Doctor shopping before and after a visit to a paediatric emergency department

Abstract

BACKGROUND:

Use of multiple care providers is known to be associated with poor continuity of care.

OBJECTIVES:

To estimate the prevalence of and identify risk factors for doctor shopping by parents of children with common acute illnesses seen in the emergency department (ED) of a children’s hospital.

SETTING:

ED at the Montreal Children’s Hospital (MCH), Montreal, Quebec.

METHODS:

Doctor shopping was defined as visiting three or more different care sites (the MCH ED, other EDs, outpatient clinics or private offices) for a single illness episode, including all visits occurring within successive 72 h periods up to a maximum of 15 days before and after an ED visit from April 1995 to March 1996. Logistic regression was used to compare characteristics of illness episodes with doctor shopping versus those without.

RESULTS:

Of the total 40,150 visits during the study period, doctor shopping was observed in 18% of the visits. The risk of doctor shopping was positively associated with an initial visit at other EDs (odds ratio [OR] 9.08, 95% CI 7.16 to 11.52), outpatient clinics (OR 4.47, 95% CI 3.71 to 5.37) or private offices (OR 1.71, 95% CI 1.48 to 1.96) versus those who visited the MCH ED first. The risk did not differ according to whether a paediatrician versus a general practitioner saw the child during the initial visit (OR 0.99, 95% CI 0.86 to 1.15). Some diagnoses (the reference category was upper respiratory infection), including urinary tract infection (OR 3.31, 95% CI 2.58 to 4.23) and gastroenteritis (OR 1.59, 95% CI 1.35 to 1.88), were associated with an increased risk of doctor shopping, while asthma was associated with a reduced risk (OR 0.71, 95% CI 0.60 to 0.86).

CONCLUSION:

Doctor shopping is common among parents of children with acute illnesses. Parents of children who were seen in the MCH ED first were less likely to doctor shop, perhaps because the parents were more confident about the advice and treatment received. Further research should investigate the underlying reasons for doctor shopping, eg, services other than an ED were not available and parents’ perceptions of the quality of health services.

The use of multiple care providers is associated with poor continuity of care and may add excess costs to the health care system. Seeking care in different sites leads to discontinuity of care, a factor that is found to be a significant predictor of hospitalization (1). Patients (adults and children) in Delaware who sought treatment from the same doctor had a significantly reduced chance of being hospitalized compared with those who were treated by different physicians or who sought care at different sites. Paediatric studies on continuity of care report that children who seek care from one primary source have fewer visits per illness episode and the cost per episode is less (2). Canada’s Medicare system offers universal health care insurance with no financial barriers to seeking treatment at the most convenient or accessible site (physician’s office, clinic or hospital).

Given the relative lack of barriers to seeking treatment at the most convenient or accessible site, concern has been expressed about the overuse of the health care system and particularly about doctor shopping (seeking care frequently and at different sites). An article in the Canadian Medical Association Journal (3) expressed concern about the financial cost of doctor shopping. The study, conducted in Manitoba, provided information on adult patients who doctor shop (the highest user had 247 visits to 71 different physicians in one year), but no information was given about the paediatric population nor even about the extent of the problem in the adult population. The patterns of and reasons for the use of health services among children may be very different because parents may seek care for their children differently than adults do for themselves. The Manitoba study (3) focused on the patient as the source of the problem and provided little information on other factors that influence the use of health services such as physician referral and follow-up, and system-related difficulties (eg, on-call availability). These other factors, however, may have as great an influence on doctor shopping as individual factors (eg, chronic illness) that are related to frequent visits to a doctor.

The objectives of the present study were to estimate the prevalence of doctor shopping, defined as seeking care frequently and at different sites, among children with common acute childhood illnesses treated at a paediatric hospital; and to identify the determinants of doctor shopping.

