Abstract
The factors that affect physicians’ prescribing remain unclear. Although previous reports suggest that prescription decisions are associated with various clinical situation, most of these studies analyzed simulated patient models rather than actual clinical practice. Here, we retrospectively analyzed actual cases of statin prescription for hyperlipidemia at Tohoku University Hospital between Apr 1, 2004 and Mar 31, 2008. Twelve physicians (6 cardiologists, 3 nephrologist, and 3 diabetologist) made decisions on whether to prescribe statins to 187 patients in 788 visits. As expected, cardiologists started prescribing statins at significantly lower serum total cholesterol levels than other specialists (221.7mg/dL vs. 244.7mg/dL, P<0.05). Interestingly, the total cholesterol levels that triggered prescribing differed significantly among cardiologists (p<0.05). These results suggested that prescription decisions differed not only among specialties but also among individuals.
Introduction
Delivering high-quality care is important for all health care providers. Health care purchasers are now focusing on identifying which individual physicians deliver good care most efficiently1,2. Physicians’ efficiency has been evaluated in terms of quality of prescribing3–5, treatment method6,7, and adherence to clinical guidelines8–10. The evaluation of physicians’ prescribing decisions is worthy of attention because physicians need to choose effective medicine for appropriate patients according to guidelines.
In most developed countries, hyperlipidemia is a major public health problem because it causes cardiovascular disease. Statins, which are inhibitors of 3-hydroxy-3 -methylgutaryl coenzyme A reductase, are currently recommended for patients with hyperlipidemia. However, in the United States, less than half the patients who should receive statin treatment to reduce coronary artery disease risk are receiving such treatment11.
A large number of studies have reported physicians’ prescribing decisions. Most of those studies used a patient simulation model, a postal questionnaire, or interviews. However, in the actual clinical situation, physicians have limited time to deliberate on all issues a patient may face. In this study, we used actual clinical data (serum total cholesterol levels) to analyze differences among individual physicians in prescription of statins to patients with hyperlipidemia.
Methods
Study design
The pilot study used a retrospective analysis of physicians’ prescribing information in a computerized physician order entry (CPOE) system and paper-based medical records.
Data Collection
From Apr 1, 2004 to Mar 31, 2008, we studied 12 specialist physicians (6 cardiologists, 3 nephrologists, and 3 diabetologists) who had treated outpatient with hyperlipidemia (n = 187) at Tohoku University Hospital in Sendai, Japan. Of all visits, we analyzed 788 in which patients had blood tests including serum cholesterol level. We recorded serum total cholesterol levels, and the levels at which physicians prescribed statins, on the basis that physicians used this information for clinical decisions. These data were derived from laboratory test results and prescription information from the CPOE system. Paper-based medical records were also used to retrieve medical information including diagnoses of hypertension, diabetes mellitus, hyperlipidemia or ischemic heart disease, and smoking history. We also reviewed physicians’ notes to check pivotal factors that influenced their decision to prescribe statins and whether they had known the serum cholesterol results before prescription.
The data collection strategy provided multiple information sources for a comprehensive evaluation of physicians’ screening and management of serum total cholesterol levels.
This study was conducted with the approval of the Ethical Review Committee at Tohoku University School of Medicine.
Statistical analysis
Descriptive statistics were determined and multiple regression analysis was used where appropriate. Comparisons of total cholesterol levels at which each physician initiated statin treatment were conducted with the Welch test for categorical variables. For comparison of average values of serum total cholesterol according to statin treatment status for each physician ANOVA was used. In addition, ANOVA was used to compare serum total cholesterol levels among physician specialties. All analyses were conducted with SPSS 17.0 (SPSS Inc, Japan).
Results
Physicians characteristics
The characteristics of the physicians are displayed in Table 1. Physicians had spent an average of 20 years in practice (SD = ±5.2). Of the 12 physicians, 50% were cardiologists, 25% were nephrologists, and 25% were diabetologists. There was no significant correlation between the total cholesterol levels and physicians’ years in practice (p = 0.503). However, physician specialty was significantly associated with serum total cholesterol levels (diabetologists > nephrologists > cardiologists, p < 0.001).
