,471 12.19 247,240 25.88 238,677 24.98 217,558 22.77 135,502 14.18 59.84 12.66
314 15.51 589 29.10 399 19.71 342 16.90 380 18.77 59.20 13.513.58
3301 16.30 6497 32.08 4527 22.35 2870 14.17 3056 15.09 57.40 12.95
2996 16.27 5914 32.12 4141 22.49 2601 14.13 2758 14.98 57.33 12.91
305 16.57 583 31.67 386 20.97 269 14.61 298 16.19 58.04 13.313.32 0.088
<0.001
260,139 432,845 173,541 90,947
27.17 45.21 18.12 9.50
259,403 431,926 173,275 90,844
27.15 45.21 18.14 9.51
736 919 266 103
36.36 45.41 13.14 5.09
8087 9436 1992 736
39.93 46.60 9.84 3.63
7382 8582 1785 661
40.10 46.62 9.70 3.59
705 854 207 75
38.29 46.39 11.24 4.07
0.006
403,619 331,549 138,570 83,734
42.15 34.63 14.47 8.75
402,707 330,847 138,305 83,589
42.15 34.63 14.48 8.75
912 702 265 145
45.06 34.68 13.09 7.16
930,035 27,437
97.13 2.87
928,076 27,372
97.14 2.86
1959 65
96.79 3.21
0.117
9255 7376 2433 1187
45.70 36.42 12.01 5.86
8429 6725 2196 1062
45.78 36.53 11.93 5.76
826 651 237 127
44.87 35.36 12.87 6.90
19,698 553
97.27 2.73
17,815 495
97.31 2.69
1783 58
98.85 3.15
0.386
0.278
0.762
0.002
932,419 25,053
97.38 2.62
930,425 25,023
97.38 2.62
1994 30
98.52 1.48
Statistics
We described the basic demographics of the physicians and general patients with DM, as well as in those with DM who were receiving dialysis. We developed Cox proportional hazards models to estimate the relative risk of dialysis and death after the DM diagnosis. Although we did not link to the Cause of Death Registry because of an inability to match patients between the databases, we used insurance coverage drop-out as the date of death, a method that was introduced by Lien et al. [25] and is an effective proxy. In the Cox model for dialysis, we censored observations at the end of 2009 or if the
P-value
0.399
<0.001
had DM from 1998 to 2005. The majority of the variables (basic characteristics) were significantly different between the two groups. After using the PS matching method with a ratio of 1:10, there were 1881 physicians and 18,654 general patients remaining. There were no significant differences in the covariates after matching.
2.4.
P-value
19,986 265
98.69 1.31
18,171 239
98.70 1.30
1815 26
98.59 1.41
subject died without dialysis. After adjusting for patient characteristics, we compared the relative risk of dialysis and death between the general patients with diabetes and the physicians with diabetes. All of the analyses were conducted using SAS 9.2 (SAS Institute, Cary, NC, USA).
3.
Results
The characteristics of the subjects were not significantly different after matching, as shown in Table 1. Among our selected matches, DM diagnoses occurred at an average age of 58.0 years in the physician group, which was similar to that of the general patient group (57.3 years; P = 0.430). Both groups had small numbers of patients with moderate or severe co morbidities (CCI 3–4 and over 5), as well as with cancer and catastrophic illnesses. Matched physicians were slightly younger than unmatched physicians, but appeared very similar on other characteristics.
