Abstract
Background: Approximately 13% of consultations in general practice involve patients with unexplained complaints (UCs). These consultations often end with general practitioners (GPs) ordering blood tests of questionable diagnostic informativeness. Objective: We studied factors potentially associated with GPs’ decisions to order blood tests. Methods: Cross-sectional study. Twenty-seven GPs completed registration forms after each consultation concerning newly presented UCs. Results: Of the 100 analysable patients, 59 had at least one blood test ordered. The median number of ordered tests was 10 (interpercentile range [IPR-90] 2–15). Compared to abdominal complaints, the blood test ordering (BTO) probability for fatigue was five times higher (relative risk [RR] 5.2). Duration of complaints for over 4 weeks also increased this probability (RR 1.6). Factors associated with a lower BTO probability were: likelihood of background psychosocial factors (RR 0.4) and GPs having a syndrome rather than symptom type of working hypothesis (RR 0.5).
Conclusion: We found a high rate of BTO among GPs confronted with patients with UCs. Furthermore, a considerable number of tests were ordered. The selectivity in BTO behaviour of GPs can be improved upon.
Introduction
In 3–39% of their consultations, general practitioners (GPs) are confronted with patients whose complaints remain of unclear origin, even after a thorough initial consultation Citation[1–4]. For the purpose of this paper, these complaints are called unexplained complaints and may be defined as those complaints for which the GP, after clarifying the reason for the encounter, taking the patient's history, and performing a physical examination, is unable to establish a diagnosis. These complaints, with only few exceptions, are usually not caused by a condition that requires intervention and they tend to fade in time Citation[5–7]. Therefore, the Dutch College of General Practitioners in its guideline on blood test ordering recommends a watchful waiting policy of 4 weeks with these patients Citation[8]. If blood tests are ordered anyway, it is advised to order a limited set of tests (haemoglobin [Hb], erythrocyte sedimentation rate [ESR], thyroid stimulating hormone [TSH], and glucose). The Dutch College guideline is consensus-based on these aspects because there is no evidence regarding either the effects of a watchful waiting policy or the number and type of tests ordered.
However, GPs often do not postpone blood test ordering or limit the types of tests ordered Citation[9–13]. They often report having strategic, non-medical reasons for blood test ordering, e.g., in order to reassure patients, to meet patients’ expectations, or to avoid referral to specialists Citation[1], Citation[14]. Although these reasons for ordering blood tests may be valid in themselves, superfluous non-selective test ordering for patients with unexplained complaints only rarely contributes to history taking and physical examination in medical decision-making. Non-selective test ordering may even produce negative effects such as false-positive test results, in turn leading to unnecessary extra investigations, triggering patient anxiety, and the risk of somatization and higher costs Citation[15], Citation[16].
A variety of factors influencing GPs’ decisions on blood test ordering in general have been investigated by others Citation[17–21]. Most of these factors were practice- and professional-related determinants. From related research areas, such as research on determinants influencing physicians’ decisions concerning prescribing, referral, or ordering screening tests, it is known that patient factors (e.g., anxiety, expectations, and family history), physician factors (e.g., perception of guidelines, clinical practice experience, influence of colleagues), and the amount of time and financial costs are involved in decisions about performing these manoeuvres Citation[22–25]. We found only one other article that took into account patient-related factors specifically influencing GPs’ decisions to order blood tests (in a non-screening situation) Citation[26]. In this article, dating from 1984, the authors concluded that the diagnostic category, the age of the patient, the identity of the doctor, and the social class of the patient were each independently related to blood test ordering.
We wanted to look for determinants of blood test ordering in patients with unexplained complaints as anchor points for behaviour-changing strategies to diminish the superfluous non-selective test ordering for these patients. If these determinants can indeed be identified, they may be worth focusing on in educating GPs on diagnostic decision-making. We therefore conducted an observational study in general practice to investigate potential patient- and consultation-related determinants of GPs’ decisions to order blood tests, and examined the number and type of tests ordered by GPs for patients with unexplained complaints in general practice.
Methods
Setting
Dutch GPs were recruited from two regions in the Netherlands, one in the western and one in the southern part of the country. No specific participation criteria for GPs were formulated, other than the location of their practices. Prior to the study, an expert panel consisting of two experienced GPs, two GP trainees, and an epidemiologist made a pre-selection of eligible complaints. Criteria for pre-selection were: frequent reason for encounter in general practices, often remaining of unclear origin after the first consultation, and reason for GPs to order blood tests. Based on these criteria, the panel selected fatigue, abdominal complaints, and musculoskeletal complaints as eligible complaints. From November 2001 through January 2002, participating GPs were asked to complete a case record form (CRF) directly after a consultation with a patient that had presented with a new unexplained complaint from one of the pre-selected complaint groups. An unexplained complaint was defined to be new if it had not been presented to the GP within the previous 6 months. GPs were given the Dutch College of General Practitioners definition of unexplained complaints, as stated in the Introduction. The final decision as to whether the complaint was unexplained or not was left to the GP.