Table 1.
physician and Patient Characteristics
| Number, Mean ± SD, or Percentage | |
|---|---|
| Physicians (n = 12) | |
| Professional | |
| Years in practice | 20 ± 5.2 |
| Cardiologist | 6 (50%) |
| Nephrologist | 3 (25%) |
| Diabetologist | 3 (25)% |
| Male | 12 (100%) |
| Patients (n = 187) | |
| Age | 65.7 ±13.5 |
| Male | 96 (51.3%) |
| Female | 91 (48.7%) |
| Smoking | 65 (34.8%) |
| History | |
| Hypertension | 129 (69.0%) |
| Diabetes | 111 (59.4%) |
| Hypercholesterolemia | 135 (72.2%) |
| Coronary heart disease | 105 (56.1%) |
Patient characteristics
The sociodemographic and clinical characteristics of the patients are displayed in Table 1. Mean age was 65.7 years. Diagnoses were as follows: hypertension, n = 129; diabetes, n = 111, hyperlipidemia, n = 135; and coronary heart disease, n = 105. Sixty-five patients were smokers.
Comparison of total cholesterol levels in decision making regarding statin treatment
Serum total cholesterol levels for each physician’s statin treatment choice are listed in Table 2. Among individual physicians, there were significant differences between serum total cholesterol levels at which physicians prescribed statins and those at which they did not (P < 0.001 except for physicians 8 and 9). Five statins were prescribed: pravastatin, simvastatin, pitavastatin, atorvastatin, and rosuvastatin.
Table 2.
Total cholesterol levels with respect to statin prescription status for each physician
| Physician | No Statin treatment (n = 542) | Statin treatment (n = 246) | P value | |||||
|---|---|---|---|---|---|---|---|---|
| Profession | Total cholesterol (mg/dL) | Total cholesterol (mg/dL) | ||||||
| n | Mean | SD | n | Mean | SD | |||
| 1 | cardiologist | 37 | 188.7 | 20.8 | 26 | 227.0 | 30.9 | < 0.001 |
| 2 | cardiologist | 30 | 178.0 | 16.8 | 14 | 220.3 | 35.3 | < 0.001 |
| 3 | cardiologist | 33 | 168.0 | 25.3 | 26 | 216.3 | 40.3 | < 0.001 |
| 4 | cardiologist | 24 | 182.1 | 39.5 | 12 | 221.8 | 21.6 | < 0.001 |
| 5 | cardiologist | 32 | 185.3 | 37.4 | 25 | 222.8 | 39.6 | < 0.001 |
| 6 | cardiologist | 37 | 179.7 | 27.6 | 17 | 221.7 | 25.1 | < 0.001 |
| 7 | nephrologist | 75 | 216.6 | 15.2 | 21 | 246.9 | 39.3 | < 0.001 |
| 8 | nephrologist | 52 | 202.9 | 40.1 | 19 | 203.7 | 42.1 | 0.935 |
| 9 | nephrologist | 37 | 193.9 | 38.3 | 23 | 226.6 | 35.1 | 0.002 |
| 10 | diabetologist | 71 | 214.1 | 26.2 | 21 | 268.2 | 57.1 | < 0.001 |
| 11 | diabetologist | 88 | 215.7 | 18.6 | 22 | 242.1 | 45.8 | < 0.001 |
| 12 | diabetologist | 26 | 202.2 | 15.7 | 20 | 223.0 | 20.1 | < 0.001 |
Table 3 shows the number of visits that had physician notes in the paper-based medical record (e.g., about awareness of cholesterol high level) that seemed to have influenced the decision of whether or not to prescribe. The greatest proportion of visits that had such notes was 59.6%, for physician 4 (a cardiologist), the lowest was 4.2%, for physician 7(a nephrologist).
Table 3.