METHODS

The patterns of visits to the emergency department (ED) at the Montreal Children’s Hospital (MCH), a paediatric teaching hospital in Montreal, Quebec, were analyzed. All children with a Quebec Medicare number who were aged three months to 12 years, and who presented to the MCH ED between April 1995 and March 1996 and received a discharge diagnosis of one of the following common acute illnesses were included: upper respiratory infection; otitis; urinary tract infection; fever; colic; abdominal pain; gastritis; food poisoning; rash; viral infection; cervical adenitis; infectious mononucleosis; croup; healthy child; or asthma or bronchiolitis. These diagnoses, selected a priori, were chosen because children frequently present with such conditions and many of them can be treated with visits to one site (eg, ED or doctor’s office).

Children from other countries and provinces, and infants younger than three months of age (9% of the patients in the sampling frame) were automatically excluded because they did not have a Quebec Medicare number. A database including the child’s age, discharge diagnosis, date of the ED visit and postal code was created from the ED computerized records, and sent to the provincial medical billing registry, the Régie de l’assurance maladie du Québec (RAMQ). RAMQ provided a database with the number of visits to a physician 15 days before and after the index visit to the MCH ED that was linked to the ED database.

Visits were analyzed as sets of visits, corresponding to a (assumed) single illness episode. The MCH ED visit was considered to be the index visit; the diagnoses used in the present analyses were based on those given at discharge during that visit. The following algorithm was used to define the number of visits for a set: if a child had a billed visit in the 72 h before or after the ED visit, then the next (previous or subsequent) 72 h were searched. If an additional visit was made within the next 72 h, then the subsequent 72 h were searched, and so on up to a maximum of 15 days, ie, five 72 h cycles, before and after the ED index visit. The algorithm was designed to increase the likelihood that previous or subsequent visits were for the same acute illness episode rather than for two or more separate episodes. Based on the clinical experience of the investigators, 30 days (15 days before or after the ED index visit) was considered to be an adequate time period for the resolution of the acute symptoms of the illnesses included in the present study. The algorithm defined 35,008 sets of visits.

Children who had only one visit (n = 20,227) were excluded from the analysis because that one visit was to the MCH ED (the sampling frame). Including children with only one visit would overestimate the prevalence of single visits to the ED. The investigators were not able to obtain information on single visits to other sites (doctors’ offices, other EDs or clinics); thus, the information would not be comparable with that from the MCH ED. All visits to the same institution on the same day were counted as a single visit. Children may have seen more than one physician at an institution during the same visit, but these consultations were grouped together as a single visit. Visits included in this study capture both referrals to the ED and specialists, and self-referrals. Physicians at local community service centres (a network of primary care clinics) in Quebec are paid by salary rather than by fee for service; visits to these centres are not captured in the database produced by physician billings and are, thus, excluded from the analyses. The sample size after exclusions was 14,781 sets of visits. Of these 14,781 sets, 60.2% of patients had two visits, 23% had three visits, 9.3% had four visits, 4.1% had five visits and 3.4% had more than five visits.

Doctor shopping was defined as visits to three or more different sites for the same illness episode, ie, within the same set. In addition to describing the overall patterns of visits and the prevalence of doctor shopping, the characteristics of the children whose parents doctor shopped were compared with those of children who visited only two different sites. All comparisons of place of visit and type of physician are based on the first visit in each illness episode. Stepwise multiple logistic regression analysis was used to examine the independent effects of age, sex, ED discharge diagnosis, socioeconomic status (based on percentage of population below poverty level in residential postal code area), type of health care facility (hospital, private office or clinic), physician specialty (specialist or general practitioner) and whether the first visit was on a weekend or weekday. Socioeconomic status and physician specialty were retained in the final regression model because of a potentially confounding effect on other variables in the model.