Instruments
Case record forms included questions on patient/complaint characteristics and GP/consultation characteristics (). When the GP decided to order blood tests, a copy of the blood test ordering form was added to the CRF, or, in case of in-practice testing, the performed tests were stated on the CRF by the GP. CRFs and blood test ordering forms were processed anonymously.
Table I. Determinants, answer categories, and cut-off points for dichotomization.
Statistical analysis
For the purpose of data analysis, various variables were dichotomized. Items, answer categories, and cut-off values of dichotomized variables are presented in .
In addition to age, the scales indicated the unexplained character of the complaint, the level of certainty regarding the likelihood of some serious underlying disease, the level of certainty about the working hypothesis, and the satisfaction of the GP with the consultation. The working hypotheses were dichotomized by a procedure in which two of the authors (H.K., G.t.R.) independently scored every working hypothesis as a “symptom type” of working hypothesis (e.g., fatigue or limb pain) or a “syndrome type” of working hypothesis. A working hypothesis was referred to as a “syndrome type” instead of “symptom type” if the hypothesis was more than a mere description of the symptoms presented by the patient, in the sense that clustered symptoms were denoted as a syndrome by the GP (e.g., irritable bowel syndrome or fibromyalgia). The authors disagreed in only three cases; consensus in these cases was easily reached by discussion.
The type of complaint and the presence of psychosocial factors (both three answer categories) were modelled as indicator variables.
Results of the univariate analysis are presented descriptively and as odds ratios with their 95% confidence interval (CI). The dependent variable of the multivariable analysis was blood test ordering (yes or no).
A multivariable stepwise backward logistic regression was performed in order to investigate the conditional associations between the independent variables and the dependent one. A backward rather than forward model was chosen because of the explorative character of the analysis. For the same reason, modest levels of p values were accepted: p(entry) < 0.15 for inclusion in the model and p(removal) > 0.20 for removal of the determinant from the model.
Analyses were conducted using SPSS version 11.5 and Stata version 8.2. In particular, Stata's cluster option to the logistic analysis command, which uses the Huber-White sandwich estimator to calculate robust variance estimations, allowed us to take into account potential non-independence of consultations for a single GP, thus providing more valid confidence intervals compared to an analysis that assumes statistical independence of consultations. The results are presented as risk ratios (RRs) with their 95% CIs.
Risk ratios were computed on the basis of the adjusted odds ratios that were derived from the logistic regression analyses, and are better estimates of association in case of a frequent incidence of the outcome of interest (here, blood test ordering) Citation[27].
Results
General characteristics
Of the 27 participating GPs, four (15%) were female. On average, the GPs were 48 years old (SD 5.7). Practices were mostly situated in urban areas (15/27), 17 were non-solo practices (group practices, health centres, or otherwise). Practices had an average number of 2394 patients (SD 540).
In total, from November 2001 through January 2002, 100 CRFs were completely filled out. The median number of completed CRFs per GP was 3 (interquartile range 2–5).
Patient characteristics and other determinants of the decision to order blood tests are shown in . Blood samples were taken in 59 consultations (59%).
Table II. Patient characteristics and other potential determinants, and their univariate odds ratios (OR) of the decision to order blood tests (100 patients, 27 GPs).
Fifty-nine patients presented with fatigue, 22 with musculoskeletal complaints, and 19 with abdominal complaints. The likelihood of blood test ordering was 83%, 32%, and 16% in these complaint groups, respectively. The small numbers of blood test orders in the latter two complaint categories prevented us from investigating the presence of any interaction between the type of complaint and the other independent variables. For example, it is possible that the likelihood of blood test ordering was different between the sexes depending on the type of complaint.
The mean age of the population was 44 years (SD 19). The difference between the mean age of the group of patients for whom blood tests were ordered (40 years [SD 18]) and of the group of patients for whom no blood tests were ordered (50 years [SD 20]) was significant (mean difference 9.5; p=0.017).
There were more women than men presenting with unexplained complaints (63/100). This held true for the group of patients for whom blood tests were ordered (58% women) as well as for the group without blood test ordering (71% women).