The number of visits with physician notes in the paper-based medical record that seemed to have influenced the decision of whether or not to prescribe statins
| Physician | Number of visits that had physician notes | Total number of visits | |
|---|---|---|---|
| Profession | Number, Percentage | n | |
| 1 | cardiologist | 25 (39.7) | 63 |
| 2 | cardiologist | 4 (9.0) | 44 |
| 3 | cardiologist | 21 (35.6) | 59 |
| 4 | cardiologist | 20 (55.6) | 36 |
| 5 | cardiologist | 34 (59.6) | 57 |
| 6 | cardiologist | 25 (46.3) | 54 |
| 7 | nephrologist | 4 (4.2) | 96 |
| 8 | nephrologist | 8 (11.3) | 71 |
| 9 | nephrologist | 3 (5.0) | 60 |
| 1 | diabetologist | 22 (23.9) | 92 |
| 0 | |||
| 1 | diabetologist | 6 (5.5) | 110 |
| 1 | |||
| 1 | diabetologist | 4 (8.7) | 46 |
| 2 | |||
Figure 1 shows the average values of serum total cholesterol at which statins were prescribed and those at which they were not. There were significant differences between cardiologists and both nephrologists and diabetologists with respect to cholesterol levels at which they did not start statins (P < 0.05). Average serum total cholesterol was 180.4 ± 28.8 mg/dL in the cardiologist group, 207.1 ± 31.9 mg/dL in the nephrologist group, and 213.2 ± 21.9 mg/dL in the diabetologist group. There were also significant differences between cardiologists and diabetologists in the average values of serum total cholesterol that triggered statin prescription (221.7 ± 33.7 mg/dL vs. 244.7 ± 47.2 mg/dL, P <0.05). According to Fig 1, there were also significant differences between nephrologists and diabetologists in the average values of serum total cholesterol that triggered statin prescription.
Figure 1.
Average serum total cholesterol with respect to statin prescription status for each physician specialty. Error bars represent one standard deviation
Figure 2 shows serum total cholesterol levels according to statin treatment status for each physician. These differed significantly among physicians. Physicians 7, 10, and 11 refrained from prescribing statins at significantly higher serum total cholesterol levels than other physicians (No. 7, 216 ± 15.2 mg/dL; No. 10, 214.1 ± 26.2 g/dL; and No. 11, 215.7 ± 18.6 mg/dL; p < 0.05 when compared with other physicians). In addition, the average values of serum total cholesterol that triggered statin prescription were significantly higher for physicians. 7 (246.9 ± 39.3 mg/dL and 10 (268.1 + 57.1 mg/dL) than for other physicians (p < 0.05).
Figure 2.
Average serum total cholesterol with respect to statin prescription status for each physician. Error bars represent one standard deviation
Discussion
We found significant differences not only among specialties but also among individual physicians regarding serum total cholesterol levels and the decision to prescribe statins. As expected, cardiologists started prescribing statins at significantly lower serum total cholesterol levels than other specialists. In addition, the results of our analysis suggested that prescription decisions differed not only among specialties but also among individuals. To date, various studies have focused on the process of judgment and decision making in practice3–7. An analysis of physicians’ prescribing habits indicated that physicians use very different judgment strategies. However, most of these studies used a patient simulation model, a postal questionnaire, or interview. Using the actual clinical data (serum total cholesterol levels) and physician notes (e.g., regarding awareness of high serum total cholesterol levels), in the present study, we found differences between physicians who prescribed medication and those who did not.
Individual physicians differ in their knowledge of which patients require lipid screening and which require treatment. Computerized clinical decision support systems (CDSS) are information systems that aim to optimize physicians’ clinical decision making12. It is important to investigate the design of these systems.. Furthermore, it is necessary to investigate how CDSS affect lipid levels and outcomes, and distribute algorithms for constructing CDSS to alert the physician about patients with a high cholesterol level.
