Main results
Overall, our results reveal a very important role for GPs in accident care. In 2014, GPs were involved in 70% of all injury cases requiring medical care but no inpatient stay and figured as initial care provider in 56% of them. While involvement stayed at about the same level from 2008 to 2014, cases in which GPs figured as initial care providers have decreased by 4 percentage points during that period. At the same time, accident cases involving care from an emergency department (ED) increased from 38 to 46% and the share receiving initial care from an ED from 30 to 35% – apparently substituting for the declining involvement of GPs in initial care. Because of the decline in providing initial care, GPs acted less as sole care providers (decrease from 51 to 44%) and increasingly as follow-up carers only (increase from 10 to 14%). Also, patients who received initial care from GPs increasingly visit EDs or a medical specialist for follow-up care, which indicates that GPs have become more inclined to refer patients to other providers.
While GPs’ general involvement in accident care remained stable over time, they acted less and less as initial care provider. Apparently, EDs stepped in as a replacement resulting in the observed complementary increase in both involvement and initial care provision by EDs. Neither an increase in the non-Swiss population nor an increase in people living in urban areas can explain these changes over time because adjusting for these factors barely affects the trend identified (see Supporting Information S2 for a comparison of raw and adjusted estimates). The share of non-Swiss in the analysis sample only slightly increased from 26 to 28% between 2008 and 2014; the urban-rural ratio stayed constant.
We also found considerable variation in GPs’ role by region and patient characteristics: GPs are less involved in accident care in more urban compared to more rural regions. Males, younger patients, and non-Swiss citizens have a lower probability of receiving initial care from GPs and – as a result – also show a lower probability of having GPs as their sole care provider.
Comparison with related literature
Our results are consistent with other studies on the topic that report that patients in Switzerland are increasingly treated in hospital EDs [6, 7]. Our study provides systematic evidence of this development from 2008 to 2014 and clearly shows that this is at the cost of GPs, who have been providing less accident care, especially as initial care provider. Hence, Switzerland seems to increasingly suffer from “ED crowding” [20] as most countries in the industrialized world do [21]. Similar developments in the changing role of GPs have been reported for other countries with a comparable health care system relying on a strong GP-led primary care network such as The Netherlands [9] or Germany [22].
Regarding regional variations and differences between patient groups, our results are in line with existing findings from both Switzerland and abroad showing that, for instance, in urban regions, patients rather seek help directly at EDs because EDs are more accessible and convenient [22, 23]. Younger people, as well as non-nationals, might – on average – be less attached to their GPs or might not even have a personal GP, which leads to them seeking help at an ED rather than at GPs’ practices when they need care urgently after an accident [24].
Implications
Our results provide detailed evidence of a profound and rapid ongoing change regarding GPs’ role in accident care. Potential causes include: changing patient behavior; GPs’ changing skills, preparedness and willingness to treat accident patients; structural factors such as GPs’ opening-hours and availability; and the introduction of new or the more frequent application of special diagnostic tools that are not at easily at GPs’ disposal. In our view, rather than just changing patients’ preferences alone, also structural reasons, such as a change in the way healthcare services are operated and provided to patients, are important drivers of the observed development. The rapid pace of the development suggests this, as does the finding that changes in the patient population over time, such as an increase in urban residents or non-Swiss citizens, are not a driving force behind it.
The fact that the time-of-day and day-of-week of an accident substantially influence whether GPs act as initial care provider indicates that the choice of the initial care provider is shaped by the actual or perceived availability of GPs. One strategy to improve both the actual as well as the perceived out-of-hour availability of GPs is the cooperation of GP-networks with hospitals to create hospital-integrated primary care emergency centers. GP-led, hospital-integrated primary care centers have the potential to improve service quality for patients [25], to increase the job satisfaction of GPs with respect to their out-of-hours duty periods [26, 27], and to reduce costs in hospital EDs [13]. The introduction of primary care physician cooperatives located within hospital EDs led to a sizeable reduction in ED use in the Netherlands [28]. Extending GPs’ opening hours to up to 12 h per day reduced the inappropriate ED use in Italy’s Emilia-Romagna Region by between 10 and 15% [29].
The substantial substitution of services provided by GPs with EDs might impact the quality of care, patient satisfaction and health care costs. A preliminary cost-increase-decomposition (Blinder-Oaxaca, not reported) shows that about a third of the average costs-per-case increase for outpatient cases from 768 CHF (2008) to 853 CHF (2014) is associated with the changing role of GPs.
There are also implications for the education of future GPs. Despite their decreasing involvement in accident care overall, GPs are still confronted with a wide range of injuries and need the corresponding traumatological skills (sutures, treatment of fractures). Our data can contribute to the design of educational tracks that are “evidence-based”, tailored to the most frequently treated injury patterns by GPs [30].
Strengths and limitations
Our administrative data has some clear advantages over, for instance, self-reported data from population surveys: the data generating process was quite constant over the analysis period and the variables of interest are based on claims and, hence, very valid. Compared to hospital data on ambulatory and inpatient care, our data has the major advantage that it allows for reconstructing care pathways related to one particular accident. Lastly, we have information on all types of providers and can clearly show how, for instance, the decline in GPs’ involvement is related to an increase in care provided by EDs.
Our study has several limitations. First, because of only partially reported data on patients’ beginning of hospital inpatient stays prior to 2014, we restricted the analysis to accidents with no inpatient stay at any point during the care pathway. Accidents with no inpatient stay account for 91.5% of all accidents and its share remained quite stable during the analysis period (91.1% in 2008, 92.0% in 2014, see Figure 8 in the Supporting Information). Also, GPs, the focus of the analyses, provide initial care mostly to outpatients. Results on all patients including inpatient cases, where possible, are reported in the Supporting Information S 5. Our main conclusions regarding the changing GPs’ role are supported in these analyses. Second, we have no information about the reasons why patients are treated by a particular provider. We do not know whether it was the patients’ choice to seek care directly at an ED, whether they were told to do so by their GPs when trying to arrange an appointment, or whether their GPs simply weren’t available. Third, patients’ care pathways could only partially be reconstructed because we only know the date of the initial treatment of a provider group (GPs, medical specialists, EDs). Fourth, our information regarding the patient’s injury is based on self-administered accident report forms and might not always be completely accurate. Consequently, there might be some unobserved injury heterogeneity we cannot control for in our analysis. However, it can reasonably be assumed that, over the analysis period, overall injury patterns did not change substantially. Hence, changes in service provision are likely not due to an (unobservable) change in the composition of the injuries. Fifth, regarding differences in patients’ behavior, we, unfortunately, have no information on relevant patient characteristics such as education, professional background, family situation, or the existence or not of a personal GP. Finally, our results on the SUVA insured are not straightforwardly generalizable to the general population because our data includes only the working population, and no patients younger than 18 and older than 65 years. In addition, patients from the service sector are underrepresented, which leads to an underrepresentation of female employees in our analysis sample. As we have shown, females in our sample have a considerably lower rate of ED use (2014, outpatients only: 33% females vs. 36% males, see Fig. 2). For both males and females, however, there is an identical decline of 5 percentage points between 2008 and 2014, as additional analyses show. Hence, while there is a difference in levels, the observed trend over time is the same for both genders.