385
diabetes research and clinical practice 105 (2014) 382–390
Table 2 provides the patient characteristics of the patients with diabetes in both groups who received dialysis. From our calculations, the physicians with DM had a slightly higher rate of dialysis than did the general patient group with diabetes (3% vs. 2%, P = 0.026; data not shown). Comparing the DM cases with/without dialysis, the age distribution was significantly older in the dialysis group with an average age of DM diagnosis 7 years greater than in the non-dialysis group (64.2 vs. 57.7; P < 0.001). Those who used dialysis had a higher CCI score (P < 0.001), with more severe comorbid conditions. Apart from age, the CCI score and more other catastrophic illnesses, there were no significant differences in the other variables between the groups. In Table 3, we compare the HRs of dialysis or death between the diabetic physician patient and the general patient groups with diabetes. The physicians with diabetes were more likely to start dialysis services than the general patients with diabetes with a 48% increase [HR: 1.48; 95% confidence interval (CI): 1.12–1.95]. A significant difference in the hazard risk of dialysis between the two groups was seen after adjustment (P = 0.006; Fig. 1). The physicians with DM (with/without dialysis) had a significantly lower risk of death (HR: 0.88; 95% CI: 0.78–0.98) than the general patients with diabetes. This finding indicates that the physicians with DM had a higher survival rate than the general patients with this disorder, as illustrated by the survival curves of the two groups (P = 0.025;
Fig. 2a). In general, the death HRs of the DM patients increases linearly with age (HR: 1.81–5.90; P < 0.001). The HRs of dialysis treatment increased dramatically with increasing CCI scores (HR: 7.89–29.40; P < 0.001). The DM patients with cancer showed increased death HRs (HR: 2.41; 95% CI: 2.22–2.61; P < 0.001) but decreased HRs of dialysis (HR: 0.47; 95% CI: 0.33– 0.64; P < 0.0001). We speculate the possible reason that the patients may suffer from other catastrophic conditions such as cancer before dialysis for the risk of death. The hazard ratios of dialysis or death for the physicians with diabetes in shown in Table A and for the general patients with diabetes in Table A.1 in the Appendix. The physicians with DM who used dialysis services had slightly elevated death HR (HR: 1.42; 95% CI: 0.92–2.19; P = 0.117) compared with those who did not. When we compared three groups together: physicians with diabetes only (Table A in the Appendix), the general patients with diabetes only (see Table A.1 in the Appendix), and all the dialysis subjects (Table 3), death HRs were 1.42 (P = 0.117), 1.34 (P < 0.001), and 1.35 (P < 0.001), respectively, showing that the increased HR of death after receiving dialysis is typically attributable to the general patient group with diabetes. Similar to Table 3, the findings in Tables A and A.1 (in the Appendix) show that aging, as well as cancer and other catastrophic conditions, increased the HR of death and that higher CCIs increased the HR of dialysis. After adjustment, no
Table 2 – The characteristics of general patients with diabetes and physicians with diabetes and using dialysis treatments or not. Variables
Total Physician No Yes Gender Female Male Age (years) at DM diagnosed 35–44 45–54 55–64 65–74 75 Mean age of DM diagnosis Urbanization of residence area Level 1 Level 2 & 3 Level 4 & 5 Level 6 & 7 Charlson Comorbidity Index 3 4–5 6–7 8 Cancer No Yes Other catastrophic illness/injury No Yes
No dialysis
Dialysis
P-value
N
%
N
%
19,790
97.72
461
2.28
18,005 1785
90.98 9.02
405 56
87.85 12.15
754 19,036
3.81 96.19
15 446
3.25 96.75
3262 6397 4405 2783 2943
16.48 32.32 22.26 14.06 14.87
39 100 122 87 113
8.46 21.69 26.46 18.87 24.51
0.026
0.621
<0.001
57.7 12.92
64.20 12.71
<0.001 0.602
7917 9210 1945 718
40.01 46.54 9.83 3.63
170 226 47 18
36.88 49.02 10.20 3.90
7918 5167 3261 3444
40.01 26.11 16.48 17.40
15 88 148 210
3.25 19.09 32.10 45.55
17,891 1899
90.40 9.60
415 46
90.02 9.98
18,791 999
94.95 5.05
423 38
91.76 8.24
<0.001
0.783
0.003
386
diabetes research and clinical practice 105 (2014) 382–390
Table 3 – The hazard ratio of dialysis or death among physicians with diabetes and general patients with diabetes. Variables
Dialysis HR
Physician No 1.48 Yes Dialysis No Yes Gender Female 1.12 Male Age (years) 35–44 1.22 45–54 1.62 55–64 1.11 65–74 1.71 75 Urbanization of residence area Level 1 1.13 Level 2 & 3 Level 4 & 5 1.07 0.99 Level 6 & 7 Charlson Comorbidity Index 3 4–5 7.89 19.11 6–7 29.40 8 Cancer No 0.47 Yes Other catastrophic illness/injury No 1.05 Yes
Death
95% CI
1.12
P-value
1.95
0.006
HR
95% CI
P-value
0.88
0.78
0.98
0.025
1.35
1.17
1.55
<0.001
0.66
1.87
0.679
1.28
1.06
1.56
0.012
0.84 1.12 0.74 1.11
1.77 2.33 1.69 2.64
0.294 0.010 0.613 0.014
1.19 1.81 2.89 5.90
0.99 1.51 2.41 4.92
1.43 2.16 3.47 7.08
0.066 <0.001 <0.001 <0.001
0.93 0.77 0.61
1.38 1.48 1.61
0.224 0.686 0.965
1.07 1.27 1.63
1.00 1.13 1.42
1.15 1.43 1.88
0.070 <0.001 <0.001
4.56 11.19 17.20
13.65 32.62 50.24
<0.001 <0.001 <0.001
1.02 1.04 1.22
0.92 0.93 1.10
1.14 1.17 1.36
0.716 0.457 <0.001
0.34
0.64
<0.001
2.41
2.22
2.61
<0.001
0.75
1.48
0.764
2.12
1.94
2.31
<0.001
Notes: N = 20,251. The models also have adjusted for monthly insured salary. Abbreviations: HR, hazard ratio; CI, confidence interval.