Several limitations should be noted in interpreting the findings of this study. First, although this study was conducted at a single university hospital, it does not necessarily follow that individual patients were attended to by the same doctors. However, our analysis suggested that individual physicians were influenced differently by serum total cholesterol when deciding on statin treatment. A second possible limitation of our study is that we did not analyze other decision making factors such as patient health status (i.e., age, gender, previous medical history, and smoking). The ways in which each of these factors influences an individual physician’s decision making regarding statin prescription should be further explored.
Conclusion
In this study, we found significant differences not only among specialties but also among individual physicians regarding the influence of serum total cholesterol levels on the decision to prescribe statins for hyperlipidemia patients. In addition we showed individual prescribing decisions extracted from actual clinical data. Our results could contribute to strategies that improve the quality of prescribing decisions in practice. Such strategies should be informed by the effects of CDSS on clinical performance.
Acknowledgments
We express sincere thanks to Mr. Manabu Endo and Mr. Tatsuya Onodera at Tohoku University Hospital for helping with the data collection and to the staff members of the Division of Medical Informatics, Tohoku University Graduate School of Medicine and IT Medical Center, Tohoku University Hospital for their cooperation in conducting the survey.
References
- 1.Milstein A, Lee TH. Computer Physicians on Efficiency. N Engl J Med. 2007;357:2649–52. doi: 10.1056/NEJMp0706521. [DOI] [PubMed] [Google Scholar]
- 2.Bierman AS, Clark JP. Performance measurement and equity. BMJ. 2007;334:1333–34. doi: 10.1136/bmj.39251.660127.AD. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bradleyt CP. Factors which influence the decision whether or not to prescribe: the dilemma facing general practitioners. British Journal of General Practice. 1992;42:454–58. [PMC free article] [PubMed] [Google Scholar]
- 4.Backlund L, Danielsson B, Bring J, Strender LE. Factors influencing GPs’ decisions on the treatment of hypercholesterolaemic patients. Scand J Prim Health Care. 2000;18:87–93. doi: 10.1080/028134300750018963. [DOI] [PubMed] [Google Scholar]
- 5.Bowling A, Harries C, Forrest D, Harvey N. Variations in cardiac interventions: doctors’ practices and view. Family Practice. 2006;23:427–36. doi: 10.1093/fampra/cmi125. [DOI] [PubMed] [Google Scholar]
- 6.Brehaut JC, Poses R, Shojania KG, et al. Do physician outcome judgments and judgment biases contribute to inappropriate use of treatments? Study protocol. Implementation Science. 2007;2:18–28. doi: 10.1186/1748-5908-2-18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Weiner SJ, Schwartz A, Yudkowsky R, et al. Evaluating Physician Performance at Individualizing Care: A Pilot Study Tracking Contextual Errors in Medical Decision Making. Med Decis Making. 2007;27:726–34. doi: 10.1177/0272989X07306113. [DOI] [PubMed] [Google Scholar]
- 8.McBride P, Schrott HG, Plane MB, et al. Primary Care Practice Adherence to National Cholesterol Education Program Guidelines for Patients With Coronary Heart Disease. Arch Intern Med. 1998;158:1238–44. doi: 10.1001/archinte.158.11.1238. [DOI] [PubMed] [Google Scholar]
- 9.Cabana MD, Rand CS, Powe NR, et al. Why Don’t Physicians Follow Clinical Practice Guidelines? A Framework for Improvement. JAMA. 1999;282:1458–65. doi: 10.1001/jama.282.15.1458. [DOI] [PubMed] [Google Scholar]
- 10.Christian AH, Mills T, Simpson SL, Mosca L. Quality of Cardiovascular Disease Preventive Care and Physician/Practice Characteristics. J GEN INTERN MED. 2006;21:231–7. doi: 10.1111/j.1525-1497.2006.00331.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) Circulation. 2002;106:3143–421. [PubMed] [Google Scholar]
- 12.Wyk JTv, Wijk MAMv, Stuurkenboom MCJM, et al. Electronic Alerts Versus On-Demand Decision Support to Improve Dyslipidemia Treatment. Circulation. 2008;117:371–8. doi: 10.1161/CIRCULATIONAHA.107.697201. [DOI] [PubMed] [Google Scholar]