RESULTS

Of the total 14,781 sets of visits to two or more sites, 1728 (11.7%) were classified as doctor shopping (visiting at least three different sites). A total of 40,150 visits were included in these sets and 7203 (18%) of these visits were within a set in the doctor shopping category. The frequency distribution for doctor shopping, according to each of the variables studied, is presented in Table 1. Results from the logistic regression analyses of associations between each of the independent variables (potential determinants) and doctor shopping are presented in Table 2. The results include a univariate analysis for each independent variable and the results generated from the multiple logistic regression.

TABLE 1:

Frequency of doctor shopping by parents of children with common acute illnesses seen in the emergency department of a children’s hospital, according to study determinants

Characteristic	Number of episodes	% of visits at three or more sites
Site of initial visit
Montreal Children’s Hospital Emergency Department	8317	7.9
Private office	5231	12.6
Outpatient clinic	768	28.8
Other emergency department	465	41.3
Type of physician seen
Specialist	11,329	10.6
Generalist	3452	15.4
Diagnosis
Upper respiratory infection or otitis	4348	10.0
Asthma or bronchiolitis	2813	7.2
Gastroenteritis or food poisoning	2013	14.8
Viral infection	2002	12.2
Abdominal pain	955	15.8
Croup	663	9.4
Fever not yet diagnosed	597	17.1
Rashes	504	12.5
Urinary tract infection or cystitis	482	26.6
Cervical adenitis	169	11.8
Infectious mononucleosis	124	11.3
Healthy child	99	9.1
Colic	12	16.7
Age (years)
Birth to 1	5202	12.1
2	1776	11.7
3 to 5	3625	12.2
6 to 13	3434	12.2
14 and older	744	12.5
Socioeconomic status
Lowest	4090	11.9
Middle	6065	12.0
Highest	3764	12.3
Day of the week
Weekend	3826	11.4
Weekday	10,955	12.6
Sex
Female	6708	12.3
Male	8073	11.2

TABLE 2:

Univariate and adjusted odds ratios (OR) of the associations between each independent variable and doctor shopping by parents of children with common acute illnesses seen in the emergency department of a children’s hospital

Characteristic	Univariate OR	95% CI	Adjusted OR	95% CI
Site of initial visit
Montreal Children’s Hospital Emergency Department	1.00	Reference category	1.00	Reference category
Other emergency department	8.23	6.73 to 10.06	9.08	7.16 to 11.52
Outpatient clinic	4.73	3.97 to 5.63	4.47	3.71 to 5.37
Private office	1.69	1.51 to 1.89	1.71	1.48 to 1.96
Type of physician seen
Specialist	1.00	Reference category	1.00	Reference category
Generalist	1.54	1.38 to 1.72	0.99	0.86 to 1.15
Diagnosis
Upper respiratory infection or otitis	1.00	Reference category	1.00	Reference category
Urinary tract infection or cystitis	3.28	2.62 to 4.11	3.31	2.59 to 4.23
Colic	1.81	0.40 to 8.30	2.78	0.59 to 13.02
Fever not yet diagnosed	1.87	1.48 to 2.36	1.77	1.37 to 2.28
Abdominal pain	1.70	1.39 to 2.08	1.67	1.33 to 2.09
Gastroenteritis or food poisoning	1.58	1.35 to 1.85	1.59	1.35 to 1.88
Cervical adenitis	1.22	0.76 to 1.96	1.39	0.84 to 2.29
Viral infection	1.26	1.07 to 1.49	1.28	1.07 to 1.52
Rashes	1.30	0.98 to 1.72	1.25	0.93 to 1.69
Healthy child	0.91	0.45 to 1.81	1.12	0.55 to 2.27
Infectious mononucleosis	1.15	0.66 to 2.03	1.04	0.55 to 1.94
Croup	0.94	0.71 to 1.24	0.99	0.74 to 1.35
Asthma or bronchiolitis	0.70	0.59 to 0.84	0.72	0.60 to 0.86
Age (years)
Birth to 1	1.20	1.05 to 1.37	1.19	1.03 to 1.38
2	1.15	0.96 to 1.38	1.11	0.91 to 1.34
3 to 5	1.00	Reference category	1.00	Reference category
6 to 13	1.21	1.04 to 1.40	1.15	0.98 to 1.35
14 and older	1.24	0.97 to 1.58	1.04	0.80 to 1.36
Socioeconomic status
Lowest	1.00	Reference category	1.00	Reference category
Middle	1.11	0.98 to 1.29	1.15	0.99 to 1.33
Highest	1.13	0.98 to 1.27	1.09	0.95 to 1.25
Day of the week
Weekday	1.00	Reference category
Weekend	1.13	1.01 to 1.26	N/A
Sex
Female	1.00	Reference category	N/A
Male	0.89	0.81 to 0.99