significant differences were found between the physicians treated with dialysis and the general dialysis patients in terms of the HR of death (Fig. 2b). Table 4 compares the HRs of death among the physicians with diabetes and the general patients with diabetes with/ without dialysis. The physicians with DM who did not require dialysis had a decreased risk of death (HR: 0.87; 95% CI: 0.77– 0.98; P = 0.018) and a non-significant increased risk of death (HR: 1.19) when dialysis was used. There was an increased HR of death with aging (HR: 1.81–5.94; P < 0.0001), as well as with cancer or catastrophic illness in the no-dialysis group. The CCIs did not play a significant role in the HR of death in the DM patients with/without dialysis, except for those with other severe health issues (CCI > 8) among the patients not receiving dialysis.
4.
Discussion
In this study, in general, the physicians with DM survived longer than did the patients in the general patient group with diabetes, but the physicians displayed no advantage from their medical backgrounds compared with the general patients with diabetes if they developed ESRD. We found that the physicians with DM, compared with the general patients
with diabetes, were diagnosed with DM at a similar age but displayed an increased survival (HR: 0.88). The physician group with DM had a higher risk of receiving dialysis and a slightly, although not significantly, higher HR of death. When analyzing only patients with diabetes needing dialysis, physicians had shorter durations from initial DM diagnosis to initial dialysis and from initial dialysis to death compared with the general patient group (52 vs. 60 months, and 13 vs. 14 months, respectively; data not shown). Moreover, severe comorbidities affected physicians with DM dramatically, and the physicians with comorbidities more often required dialysis. This study found that the physicians with diabetes had an increased risk of death if they were male, older, required dialysis, or suffered from cancer or other serious health conditions. A limited number of studies have examined professional groups and investigated their performance-related behavior, and the majority of these studies used small samples or partially considered the health worker effects [26]. One Norwegian study, for instance, reported that physicians were healthier than the general population [4], and other examples from Harvard (the physicians’ health study II) described the risk factors and the benefits associated with the primary prevention of cardiovascular disease [27,28]. Other reports described the behavior of physicians as patients [29,30], which revealed useful information for further studies, while the
diabetes research and clinical practice 105 (2014) 382–390
Fig. 1 – The hazard risk of dialysis between the diabetic physicians and general diabetic patients. Survival curves of matched diabetic physicians and general diabetic patients. The adjusted survival curves were controlled for gender, age, insured salary, urbanization of residence area, and Charlson Comorbidity Index (HR = 1.48; 95% CI = 1.12–1.95; P = 0.006).
387
majority of the studies adopted the survey method to explore the determining factors instead of using longitudinal databases. This study is the first to use national data to explore the hypothesis that a professional group benefits from professional knowledge and that the key to better performance relies on persistent adherence. We have clearly defined the DM population and used the PS method to match the physicians and general patients to avoid selection bias. The longitudinal database provides a better opportunity for accumulation of data concerning physicians and survival analyses. Health-related behavior is influenced by the knowledge and attitudes of an individual. It has been thought that professional groups such as physicians should have a more comprehensive concept of disease as a result of their education and knowledge, thereby leading to better performance in disease control and prevention [31], such as lower likelihood of cesarean section in female physicians [6]. Other research highlighted the lower mortality rate or morbidity rate in professional groups in comparison with the general population [8,32–35], which was attributed to their knowledge, economic resources and/or lifestyle. Given their knowledge, physicians should have a more positive attitude toward their patients and themselves. They should have a better understanding of the determining factors for chronic diseases, such as DM, and should understand that there is no medication for effective DM control that does not require diet, exercise, and life style modification. Most studies related to physicians’ attitudes and practices emphasized the patients’ treatment and did not focus on the physicians’ awareness [36]. In this study, the results imply that, except in severe health situations, physicians may have a higher probability of better survival outcomes. This finding may
Fig. 2 – The hazard risk of survival between the diabetic physicians and general diabetic patients. The adjusted survival curves were controlled for gender, age, insured salary, urbanization of residence area, and Charlson Comorbidity Index. (a) Survival curves of matched diabetic physicians and general diabetic patients (HR = 0.88; 95% CI = 0.78–0.98; P = 0.025). (b) Survival curves of receiving dialysis care among matched diabetic physicians and general diabetic patients (HR = 1.25; 95% CI = 0.86–1.81; P = 0.240).