More than 60 different types of tests were ordered in all. In , the top 10 laboratory tests ordered is shown. The tests were not ordered in panels, so each test could be counted individually. The median number of tests ordered per patient was 10 (interpercentile range [IPR-90] 2–15) for those who had tests ordered.
Table III. Top 10 ordered blood tests (n=59).
Haemoglobin and ESR were most frequently ordered, but tests such as bilirubin, urea, aspartate aminotransferase (ASAT), folic acid, triiodothyronine (T3), and vitamin B12 were ordered more than once as well. Note that is predominantly influenced by blood test ordering in patients with fatigue (49/59). For the abdominal complaint group, ESR, Hb, leukocyte count, gamma glutamyltransferase (GGT), and bilirubin were the top 5 orders. In the musculoskeletal complaint group, ESR, leukocyte count/differentiation, and rheumatoid factor were ordered most frequently.
Multivariate analysis
Possible independent effects of all but three of the determinants mentioned in were further investigated by means of a multivariate backward stepwise logistic regression model with blood test ordering (yes/no) as the dependent variable ().
Table IV. Factors most strongly associated with blood test ordering in case of unexplained complaints (multivariable analysis).
The variable “perception of complaint intensity by the GP” was not analysed further because there was hardly any variability within this determinant. The level of certainty about the working hypothesis was not analysed either, because we observed it to be dependent on the type of working hypothesis (syndrome type or symptom type of working hypothesis). The variable “type of working hypothesis” was added instead (see Methods). “Satisfaction with the consultation” was not added as a variable to the multivariate analysis, because this is a determinant that may be influenced by the decision as to whether or not to order blood tests in itself.
As may be seen from , fatigue as a complaint was an independent predictor and had a great impact on the chance of blood test ordering (RR 5.2, 95% CI 2.5–6.1), as did longer-standing complaints (>4 weeks) (RR 1.6, 95% CI 0.8–2.2). Where known psychosocial factors were thought to have influenced the complaints, there was a decrease in blood test ordering (RR 0.4, 95% CI 0.03–1.1). When the GP had a syndrome, rather than a symptom, type of working hypothesis, blood tests were ordered less often (RR 0.5, 95% CI 0.2–0.9). For all these relative risks, the potential non-independence of patients of a single GP was taken into account.
Discussion
Summary of main findings
In our search for determinants influencing the physicians’ decisions to order blood tests for patients with unexplained complaints, we found that fatigue as a complaint, a duration of the complaint of over 4 weeks, the absence of possible explanatory psychosocial factors, and the absence of a syndrome type of working hypothesis independently increased the chance of blood tests being ordered. Furthermore, we found that fatigue was the most frequently encountered unexplained complaint, followed by musculoskeletal complaints. Sixty-three (or 63%) of the patients were female. Blood tests were ordered in 59% of the consultations. The median number of blood tests ordered per patient was 10 (IPR-90 2–15).
The fact that fatigue as a complaint (as compared to musculoskeletal and abdominal complaints) increased the chance of blood tests being ordered did not come as a surprise. Fatigue is a more elusive complaint than unexplained abdominal or musculoskeletal complaints. There is a variety of opinions on what to ask, examine, and test Citation[28], Citation[29]. Furthermore, with fatigue, there is no well-defined physical examination providing additional diagnostic information, so it does not play an important role in reassuring patients with fatigue as it may for patients with unexplained abdominal or musculoskeletal complaints. This may explain the added value of blood tests for patients with unexplained fatigue.
The fact that duration of complaints of over 4 weeks increased the chance of blood test ordering may be expected as well. When patients have suffered from complaints for over 4 weeks at the time of their presentation, the GP and the patient will want to gain more certainty about the innocent nature of the complaints. In such a situation, it will be harder for the GP to postpone blood test ordering. In a way, GPs in fact adhere to the Dutch College guideline (or patients follow their own wait-and-see policy before seeing the physician).
The presence of explanatory psychosocial factors decreasing the chance of blood test ordering can be explained as follows. In the presence of a potential explanatory psychosocial factor, the GP may not feel the need for blood test ordering. When this explanation is lacking or less pronounced, one can understand the GP's blood test ordering as an additional diagnostic tool.
The same kind of reasoning may be adopted for the presence of a syndrome type of working hypothesis (such as irritable bowel syndrome or fibromyalgia). Symptom types of working hypothesis (like fatigue or limb pain) are less clear, both to the patient and to the GP, and may still be, albeit with a very low probability, first symptoms of a more relevant disease (e.g., fatigue as a first symptom of diabetes).