N/A Not applicable

The place of the first visit was highly associated with doctor shopping. Compared with the MCH ED, children who visited other EDs, outpatient clinics or private offices for the initial visit all had a significantly higher risk of doctor shopping than those who presented first to the MCH ED. Children seen initially by generalists were at increased risk for doctor shopping, but not after controlling for confounding variables in the multiple logistic regression model (adjusted odds ratio [OR] 0.99, 95% CI 0.86 to 1.15).

Several diagnoses had a significant impact on the risk of doctor shopping. Using upper respiratory infection as the reference category, six other diagnoses were significantly associated with doctor shopping in the multivariate model. Five of these diagnoses (urinary tract infection or cystitis, fever not yet diagnosed, abdominal pain, gastroenteritis and viral infection) had a positive association, while asthma or bronchiolitis had a protective effect.

Parents of children younger than one year of age had a slightly elevated risk of doctor shopping compared with parents of older children. Using areas with more than 30% of persons below the poverty line as the reference category, children residing in either of the wealthier neighbourhood categories had a greater likelihood of doctor shopping (not significant).

Other variables that were analyzed but were eliminated from the multiple regression analysis because of nonsignificance were the day of the week of the first visit and the sex of the child.

DISCUSSION

Doctor shopping for common acute childhood illness episodes can increase health care costs and have negative implications for continuity of care. To our knowledge, this is the first study to quantify the prevalence of doctor shopping in a paediatric population. That there were 7203 visits to three or more different sites during one year suggests that this phenomenon warrants further study.

The major determinants of doctor shopping appear to be related to the delivery of health services and to the diagnosis given at the ED. In our study, the influence of the ED diagnosis on visits to different sites raises several questions. While all of the diagnoses recorded are fairly common childhood illnesses, only six diagnoses had a significant association with doctor shopping. Asthma accounted for a large number of sets of visits (n = 2813) but had a protective effect on visiting different sites. This may be related to the availability of treatments for asthma (eg, masks are not available in all doctors’ offices) or to the perception that asthma is ‘fixed’ in the ED, with no need for further medical treatment.

There may be many reasons why parents take their children to different sites for diagnosis and treatment of the conditions associated with doctor shopping. Reasons for doctor shopping may include conditions that require specific laboratory tests (eg, urinary tract infections), and conditions with symptoms that may take a long time to resolve and cannot be ‘fixed’ by a single visit to any site (eg, gastroenteritis). An additional factor that may influence doctor shopping is communication with parents about the natural course of a disease; perhaps physicians working in a children’s hospital ED have more experience with such communication. Another such factor is misdiagnosis of a paediatric illness. Parents may have reasons to continue to change sites of care if the child’s symptoms fail to improve (eg, a child is given antibiotics for an ‘infection’ that turns out to be a viral illness). Thus, visiting different sites (defined as doctor shopping in this study) until the symptoms resolve may sometimes be the most appropriate way for parents to deal with an ongoing illness.