388
diabetes research and clinical practice 105 (2014) 382–390
Table 4 – The hazard ratios of survival among physicians with diabetes and general patients with diabetes with or without dialysis. Variables
Death (for dialysis patients; N = 461) HR
Physician No Yes 1.19 Gender Female 0.97 Male Age (years) at DM diagnosed 35–44 1.39 45–54 1.52 55–64 2.24 65–74 3.56 75 Urbanization of residence area Level 1 1.03 Level 2 & 3 0.85 Level 4 & 5 0.80 Level 6 & 7 Charlson Comorbidity Index 3 0.63 4–5 0.60 6–7 8 0.59 Cancer No 1.07 Yes Other catastrophic illness/injury No 0.58 Yes
95% CI
Death (for no dialysis patients; N = 19,790)
P-value
HR
95% CI
P-value
0.81
1.75
0.371
0.87
0.77
0.98
0.018
0.47
1.98
0.924
1.28
1.05
1.57
0.017
0.74 0.82 1.21 1.91
2.63 2.81 4.14 6.64
0.306 0.184 0.010 <0.001
1.17 1.81 2.91 5.94
0.97 1.50 2.41 4.90
1.43 2.19 3.52 7.19
0.105 <0.001 <0.001 <0.001
0.79 0.53 0.37
1.36 1.35 1.72
0.814 0.478 0.563
1.07 1.30 1.64
0.99 1.15 1.42
1.15 1.47 1.90
0.071 <0.001 <0.001
0.18 0.18 0.18
2.19 2.02 1.96
0.465 0.410 0.390
0.99 0.99 1.17
0.89 0.88 1.05
1.10 1.11 1.31
0.837 0.809 0.005
0.76
1.49
0.706
2.54
2.33
2.77
<0.001
0.44
0.78
<0.001
2.28
2.08
2.49
<0.001
Notes: The models also have adjusted for monthly insured salary. Abbreviations: HR, hazard ratio; CI, confidence interval.
partly be attributed to their educational background or access to medical resources and is consistent with other studies. For the patients with DM on dialysis treatment under the universal health insurance coverage in Taiwan, the barriers of accessibility and finances of adhering to the treatment protocol were considerably lower than in other countries. Optimum healthcare for health care professionals should include the application of medical knowledge, adoption of a good attitude, and medical practice identical to that given to patients in the general population. The findings between physician with DM and physician with DM using dialysis services imply that deficiencies remain between the attitudes and practices. A possible explanation is that DM initially develops during the peak time in the professional life of a physician and that career involvement prevents physician patients from adhering to the principles of DM control to delay the need for dialysis. Based on our findings, another condition that should be considered is the behavior of unhealthy physicians. The behavior includes a broad spectrum of factors, ranging from self-care and self-prescription of drugs to informal consultation and formal treatment by another physician [29]. Physicians appear to be low users of formal services but high users of preventive care [5]. In addition, physicians, who should have the best medical knowledge, should have positive outcomes of illness, especially with chronic diseases that require regular clinical follow-up and lifestyle changes.
However, medical knowledge and access to medications increase the potential for self-treatment, which is common among physicians but also raises concerns in the profession [37,38]. Female physicians, for example, are generally involved in disease prevention and have advanced knowledge of the benefits of regular PAP smears, but they may not adhere to the recommendations to a higher degree than the general population [39]. A physician with DM may not initially delay receiving diabetes services because he/she is quicker to recognize the relevant factors and care for themselves better than patients in the general population. Nonetheless, if a physician with DM is neglecting his/her personal health by failing to adhere to the DM treatment protocol, his/her condition may worsen more rapidly. It has been shown that low help-seeking is a personality trait that is a risk factor connected with severe disorders [40]. In this study, the best strategy in chronic diseases for an improved outcome is to strictly follow the appropriate protocol, regardless of whether the patient is a doctor or a general patient. There are several limitations to our analyses. We did not have additional information to identify the factors associated with physician behavior; further studies should be performed to include these data. Lifestyle is typically a strongly contributing factor in patients with chronic diseases, especially in DM or in dialysis treatment. The NHI database records insurance claims and does not provide patient lifestyle
diabetes research and clinical practice 105 (2014) 382–390
information. We were unable to link to the Cause of Death Registry because of restrictions in data application. Using the NHI withdrawal data as an approximation of the date of death presents a low risk for errors, including the possibility of using the data of patients who dropped out from NHI but who may have used medical services. Patients with DM or dialysis treatment would not leave their medical care service insurance, especially under a universal health insurance program. Health professionals may have more knowledge regarding disease control and prevention than the general population. In the case of chronic diseases, such as DM or dialysis, the fundamental strategy in controlling the disease and obtaining the best outcome for patients and physician patients is adherence to the treatment guidelines. Professional groups should be aware of these findings to decrease the risk of health care providers’ failing to comply with medical advice.