Strength and limitations of this study
Since we looked at patient- and consultation-related factors, this study adds a missing piece to the research on determinants of blood test ordering. There are, however, a few limitations of this study worth noting. First of all, selective inclusion may have taken place, since the 27 participating GPs only identified about three patients each during a 3-month period, but probably saw more of them. This may have caused selection bias. It was infeasible to perform a non-inclusion analysis to check this, because unexplained complaints are not registered as such in patients’ records by the GPs. A study by Vinson et al. suggests that doctors are more likely to provide a diagnosis when they prescribe antibiotics, even if the symptoms are not suggestive of the diagnosis Citation[30]. Consequently, the act of ordering blood tests may have induced physicians to classify the patient differently and not to fill out a CRF for these patients. This may have been a source of selective inclusion, although the fact that GPs having a symptom type of working hypothesis (as compared to a syndrome type of hypothesis) increased the chance that they ordered blood tests does not support this hypothesis. Secondly, GPs were provided with the definition of unexplained complaints as described in the Dutch College of General Practitioners guideline. The final decision as to whether the complaint was unexplained or not was left to the GP. It is therefore possible that different GPs included different “kinds” of patients. However, we took this potential non-independence of patients of a single GP into account in the analyses.
GPs were not instructed to take a structured history or perform a predefined physical examination, but were asked to fill out a CRF after the consultation, which contained questions on history taking and physical examination. This may have caused bias, because not every GP asked every patient all the questions in the history-taking section of the CRF or because GPs may not have been able to recall the answers to specific questions.
The limited number of patients included in our study may have influenced the analysis. If the number of included patients had been higher, other determinants might have stayed in the final model as well.
In the context of existing evidence
Our finding that fatigue is the most frequently seen unexplained complaint is in accordance with the literature Citation[31]. The fact that GPs ordered blood tests in 59% of first consultations and did not postpone blood test ordering for 4 weeks does not seem to be in accordance with the advice in the guideline on blood test ordering Citation[8]. However, many GPs took the duration of the complaint at the moment of consultation into consideration in their decision to order blood tests. When taking this into account, the high percentage of first-consultation blood test ordering is understandable. The median number of blood tests ordered, if any, was 10 (IPR-90 2–15); this is not in line with the four tests recommended in the guideline. In most cases, GPs did order the recommended tests, but added quite a number of rather uninformative tests, such as urea and T3. The subsequent higher risk of false-positive test results and their consequences (unnecessary additional investigations, increased patient anxiety, the risk of somatization, and higher costs) indicates that there still is much to be gained by more selective test ordering by GPs for patients with unexplained complaints.
Implications for research and practice
We found plausible patient- and consultation-related determinants of GPs’ decisions on blood test ordering for patients with unexplained complaints. In a larger sample, other determinants may be found to be of interest as well. The main question, however, will be whether the determinants for blood test ordering are modifiable. Only in that case could one try to influence the blood test ordering behaviour of GPs. For instance, in our study, type and duration of the complaint were factors associated with blood test ordering and these cannot be modified.
Another interesting research theme might be to study what the additional value of immediate or postponed blood test ordering is in the work-up of patients with unexplained complaints. Does immediate or postponed blood test ordering in patients with an unexplained complaint have any impact on the GPs’ policy or on the early or late diagnosis of a disease? What is the influence on the anxiety and satisfaction of the patients? Until the results of these kinds of studies are available, there is room for improvement in the number and rationale of tests ordered by GPs.