Other studies have examined the importance of diagnosis on care-seeking behaviour. A study in an American public hospital (4) examined the factors that influenced adult patients’ return visits to the same ED. Surprisingly, only 25% of return visits were due to disease-related factors. The remaining factors were patient-related, physician-related or system-related. Another survey in an American hospital found that patients were not willing to seek care in an environment other than the ED (5). Over one-third of patients with the lowest severity of illness claimed that they would not trade ED care for care offered in a clinic setting, even if the latter setting were available. Perhaps parents who visit the ED of a paediatric hospital believe that the best care is available there and do not seek care elsewhere. Our findings appear supportive of this hypothesis. Our results suggest that although a child’s diagnosis may be an important factor related to doctor shopping, it is certainly not the only one.

We did not find specialty training, which given our sampling frame would have usually been in paediatrics, to reduce the risk of subsequent doctor shopping. The lack of association between specialty training and quality of care outcomes is not unique to our study. Dale et al (6) compared satisfaction and outcomes among general practitioners, house officers (‘staff physicians in a hospital’) and registrars (‘residents’), and found that, although general practitioners were less expensive, patient satisfaction was high with the three types of providers, and the outcomes were not significantly different. Conversely, outcomes for children have been reported to be better if they are treated by paediatricians rather than family practitioners (7). Continuity of care (the presence of a single primary care practitioner, regardless of training) is, most likely, an important protective factor for doctor shopping. This hypothesis is consistent with previously published research on paediatric ED use that suggests that parents of children with a regular physician are less likely to seek inappropriate care at an ED (8–11).

While socioeconomic status was not significantly related to doctor shopping in the multivariate analysis, univariate analysis suggests that patients living in wealthier areas are somewhat more likely to doctor shop, a trend in the opposite direction to that suggested by other studies (12).

Our study was based on existing databases for a single ED, without any direct contact with (or identification of) patients or physicians. While this was an efficient way to examine variables related to doctor shopping, several limitations are evident. Because RAMQ data exclude children without a Quebec Medicare number (about 9% of total visits), these children were not included in our study. Children younger than one year of age are covered by a parent’s card, but families are encouraged to apply for a card as soon as the child is born. Perhaps families who receive a Medicare number early are different from those who do not. Thus, the interpretation of our findings must be cautious, especially for children younger than one year of age. Data from local community health centres were not included in the RAMQ database, and visits to these centres are, therefore, not included. The total number of visits is certainly underestimated, but no information is available on the distribution of this underestimation within the other variables. Finally, we emphasize that our study targeted resource use patterns of children who visited one children’s hospital ED, and, thus, the results are not necessarily generalizable to children who seek care at other institutions.

CONCLUSIONS

Clinicians have concerns about the implications of doctor shopping for patient care and appropriate resource use. Concern has been expressed about the ability of primary care facilities, particularly walk-in clinics, to provide high quality paediatric care in a predominantly adult setting (13). If children use other primary care services and subsequently visit the hospital ED either because physicians refer them for specialty care or because parents have residual concerns about their child’s condition, the net result is clearly increased rather than decreased expenditures. The results of the present study suggest that seeking initial care for some common childhood conditions at a paediatric ED rather than at a clinic may conceivably reduce rather than increase costs.

The reasons underlying parents’ decisions to seek care in different sites should be examined in future studies. Of interest from a health policy perspective is the influence of the site of the initial visit. Parents of children who are seen first in the ED of a paediatric hospital appear less likely to doctor shop afterward. Many hypotheses can be generated about the reasons for this finding. For example, doctors in the authors’ ED are not paid per visit and have no shortage of patients. Alternative acute care health care services for children may not be available at a time when parents need them, so they seek care at a site and a time that is suitable for them. Finally, doctor shopping may be related to the perceived (or actual) quality of care. Parents may have greater confidence in an ED at a paediatric teaching hospital where the staff has expertise in the care of acutely ill children. Physicians working in walk-in clinics may be more likely to refer patients to an ED if they are unsure of the most appropriate treatment for a child or if diagnostic tests (eg, urinalysis) are not readily available. If a child is treated in an environment that caters predominantly to adult patients, common childhood conditions may be misdiagnosed. Further investigation of all these possibilities is warranted.