Competing interests
[7] [8]
[9]
[10]
[11]
[12]
[13] [14]
The authors declare that they have no conflict of interest.
Acknowledgments
[15] [16]
This study was supported by China Medical University and Asia University (Grant no. CMU99-ASIA-18) and used data from the National Health Insurance Database provided by National Health Research Institute. The interpretations and conclusions contained herein do not represent those of the National Health Research Institutes or Department of Health in Taiwan.
[17]
[18]
Appendix A. Supplementary data
[19]
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.diabres.2014.07.004.
[20]
[21]
references [22] [1] Andersen R, Aday LA. Access to medical care in the U.S.: realized and potential. Med Care 1978;16:533–46. [2] Rosenstock IM. Historical origins of the health belief model. Health Educ Monogr 1974;2:328–35. [3] Mechanic D. Correlates of physician utilization: why do major multivariate studies of physician utilization find trivial psychosocial and organizational effects? J Health Soc Behav 1979;20:387–96. [4] Stavem K, Hofoss D, Aasland OG, Loge JH. The selfperceived health status of Norwegian physicians compared with a reference population and foreign physicians. Scand J Public Health 2001;29:194–9. [5] Wachtel TJ, Wilcox VL, Moulton AW, Tammaro D, Stein MD. Physicians’ utilization of health care. J Gen Intern Med 1995;10:261–5. [6] Chou YJ, Huang N, Lin IF, Deng CY, Tsai YW, Chen LS, et al. Do physicians and their relatives have a decreased rate of
[23]
[24] [25]
[26]
[27]
389
cesarean section? A 4-year population-based study in Taiwan. Birth 2006;33:195–202. Frank E, Biola H, Burnett CA. Mortality rates and causes among U.S. physicians. Am J Prev Med 2000;19:155–9. Carpenter LM, Swerdlow AJ, Fear NT. Mortality of doctors in different specialties: findings from a cohort of 20000 NHS hospital consultants. Occup Environ Med 1997;54:388–95. Wu CY, Hu HY, Huang N, Pu CY, Shen HC, Chou YJ. Do the health-care workers gain protection against herpes zoster infection? A 6-year population-based study in Taiwan. J Dermatol 2010;37:463–70. Gagne P, Moamai J, Bourget D. Psychopathology and suicide among Quebec physicians: a nested case control study. Depress Res Treat 2011;2011:936327. Schernhammer ES, Colditz GA. Suicide rates among physicians: a quantitative and gender assessment (metaanalysis). Am J Psychiatry 2004;161:2295–302. Gastfriend DR. Physician substance abuse and recovery: what does it mean for physicians – and everyone else? JAMA 2005;293:1513–5. Ingstad B, Christie VM. Encounters with illness: the perspective of the sick doctor. Anthropol Med 2001;8:201–10. Hem E, Stokke G, Tyssen R, Gronvold NT, Vaglum P, Ekeberg O. Self-prescribing among young Norwegian doctors: a nine-year follow-up study of a nationwide sample. BMC Med 2005;3:16. Hall PM. Prevention of progression in diabetic nephropathy. Diabetes Spectr 2006;19:18–24. Centers for Disease Control and Prevention. Incidence of end-stage renal disease attributed to diabetes among persons with diagnosed diabetes – United States and Puerto Rico, 1996–2007. MMWR Morb Mortal Wkly Rep 2010;59:1361–6. Collins AJ, Foley RN, Herzog C, Chavers BM, Gilbertson D, Ishani A, et al. Excerpts from the US Renal Data System 2009 Annual Data Report. Am J Kidney Dis 2010;55. S1–420, A6–7. The Diabetes Prevention Program. Design and methods for a clinical trial in the prevention of type 2 diabetes. Diabetes Care 1999;22:623–34. Dorr DA, Wilcox A, Donnelly SM, Burns L, Clayton PD. Impact of generalist care managers on patients with diabetes. Health