References
- van der Weijden T, van Velsen M, Dinant GJ, van Hasselt CM, Grol R. Unexplained complaints in general practice: prevalence, patients’ expectations, and professionals’ test-ordering behavior. Med Decis Making 2003; 23: 226–31
- Peveler R, Kilkenny L, Kinmonth AL. Medically unexplained physical symptoms in primary care: a comparison of self-report screening questionnaires and clinical opinion. J Psychosom Res 1997; 42: 245–52
- Kirmayer LJ, Groleau D, Looper KJ, Dao MD. Explaining medically unexplained symptoms. Can J Psychiatry 2004; 49: 663–72
- Pilowsky I, Smith QP, Katsikitis M. Illness behaviour and general practice utilisation: a prospective study. J Psychosom Res 1987; 31: 177–83
- Knottnerus JA, Starmans R, Vissers A. Diagnostic conclusions of the GP in unexplained fatigue. [Diagnostische conclusies van de huisarts naar aanleiding van onverklaarde moeheid.]. Huisarts Wet 1987; 30: 9–12
- Kroenke K, Jackson JL. Outcome in general medical patients presenting with common symptoms: a prospective study with a 2-week and a 3-month follow-up. Fam Pract 1998; 15: 398–403
- Kenter EG, Okkes IM, Oskam SK, Lamberts H. Tiredness in Dutch family practice. Data on patients complaining of and/or diagnosed with “tiredness”. Fam Pract 2003; 20: 434–40
- Dinant GJ, Wijk van MAM, Janssens HJEM, Somford RG, Jager de CJ, Beusmans GHMI, et al. The DCGP guideline on blood test ordering. [NHG-standaard Bloedonderzoek. Huisarts Wet 1994; 37: 202–11
- Connelly D, Steele B. Laboratory utilization. Problems and solutions. Arch Pathol Lab Med 1980; 104: 59–62
- Hartley RM, Epstein AM, Harris CM, McNeil BJ. Differences in ambulatory test ordering in England and America. Role of doctors’ beliefs and attitudes. Am J Med 1987; 82: 513–7
- Zaat JO, van Eijk JT. Bonte HA. Laboratory test form design influences test ordering by general practitioners in the Netherlands. Med Care 1992; 30: 189–98
- Lyon AW, Greenway DC, Hindmarsh JT. A strategy to promote rational clinical chemistry test utilization. Am J Clin Pathol 1995; 103: 718–24
- Hindmarsh JT, Lyon AW. Strategies to promote rational clinical chemistry test utilization. Clin Biochem 1996; 29: 291–9
- van der Weijden T, van Bokhoven MA, Dinant GJ, van Hasselt CM, Grol RP. Understanding laboratory testing in diagnostic uncertainty: a qualitative study in general practice. Br J Gen Pract 2002; 52: 974–80
- Vafiadis P. The dilemma of the unexpected result. Aust Fam Physician 1996; 25: 971–3
- Sherwood P, Lyburn I, Brown S, Ryder S. How are abnormal results for liver function tests dealt with in primary care? Audit of yield and impact. BMJ 2001; 322: 276–8
- Goold SD, Hofer T, Zimmerman M, Hayward RA. Measuring physician attitudes toward cost, uncertainty, malpractice, and utilization review. J Gen Intern Med 1994; 9: 544–9
- Bugter-Maessen AM, Winkens RA, Grol RP, Knottnerus JA, Kester AD, Beusmans GH, et al. Factors predicting differences among general practitioners in test ordering behaviour and in the response to feedback on test requests. Fam Pract 1996; 13: 254–8
- Malcolm L, Wright L, Seers M, Davies L, Guthrie J. Laboratory expenditure in Pegasus Medical Group: a comparison of high and low users of laboratory tests with academics. N Z Med J 2000; 113: 79–81
- Cherkin DC, Deyo RA, Wheeler K, Ciol MA. Physician variation in diagnostic testing for low back pain. Who you see is what you get. Arthritis Rheum 1994; 37: 15–22
- van Merode GG, Stroink AE, Maarse JA, Goldschmidt HM. Impact of insurance coverage type on laboratory test ordering behaviour of general practitioners. World Hosp Health Serv 2000; 36: 7
- Tudiver F, Brown JB, Medved W, Herbert C, Ritvo P, Guibert R, et al. Making decisions about cancer screening when the guidelines are unclear or conflicting. J Fam Pract 2001; 50: 682–7
- Haggerty J, Tudiver F, Brown JB, Herbert C, Ciampi A, Guibert R, et al. Patients’ anxiety and expectations: how they influence family physicians’ decisions to order cancer screening tests. Can Fam Physician 2005; 51: 1658–9
- Butler CC, Rollnick S, Pill R, Maggs-Rapport F, Stott N. Understanding the culture of prescribing: qualitative study of general practitioners’ and patients’ perceptions of antibiotics for sore throats. BMJ 1998; 317: 637–42
- Webb S, Lloyd M. Prescribing and referral in general practice: a study of patients’ expectations and doctors’ actions. Br J Gen Pract 1994; 44: 165–9
- Hartley RM, Charlton JR, Harris CM, Jarman B. Influence of patient characteristics on test ordering in general practice. Br Med J (Clin Res Ed) 1984; 289: 735–8
- Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998; 280: 1690–1
- Solberg LI. Lassitude. A primary care evaluation. JAMA 1984; 251: 3272–6
- Valdini AF. Fatigue of unknown aetiology—a review. Fam Pract 1985; 2: 48–53
- Vinson DC, Lutz LJ. The effect of parental expectations on treatment of children with a cough: a report from ASPN. J Fam Pract 1993; 37: 23–7
- Pop, P, Beusmans, GHMI. Laboratory test ordering in vague and general complaints. [Laboratoriumonderzoek bij vage en algemene klachten.] Practitioner. 1990;19 suppl:75–81.