Serv Res 2005;40:1400–21. Chen Y-C, Yeh H-Y, Wu J-C, Haschler I, Chen T-J, Wetter T. Taiwan’s National Health Insurance Research Database: administrative health care database as study object in bibliometrics. Scientometrics 2011;86:365–80. Cheng CL, Kao YH, Lin SJ, Lee CH, Lai ML. Validation of the National Health Insurance Research Database with ischemic stroke cases in Taiwan. Pharmacoepidemiol Drug Saf 2011;20:236–42. Chang CH, Shau WY, Jiang YD, Li HY, Chang TJ, Sheu WH, et al. Type 2 diabetes prevalence and incidence among adults in Taiwan during 1999–2004: a national health insurance data set study. Diabetic Med 2010;27:636–43. Ahmed A, Aban IB, Vaccarino V, Lloyd-Jones DM, Goff Jr DC, Zhao J, et al. A propensity-matched study of the effect of diabetes on the natural history of heart failure: variations by sex and age. Heart 2007;93:1584–90. D’Agostino Jr RB, D’Agostino Sr RB. Estimating treatment effects using observational data. JAMA 2007;297:314–6. Lien HM, Chou SY, Liu JT. Hospital ownership and performance: evidence from stroke and cardiac treatment in Taiwan. J Health Econ 2008;27:1208–23. Radon K, Goldberg M, Becklake M. Healthy worker effect in cohort studies on chronic bronchitis. Scand J Work Environ Health 2002;28:328–32. Lotufo PA, Gaziano JM, Chae CU, Ajani UA, Moreno-John G, Buring JE, et al. Diabetes and all-cause and coronary heart
390
[28]
[29] [30] [31]
[32]
[33]
[34]
diabetes research and clinical practice 105 (2014) 382–390
disease mortality among US male physicians. Arch Intern Med 2001;161:242–7. Djousse L, Gaziano JM. Egg consumption in relation to cardiovascular disease and mortality: the Physicians’ Health Study. Am J Clin Nutr 2008;87:964–9. Rosvold EO, Bjertness E. Illness behaviour among Norwegian physicians. Scand J Public Health 2002;30:125–32. Rosvold EO, Tyssen R. Should physicians’ self-prescribing be restricted by law. Lancet 2005;365:1372–4. Rimpela AH, Nurminen MM, Pulkkinen PO, Rimpela MK, Valkonen T. Mortality of doctors: do doctors benefit from their medical knowledge. Lancet 1987;1:84–6. Juel K, Mosbech J, Hansen ES. Mortality and causes of death among Danish medical doctors 1973–1992. Int J Epidemiol 1999;28:456–60. Torre DM, Wang NY, Meoni LA, Young JH, Klag MJ, Ford DE. Suicide compared to other causes of mortality in physicians. Suicide Life Threat Behav 2005;35:146–53. Ullmann D, Phillips RL, Beeson WL, Dewey HG, Brin BN, Kuzma JW, et al. Cause-specific mortality among physicians with differing life-styles. JAMA 1991;265:2352–9.
[35] Liu SH, Li TH, Lin YL, Shiao YJ, Wu SC, Li CY, et al. Lower morbidity and disease risk among the Chinese medicine physicians in Taiwan. Tohoku J Exp Med 2009;219:207–14. [36] Mousa HA, Mahmood TA. Do practice guidelines guide practice? A prospective audit of induction of labor three years experience. Acta Obstet Gynecol Scand 2000;79: 1086–92. [37] Christie JD, Rosen IM, Bellini LM, Inglesby TV, Lindsay J, Alper A, et al. Prescription drug use and self-prescription among resident physicians. JAMA 1998;280:1253–5. [38] Montgomery AJ, Bradley C, Rochfort A, Panagopoulou E. A review of self-medication in physicians and medical students. Occup Med 2011;61:490–7. [39] Chen LS, Huang N, Tsay JH, Wang PJ, Chou YJ, Chou P, et al. Screening for cervical cancer among female physicians and their relatives in Taiwan: a population-based comparative study. Prev Med 2007;44:531–5. [40] Tyssen R. Health problems and the use of health services among physicians: a review article with particular emphasis on Norwegian studies. Ind Health 2007;45: 599–610.